V )'i ■^: GEIMl'S LAW: A STUDY. '4^ GRIM M'S LAW A STUDY OE HINTS TOWARDS AN EXPLANATION a OF THE SO-CALLED LAUTVERSCHIEBUNG " TO WHICH AEE ADDED SOME EEMAEKS ON THE PRIMITIVE INDO-EUROPEAN K AND SEVERAL APPENDICES. BY T. LE MAECHANT? LONDON: TRiJBNER & COMPANY, LUDGATE HILL. STRASSBURG: KARL I. TRUBNER. 1876. {All rights reserved.) PKINTED BY TAYLOR AND FBANCIS, RED LION COURT, FLEET STREET. 4 ,m. P 60^ PEEFACE. The subject of this little Treatise came unavoidably in my way almost on the threshold of a somewhat more extensive investigation upon which I had pro- posed to myself to enter. Among other preliminary inquiries, it was necessary for my immediate purpose to see whether certain groups of I-E. roots, the mem- bers of each of which seem to offer the marks of a close relationship to one another, could, with any approach to certainty, be severally traced to a single parent form ; and this inquiry, as it advanced, became at length inextricably involved with another arising out of a very obvious fact, namely, that the varie- ties of form exhibited by cell the members of a group within one of the great dialects correlated by Grimm's Law are identical with the varieties of one member of a group'-as exhibited by all the great dialects so cor- related. Looking about for the various explanations of this fact, I found it scarcely possible not to surmise that all those varieties of form, whether collected in one dialect or dispersed among several, must have originated by phonetic variation in the same way and at the same time, and that their distribution in each and among all of the aforesaid dialects must have VI PREFACE. been due to the relationship that subsisted among those dialects in primeval times. The pursuit of these and such-like conjectures led me farther and farther astray from the line of inquiry originally marked out ; and being (like most people who have thought on the matter) quite dissatisfied with the current hypotheses of Grimm's Law, I resolved at last to detach this sub- ordinate subject altogether, and see what conclusions respecting it were attainable by a rigid application, to the facts before me, of linguistic principles that were either already demonstrated, or that appeared to me demonstrable. I did this the more willingly because the limited time at my disposal for any studies of the kind seemed more likely to suffice for a monograph upon a single and definite problem than for an investigation into larger questions. It soon became manifest, however, that to elaborate all the points of interest that offered themselves at almost every step of even my more limited way would tax the powers of an inquirer who could devote his whole time to the work. Hence, although I have found myself able to consider a few of the more important of such points in separate Appendices, yet as regards many others I have been compelled to rest content with mere hints or bare statements, in the hope of resuming the study of them at a future time. But neither inevitable shortcomings of this kind, nor some other defects both in form and in matter, of which I am sufficiently conscious, will very greatly grieve me, if I shall only have succeeded in clearly PREFACE. Vn working out the principal features of the theory here- inafter propounded. Looked at in the most general and comprehensive way, the characteristic of this theory will be seen to lie in its treatment of Grimm's Law as a compound, instead of a simple, phonetic problem, combining in itself (not of any natural necessity, but through a unique conjuncture of lin- guistic conditions) two distinct problems ; of which one involves the origin or generation^ and the other the distribution or arrangement, of the sounds com- posing the several Mute-systems covered by the Law. It will be found, I hope and believe, that this method of treatment, whatever else it may do or leave un- done, will avoid the most glaring objections to the prevalent hypotheses on the subject. For, on the one hand, by the origin assigned to the weaker Mutes, the evolution of the German Mute-systems, instead of appearing to invalidate the almost universal Prin- ciple of Debilitation (or of " Least Effort," § 6), is actually accounted for by that principle ; while, on the other hand, in favour of the process to which the symmetrical arrangement of the Mutes in the related systems is attributed it becomes possible to produce some evidence, at any rate, from similar phonetic pro- cesses actually observable. What is here attempted, therefore (to borrow an illustration from a sister science), closely resembles, in a small way, what has been achieved in Astronomy on a grand scale ; for as it was only when the orbital motions of the planets were no longer regarded as simple and uncompounded, but were treated each as Vill PREFACE. the resultant of two rectilinear motions, that both they themselves came to be properly understood, and that the planetary masses were shown to be as obedient as all other matter to the universal sway of attraction ; so, in the present case, I have ventured to think that Grimm's Law, by the analysis here proposed, will be- come susceptible, as a whole, of a satisfactory explan- ation, and will, in particular, be reduced to subjec- tion under the Principle of Least Effort, — as, indeed, it ought to be. For this Principle is, in Phonology, exactly what Gravitation is in the system of the Uni- verse ; and no case of sound-change can with perfect safety be represented as in direct opposition to it, unless counteracting agencies can be actually detected at work. Finding nowhere the slightest ground for thinking that any such agencies ever contributed to the phenomena of Grimm's Law, — finding everywhere, in fact, good grounds for thinking the reverse, — I have deemed it of prime importance to reconcile the Law with the aforesaid Principle ; and to effect this re- conciliation is the first, and, while really the more important, not perhaps the more difficult, of the two problems which go to make up the compound one. But the most striking and interesting, although at the same time the most perplexing, aspect of Grimm's Law lies in that distribution of its implicated sounds which will here furnish the subsidiary component problem. It is this that differentiates it from other cases of sound-change ; and by this it stands unique. There are not wanting, indeed, other and simpler cases of symmetrical distribution (§§ 18, 39); but they are PREFACE. IX only approximations to this case ; and the reasoning from them to it must needs be constructive only. In the treatment of this problem accordingly we shall very likely have to venture on untrodden' ground and to enunciate new principles; and if it shall appear that the evidence adduced to establish these principles is never so copious, and sometimes not so clear, as could be wished, let it be remembered that such a drawback is inseparable from the necessarily narrow sphere of observation of any individual inquirer. If the attention of other observers, over wider linguistic areas, shall ever be directed to a search for more and better evidence of a similar kind, I cannot b^t think that much will be forthcoming. The mention of evidence leads me to remark that the later sections of this Essay (unless it be the last of all), which treat of the phenomena exhibited by the I-E. ^, are not to be dissociated from the body of the book. Those phenomena, it is true, are not without great interest of their own ; but this fact would of itself by no means justify an examination of them here. They are introduced ostensibly as being directly ex- plicable by the principles previously laid down in the book ; but they may really be looked upon as offering another and a very important example (albeit on a less extensive scale) of the very same modes of phonetic evolution and distribution as those exemplified by Grimm's Law itself, and therefore as a valuable addi- tional support to the line of reasoning by which an explanation of the Law has been attempted. I beg leave to state, in conclusion, that in em- PREFACE. ploying the title of " Grimms Law " I do not intend to express an opinion in favour of the great German's right of paternity in the Law. My own feeling is rather one of regret that Eask's prior claims are not more generally acknowledged. For this is a case in which the Hesiodic paradox holds good, that the half is more than the whole ; in other words, it was a greater achievement on Eask's part to demonstrate, in the first instance, the relationship between the Clas- sical and L.G. Mutes, than on Grimm's part to elabo- rate and extend the demonstration afterwards. But I am afraid it is now too late to make a change. The title I have adopted has taken possession of the ear of the world of letters, and will not very easily be superseded. London, July 1876. THE AUTHOR. ABBREVIATIONS. A-S. = Anglo-Saxon. L.G. CI. = Classical. M.H.G.: C.G. = Comparative Grammar. N.H.G. : Corap.= Compendium. O.H.G. = D. = Deutsch -e, -er, &c. O.N. D.G. = Deutsche Grammatik. ^ Gesch.= Geschichte. G.C. = Grammaire Comparee. Spr. H.G. = High German. Y.G. I-E. = Indo-European. W-B. Li-Sl. = Lithu-Slavonic. Z-S. ={ Low German. Middle High German. New „ „ Old Old Norse (Icelandic). Old (Ecclesiastical) Slavonic. Sprache. [matik. Yergleichende Gram- "VYorterbuch. Zeitschrift. The Titles of Books are mostly abbreviated ; but none of the references will offer any difficulty, unless it be " Whitney (Jolly)," which is put for the almost interminable title of Dr. Jolly's German adaptation of Prof. Whitney's well known Lectures. CONTENTS. Section Page 1. Importance of Grimm's Law: an explanation of it still wanting 1 2. The principal I-E. Mute-systems formulated; their "Cyclo- functional " relationship 2 3. Application of the Principle of Sufficient Eeason to the Re- sults of § 2 6 4. The Historical Hypothesis of the Law irreconcilable with those Results * 6 5. The "Principle of Least Effort"; Grimm's Law to be brought under it 7 6. Relative strength of the Mutes adjusted 9 7. Discrepancy between the Historical Hypothesis and the Principle of Least Effort 11 8. Some attempts to remove this Discrepancy examined .... 12 9. Imaginary "Contact" of the Germans with alien races . . 15 10. Suggested Contacts of the Old Germans with the Kelts and with the Einns tested by the known effects of Contact upon Phonology, Inflexion, and Vocabulary 17 11. Discordant views respecting the first supposed " Substi- tation " 22 12. Curtius's Theory of Aspiration 23 13. His modes of accounting for the other supposed Substi- tutions 24 14. An asserted Reason for the supposed Initial Substitution examined 25 15. Incongruities involved in the Assumption that CI. A be- came L.G. S 29 16. C\irtms^a *^ Unterscheidungsirieh" deaciihed and exa,mmed, 30 XIV CONTENTS. Section Page 17. This imaginary principle undiscriminating, partial, and un- supported by observation 33 18. Compensation: "Cross Compensation" and its varieties investigated 36 19. Summary of the chief features of Cross Compensation; Curtius's " Eeciprocal Compensation " inconsistent therewith 40 20. Attempt of Mr. Sweet to explain Grimm's Law by Cross Compensation : how and why unsuccessful 42 21. Rationale of Cross Compensation 44 22. Physical Illustration of the process ; resume of its -Condi- tions and Phenomena 46 23. Eeturn to the Conclusion from Sufficient Reason : this Con- clusion virtually accepted by Prof. Max Mtiller 49 24. His general hypothesis, how far true 51 25. The Tenues the only Primordial Mutes 53 2Q. Bearings of this Proposition 55 27. Light thrown by it upon the Cause and Purpose of the I-E. Mute-changes .... 57 28. Fertile Phonetic Yariation : Eesults of its Action in two commingled Dialects fojmulated 59 29. Pertile Yariation as exhibited by three Dialects formu- lated 62 30. Sterile Yariation formulated 63 31. Illustrations of Fertile Yariation in primitive times 64 32. Dialectic Place of Origin of Aspiration 67 33. Effects of Aspiration in L.G. compared with its effects in cognate dialects 69 34. Bearing of §§ 32, 33 upon the quality of the original Aspirates 71 35. Occurrence of Mediae for the CI. Aspirates in other than the L.G. dialects. (See also Appendix G.) 72 36. Dialectic Place of Origin of the Media 76 37. The Symmetrical Distribution of the related Mutes, and the Dialectic Conditions on which it depends 77 38. Example of such Distribution in the case of two related • sounds in two dialects 79 39. Minute examination of the same example 81 40. Summary of Phonetic Processes involved therein 83 41. Non-essential Conditions and Phenomena involved ...... 84 CONTENTS. XV Section • Page 42. " Reflex Dissimilation " defined : other Definitions 86 43. Extension of Reflex Dissimilation to the ease of three re- lated sounds in three dialects (Grimm's Law) 87 44. Remarks and Explanations 90 45. Contemporaneous origin and spread of the Aspirate and the Media 92 46. Explanation of dual (Li-Sl. &c.) Mute-systems. (See also Appendix G.) 94 47. Retrospect and Prospect 97 48. Principal forms of Irregularity in Phonetic Variation and Distribution 99 49. Systematic Investigation of the General Principle under- lying the Exceptions to Grimm's Law 102 50. Remarks and Explanations 108 51. The External Evidence supposed to be in favour of the Historical Hypothesis 110 52. Re-examination of the Dacian words preserved by Dios- corides 113 53. Historical and Phonetic Characteristics of O.H.G. exa- mined 121 54. Explanation thereof, and of the Phonetic Difficulties of N.H.G., on the Theory of this Treatise 126 55. Proposed extension of this Theory to the Solution of other linguistic Difficulties : e.g., the phenomena presented by the primitive I-E. h 134 56. Dr. J. Schmidt's Interpretation of those Phenomena .... 136 57. Dr. Pick's Hypothesis on the subject stated 138 58. Objections to Pick's Hypothesis 140 59. M. Havet's Hypothesis stated 144 60. Objections to Havet's Hypothesis 145 61. Application of the Principle of Reflex Dissimilation to the same Problem 149 62. Dialectic Place of Origin of the Labial and of the Sibilant Affection of the I-E. h : Ethnic and Phonetic Explan- ations 154 63. Relative Chronology of those Affections and of the greater so-called Lautverschiehung 159 64. Attempt to connect the ^-Affections with the Evolution of i and u from a 169 XVI CONTENTS. APPENDICES. Page A. — On the Affinity of B in English for the open J. -sound . . 176 B. — On some current Opinions respecting the Indo-European Aspirates 181 C. — On the Extension of Language by Phonetic Yariation. (Opinions and Illustrations.) 185 D. — On Mr. Sweet's Scheme of the " LautverscMehung" 189 E. — On the possible Parent Forms of certain related Indo- European Roots 205 F. — On Grassmann's Treatment of some Exceptions to Grimm's Law 209 G. — On the interior Mediae for Aspiratse in Latin 214 Adde:nda et Cokrigenda 220 Index 223 CA]JF()j>xf A GEIMM'S LAW: A STUDY. 1. — Next in point of philological importance to the revela- tion of Sanskrit to the scholars of Europe_, Curtius justly places the discovery of that phonetic relationship between cer- tain Indo-European languages whose expression in set terms is known as ^^ Grimm's Law/^ For not only did this dis- covery at once enable men to demonstrate, with almost mathe- matical precision, the real identity underlying the apparent differences in the vocabularies of those languages, but it led the way to the investigation of the whole body of Indo-Euro- pean articulate sounds ; the systematized results of which in- vestigation, under the name of Phonology or Sound-Lore, serve in turn as an indispensable basis for the entire super- structure of Comparative Grammar. Nor is this all; for while, on the one hand, Grimm's Law has proved for the philologist a potent instrument of Scientific Inquiry, on the other hand the concise formulation of which it is suscep- tible, and the ease with which it may be applied on a small scale, have rendered it perhaps the most popular and widely known of all linguistic inductions. It supplies matter now- a-days for a section in many of our more pretentious English Grammars. It is even becoming a choice article in the regu- lar stock-in-trade of Examiners ; and boys and girls in their teens are sometimes required not merely to state, but to e^- plain the Law in question ; that is, either to show how the phenomena which it summarizes are reconcilable with known linguistic processes, or else to establish some hitherto unob- B A STUDY. t§2(«)' served principle to which they ought to be referred. Now this is the very thing that is wanting ; for^ in the opinion of competent judges, the construction of a satisfactory theory of Grrimm^s Law is a problem which has long been_, and still is_, awaiting solution^ But as none of the many original inves- tigations called into being by such requirements have, so far as I am aware, ever been made public, it is possible there may yet be room for the few hints which I here intend to offer. 2. — (a) This widespread knowledge of the leadrng facts in- volved in the problem before us, be it deep or be it shallow, is so far to the advantage of an inquirer after principles as to allow hiia to dispense with a detailed statement of those facts ^. In this place, therefore, and simply as a provision for future reference, I shall only formulate the Law as concisely as possible. !Por such formulation, as well as mnemonically, no nomenclature is, in our language, so convenient as that which describes the three main classes of Mute-consonants as Hard, {k, t, p)i Soft {ff, dj b,) and Aspirate {kh, th, ph, or gh, dh, bh). It is quite immaterial what objection may be urged against these designations on abstract or physiological grounds. They will be employed here as symbolic rather than descriptive. Indeed, I shall for the most part put aside the complete words, and, taking ftierely their initials, H, S, A, manipulate these pretty much as if they were algebraical sym- bols, — without, however, precluding myself from having re- ^ ' ' Die Lautverscliiel3img ist ohne Frage die Ibedeutsamste lautliche Tliat- saclie in der Gescliiclite des Sprachenlireises in den unsere Miitterspraclie liineingeliort, und bildet zugleicli eine der allermerkwiii'digsten und schwierigsten Ersclieinungen ihrer Art, deren Erklarung die Spracliwisseu- schaft zii untemelimen hat. Auch ist his jetzt noch keine hefriedigende Erkliirung dafur (jefundenP — Whitney (Jolly), p. 155. " One of the most important problems, which now awaits solution, is to explain the causes of that regidar shifting of sounds which words undergo in different cognate languages." — Sayce : " Principles of Philologv," p. 47. ^ Such a statement will be found in Baudry's '^ Grammaire Compart" pp. 131-150 ; or Max Miiller's Lectures, Second Series, Lect. v. : and a more concise summary in Monis's " Historical Accidence," Chapter ii, j or Earle's " Philology of the English Tongue," § 1 . § 2 (c).] GRIMM^SLAW: A STUDY. 3 course to other symbols or nomenclature if occasion should require it. . (b) Now the phenomena summarized in Grimm-'s Law stand out most strikingly upon a comparison of three prin- cipal Mute-systems. One of these^, shared in (with certain known variations) by Sanskrit, Zend, Greek, and Latin, may for brevity be called the Classical (CI.) system. The second, of which Gothic is generally taken as the best representative, but in which the Old Norse, Old Saxon, Anglo-Saxon, and other dialects shared, may be called the Low German (L.G.) system. The third, as characterizing High German tribes, may be called the High German (H.G.) or (inasmuch as it ap- proached more nearly to complete regularity a thousand years ago than it does now) the Old High German (O.H.G.) system. Other Mute-systems, as for example the Lithu-Slavonic (Li-Sl.) and the Keltic, which partake of the characteristics both of the first and the second, are really not less important than these ; but for the present they may be left out of sight. As regards the three first-named systems, Grimm^s Law requires that a given mute in any one shall be represented in each of the others by a different mute of the satne family, whether gut- tural, labial, or dental, as the case may be. Thus — Classical H = Low German A = High German S ; A— S — W. }} ** — )y a " — jj i) *** S— TT— A . or, grouping each system^ horizontally, — If the Classical system is H A S or A S H or S H A, the corresponding L.G. is ASH or S H A or H A S9 the corresponding H.G. is SHA or HAS or ASH. (c) That these three tabulations are of identical value, or severally represent precisely the same set of facts (differing only in order of sequence) , will be seen by comparing the ver- tical columns of any one with those of any other. It is conse- quently quite indifferent which of them be taken to symbolize ^ In order to avoid turning aside into explanations and qualifications, I assume for the present that all the three systems are ideally perfect. Their deviations from regularity will he considered at the proper time. B-2 4 GRIMM^S LAW : A STUDY. [§ ^ (^ • the Law. For our own convenience, however, it is desirable to select some one as a standard for future reference : let us therefore take the first. This done, a glance will detect the remarkable symmetry which characterizes the relationship be- tween the three systems. As the eye passes down the lines of initials, each succeeding system appears to be derived from the foregoing one by precisely the same amount and precisely the same kind of change : in other words, whatever phonetic operation, as it were, is executed upon the CI. system to produce the L.G. system, must also be executed upon the L.Gr. to produce the H.G. system; and (what is equally important, but is rarely if ever made prominent) the very same operation, when executed upon the H.Gr. system, brings us round again to the CI. system. That is, descending to the individual sounds, if we pass (in the horizontal lines) from H to A and from A to S in any one system, such transition uniformly requires a corresponding transition from A to S and from S to H in the representative sounds of the following system. And a similar phenomenon presents itself if we read the horizontal lines from the lowest upwards ; or, again, if we read them alternately {i. e. from CI. to H.G., from L.G. to CI., and from H.G. to L.G.). Thus, take them in what order we may, each system regularly varies with each of the others ; so that, to borrow the language of Mathematics, each system may be called sl function of each of the others ^ (d) This will allow us to represent the systems and their relationship with much greater brevity than can be done even by the above tabulations. For as each system is made up of the same phonetic quantities, so to say, as the others {viz. H, A, and S), although in a diflPerent order, this general agreement may be indicated by using one and the same sym- bol (say '2, = Xv(TT7]fjLa) for them all; while their variation by sequence, and undetermined priority, may be indicated by the letters x, y, z. Putting therefore S^, S , 2^, for the three sys- ^ Tlie analogy is perfect : compare Todhunter's Definition : — '< Suppose two magnitudes which are susceptible of change so connected that if we alter one of them, there is a consequent alteration in the other, this second magnitude is called a, function of the first." ("Diff. Calc." tnit.) §2{e).] tems^ we see by the relationsliips just now described, that 2^ (whichever it may be) is a " function '"' both of 2 and S^; 2y both of S^ and S^ ; and S^. both of S"*' and X^. Thus, taking the systems in succession, and putting / for the func- tion which 1^ is of S , we have the series of equations — 2.=/(2;;2^=/(2,); S,=/(SJ; ... (a) which leads to a repetition of the whole series. Again, alter- nando, putting for the function which 2^ is of S^ we simi- larly have — _ S,=<^ (SJ ; t= (SJ ; t=

(Sj, and so on), the second series of equations does not really give a new value of any one system in terms of any other, but merely represents a back-reading from 2 to S^, &c. ; so that, instead of ^^ alternando/^ we may equally well read ^' invertendo ; '' (3) that, as far as phonetic relation- ship is concerned, it is perfectly immaterial, and only a subject for arbitrary convention, which system shall be called the first; for 2^ is evolved from ]S^ by precisely the same amount and kind of variation as 2 from S and 2 from 2^. Conse- quently the following single tabulation might be substituted for all the three tabulations in subsection (6) above, of which indeed it is but another reading : — CI. or L.G. or H.G. system = H A S = S^; L.G. or H.G. or CI. „ = A S H = :2^ ; H.G. or CI. or L.G. „ = S H A = 2^. And, lastly, by fact (2) it is immaterial whether the series of equations marked (a) or that marked (/8) be taken to represent the relationship between the three systems. A STUDY. [§ 3. 3. — The principal point brought out in the preceding section is what may be called the Cyclo-functional relationship between %^j Sy, and 2^. Applying to this the Principle o£ Sufficient E,e,ason_, we arrive at the conclusion that no one of the three systems so related may, in preference to either of the others, be assumed as the normal and primitive system, from, which the others are only deviations; for it is manifest that no reason can be urged for or against the priority of one system which may not be urged with exactly equal force for or against the priority of each of the others. The only alternative is, as Max Miiller says, that ^^none was before or after the other ^'' ^, to which may be added that ^^ none is greater or less than another : ^^ that is, in short, none is anterior to the others in time or superior in importance. This negative conclusion stated in positive terms means that the evolution of 2^, 2 and 2^j was the result of simultaneous or contemporaneous phonetic action ; and I shall hereafter show that that action, as between each system and each or both of the others, must also have been mutual and reciprocal. Hence the symbol 2^ (as was shown in § 2 (e)) belongs as much to one system as to an- other ; and so do similarly 2 and 2^ : so that the symbols we have selected are in danger of losing their value through vagueness of application. In order, therefore, that they may be conveniently employed, it is desirable to arrange them ar- bitrarily in some definite order. Let us then agreB that 2, shall denote the CI., 2 the L.G., and 2^ the H.G. system ; no harm will be done, provided we bear in mind all along that a?, y, z indicate, not a necessary, but merely a conventional order. 4. — By the conclusion reached in the foregoiug section we are brought face to face, in point-blank opposition, with the prevalent hypothesis of Grimm's Law. This may be called the Historical or Chronological Hypothesis. It was laid down by Grimm himself, and has been adopted with slight variations by Curtius and most of the leading Continental Philologists. ^ Lect. v., Second Series; where, however, the statement has too much the aspect of a mere dictum. § 5 («).] GRIMM^S LAW : A STUDY. 7 It supposes that the CI. system (our Sj was the only pri- mitive and original phonetic system^ shared in by the ances- tors of all the Indo-European nations alike; that the L.G. system (2 ) was^ in point of time^ a subsequent deviation from %^; and the H.G. (2^) a subsequent and precisely similar deviation from % . It necessarily leaves completely out of sight the functional relationship between 2^ and %^ {viz. S^ =/(Sj) ; for to suppose that ^^ in its turn was a subsequent and similar deviation from ^^^ would obviously be such a ghiring reductio ad absurdum as would at once overturn the whole hypothesis. But as the cyclo -functional relationship between the systems is a patent fact^ the said hypothesis is driven to assume a successive series of phonetic move ments of such a kind that that relationship is accidentally caused to simulate contemporaneous and reciprocal action. It cannotj however,, be^ that the phonetic movements under consideration were simultaneous and yet successive — reci- procally or mutually related^ and yet totally independent. Of propositions thus irreconcilably hostile one or the other must give way. But before we^ on our side^ relinquish in favour of the Chronological Hypothesis the conclusion of § 3 w;hich was laid down as our starting-pointy it may be worth while to inquire whether the agreement of that hypothesis with ascertained linguistic principles and the internal harmony of its own leading features are so complete as to justify our sur- render. Such an inquiry will^ at any rate^ have the advantage of directing attention, not only to the difficulties involved in our main problem, but also to the fundamental processes in accordance with which a solution of that problem ought to be effected. 5. — («) We have to remember, at the outset, that Grimm's Law does not express any primary and independent principle of language, but merely a particular case of Sound -change. The case, as regards its form, is certainly very complex. In this respect it may even be, and as far as we know is, quite unique ; but its nature nevertheless remains unaltered. It should therefore take its place in the same rank as other 8 grimm's law : a study. [§ 5 {b) . cases of Sound-change. It should be subjected to the wider law^ or principle under which they are generally reducible. And the first thing we should expect to find in any Theory of Grimm^s Law would be an attempt^ at any rate, to harmon- ize the two. (b) The wider law or principle just referred to is that which, following French writers on the subject, I shall call the " P rinciple of Least Effort.^^ The sway of this principle really extends far beyond the realm of language; for it is hut another name for the general tendency of mankind to make all labour as easy to themselves as possible. In its lin- guistic application it has been so well put by Professor Whit- ney that, to avoid stating it worse, I will venture on a some- what hackneyed quotation. '^AU articulate sounds,^^ says he, ^^ are produced by eflPort, — by expenditure of muscular energy in the lungs, throat, and mouth : and this effort, like every other which man makes, he has an instinctive disposition to seek relief from, to avoid^^^. This tendency, or '^ instinctive disposition,^^ is as universal and continuous in its operation ^ The word " law," whose associations used to be only noble, has been so hackneyed of late bj writers on philological and other topics, in order to give factitious dignity to all sorts of trifling rules, that one hardly recognizes an old friend when occupying a position of importance. ^ I quote the following version for the sake of Dr. Jolly's concurrence : — " Die erste und wichtigste Erscheinung .... die Ursache fast aller Laut- veranderungen. . . .ist jener Trieb des Menschen . . . . das Streben, den Sprachorganen die Sache leicht zu machen, — die schwerer sprechbaren Laute und Lautverbindungen durch bequemere, weichere zu ersetzen, und alien unniitzen Ballast in den Wortern iiber Bord zu werfen. Alle arti- culirten Laute werden mit einer gewissen korperlichen Anstrengung her- vorgebracht, indem dabei die Muskelthatigkeit unserer Lunge, unserer Kehle, und unseres Mundes, in Anspruch genommen wird. Diese, gerade wie jede andere Anstrengung, sucht sich der Mensch, kraft eines natiir- lichen Instincts, vom Halse zu schaffen, cder doch zu erleichtern : eines Instincts, den man nach Belieben als einen Ausfluss der angeborenen Tragheit oder der Sparsamkeit, d. h. des Selbsterhaltungstriebes des Menschen betrachten mag ; er fliesst in der That bald aus der ersteren, bald aus der letzteren Quelle, je nach den Umstanden; er ist Tragheit, wenn dadurch mehr verloren als gewonnen wird, — weise Sparsamkeit wenn der Gewinn die Einbusse iibersteigt. Kein TuUelchen, kein Jota 4 6 [a) .] GRIMM^S LAW : A STUDY. as the tendency of water to run down hill. To it may be re- ferred nearly all the phonetic changes which have taken place in the Indo-European languages within historic times ; and as there is no reason to suppose that human nature (or_, to be safe, let us say, our variety of it) was ever different in the main from what it still is, we may fairly conclude that the same principle began to operate as soon as there was any linguistic material to operate upon. 6. — (a) An important division of Phonology, therefore, is that which concerns itself with the relative ease and difficulty (where these are ascertainable) of both vowels and conso- nants j and which demonstrates that, apart from counteracting causes, the course of phonetic change uniformly proceeds from the harder to the easier sound. Thus Bopp, Corssen, and others make it an important preliminary to their investigations to adjust the weight or strength of the primitive (or classical) vowels ; and to show that while the heaviest or strongest, a ( = English ah) J may descend to u { = 00) and i ( = ee), neither i nor u ever reascends (sua sponte) to a. From this and similar considerations Grimm ^, Leo Meyer, and others have hazarded the bold conjecture that, at a very remote period in the his- tory of the parent Indo-European Speech, a was the only vowel ; and Tick (Worterb., Nachwort I.) has attempted with considerable success to raise that conjecture to certainty^. kann sich dem Eirvftuss dieses Triehs entziehen, imd er kommt auf die mannigfacliste Weise zum Durchbruch." — Whitney (Jolly), pp. 105-6. So too M. Baudry : — " En resume, comme il arrive pour tout acte hu- main, le langage livr^ a lui-meme tend a s'exercer avec la moindre action j ou, ce qui revient au meme^ avec Taction la plus commode possible." — G. C, p. 85. I add but one more statement : — " AUe Veranderungen der Laute, die im Verlaufe des spracblicben Lebens eintritt, ist zunachst und unmittelbar Folge des Strebens, unseren Sprachorganen die Sache leicbt zu macben ; Bequemlicbkeit der Ausspracbe, Ersparung an Muskelthatigkeit ist das hier wirkende Agens." — Schleicher : Die D. Spr., 3rd ed., p. 50. 1 ^'Wiederum ist von den drei vocalen A der edelste, gleichsam die mutter aller laute, aus dem zimachst J und ?7 hervorgegangen sind." — Gesch. der D. Spr., p. 274 (ed. 1848). 2 He states (somewhat too unhesitatingly perhaps) the results of his 10 grtmm's law : a study. [§ 6 {b), (b) In the case of tlie Mute-consonants the strongest family are the gutturals ; next come the dentals ; and lastly the labials. Gutturals, therefore, by the same principle may descend to dentals or labials; but, as a rule, neither dentals nor labials may reascend to gutturals. The line of debilita- tion for the different families of mutes, and mostly for the vowels, is mainly determined by the simple physical rule, that sounds produced in the fauces require the greatest muscular effort, and that such effort diminishes as the sounds are pro- duced further and further along the vocal passage towards and up to the lips. (c) But it is with the members of the separate families that we are now more particularly concerned j and here, in- asmuch as all the members of each family are necessarily pro- duced at or about one and the same point of the vocal passage, a different rule must hold good. The comparative strength of these is determined partly by the energy and partly by the completeness of the contact of the vocal organs at the point of production. The difference between the Hard and Soft mutes (Tenues and Mediae) resolves itself roughly into the greater or less energy or muscular tension* with which they are p'roduced ; so that their relative strength is well enough indicated by their names. The Aspirates seem to have differed originally from the Pure {i. e. unaspirated) mutes by giving way very slightly — and, except to a fine ear, perhaps imper- ceptibly — to the vis a tergo supplied by the puff of breath which, in the utterance of the Pure mutes, was completely arrested^. Among many of the Indo-European peoples this aspiration soon advanced to a very perceptible interruption. investigations as follows: — ^^ Sammtliche Wurzein auf i sind jiingere Formen von solchen auf a " ; and again, *' Ebensowenig wie i ist u ur- spriinglich, sondem durchweg aus a, in einigen Fallen audi aus va, eutstan- den." Hence in explaining Vowel-strengthening he postulates a time "alsder a- Vocal noch allein herrschte." — W-B., pp. 950, 951, 95G. ^ " Czermak, by using his probe, found that hard consonants drove it up much more violently than soft consonants." — Max Miiller: Lect. iii., Second Series. ^ See Appendix B. §7.] GRIMM^SLAW: A STUDY. 11 or rather prevention^ of complete contact; and the result^ under conditions of uniform development, at last yielded the three spirants f, th, and h, which are reckoned among the feebler sounds, — the last, indeed, being in the lowest^stage of debilitation. {d) The Holethnic ^ language is generally thought to have possessed, like the Greek, only one aspirate of each family, — whether a hard or soft aspirate has been vehemently con- tested ^ : a complete family, however, should possess two, — both a hard (A) and a soft (A'). These four, arranged in pairs according to their energy, stand thus : ff ,S , — ^4, A'; and, ac- cording to their completeness of contact, thus : H, A — S_, A'; A and S therefore maybe considered physiologically as debi- litations by different agencies, and oii opposite sides, so to say, of H; they start from it on divergent lines of descent; they are naturally deducible from it alone, and not from each other; and, in fact, like dissimilar quantities, they are hardly com- parable. A', however, may be regarded as producible in two ways, — either by the softening of A, or by evolution from S in precisely the same way that A is evolved from H. In re- presentations of the primitive alphabet the order generally given is H, A, S ; an order which seems to imply that the primitive Aspirate was the hard Aspirate. That it really was so I have no doubt : the inconsistency lies in this, that in the prevalent opinion it was the soft Aspirate. But it is not now essential for us to discuss more minutely the relative strength of the mutes of each family : enough for the present if it be admitted that both A and S, or at any rate and especially the latter, are weaker than H. 7. — If we test the Chronological Hypothesis of Grimm^s Law by this wide-ruling Principle of Least Effort, we shall ^ I shall venture, for brevity, to call the primitive undivided I-E. people the "Holethnos " (to okov Wvos=gens integrd) ; whence the adjec- tive "Holethnic " by correct derivation. The term " Proethnic", sometimes employed, besides being destitute of a related noun (for to talk of a "Proetli- nos " would be nonsense)^ implies that the primitive people were not them- selves an ethnos or gens. '' See § 34 infra. 12 GRIMia's LAW : A STUDY. [§ 8 (fl). find at each supposed phonetic evolution a glaring example of disobedience to the latter on the part of the former. For the S of ^/ is represented by H in S , and again the S of ^ by the H of 2^; that is_, every CI. Media must, by the hypo- thesis^ have become a L.G. Tenuis,, and every L.G. Media an O.H.G. Tenuis. Of the total number of mutes involved in the CI. vocabularies, between 20 and 30 per cent.^ are Mediae ; and these, by the hypothesis, must have become Tenues during the evolution of S^. The L.G. mutes exhibit pretty nearly the same percentage of Mediae (supposed to be derived from the CI. Aspiratse) ; and these in turn must have be- come Tenues during the evolution of %^ : so that, altogether, about half the total number of mutes \^hich form the common property of the I-E. peoples must^ by the same hypothesis, have undergone^ at one stage or the other, a change directly opposite to that required by the Principle of Least Effort. 8. — {a) This incompatibility of the Chronological Hypo- thesis with the principle just named has not escaped the keen perception of the many able men who have adopted that hy- pothesis; and it becomes interesting to discover how they propose to remove it. Here and there a straightforward writer finds no other method than that of impeaching the universal action of the debilitating principle itself, — a method which merely amounts to a confession that his hypothesis cannot be brought into harmony therewith. There is besides a prevalent but very loose way of talking of the successive substitutions as if they were the result of the inherent nature, or of a neces- sary tendency, of the individual mutes of each family. Several remarks of Grimm himself seem to countenance this notion ^ ; so also do the descriptions and '^ schema ^' of Schleicher '*. It is scarcely to be wondered at, therefore, that the retailers of ^ See the final tabulation of § 2 (e). ^ More exact statistics "will be quoted presently (§ 14). ^ " In der media liegt gleichsam seine naturliche kraft [i.e., that of the Lautverschiebung], die sich zur tenuis verdiinnt und hernach wieder zur aspirata verdickt : aus der aspirata muss darauf die einfache media abtropfen, und dann der umlauf neu heginnenJ'^ — Gesch., p. 416 (et sic alibi). * Die D. Spr., p. 97. §8(c).] GRTMM^SLAW: A STUDY. 13 oracular utterances speak with greater boldness on the sub- ject K {b) All this is^ in fact, only the virtual assumption of a prin- ciple precisely the opposite of the Principle of Least Effort. It is as if one should declare that, because water is found at different altitudes, therefore half the rivers of the world must once have run up hill. Under special circumstances, it is true, and on a comparatively small scale, water may be urged above its ordinary level, — as, for example, by force of wind, or by me- chanical means ; and similarly may sounds be ^^ raised '' by some extraneous influence, such as the force of accent or em- phasis or the ^^ attraction '^ of a neighbouring sound ^. But in the latter case, no less than in the former, we look about for an influential agent ; and if none is at first discoverable, it is hardly consistent with scientific caution to declare that none exists. The leading supporters of the Chronological Hypothesis have evidently felt this, but have not always been content to wait for a suitable explanation. Indeed, the too hasty attempts to grasp at any means for propping up this weak side of the hypothesis have led to some of the wildest, I had almost said the drollest, writing upon the subject. (c) In this category (but with many apologies) may we not venture to place the supposed Muth und Stolz {'^ pride and pluck ■'^ as Max Miiller translates it) of the ancient Germans, to which Grimm is inclined to attribute the supposed ver- schiebung^'? In such a conjecture we may recognize, with Baudry, an amiable and simple patriotism; but the great Teutone could hardly have been unaware that his explanation suggests too much; for, as Curtius remarks, with some humour, if we admire the courage of our ancestors in ^' raising '^ some ^ E.g. : — " . . . What, for example, was a j!? in the original form of a word, or at least in the oldest form known, is found at a later stage transformed into/, which next passes into h ; and this again tends to be- come p, and go through the cycle anewT — Chamhers's Encyclopaedia. ^ See Appendix A. ^ " Liegt nicht ein gewisser muth und stolz darin, media in tenuis, tenuis in aspirata zu verstarken ? " — Gesch., p. 437. 14 GRiMM^s LAW : A study/ [§ 8 {d) . o£ the mutes, we should equally bewail their pusillanimity in '' lowering '^ the others ^ (d) Closely akin to this conjecture is Grassmann's visionary Heroic Age, which he would elevate far above the reach of the great phonetic laws to which ordinary men are subject. What Grimm, however, modestly suggests in the form of a question, Grassmann decides sans if or but ^ ; although he is equally forgetful to explain how his potent heroes could at once conquer their own innate physical tendencies (as in the change from S to H) and be conquered by them (as in the other and fourfold more extensive changes, — H to A and A to S). All, therefore, that the hypothesis gains is a new element of bewilderment. (e) Any one who thinks it worth while may easily test the value of such fancies by selecting the most heroic age of the proudest people known to history (for there have surely been as brave men since as there were before Agamemnon) and investigating the amount and kind of corresponding variation in the phonetic character of the language of that people. Neither Grimm nor Grassmann has ventured to apply this simple test; and whoever does will find that no variation contrary to the ordinary natural rule is anywhere^ during any such period, discoverable. The great deeds and series of deeds that we connect with exceptionally heroic ages are generally achieved in comparative silence ; and where men have used language as an auxiliary, their mind and soul have stamped themselves, not upon their alphabet, but upon their stijle. Now we have no other safe guide to what might or should have been than what has been, — to the unknown than the 1 '' Wenn wir den mutli unserer vorfahren in der erliebung von d in t, t in th verehrten, so miissten wir ihren kleinmuth in der senkung von dh in d bedauern."— Kulin's Z-S., ii. 330. ^ " Es gilt diese regel [our Principle of Least Effort] docli nur fiir solche perioden der entwickelung, in welchen die urspriinglicbe scbopferiscbe kraft erscblafile, und einer allmablicben entartung und verweicblicbung des volkslebens und damit aucb der spracbe platz macbte, .... nimmermebr aber fiir eine zeit energiscber kraftentwickelung, nicbt fiir die beldenzeit eines volkes," &c., «&c.— Kubn's Z-S., xii. 100. § 9 {b).'] GRIMM^S LAW : A STUDY.* 15 known. A belief in the uniformity of nature is as essential in Comparative Philology as in any of the natural sciences. To refer phenomena to a totally different order of things^ supposed to have prevailed at an epoch which lies far beyond the reach of investigation, is not to explain difficulties, but only to slur them over ; it is to trust for the acceptance of a doctrine to the possible absence of evidence to the con- trary. Such conjectures, therefore, as those just quoted merit only so much attention as respect for their authors may prompt. 9. — {a) Other adherents of the Chronological Hypothesis have naturally been dissatisfied with these attempts to explain the discrepancy under consideration, and have looked about for some external agency which may account for it. One favourite resource of these is an imaginary action upon the primitive German tongue of some foreign race with which the German tribes may have come into contact. Thus M. Baudry ^, in combating a different opinion, remarks : — '^ Nous aimerions supposer que la confusion n^existait pas au com- mencement, mais qu'elle s^est produite h un instant donne, par suite d'une circonstance perturbatrice, telle que le contact cVune race etrangere/* {b) Now, that the Germans, as well as the other divisions of the Indo-Europeans, may have come into contact with alien races, is highly probable. But when we consider the thoroughness of the phonetic changes to be accounted for, this explanation must be stretched to imply very much more than mere contact. Even extensive absorption, as the result of conquest, would, so far as our observation reaches, not be too much. Against this, however, the arguments seem to me overwhelming ; but as it is hardly worth while to occupy much space in discussing the suggestion in its present vague and general form, I will merely point, by the way, to the singular homogeneity and strongly-marked individuality of the German dialects. High and Low alike. Except the English, which has been disintegrated during 1 G. C, p. 146. 16 GRIMM^S LAW : A STUDY. [§ 9 (c) . comparatively recent times, these all iu every department — Grammar, Syntax, Vocabulary, Idiom, Phonology — convey the impression of uniform development under the guidance of a well-marked type of mind accustomed, in the expression of its thoughts, to rely above all things upon its own native re- sources. And even the English — which has been subjected to a succession of "contacts,^^ and, indeed, to convulsions of a violence and duration that the advocates of *^ contact ^^ would scarcely postulate in the case to which they apply it — shows how, notwithstanding all modifications of grammar, loss of inflexion, and disintegration and renovation of vocabulary, the Mute-system of a language maintains its place. Among the Vowels, however, (to judge from the history of the English vowel-system,) we should expect to find extensive traces of such contact. On the contrary, we find the primitive vowel- system maintained by the Germans with singular purity ^ Among the conditions, therefore, to be satisfied by the alien race, it appears that their vowel -system must have been identical with that which the primitive Germans brought into Europe, while their mute-system must have differed in such a remarkable way as, by " contact ^^ with the German system, not to corrupt it, but to transmute it, and not merely to transmute it, but to transmute it symmetrically with respect to the primitive or classical system with which it was, by the hypothesis, previously identical. (c) But this is not all : upon one improbability is to be superimposed another still greater ; for so far we have only taken the first (L.G.) supposed substitution into account: the second (H.G.) yet remains to be explained. And if '*^con- tact^^ is to account for the former, it should also account for the latter ; or else the suggestion, being only applied to one ^ See the Gothic system in Sclileiclier's Comparative Vowel-tabulation, Compend., pp. 156, 157 ; to which may be added the following from "Die D. Spr." (p. 91): — "Die hohe lautliche und formliche schonheit, die das Deutsche auszeichnet, und die, was das wunderbar lebendig erhaltene, ja weiter als in der Ursprache entwickelte vocalsystem betriffl, von keiner andem indogermanischen sprache erreicht wird, hat das Gotische am treuesten und reinsten erhalten, obwohl kein deutscher sprachzweig dieser vorziige vollig entrath." § 10 (a) .] GRIMM^S LAW : A STUDY. 17 phonetic movement, is worthless. Now in the second case we must suppose contact with another alien race different from the first; for as the German mutes had already, by hypo- thesis, been adjusted to those of the first race, it would be absurd to suppose that that adjustment would be completely perturbed by a continued or renewed contact with the same race. Nevertheless we have here again to explain the con- tinued existence of a vowel-system of remarkable purity (for the Oldest High German in this point differs very little from the Gothic) side by side with a transmutation of consonants, not merely symmetrical with the L.G. system (from which the H.G. system is supposed to be derived), but symmetrical therewith in precisely the same order and pro- portion as the L.G. system is with respect to the CI. system. The improbability of an accident of this kind happening, as the result of contact, on the top of a previous similar accident really amounts to an impossibility. 10. — (a) But in order that this '^ contact ^^ hypothesis may be subjected to a fitting scrutiny, it ought to specify the language or languages to which the several substitutions are to be attributed : we might then inquire whether the German dia- lects exhibit any of those changes which our observation and experience of linguistic commixture lead us to expect. Ac- cordingly, there are not wanting philologists who have ven- tured to be thus specific. Dr. E. Forstemann, for example^, attributes the substitution of S for A to Keltic, and that of H for S to Finnish, influence. The Germans, in their migration westward, overran and subdued extensive territories occupied by the Kelts, whose phonetic system was destitute of Aspirates. This contact, it is said, led to the deaspiration of all their pri- mitive Aspirates on the part of the Germans. In spreading northwards the Germans similarly overran and subdued lands occupied by the Finns, whose Mutes then consisted of Tenues ^ Gesch. des D. Sprachstammes. See also Prof. March, Compar. A-S. Gram., 1st ed., p. 29, note (a). The latter, however, appears inclined to attribute to the Kelts the two substitutions which are not attributed to them by Forstemann. c 18 grimm's law : a study. [§ 10 (b), only. This contact had the effect of raising to Tenues aU the primitive German Mediae. (b) Here unfortunately terminates the list of specified contacts ; and it is consequently either too short or too long. For if contact was or could have been the cause of the sup- posed substitutions, then, to be consistent, we want a good deal more of it. The two substitutions attributed to it do not cover half the L.G. Mute-system alone (§14 infra) : to account for the remaining substitution (A for H) a far more forcible and effective contact remains to be discovered ; while for the series of H.G. substitutions a fresh series of successive con- tacts is required. On the other hand, if the larger part of the L.G. and the bulk of the H.G. consonants were trans- muted by some other agency than contact, where is the use of imagining the two contacts just specified in order to account for so comparatively limited a part of the whole phonetic movement ? (c) But let us accept the explanation so far as it goes : then, considering the very thorough transmutation effected by each of the two suggested contacts in that section of the mutes which it selected for its operation, we have a right to expect that some of the other and more usual effects of linguistic mixture should likewise be distinctly pointed out. These effects are well known; indeed, they may almost be formulated, in outline, thus : — If two different races, r and r , whose languages are I and /', undergo fusion, the resulting language L lies intermediate to I and Z', and, ceteris paribus, approaches more nearly to Z or T as r, in point of number and importance, is greater or less than /. Of course, if r and r' be nearly re- lated, / and I' approximate in corresponding ratio ; and L will differ less from either I or /' than each of these differs from the other. These results hold good in every department of language ; but to apply and exemplify them fully would re- quire a volume to itself. My time and space allow me only to indicate, as briefly as possible, two or three leading depart- ments in which they ought to appear. (d) And first as to Phonology. Whatever sounds in / and r are exactly alike, can, of course, be readily pronounced by § 10 (e).] grimm's law: a study. 19 both r and r\ But where the sounds of /.(say) differ from those of V , the attempts of r' to acquire them will, in some cases at least (i e. when they offer any difficulty), result in approximations only, — that is to say, in modified or merely imitative sounds which r' can produce with greater ease than the genuine ones. And as the natural tendency is towards debilitation, r will readily stoop to the infirmities of r', until at last some intermediate sounds are universally accepted by both races. The same is true of r' and r in reverse order ; and thus phonetic contact means phonetic compromise ; and one of the most striking results of linguistic fusion is an acce- lerated phonetic debilitation. If, then, the most ancient Ger- man underwent a twofold admixture, the results of which are so striking among certain sections of the mutes, we should naturally expect to find likewise some signs of a general cor- ruption among the whole body of sounds. In point of fact we find nothing of the sort. Even the changes attributed by the hypothesis to the specified contacts are not corruptions at all, but systematic transmutations, one of which (that from S to H) is exactly the reverse of debilitation. {e) Next, as to Grammatical Structure, — or, at least. De- clension and Conjugation ^ Here the prevalent doctrine is that in the Grammar of a language there can be no mixture^. This may be true ; but I am not sure that the proofs are per- fectly satisfactory. In all the favourite examples (modern English to wit) it will be found that there has been a con- siderable preponderance, either of numbers, or of political and social importance, or both, on the side of one or other of the commingled races. We are not yet in possession of any in- duction from a sufficient number of cases in which r and r' were pretty nearly equal in those respects. From what is to be seen where they have been unequal, I am inclined to think that if two dissimilar languages, in what may be called their natural state, could be fused upon precisely equal terms, the ^ I take the case of Inflexional Languages only ; our observation and experience are but just beginning to travel beyond these. ^ See an excellent chapter on the subject in Mr. Sayce's " Principles of Philology." 20 grtmm's law : a study. [§ 10 (/). • inflexions of both would at last entirely disappear^, and totally- new methods (like the French future tense and prepositional noun-cases) would be invented for expressing all grammatical relationships. In this case, it would still be true that there would be no mixture of Grammar; but the possibility of mix- ture would be prevented by mutual destruction. To this state of things the grammar of some modern mixed languages (English itself, for example) offers no very distant resemblance. Nor is the reason past finding out. For the inflexional appen- dages of I mean nothing to r'_, those of V nothing to r ; it is as much as the people on each side can do to catch up the cen- tral word with its general meaning. Soon, therefore, those appendages begin to be peeled away, and their places to be supplied by separate particles and notional words, which are few in number, easily acquired, and useful for many additional purposes. In the matter of Inflexion, accordingly, still more than in Phonetics, contact means accelerated corruption. Consequently, among the Old Germans, on the hypothesis under review, both the Kelts and the Finns should have ini- tiated a period of grammatical decay, and the efi'ect of the two together ought to have been very striking. What is more, the O.H.G., which for consistency's sake (See subsection {b) above) should exhibit the effects of a sixfold contact, ought to have been completely stripped of its inflexions. But the actual state of the facts is just the reverse. In the Oldest L.G. no corrupting agency, beyond that of uniform and cha- racteristic native debilitation, is to be detected at work ; while, more remarkably still, the Oldest H.G. is particularly rich in inflexions, and its descendent dialects are at this day in- flexional to a higher degree than most of their contemporary sister dialects. (/) But the state of the Old-German Grammar and Pho- ^ Or, if there were any remains, I see no necessary reason why / and V should not both contribute thereto. We have an indication of the pos- sibility of such a result in English ; where, although the structure is in the main Teutonic (as it ought to be by (c) above), yet the universal pre- valence of the noun-plural in « is due entirely to French influence and example.— See Earle's "Philol. of the Engl. Tongue," 2nd ed., pp. 360-352. § 10 (g).'] grimm's law: a study. 21 nology proves only that there can have been no extensive commixture of races. For slighter degrees of contact we re- quire another test ; and this is furnished by the Vocabulary, There is scarcely any degree,, however slight^ I will not say of actual contact J but evenof communication^ with an alien people, which is not sufficient for the introduction of some of their words. Travel^ Commerce^ and Education are here as effective as Conquest. English^ for example, abounds in words col- lected from almost every country under heaven; and such words might be arranged in a numerical scale which would roughly indicate the extent of communication with the several peoples from whom they have been borrowed. This test, ap- plied to the suggestions noticed in subsection («), is decisive against them. Forstemann, with every disposition to succeed, fails to produce from the Oldest German a single word that can certainly be traced to the Finnish^; and although, as might be expected, he produces several which appear to have come from the Keltic ^, he is not sure but that even these may have spread at a later period through the leading German dialects from such of the German tribes as are known to have really come into contact with the Kelts. To talk, under these circum- stances, of Keltic and Finnic influence as so unerringly select- ing and transmuting, as if by systematic and concerted action, first the Aspirates and then the Mediae of the Old German Mute-system, is mere trifling. {g) The truth is that the differences between S^,, S , and S^ are of a totally different character from those known to be attributable to the fusion of one race and language with another, and require a totally different treatment and expla- nation. Grimm^s Law formulates neither fusion nor con- fusion. What we must keep our eye on is the harmonious, equipollent, s ym metrical, and functional relationship of all those systems inter se ; and that relationship is adjusted upon a plan of which the interpretation is not to be read in the chapter of accidents. But before we quit this part of our subject, it is pertinent to remark that the experiment of sub- jecting one and the same language to the influence of alien • ' Gesch., p. 610. 2 i^.^ p. 607. 22 grimm's law : a study. [§ 11« races^ both unrelated one to another and of all degrees of rela- tionship^ has been tried for us on a large scale,, and^ as it were, under our very eyes. I refer, of course, to the spread of the Latin tongue among the peoples of Southern and Western Europe, whence has resulted the formation of the Romance languages and their multifarious dialects. Here, if anywhere, we might fairly expect to find, on the " contact " hypothesis, numerous and very striking examples of apparently correlated phonetic action similar to that formulated in Grimm^s Law ; but here we do not find them ; and it will hereafter appear that the condition of independent development under which those languages grew up furnishes the very reason why we do not find them. 11. — In approaching the consideration of the most generally received form of the Chronological Hypothesis we must first notice the discussion as to where, on that hypothesis. Sub- stitution originated, — that is to say, as to which of the three possible changes (H to A, A to S, or S to H,) came first in order of time. And here it is perplexing to find, at the outset, that each of the three has, by some leading philologist or other, been selected as the initial one. Thus Grimm ^ starts with the change from S to H ; and so, with muth, if not with stolz, puts one of the least likely changes in the fore- front. Bopp^, Schleicher^, and others'* take the supposed sub- stitutions in the order H to A, A to S, and S to H. The third and only remaining case possible, which begins with the change from A to S, has, in its turn, been propounded by no less eminent a scholar than G. Curtius^. Regarded from the stand-point furnished by the Principle of Least Eff'ort, the view of Bopp is, as we shall presently see, the least ob- jectionable. As, however, the third view (that of Curtius) appears to harmonize better than the others with some of the * Gesch., p. 344 et alibi. ^ V. G., § 87. 3 Die D. Spr., p. 97 et seqq. * K g., Whitney (Jolly), p. 154; and Grassmann in Kuhn's Z-S., xii. 96. » " Die Aspiraten der I-G. Sprachen : " Kuhn's Z-S., ii. 322, &c. § 13.] GRIMM^S LAW : A STUDY. 23 phenomena to be explained, as it has been expanded into a complete theory of Grimm's Law, and as it has been very widely accepted, I shall take the liberty of selecting it for a somewhat minute examination. 12. — As an indispensable preliminary, we must glance at the same learned author's theory of Aspiration, under which the aforesaid initial substitution is ranged. By this theory each of the Aspirates is assumed to have been originally a compound and soft sound ( = media + breath), capable, indeed, of combination into a simple sound, but also, under other circumstances, separable into its components, each of which might either absorb or fall away from the other, pretty much as if they were still independent sounds in mere juxtaposition. That is to say: — (i) the breath might vanish, leaving the media alone ; (ii) the media might vanish, leaving the breath alone ; and (iii) media and breath, by fusion, might produce a spirant. To these Curtius adds the startling postulate (iv) that the breath might also raise the accompanying media into a tenuis, and then, by continued association with it, give rise to the hard Aspirate. Result (iv) certainly has the aspect of an ex- crescence ; it has already been assailed by Max Miiller ^ ; and I shall endeavour hereafter to show that it is as needless as it is unlikely ". But neither this nor the results marked (ii) and (iii) concern us just now. The characteristic of the theory clearly lies in the composite aspect under which it represents the primitive Aspirates, and the loose cohesion it attributes to their components. This is essential to its connexion with the Lautverschiebung ; for it is by supposing the breath to fall away from the media, as in (i), that we are to obtain the sub- stitution of S for A, which Curtius assumes to have been the initial one. This starting-point is not without appearances in its favour ; for where A occurred, by the hypothesis, in the ^ Lectures, ii. 223, note. ^ It is introduced to serve as a basis for an explanation of the awkward fact that all the Greek Aspirates were hard Aspirates ; which quality must be reconciled, by any means, with the supposed softness of the old Aspirates. (See §§ 32, 33 infra.) 24 grimm's law : a study. [§ 13. Holetlmic speech^ S occurs in later languages on a very ex- tensive scale, as, for example : — throughout the L.G. dialects, throughout the Li- SI., generally in the Old Keltic and Old Persian; often in the Zend; and nearly always in Latin, and sometimes in Greek, in the case of internal consonants. Besides, Lottner, towards the end of his well-known article on the Exceptions to the First Lautverschiebung^, has acutely pointed out that the change from A to S, as between 2^ and 2 , is by far the most thorough and complete of the three changes — ^which seems ^ to warrant the conclusion that that change had set in soonest, and had therefore been in opera- tion longest ; and Grassmann {" Ueber die Aspiraten,^^ &c.) ^ claims to have furnished fresh evidence in favour of the same view by establishing the prevalence of similar deaspirating movements both in Greek and Sanskrit. 13. — Now, in spite of the suspicious looseness of composi- tion and the protean changes attributed by this theory to sounds which might rather be expected to exhibit somewhat of consonantal unity and rigidity, let it be provisionally granted that an initial substitution of S for A may thus be accounted for : it next becomes necessary to see what advantage accrues therefrom to the other substitutions. To explain these, Cur- tius supposes that when the first substitution was completed, or at least as it advanced towards completion, a Discrimi- nating Impulse or Instinct {Unterscheidungstrieb), a sort of Differentiating Principle of language, came into play. Men were conscious that the old Mediae were not the same as the new Mediae which sprang from the primitive Aspiratae, and so the old g, d, b, were gradually strengthened to k, t,p. But these new Tenues, iu their turn, clashed with the old.Tenues; and the latter therefore began to differentiate themselves from the former by severally assuming a thick breathing after them ; and thus were produced the genuine Aspirates M, th,ph'y which afterwards became weakened to their related Spirants. Such 1 Kuhn's Z-S., vol. xi. " But perhaps only seems. See § 60 (c) infrd. • Kuhn's Z-S., vol. xii. See Appendix F, infrh. § 14 (a).] GRIMM^S LAW : A STUDY. 25 is the original statement of Curtius ; and it agrees substan- tially with the views expressed in his Greek Etymology^, — the main difference being that, instead of the Differentiating Impulse, he here invokes Reciprocal Compensation as the operative cause of the later substitutions. 14. — {a) Now, however ingenious a theory may be (and the ingenuity of the theory before us is beyond dispute^), its real value is unfortunately to be estimated, not so much by the dif- liculties it solves as by those it leaves unsolved. It may har- monize with ninety-nine facts, but it may be irreconcilable with the hundredth ; in which case we have no option but to suspend our assent on account of its failure in the one point, rather than to accept it too readily for its success in the ninety- nine. In the theory of Curtius I think I see more than one essential point which is scarcely of a kind to command accept- ance. I do not say much at present on the apparent absence both of motive and of purpose in the assumed initial substitu- tion, and indeed in all the rest — although it is, on the face of it, more than extraordinary that a dialect to which, by the hy- pothesis, the CI. mutes were originally native should spontane- ously modify the whole series with scarcely any other ultimate result than a mere change in their distribution. There was, it seems, no addition to their number ; and as the CI. system still sufficed for the CI . tribes, we may assume that the pho- netic requirements of the Germans likewise must already have been fully supplied. And similarly, as to the second Verschiebung , the L.G. mute-system, in its turn, is supposed to have been at one period native to the H.G. tribes; and as that system continued to suffice for the L.G. tribes, and as the 1 Third edition, pp. 425-6. ^ "Das Lautverschiebungsgesetz der germanischen Sprachen ist eine viel zu wichtige sprachliche Erscheinung, als dass sie nicht seit ihrer Ent- deckung schon eine Menge Versuche sie zu erklaren hatte hervorrufen miissen; unter diesen diirfte aber die von Curtius in Kuhn's Z-S. ii, 322 IF. aufgestellte als eine wohl gelungene zu bezeichnen sein." — Whitney (Jolly), p. 155, note. Curtius's views have been adopted by Lottner, Grassmann, Forstemann (throughout his important " Gesch, des D. Sprachstammes "), and a host of other writers. 26 GRIMM S LAW : A STUDY. [§ 14 (^). second Verschiebung, like the first_, merely led to a revoliition, and not to an extension or any perceptible improvement^ of the L.G-. system, we may assume that the latter system was amply sufficient for the requirements of the H.G. tribes also. No other example, I suppose, in the whole world can be pro- duced, in which a people has thus — I will not say twice, but even once — spontaneously revolutionized by far the firmest section of its articulate sounds. [b) To assert that all this supposed change really was as wanton and even mischievous as it appears, would perhaps at this point be inconsistent with the caution befitting our in- quiry ; for both cause and purpose may lie beyond our ken. But what we may safely say is, that, on the hypothesis under review, the first substitution appears causeless, purposeless, and (if it really necessitated the others) mischievous. It may deserve passing mention, however, that some adherents of this theory have discovered for us a cause so widely dissonant from the facts of the case, that one finds it hard to believe that sufficient attention has been paid to those facts. We are told that the Aspirates were "objectionable^^ *. No doubt they were so in some quarters; and this very fact will be here- after applied to elucidate a totally different set of phenomena. But to the assertion that the Aspirates were objectionable to the ancestors of the Low Germans, I altogether demur. From the evidence at our disposal I should infer precisely the reverse. But to make good my point I will fall back on independent authority. (c) Nearly a quarter of a century ago Dr. E. Forstemann executed a series of calculations to determine the relative proportions in which the several articulate sounds, vowel and consonantal, enter into some of the leading I-E. languages^. He seems to have had no immediate object in view beyond establishing facts; and as he could not possibly foresee to what use those facts might be turned, his testimony must be ^ " The first variation \i. e. the supposed L.G. Lautverschiehung] rose, as all (?) agree, from the objectionable Aspirates." — Peile: "Grk & Lat. Etym.", 2nd edit., p. 157. ^ Kuhn's Z-S., i. ; and more particularly ii. 38. § 14 (c) .] grimm's law : a study. 27 held to be completely free from bias. Taking first tbe prin- cipal Classical languages, lie obtains the following results as to the Mutes^: — Skt. Grk. Lat. Tenues .... 52'6 per cent. 71*4 per cent. 72*0 per cent. Medics 29-0 „ ,, 14-3 ,, ,, 25-5 „ „ AspiratcB . . 18'4 jj ,y 14*3 „ „ 2*5 ,, ,, Striking the average, we may attribute to the Classical Mute- system 65*6 per cent, of Tenues, 22-7 ,j „ Media, but only 11*7 „ „ Aspiratce. Compare with this the Gothic Mute-system, which, as he next shows, comprises 14*3 per cent, of Tenues, 22*9 „ ,/ Medice, and no less than 62*8 „ „ Aspiratce'^; ^ I have, for convenience, reduced his proportions to percentages ; and I am responsible for the averages. ^ It does not appear from Forstemann's articles on what lines of inquiry he advanced — on how many words, for example, he based his calculations — ^whether he took the words from a dead vocabulary, or from living compositions — and whether, in the case of Gothic, he includes not only actual substitutions, but those which would have been made but for some overruling circumstance, such as " protection " by s. I raise these doubts because, before I had unearthed those articles, I had commenced a similar series of calculations in relation to the mutes alone, but with results some- what less strikingly in favour of my argument in the text, although still striking enough. Thus, for the CI. system, taking the primary words (above 900) in the primitive I-E. vocabulary forming the first division of Fick'sW-B.,Ifound Tenues 565 per cent., Medics 24^0 per cent. Aspirates 20'5 per cent.; and results almost identical were obtained by digesting the initials of all the words in the same vocabulary ; while in Homer, as representing the older Greek, out of above 700 mutes examined, there were Tenues 59*7 per cent., Medice 22*3 per cent., Aspiratce 18 per cent. In elucidation of the L.G. system, I took the initials (as these are in general most regularly verschoben) of all the words (about 1100) in Fick's primitive German vocabulary ; among which I found Tenues 17 per cent., Medice 35 per cent., Aspiratce 48 per cent. With this result may be compared that of an examination of the mutes 28 grimm's law : a study. [§ 14 [d). that is to say, among the people to whom the Aspirates are asserted to have been *^' objectionable/^ Aspiration prevailed five or six times as extensively as it did among those to whom they were not so. (d) Never, surely, was there suggested a motive for any given line of action so completely inadequate to explain the same, or so flatly contradicted by facts. It would have been a far more cogent argument to attribute the change from S^ to S to the very marked, the really extraordinary, natural aptitude and affection for Aspiration on the part of the L.G. peoples. This would bring us over to the second view (that of Bopp and others) stated in § 11, which thus accounts, at a stroke, for three fifths of all the supposed consonantal trans- mutations. But some mystery would still overhang such an explanation; for one would have thought that so strong a love for Aspiration would certainly have led the Germans to maintain the Aspirates they already had, and to increase their number, if impelled to do so, by modifying other mutes. (e) It cannot be argued that this large percentage of the newer mutes thus finally obtained were not really Aspirates but only Spirants ; for, not to mention the distinct assertion of Curtius to the contrary ^, such an argument would deprive the theory under review of much of its value, and would in fact render a second necessary in order to account for the evolution of ^^ from S . To maintain one and the same hypo- thesis consistently throughout both of the supposed evolu- tions, it should be maintained that the L.G. Aspirates (not Spirants) were similarly objectionable to the H.G. tribes. (nearly 1800) in the episode of Ohthere in King Alfred's Orosius, where there appear Tenues 22*5 per cent., Medice 29*2 per cent., Aspiratce 48'3 per cent. I had intended to extend my examination to the mutes, both in vocabu- lary and in composition, of all the leading CI. and L.G. languages j but it is much easier to borrow statistics than to compile them. ^ " Das alte k, t, p konnte es sich nicht gefallen lassen mit dem neuen auf einer stufe zu stehen. Ihm stiirzte ein dicker hauch nach, der anfangs ticherlich toahre aspiraten (kh, th, ph) erzeugte ; von denen kh und ph zu h und/sich verflUchtigten."— Kuhn's Z-S., ii. 331. § 15 {b).] GRIMM^S LAW : A STUDY. 29 and that tlie desire to avoid them led to the initial movement towards the second Verschiebung ^ Besides, if the new L.Gr. sounds had been Spirants, their supposed components would have been fused beyond the power of the H.Gr. tribes to sepa- rate the breath from the other element (which, by the way, in this case, is not the Media ^). 15. — {a) Returning from this digression, I venture to sug- gest that the nature of the supposed change (A to S)^ which is the corner-stone of the whole hypothesis, is, in itself, not altogether beyond challenge. As was pointed out by Schleicher many years ago^, the natural tendency of the Aspirates within historic times and in individual languages has been to become, not unaspirated mutes, but Spirants ; and this tendency he went so far as to call '^ o. law arising from physiological causes.'^ Now, although there is room to doubt whether the genuine Aspirates ever passed into Mediae by natural and unconstrained debilitation, I would not point-blank deny the possibility of such a transmutation, at least under the influence of neighbouring sounds. But what cannot fail, in Curtius^s theory, to strike us at a glance, is the incongruity that, up to a certain limit of time, the tendency of the whole body of Aspirates ran in one direction (i. e. towards the Media), while from that limit it has run in a different one (i. e. towards the Spirant). This would arouse suspicion if it were chargeable upon two different races of people ; but when it is charged, and that twice over, upon one and the same race, it approaches the incredible. {b) For observe what we are really asked to believe, namely : — (a) that the original CI. Aspirates among all the primitive Germans became Medise ; (/S) that the second crop of Aspirates (supposed to result from the third substitution, A for H^ in S ) generally became, among the Low Germans, Spirants; (7) that, among the High Germans, of this same crop part 1 See § 16 (c) infrd,. * See the following § (c). 3 " Formenlehre der K-Sl. Spr." — quoted, in fact, by Cnrtius in the article so often referred to. 30 grimm's law : a study. [§ 15 (c). also became Spirants ^_, but the greater number passed into Mediae ; and (S) that of the third crop of Aspirates (supposed to result from the third substitution, A for H, in 2^) some have in H.G. become Spirants, and the others are arrested at a stage which is much nearer to the Spirants than to the Mediae. It is more than odd that one and the same sound, among one and the same race of people, and, so far as we know, under one and the same set of circumstances, should have exhibited these incongruous modes of development. Of the passage of Aspirates into Spirants we find ample historical evidence — in Skt, in Grk, in Latin, in L.G., in H.G. ; but in favour of their transmutation into Mediae we have only a fact which may mean something quite different — ^namely, that while Aspiratae appear in certain words of some languages, Mediae appear in the corresponding words of related lan- guages. (c) In the way of the contested transmutation (A to S) indeed, as between 2 and 2^, Curtius^s own hypothesis intro- duces a new and unexplained difficulty. For by that hypo- thesis the L.G. Aspirates, forming 60 per cent, of all the L.G. Mutes, were hard Aspirates (§ 14 (e), note ) ; and if the CI. Aspirates (supposed to be soft) were severally equal to media 4- breath, and by loss of breath gave rise to the L.G. media ; the L.G. hard Aspirates ought severally to he = tenuis-^ breath, and, by the loss of the breath, ought to give rise to tenues in H.G. ; which, when deaspirated, as supposed, they do not. There is consequently wanting somewhere an inter- mediate stage of debilitation. Or are we to suppose that the breath, which in Greek raised the media to the tenuis (§12 ante) and then combined with it into the hard aspirate, here lowered the tenuis to the media, and combined with it into the soft aspirate, before the latter passed into the H.G. media ? 16. — (a) So much with reference to the initial substitution; ^ This qualification is inserted to meet the view which represents as unverschoben those H.G. fs and A's which agree with the same Spirants in L.G. § 16 (b).'} grimm's law : a study. 31 w]iich_, on the hypothesis under review^ is apparently causeless and purposeless ; which attributes to the primitive Aspirates a bisonant and oscillating character_, open, to say the least, to considerable doubt ; which, in its execution, involves more than one striking inconsistency ; and which, as we next have to notice, is represented as directly or indirectly necessitating other changes which, with it, amount to nothing less than a phonetic revolution. For the supposed debilitation of the CI. Aspirates had, it is said, the immediate effect of inciting to the strengthening (in the second substitution, H for S) of twice as many Medise into Tenues, and ultimately involved the change (in the third substitution, A for H) of four times as many Tenues into Aspiratae. (b) As no one has asserted more distinctly than Curtius* that ^^ Weakening '^ is the ruling principle of phonetic change, and as he even rejects^ Max Miiller^s counteracting principle of '^ Dialectic Regeneration '^ on the ground of its inconsistency therewith, we are naturally curious to learn the nature of those more potent forces to which he would attribute such changes as the above-mentioned second substitution (H for S). Their names, as we have seen, are Discrimination and Compensation, — names which may really be intended to denote different aspects of the same process^ — the former perhaps directing attention rather to some unperceived mental instinct originating the second and third substitutions, the latter to the relationship which those substitutions assumed in course of execution. 1 Grk Etym., p. 412. = -^ ^ ^^^^^ ^ Compare the following passage from the " Grk Etym." (pp. 425-6) with the summary of Om-tius's older explanation in § 13 ante : — " Der Uehergang von g, d, b in k, t, p in den Germanischen Sprachen erklart sich aus jenem Zusammenhange, der zwischen sammtlichen Lauten einer Sprache in der Art stattfindet, dass sich diese wechselseitig compensiren. Die einmal eingetretene Verwandlung eines dh in d trieb auch das urspriing- liche d aus seiner Stellung, so dass das alte d zu t ward, und endlich das neue t wieder das schon langst vorhandene alt iiberlieferte zu ^A verschob." The repeated statement that one sound ^^ verschoV another is, of course, no explanation of the " VerschiebungP Nor is it quite self-evident how the change of ^ to ^A can, properly speaking, " compensate " for that of d to t, or the latter change for that of dh to d. But see §§ 18-22, infr^. 32 GRiMM^s law: a study. [§ 16 (c). (c) Of what Curtius means by the former (to speak of this first) we can acquire some conception from its supposed mode of operation. This^ as it affects the evolution of S from S^., is represented as follows : — When by phonetic change in a given language a whole series of consonants of a certain degree became assimilated to another series of another degree^ the latter series was made to change its degree throughout ; and when this series consequently became assimilated to a third series_, this third series was also made to change its degree throughout. The singular result in the case before us was, that the language thus acquired, in fivefold measure, the very sounds which, for some unknown reason, it endeavoured at the outset to get rid of. To put this process in another light — a language which, under the sway of its debilitating tenden- cies, was powerless to maintain a single distinction (viz. that between A and S) already existing on a small scale, is never- theless supposed to have elaborated two other fresh distinc- tions on a much larger scale, and partly in direct opposition to such tendencies. To account for the evolution of 2^ we ought, of course, to suppose a repetition, in different propor- tions, of the same set of changes. But here we meet with a difiiculty. For, as is well known, the existing H.G. labial and guttural families, in addition to their own Spirants which properly answer to the L.G. Tenues, also exhibit Spirants (/ and h) in correspondence with the L.G. Spirants/ and A, and where, on Curtius's hypothesis, they should exhibit the Mediae b and ^^ Nevertheless, although in these cases the A of 2 is said not to have proceeded to S in S^, we still find the S of % apparently raised to H in %^ {i. e. b and g to p and k) . By what power, then, was this transmutation effected, or at least necessitated?^ ' But see §§ 52-54 infra. * It was Dr. (now Prof.) W. Scherer, in his acute work " Zur Gesch. der D. Spr.," who, if I remember right, first raised this objection to Curtius's hypothesis. Scherer's own view is that the successive substitu- tions were totally independent of one another ; and he gives an elaborate plan of the way in which he supposes the evolution of one sound from another to have taken place. I am not aware that this view has gained any converts, except perhaps Dr. Schweizer-Sidler (See Kuhn's Z-S., § 17 (a).] grimm's law : a study. 33 [d) But if all such difficulties were obviated, the whole process would still savour little of the economy of nature. If the initial change actually had taken place as is supposed, no great harm would have been done ; and we are at a loss to see either why any remedial measures were necessary, or, if necessary, why they should involve other changes sevenfold more extensive and partly in direct contravention of nature^s ordinary modes of procedure. One would think it much simpler, and almost infinitely easier, for a discriminating principle to have exerted itself in causing the immense majority of Mediae and Tenues to resist the supposed tendency of the Aspiratse, and thus to keep them in their original places. Such strange movements of language, however, are appropriately relegated to a period beyond observation ; for they seem quite out of harmony with any of the movements which history has yet enabled us to watch. These, with few and generally explicable exceptions, all tend in one direction. When certain results have been achieved, men appear quite indifferent to the nature of such results, and incapable of altering them, however inconvenient they may be^ 17. — (a) Now the chief if not the only practical inconveni- xviii.; p. 289). It is partly for this reason that I do not select it for special examination, and partly because, as my own central idea is precisely the reverse of it, all the arguments and evidence I may hereafter adduce will cut directly or indirectly against it. ^ One further diiSculty I have not stated in the text, because Ourtius himself makes no allusion to the point that gives rise to it. I should judge, from his description of the supposed successive substitutions, that for their complete execution a very long period of time would be necessary. Thus the original Aspirates would have died out long before the Low Germans proceeded to evolve the second set of Aspirates out of their older Tenues. By what means, then, was the memory of the Aspirates, or at least the impulse to reproduce them, kept alive from gene- ration to generation? Fdrstemann (Gesch., i. 14) manifestly feels this difficulty in reference to a long period : for a short period he holds the necessary Spracligefuhl ^^ nicht fiir unmoglich". All 1 would here remark is, that if a given sound has completely died out of a language, it can only be recovered when it forms the terminus to which, by natm'al debilitation, some still existing stronger sound tends. 34 GRIMM^SLAW: A STUDY. [§17'(^). ence arising from the assimilation of two series of consonants, sncli as the Aspiratse and Mediae of %^, would perhaps be the formation of a small number of homonyms. In this respect the primitive German language, on the hypothesis in debate, would only have shared the fate o£ most, if not all, other lan- guages, and to a far more limited extent than some of them. A monosyllabic language, indeed, like the Chinese, is but, as it were, a cluster of homonyms. A word of a single syllable may have to serve for a score of meanings ; and its application in any given case has to be determined by such indirect means as emphasis and intonation^. The reduction of originally different words to the same sound displays itself, though far less extensively, in modern European languages, especially those which, like the French, have been reconstructed out of the debris and detritus of an older language. But little or no mischief is done ; for, in the first place, the meanings of the several primitives are in general so widely different, that the homonymous derivatives remain to all time clearly distin- guished in use; and, in the next place, since modern lan- guages appeal quite as much to the eye as to the ear, means have been offered, by the fluctuating value of letters, for per- petuating symbolic distinctions after phonetic differences have disappeared^. {b) Nowhere, however, do we find (where we can observe) that Nature has been so busily fastidious as to remedy any such apparent confusion ^: why, then, are we to assume that ^ See Mr. Sayce's " Principles of Philology," p. 103. * U. g., in our pm^e, pear, and pair ) or the French ver, vert, veire, and vers; etc. Here, if anywhere, in connexion with the purely artificial process of writing, a sort of rudimentary differentiation may be observed in operation — not, however, actively, by special adaptation of varieties of spelling to varieties of meaning, but (if it is not a contradiction to say so) pavssively, by a refusal to aUow the spelling to proceed to complete assimi- lation, even after the fusion of the sounds. ^ " Quand une langue d^riv^e perd la faculty de prononcer certains sons qui faisaient partie de celle d'ou elle sort, il en rdsulte tout simplement des homonymes, et la langue nouvelle rHy 6chajype pas. Ainsi, le Zend con- vertissant les aspir^es du Sanscrit en pures, les deux racines dhd- (poser) et da- (donner), rl6r]}ii et 6t8co/xt, s'y confondent dans im meme verbe da- qui signifie a la fois ^ donner ' et ^ faire.' H en est de meme de -^ere dans § 17 (c).] GRIMM^S LAW : A STUDY. 35 she interfered where our observation cannot reach? If, indeed, differentiation had set in among complete homonyms only, the limitation of her interference and the apparent reason for it might have passed for evidence that she really had for once relaxed her passionless indifference ; but the interference is supposed to have gone on right and left, even between words which, after the first substitution, had but one letter in com- mon ^ (c) But farther, the assimilation of the classical Aspirates to the Mediae would only produce (§ 14) a series of Mediae making altogether less than 40 per cent, of all the mutes; while the Tenues would still form upwards of 60 per cent. Among the words involving Tenues we should therefore conjecture that there must have already existed many more homonyms than among those involving Mediae, new and old together. Now here was a fine opening for the application of some cor- rective process on the part of Nature ; but none was applied : those numerous Tenues were all (or nearly all) , by hypothesis, transmuted into Aspiratae, and must consequently, in their new form, have perpetuated any disadvantage arising from the existence of homonyms. The action of the discriminating principle thus invoked appears therefore to have been some- what wwdiscriminating. Nor was it quite so impartial as one could wish. The early Germans, and they alone, were the chosen race among whom it is represented as manifesting its powers. The Slavonic tribes, who are supposed to have ac- companied the Germans in the first substitution, seem, for no apparent reason, to have been left in impotence. And the lea composes latins: il signifie tantot ^donner' (red-dere) et tantot 'mettre' (ah-dere, con-dere). Le chinois surtout est plein de ces homo- nymes produils par des ressemblances de mots uses, auxquels on pent comparer le fran9ais tour, de tmris et de tornus] somme, de somnus et de sumfna, etc. Une fois que la degradation les a reduits a cet etat, le Ian- gage est bien forc4 de les garder et de les eclaircir par des moyens indi- rects. A tnoins d'une convention expresse, egalement impossible a imagi?ier et a mettre en pratique, Jamais on n^aura recotirs a, V expedient de les distin- guer par des prononciations nouvelles.''^ — Baudry : G.C., p. 145. ^ "Das junge r/ von Goth. , (c) an effort to remove the discrepancy; {d) which effort, however, is diverted by the said sentiment into a counterbalancing dis- crepancy. These fundamental conditions involve the minor phenomena stated in § 19 (a), viz. : — (4) the correlative duality ofthe phonetic movements; (5) their contemporaneity; and (6) the identical phonetic value of the interchanged sounds, both of which, previously to the movement, are common to both dialects. {d) I have purposely employed much time and some repe- tition in elucidating the characteristics of this important pro- cess — first^, because I do not know that the same thing has anywhere else been done, and, secondly, because the process itself is, on a small scale, a genuine Lautverschiebung under the only known conditions which admit thereof. Some of its phenomena, indeed, do not directly illustrate the principles on which, as I believe, a correct theory of Grimm's Law is to be based ; but others exhibit language as acting with a systematic vigour even greater than such a theory postulates. We shall hereafter have to inquire what results follow from absence or modification of any of those conditions. Our business up to the present has been to track out various hypotheses on the lines laid down by their aflthors ; to examine the forms of compensation suggested to us ; to construct the forms actually ^ No reader will, I hope, for a moment imagine that this Dissimilating Sentiment is but the reproduction, under another name, of the Discrimi- nating Impulse whose existence is questioned in §§ IG; 17. The former represents a fair induction from observed facts, and is really a hahit of mind generated by the differences between dialects in contact. The latter is a mere creation of the imagination, and is actually opposed to the habit of mind which should have been generated by the conditions under which the Chronological Hypothesis supposes the evolution of 2^ from 2, to have taken place. For, firstly, the L.G. dialect is supposed to have been com- pletely removed from the influence of the CI. dialect ', so that there seems to have been no agency for evoking the sense of phonetic discrepancy ; and, secondly, as the first movement of the whole evolution was the fusion of one series of mutes with another, the sentiment generated by that movement should rather have been an .dissimilating Sentiment. § 23 {b).] GRIMM^S LAW : A STUDY. 49 warranted by facts ; and to show that the conditions involved in those hypotheses are at variance with the results at which we have arrived. 23. — {a) We have now reached a point at which we may fairly stop to breathe and look around us. So far we have found that no explanation of Grimm^s Law which takes the Historical or Chronological Hypothesis as its groundwork is altogether satisfactory. We may have started fairly enough ; but_, whichever road we have taken^ we have sooner or later found a lion in the way. There is therefore some ground for supposing that the discordance between that hypothesis and the conclusion from the Principle of Sufficient Reason (§ 3) may be entirely the fault of the former ; so that we shall be all the better prepared to admit that that conclusion may be a truth the neglect of which surely leads to the contra- vention of other truths more directly demonstrable, and to suspect that no hypothesis of Grimm^s Law can lead to satis- factory results which is based on the successive evolution of the three phonetic systems S^, S , 'Z^, or (what is the same thing) which does not take their contemporaneous evolution for a fundamental principle. Thus we return to the point from which we diverged (§ 4) into an examination of various aspects of the Chronological Hypothesis. (b) One philologist of the first rank (Prof. Max Miiller) long ago endeavour edj as we have seen (§ 3), to enforce the view to which we now recur. But the line of reasoning (if it was not rather the intuition of a keen and practised intellect) by which he reached it_, does not stand out in his Lectures with that fulness and clearness which generally characterize the writer. It is partly for this reason perhaps that the reception accorded to his statement of the view has scarcely answered to the Professor^ s great reputation ; but it is partly also, I think, because there appears to be some inconsistency between the different parts of that statement. For as Baudry ^ has pointed out, these really constitute two separate hypotheses. The first properly belongs to the class of hypotheses which we 1 G.O., pp. 144, 146. E 50 GRIMM^S LAW : A STUDY. [§ ^^ (c). have already examined and put aside. For although it re- presents the substitutions as going on contemporaneously in S^, ^ , and S , yet it not only assumes the pre-existence of an earlier triple mute-system_, but it also represents the indi- vidual sounds of %^, Sy, ^^, as being successively adjusted to their actual place in those systems ; and the order of adjust- ment in S is precisely that adopted by Curtius ^ The ulti- mate cause of the whole movement is stated to be thisj — that " the Teutonic tribes_, in "taking possession of the phonetic in- heritance of their Aryan, not Indian, forefathers " (as if' this were something in which they had previously had neither part nor lot), ^^ retained the consciousness of the threefold variety of their consonantal checks, and they tried to meet this three-fold claim as best they could ^^'^. The first substi- tution (S for A) of each of the German systems is repre- sented to have been a matter of preference; the others, of necessity. Under the sway of "a wish to keep distinct what must be kept distinct,^' the old Soft-mutes and the new *' could not- be allowed to run together;^' and so the old ones were raised to Tenues. Then arose " pressure '^ and '^ necessity '^ which drove the Gothic nations to the third substitution ; — with much more of the same kind : all which is but a variation of Curtius^s assertion that " Das junge d verschob das alte d" &c. (c) This aspect of the hypothesis therefore represents only a perplexing compound of the simultaneous and the succes- sive — of inherent phonetic necessity (pressing, be it remem- bered, upon the Germans alone), and, apparently, of arbi- trary contrivance and convention, — which lies open to most of the objections applicable to the purely chronological hypo- thesis, and to others besides. These we need not here pro- duce or reproduce ; for the learned author presently broaches his hypothesis under quite another aspect : — " From the very beginning different branches of the Aryan family fixed the three cardinal points of the common phonetic horizon differ- ently. While the Hindus fixed their East on the gh, dh, and bh, the Low Germans fixed it on the g, d, and b, the High ' See §§ 11, 12, ant^, » Lectures, ii. 225, 226. § 24.] GRIMM^S LAW : A STUDY. 51 Germans on k,t, p. All the rest was only a question of what the French call s'orienter"'^. This statement is intimately connected with a previous one,, which refers to a condition of things anterior (it would seem) to ^^ the very beginning ^^ — ^^1 feel strongly inclined to ascribe the phonetic diversity which we observe between Sanskrit, Greek, and Latin, to a previous state of language, in which, as in the Polynesian dialects, the two or three principal points of consonantal contact were not yet felt as definitely separated from each other ^^^. And again : — ^^ I can conceive diflPerent definite sounds arising out of one indefinite sound ; and those who have visited the Polynesian islands describe the fact as taking place at the present day. What then takes place to-day can have taken place thousands of years ago ; and if we see the same word beginning in Sanskrit, Greek, and Latin, with k, t, or p, it would be sheer timidity to shrink from the conclusion that there was a time in which that word was pronounced less distinctly ; in short, in the same manner as the k and t in Hawaian^^^. 24. — This view, however, seems to me to go too far in the opposite direction — farther than any known facts will war- rant. It means, if I understand it aright, that not merely the three related sounds of each family, but also the three families themselves — i. e. no less than nine sounds — were all evolved out of one primitive *^ indefinite ''^ sound. Cur- tius may well ask'* what that protean and indescribable sound could have been. But it is still more to the point to remark, as • he further does, that (not indefiniteness but) vigour and distinctness must have been, so far as we can judge, the characteristic of the Holethnic phonetic system °. In the most ancient remains of the separate dialects, the distinction between the different families of mutes is quite 1 Lectures, ii. 231. ^ Id., p. 197. 3 Id., p. 200. * Grk Et., p. 412. ^ Max Miiller's illustration from the Hawaian is of dubious application. The confusion there existing is far more like a late corruption of an ear- lier and purer condition of the mutes, than an embodiment of the primi- tive state of the language. i;2 52 grimm's law : a study. [§ 24. as sharply marked as in the younger dialects, or even more so ; while the conditions regulating the transition of a pri- mitive k into t, Pj and still feebler sounds, have been pretty clearly ascertained, and are almost certainly independent of any phonetic instability in the Holethnic period ^ What- ever may have been " indistinct,^^ the lines of demarcation between k, t, and j9, in the period immediately preceding the so-called Lautverschiebung , were certainly not so. Nor, so far as the most ancient remains of language enable us to judge, was there any confusion, within historical memory, between the much more closely related sounds forming the three mutes of each family. Nevertheless, as the members of each family are produced at one and the same point of the vocal passage under but slightly differing conditions, it is, on physiological grounds, not improbable that all three of each group may at a still remoter period have radiated from a single sound ; so that, if we may make Max Miiller's sug- gestion refer to each family only ^, I believe we shall find in it a close approximation to one of the fundamental truths on which, as I make bold to think, a correct theory of Grimm^s Law must be based. But in descending from the general statement to the working out of its details, we shall be com- pelled at once to part company with the learned Professor. For if, without going beyond his own hypothesis, we test his suggestion by the ruling Principle of Least Effort, we shall at once perceive the necessity of an important modification ; for a single sound that should generate three related sounds ^ See §§ 65-63 infra. I am not denying that future investigators may find some justification for referring all three families to a single primor- dial mute. But if they do, I venture to predict that that mute will not be a flabby and " indefinite " one, but the hardest and firmest of them all, — i. e., K (unless a firmer should yet be discovered). ^ G. Michaelis (" Ueber den Unterschied der Oonsonantes tenues und mediae," &c.), while thus limiting the suggestion, finds the parent sound of each family in the Sanskrit Soft Aspirate, which seems to him a sort of chaotic jumble ofifering no clear distinction between articulation, breath, and voice. C. Arendt, in reviewing the tract (Kuhn's Z-S., xii. 441), anxiously (and, I think, successfully) defends those sounds against such an imputation. § 25 {a).] GRIMM^S LAW .* A STUDY. 53 differing in strength could not be weaker than the strongest of the three^ or else^, by that principle, it could not have given off the strongest. We are unable, however, not only to produce but even to conceive of any sound or sounds stronger or purer than the Tenues; and as the other sounds to be accounted for are weaker or less pure than their related Tenues, and could be directly derived from them ^, what ground or what need is there for assuming the existence of parent sounds for the three families differing in any respect from the three Tenues themselves ? 25. — (a) I propose then, that, instead of some unknown, ^^ indistinct ^% and undefinable sound, we take the purest, the strongest, the most distinct sound of each of the three families as the primordial one of its family, and as the parent of the other two. By this means we shall, I think, at once put ourselves in harmony both with the Principle of Least Effort, and with the acknowledged vigour and distinctness of the earliest I-E. mutes. What is more, we shall effect a junction with the conclusion from Sufficient Reason, by which we are forbidden (§§2, 3) to ascribe pri- ority or superiority of any kind, to one of the three leading phonetic systems over the remaining two. For it is certain that the symmetrical ter-trinal trinity constituted by those three systems together cannot have existed from all time. Indeed the very object of every inquiry into Grimm's Law is to trace all three to some single primitive system. But if no one of the three may be deduced from either of the others, we must assume that ^. %. and 2, were evolved within some fixed period from some pre-existent system (let us call it Sj, which must have differed both from 2^, S , and S^ (or else that one of the three which coincided with it would really be 2p and therefore prior to the others, which is contrary to the hypothesis), and yet must have been symmetrically related to them all, and they to it ; for each of them simultaneously or contemporaneously derived from it precisely the same sounds as the others, and in such a way that to the same ^ See Appendix B. 54 - , GRIMM^S LAW : A STUDY. [§ ^^ (b). sound in each there always corresponded the same sounds in the other two : that is to say^ Where 2^ derived H, 2^ derived A, and X^ derived S ; » X^ ■ )f H, X^ „ A, „ 2^ ij S. [b) This tabulation may perhaps enable us to discover what that symmetrical relationship must have been ; for^ taking the totals of the three vertical columns^ we observe that all the three systems together derived from S^ three complete and co- extensive series of mutes — one of one degree (H)^ another of another (A), and another of the third (S) — each compri- sing as many sounds as each of the three fragmentary series (H A S_, or A S H^ or S H A) which severally make up 2^, 2 , and 2^ ^ ; or^ as the same fact may be stated^ 2^ was of such a nature as to give oflP, or allow to be given off, for distri- bution among 2^, 2 , 2,, three coextensive series of mutes — one consisting of Tenues only, another of Aspiratse only, and the third of Mediae only. 2^ therefore must have com- prised either three perfectly coextensive series of mutes {i. e., 2 1=2^ + 2 + 2j, or else only one uniform series ^, from which the other mutes must have been derived. Now to suppose (as nobody, however, can) that such a triplicate series of mutes ever existed would really be to suppose that the vocabulary of the older Holethnic speech was tripled in that part (and that only) in which the mutes were involved ; so that every mute-built word would and did, in one and the same lan- guage, appear in three precisely equivalent forms constructed by merely changing its consonants. This supposition may be at once dismissed on the ground of its own inherent absurdity. The alternative is that the older system comprised only one ^ ^ The symbols 2^ &c. have hitherto indicated the arrangement of the three classes of mutes in the dialects to which they severally appertain ; their meaning is now extended to denote the whole series (or quantity) of mutes in those dialects. ^ See the following note. ' 2^ cannot represent two such series of mutes j for as the numbers 2 and 3 are prime to each other, 2j could not in this case have stood in any sym- metrical relationship to the younger systems either singly or together. It §26.] grimm's law: a. study. 55 of those uniform and coextensive series^ and that the triform series constituting each of the younger systems resulted from a twofold variation affecting different sections of that one older series^ but leaving a third section unaffected_, — the sections altered and the sections retained by the several systems being reciprocally adjusted among those systems upon a plan which will presently, I hope, appear to be not altogether inexpli- cable. But when we speak of the ^ Variation •'^ of articulate sounds, we are compelled by the Principle of Least Effort (unless some counteracting principle is shown to be in action) to understand "debilitation." Consequently we have to assume that the one unbroken parent series was made up of mutes of that degree which is susceptible of debilitation into mutes of the other two degrees. Which of the three degrees that must have been, admits of little discussion. The Mediae, by common consent, are out of the field. The Aspiratse are held, even by those who seem to teach that they are stronger than the Tenues, to result from an " affection" of the latter^; while the latter are never represented as derived by natural debilitation from the former. The three Tenues therefore alone remain as the primordial mutes, beside which there were originally no others, and from which all the others, at a later period, contemporaneously radiated by systematic debilitation^. 26. — The fundamental proposition, then, on which alone it seems to me possible to build a satisfactory theory of Grimm^s Law is, that the two weaker degrees of mutes were evolved from ■would have partially coincided witli some one of them ; and the evolution of the others would be even more difficult to account for than on the ordinary Chronological Hypothesis. Besides, it would still imply thfe exist- ence of a duplicate vocabulary; and the objection to this differs only in degree from the objection to a triple vocabulary. ^ See Appendix B. ^ The line of argument intended to lead up to the sole existence (i. e. for linguistic purposes) of the strongest sound of each of the three families, is of course to be kept distinct from the claim of the three Tenues to be those sounds. The former will hold good, even though this claim should ever be disputed. 56 gAimm's law : a study. [§ 26. the remaining stronger mute of each family, which stronger mute could scarcely have been any other than the Tennis. This proposition is of the very highest philological importance, andmay be exhibited under several different aspects. It means, for example, that there was once a time in the history of the Indo-Europeans when the Tenues were the only mutes appro- priated to the service of speech^. It means either that once on a time all the Aspirata3 and all the Mediae of the several I-E. Mute-systems were actually Tenues ; or, at any rate, that the words in which Aspiratse and Mediae occur radiated by pho- netic variation from words involving Tenues only. And it means that the Tenues of each system are what that system ^ This reduces the Holethnic (I-E.) speech, so far as the mutes are concerned, precisely to the ancient condition of the Finnish, and supplies substance for the rather shadowy arguments of Forstemann (Gesch., i. 236-238) in favour of a primeval connexion between the two families of speech. The proposition of the text is analogous to the important doctrine (§ 6 (a) ante) that there was once a time when our forefathers employed no other vowel than a. The bearing of such doctrines, when demonstrated, upon the problem of the Origin of Language, or at least of our great family of languages, will be manifest at a glance. For when that problem is reduced to the question, how a single vowel, two or three consonants, and perhaps three or four other sounds — in all not more than eight (which number may yet be diminished) — how these, I say, were originally appro- priated to the expression of ideas, a complete solution is surely not to be despaired of. It is in vain for learned bodies (like the Parisian Socieie do Linguistique) to refuse to listen to inquiries on this subject. The problem is one which, until solved, will always possess an irresistible fascination for speculative thinkers j but it is also one that can never be solved by mere speculation. What we want to know is, not how language might or should have originated, but how it did originate. We want a lantern for our feet, not fireworks about our head. We must advance towards the solution by steps which rest on solid facts, even though at each step it cost much labour to plant the foot firmly ; and such a step is clearly gained every time sufficient reason can be shown why the phonetic mate- rials of speech should be simplified and diminished in number. — I need hai-dly add that in speaking (in the text above) of language as '^ appro- priating " various sounds to its own uses, I do not mean to hint either that such sounds previously had an independent existence, or that no others were producible by the vocal organs of man. A few, very few, sounds of broadly marked character must, in the very infancy of our race, have been, one after the other, thus directly appropriated ; the rest were evolved nrom these by phonetic variation. § ^T" (b).] grimm's law: a study. 57 preserved for itself out of the earlier and larger common stock of Tenues; so that the L.Gr. Tenues are not substitutes for the CI. Mediae, nor the H.G. Tenues for the L.Gr. Mediae; but the Tenues of all three systems are of equal antiquity and co- ordinate primitiveness, and constitute, when put all together, the ancient common stock, of which the several tribes subse- quently allowed different parts to be differently weakened. And by this general weakening the evolution of all the sys- tems is brought into harmony with the Principle of Least Effort ; so that we have no longer to provide, at successive stages, for the raising of S to H, nor even for the doubtful change from A to S : we have only to assume the debilitation of one stronger sound in two diverse directions ; which twofold debilitation must, as has now been so often insisted upon, have gone on simultaneously in all the three dialects. 27. — {a) The leading positive conclusions which we have now arrived at are, in sum, that the three systems, S^, 2 and %^ are of coequal antiquity and coordinate importance, and that they were contemporaneously evolved from a single preexist- ent system comprising Tenues only. I shall next try to show what light is thrown by these conclusions, and by our col- lateral inquiries, upon various points which the Chronological Hypothesis left, to say the least, in obscurity. The chief of such points are (1) the cause (or agency) and the object (or purpose) of the assumed changes of the mutes; (2) the dialectic place of origin of those changes ; and (3) their symmetrical distribution among the principal tribes of the Holethnos. {b) As to the first point, we have seen not only that, on the hypothesis just mentioned, no satisfactory explanation of the individual Mute-changes is forthcoming, but also that the whole movement described as the first Verschiebung, and its assumed repetition in the second, are both, to all appearance, absolutely wanton. Each is a mere revolution, quite incom- prehensible in a series of sounds which are supposed to have been native to, and amply sufficient for, the tribes among whom the movement is supposed to have been effected ; and each exhibits phonetic changes whose possibility is more than 58 grimm's law : a study. [§ 27 {b). doubtfdl. But if the conclusion of § 25 is laid down as a fundamental doctrine, the whole state of the case is immedi- ately and completely altered. The effective agent of all the changes is identified with the ruling principle of debilitation 'under which they, as sound-changes, ought naturally to fall ; and their object becomes the enrichment of language by the multiplication of its phonetic resources : for whereas there had previously been but three mutes (viz. the Tenuis of each family), there were now nine ; that is to say, phonetic provision was made for trebling the expressive capacity of a large section of the Holethnic vocabulary. Such a provision is one which the linguistic sense would immediately avail itself of; and, in fact, it is just here, if I mistake not, that that species of Differen- tiation described in § 17 [d) found an ample field of action. If, with the constant acquisition of new ideas and subdi\dsion of old ones, the alphabet had remained fixed within its original narrow limits, one and the same verbal form would soon, in many cases, have had to do duty for a continually increasing number of meanings ; so that at last the parent speech would have become crowded with clusters of homonyms \ From ^ Baudry, who makes light of the supposed evil of homonyms in a modern language (G.C., p. 145, quoted in note to § 17 ante), treats it (p. 146) as a very serious matter in the parent tongue. " La confusion primitive entre les aspirees et les pures " (he is combating the second aspect of Max Miiller's hypothesis) "est une conjecture qui parait peu admissible, a cause des homonymies nombreuses qu'elle aurait cr^^es. Les homo- nymes par corruption se c6n9oivent et sont inevitables ; mais a I'origine, il semble qu'ils auraient tout brouill^." As nothing is stronger than its weakest part, I might content myself with denying the " il semble " of the foregoing quotation. But as the objection, if there were anything in it, would tell with much greater force against the theory of the text above than against Max Miiller's (for there can be no mistake about the unity of the Tenuis; whereas Max Miiller's uncommonly " indistinct " sound might be all things to all men, — one and yet many at the same time), I will endeavour to do for my denial what Baudry does not do for his asser- tion — i. e. give some reason for it. The objection, then, I think, implies that all the words which in all the I-E. dialects at or just before the Sepa- ration exhibited Aspirates and Mediae, as well as those which actually ex- hibited Tenues, would, if our hypothesis were true, have been found with Tenues in the parent speech ; that is to say, that that speech was always as copious in its vocabulary as all its component dialects were at the epoch § 28 {a),'] GRIMM^S LAW : A STUDY. 59 this fate a way of escape was found in the appropriation of the results of debilitation to the expression of new meanings ; so that the very imperfections of man^s constitution were, as they still are, pressed into the service of his intellect. Not that all such results were thus appropriated ; far from it. So large an increase of capacity for expression (as I have called it) was more than was actually needed ; for a considerable part of the primitive vocabulary, as well as of later ones, must have been of such a kind as not to admit of differentiation. To elucidate this statement, however, it may be desirable to devote a section or two to a somewhat fuller investigation of the mode of extending the powers of language by Phonetic DiflPerentiation than was called for in section 17. 28. — {a) In that section, only a general outline of this prin- ciple is given ; and in Appendix C an attempt will be made to confirm the accuracy of that outline by facts and opinions. The great truth involved in the principle is, that the distri- bution of meaning does not precede and govern, but follows and is subject to, phonetic variation. But although all such distribution is preceded by this variation, it is not true, con- versely, that all such variation is followed by that distribu- just specified. This, again, would really mean that the demands on lan- guage were as multifarious in the infancy of mankind, when men had all their ideas to acquire, as in later ages, when men inherited the ideas which it took very many generations of their forefathers to accumulate. To this I demur ; for whensoever and whencesoever man first appeared on the earth, it is certain that he then had both to acquire all his ideas and to construct or select for them some phonetic expression. Consequently there must have been a period in his history when the possible combina- tions of half a dozen different sounds would far exceed all his re- quirements. It was the accumulation and more precise distribution and subdivision of ideas that continually impelled him to burst through the narrow limits of expression which at first hemmed him in -, and, accord- ingly, instead of supposing that the tracing of the two feebler mutes of each family to a single stronger one involves the permanent existence of nume- rous homonyms in the parent speech, I reversely suggest that the evolu- tion of those mutes by debilitation offered the means of avoiding homo- nyms which must otherwise have come into being, and perhaps of getting . rid of some that actuallv did for a time exist. 60 grimm's law: a study. [§ 28 {b). tion. Whether it is or not, depends partly on the intellectual requirements of the people, and partly on the nature of the several sections of the vocabulary operated upon. If a whole tribe of men of dull intellect remains content with words that express more or less than would be agreeable to clearer thinkers, or if the meaning of words cannot, or need not, be made more precise, then, although the phonetic embodiment of the meaning may change, the same single meaning will cling to all the changing forms. Phonetic variation therefore, viewed in relation to its linguistic uses, may be divided into two species — the fertile (say), and the sterile. {b) Fertile Variation is that already described in § 1 7 (c?) . It occurs when, to a word whose original phonetic form gives off a second, third, or further dialectic form, there was originally attached a vague, wide, and general meaning, which similarly gives off a second, third, and so on, better-defined or spe- cialized meaning — when, subsequently, these meanings be- come severally attached to each of those forms — and when the more definite words ^ thus generated are adopted by both or all of the commingled dialects. To take the simplest case : suppose there are two dialects, D and D', in contact with each other, of which D is the standard and D' the fluc- tuating dialect (§ 21) ; and suppose that in Z)', by the side of the original sound s of any word, there grows up a secondary sound s' : then, if the original meaning m of the word be such as to give off a secondary meaning m', it may (not must) happen that sooner or later the two meanings will become severally assigned to the two sounds. There are, of course, four possible ways in which these two pairs of elements may be combined into words ; viz., m + Sj m' -{-s', m + 's, m + s' ; that is to say, the old sound may cling to the old meaning, and the new sound and meaning may diverge in company (as represented by the first two of these formulae), or (as repre- sented by the second two) the old sound may gradually ^ I use the term "word " to cover sound 4- meaning. It is often used of sound alone, as by Locke, generally, in the Third Book of his great Essay. § 28 {d) .] GRIMM^S LAW : A STUDY. 61 change its meaning, while the old meaning may gradually pair off with the new sound. (c) The distribution of these combinations between the dialects will constitute two different systems, according to the mutual relationship subsisting between D and D'. If that relationship be very close and intimate, so that in point of fact the dialects are still one and the same, and unable to carry their differences beyond the origination of new sounds, then the same pair of words will finally be current in both ; namely, either m + 5 and m! + s' oy m' + s and m + s' ; but if, their co- hesion being looser, they exhibit the partially independent re- lationship described in § 21, then the final distribution may be represented, for one dialect by the first two of these for- mulae, and for the other by the last two. Dj for example, will maintain the parent word m + s, and assign the new meaning m' to the new sound / : while D' gradually changes the sound attached to the old meaning ; and when, through continued intermixture with J), it recovers (if indeed it ever lost) the older sound, it assigns thereto the new shade of meaning. Thus, finally, colligating like meanings, we have — m +s in D=m -\-s' in D') , .^. dV ■ ■ ■ ^^ m'-\-s' in D = m' -\-s in or, neglecting the meanings, D exhibits s and s''\ : : r ^ •. ])' ,) s' and s J where the relationship between the two sounds is^ apparently j Cross Compensation pure and simple (§ 18). (d) Apparently, but not really; for some of the most import- ant conditions (§ 22 (c)) of such Compensation are wanting. Thus, instead of two sounds {s and «'), there is, at the outset, but a single sound {s) common to both D and D'. This may seem a slight matter ; but it is really a most important one ; and correct views respecting it will, I think, elucidate more than one obscure linguistic phenomenon. For in Cross Compensation both s and s' are already native to D and D' ; and the mere interchange of their places in the vocabulary 62 Grimm's law : a study. [§ 29. of jy does not in the least alter their phonetic value. But, in the case we are now discussing, s alone is native to both D and H : 5' is a variation indigenous to D' only ; and it may really be regarded as a foreign sound to D. To the section of people speaking D, therefore, 5' may in fact offer some difficulty ; and, in attempting to catch it up, being unable to reproduce it exactly, they may actually produce only some more or less close approximation to it. But further, as a consequence of the unity of the parent sound s, the phonetic change in each dialect is one of debilitation only, and falls completely under the Principle of Least Effort. D', therefore, is relieved, as it were, of the labour of raising a weaker sound to a stronger, in order to compensate for the descent of the stronger to the weaker : so that the whole process, when tested by the Principle of Least Effort, is less open to excep- tion than the process of Cross Compensation itself. 29. — If now we suppose Fertile Variation to extend to three dialects, 2), Z)', D", of which D' (suppose) gives off a variety s\ and D" a variety 5", of the parent sound 5, and in which the original meaning m gives off two related meanings m' and m" ) then, supposing the dialects, as before, to be closely fused, there will be in them all an exact agreement in both sound and meaning ; or the three resulting words, m-\-Sy m! -\- /, and mf' + s'' , will (if required) become current in them all. But if the re- ciprocal relationship among them is similar to, and, indeed, merely an extension of, that between the two dialects, D and iX, just now described, and tabulated in formula (f), then the verbal relationship may assume the following perfectly regular and symmetrical form : — m -)rs in D=m +/ inD'=m -\-s''mD''^ m' -\-s' „ =:m' +/' „ =^m -\-s „ h . . (tt) m"-\-s" „ =m"-\-s „ =m''H-/ „ J or, the meanings being neglected, D exhibits 8, s\ s" j ly „ 8\ § 30.] GRIMM^S LAW : A STUDY. 63 If the symbols H^ A^ S be substituted for s, s' s", this tabu- lation becomes identical with that of Grimm's Law in § 2 (e). The case it represents is certainly a complex one ; but, given the dialectic conditions, the case^ is one that must occur. The conditions are, it is true, extraordinary, but by no means impossible ; for what is certainly true (as I shall presently show by an example) of two sounds in two dialects may be true of three sounds in three. Indeed everybody who takes in hand to account for the unique character of the phenomena summarized in Grimm-'s Law, expects to find that they result from extraordinary conditions ; in proof whereof we have seen that those who have failed to discover, have sometimes not failed to invent, such conditions. 30. — ^Let us now pass on to Sterile Variation, so far at least as it is due to the nature of the section of the vocabulary ope- rated upon. This would most naturally be exhibited in the case of words already denoting specific things and actions {e. g., patar, irarep-, pater —vater=f adar= father). In such in- stances the parent word, though gradually diverging in sound in difierent dialects, retains the same meaning in them all. If the dialects, after incipient divergence, again become completely fused, one form drives out the other or others (as, e. g., the English father has served the twin-form fader) : but, among dialects tending to division, each preserves its own special variety, whether it be the primitive or a degenerate form; while, as there are no new varieties of meaning to express, no dialect needs to adopt any varieties of form current in the commingled dialects. Hence the phonetic varieties are dis- tributed individually through the individual dialects. That is, supposing D, D', D" , &c. to be the dialects, s, s', s", &c. the corresponding varieties of the sound, and m the unvarying meaning, of a given word, then D will exhibit m-\-s ^ D' „ mi-s' \; . . . . (p) D" „ m^s") and so on, to any number of dialects and sounds. Or, as m 64 >• GRIMM^S LAW : A STUDY. [§ 31 (fl). remains constant throughout^ the variations in D, D' , &c. are confined to sound alone_, and may be briefly represented by s, s'j s"y &c. If therefore for D, D', D" we put our three leading I-E. dialects, then instead of s, s, s'\ we may read, as before, H, A, S, — the CI., or L.G., or H.G. dialect being placed first (§ 2 (e)) according as one or the other maintained the parent sound. But as there was no special reason, outside phonetic action, why, in the case of Sterile Variation, any one dialect should catch up the forms generated in any other, I prefer to think that the fashion of symmetrical distribution was set by the action of Fertile Variation, on the plan de- scribed in §§ 28, 29, and that all sounds were ultimately drawn into the movement, and adjusted according to that fashion or pattern, by the principle of Analogy. 31. — (a) On the hypothesis, then, that A and were ori- ginally but dififerent debilitations of H, we have found, I think, a theoretically satisfactory explanation of the origin and object of the great phonetic movement represented by Grimm^s Law. Its origin is traced to the well-known cause of nearly all sound-change; its object was the extension and variation of the power of expression in accordance with an equally well- known procedure of language. The question arises, by the way, whether any facts in support of such an explanation are producible, not from modern languages, where they are in- numerable, but from the Holethnic speech itself, so far as we are able to restore it. I think there are, and not a few ; for among the primitive forms into which the common I-E. verbal stock has been resolved, there appear many groups or clusters of two or more roots, the members of which differ so slightly, both in form and meaning, that it is scarcely possible to avoid inferring their radiation from some single original centre. Many good examples of such radiation by simple i^oi^e/- weakening may be found in Fick^s Worterbuch *. What we are here concerned with, however, is radiation by con^owflw/- change ; and of this likewise, the traces, if less numerous, are not less distinct. ^ Pp. 943-965. §31(5).] GRIMm's LAW: A STUDY. 65 (b) To quote an example or two : Prof. Max Miiller reflects somewliat severely upon those ^' wlio think it mere pedantry to be restrained by Grimm^s Law from identifying such words as KaX-elv and '^to calV — corvus and 'crow';'' for call and crow imply a (CI.) primitive gar- (whence yrip-v<;, gar-ruhis, gal-lus, &c.), while KaX-elv and corvw.^ imply a primitive kar- (whence Kop-ayvr), Krjp-v^, &c.)^ Kar- and gar-, therefore, must be kept distinct to a period beyond the Sepa- ration. But, teste Max Miiller_, what does kar- mean ?' '' To shout, to praise, to record :^^ and what does gar- mean? "To sound, to praise/^ i.e., two roots almost identical in form are used to express the slightest possible varieties of mean- ing; and some of their remotest descendants still dififer just as little (compare our " to call," which represents gar-, with '^ to hail," and '^ to halloo " from kar-) . Now, what simpler explanation of such a remarkable twofold similarity can be conceived than that gar was originally a mere phonetic varia- tion of kar, and that each was gradually appropriated to express one of a pair of scarcely divergent meanings ? This is not all, however. There was certainly a thu'd CI. root of the same group, — viz. khar-^ iX'^P'y ^^^-)> whence the Skt. reduplication gharghara '^ laughter " and the Gk. ')(e\-Lh(ov= Lat. ^ir-undo; to which answered the O.N. gala "sing^^; A-S. galan; O.H.G. kalan or galan. The parent form, kar- {s) seems therefore to have given off two phonetic varieties, khar {s') and gar {s"). These were caught up by the com- mingled Holethnic dialects and applied to designate varieties (w', m") of the one action of shouting or using the voice (m), which was more vaguely denoted by the parent form kar ; so that from this group we can construct an almost complete exemplification of the symbolical scheme (tt) in § 39 : — Chkar- {KaX-elvY =L.G.Aflr-(^flZ-loo) = H.G.[^«Z-6n4] „ khar-{xe\-L8cov)= „ gar-{gal-Sin) = „ kar- {kal-Q,n) „ gar- {yrjp-vs:) = ,, kar- {call) = „ khar-{chal-l6n) 1 Lectures, ii. 27, i. 416. ^ j^ defence of kh see § 33 infrd. ^ L or X, in the separate languages, generally represents a primitive r. * By the hypotliesis of §§ 52-54 infra, this form in h- would represent an older form in g- (gal-on), which completes the scheme. F 66 GRIMM^S LAW : A STUDY. [§ 31 (c) . (c) Another example is' furnished by the guttural demon- strative. The form of this root was originally perhaps aka ^, which, by varieties of accent, would give off the two forms ak and koj or with vowel-weakening ik and kij and others. Of all these the consonant is the characteristic part ; so that we may put s=k, s' = kh, s" =g. In meaning the word origin- ally covered all deiktic action irrespective of direction ; but afterwards it and its varieties were appropriated to indicate three special directions, — that (yonder), this (by me), and this person (myself: cf. the Americanism "this child,'' or the negroism "dis nigger ''). Thus — CI. a-ki (eVet-vo?, Lat. -ci)=L.G. hi {he) =H.G. (wanting) „ gha {hi'Cj ho-c) = „ ga-, ge- = „ ka-, ki- ,j agd {eycoj &c.) = „ ik (/) = ,, ich^. {d) Perhaps not many cases can be produced in which a triform variety runs nearly through all the three dialects. This is hardly to be wondered at, seeing that every indivi- dual language inclines to certain favourite modes of word- formation and composition, the results of which ultimately drive out older and simpler forms. But any one who may wish to pursue the subject will find many examples of less regular differentiation in Holethnic times. Thus Fick^ points out that nearly half the primitive (CI.) roots ending in -g are by-forms to roots in -k of closely related meaning (e. g., arg- "be bright '^=flr^- "blaze, gleam,'' nearly) ; and again, further on '*, that many in -c? are similar by -forms of roots in -t {e. g., kad- = kat- "faU") ^ ^ See Appendix E. * In tabulating the demonstratives we must bear in mind (1) that the guttural demonstrative threw off the relative series (A^-, kFo- or tto-, quo- or qui- = lj.(j(. hva-,who; as to the tv of which, see itifrcl §§ 55-63) ; (2) that there were other demonstratives, which, especially those from the dental ata (at and ta), ultimately shared with the guttural the somewhat narrow area of meaning to be occupied. » W-B., p. 982. * P. 1000. « Fick, in his treatment of the " detenninative " or final element d, appears somewhat inconsistent both with himself and still more with the probabilities of the case. After expending much labour to show that all the host of roots involving t, w, d, i, u, aiy and au, are derived by pho- § 32 (a) .] GRIMM^S LAW I A STUDY. 67 32. — («) The second point proposed for examination in section 27 is tlie dialect or dialects of the Holethnos in which the phonetic movements leading to the evolution of the systems 2^., 2 , 2^ may have arisen. In the prevalent theory of Grimm-'s Law this question is_, of course^ predetermined. One system (S^.) being assumed as the parent of the rest^ its sounds (Hj A, S) become an impassable barrier at which all inquiry necessarily stops. But if, at an indefinitely remote period, the Indo-Europeans, as I have tried to show, can have had no other Mutes than the Tenues, it is clear that the Mediae and Aspiratse must severally have made their appearance among them at a given time and in a given place. The time must be left out of the question. By the place we must understand (not a geographical, but) a dialectical posi- tion. Beginning with the Aspirate, I observe that the in- fection of Aspiration could not have broken out over all the parent language at once ; for it would then be difficult, if not impossible, to explain why it did not affect the same part of the vocabulary throughout, but affected different parts in different dialects. Further, it would in that case have been netic variation from comparatively few roots involving a, and after sug- gesting" no other origin for the determinant ff, than that it arose from a weakening of k, he proceeds to argue against a similar suggestion in respect of d (viz., that it is a weakened t) chiefly on the gi'ound of the rigid firmness of the Holethnic speech : — " Wir miissten dann fiir eine feme Vorperiode unsrer Grundsprache ein Schwanken der Laute annehmen, wozu uns ihr sonstiger Granitbau gar nicht berechtigt. Alle starkeren Consonantenwechsel und -wandel sindauf die Einzelsprachen beschrankt; erheben wir uns nur eine Stufe hoher .... so hbrt schon fast alle Laut- affection auf . ... So also weiter zuriickschliessend kommen wir zu dem Ergebniss dass die Ursprache vollig lautfest, keinerlei Wandel und Wechsel von harten in weiche oder gehauchte, von Gutturalen in Labiale, u. s. f., unterworfen gewesen sei" (W-B., p, 1000). From all this it would seem that a hard and fast line is to be drawn at a certain epoch, behind which language must be supposed to change no more j which is as reason- able and as probable as if a geologist should select one of his strata as the ultimate limit of cosmical change. The fallacy underlying such views is not far to seek ; but it must here suffice to remark that if language ever presented the rigidity of form imagined by Fick^ man must have been a very dijSferent being from what he now is. f2 68 grimm's law : a study. [§ 32 {b) . native to all those dialects_, and we should have expected to meet with a homogeneity of character and a uniformity of de- velopment which are quite at variance with the known nature and history of the sounds which represent it. Besides,, there is the patent fact that certain tribes^ whose Mute- system, iii other respects, agrees with ^^, possess no Aspirates at all ; which means, on our view, that they never did possess any. For it would surpass belief that those tribes should have taken part in a great aspirating movement, and yet that no trace of Aspiration should afterwards be discoverable among them. (b) In suggesting an answer to the question. Where then did this phonetic infection originate ? it is not without hesi- tation that I venture to diflPer from universal opinion on the point. Nevertheless I think the facts displayed in section 14 leave us no choice but to assign the parentage of Aspiration to the clan or division afterwards represented by the Low Germans ^ I see no better way of explaining the overwhelm- ing share which the spirants (A, th^f) take in the L.Gr. pho- netic system than to refer it either to the great virulence of the original infection (indicating special affinity between the dia- lect and the sound), — or to the earlier period at which it set in, and consequently the greater length of time during which it operated, — or, still better, to both these causes put together. Subsidiary evidence to the same effect is deducible from the regular and early decline of all three Aspirates to Spirants in the purer L.G. dialects, such as the Gothic. There is a harmonious relationship between the latest forms of all the three which well accords with the supposition of a uniform process of debilitation operating on sounds whose original relationship was similarly harmonious ^. ^ I shall continue to apply the abbreviated descriptions CI., L. G., &c., to those divisions of the Holethnos which must have represented the ancestors of the later peoples so designated. ' An examination of the history of the three spirants, h, th,/, through- out the I-E. languages suggests that they are nowhere primitive sounds. Hence it is a weak point in the first-mentioned aspect of Prof. Max Miiller's hypothesis (§ 23 anti) that he supposes the Germans to have had at their disposal — lying by them unemployed as it were— sets of Spirants which, according to him, could never have been Aspirates; but for whose § 33 [b) .] grimm's law : a. study. 69 33. — {a) If, again, we take our stand upon tlie L.G. Aspi- rates as a centre, as it were, and look round upon the varieties of Aspiration furnished by the other I-E. tribes, the scene presented is one of complete phonetic bewilderment. The new sound to these other peoples was an apple of discord. The ancient Indians show us, in the main, a set of three Soft Aspirates {gh, dh, bh), which may all in some cases become the guttural breathing h ; the ancient Greeks, a set of three Hard ones (%, 6, ^) , which in later times became Spirants ; the Old Bactrians generally the three MedicB, and some peculiar Spirants (/, z, w) ; the ancient Italians a couple of Spirants {k and /) as initials, and generally the three Medics within words; the High Germans Old and New exhibit an extensive variety of symbols {ph, pf, /, v, z, tz, ss, ch, A, and others), although the phonetic values of these may have been less various ; while the Old Kelts and Lithu- Slaves exhibit no Aspirates at all, their place being taken by the Mediae. Thus, to sum up : — The Guttural Aspirate, represented in L.G. by hj is represented in the other dialects by kh, gh, X) hj ch, /, g, z' y z. The Dental Aspirate, represented in L.G. by th (]?), is represented by th, dh, 6, h, f, ss, ts {z) , d. The Labial Aspi- rate, represented in L.G. by/, is represented hj ph, bhf<^, pfj /, Vy h, b, w. Most of these incongruities are attributable to the different CI. tribes ; and yet, on the Chronological Hy- pothesis, it is among these tribes we are to look for the pho- netic system which is to be taken as the standard and primitive system to which the others are to be adjusted. {b) The explanation of all this phonetic diversity lies, I think, in the principle laid down in § 28 {d) . A new sound. A, in each family of mutes, sprang up and spread widely in the L.G. dialect. The original phonetic value of the sound may not now be* exactly recoverable ; but whatever it was, it must have been strange and somewhat difficult to the surrounding sudden appearance in their actual form lie does not otherwise account : — <' Aspirates, whether hard or soft, they [the Low Germans] had none; . . . and in order to distingmsh the third series [of mutes] both from the g, d, 6's and k, t, p'a, which they had used up, they had to employ the corre- sponding hard hreaths A, thj and/."— Lectures, ii. 225, 226. 70 grimm's law : a study. [§ 33 (b) . tribes; so that not only the wider CI. and H.G. dialects^ but their component subdialects, although otherwise in unison^ were unable^ when adopting the sound, to agree among themselves in giving to it its precise phonetic value, or indeed any uniform value at all ; and only produced a set of more or less successful imi- tations, except where they gave up the attempt altogether, and took refuge in the Mediae. That CI. subdialect which seems to have approached more nearly than the rest to an exact repro- duction of the Aspirate, if indeed it did not completely succeed. Was the Hellenic. This is no more than might be expected from the fine ear which distinguished the later Greeks, and which was, no doubt, a characteristic of their early progenitors; and it is actually deducible partly from the history of the Greek Aspirates (which, like the L.G. Aspirates, have ended in Spirants) ; and partly from their nature as Hard Aspirates : for whoever will carefully consider the L.G. Spirants, both in themselves^ and as reflected in their H.G. representatives, can scarcely come to any other conclusion than that the parent sounds must have been hard sounds. If then, the origin of the Aspirate is to be assigned to the CI. tribe at all, the Hellenic subsection alone can claim to be its parent. This, however, would still be inconsistent with the small part which the Aspirates play in the Greek phonetic system, and would render it more difficult to explain why the other CI. subsections should have differed so widely from the Hellenic. Those subsections moreover differed among them- selves not only (apparently) in the original character of their Aspirates, but also in their modes of debilitation. These modes do not comport with the supposition that the Aspirates were indigenous; for while the purer forms of L.G. have preserved the gutturalism, dentalism, and labialism of the several Spirants down to the present moment, the Skt, for example, in cases where it debilitated the Aspirates, reduced them indiscriminately to the guttural Spirant h ; and the Latin reduced the (initial) dental, and sometimes even the guttural, as well as the labial. Aspirate (or its imperfect representations of them) almost as indiscriminately to the labial Spirant /. * See Appendices B and D. § 34.] GRIMM^S LAW : A STUDY. 71 34. — The doctrines of the last two sections, if accepted, will settle some very lively controversies,, — settle them, how- ever, by completely cutting away the hypotheses which give rise to them. Perhaps the most vehement of all is that which has raged about the questions, whether the Hard or the Soft Aspirate was the older, and which was the parent of the other. Kuhn^, in reviewing Schleicher, maintains, with his usual sound judgment, that the Hard Aspirate was the older ; and indeed this view is clearly the only one which is consistent with the Principle of Least Effort. Curtius, on the other hand, as we have seen (§ 12) ^, makes the Soft Aspirate the older, and derives the Hard from it. Once more, Grass- mann^ attributes both Hard and Soft Aspirates to the primitive speech, — rightly enough, on our hypothesis ; but on the Chro- nological Hypothesis this still leaves the main difficulty un- solved — viz., why Hard Aspirates in Greek should almost uni- versally answer to Soft Aspirates in Sanskrit. With one or other of these three leaders, but mostly with Curtius, nearly all subsequent writers on the subject hgive sided. For us, however, it is now needless to enter into the controversy. The various sounds of § 33 {a) not being derived one from another, and all being but so many dialectic reflections or attempted reproductions of a new and difficult sound which originated in a commingled dialect, their history and ultimate form differ in various directions from the history and ultimate form of the original sound. Minute phonetic twists in the several adopting dialects, imperceptible at the outset to the finest ear, might still wrench the sound on to widely divergent lines of debilitation; and hence the various termini {which, alone we know) may be considered to be connected by those various lines, not actually with the genuine Aspirate, but with so many imitations standing in various degrees of proximity to it. The Aryan in both branches, and perhaps the H.G. and the Grk, erred, it would seem, on the side of too nearly complete a con- tact, although they may have differed one from another in point 1 Z-S., xi. 308, &c. ^ He discusses the question at length in his Grk Et., p. 423, &c. 3 Kuhn's Z-S., xii. 81, &c. 72 GRIMM^S LAW : A STUDY. [§ 35 («) . of energy ^ . The Italians^ and notably the Latins, seem to have possessed a very defective capacity for acquiring, or at any rate preserving, either the Aspirates themselves, or even their own substitutes for them. These people were perhaps of a dull ear and an unwieldy tongue. Their representations of the Aspi- rates accordingly exhibit not merely too great relaxation, but relaxation of diverse kinds — sometimes of contact and some- times of energy. Not only did the initial Aspirates become Spirants, without exception ; but these Spirants themselves, being indistinctly formed, mostly ran together into the one spirant/; while within words the less decisive efforts to acquire the Aspirate, or, rather, the influence of the vowels between which internal mutes generally fall^, led to the production of the Media. 35. — [a) This phenomenon of the representation of the CI. Aspirate by the Media, in other than the L.G. dialects, calls, perhaps, for a little further remark. The doctrine that the latter sound is the terminus to which the former, by natural debilitation, ordinarily tends, has already been controverted. It is not, even on the hypothesis of Curtius (§ 15), universally ^ The Grk Aspirates could hardly at any time have been weaker than they were in 01. times j but the Skt Aspirates may very well have been stronger. A general and uniform diminution of energy would not appear too violent in the latter language, in which whole families of consonants, imknown to the sister tongues, have been evolved out of the limited set which it once shared with these. But see infrd,, § 45 note. ^ The effects produced upon consonants by such bi-vocalism are in some languages very violent — a fact which is sufficient to render it altogether unsafe, when a consonant so circumstanced presents a different form from what is expected, to venture upon any decisive assertion respecting its history and its earlier forms. In the older stratum of French, as is well known, such medial consonants, even the strongest, have completely disap- peared. In Greek we observe, on a smaller scale, a similar phenomenon in the case of internal a and sometimes v. The corresponding s of Latin did not indeed disappear ; but it clearly indicates the influence of the con- joined vowels by its passage into r. The difference in the termini (r and nothing) rea(ihed by the Latin and Greek s respectively offers an illustra- tion of the e ffect of that slight organic difference in the phonetic value of the same 1 etter in two related languages which I have roughly described as a " minute phonetic twist " (See the first note to the next section). § 35 («) .] grimm's law : a study. 73 true ; and the suggestion (§ 25) that both A and S were directly and contemporaneously evolved from H^ cuts away the main support both of that doctrine and of this hypothesis, which lay in the supposed passage of the A of %^ into the S of ^ But the appearance of Mediae for Aspiratse in Latin, for ex- ample, still leaves the question undecided whether genuine Aspirates, which have certainly become Spirants on a large scale, could also descend to Mediae by a divergent line of de- bilitation. For, according to the hints offered in § 33 above, the sounds representing the Aspirates in most, if not all, sub- sections of the CI. tribe, would not be genuine Aspirates, but only the results of so many more or less successful efforts to acquire a new and strange sound : consequently, according to my views, the internal Mediae of Latin would be evolved not from the genuine Aspirates, but from the approximative imi- tations of them, which alone the Italians, and especially the Latins, were able to produce, such evolution being modified by the potent influence of the conjoined vowels ^ ^ The variations of phonetic value wliich a sound, generally considered one and the same, may really imdergo, according as it is pronounced by natives of various districts and countries, and the effects of such variations upon the destiny of the sound, form a subject which has scarcely been suificiently taken into account by writers on Vocal Phonetics or the Physiology of Speech. Yet these variations form an appreciable, nay a very important, factor in Phonology ; to which, indeed, should be referred, for example, the various forms of attraction which a sound treated as one and the same exhibits for other different sounds in different languages, and ultimately the different history and fate of the same sound even in the related branches of one older language (See the previous note). The training of the vocal organs and the subsidiary muscular apparatus in difi'erent countries generally differs widely in a few points and slightly in a great many — as every one will admit who, whatever his nation, has listened to the early attempts of foreigners to speak his language, and has noticed the strangely different powers, the curious and inimitable twists, frequently given even to the simple vowels, and sometimes to con- sonants. The same is true to a less extent of dialects of the same lan- guage. The lisping Cockney, for example, can no more produce the fine brogues of our northern counties than he can the clicks of the Hottentot. Hence, if a language be imposed on, or adopted by, an alien race, some of its sounds will probably be at once wrenched from their native position ; many more will undergo some less violent, and at first scarcely percep- 74 grimm's law : a study^. [§ 35 {b) (b) But there may have been other modifying causes which ought to be taken into account. I believe too little weight has been given by philologists to the former reciprocal in- fluence of commingled dialects. When two younger dialects in contact (D and D') exhibit related sounds (say s and s'), it is sometimes too readily assumed that one sound (say s') is directly derived from the other {s), even when it is well known that D' (in which s' occurs) must once have been, for an indefinite period, in contact or commingled with some third dialect which also answers by s' to the 5 of D. Now, both the Li-Sl. and the oldest Keltic dialects, as regards their mute-system, belong in the main to the CI. section ; but that they ever possessed any Aspirates may be with absolute safety tible deflection ; and both these and those may experience a totally different destiny from that which awaits them on their native soil. These are conditions which, owing perhaps to the prevalence of such views as are animadverted upon in the last note to § 31, have often been strangely overlooked, but which play an important part in the theory of this Treatise (See §§ 10 (d), 28 (d), 33, 34, 41 (6), 61, ef alibi). [After this note was in type, there appeared in the " Educational Times " for Nov. 1, 1876, a Lec- ture " On the Acquisition of Languages," by perhaps the first English authority on Vocal Phonetics, A. J. Ellis, Esq., M.A., F.R.S., which in some parts offers such an apposite illustration of the foregoing remarks, that I have removed some additional ones I had made in order to find room for a quotation. Mr. Ellis is urging the importance of acquiring in childhood the correct pronunciation of foreign languages ; and, in the case of German, he endeavours to bring out the veiy different phonetic value of letters which seem to the eye identical with English letters, by a minute analysis of a few simple phrases — Brod und Milch ; das Fleisch und die Sauce ; etwas Bramitwein und Wasser—^hich. really " bristle with sounds which few.Englishmen appreciate, much less reproduce ": — " In Brod the trilled r, the long o without any tendency to end in oo, the dental d (that is, with the tongue against the teeth, not against the palate, as in English), going off immediately into a dental t, so that Brodt was the older spelling. In und the u is our common u in full, but we do not have it before n, the long n going into dental dt, as before. In Milch the difference between mil and our mill in the vowel (very short ee) and dental /, and finally the palatal ch after /, which is not ah, nor a voiceless y, though something like both. In das the broad and very short a, not lengthened as in our grass. In Fleisch the very broad diphthong et, almost our aye, the very hollow «cA, hollower than our sh. In Sauce, the final e, more like our a in idea § 35 (b). GRiMM^s law: a study. 75 denied. Instead of those sounds,, both the Li- SI. and the Keltic exhibit a series of Mediae which, as I shall try to show (§ 46), originated independently. Such sections of the Holethnos as the Italian^ therefore, among which Medise occur partially or irregularly instead of Aspirates, may very well have had their earliest and imperfect imitations of the new sounds warped, as it were, or determined by the corresponding pronunciation of other sections, such as the Keltic or Li-SL, which were still in contact with them ^. Precisely the same thing occurs, especially as regards the Dental Aspirate, in many of the L.Gr. dialects, where d occupies the place which in the than anything else, and the peculiar pronunciation of the initial s as a slight s, followed by a full z. In etivas the broader e, dental t, lip w (a, v without touching the teeth, not our w, nor v, nor vw), the distinct un- accented syllable was. In Branntwein the broad a and ei, and the same w as before, the long n before t, the dental t, really dt as before, and the final long n. In Wasser the w and a as before, the final trilled r. It is a very long while, indeed, before English organs can form these sounds readily and without efibrt. The English w is so especially difficult to Germans that I have seldom or ever heard one, even after years of practice, who could produce it. The German lo offers the same difficulties to English- rhen." — As if by way of praxis on these remarks, the public were favoured by " Pimch/' a week or two later, with a sketch of a German tenor warbling " The Last Kose of Summer"; and this is how he is represented as "trans- planting " one of the stanzas : — "I'll not leaf zee, sow lone von. To bine on ze schtem ! Zins ze lofly are szchleeebingk, Coh ! szchleeeb sow fiz dem ! Zos ghyntly I schgadder Zy leafs on ze bet, Vair zy maids of ze karrten Lie schentless and tet ! "] 1 See note 1 to § 32. * The solitary case of mihi, although insufficient for the basis of an ar- gument, because it is solitary, is nevertheless of some interest as indicating that an attempt really was made to pronounce an internal Aspirate. In this particular case it was probably (as Curtius, I think, somewhere sug- gests) a species of dissimilation, i. e. an effort to avoid the close proximity of two labials (niihi) which operated to prevent the formation of h. Pro- bably, too, the effort in this case likewise was assisted by the influence of another closely related dialect, viz., the Skt., where we find mahja (for mahhjd). On the theory of the text, however, mihi proves nothing in re- ference to the Mediae which represent internal Aspirates in other Latin words. 7Q GRTMM^S LAW : A STUDY. [§ 36. purer forms of L.G. is occupied by th (j?, "S) ^ And thus, as the Italians consorted with the Keltic or Li- SI., so various L.G. subdialects which were unsuccessful in catching up the correct sound of the dental Aspirate, consorted with the closely- commingled H.G. dialect^. In short, if we base our idea of the connexion and relationship of the dialects and subdialects of the Holethnos upon our observation of the dialects and subdialects of living languages (and surely this is our safest method), we shall not imagine them to have constituted two or three great dialects perfectly uniform and homogeneous within themselves, but broadly marked off from one another ; much less shall we dream of the Holethnic speech as of one rigid and granitic whole, whose constituent elements were exempt from all change either relative or absolute : we shall rather judge that at some few centres the characteristic dialectic differ- entiae were developed to the utmost point consistent with general linguistic unity, while intervening dialectic areas ex- hibited every possible variety and combination of agreement and difference with each and all of the principal centres. 36. The same line of argument which assigns the origin of A to the L.G. dialect, would assign the origin of S to the H.G. dialect. Its range, both in Sj; and Sy is too limited to justify us in placing its origin in either the CI. or the L.G. dialect, whereas in Sz its range corresponds (or probably once corresponded) to that of A in ^y and to that of H in S^;. Besides, the main characteristic of ^, was (as we shall presently see) Re- sistance to infection ; and as to ^y, the ravages of its own single infection were so extensive that to superimpose a second there- upon would be to represent it as most exorbitantly and even in- credibly corrupt. Being an easy sound, the Media offered no opportunity for that diversity of pronunciation which presents itself in the case of the Aspirate. It might be proposed, however, as a question for speculation, whether S thus origi- nated in H.G. quite independently of A, or whether A may ^ See Appendix D. ' But in this case the possibility of subsequent assimilation must be borne in mind (See §§ 52-54, infrd,). §37(«).] GRTMM^S LAW : A STUDY. "71 not first have made some little progress among the L.G. tribe, and S have arisen from the earliest crude efforts of the com- mingled H.G. section to imitate that sound_, — their efforts, like those of the Latins (§§ 34, 35), being diverted by extra- neous causes to a different line of phonetic development. In this case the two dialects would have been previously in inti- mate combination, their tendency to loosen out into two would have been quite latent, and this new phonetic diversity would simply have made the tendency overt. The dissimilating effect of the correspondence between S in one dialect and A in the other, would still, of course, be precisely the same as if S had originated independently (§ 41 (5)) ; and the sentiment created would be strong enough to excite, at a later period, a more successful attempt to adopt the Aspirate. This view is sup- ported chiefly by the fact that the H.Gr. medial affection at- tacked the same section of the original series of Tenues as did the L.Gr. aspirate affection ; but it appears at first sight inconsistent with the probability that A and S (as will pre- sently be shown) began to infect the principal CI. dialects almost or quite simultaneously. Still this does not neces- sarily prove that they originated at precisely the same moment ; for dialectic peculiarities may exist and even spread in their native dialect for a long time before they acquire the power to affect a neighbouring one. 37. — {a) The third poipt proposed for examination in § 27 is the symmetrical distribution of the related Mutes among the principal tribes of the Holethnos — or, what is the same thing, the cause of the functional relationship existing between the principal Indo-European phonetic systems. This is to be looked for in the dialectic condition, and in the relationship of the dialects, of the Holethnos. Upon the Chronological Hypothesis that condition, up to the very moment of the original fissure, may be described as a homogeneous unity, im- plying a corresponding unity and close cohesion of the people up to the same epoch. Our hypothesis, on the other hand, requires that the dialects of the Holethnos should have already displayed the expansiveness and variation consonant with the 78 grimm's law : a study. [§ 37 (b). character of a primitive language spread over a wide area, and should in fact have reached the stage befitting a people ripe for the division and subdivision to which they actually became subject^ In the latter condition, the rival dialects of a language assume, or tend to assume, coordinate im- portance : and if no external agencies operate to check their expansion and divergence, they ultimately become distinct languages. Such counteracting agencies appear, historically, in various causes of social and political compression, which reconsolidate the diverging sections of people, give to one dialect the paramount importance of a standard or polite dia- lect, and at last bury the rest in obscurity. Every great language of a great people ofiers an example of this process. (b) But what we are here concerned with is the opposite process of ethnic and linguistic repulsion, in virtue of which divergent sections of a nation or tribe tend to become distinct peoples, and their dialects to become distinct languages. In watching this process we observe two linguistic stages which are (or may be) marked by different characteristics : the first extends over the time that the tribes remain in presence of one another ; .the second begins from the epoch at which they part company. In the latter stage (which we will first dismiss, as not much concerning us here), the dialects necessarily cease to have any direct mutual or reciprocal influence. Each now foUows its own line of development as an independent lan- guage; and this (phonetically) consists in the modification and sometimes complete rejection of sounds which, under new ^ " Das spatere Indische, Eranisclie, Griechische, Italische, Keltische, Lituslavische, Deutsche, so wie die iibrigen fiir uns verschwundenen Spracli- zweige, waren bereits auf engem Raume leise von einander gesonderte Mundarten, die noch stets mit einander in enger Btjriilirung standen und vielfach gemeinsame Schicksale hatten. Nicht durch Auswanderung in weite Fernen haben sich diese Sprachen gesondert, sondem durch Dia- lekthildung haben sich diese Volkstdmme einander entfremdet und erst in Folge dieser Entfrenidung haben sie ihre weiteren Wanderungen angetreten " (E. Forstemann : Gesch., i. 241). These sensible remarks, however, are offered merely as an alternative hypothesis in explanation of the rela- tionship of the Li-Sl. to the German on the one hand and to the Eranian on the other. § 38 (a) .] grimm's law : a study. 79 conditions, it finds troublesome to produce. Even in this case the operation of common physical and mental charac- teristics occasionally leads to similar or identical results in the different related languages ; but those results are not corre- lated in a manner implying systematic reciprocal action. In the former stage, however, it may be otherwise; for com- mingled dialects, as we have seen (§§ 18-22), and as we shall again see, have the power of exciting in one another reac- tionary and counterbalancing changes ; so that, when each at length becomes an independent language, it may still show deep- scored marks of its long conflict and ultimate instinctive compromise with the others. Hence, conversely, wherever, among a number of related languages, we discover reciprocal, symmetrical, or compensatory phonetic relationships on any considerable scale, we may conclude that the movements leading thereto must have taken place whea those languages existed as dialects in presence of one another. And thus, in the case of the related systems X^, S , 2^, it may be asserted on this ground alone, apart from any other objec- tion, that the doctrine of independent and successive evolu- tion lies quite outside the bounds of probability. 38. — {a) But in order to conceive with clearness the rationale of the symmetrical distribution of the mutes ex- hibited by those three systems, we must revert to our inquiry into the process of Cross Compensation (§§ 18-22), and to the modification of its results subsequently arrived at in sec- tions 28-30. The former proves language to act in the very manner, but with even greater vigour than, we require ; the latter gives the effect which would be produced upon phonetic distribution if certain conditions of the first-named process were modified. With all this in mind, I propose to resume, from § 18, the examination of the interchange between v and w in so-called " vulgar ^^ English, as compared with the standard and " polite " dialect. To understand its origin, we must recollect the violent assaults to which our lan- guage has been subjected, and the composite nature of the English Vocabulary which is a result thereof. A glance at 80 GRIMM^S LAW : A STUDY. [§ 38 {b) . any English dictionary shows that nearly all our words begin- ning with V have come to us from or through the French as the result of the Norman Conquest ; while all those begin- ning with w are of native origin^ and there fore^ for us, of far greater antiquity. We are thus carried back to a time when in the popular language the sound represented by w was the only labial semivowel. For a long time after the Conquest, as is well known, the ruling class and the subject class spoke distinct languages. But social and poli- tical conditions made it inevitable that both an ethnic and a linguistic fusion should take place. The two languages, therefore, at first gradually, and at last rapidly, approximated to each other, until they assumed the form of two com- mingled dialects. (b) Now observe what must have taken place over a certain area with respect to v and w. The standard dialect, while main- taining the native sound in its place, adopted also the related sound V from and with the language of the court and nobility. The immediate effect of the appearance of the new sound in one stratum of language was to evoke instinctive resistance in the close-lying popular stratum. And this resistance was not content to manifest itself passively by a mere retention of the native w and refusal of the foreign v ; it further proceeded actively to turn (probably one by one, as it acquired them) all the intruding i^-sounds into w^s. But as the two dialects continued in presence of, and active communication with, each other, this transmutation of t; to it; generated (§ 22) a sense of incongruity in the popular ear, which incited to an effort to correct it, — and at the same time a Dissimilating Sentiment, which diverted that effort to the production of a counterbalancing incongruity, in the change of the native w into V : and thus phonetic resistance gradually passed into a new stage whose results appear to be in direct contradiction thereto, and which may be designated phonetic retreat. It might certainly have been expected that a dialect which had evinced such phonetic vigour as to transform a foreign to a native sound, would endeavour still more obstinately to pro- tect its native sound from change. But it did not : the gra- § 39 {b).] GRIMM^S LAW I A STUDY. 81 tification of the Dissimilating Sentiment overrode even so apparently natural a tendency. 39. — («) Examining this example more minutely^ we ob- serve that the polite or standard dialect distributed the two sounds^ w and v, to their proper places in the vocabulary with almost complete exactitude, while the lower dialect re- sponded by an order of distribution precisely the reverse. Hence in the case of the latter we are presented with the curious phenomenon of a cross-transfer of a foreign sound to native words and of a native sound to foreign words. Thus we have ^ (initial) Fiin (one), Funce, Fensday, Fy (why), &c. ; but Voters, ?Fillin (Villain), J^agabone (Vagabond), &c; and (within words) Picki;ick, Artervards, Cambert;el, &c.; but Pentonz^;il, Aggeraz^;ating, Adt^antage, Diz^ulge, Inconz^;eni- ence, Suri«;ive, Ini^ent, Perz^erse, &c. This interchange appears most striking in the case of pairs of words derived from the same ultimate roots ; as Form for ^orm, but Varmint for Fermin ; Ford „ ^ord, „ Wevh „ Ferb; Fine „ ?^ine, „ ^Finegar „ Finegar ; Fay „ ^ay, „ TFojSL^e „ Foyage ; Fant „ ^ant, „ Wain „ Fain; Faggin,, ?Fagon, „ ?Fehicle „ Fehicle ; &c. • {b) The infection of the native w at last became so virulent as to attack it in cases where we might have expected it to be protected by a neighbouring consonant ; as, sz;ig, sz;indle, per- suasion, shallow, ti;enty, &c. It even operated upon the sound of w irrespective of the symbol and of the origin of the words in which the sound occurred; e. g., reki;ire, reki'cst, kvite, kvestion, for reg'wire, &c. ; and so vigorous were its assaults, that where u is the usual diphthongal u [yoo = French iou) it ^ Nearly all tlie examples in the text are " Sam. Veller-isms " taken from " Pickwick." The following is from an amusing Introduction (by " Jacobus Bally, M. A.") prefixed to some old editions of Dalzel's "Analecta Minora " : — " Literas WetV . . . affinitate inter se conjungi nemo bodie opinor denegabit. Rem dudum ad liquidum perduxerunt Cockneyenses, . . . ' TTeal, Fine, and Vinegar are ^ery good Victuals I Wow I ' " 6 82 grimm's law : a study. [§39 (c). first resolved the diphthong into its constituent vowels, y ( = i) and 00 {=w)j and, while preserving the former intact, trans- muted the latter into v : e. g., [Samwel, Samiz^el] Samit^el ; [situation, site?^;ation] , sitivation ; [gradual gradfz^^al] , grad- ival; [punctual punctit^al] punctival; [January JaniM;ary] Janivary, &c. Some of the intermediate forms, however, were used almost as often as the final ones j and a few {e. g,j sitfz^a- tion, actii^ally, Janiz^?ary) seem in some quarters to have been completely arrested in transitu. {c) No doubt there were many exceptions to this substitu- tion even during the time it was in most vigorous action ; and if any competent observer had investigated them at that period, they might probably have been satisfactorily reduced to some rule j e. g.y the pronoun we was seldom if ever altered, perhaps because its continual employment by both dialects had led to their assimilation in this point. So, ^' werry weW/' not " werry vell,'^ the second w being perhaps preserved by attraction of the first. Again, internal v^s in monosyllables were never changed (move, save, &c.), nor yet the second of two successive v's (coni^ivial, surz^ive, &c.) j for a change in such words would have completely destroyed their character, and prevented their recognition by the commingled sections of people. {d) With these exceptions, however, we are not greatly con- cerned ; our business is with the normal process of interchange. It is clear that the maintenance of this is dependent on the mutual relationship between the two dialects affected ; and it becomes interesting to conjecture what must have happened upon given perturbations of that relationship at successive periods in the history of the phonetic movement. If the two dialects had been completely separated after the assimilation oi V io w had commenced, and before the dissimilation of w from V had set in, we should subsequently have found in the one (the upper or polite) a series of v*s and a series of w*^', in the other (the lower and popular) a series of w*% only, which would have had to do duty both for the w'b, of the first dialect and for so many of its v's as had been adopted and transmuted by the lower dialect before its separation from the upper. Again, if such separation had taken place §40 (h3 100 grimm's law : a study. [§ 48 (c) . that are to be found in one or more of the other languages. (2) One dialect may adopt or retain a form which should have been resigned to another, while also retaining or as- suming that which belongs to itself — and this, without having yet distributed the common meaning : hence it may exhibit twin or duplicate roots. A case in point is furnished by the retention, on the part of the CI. dialect, both of the form kar- [kal-), whence Skt ^i-qtr-^. = Zend ^ar-eta = Lith. szal-u, ^^cold'^; and of the form gar- (gal-), whence Skt jal-a, = Lat. ^e/-u=O.H.G. chuol-i=Goth. kal-A-as, and our ^^ cool, cold/' &c. So, too, the CI. dialect appears to have maintained the triple root mak-, mag-, and magh- {mah-), but with an inci- pient appropriation of mak- to denote ^^ length '' {fjL7]K-o<;, fuiK-p6<;)j one species of the bigness or greatness denoted by tiiy-a^, Skt. mah-as (but m«/-man = majestas ; where y=^), Goth. mik-Hs, Scotch muck-le. This irregularity may explain the numerous duplicate roots in the individual languages, such as BeK- and Bex-, '^^'^' ^^^ "^^X'i ^^0- ^^^ '\a(f>-, in Greek. (3) One dialect, while thus adopting or retaining a form strictly belonging to another, may drop or neglect to acquire its own. Thus, in Europe, the 01. form of the first personal pronoun is based on the stem agam ( = eyo), ego, to "which regularly correspond the Goth, ik, H.G. ich), whereas the Aryan (also CI.) pronoun is based on the stem agham (Skt aham, Zend azem), where the aspirate indicates an ap- parent agreement between the Aryan and H.G. dialects ^ (c) None of these cases, however, need give us much trouble, so long as one of the cognate subdialects exhibits the correct form ; for it is more likely that the rest are wrong in exhi- biting an abnormal form than that the one is wrong in exhi- biting the normal form. But when all the subdialects of one great dialect (say the L.G.) are irregular as compared with all the subdialects of another great dialect (say the CI.), or these as compared with those, then arise the questions, Which ^ Other illustrations will be found in Grassmann's learned article before referred to, or in Fick's " Spracheinheit," Abliandlung iv. I purposely abstain, here and elsewhere, from piling up examples which will at once etrike the reader, or may at any rate be found in books easily accessible. § 48 (c) .] GRIMM^S LAW : A. STUDY. 101 are wrong? and In what way are tliey wrong ? Now the irre- gularities which chiefly concern ns here are those comprised under case (3) , with the additional condition that all the subdia- lects on one side or other agree therein. But case (3) may arise from two causes : (i) the retention o£ a sound which should have been changed ; and (ii) the changing of a sound which should have been either retained or else changed in another way^ For the former to occurj a dialect need only cling to the sounds which it actually possesses and has possessed from all time ; for the latter to occur_, a dialect must go out of its way, as it were, to be irregular : it must alter its sounds on a foreign instead of a native pattern. Although, therefore, the latter case, owing to special phonetic propensities, may be frequent enough in the later history of the I-E. subdialects, the former is much more likely to have ruled at the outset of a series of great phonetic movements. Besides, as Dissimilation would not operate instantaneously over the whole of the several dialects, but would permeate gradually through them, the divergence of those dialects may have removed them severally beyond the influence of the rest before the vocabulary of each had been completely transmuted ; and in that event each main dialect would retain some few consonants unchanged. Such retention will also appear highly probable if we remember (§ 36) that each of our phonetic systems, in relation to the other two, was a system of Resistance, — 2^, indeed, almost entirely so. Now, by the hypothesis of this treatise, (1) the term ^'^ retention " can only be used of the Tenuis; and (2), as 2^, S , and S^ are supposed to have radiated symmetri- cally from that single sound, there is no reason why one ^ This division seems nearly to agree with that of Dr. E. Forstemann, whose recent work (" Gesch. des D. Sprachstaninies," vol. i.) did not come into my hands until long after the above remarks were written. He reduces all the exceptions under Stillstand ("suspension") and BescTileunigung (" acceleration ") of the process of Substitution. By the former the L.G. (which with him is, of course, the chief sinner) exhibits the same sound as the 01. dialects ; by the latter it exhibits a sound which should properly belong to the H.G. But his reference of the other systems to 2«^as their primitive, and my reference of all three systems to one set of parent sounds (the Tenues), will be seen to make a radical difference between our views. 102 Grimm's law : a study. [§ 49 (a) . system should be more or less irregular than either of the others. Accordingly, instead of throwing nearly all the irregularities upon one system (2 ) and employing two pre- cisely opposite priQciples^ to account for them, I propose to distribute them, as I have distributed the phonetic changes, among all the three systems, and to show that they may be simply and uniformly accounted for by the single principle of the Retention of the Tenuis on the part of one or other of those three systems. 49. — {a) In order to proceed systematically, let us direct our advance by the well-known article of Lottner before re- ferred to. This article, being based on the Chronological ' Hypothesis, assumes that the exceptions to the first " Laut- verschiebung " {i. e. the supposed transmutation of 2^ into S ) are attributable almost exclusively to the L.G. dialect. In accordance with our hypothesis, however, we will at first vary the assumption so far as to suppose that the irregularities lie between S^ on the one side and S^ on the other, S^ being temporarily left aside. Then, as each of the sounds H, A, S, in each of those two systems, has but one normal representative in the other, it is clear that each sound in the one might, if all possible irregularities occurred, be abnormally represented by the remaining two sounds of the other. This, of course, yields six possible irregularities, making up three groups of a pair each. Thus : I. Instead of A in S^ for H in 2^, there might occur either SorH; II. Instead of S in 2^ for A in 2^, there might occur either AorH; III. Instead of H in 2^ for S in 2^, there might occur either A or S. These groups of irregularities may be represented to the eye as follows, the possible irregular representatives being coupled by oblique lines : — * See the preceding note. § 49(c). GRIMM^S LAW : A STUDY. 103 Group I. Group II. . (yrowi) III. S \ty.. ASH i^Sy.. Ash Ls,..A ^ H (^) Now^ as long as S^. is held to be the original standard and unchanging system from which 2 was verschoberij all irregularities (except where the CI. subdialects dijffer inter se) are necessarily charged upon the latter system only. But if, a^ we have maintained, S^. itself enjoyed no immunity from change, but was evolved out of the one primitive series of Tenues on precisely the same general plan as S and 2^ (although with much less extensive debilitation), then, in examining those irregularities, we are manifestly unable to leave S^, out of the account. Hence the irregularities may, with equal justice, be grouped in such a way as to make 2 the normal system, and S^. a deviation therefrom : that is to say, every pair of the mutes in 2^ may be represented as answering irregularly to one or other of the mutes of 2 . This grouping may be pictured to the eye as follows : — Group IV. j-S.-.H A Group V. ^2 . . H A S is, Group VI. £ .. » A S [c) Each possible irregularity, if supposed to lie between 2^ and 2 only, now becomes susceptible of two interpreta- tions : thus — I. (a) Either ^^ retains Tenues which it should have weak- ened to Aspiratse ; Or 2 weakens (abnormally) some Tenues to Medise instead of Aspiratse (V(7)). (^) Either 2^ retains Tenues which it should have weakened to Medise ; Or 2 retains Tenues which it should have weakened to Aspiratse (VI (7)). 104 GRIMM's LAW : A STUDY. [§49(«?). II. (a) Either S^ weakens some Tenues to Aspiratae instead of retaining them as Tenues ; Or S weakens (abnormally) some Tenues to Aspiratae in- stead of Medi» (IV (7)). (/3) Either S^. weakens (abnormally) some Tenues to Aspi- ratae instead of Mediae ; Or 2 retains Tenues which it should have weakened to Mediae Vll^))- III. (a) Either S^ weakens some Tenues to Mediae instead of retaining them as Tenues ; Or S weakens some Tenues to Aspiratae instead of retaining them as Tenues (IV(S)). (/5) Either %^ weakens (abnormally) some Tenues to Mediae instead of Aspiratae ; Or 2 weakens some Tenues to Mediae instead of retaining them as Tenues (V(S)). (d) We have next to notice that of these possible irregu- larities those comprised under II (a.) and III (a) — indicated by dotted lines in the figured groups 11.^ III., and IV. — do not occur. If it be asked, Why, then, were they introduced? I reply. Because their absence is instructive. It shows, at any rate, (1) that X^ did not admit debilitation at all — either to Aspiratae or to Mediae — when it should have retained the Tenues ; (2) that X did not admit abnormal debilitation to Aspiratce ; and (3) that X did not admit debilitation to Aspi- ratae when it should have retained the Tenues, Let us see how these facts affect the four remaining pairs of alternatives. Owing to the symmetrical relationship of the systems, what- ever is true of one ought, by our hypothesis, to be true of another. We may therefore conjecture, from fact (1), that 2 also did not admit debilitation at all when it should have retained Tenues ; and fact (3) proves our conjecture correct as regards debilitation to Aspiratae, although this is the very form of debilitation to which, by our hypothesis, X should be most prone. Fact (2) gets rid of abnormal debilitation from S , also in the very direction in which it was most likely to occur, i. e. towards the Aspiratae. Hence the probabilities against irregular debilitation to the Mediae on the part of 2 § 49 {e).] GRiMM^s LAW : a study. 105 are very strong. Still stronger is the probability against any abnormal debilitation on the part of ^^j the system of greatest resistance. If these probabilities are to be relied on_, the only cause of irregularity now left^ as between X^ and ^ , is the Retention of the Tenuis, And here we start with the very strong point that group I (/9) (=Vr (y)) admits of no other explanation. We need have no hesitation therefore in applying that explanation^ at least provisionally^ to I (a) (=V (y)) and to II (/S) (=VI (8)) j under which latter case^ however^ there can be ranged only one pretty certain exception (viz. the Skt graM-j Grk 7pt0-o9; Goth, grip-an; A.S. grijo-an; H.G. grei/-en) and about four doubtful ones. {e) But this_, apparently _, leaves the case III {0) (=V(S)) unexplained; for neither of the alternative interpretations of that case in subsection (c) accords with the principle of Retention of the Tenuis. Are^ then^ our probabilities worth nothing ? By no means. The apparent discrepancy is merely an indication that we have not taken into account all the facts we ought. We have so far considered the action of only tivo systems in relation to their mutual irregularities ; whereas our hypothesis requires us to consider the reciprocal action of three. In truth, we can no more neglect the influence of S in the production of irregularity, than the astronomer, in examining the perturbations of the solar system, can neglect the attraction of one of the larger planets. If things so unlike may be compared, each of our phonetic systems may be described, in reference to the other two, as an outlying and perturbing or attracting system, the core, the centre of gravity of which lies in the Tenuis ; while the derived Aspirate and Media are but, as it were, satellites, whose perturbing influ- ence may be taken as nil. Especially is this true as regards the action of 2^ upon 2 , at least according to the hypothesis of this book, which provides (§§43, 45) that the Dissimi- lation between those two systems may have set in earlier than the like influence of either upon 2^, or of S^ upon either, that it may have been more vigorous, and that it may even have continued after the connexion of the German and Classical tribes had been severed (§ 53, infra). Let us then introduce 106 GRlMM^S LAW : A STUDY. [§49 (e). S^ into our system of phonetic perturbations; and let us combine with it, in the following single plan, all the rela- tionships exhibited in the above groups (I -VI.) ;— if) Now let it be granted (1) that S^ retained a number of Tenues which, in order to produce perfect dissimilation with respect to S^ should have become partly Media and partly Aspirates -, and (2) that the Mutual Dissimilation between 2 and 2^ was originally carried further towards perfection than that between 2^ and 2^, or even between 2^. and 2 . What follows from these postulates ? Just this, — that to explain the case III (/3) (or V (S)), 2^ may be substituted as the real perturbator in place of either 2^ or 2 ; and that S in the latter, which really represents a certain number of sounds that were wrenched out of their proper orbit, so to say, by the influence of H in 2^, is thus made to appear (when 2^. and 2 alone are considered) as the irregular representative of S, and partly perhaps of H, in 2^^. ^ For example, the g of hhug- or ^t^y-, whence €vyciv ajidfuffe7'ef is re- presented by ^ in Goth, biug-ajij O.N. hjug-B.^ A.S. bug-aHy — a case of Stillstand, as Forstemann would say. The doctrine of the text above, how- ever, would attribute the irregularity to the retention of k by the oldest H.G. {pioh-2in), and to the adjustment of 2y to 2« rather than to 2x. Again, the t in hort-u&, xopr-os, is represented by d in Gotli. gard-i, A.S. gard, our yard and garden ; and this may similarly be attributed to the influence of t in the O.H.G. cart- or gart-, N.H.G. Gart-en. This form of exception runs, in some cases, through a whole series of words or inflexional forms ; e. g.j to the CI. passive participial termination -ia the L.G. answers hj-da (except the Goth. nom. sing, in -ths) j and here again the H.G. exhibits t (Forste- § 49 (/) .] grimm's law : a study. 107 (/) Hence^ in sum^ we arrive at simple and symmetrical •expressions of all the cases of irregularity treated of in Lott- ner's article ; that is to say_, (i) S^ maintained a number of Tenues which, to make that system exactly correspondent to the other two, should have been debilitated partly to Medics and partly to Aspiratce ; con- sequently its H is irregularly represented in S by H and S (Group I.). (ii) % maintained a number of Tenues which, to make it exactly correspondent to S^., should have been similarly debi- litated in both ways ; consequently its H is irregularly repre- sented in %^ by H and A (Group VI.). (iii) 2^ also maintained a number of Tenues which, to make it exactly correspondent to 2^, should have been similarly debilitated in both ways ; consequently its H^ which attracted to itself S in 2 , is irregularly represented in 2^ by H and S (Group V. and Scheme (^Ir)). mann, Gesch., i., p. 372 ; Helfenstein, C.G., p. 407). Now what reason can be given, on the Chronological Hypothesis, why, on the one hand^ the L.G. dialects, generally so regular with respect to the Classical, should select this line of deviation for their principal exceptions j or why, on the other, the H.G., which, on that hypothesis, is so irregular and incomplete with respect to L.G., should select the very cases in which 2y is irregular with respect to 2x for the exhibition of regularity ? — It should be added, however, that the usage of the L.G. dialects is not strictly uniform. It not seldom happens that some L.G. dialects, in spite of H.G. influence, are regular with respect to 2xf while others have yielded to such influence. A fa- miliar and distinct example of such dialectic disagreement occurs in the forms /adar and mddar ; in which, if they had been regularly differentiated against the CI. patar, mdiar, J? or "S should have appeared instead of d ; and the H.G., if similarly differentiated against 2a-, should have exhi- bited d. But here, too, the H.G. maintained the primeval tenuis {faiar, muoter now Vater, Mw^er) ; and the more active differentiation between the L.G. and H.G. dialects led most of the former dialects (the latter remaining obstinate) to adjust themselves in these cases to the latter. Nevertheless other dialects, especially the Old Norse, appear regular with respect to 2z {fa'^ir, motir). And it is not to the passage of d into ^, but to the existence in this country of dialects closely related to the O.N., that I would attribute the somewhat recent appearance of the spirant in our "fa^Aer" and "mo^Aer" (See Appendix D). Compare the instructive difference in the case of CI. bhrdt-ar, L.G. broth-ar, in relation to O.H.G. pruod'ar=JBrud'erj the last being regular with yespect to the first. 108 GRIMM S LAW : A STUDY. [§50 (a) Thus tlie simplest summation of the whole series of irregu- larities is represented to the eye in the following scheme : — («) 50. — Remarks on the foregoing section. {a) It will be seen upon an examination of Scheme («) that S^ and S„, and likewise 2^ and 2. exhibit a number of X y X z Tenues in common; for to say, as in expression (i) above, that H in ^^ is represented by H in S , — and, as in (ii), that H in 2 is represented by H in 2^, — is saying the same thing. And similarly of 2^ and S^, as correlated in expression (iii) . The nature of the case precludes distinct evidence as to the share contributed by each system to each little common stock. It is not impossible — indeed, on the hypothesis of this book, it is very probable, that, in each case of irregularity the whole may be due to Retention of the Tenuis on the part of one system only. At any rate, such a distribution of the ex- ceptions (scheme (cb)) would exactly fit in both with the relative mutability and with the direction of change which we have assigned to the several systems. Expression (i) would still stand as it does; for 2^ being the system of greatest resistance, and neither A nor S being native to it, the Tenues which it exhibits in common with 2 and 2^ would then be entirely due to retention on its own part, and in no case on the part of the latter systems. Consequently, in (ii) it would suffice to say merely that S retained one or more Tenues which should have been debilitated to Mediae (the character- istic debilitation of 2 J ; and in (iii) that 2^ retained Tenues § 50 (c) .] grimm's law : a study. 109 which should have been debilitated to Aspiratae (the charac- teristic debilitation of S ) . In short, the resistance of each system was exerted (as might be expected) against that sound or those sounds which were foreign to itself. (b) In Group V. the L.G. S appears as representing both H and S of S^. The former irregularity has just been assigned to the retention of the Tenuis by 2^, and the whole of the latter irregularity to the retention of the Tenuis on the part of S^, Hence one might conjecture, independently of other reasons (§ 45, note), that the L.G. had more influence than the H.G. upon the CI. dialect ; also that the mutual action of the two German dialects was at one time more vigorous than their action against any third system, and that it was con- tinued to a later period. For the simplest way of explaining the extensive representation of the S of S^ by S in 2 will now be to suppose that S^. first adjusted its Mediae to the Tenues of % , and that some of these Tenues were subse- quently debilitated to Mediae by way of adjustment to the more obstinate Tenues of S^, which latter Tenues, in order to effect exact correspondence between 2^ and 2^, should, if the mutual action between these two systems had been sufficiently strong, have been weakened to Aspiratae. It is manifest that 2^ could not again follow 2^ ; for that would have involved the impossible change (§ 46 {a) ) from S to A. (c) It will now be obvious that the fact adduced by Lottner in support of Gurtius''s hypothesis (§ 12 ante) — viz., that the substitution of the L.G. S for the CI. A is more thoroughly executed than either of the other substitutions— is susceptible of a totally different explanation. This fact, indeed, is to a great extent explicable by the very small part played by the Aspirates in the CI. Mute-system (§ 14), which of course leaves room for proportionately few exceptions. That the proportion is still smaller is due to the position which A in 2^ and S in 2^ hold in relation to H in 2^ (Schemes (yjr) and (ft))), and to the dissimilating effect of the latter upon 2 (§ 49 (e), and {b) above), whereby not only the Mediae of the L.G. system corresponding to the Aspiratae of %j. are kept in their places, but others also are added to their number. 110 aEiMM^s law: a study. [§ 51 (a). Yet even this powerful influence is counteracted by that of H in S . For,, be it observed^ where irregular representation of the CI. A can occur as the result of Retention of the Tenuis on the part of % , there it does occur (Group II(/S)). Where it could only occur (Group 11(a)) as the result of ir- regular debilitation on the part either of S^. or S^, there it does not occur. That is to say, in accordance with our former interpretation of the facts, ^^ did not admit debili- tation to A when it should not, nor did S admit debilitation to A in lieu of debilitation to S. Or, from another point of view, the Media of S^ was powerless to wrench any of the mutes of 2 out of their places. 51. — {a) The more intimate relationship between the two German systems which I have tried to demonstrate has given rise to facts which have been diverted to the support of the Chronological Hypothesis ^ Some of these, viewed apart from our general scheme, appear to fit in well enough with that hypothesis, so far as it assumes the direct evolution of Sg out of 2 . But with this iuterpretation of the facts are combined arguments drawn from the apparently imperfect evolution of the H.G. system; for it is on this ground that 2^ is generally held to be a comparatively modern and incom- plete variation from 2 , from which (we are told) it was unable to differentiate itself more than partially, with which it pre- served a great deal in common, and to which it has mani- fested a strong tendency to return^. But here the only state- ment that is approximately true is the last : the others are . merely inferences from phenomena which may be interpreted in a precisely opposite way. For there are two stages in the ^ " Le haut-allemand se plie aux exceptions du gotliique, et c'est en prenant celles-ci pour point de depart qu'il fait son Evolution propre. Par exemple, lorsque, par une derogation sans motif appreciable, la racine vedique grabh (' prendre ') se pr^sente en gothique sous la form greipan au lieu de greiban, le haut-allemand greifen a op^r^ la substitution en subissant I'influence de greipan et non celle de grahhj qui aumt donn^ greipen, C'est done au gothique qu'il est subordonn^." — Baudry : G. 0., pp. 148, 149. (But see the preceding note.) » Baudry : G. C, p. 148 ; and Peile : « Grk and Lat. Etym./' p. 157. § 51 (5).] GRiMM^s law: a study. Ill history of commingled dialects at wliicli the phenomena referred to may present themselves — namely^ not only when such dialects are passing from absolute identity, through mere commixture and contact, onward towards complete separation, but also (§ 39 {d) , ad fin.) when, before separation, there arise external causes of sufficient potency to counteract and reverse their tendency to divergence, and to weld them again into one. It is important, therefore, to inquire whether the Mutual Dissimilation between S and S^ may not have been at one time much more completely executed than it appears to have been in the eighth or ninth century, so far as the oldest remains of the H.G. dialect enable us to reconstruct its mute-system at that period ; and whether the extensive and violent changes of a social and political nature among the Germans, High and Low alike, may not, even within our era, have destroyed the phonetic equipoise upon which the maintenance of the completely dissimilated mute-systems depended, and have produced a partial, but now arrested, re- assimilation. Questions o£ this kind require that we should further ask what and what sort of external evidence can be adduced in support of the successive phonetic evolutions as- sumed by the Chronological Hypothesis. I will consider the latter question first. [b) The small amount of external evidence discoverable was originally collected by Grimm himself; and, indeed, it is his interpretation thereof which really constitutes the His- torical Hypothesis of the Law which goes by his name. His object, of course, is to show that there was a time when the L.G., and a still later time when the H.G., consonantal stage had not yet been reached. The former of these periods he brings up to the middle of the first century of our era^, and 1 At least for the eastern Low Germans; for lie remarks (Gesch., p. 483) : — " Was uns von deutschen eigennamen bei Caesar, Plinius, Tacitus, iiberliefert worden ist, halt in der regel den gothisclien con- sonantismus nacii der ersten lautverschiebung fest "; and then follows a list of nearly forty such names. E. Forstemann (Gesch., i. 356), with good reason, throws back the supposed Verschiebungen to an indefinite antiquity : — " Wir behaupten, dass Tenuis zur Spirans [the latest of the three substitutions on Curtius's hypothesis, which F, adopts] bereits in so 112 - grimm's law : a study. [§ 51 {c). the latter up to the seventh century. But after a careful consideration of all that he urges_, I confess myself unable to find in it any justification for such bold and precise con- clusions. Instead of constructing his hypothesis out of facts, he seems to me to deflect facts into subordination to a pre- constructed hypothesis. Indeed most of the supposed evi- dence he quotes seems to me to point towards totally different conclusions. {c) To prove the absence of change from the mute-system of the (L.G.) Getae and Daci in the earlier part of the first century, Grimm relies chiefly on the once popular work of l)ioscori(Jes upon Materia Medica. In that work Dacian equivalents are given for the names of some thirty-two plants ; and of those Dacian words six are considered by Grimm to exhibit still '^ unverschoben '' mutes, viz. : — (1) KpovcndvT], krustane, equivalent to the Greek %e\tSovtov, which Grimm connects with the Lithuanian kregzde, " swallow^'; (2) dirpov^, apruSj equivalent to the Greek ^vp{^, for which he inter- rogatively suggests a connexion with the Latin aper ; (3) BuVf dyUj equivalent to the Greek dKaXr](f>r], and to the Latin wrtica or ^' nettle '' (the Welsh is dyn-ad) ; (4) irpiaBrjXa, priadelttj equivalent of dfiireko^ ixeKaLva, and related to the Old High German /riec^eZa = '^^ arnica ^^; (5) ire^piva, pegrina, another name of the same plant, which Grimm assumes to be related to an arbitrary Low German fagreina, ^^ pretty one,-" from the "L.Q, fagr-, A.S. /«^r, our ^'fair^-*; (6) Box^^d, docheld, equivalent to the Greek 'xafxal'inrv^, ^^ ground-pine,'^ which is made a case in point by a " verschiebung^^ into Gothic tagl, ^'hair, tail,^^ only the meaning cannot be made to suit the plant (^^nur weicht geschlecht ab und schilderung der pflanze,'' as Grimm naively adds). In these words the kjp, djp,py and d are supposed to have been still " unverschoben '^ from the CI. stage. But as it will be necessary to examine in detail some of Grimm^s explanations both of these and of alter Zeit verwandelt wurde, dass selbst unsere altesten Namen keine Spur mehr von der alten Tenuis aufweisen." This is, of course, so much to the advantage of the views put forward in this Treatise. § 52 (b).] grimm's law: a study. 113 other words, and as this can best be done by analysing and examining the whole list, I shall discuss these six when they come before us in their proper order. 52. — The said list then may be divided into several groups^ as follows ^ : — (a) Words useless for our purpose : as opfjuia, a-aXla, oXfia, which contain no mutes ; a-Kiaprj^ where the k is ^^ protected '"'; fjLo^ovXaj " thyme/^ which Grimm connects with our mosSy treating the ^ as 5 (see ^ovoarr], in [d, 2_, ii.) below) ; dviaa-- cre^i (see {d, 3, ii., note)), and KVKcoXlB-a, which, if native and connected with ^'^ cuckoo/' may still exhibit the same consonants as {e.ff.) the Latin cuculus, inasmuch as onoma- topoeias escape '^ verschiebung '^ so-called. (b) Borrowed words : as /8Xr;9 (= Grk ^rjrov) , opfiia (= Grk opfJLLVov) , puavreia, and perhaps fiovSdXXa ( = Grk ^ovyX(oa(rov) and ryovoXrJT-a (= Xidoairepfiov), all apparently from Hellenic sources ; irpoirehovXa is, no doubt (as Grimm suggests), a mere error for irefiirkhovXa, pempedula, the Gallic name (equivalent to the Greek Trevrdcj^vXXov, or Latin quinquefolium) ; to which should be added dyn, also apparently derived from a Celtic source (§51 (c)). In fact, although this word dyn is one of the six pillars of Grimm's hypothesis, he is nevertheless so uncertain about it as to suggest (Gesch., p. 217) the reading of TdXXoij instead of AdKocj for the name of the people to ^ Throughout this examination we must continually bear in mind three circumstances : (1) the popular corruptions of names of common plants, which are often extensive and grotesque, — cf. our dandylion, sparrow-grass, fever-/et<;, &c. j (2) the probable introduction of errors in the successive transcription of ancient MSS., even where the original author may himself have been strictly correct (and here the printed editions of Dios- corides, which alone I have been able to consult, give us no assistance) ; (3) the propensity of the Greeks to give an apparently Grecian cast to foreign words by slight modifications of the spelling ; e. g., 'lepo-aokvjxa for Jerusalem, as if from Up6s ; AevKavia, as if from Xcvkos (See an article — "Die Veranderung lateinischer Eigennamen im Griechischen " — by Dr. F. Strehlke in vol. i. of Kuhn's Z-S.). After making due allowance for these qualifying circumstances, we must esteem ourselves fortunate if we discover that a moderate percentage of the words in the List still enable us to assign them their parentage with an approach to certainty. I 114 GRlMM^S LAW : A STUDY. [§ 52 (c) . whom it belonged. To the words borrowed from the Greek he would add K6pKepav (= Grk avayaWl^), as if from d(j)pci)v ; but Kcpacppcov is so suspiciously like the Latin cere- brum (-^rum for -^Aarum) , that I can hardly help attributing the word to a source more nearly akin to the Italian dialects: the initial xep- then becomes a mere reduplication. To a simi- lar source (Aedic?) should probably be assigned KoaBafia, where the Koa- ^ is almost certainly truncated from aqua, and Bdfia is perhaps =^^ dweller '' and akin to domus : compare the Greek name for this very plant {iroTa/jLo-ryeiTcov) . (c) Words hitherto quite undetermined, and perhaps un- determinable ; as Koriara ( =gramen) , BceXeia ( = voaKvaiiio<;), and KapoirldXa {=KaTavdyK7]=herbafilicula). {d) Words that have a distinctly German aspect. These may be arranged in three clusters : (1) Low German. — (i.) ir6\'irovL6oi6o-e6e\d, fitho-fethelay — an un- mistakable approximation to a reduplication {cf. " rathibida,^' (1, ii.,) amd K€pK€papcoy in (b) above), based, I would suggest, on the L.G. form fithra or fethera = our feather = H.G. feder. A name so derived would at any rate be admirably descriptive of the plant. (vi.) irpohlopva = Grk iWe^opot; fieka^ = Lat. veratrum nigrum. Here -orna is a mere termination. For prodi- its connexions are quite uncertain. ANIA2- may indeed have been easily corrupted from AMA2, or even from AMIA2 ; and in A-S. amas denotes a weaver's rod (the plant was rushy) ; and ameos is already the name of a plant, perhaps bullwort ; which again brings to mind the umbelliferous genus Ammi, — a name which occurs in Pliny. The Grk name of the plant, however, was opoQpvxis, the smell being supposed to " make asses bray " (Donnegan) ; so that some mutilation of Svos or astnus (or a Ger- man equivalent) may lie hidden under dpiaa--. § 52 (c?).] GRiMM^s law: a study. 119 Grimm refers to Goth. fr6)7s ^^ sapiens '' (although the quan- tity of the 6 is against it) ; so that prodi-orna would be ^^ das klug machende " — an ingenious allusion to the employment of hellebore among the Greeks and Romans for the cure of insanity. Whether the Dacians similarly used the plant we do not know. But if this derivation is correct,, the d is in the H.G. stage^ and the p apparently in the CI. The ques- tion therefore arises^ Which letter shall be adjusted to the other? Grimm^s mode of treatment requires the d to be considered an error for a CI. t. But as we have not yet found in this list of words any other clear example of CI. con- sonants^ while many German ones have presented themselves, we are bound, I think, on the contrary, to regard the d as at the H.G. stage, and to adjust the p thereto. It seems, however, at first sight, difficult thus to reduce the p ; for, as is well known (§ 16 (c)), where the L.G. exhibits /, the oldest extant H.G. generally exhibits /also; hence the L.G. frd]f- would, in the known forms of O.H.G., appear asfrod-, I must here confess, however, that I propose, in the sections next ensuing, to show (what is in fact an essential part of my hypothesis, namely) that there was a remoter time, behind that of the oldest known H.G., when the ideally correct H.G. forms with initial b and ff (like those with initial d) were actually in use. Our difficulty here will then be reduced to a much smaller one — viz., that tt occurs where we should expect /3. But this may be (1) because the Greek jS (which often represented the Latin v) was too soft and unsettled a sound to represent the firm German b, and hence (See (ii) and (iii) above), tt may here have been employed for that pur- pose : or (2), still more probably, the tt may simply be due to the unconscious assimilation, on the part either of the author or of a transcriber, of the first syllable to the Greek preposi- tion TTpo-y as in irpo-TreSovXa (See (b) above). I should take the word therefore, if derived as Grimm suggests, to be really Brod-ioTna, ; where bi^od- is the primitive and ideally correct H.G. for the L.G. froth-. It is not impossible, how- ever, that prod-iorna may really involve reference to a de- coction made of the plant for medicinal purposes; in which 120 GRIMM^S LAW : A STUDY. [§ 52 {d). case prod is the O.H.G. equivalent (actually extant) of the O.N. and A-S. bro« = our ^^ broth '\ (vii.) 7rpLaBrj\a= Grk a/zTreXo? fjieXacva: apparently a sort of wild vine. This word offers, perhaps, a clearer case of relationship to a German word th^n any other in the whole list. The oldest extant H.G. related word is of the form friedilaj ^' arnica/^ which agrees with the O.N. fridill as to its consonantal stage. The related CI. forms exhibit p and t; as Skt jore^ar = O.Sl. jorija/eli, " amicus ■'^j so that the L.G. forms should exhibit / and th {cf. the kindred words Skt jt?riya/va=Goth. /rija^Ava), and pure H.G. should exhibit b and d. Consequently (excluding the notion of Lithu- Slavonic influence) we meet with precisely the same pheno- menon in joriac?ela as (according to Grimm^s view) in the fore- going joroc?iorna ; i. e., the internal consonant is at one stage, the initial at another. To harmonize the discrepancy I would again, for the same reasons as before, make the tt = a hard b, and read Briadela as the primitive H.G. form, the b of which in later times was assimilated to the L.G./. (See §§ 53, 54.) (viii.) ireyplva: another name for the plant designated by the preceding word. Grimm refers us to the Goth, and O.N. fagr-j A.S.f(B(/r-j (our " fair,^^) and the O.H.G./«^ar. Here again the inner consonant {ff) is properly L.G. ; but by the influence of the r would become H.G. also; while the initial p is to be supposed " unverschoben," or arrested at the CI. stage. This is the fourth or fifth discrepancy of the kind ; so that, if Trey/)- really represented fagr-, I should say once more that tt is for a hard /3 ; and that begr- represents the primitive H.G. stem. But the connexion is far from indubit- able, especially as it gives a second (and somewhat inap- plicable) figurative name to the same plant. I am rather disposed to regard this name as a literal and concrete one, which alludes to the fruit or '^ berries " of the plant. In most of the old German words for ^*^berry,'^ as the Goth. {basja)j O.N. [ber), and O.H.G {peri and beri), no g indeed appears ; but the A-S. is berie or berige, which means ''grape'' as well as " berry '' ; and the genitive plural is berigena or § 53 {a).] GRIMM^S LAW I A STUDY. ]21 bergena. Now we have already seen a remarkable coinci- dence or two between tbe Dacian and the A-S.^ as in the case of Bo'x^eka = A-S. )7acele {2, iii), and perhaps of rovX^rjXd = A-S. delf-an (3, i). Possibly^ therefore^ in the present case likewise TreypLva may be (by an easy and probable trans- position of P and r^) for j9er^-ina, a H.G. form exactly equi- valent to a L.Gr. berg -m.2iy i. e. berig-insi = "grape- or berry- plant/' The appropriateness of such a connexion is obvious. It should be added that Bopp ^ traces the word ^^ berry " to the Skt bhaksjam {i. e. bhag-s-ja-m^^), akin to the Gk (f>ay-f which accounts for the guttural^ although it introduces irre- gularity on one side or the other. {e) Thus^ with fewer and less violent assumptions than are necessary in order to reduce the mutes of six refractory words to the CI. stagCj we have found, I think, two [itoXttov^ and padi^iha) which are distinctly L.G., three which are probably so {artKovTTvoe^, BaKiva, and ^ido^OeOekd) j five which are almost certainly H.G. {KpovaTavrj, ^ovoarrjj 8o')(^6\a, d-rrpov^j and cre^a), and five which are less certainly so {rovX^rjXdf TovTaa-rpa, irpohiopva, irpLO^rfka, and ireypivd). The very words, therefore, by whose aid Grimm proposed to prove that the L.G. consonantal stage had not in the first century been reached among the Dacians, seem to have turned traitors, and to show, if they show anything, not only that L.G., but even, and still more [mirabile dictu), that H.G. sounds were prevalent among that people. 53. — (a) This, of course, is completely at variance with Grimm's proposition that the H.G. phonetic system was not evolved until the seventh century after Christ. It behoves ^ Transposition is almost a characteristic of the Greek language within itself. Thus ^ is often = o-k, ^ = o-7r, ^=a-S ; stems in aiv-, aip-, &c.,plus a vowel, are for older forms av-ya, ap-ya, &c. The form begera (=beri- gena), however, in which g precedes r, actually occurs in A-S. (See Bosworth). 2 V. G., iii. 343 (on the suffix -yd) : — " Von Substantiven gehort der Neutralstamm hasja ^Beere' hierher, wenn es, wie ich vermuthe, dem Skr. Vdhi-ya-m ' Speise/ eigentlich ' zu essendes ' entspricht (von Vaks' * essen,' Gr. <\)dy-(o) und des Gutturals der Wurzel verlustiggegangen ist." 3 Fick, W-B., p. 133. 122 GRIMM^S LAW : A STUDY. [§ 53 {b)- US therefore to examine the additional arguments by which he attempts to support his proposition. The chief of them amounts merely to this, — that no traces of that system are anywhere to be found ^. An obvious reply is, that non- appearance and non-existence are very different things, and that to reason from one to the other is quite fallacious, unless the reasoner is prepared to assert his own omniscience. As re- gards the case before us, any such assertion would be strangely inconsistent with the scantiness of our information respecting the early Germans, and especially respecting their language. Here is a numerous and powerful cluster of tribes that early in our era spread over half Europe ; and yet little remains, beyond a score or two of proper names, to shed light upon the texture of their speech. Who knows what diversities of dialect prevailed, not only among these scattered tribes, but even among different sections and strata of one and the same tribe ^ ? Such an argument as the one just referred to can only, at the best, hold its dubious footing until the smallest piece of distinct evidence to the contrary is producible. Such evidence, I think, is furnished by the Dacian terms as explained in the preceding section ; and it certainly seems to me that the accumulated force of all of them together is not easily to be evaded. {b) Two points in connexion with those words may perhaps offer a diflBculty : — first, the fact that, for some of the foregoing explanations to hold good, certain consonants must have been ideally correct H. G. forms rather than the really L. G. forms which now to a great extent characterize H.G-. This, how- ever, I hope almost immediately to show to be in strict har- mony with the probabilities of the case. The other point is the mia?ture of H.G. andL.G. words among the same people; ^ " Niemals aber erscheinen die laute der zweiten, d.i. ahd. verschie- bung, die also im ersten jh. sicker nicht entfaUet war." — Gesch., p. 483. * " Von den ersten Anfangen der geschichtlichen Zeit an hat es in Deutschland eine Menge stark ausgepragter Dialekte gegeben, von denen jeder in einem beschrankten Bezirk Alleinberrscher war, und ebenso viel, richtiger ebenso wenig, Anspruch auf den Namen der Deutschen Normal- sprache hatte als alle iibrigen." — Whitney (Jolly), p. 242. § 53 (c).] grimm'slaw: a study. 123 for while the bulk of those words apparently exhibit H.G. consonants, a few apparently exhibit L.G. consonants. But, remarkably enough, the earliest extant specimens of O.H.G. present unmistakable traces of a similar mixture. The ex- amples from the Prayer of Wessobrun and the Song of Hil- debrand, and the curious rules of Notker, are well known. Grimm-'s interpretation of this phenomenon is that in the ninth century the H.G. substitutions were not yet completed) and, asks Baudry triumphantly, ^^ des faits de cette nature [e, g. dat for daz] ne tranchent-ils pas la question ^^ ? ^ By no means. That interpretation, as has been said (§51 (a)), is only one of two of which those facts are susceptible. Instead of indicating incomplete evolution of one dialect from another with which it was previously identical, they may indicate incomplete co- alescence of two dialects that were previously quite distinct. (c) The supposition (to treat of this first) that %^ was less completely evolved from 2 than 2 from S^,, is arrived at by combining what is known of the H.G. Mute-system since the ninth century, with what is imagined about it before that time. All we can rely upon, however, is what we know ; and that is, the recent and partially-executed tendency of S^ to assimilate itself to S . In popular views on the subject that tendency is sometimes represented as innate and spontaneous, and (like the original evolution^) as the result of some inherent necessity. And so the supposed original movement of S^ away from S and its later movement back again towards S , although ex- actly opposed to each other, are both, it seems, to be assumed to have resulted from equally natural impulses. In order to discover whether such assumption stands upon a secure founda- tion, we must, instead of giving undue importance to individual words, rather take a wide view of the general tendency of the H.G. Mute-system within historic times. This, if I mistake not, points to quite a different conclusion. That there has been a considerable approximation of the H.G. to the L.G. Mute-system during such period, is a fact of the utmost im- portance in relation to our discussion; but its importance lies in 1 G. G., p. 149. * See § 8 anU, especially the notes j see also § 64. 124 GRIMM'S LAW : A STUDY. [§53(rf), ttis, that that approximation was gradual and continuous from the very first appearance of the H.G. dialects in literature up to the comparatively recent time when those dialects began to assume the unity and fixity of a mature language. As to any previous phonetic movements, there is, of course, a total absence of external evidence. This being the case, one would have thought that sound reason would have led men to infer that that phonetic movement extended uniformly backwards into remoter ages, and not that it was at once reversed — ^unless, indeed, we are to act on the dangerous principle already abjured (§ 8 (e)), that what takes place when history is silent is just the opposite of that which takes place when history speaks. [d) To make clear what has just been urged, let us expand S^ and S^ into the two series of individual consonants which they symbolize, appending the correspocding forms assumed at later epochs by the H.G. Mutes, thus : — 2, k h [S^ (ideal) . ch g O.H.G-. (9th cent.) , . ch H M.H.G. N.H.G. (init.) KH (med.) ch H (final) ch CH C (init.) KH G (med.) chH G (final) chH G This table ^, if I mistake not, tells a very distinct tale. Omit- ting the dental series for the present, we see that the latest H.G. forms agree with the L.G. to the extent of two thirds or six ninths as regards the labials, and, as regards the gutturals, to the extent of seven ninths; but going backwards in time to the ninth century, the degree of assimilation is reduced to one third or three ninths in each of those two families. As, then, we ^ It is really Grimm's (Gesch., pp. 424-426). The ideally correct 2« is introduced for comparison. The capitals in the lower groups indicate points of assimilation to 2y. § 53 (e).] GRIMM^SLAW: A STUDY. 125 have no evidence of any contrary movement immediately previous to the ninth century, the obvious inference is that, if we could trace the dialect still further and further back in time, even this amount of approximation to S would at last disappear, and we should arrive at the ideally perfect %^. (e) This coincidence of H.G. with L.G. in respect of h and /, when interpreted to mean incomplete evolution of %^ from 2 , is a serious stumbling-block in the way of some forms of the Chronological Hypothesis. Its antagonism to the views of Cur tins has been already pointed out (§16 (c)) ; and its bearing on those of Mr. Sweet will be considered in Appen- dix D. I may remark in passing, however, that it is equally inconsistent with the direct evolution of S^ from 2 by Cross Compensation (§§ 18-22). For if, in both the guttural and labial family, one member (the Aspirate) was never altered, then, in a case of evolution of the kind just men- tioned, the other two members should have been simply in- terchanged j that is, to the L.Gr. k g and should have answered \ \ H.Gr. g k and h p. whereas we actually find L.G. k g and p answered to by i i H.G. ch k and ph Now why should or how could 2^, which duly answered to g by k, answer to khj ch? And if it already possessed the L.Gr. aspirates h and /, why should it go out of its way to create an incongruity by evolving a second set of aspirates ? The incongruity becomes stiU more incongruous by compari- son with the exact regularity of the dental family, in which the third member occupies its proper place, and which there- fore is ideally perfect with reference to the dental family of S . Add to all this the remarkable fact that the purest extant O.H.G.^ must thus have spontaneously left itself without ' " StreDgalthochdeutsch " or " Alamannisch " (Schleicher, "Die D. Spr.," p. 97). )h pi 126 grimm's law : a study. [§ 54 (a) either g or b^. It will hardly be urged that the old h and /were on their way towards these mediae. For, at the outset, the possibility of such a transition, as a matter of natural sound-change, may be safely denied ; and, again, if they were on the way, why did they not proceed therein, as, by hypo- thesis, the dental aspirate must be represented as doing ? The only satisfactory solution of all the difficulties seems to be, that at a remoter period both the guttural and labial family, just like the dental family, were completely differentiated at the same time, and in precisely the same way, as the corre- sponding families of 2^ and 2 , and consequently that the appearance of L.Gr. mutes in H.G. is the result of a later partially-executed assimilation of the latter to the former, — in other words, that the gradual assimilation of 2^ to 2 which is traceable during comparatively recent times is but a continuation of a movement which must have set in much earlier. 54. — (a) This proposition, in its relation to historical evi- dence, requires perhaps a little further discussion ; for it has very important bearings in many directions. Let us begin by summing up the leading facts or phenomena which we have already met with in connexion with the subject. These are, I think: — (1) The existence of (apparently) H.G. forms at the beginning of our era (§ 52) ; (2) The appearance of L.G. and H.G. forms side by side, both then and centuries later, in the same localities (§ 53 {b)) ; (3) The non-appear- ance {i. e., so far as our scanty knowledge reaches) of any H.G. tribes in history in the earliest part of our era ; — (4) The pro- gressive, but at length arrested, assimilation, within recent times, of the H.G. to the L.G. Mute-system. {b) Now, with all these facts the hypothesis of this treatise is in strict harmony, and they with it. As to the first, it is clear that if 2^ was evolved simultaneously with 2^. and 2 out of a common central mute of each family many centuries, perhaps many chiliads of years before Christ, and was found existing many centuries after Christ, it must have existed at 1 Schleicher, "Die D. Spr.," p. 100. § 54 (c).] grimm's law : a. study. 127 all intermediate periods. The discovery that H.G. words existed during such periods is therefore just what may he expected whenever any fresh sources of information become available. (c) As to the second^ it is a fundamental condition of our hypothesis that the dialects in which a functional phonetic relationship is mutually and simultaneously evolved should be actually in presence of one another and even commingled ; and the phonetic equipoise^ once established, may be main- tained as long as that commixture remains at or about the same degree of intimacy. It may happen, as I have more than once remarked, that the dialects, after the evolution of their reciprocally adjusted Mute-systems, may not only tend to sepa- ration, but actually separate ; and this was the case as regards the CI. on the one side and the German, both High and Low, on the other. Or, on the contrary, social and political causes may check the tendency towards separation of the peoples using such dialects, and may have the effect of reducing or tending to reduce linguistic diversities to unity; and this seems to have been the case as regards the High and Low Germans with respect to each other. The incipient divergence between these two peoples indicated in the figure to § 46 (c) was carried no further ; but in their migrations westward the one tribe or cluster of tribes threw in their lot with the other ; and the resulting commixture ultimately became too intimate for the preservation of the phonetic equipoise originally esta- blished. I would suggest, in fact, that X^ and S^, within and even before our era, were the characteristics, not so much of distinct tribes, as of different sjfata, so to say, of one and the same tribe or group of tribes : just as the Cockney dialect (§§ 38. -41) and the polite English dialect are (or were) spoken by different, but overlapping and even commingled, strata of the inhabitants of one and the same town. Nay, further, as polite English is to the dialect just named, and as the CI. dialect was both to the L.G. and to the H.G. (§ 44 {a)), so, I am inclined to think (in contradiction to their designations), was the L.G. to the H.G., the Low German being the language of the higher stratum of the people, and the High German of the lower. The 128 grimm's law : a study. [§ 54 (d). further divergence of the dialects having been thus arrested, the linguistic difference between the different strata of people became, for the most part, merely phonetic, — and phonetic, for the most part, only so far as the mutes were concerned. And this difference is perfectly consistent both with phonetic agreement in the other sets of vocables, — liquids, semivowels, vowels ; and still more with that close relationship of grammar and vocabulary which must, early in our era, have characterized all the German tribes. I am not suggesting (for it would be going beyond all probability) that a H.G. substratum extended with and among all the L.G. tribes. Many of these were pro- bably almost or quite free from such admixture. But there are others to which it may be more especially attributed ; and such, perhaps, were the Goths and the nearly related, if not identical, Daci. For close as is the grammatical relationship between the German dialects iu general, that between H.G. and Gothic is the closest of all^. Hence it comes to pass that in the list of Dacian words from Dioscorides, the number of those which have apparently a H.G. form is so comparatively large. These would probably be the popular names of plants current among the lower people. (d) The most important bearing of the last subsection is upon the third and fourth of the facts in subsection {a). As to the third, it may now be conjectured why the H.G. dialect has left no evidence of its existence in those far-off ages. Its footing, if we are right, would be that of an obscure patois widely prevalent, perhaps, but among the humblest classes; and its phonetic peculiarities^, if noticed at all, would be despised as vulgarisms^. The consonant-system, of the lan- guage used by kings, nobles, priests, and warriors would be that of the L.G. dialects; and the names of the leading men, tribes, towns, and districts, which have been preserved for us ^ Helfenstein : C. G., p. 7, and throughout the book. ^ The old-world character even of N.H.G. harmonizes well with such a view of its former position. " German is, in many respects, much more archaic than Middle English, and may be said to stand to it in almost the same relation as Old English [Anglo-Saxon] does." — Mr. H. Sweet, " Hist, of Engl. Sounds," p. 40. § 54 (C?).] GRIMM^S LAW : A STUDY. 129 by the Romans^ naturally exhibit the same phonetic charac- teristics. How then came S^ to the front ? The cause may easily be found in the wars and still more in the great migra- tions of the different German tribes during the earlier cen- turies of our era^ England, Denmark, Scandinavia, France, Italy, and Spain were all more or less completely occupied by German tribes, whose original home was nearer the centre of Europe. Both war and migrations (which at the outset were really but military expeditions) would thus draw off, never to return, the very classes^ among whom L.G. had its stronghold. The effect, at least in southern and south-eastern Germany, must have been nothing less than a social revolu- tion : and in those tribes which originally contained a con- siderable H.G. substratum, the corresponding dialect would forthwith acquire a decided preponderance^. ^ " So viel steht feat, dass alle germanisclien Dialekte einen selir starken Ortswechsel. . . .in jener Epoclie erfahren haben, als die germanische Welt in ihren Vernichtungskampf mit der romischen eingetreten war, und die Volkerwanderung die deutschen Stamme von dem einen Ende Europaa bia zum andern fiihrte. Ganze Stamme, darunter die einat so machtigen und beriihmten Goten und Vandalen, verloren ihre nationale Selbstandigkeit und gingen in anderen Volkern unter ; mit ihnen starben ihre Sprachen aua. Aua Biindnisaen und Wanderungen, aua inneren Kampfen und auawartigen Eroberungen ging im Laufe der Zeit eine Reihe von Ver- scbmelzungs- und Austilgungavorgangen, von Trennungen,Vereinigungen und Fusionen hervor, aber ohne daaa man sich dadurch der Einlieit gena- hert hatte; und vor vierthalb Jabrhunderten, zu der Zeit ala daa jetzige Deutscb sich zuerst als die allgemeine Landesspracbe geltend zu macben anfing, berrscbte in unserem Vaterlande nocb ebendieselbe babyloniscbe Sprachverwirrung wie um den Beginn der cbristlicben Zeitrechnung." — Whitney (Jolly), pp. 242, 243. ^ Not all, of course, but perhaps the bulk of them; and, on the other hand, a considerable fraction of the H.G. substratum may have accom- panied some of the migrations. ^ May not some of the earlier movements among the Germans, of which we obtain an occasional glimpse in history, be reasonably attributed to the coexistence on the same soil, but with imperfect amalgamation, of such dia- lectically divided strata as the above remarks suppose ? Take, for ex- ample, the well-known case of the Batavi, ^^ pars Chattorum seditione domestica pulsi," as Tacitus says (Hist., iv. 12), who settled on the Low- lands about the mouth of the Rhine, where to this day the people are 130 GRTMM^S LAW : A STUDY. [§ 54 (e). (e) If now a subsequent amalgamation of the remaining people took place, what could ensue but the process repre- sented by our fourth fact ? To talk of a language or dialect (as some do of H.G.) setting out on a given line of phonetic de- velopment and, when at or near the end, turning round and, proprio motu, retracing its own path, may not be sheer non- sense ; but it is certainly at variance with all that we know of unconstrained linguistic movement. A homogeneous lan- guage, undisturbed by external influence, should move on in one and the same direction, in accordance with the physical or metaphysical tendencies which have contributed to set it in motion ; and when a language does not so move, we may al- ways expect to find some source of perturbation external to it. If then in the 9th century the H.G. Mute-system ap- proached much more nearly to the ideal %^ than it did in later times, and if in the 19th it exhibits a considerable diver- gence therefrom and an extensive assimilation to ^ , we may be absolutely certain that this phenomenon can have re- sulted from no other cause than the action of the dialect of which ^ was the characteristic upon the dialect of which %^ was the characteristic — in short, from a fusion of dialects previously distinct ; and the equilibrium ultimately reached in N.H.G. is the result of a compromise between the two. In point of fact, we have here an admirable exemplification of the formulation of § 10 (c) . (/) The general character of this compromise is stamped upon the table in § 53 {d). In the case both of the Gutturals and Labials, the H.G. gave way. Its first concession in both families was clearly the adoption of the L.G. spirants in place of purely L.G. ; while those (or the bulk of those) who remained behind were the ancestors of the H.G. Hessians of later times. So again perhaps with respect to the rising that drove Ulfilas and his Gothic fol- lowers across the Danube. In this case, indeed, religion rather than race may have led to the tumult ; but although history tells us absolutely no- thing further about the persecutors, these certainly occupied the territory, and were doubtless the descendants, of the very Daci among whom (if I have understood Dioscorides aright) both H. and L.G. forms were current side bv side. § 54 (/).] grimm's law: a study. 131 its own mediae. In the case of h for g it is difficult to deter- mine whether any, and^ if any, what, intermediate phonetic stage was required, or was possible. But the passage from b to / could hardly, by the Principle of Transition^, have been immediate : for 6 is a stopped, / an unstopped, and therefore (in this respect) weaker, sound -, on the other hand, 6 is a soft, / a hard, and therefore (in this respect) stronger, sound. Consequently the first stage in the passage must have been the unstopping of b — a debilitation, yielding the soft spirant V ; and the subsequent raising ofvtof was due, of course, not to natural development, but, as it were, to external violence — that is, to the assimilating influence of the com- mingled L.G. dialect. Here, I think, we have the explana- tion of a phenomenon which, although it has hitherto re- ceived but scanty attention, is really very remarkable : I allude to the frequent appearance of the labial v in H.G. in the place and now with the power of/, — its appearance, moreover, always, I believe, where the ideal S^ should exhibit b (cor- responding to /of S ), and never where 2^ exhibits its native spirant in correspondence to the p of X . These v^s, I would suggest, originally represented older b's which, when H.G. became a written language, and indeed at a much later period (for, as far as the written language goes, they are really more frequent in M.H.G. than in O.H.G.), had only advanced (at least apud plebem) halfway towards the / of the assimilating dialect ; they had become unstopped, but not yet hard or voiceless. The symbol, thus fixed, remained unaltered ; but the sound, as often happens, left the sign behind, and proceeded to complete assimilation with the sound of /. If now the commingled Low Germans had held firm to their k and p, the labial and guttural families of the fused dialects might have been completely assimilated to the L.G. type : but in these points the latter dialect went over to the H.G. The L.G. dental spirant, however, seems then to have ofiered (as it still ^ "Le Principe de Transition consiste en ce que la permutation ne marche que pas a pas et ne fait qn^un pas a lafois" (Baudry : G.C., p. 82.) It is therefore identical with the famous Linnean doctrine " Natura nonfa4iit saltus,'" so widely recognised in the natural sciences. k2 132 grimm's law : a study. [§ 54 {g), offers) considerable difficulty to those unpractised in it^ Here, accordingly, the H.G., holding fast to its media, preserved the relative phonetic position of the whole dental family ; and the L.G. had again to adjust itself to the H.G.^ (g) Upon the whole, the preponderance of the H.G. is, in the oldest remains of the dialect, very decidedly marked. One may easily imagine, however, that, long after the great migra- tory movements before referred to, L.G. would maintain its prestige as the old polite and literary dialect, and that, be- fore the final fusion of any two related mutes (or, rather, in the present case, absorption of one by the other), duplicate forms exhibiting both would remain in use side by side. Such commixture is abundant in M.H.G.-^, and indeed remains to the present time. It may not now^ appear in literature ; for a primary requirement of written language is orthographic uni- formity. But if we turn to the spoken language, we find that extensive areas in North Germany are still occupied by various Piatt -Deutsch dialects; while the purest H.G. is almost confined to the south and south-east. And, again, where the written language exhibits assimilation to Low German forms, the pronunciation often refuses to follow. Thus, we write raub, sand, sammlung, &c.; but in pronouncing such words we -finish, in each case, with a distinct tenuis'* ; that is to say, aflthough these final letters have assumed L.G. forms to the eye, the pronunciation remains pretty nearly the same as if the ideally correct H.G. consonants [p, t, k,) were actually written^. (h) Thus does the evidence which is adduced to prove in- ^ See Appendix B, last note. '^ It is worth noting that nearly all the supposed " unverschoben " forms .quoted by Grimm (Gesch., p. 485) are dentals, indicating, as we might expect, a somewhat stubborn resistance on the part of L.G. to complete absorption into H.G* 3 See Schleicher, "Die D. Spr.," p. 201. * See Note 1 to § 35. ' Max Miiller employs this discrepancy (Lectures, ii. 226) as an illustra- tion of the way in which he supposes the Goths to have " fixed the second series, the ff, d, 6's, in their national utterance, as k, t, p" But for every- thing there is a. reason. § 54 (i).] grimm's law : a study. 133 complete dissimilation of S^ from % (whicli are assumed to have been previously identical) join with various other con- siderations in proving rather the incomplete assimilation of one of these systems to the other^ — the two having previously been totally distinct. And now, in concluding this part of my subject, I may be permitted to repeat that Reflex Dissimila- tion is a very different process from a simple phonetic Verschie- hnng. It does not set in, run its course, produce given efl'ects within a given (and comparatively limited) time, and then cease. As has already been often stated, when once the Dis- similating Sentiment has been, by whatever means, evoked, it may remain vigorous for an indefinite period of time. The length of that period is determined by a set of circumstances, political and social, totally independent of language. But as long as two strata of people (to take the simplest case) mutually maintain that relative position, or degree of contact or com- mixture, which renders correlative phonetic variation possible, so long will this incessant translation and retranslation of cor- related sounds, when once established, continue. What hap- pens if that relationship is afterwards destroyed has likewise been pointed out, but may be again stated : — If one stratum be separated from the other (as was the CI. from the N. European), each will take with it its own phonetic system, and the Dissimilating Sentiment will gradually or rapidly die out, according as the separation is gradual or instantaneous, and as intercourse is or is not thereafter for a time kept up. But if one stratum be absorbed or welded into the other, either its system will completely disappear in the system of the other, if that other is socially or numerically far superior; or, if there be some approach to social and numerical equality, the result will be a phonetic compromise, such as is exhibited by the H.G. of to-day. {%) Even while such absorption is going on, dissimilation may continue active, although in fewer and fewer cases, up to the very moment of complete amalgamation : and as long as it is active, it seizes and alters any suitable word presented to it by the commingled dialect, without inquiring whence that word may have come. Thus, formerly, members of the London, 134 grimm's law; a study. [§ 55 («). ^'residuum^^j on acquiring, from any source, a new word com- mencing with V, would at once, as a matter of course, trans- form that V into w. For the Dissimilating Sentiment operates with reference only to the standard dialect in its presence, and takes no note (how should it ?) whether the words which call it into play are native to that dialect or imported from abroad. And some of the examples quoted by Grimm seem to illustrate a precisely similar mode of action on the part of the Old High Germans. Thus the words curt -us and moneta were borrowed from the Latin at an early date^ But the mass of the High Germans no longer knew anything of the cognate dialects of Italy. The reaction of H.G. against the CI. dialects, which would at one time have turned curt into gurd — and moneta into monida, had ceased for many a century. Not so, how- ever, the reaction of the popular (H.G.) dialect against the polite and literary; which, if not still L.G., inherited, in the popular ear, its prestige. This process, though dying, was not yet dead ; and on the appearance of foreign words in the latter dialect the former seized and treated them as if they were really L.G. forms, producing, in the above exam- ples, kurz and muniza. In this case, indeed, the popular dialect, over a certain area, ultimately absorbed the standard one ; so that these and such- like forms came to the surface; nevertheless hurt survived in hterature to a comparatively late stage of the language. 55. — (a) And here I shall desist from further treatment of the main body of my subject. Of the imperfections of the fore- going sketch no one can be more sensible than myself. But I am consoled by a confidence that the principles laid down in it, whatever may become of my manner of applying them, will furnish a safe and satisfactory foundation for a correct theory of Grimm's Law. Of one thing, at any rate, I am quite sure : — that problems of this kind are not to be solved by shutting one's self up in one's study, and dreaming of '' Heldenzeiten ** and the like fallacious explanations of one aspect of the puzzle, ^ Forstemann, however (Gesch., i. 370), following Lottner (Z-S., xi. 186), connects kurz directly with the A-S. scort (our short). §55(6).] GRIMm's LAW. A STUDY. 135 which only make the other aspects more perplexing than before. For we have here to deal not so much with material as with movement, not so much with fossils as with life ; we have not to reconstruct an earlier language, but to account for a given set of changes. Our business, therefore, is to observe, as well as to theorize — to observe, not only what is petrified in written remains, but what is going on in the living spoken languages and dialects about us ; and although we may never discover another example of the remarkable triplicate relation- ship which characterizes our S^r^ 2y, and S^., and which pro- bably arose from a combination of circumstances unlikely to recur, I have no doubt that many examples of the simpler ease of dual relationship will be detected ; some of which may tend, still more directly than those I have collected, towards elu- cidating Grimm^s Law. Or, conversely, other linguistic pro- blems may be met with, towards the solution of which some or other of the principles laid down in this book may be made to contribute. I shall conclude my little treatise by noticing a problem which seems to me to offer a case in point. {b) Among the minor philological questions recently dis- cussed on the Continent is one relating to the nature of the primitive I-E. ^-sound. The leading facts involved in the question are these : — The Aryan languages employ not only a hard or pure k, but also, in many words, a k which, by a known affection, has degenerated almost into a sibilant {q) . Let us call this the Aryan or Sibilant affection. Now, in the corresponding words of most of the European languages, the representative of this gutturo- sibilant is invariably a pure k. On the other hand, some European languages, especially Greek and Latin, exhibit another affection of k, which, in its earliest known stage, resembles a combination of k with the labial semivowel w. Let us call this the European or Labial affec- tion. In Latin it appears in the form qu — the guttural still maintaining its preponderance ; but in Greek (as in some old Italian and Keltic dialects) the labial semivowel seems to have gradually acquired the supremacy — first becoming conso- nantized and then casting off the guttural altogether ; so that (omitting dialectic forms, and a few cases of dentalization) we 136 grimm's law : a study. [§ 55 (c). have to suppose the series — kw {=/cF) , kb, kp,p [ir) . But here again_, just as to the Aryan or Sibilant affection most of the European dialects answer with a pure k, so, in turn, to this European or Labial affection the old Aryan dialects answer with a pure k^. [c) This singular behaviour on the part of h might still per- haps have remained uninvestigated but for another remarkable fact. The Lithuanian and Slavonic, which are European dia- lects (once, and probably long, in contact with the old German dialects), exhibit in this matter of the A:'s, a nearly complete agreement with the Aryan : that is, to the European or Labial affection they answer with a pure k; and where the other Euro- pean dialects maintain pure k, they answer, in precisely the same cases as the Aryan (so far as the vocabularies agree), with a sibilant ; thus : — Skt. O.Sl. Li. j Lat. Grk. ^ta =suto =szimtas I ^centicm =6KaT6v: gvan = su-ka = s^u {szuns) =■ canis = kvoov : nag-ati =:nes-ti =nesZ'ti i =nanciscor =ev€yKa: prac- =prositi =pirsz-ti =precor = irpaK-^co : which may suffice as examples of the Sibilant affection an- swering to the European pure k. On the other hand : — Grk. Lat. Skt. iriTTTco, TToiravov = coquo,popina =pa¥-,pak/ati = O.Sl.joe^a: ire/jLTre (Trevre) =quinque =-panMan =Li. penki : eirofiac —sequor ==sak!-,salzate= ,, seku: '^TrapT- =jecur =yakan = „ jekna : which may suffice as examples of the Labial affection answer- ing^ to the Aryan pure h It must be added, however, that a large number of Ar's appear both in Europe and in Asia as pure Fs. 56. — {a) Before making any remarks of my own upon these peculiarities, I shall glance as rapidly as possible at two or ^ For our present purpose the Aryan pure k must be assumed to include (as it once actually did) the Skt palatal k, generally represented, in our type, by k'^ but sometimes by c. 2 See the preceding note. § 56 (b).] Grimm's law : a study. 137 three of the leading attempts which have already been made both to elucidate them and to assign them their proper place in a general scheme of scientific Indo-European philology. It is a necessary preliminary to observe that their principal interest has hitherto been assumed to lie in their bearing upon what may be called the genealogy of the successively- diverging languages of the Indo-European family. With a view to precision of treatment, it has been found convenient by philologists to assign the languages, which thus successively branch ofp, to a series of linguistic " periods "_, during each of which the parent language of a subsequent group of lan- guages is assumed to have existed in the form o£ a single and homogeneous language ^ Such an assumption, if not pushed to the extreme of denying the existence of dialectic differences -within the successive parent tongues, may probably in some cases be found to correspond with real historical periods, du- ring which the tribes speaking those languages exhibited a general ethnic unity ^. {b) But with the later '^ periods '^ we have in this place not much to do : our subject carries us back to the earliest of all, ^ The following genealogical tree is based on Fick's in his I-G. W-B., p. 1051 :— The Holethnic Speech European Ar\'an North-European South-Eiu-opean Iranian Indian I I ■;_ - . German Lithu-Slavonic Keltic Grseco-Italian I have substituted the assumed languages for the corresponding peoples, as given by Fick, and have omitted the later subdivisions. Here and else- where I confine the term Aryan to the Asiatic branch of the Holethnos — an application of the term which is, I believe, universal on the Continent. Its extension to include the Europeans seems to me to be justified by no sufficient reasons. ^ But see on this point Max Miiller, Lectures, i. 205, 206 ; who, how- ever, if I understand him aright, runs to the opposite^ extreme of refusing to all such periods any historical reality whatever. 138 grimm's law: a study. [§ 57 (a). — to the Urvolk or Holethnos^ with its primitive speech, and to the original Separation of the people into two sections, one of which remained in Asia, while the other moved off towards, and ultimately reached, Europe. In virtue of this ethnic sepa- ration there must, in course of time, have ensued (supposing each of the sections o£ people to have maintained a general unity) a corresponding linguistic division into a clearly marked European, and a clearly marked Asiatic (Aryan), ^^ period^'. In point of fact the old European dialects on the one side, and the old Aryan dialects on the other, actually do exhibit grammatical and phonetic characteristics of the ir own. B utj first application^ of the phenomena exhibitedj^heprimitive k was to the denial of any such clear and decisive original Sepa- ration as the ^^period^^ theory asserts, and to the bridging over, as it were, of the broad and deep gulf which was supposed to di- vide the one group of dialects from the other. For, it was urged, ( "'"the Li- SI. really agrees in some important points (as, e. g., in j the splitting or radiation of a into «, e, and o, and in the evo- I lution of / from r) with the European division ; but in its treat- 1 ment of k it agrees just as completely with the Aryan division. I If, therefore, on the former ground the Li- SI. must be assigned to the one division, it must on the latter ground be just as decisively assigned to the other ; that is to say, it belongs to j both at once, and forms, as it were, a bond of union between ) them; so that we are no longer justified in imagining any I such broadly-marked separation between the two as the / *' period " theory requires ^. 57. — (a) To such views as these an elaborate reply has been offered by Dr. Eick^ ; and so far as they are necessarily ini- ^ By Dr. Joh. Schmidt : ^'Die Verwandtschaftsverhaltnisse der I-G. Sprachen ": Weimar, 1872. ^ Bopp was not hampered by the rigid " periodism " of the younger Continental Philologists. The relationship of the Li-Sl. to the Aryan had attracted his attention ; but he attributes it to the longer and later connexion between the two than between other P]uropean dialects and the Aryan. — See his V. G., p. 39 ; especially the note. ^ " Die ehemalige Spracheinheit der Indogermanen Europas " : Gcittin- gen, 1873. § 5^ wo GRIMM'S LAW : A STUDY. 139 mical to a period o£ general linguistic agreement of all the European tribes^ his reply must be considered altogether suc- cessful. But his treatment of the purely phonetic question is seriously affected by two antecedent considerations : — first, his determination to vindicate the '' period" theory in its most uncompromising form_, so that his phonetic hypothesis holds a place completely subordinate thereto ; and_, secondly, his as- sumption throughout that the guttural peculiarities in question, although they did not originate, were yet developed, subse- quently to the original Separation. His fundamental propo- sition is that the Holethnic speech comprised two distinct A;-sounds^ . Let us designate these^ by ^^ and k^. Then k^ is that k which is represented in Asia by k pure, and in Europe, except among the Lithu- Slaves, sometimes (and formerly, Pick thinks, always) by kw, &c. ; and k^ is that k which is represented in Asia by q and in Europe (except as before) always by k pure. {b) Consequently there requires explanation a complex series of phenomena, which may be tabulated thus : — ^' [k^ appears in Europe generally as k pure; Iky „ ,, sometimes SiS kw=qu=7r ; ^ A^ k^ „ „ always as k pure : ky appears in Aryan always as k pure^ ; k^ ,, „ f, as g. or reversely : — r European^ pure represents k^in all cases and k^ in most cases ; I J J kw ( = 5'w,&c.) represents k^ in the remaining cases. r Aryan k pure represents k^ in all cases ; \ „ <^ represents k^ in all cases. ^ Ascoli, in Italy, was, I believe, the first to enunciate this view. He was followed by M. Havet in France (" Revue critique," Nov. 23, 1872). Fick concedes priority to both, although he appears to have conceived the idea independently. ^ I borrow the convenient notation of M. Havet ; that of Fick would require special type for one of the A;'s ; while his use of a bare k for the other is sometimes misleading. ^ See note to § 55 (h). 140 grimm's law : a study. [§ 57 (c). This complexity is^ as we have seen, further complicated by the circumstance that one old European dialect (the Li- SI.) sides with the Aryan. But two important facts which immediate- ly require notice are: — (1) that the characteristic affection of \ has disappeared in the other dialects of Europe ; and (2) that a great majority of the ^^s which, in these dialects, correspond to ^^ and should therefore on Fick^s hypothesis exhibit the labial affection, actually exhibit no affection at all, but are, in fact, like the ^'s representing h^^ 'pure k's. (c) To account for this, Fick invokes a summary process which he calls Verwischung or *^^ Obliteration '^, in virtue whereof the Aryans in Asia and the Lithu- Slaves in Europe are supposed to have independently cleared away the Labial affection ^ ; while the other Europeans are supposed to have similarly cleared away the Sibilant affection ; so that k^ in the former case, and \ in the latter, became pure h. These other Europeans, moreover (as was remarked in § 55 ad fin. and [b) above), must, on this hypothesis, have treated the majority of the labialized ^'s in precisely the same way. But here, again, Fick is ready for us with a general Verschmelzung or '^ Fusion ^^, whereby most of the descendants oi\ and all the descendants of h^ have in Europe become verschmolzen into a single sound, viz. the pure k, 58. — (a) Of all such summary modes of cutting away diffi- culties there are who harbour a not unnatural suspicion; for nothing is easier than to imagine that we understand or have accounted for a phenomenon or a process when we have merely learnt or invented a specious term which only conceals the fact that we know little or nothing about it. I should like, therefore, humbly to inquire what may be meant, for exam- ^ If, at least, we are to attribute that affection to Holethnic times. Fick's opinion on this point, however, is obscure ; indeed he seems to leave it at last for the reader to determine : — " Hierbei ist vollig gleich- giiltig in welcher Periode man sich diese Afficirung voUzogen denkt" (p. 28). I shall assume, for simplicity, that Fick's k^, representing an incipient la- bial affection of k pure, was primitive ; otherwise it will be necessary to suppose that the bulk of the Europeans introduced that affection merely to get rid of it again ; which is absurd. § 58 (b).] GRiMM^s law: a study. 141 pie, by this Verwischung. And our first busiiiess_, of course, when any such new or doubtful principle presents itself, is to see whether or not it harmonizes with old and well-established ones. Thus the weak points of the prevalent conjectural hy- potheses of Grrimm^s Law became apparent when these were tested by the indubitable Principle of Least Effort. And in the case before us we may again use this very same principle as a touchstone. For if (adopting the line of argument em- ployed in Appendix B in reference to the Aspirates) we con- sider the ultimate forms assumed by hx and h^, we can hardly arrive at any other conclusion than that both of these, in comparison with k pure, must originally have exhibited at least an incipient corruption or debilitation. For the ulti- mate form, or at least one ultimate form, of k^ is p, and the ultimate form of k^ is g : both p and g (the latter, of course, much more than the former) are weaker sounds than k pure; and consequently the primitive affection whose development gave rise to those weaker sounds can hardly have been any- thing else than an incipient debilitation. In the case of k^, whose characteristic affection was clearly the more virulent, and whose history is simpler than that of k^, we observe that the debilitation was, on Fick^s hypothesis, developed both by the Aryans in Asia, and (independently) by the Lithu-Slaves in Europe ; while all traces of it were ^^ obliterated '' among the bulk of the Europeans. (b) Here therefore Fick^s Verwischung means that all the Europeans, except the Lithu-Slaves, on no limited scale^ and for no apparent reason, raised a weaker sound to a stronger. As to \, which is represented to have made up the remainder of the series of kh, Fick leaves us on several points in considerable doubt. If we are to treat it as having been characterized by the labial affection from all time, then the Holethnos must have been totally destitute of a pure k, — which is hardly cre- dible (§§ 24, 25, ante) ; then, too, the Aryans in Asia and the Lithu-Slaves in Europe (in spite of the adverse influence of the other Europeans) must have independently changed this affected k to pure A;, just as these other Europeans are supposed to have changed k^. In addition to this we have the further 142 GRIMM^S LAW : A STUDY. [§ 58 (c) . remarkable fact that the other Europeans also represent \ in the great majority of instances by pure h : so that ifk^ was affected from all time, or at least in Holethnic times (as I think Fick intends us to suppose), then the bulk of the Eu- ropeans must have followed two incongruous modes of treat- ment of k^, by one of which it became k pure, while by the other it became qu and tt or p^. (c) The incongruities involved in these phonetic movements are such as one must hesitate to believe in. For k^ and k^ are supposed to have had each its own domain in the Holethnic vocabulary, and therefore in the vocabularies of all the consti- tuent dialects at the Separation. And though (§§ 33-36 ante and notes) all the dialects, when separated, need not have fol- lowed precisely the same lines of development, yet their several lines could hardly have run in opposite directions, or their final forms be inconsistent with one another, especially among those dialects which in other respects evince the closest reci- procal affinity. Nevertheless, on Fick's hypothesis, we have to attribute such diversity and inconsistency twice over to groups of dialects thus nearly related; i. e., to the South-European as opposed to the Aryan, and the German as opposed to the Litliu- Slavonic, — the one group in each case invariably sending k^ up to 7c pure (or, in German, its eqiiivalent) , the other as invariably sending it down to a sibilant — the one in some cases maintain- ing kw (or its equivalent) sometimes lowering it top, sometimes raising it to k pure, the other invariably raising it to k pure. These results are not merely diverse ; they are mostly irre- concilable. Yet the antagonism between the more nearly allied dialects is perhaps even less wonderful than, on Fick's hypothesis, the independent identity of action on the part of the Aryan in Asia and the Lithu- Slavonic in Europe. * " The old qu " in Latin is represented as " in full retreat " before the the simple c=k (Spracheinheit, p. 13) ; and again, kwj both in Latin and Greek, as a " receding and continuously vanishing sound " (p. 21). But for a flatly contradictory view, see Corssen, ^' Aussprache ", &c., i. 70 : — ** Dass auch innerhalb des Lateinischen, und zwar in der alteren und in der klassischen Zeit der Sprache, qu sich aus c entwickelt hat, zeigen die Wortformen * oywoltod ' * hujus^Me ' (=ce) " — and others. Compare i. 356, tiofe. § 58 (e).] GRIMM^S LAW : A STUDY. 143 {d) With a view to obtain a definite conception of the ex- tent of the supposed phonetic movements, at least in Europe, I have selected what appear to be primary words^ involving Jc^ and \, to the number of about 230, from Fick^s second Ab- handlung^. Dividing these proportionally, I find that about 30 per cent, involve h^ and exhibit a sibilant either in Lithu- anian or in Slavonic, or both. These among the rest of the Europeans were .raised to pure A;'s. Of the remaining 70 per cent., which should involve h^, nearly alP exhibit pure h in Li-SL, but in the rest of Europe they have to be divided into two sets or series : in the one set, comprising only about one fifth of them (or about 14 per cent of all the h'^) the labial affection is actually traceable ; but in the other set, comprising no less than four fifths of them (or about 56 per cent, of all the h*^) such affection nowhere appears. Thus, by the hypo- thesis in debate, the bulk of the Europeans seem to have sent but a very small number of these ^-'s on to or towards the labial terminus ; while they sent the vast majority in a totally different direction, — viz., towards k pure. (e) The main objections to Fick's hypothesis therefore seem to me to be these : — that it would leave the Holethnic speech without a pure 1c ; that it would doubly offend against received methods of philological reasoning, — first, by refusing to the Holethnos a pure sound which is common to both of its great divisions, and, secondly, by assigning to it two affections of that sound, which we only know as special and character- istic, one of the European and the other of the Aryan divi- sion; that it would attribute to the most closely related dialects irreconcilably diverse modes of treating the same sounds, and, on the other hand, identical modes to dialects more remotely related — these identical modes being sup- posed to exhibit themselves quite independently of each ^ The proportional values which follow will remain very nearly the same if all the secondary words be likewise included. ^ '^ Das Vorkommen von ^, und ^2 i^^ Wortschatze der europaischen Spracheinheit." ^ The two or three exceptions scarcely affect the case. They are pro- bably due to contact with the other Europeans. 144 GRTMm's LAW : A STUDY. [§59(«). other and under circumstances of a widely differing cha- racter; that the assignment of two ^'s to the Holethnos would only account, in reality, for these identical modes of treatment ; and that the suggestions intended to account for the diverse modes of treatment in the more nearly related dialects involve, firstly, incongruous developments of one and the same sound {k^) in one and the same set of dia- lects, and, secondly, in the case of both ^'s, a general disobe- dience to the Principle of Least Effort, in which disobedience, indeed, the Aryan and Li-Sl. dialects concur, so far as Jc^ is concerned. To meet the last- cited objection, it should be shown either that, if k pure is to be derived from ^^ and k^, or either of them, these sounds (one or both) are stronger than k pure, or else, on the other hand, if k pure is stronger than either or both of them, that they are (one or both) derived from k pure. 59. — {a) This alternative has been boldly met by M. Havet, who has recently returned to a consideration of this question . To k^ and k^ he gives distinct and known phonetic values. He considers k^ to have been originally = oar kw or the La- tin qu — exactly, always, and everywhere — and to have been stronger than k pure ; which, indeed, he represents as deri- vable from the former by natural debilitation. K^, again, was originally nothing else than k pure itself; and from it the si- bilated guttural is represented as having descended. Such, it is said, were the values of k^ and k^ in the Holethnic speech ; and although the various descendent dialects may have modi- fied those values, they have all, in one way or another, pre- served the distinction between the two ^'s. (b) By the provision that k^ is essentially stronger than k pure is meant that kw (qu) may (and does) descend to k pure by natural debilitation, and that, " in spite of current opinions, the contrary change cannot take place.^' Hence (to omit finer degrees of debilitation) kw, k pure, k' (or c), '■ " L'unit6 europ^enne, et la question des deux k Ario-europ^ens " : in the Mimoires de la Society de Linguistiqu£, ii. 4, 1874 ; an article brimful of condensed facts and reasonings. §60(«).] GRIMM's LAW : A STUDY. 145 and g are successive stages of phonetic descent. At the Separation each of the two great sections of the Holeth- nos took with it the old 'kw and k. Among the Aryans the original h^ {kw) descended to k pure, and the original k^ {k pure) to k' (c) and ultimately to g. The bulk of the Eu- ropeans, on the contrary, maintained k^ {k pure) as it was, and brought down k^ (kw) in most cases to the same level ; but the Lithu- Slaves, by a merely '^fortuitous coincidence^^, reduced k^ in all cases to k pure, and k^ ( = k pure) in all cases to a sibilant, in precisely the same way, and to precisely the same extent, as the Aryans, so far as the two vocabu- laries correspond. To this " coincidence ", and to some other minor features of the hypothesis, objection might fairly be taken. But I pass on to the corner-stone of the whole, viz., the relative phonetic value attributed to the two ^^s, or, at least, the value given to k^. 60. — («) The merit of this hypothesis consists in the respect it appears to pay to the Principle of Least Effort. Whether or not that respect is real as well as apparent depends on the correctness of Havet's determination of the relative value of hw and k pure. On this point I confess I do not feel that cer- tainty which M. Havet's precision of language is intended to inspire. We have already seen (§ 58 («)) that kw [qu) in several dialects ultimately degenerates into p. This pho- netic characteristic (implying that the parent sound had set out on a given line of debilitation) is in accordance with the physiological characteristic of kw as indicated by its vocal formation. For in pronouncing this combination the point of contact of the tongue with the palate is much forwarder in the vocal passage than the point of contact at which the purest k {k before a=ah, for example) must be pro- duced ; and this is the physical characteristic of a weaker sound (§ 6) . Authority on this point may not count for much, or else that of Bopp^, Curtius'^, and Corssen' might be quoted as in direct opposition to the views of Havet : in fact, their ^ Y. G., § 84, 1. ^ Grk Et., p. 451. ^ u Aussprache," &c., i. 70. L 146 GRIMM's LAW : A STUDY. [§60(6). opinions are, no doubt, the ^^ current opinions " which Havet expressly controverts. Nor, perhaps, is the point in. dispute to be decided by the analogy supplied by the combination [ky] of the palatal semivowel [y) with k, which is universally admitted to be an initial debilitation, and the physical formation of which differs from that of kw almost entirely in the absence of protrusion of the lips. [b) But neither, on the other hand, does the evidence ad- duced by M. Havet in favour of his own view seem to me to be decisive ; for he bases his doctrine respecting the relation- ship of k^ [kw) to k pure upon examples taken from the Ro- mance languages as compared with the Latin. Now, in ac- cordance with the principle laid down in § 28 (d) ante, § 35 note 1, et alibi, and worked out in relation to the Aspirates in § 33, this circumstance may very seriously weaken, if not en- tirely destroy, the foundations of his doctrine. For in the Romance languages we meet with Latin sounds, not after they have pursued a uniform line of natural development or decline among one and the same people, but after they have been sub- jected to the mispronunciation of foreign and semi-barbarous races, to whom some of them would probably be strange and difficult and who would therefore modify them to suit the pre- vious education or want of education of their own vocal organs. (c) In France, especially, whence Havet draws most of his examples, it is perfectly well known that there was con- siderable difficulty with one element in the combination kw {qu), viz. the labial semivowel. The imcombiried semivowel in Latin, indeed, had probably passed into the spirant v before the extensive spread of that language over western Europe. But the old semivowel [w) was at a later period introduced over the same area, as well as into Italy, by invading Ger- mans (§53 [d)). The natives, however, seem then to have been unable to pronounce it alone ; and their attempts to do so led to the combination therewith of the guttural^ *. And the physiological peculiarity which led, at the outset, to this ^ For some examples, see Max Miiller, Lectm-es, ii. 295-297. The process is ust the reverse of what took place in the oldest Latin ; where the semi- vowel overpowered and at last annihilated the conjoined mediae and as- § 60 {d).] grimm's law : a study. 147 accretioii_, also led subsequently to the complete preponderance of the new guttural and to the disappearance of the old semi- vowel, — often to the eye^ and nearly always to the ear^ But no one_, I presume^ would argue from this irregularity^ or pho- netic contortion,, that w represents a naturally stronger sound than g, and that the sound of g is the legitimate descendant of the other. Now^ turning to the closely related combination kw {qu)y we observe^ firstly, that it already existed, and was im- posed on the Gauls and others with the language in which it was indigenous, or at least naturalized (See next section) ; and, secondly, that the guttural thus provided for them in advance was more powerful than they with which they provided them- selves. We may fairly argue therefore, a fortiori, that, if a guttural could be inserted where none before existed, and, when inserted (although thus inorganic, and not the strongest sound of its family), could overpower the conjoined semivowel, much rather might it be expected that k, the strongest of all mutes, already in combination with w, would, among the same people, decisively overpower and annihilate the same semivowel. [d) It seems to me therefore altogether unsafe to base upon the history of kw fqu) in any of the Romance languages the doctrine that that combination is naturally and essentially stronger than k pure, and that the latter is always and every- where the legitimate descendant of the former. The case is rather one of those in which the natural operation of the Principle of Least Effort has been forcibly set aside by the physical peculiarities of a people on whom a foreign language has been imposed from without. If it is not an exception that proves our great rule, it is, at any rate, one that does not in- validate it. Admirable therefore as is M. Havefs article in piratae ; as, e. g., in vivere {=wiwere) for gwi-gio-eie ; sudvis for suadwis levis for leghwis, &c. In spite of the caution suggested by these and such like cases, Benfey, in his Shorter Grammar, is inclined to lay down the general rule that, where the combinations kio, gw, &c., occur in the older I-E. dialects, tbe semivowel alone was the primitive element, and the con joined mute a later accretion. 1 For many familiar examples, see Brachet's " Dictionnaire Historique. . s. V. ^'■Gacherr l2 148 GRIMM^S LAW ! A STUDY. [§ 60 (c). many points^ it is, I think, infirm in just that one on which, above all, the value of most of his other opinions depends ; in a word, he attributes to natural development a change due rather to what may be called the violent action of external forces. It is as if we should argue that the old aspirate th { = 6, ]}) is stronger than, and the natural parent of, the tenuis t, on the ground that the former is, in the Romance languages, represented by the latter — or even that our soft spirants v and th (="8, dh) hold a similar relation to b and d, because the negroes of America say bery, brudder, dis,hc.,ioY very, brother, this, &c. M. Havet^s illustrations are indeed of great value, but in support of a conclusion of a very dififerent kind ; for if the combination kw made its appearance among any of the oldeF Indo-Europeans, and if it was afterwards in any cases replaced by k pure, this similarity of treatment implies simi- larity of history ; and one might infer that the combination was no more native to those people than to the Gallic races of our own era. (e) On the whole, then, M. Havet's view of the relative strength of kw and k pure seems to me to be inconsistent both with the comparative physiology of the two sounds, with the analogy supplied by the relationship between ky and k, and with the tendency of kw to become p ; while the exam- ples by whose aid he endeavours to prove the'natural deri- vation of the pure from the affected k seem to me to prove something altogether diverse. Apart from these points, his hypothesis difi'ers very little from that of Fick, and has little or no advantage over it ; so that what has already been urged against the one (§ 58 {e)) applies pretty nearly as much against the other. ' M. Havet does, indeed, appear to endow the Holethnos with a pure k in his k^ ; but as he represents the k pure of later times to be the first stage of descent from kw (except where the latter branches ofi" towards p) , and as we can hardly suppose both a stronger sound and a weaker and immediately derivable sound to have coexisted from the birth of language, it appears to be an almost necessary exten- sion of his hypothesis that the same mode of phonetic gene- ration characterized still earlier times ; so that, going farther § 61 (^).J GRiMM^s law: a study. 149 and further back_, we shall at last reach a period when kw alone, and not k pure, existed^ 61. — {a) It will have been seen from the preceding sections (55-60, and the notes) that scarcely any philological problem can have given occasion to more diverse and hostile opinions than that involving the genealogy and relationship of the various I-E. A;-sounds. If to those opinions I here venture to add one more, it is because I have grounds for thinking that some of the principles enunciated in the former part of this treatise will enable me to attack the problem upon a totally different side from any towards which previous inquirers have advanced. I must remark at the outset that the ethnic question, involving the relationship of the Lithu- Slaves to the Aryans on the one side and to the rest of the Europeans on the other, is distinct from, and as regards our present inquiry subordi- nate to, the linguistic question, and not this to that. If, indeed, we had trustworthy records of the primitive history of that section of the Europeans, we should no doubt be able to account for all the linguistic phenomena which now appear so perplexing. As things stand, however, the only means we have of obtaining, conversely, a glimpse or two of the history of the people is furnished by language ; and hence I shall for a time leave the ethnic question altogether out of sight. {b) Referring now to §§ 18-22, 28 {d), and especially to §§38-43, we can hardly fail, I think, to be struck by the great similarity between the phenomena presented by the distribu- tion of the various I-E. ^-sounds and those presented by the cases there discussed and formulated ; for the most striking feature of this distribution is the diagonal or cross relationship between the pure and impure ^^s of one set of I-E. dialects and those of the other set ; by which, of course, I mean that ^ Grassmann, inKiihn's Z-S., vol. ix. (" Ueber die Verbindung der stum- men Konsonanten mit folgendemF") bad already laid down the doctrine tbat, wherever such a combination as kw can be traced in any of the I-E. dialects, it was in every case the primitive and pai-ent sound, whatever variations might represent it in the related dialects. — The older conflict- ing opinions on this perplexing Verbindung are admirably summarized by Baudry, G. 0., pp. 114, 115. ]50 GRiMM^s law: a study. [§ 61 (c). to the European pure h on the one side answers the Aryan and Li-Sl. impure k on the other ; while to the European impure A" (so far as it prevails) answers the Aryan and Li-Sl. pure k. To explain this has clearly been the ohief difficulty of previous hypotheses. Nevertheless the general phenomenon appears to me to be referable to a class which will, by this time, be quite familiar to the attentive reader; and it may be analysed into a number of particular and constituent phe- nomena, which will, I hope, be equally familiar. In the maintenance o£ the pure k, for example^ in opposition to the impure ones, we recognise phonetic Resistance on both sides ; while the respective debilitations or corruptions seem almost as clearly to exhibit phonetic Retreat (§ 42) . But these correlative processes necessarily imply that, while they were in action, the dialects exhibiting them were in presence of each other (§ 41 [a)) i hence the movements among the Fs must be thrown far back into Holethnic times, when the (subsequent) European and the (subsequent) Aryan groups of dialects stood to each other in the relation of single dialects in contact. (c) Again, of the three sounds in debate (viz. k pure and the two affected Fs) , the first alone is common to, or has the same phonetic value in, both sets of dialects : hence, on the hypothesis of this book, we must suppose either that, before the setting-in of Dissimilation (that is, the Mutual Resistance and Retreat just mentioned), the Holethnos possessed two impure Fs, but no pure ^, and that the evolution of a pure k out of one impure k by one section of the Holethnos led to the evolution of an exactly equivalent pure k out of the other impure k by the other section : all which, when followed out into detail, appears to me unlikely and even absurd; — or else that the entire Holethnos once possessed but a single and uniform k, which was no other than the purest and strongest k ; and that a dialectic fissure, as it were, was originated by the debilitation, on the part of one section of the people, of a number of these pure k'9, by an incipient sibilant or labial affection, as the case might be ; and that the other section, while preserving those ^"s pure. § 61 {d).] GRIMM^S LAW .' A STUDY, 151 nevertheless proceeded ultimately to respond to that debili- tation by an incipient labial or sibilant affection of some of the pure ^^s which were not attacked in the first-mentioned section : all which constitutes a case precisely similar to those abeady treated of in former parts of this book. And^ once more^ we have to observe that the two impure k's differ pho- netically from each other. This implies that the pure and im- pure k of one set of dialects are not the result of interchange^ by Cross Compensation, between the impure and pure k of the other set (§ 22 (c))_, but that, if those k's are to be correlated at all, it must be by way of Reflex and Mutual Dissimilation (§§28 {d), 33, 40 {b), &c.) ; for this process does not neces- sarily require that the mutually counterbalancing impure sounds derived from the stronger parent sound should be exact phonetic equivalents of each other. I propose therefore to maintain that k pure must have been (as by its nature it ought to have been) the original single parent sound from which the impure k's were derived — one by ordinary sound-weakening, and the other by Reflex Dissimilation. (d) But a question that may here fairly arise is, whether the Aryan and the European debilitations did not differ too widely to allow of their correlation. If, indeed, we conside only the ultimate forms {p and q) assumed by the one and the other, we shall of course answer at once that they did, But p and q are really the extremities, as it were, of divergent phonetic radii. Remounting from those extremities we find that these radii converge towards, if they do not actually meet at, a common phonetic centre, which is nothing else than k pure ; so that the phonetic distance between the nearest known points to k pure is vastly less than that between the extremities. In the case of the radius extending to q we may consider the centre as actually reached ; that is to say, we can, out of the facts of language, establish a direct line of descent from k pure to varieties of the sibilant. In truth, scarcely any line of phonetic debilitation is better known in our family of languages than this. And in every case the first step downwards is traceable to the accretion after k pure of the palatal semivowel y ( = German j) or to the combination with 15.2 grimm's law: a study. [§ 61 (d). k of the palatal vowel i (=Engl. ee) — the further steps being represented by k^ { = tsch = 'Eng\. ch in church, nearly), by ^=Germ. ch in ich nearly, and by s^ ( = Engl. sh) ; from which the passage to s is not uncommon ; and that to z, and even to complete disappearance, is perfectly possible. The initial movement down the other radius should, by analogy, be caused by the accretion after k pure of the labial semivowel w, or of some closely-allied " parasitic " sound (to borrow Curtius's picturesque term) which might readily pass into w. In this case the two phonetic radii might be represented to the eye as follows : — k pure qu=Tcw/ \Jcy p, it/ \^' {tschj ch) sh, s, &c.. This order of evolution for Jew is, however, as we have seen, reversed by M. Havet; and it must be admitted that the negative evidence arising out of the difficulty of establishing our assumed order from facts supplied by the younger lan- guages seems to be in his favour. But it is really quite as much in my own ; and my remarks in § 60 (d) are intended to hint a reason why, even though hv (qu) should sometimes appear, in the ancient, as well as in the Romance languages, to have been supplanted by Tc pure, the former may nevertheless represent a primitive corruption of the latter. For if we suppose the evolution of lew out of k pure to be due to the action of Reflex Dissimilation, we can hardly look upon the former as an indigenous, spontaneous, and unconstrained debilitation of the latter. It would, in that case, be due to the instigation of the Dissimilating Sentiment, whose vigour in inciting to phonetic movements which would not otherwise have taken place we have already had ample opportunities of observing (§§ 18-22, 38, 39, et alibi). It might therefore (§ 28 {d)) differ in value from the debilitation which evoked it^ pretty much as § 61 (e).] grimm's law : a. study. 153 the reproduction of a strange and difficult sound by foreigners differs from the native value of that sound ; and its relation- ship to Tc pure would differ from the relationship of the same original debilitation to k pure pretty nearly as a forced con- tortion differs from a spontaneous weakening. {e) All this virtually assumes (what I shall presently show to be supported by facts) that the Aryan,, Palatal, or Sibilant affection supplied the initial movement which incited to Reflex Dissimilation on the part of the commingled dialects. Still (to return to the point from which the preceding sub- section started) , if the earliest form of the European or Labial affection is necessarily to be taken as kw {qu), the difference between it and ky may appear to be somewhat too wide to allow us, with perfect satisfaction, to correlate the sounds in the way suggested. Considering, however, the fluctu- ating value of the closely-related labial vowel w, — in Greek, for example, on the one side, and French on the other, as compared with Latin, — we shall not be transgressing proba- bility if we conjecture that the Labial affection may originally have had a value intermediate to k and kw, such as may be represented by kit. This, of course, is a considerable approxi- mation to ky, but is justifled by the Principle of Transition (§54 (/), note), on the ground that the phonetic distance, as it were, from k pure to kw is too wide to be covered by a single even though forced leap. By the intermediation ofku, at any rate, the European affection is brought sufficiently near to the Aryan to allow us readily to believe that the former was intended to be a reproduction, by Keflex Dissimilation, of the latter ^ Reducing the whole movement, therefore, under the formulations of § 18 or of § 28 (c), we now obtain the correlative values — ^ The intermediate combination (kil) might perhaps almost equally well be assigned to the Aryan side, and kw be left as the original value of the European affection. But see the next note. 154 grimm's law ; a study. [§ 62 (a). which formulation correctly summarizes nearly all the phe- nomena of the distribution of the several A:-sounds ; while the European reproduction of the initial Aryan debilitation is supposed to differ slightly in phonetic value from the latter^ just as (a) and (a) in § 44 (/) differ from A^ 62 — {a) I have already forestalled the inquiry as to the dialectic place of origin of the whole movement. This has been assigned to the conjoined Aryan and Lithu- Slavonic dialect, but solely_, so far, on the ground that the sibilation of 7c has the aspect of a spontaneous and indigenous debili- tation for which We can account, while the character of the ^ I was originally inclined to trace the Aryan debilitation, as well as the European, to an earlier kii, intermediate to it and k pure. In this way the primitive values of both the impure Fs would be equalized, and to the oldest Aryan A;-system (k piire and kii) would diagonally correspond, sound for sound, the oldest European ^-system (kii and k pure). This intermediate value to k and ki/ might be defended by the frequent inflex- ional and derivational passage of u to i in our family of languages, which implies a passage through the intermediate value u. But as my theory does not really require such equalization of the primitive corruptions, I have thought it best, following what evidence we have in respect to ki/, to make this a direct debilitation of k pure ; and only to employ the inter- mediate kii on one side (i. e., between k and kw) instead of both. — I learn from a short note in M. Havet's article just now examined, that Ascoli has already suggested that the primitive I-E. k had (in all cases ?) a value which he represents by k^ (where y, I suppose, is = my w), and which gave off a k' in Asia and a k" in Europe. I may therefore perhaps be per- mitted to explain that these sections of mine on the Z:-sounds (except so much as relates to M. Havet's views) were originally written immediately after the appearance of Fick's book, and before any other part of my treatise, and, further, that I have been unfortunately prevented by a series of accidents from reading anything that the learned Italian has written on the subject. That the general idea therefore of refening the impure A;'s to a common and intermediate sound should have been independently con- ceived is reallj/ a "fortuitous coincidence." The value of the idea, how- ever, depends on the use made of it. As regards this point, I am still unaware of the way in which Ascoli proposes to explain that cross rela- tionship which, as I have above remarked, constitutes the main difficulty of the problem, — that is to say, why that k' which was labialized in Europe became pure k in Asia, while that k" which was sibilated in Asia became pure k in Europe. §62(6).] GRIMM's LAW : A STUDY. 155 European affection is just the opposite. But more may be said for the same view. For, following the line of argument adopted in § 32 with reference to the Aspirates, we observe, in the first place, that the action of this palatal debilitation is, among the Aryans, uniform and regular; and, in the second place, that it is virulent and extensive. In the Aryan Vocabulary of Fick^s I-Gr. Worterbuch, no less than some 40 per cent, of the primitive words originally involving h pure exhibit the semi-sibilant {q) ; while in Skt about 15 per cent, more exhibit the palatal {¥ or c). These Jcs or c^s, indeed, although they answer to the k of Fick or k^ of Havet, may really be considered to represent both a continued action of the Aryan impetus towards this form of debilitation, and also an arrestation of the debilitation before it had produced its full effect. {b) In contrast, on the other hand, with the virulence and wide extent of the native Aryan debilitation stands the mild- ness and narrow extent of the reflex (European) debilitation ; and in contrast with the uniformity of the former stands the diversity observable in the latter. No one dialect seems to have hit upon the correct value. In Greek, the parasitic element assumed, at an early date, the power of a consonant, at least so far as the assimilation of the conjoined consonant is concerned. Hence for the labialized k we generally find tt. In this treatment of the k some of the old Italian dialects resembled the Greek: so did the Gallo-British ; while the other leading Keltic dialect (the Old Irish) appears to have reduced all its k's. to a uniform value generally represented by c. The Latin differed from the other Italian dialects in preserving the rigid combination qu, with a decided guttu- ral preponderance. And, the normal change from kw [qu) to hw being made, the German dialects exhibit an original affinity with the Latin. These varieties of treatment I would attribute, as I attributed the varieties of Aspiration. (§ 33) to more or less (but not completely) successful efforts, on the part of some dialects, to acquire a sound indigenous only to a commingled dialect, but new and strange, and in effect foreign and difficult to the first-named dialects. And so. 156 GRIMM^S LAW : A STUDY. [§ 62 (c) . substituting for forcible imposition of a sound an attempted imitation instigated by the Dissimilating Sentiment, we may actually discover, in the early history of the European affec- tion, a close resemblance to the way in which the Latin qu made its appearance in the Romance languages (§ 60 (c)). It would therefore not be surprising if the later history of the older combination should also be found to exhibit, in some cases, a resemblance to the later history of its Latin descend- ant and representative. (c) The inciting debilitation having originated on the Aryan side, the reflex movement on the European side would natu- rally set in later and advance more slowly. This may help to explain why the labial affection is demonstrably traceable in so small a proportion of the European ^'s ; but this is not all the explanation. On the ground of the statistics quoted in § 58 (d), I should conjecture that the mutual adjustment of the Aj's between the two sets of dialects was cut short by some overruling movement, social or linguistic, pretty nearly at the stage at which we find it^ If it had been completed, we should perhaps have found, judging from the distribution of the Aspirates (§ 14 (c)), some 60 or 70 per cent, of sibilated Aj's in Asia corresponding to pretty nearly the same percen- tage of pure k'^ in Europe ; and 30 or 40 per cent, of labial- ized ^'s in Europe corresponding to pretty nearly the same per- centage of pure h'^ in Asia. Now there are two known move- ments to which such an arrestation of the dissimilating pro- cess may be attributed. One is the original Separation of the Holethnos into the European and Asiatic sections ; the other (on the theory of this treatise) is the rise and spread of the greater phonetic movement formulated by Grimm's Law. The former, by removing each of the two great dialects in which ^ I would not deny that some small number more of the European Ks, may have been partially labialized, — or, more properly speaking, that the pure k may in some words have coexisted for a time, and have been used indifferently, with a labialized k, and that, when the Dissimilating Sentiment was destroyed, the pure k may- have reasserted its rights, and have driven out the other. But I see no way of admitting that all the A's which were not sibilated in Asia were originally labialized in Europe and that from four fifths of them the labialism was afterwards expelled. §62 (C?).] GRIMM^S LAW : A STUDY. l57 action and reaction were going on beyond the influence of the other, would almost immediately destroy the Dissimilating Sentiment in the European division (see §§ 37 {b), 54 (a), et alibi). The latter would produce a similar efiiect by occu- pying more rapidly and vigorously the same ground as the minor movement (and much besides) ; and, so far as phonetic change was subservient to linguistic extension (§§ 28-30), by rendering the operations of the more limited movement un- necessary. It may be thought daring and rash to propose to settle the relative chronology of such almost primeval move- ments ; but as I have, throughout this treatise, maintained that the greater and more complete Verschiebung (so-called) must have altogether preceded the Separation; so I shall in the next section try to show some reason for thinking that the smaller and less complete interchange, affecting k only, must have preceded the greater. And (unless the cause is quite unknown) it is to the overwhelming influence of this greater movement that I would attribute the extinction of the smaller while yet incomplete. {d) The ethnic question, deferred from § 61 («), will now require but little remark. If we are justified in tracing the distribution of the A;-sounds to a mutual understanding, as it were, between (for our purpose) two dialects, we necessarily assume what has already been so often insisted on (viz., the contact of those dialects with each other) as the fundamental and indispensable condition of the symmetrical dissimilation between them. We have no option then but to maintain that, at the epoch when the changes of the Tc were in progress, the Li- SI. section of the Holethnos was in pretty close cohesion with the Aryan section, partook with it of a common debilita- tion, and acted and reacted with it upon the Europeans proper. This is perfectly consistent with the general consonantal sys- tem of the Li-Sl. dialect, by means of which I have endea- voured to adjust the relative geographical position of the tribe ; it only suggests further that the Aryan tribes must have occupied that part of the CI. district which extended to and overlapped the Li-Sl. district (see the figure, § 46 (^)). The difficulty therefore to be explained with respect to the 158 grimm's law : a study. [§ 62 (e). Lithu- Slaves is not^ as it seems to me, why, althougli they were subsequently a division of the North Europeans, their phonetic system should be what it is, but, conversely, why, their phonetic affinities with the Classical and especially the Aryan sections of the Holethnos being so close, they should subsequently have drifted away among the North Europeans. {e) In accordance with the views I have now endeavoured to propound, the phonetic movements among the ^^s, so far as they went, may be laid out in terms precisely similar to those employed in § 40 respecting another movement :- — (1) There is originally a single language (the Holethnic) employ- ing a single sound of a certain character (Jc) ; (2) this lan- guage divides, or tends to divide, into (for our present purpose) two dialects, an Asiatic and a European ; (3) in one of these (the Asiatic) a debilitation [hy) of that sound springs up and spreads ; (4) the other dialect (the European) at first resists that debilitation ; but (5) the two dialects continue in pre- sence of each other ; hence (6), by the habit of answering to ky by h pure a perception of incongruity and the Dissimilating Sentiment are at last awakened among the Europeans ; and (7), under the influence of the former, this people proceed to adjust (as they suppose) their sounds to those of the com- mingled dialect ; but, diverted by the latter, their efforts only result in a counterbalancing corruption of such of their own pure ^'s as correspond to the unaffected Asiatic ^^s, — ^the sound they actually produce, however, not being an exact reproduction of the Asiatic ky^ but differing from it in being a stage nearer to kw (say ku), from which it ultimately de- scended or advanced to kw [qu). (/) This whole movement, therefore, seems to me to be re- ducible to precisely the same general principles as those to which the phenomena of Grimm^s Law itself are referable. And our examination both of the one and the other proves, I think, that all the principal phonetic characteristics of the I-E. languages were impressed upon them in their common infancy and adolescence, and not in their independent and fixed maturity. Not only is this conclusion just what the abstract nature of things requires, but it is also one which allows the § 63 {a) .] grimm's law : a study. 159 amount of phonetic change in both cases to be greatly reduced and to be fairly distributed. Further it permits us to view the Holethnic speech under the same aspect as we view any other great language — i. e., as a conglomerate of dialects (§§ 35 (b), 44 (/ij) agreeing with and differing from one another in every possible variety of combination. If there- fore any peculiar phonetic movement set-in in one dialect, the behaviour of the rest in reference to that movement would be determined partly by the inherent genius of each, and partly by the existing relationship of each towards that one. The sides taken by some of the leading dialects in respect of these changes of k are indeed instructively different from those taken by them in respect of the greater (and later?) movement. Thus, in respect of the former, the Li-SL. sides completely with the Aryan, and the Grseco- Italian and the combined Germans are banded against those dialects ; while in respect of the latter, the Grseco-Italian coheres with the Aryan, and the Li-Sl. exhibits, in one point, an apparent assi- milation to L.G. This behaviour of the Grseco-Italian is worthy of observation ; for it clearly marks out the dialect as once occupying the front rank among the dialects of greatest resistance (§ 44 («)) . For when the Aryan admitted a head- long debilitation, the Grseco-Italian was found on the same (resisting) side as the Germans, and, like these, only admitted the much less virulent and much more limited results of Reflex Dissimilation ; but when the German dialects gave way to still more sweeping corruptions, the Graeco-Italian passed over to the resisting camp with the Aryan, and again only admitted (and to a less extent than the Aryan, — §14 (c)) the results of a like Dissimilation. 63. — {a) Let us now return to the question whether the affections of the k may not have belonged to a remoter epoch than the phonetic movement represented by Grimm^s Law. I have suggested that they did ; for the former movement, no less than the latter, probably represented an attempt to extend the powers of language by Phonetic Variation (§§28-30). But the latter movement was not only much more extensive 160 grimm's law: a study, [§ 63 (a), than the former; it was also more extensive than was required (§27 {b)). Hence^ if the latter be supposed to have come first in time, there will seem to be left no reason why the less extensive movement (supposing it had set in at all) should have been taken up by other dialects than that in which it originated ; whereas, supposing the smaller to have come first, it might, although thus taken up, still prove quite insufficient for the requirements of a rapidly growing language, and might be superseded and rendered unnecessary by the greater move- ment. But as this mode of reasoning is necessarily incon- clusive, I will venture to indicate a few facts which seem to me to furnish more cogent evidence to the same effect. The reader will remember that, by the theory of this book, the k's we have been treating of {i. e, the affected Fs of both species, and the pure k's to which I have endeavo],ired to restore them) although commonly spoken of as the " Primitive '' k's, are in reality only the " Classical '' k's, — the guttural Tenues of S^ alone. The Fs of 2 ,, however, and the Fs of S^ have been shown (§§ 24-26) to be equally primitive with the Fs of 2^; all the three sets of k contributing, in proportion to their number, towards the one unbroken series of primeval Tenues out of which the other mutes were afterwards evolved. Hence, if any phonetic movement affected the guttural Tenuis in the early era which preceded the greater so-called Verschie- bung, that movement cannot be supposed to have selected as its victims just those Fs (represented in L.G. by h), and just those only, which were subsequently to constitute the share of the CI. tribes. It should also have similarly affected (i) those ^'s (let us represent them by k) which subsequently formed the share of the L.G. tribes, and which were repre- sented by the CI. g and the O.H.G. ch-, and (ii) those Fs (let us represent them by fe) which subsequently formed the share of the H.G. tribes, and which were represented by a CI. x (gh, h) and a L.G. g. Further, these several sections of the original unbroken series of A:'s should have exhibited, in pro- portion to their extent, the same mutually dissimilating action as is exhibited by the (later) CI. section j that is to say, just as we found (§ 61 (e)) the § 63 (a) .] grimm's law : a study. 161 Aryan k . . , answered to by the f European kil (kw) so we ought to find (i) the Aryan k answered to by the European and likewise (ii) the Aryan fe "ki/ answered to by the European few (fe^^) . . fe ^ i^y\ so that, after the completion of the more extensive debilita- tions whereby k became the CI. ^ and H.G. ch^ and fe the CI. X [ff^f ^) and L.Gr. ^, we ought, in virtue of (i), to find traces of — (1) A section of labialized k'^ in L.G. and of labialized ^'s in Graeco-Italian, on the one hand, corresponding to a section of puree's in the Aryan and Li- SI. dialects on the other {i. e., L.Gr. kw {qu)=h?it. gw (i') = Grk yf (yS)= Aryan and Li- SI. g pure) ; (2) A section of pure ^^s in L.G. and of pure ^'s in Grseco- Italian, on the one hand, corresponding to a section of affected ^^s in the Aryan and Li-Sl. on the other (i. e., L.G. ^=Lat.^ = Grk y = Aryan gy (j) = Li-Sl. s) . And, in virtue of (ii), we ought to find traces of—* (3) A section of labialized Fs in O.H.G. (=labialized^'s in L.G.) and of labialized ^(^s (h's) in Grseco- Italian, on the one hand, corresponding to a section of pure Aspirates in Aryan and pure Mediae in Li-Sl. on the other (i. e., H.G. kw {qu) = L.G. gw = Lat. hw, gw {v) = Grk %f (<^) = Aryan gh = Li-Sl. ^); (4) A section of pure k's in O.H.G. (=pure ^'s in L.G.) and of pure %^s (A^s) in Grseco-Italian, on the one hand, cor- responding to a section of affected Aryan Aspirates and Li-Sl. Mediae on the other {i.e., H.G. ^=L.G. ^=Grseco-Italian x> h, ^= Aryan ghy {h ?) =: Li-Sl. z). Besides these two pairs of sections there should also be, in each case, both (i) and (ii)^ a considerable section oi k's not M 162 GRIMM^S LAW : A STUDY. [§ 63 {b) . jetj or not finally, reached by the Dissimilating Process (§ 62 (c)) ; so that both Aryans and Europeans would, as regards such sections, agree in exhibiting pure gutturals. But leaving aside these sections, from which reflex action was absent, and which therefore do not much concern us here, we have to observe, with reference to the affected and mutually dissimilated sections, that, although we ought to be able to trace all the relationships just detailed, yet, on the supposition of the priority of the movement among the ^^s to that represented by Grimm^s Law, we should be prepared for a considerable disturbance, by the later and much more thorough movement, of the original regularity of the correspondence in the CI. dialects between the pure and affected k's and fe's, such as still characterises the affected and unaffected CI. Fs, properly so called, which remained undis- turbed. Indeed a too great regularity in the former case would actually be prejudicial to the proposed chronology of the two phonetic movements ; for while it might be held to imply the total absence of any such later and disturbing movement, it would also be perfectly consistent with the hypothesis of a simultaneous affection of the CI. guttural Tenues, Medise, and Aspiratse, at a period subsequent to the evolution of the latter two sets of consonants from the other set. (b) As, then, if the affections of the k preceded the other and more extensive movement, we ought to meet with the phenomena just detailed, so, conversely, if we actually meet with those phenomena, we may, in default of any other or better explanation thereof, treat them as evidence that our chronological adjustment of the two movements is correct. But in proceeding to demonstrate the four cases of relation- ship specified in the preceding subsection, I must express my regret that I am unable either to faU back on any researches of other inquirers, or myself to devote any great amount of time and labour to the collecting of evidence. Nevertheless, considering the supplementary nature of these later sections of my treatise, I may perhaps claim to have done enough, if I shall have adduced a sufficient number of examples to place the existence of such relationships beyond dispute. Follow- § 63 (b) .] GRIMM^S LAW : A STUDY. 163 ing then tlie order already adopted^ we ought to find instances in which — (1) L.G. kw (qu) = Jjat. gw{v):s:GYk yf(/S)_, in opposition to an Aryan and Li-Sl. g pure. Accordingly : O.N. kwdn- (kon-), Goth, qen-, O.S. qudn-, A-S. cwen-, Engl, queen (and quean l)—^vv-r\ and Eoeot. ^av-a (for yfav-a); against Ski gnd^, Zend ghen-a^j (but also Sktjan-i, Zend Jen-i^y) Pruss. gann-a (but O.Sl. zen-a). O.N. kwal-j O.S. qual-j A-S. cwaU, O.H.Gr. qudl- and chwdl-, akin to O.H.G. quell-an, 'Ein^l, quell (and kill?^) (=/SaX- ?) ; against Skt gal-j or Lith. gel-u^, O.N. kwe-^-a, Goth. gi-)7-«w (g^)?)^ A-S. cwe-^-an, Engl. quo-thj be-quea-th (root ^2«?«-) ; against Skt ^a- '^'^ sing/^ ^«-c? '^'^ speak''"' (but Lith. zad-). O.N. kwam- {kom-^), Goth, gim- (qam-), O.S. kum-an, A-S. cum-an^, Engl, come^, O.H.G. quem-an, N. komm-en^ , — 112^1. ven-ire {for gwen- or gwem-ire) = Grk fiaiv-co [i. e, fiav-jco for jFa/jL-jo)) ; against Skt^ Zend^ and O. Pers. ^am-. O.N. kwerk' = O.H.G. qu'erc-a and querech-ela '^ gullet'^ (but Lat. gurgula and Grk ^ep^epo%^ are apparently irregu- lar) ; against the Skt gargara, " whirlpool^ gulf/' — no doubt a duplication of ^ar-, ^^ swallow/' which is correctly (on my ^ Fick, W-B., p. 67. I must here express my obligation to this boldly conceived work, witliout which the labour of collecting and col- lating even the few examples I cite would have been greatly increased. ^ Id., p. 257. See the first example (cennan, &c.) in subsection (d) infra. ^ Is the annihilation of the labial semivowel in kill due to Norman influence (§60 (c) ante) ? The form kill cannot be traced far back in time ] and, besides, we natives have rather an affection for the combi- nation qu, just as our Saxon ancestors had for its equivalent cw. * It is not quite clear to me that the Skt root gal-, " fall," (or as a causative " make fall ") and the Li-Sl. gaU, " to hiu*t, to pain," are con- nected. According to Fick they ought to be ; for (W-B., pp. 61, 518) he makes each equivalent to the German kwah. ^ The u and o represent a fusion of the semivowel with the conjoined vowel (See next section). ^ This word, perhaps, in Graeco-Italian, simulated onomatopoeia ; and hence the gutturals escaped affection, and appear identical with the Skt gutturals. M 2 164 grimm's law : a study. [§ 63 (c). hypothesis) answered to by Lat. vor-are (for gwor-are) and Grk ^op- {yFop')y the root of ^L-^pco-aKco, &c. O.N. kworn, Go. qairn-u, A-S. cweorn, Engl, quern, O.H.G. quirn and chwirn-a-, against Lith. girn-a (but O.Sl. zrun-y). The foregoing examples, together with that in the next subsection, leave no doubt, I think, as to the original cor- respondence, so far as they reach, of a European affected k to an Aryan pure k in that section of the /r's which after- wards became the CI. ^s. If the number of such examples is small, so also is the proportion of Tenues in the L.G. Mute-system (§ 14 (c)) compared with their proportion in the CI. system. (c) One very important and interesting example, properly belonging to the foregoing series, deserves a special examina- tion. This starts on the European side from the root kwiw- or hwih-, '^live^^, quiwa- or quiha- " living ^^; O.N. hwikvy Goth, qiwa-, O.S. quik, A-S. civic, Engl, quick, O.H.G. quec and check, = Lat. root viv- (for gwiw-) in viv-us {gwiw-us) and viv-ere {gwiw-ere) = Grk root ^lF- (for fyFiF-) in y8to-9 {i. e. ^iFo-si), &c. Here the labialism of the initial consonant is most decisively marked ; and we should expect to find the purity of the initial g just as decisively preserved on the other side. So it is in the Lith. {gyva-s) and Pruss. [giwa-s) j but the O.Sl. is Hvu, and all the Aryan derivatives are based on the forms jiv- and jlva-. It is manifest, therefore, that the last two dialects have gone astray from regularity ; but they have done so in a way which still comports with my hypothesis. For the more violent and extensive phonetic movement so often referred to, in evolving the Aryan and Li-Sl. Mediae from the primordial Tenues, would probably effect a complete arrestation of the older and but partially gratified Dissimilating process. It might well happen, there- fore, that the original distinction between the ^^s which once answered to the impure European k, and those which did not, should in sundry instances be forgotten, and that (at any rate after the Separation) the debilitation to which the latter were liable should in some odd cases attack the former also^ * That is to say : Immediately before the arrestation of the less by the § 63 (c).] GRIMM^S LAW : A STUDY. 165 As to the final consonant of the same root {(/wiw-, gvy-) , I am afraid this mnst remain, for the present,, a cruoe for etymo- logists. M. Havet's remarks on the root, though not new in substance, are neatly put, and appear at first sight to explain the whole matter. The stem jiv- (he observes) of the Skt present tense {jiv-ati=vivit) proves the existence of a primi- tive g before the i and of sl w after it ; while the Latin via^it {i. e. wig-sit) shows a w before the i and a g after it ; hence he constructs the form gwigw-, a supposed duplication of a still older gwi-, as the parent of the cluster of words above cited. Perhaps, in the absence of any suspicion of the existence and operation of such a process as Reflex Dissimilation, no other conclusion was open to him. But relying on this process, we may deny that the Labial aflPection is in any case Aryan, and that either the initial ^ or the final i? {w) of jiv- has come from gw-. And the same, of course, is true of the Li-Sl. dialects, whose evidence is, in this particular example, singularly copious and clear. The original initial of the root therefore appears to have been ^ (=k) pure; and this was preserved on the Lith. side. But as to the terminal sound, although the combination Ilw (gw) was by our hypothesis peculiarly and distinctively European, yet no trace of its guttural element greater movement, the complete series of Aryan ^'s would be divisible into : (1) palatized k's answering to European pure ^'s ; (2) pure A's answering to European labialized ^'s; (3) pure ^'s common to both dialects. Subsequently therefore to the greater movement, tbe resulting series of ^'s would be similarly divisible into : (i) palatized g^a answering to European pure ^'s ; (ii) pure ^'s answering to European labialized ^'s ; (iii) pure g^a common to both dialects. Now, although, as long as the European and Aryan dialects were in continuous presence, group (ii) might have been maintained pure by Phonetic Resistance (§ 42), it is not less likely that the extinction of the Dissimilating Sentiment, in relation to those ^'s, may have followed upon the stoppage of correlative debilitation. But how- ever this may have been, it would surely be absurd to suppose that the Aryans would for ever continue to cherish and transmit a distinction be- tween groups (ii) and (iii) after the removal of the European influence upon which the preservation of group (ii) as pure ^'s depended. If there- fore the indigenous Aryan tendency to palatization still operated in any degree— as it continued to do (§ 62 (a)) in the case of Ic, — it is difficult to see why group (ii) should be necessarily exempt fi-om its action. 166 grimm's law : a study. [§ 63 {d) . is to be found either in Gothic or in Greek. In the latter especially, considering its fondness for duplication, we have every reason to think that, if the assumed combination had ever been formed, it would have been preserved, and that we should consequently have found the word /3//3c>) ; in opposition to an unaffected Aryan gh and Li-Sl. g. Under this case, however, I know of only one example ; and in that, the traces of the given relationship are partially obliterated among the Germans. The example is represented by our word ^^ snow,^^ A-S. sndw ; where it would seem that the guttural of the combination gw has disappeared, as in Latin ^ So too in the cognate dialects : Goth, snaiw, O.H.G. snew or (in the verb) sniw-. But in Grseco-Italian the com- bination is well preserved : Lat. nix, niv-is and ningu-is (for snigw-is)j or (in the verb) ningu-ere and niv-ere [=^ snigw-ere) j where g stands for the primitive Aspirate ; while the Greek form of the root is vL' {i. e. crvvx^-), as in (t^v) vicp-a, v vl4>dei, &c. On the other side, the root in Zend is ^nigh-, in Lith. sneg-y O.Sl. sneg-j Pruss. snaig- : in all which the guttural is, as it should be, free from any affection. (/) Lastly we come to case (4), in which H.G. k = L.G. g r= Lat. h [g) = Grk ;)^, in opposition to an affected Aspirate in Aryan, and an affected Media in Li-Sl. Examples : O.H.G. car-ni, or gar-ni = O.N. gar-nir, " boweV^ akin to Grk %o\-a8€9, and Lat. hir-a (Plaut.) whence hilla [i.e. Mr-ula) and her-nia ; against Skt hir-dj and Lith. zar-na. O.H.G. gz-en, N.H.G. g dh-nen = hsit. ?ii-are, &c. = Grk ya-a-Kco, ')(ai-v-(o {i. e., %av-ya)), &c. ; against Skt M-, Zend zd-, Lith. ii-, O.Sl. zij-. O.H.G. zunJcd and zunga, O.N. tungajQotla.. tuggdn-, O.S. tunga, A-S. tungCy Engl, tongue = Lat. lingua (for dingua) ; against Skt jih-vd, Zend hiz-va, O.Pers. izdva, Lith. leztcv-is, O.Sl. jezy-Jcu. (But in the Li-Sl. dialects this word seems to have been confused with the following.) H.G. lech-eny Goth, bi-laig-on, A-S. licc-ian, Engl, lick = Lat. ling-o = Grk Xelx'O) ; against Skt rih- and lih-, Lith. Vi'ziuy O.Sl. li'zci. O.H.G. w'ek-an and w'eg-an, O.N. weg-a, Goth, ga-wig-an ^ With us, -ow (as in morrot^, borrotr, bow, talloM?, &c.) not seldom represents a final g pure : much rather might it be expected to form the terminus to which gto would tend. § 64 («).] grimm's law: a study. 169 = Lat. veh-ere = Grk o%- (for fo%-) in o'xp^ &c.j against Skt vah-, Lith. ?;ei-w^ O.Sl. vez-a. This list might be lengthened by examples which compare the Grseco-Italian forms on the one side with the Aryan on the other ; but as h is the terminus to which the Skt Aspirate ghj irrespective of any affection, naturally tends, it would manifestly be unsafe to institute such a comparison. I have consequently selected only those examples in which the Aryan forms are supported by all the known forms of the Li-Sl. dialects. 64. — [a) One other bearing of the question discussed in §§ 55 — 63 may deserve a few remarks ; and they shall be the last on the subject. I refer to the possible connexion of the affections of the I-E. h, in their incipient form, with the movement which gave rise to the vowels i and u. In the view which would trace both of these vowels to a as their parent (§ 6 ante) I feel bound to concur ; for besides other and more definite reasons, all the glimpses I have obtained of the early history of articulate sounds suggest the general conclusion that in all cases in which feebler Speech-sounds are naturally derivable from other and stronger related sounds, they were once actually so derived ^ But the phonetic distance, so to say, between a and either w or i seems too great to be covered by a single leap (§ 61 {d)) ; and there is reason to suspect that a must have passed to each of those '^ If the above conclusion may be laid down as an axiom, we shall per- ceive it to be an indispensable preliminary to any rational inquiry into the Origin of Language (or Languages) to eliminate all secondary or derived sounds from the primeval phonetic system or systems of which we are in search (§ 26, note). A minute examination and careful comparison of the linguistic force of the four or five remaining sounds, as they occur in a series of related languages, may then suggest the way in which these sounds originally became vivified with meaning. It has long seemed to me preposterous (I use the word in its strict etymological sense) to assume, as some do, when professing to pursue such an inquiry, the right of employing ad libitum the copious phonetic resources of mature languages, and of laying down, as the primordia of speech, sounds and combinations of sounds as to which it should first be asked whether they were possible at the birth of language. 170 grimm's law : a study. [§ 64 {a). feebler forms by some intermediate stage. The two stages are, in fact, those supplied by Fick ; who maintains, without re- serve, that i and u are contractions for ya and iva respectively ^ By this, of course, is not meant that a and its derived vowels appeared or existed in a detached and independent form, as we see them in modem alphabets, but that in certain of the oldest I-E. vocables (call them "words" or call them ^' roots ^^), which, like all the rest, were characterized by the vowel a, this vowel descended, through the phonetic stages ya and wa to i and u. Thus, putting k for any consonant, and supposing such vocables to have been ori- ginally of the form tea, or a/c, or even kok, then their inter- mediate palatized forms would be Kya or yaK or KyaK, leading at last kI or Ik or kIk; and their intermediate labialized forms would be Kwa, or wuk, or kwuk, leading at last to KU or uk or kuk. If now k be put for k, we obtain, in kya and kwa, combinations identical to the eye (§ 61 {d)) ^ W-B., p. 1043. " Da sich uns durchweg das Resultat ergeben, dass die Zeit der Wurzelschopfung- der Entsteliung von i und u vorausliegt, sind die . . . Wurzeln * und u in dieser Gestalt nicht zu dulden ; ihre wahre wurzelhafte Form ist ya und va [ Anglice i^«], woraus i und u bloss verkiirzt sind. Dies wu'd unumstosslick bewiesen durch den Umstand, dass in den alten und Zahlreichen Weiterbildungen von den fraglichen Wurzeln aus, nicbt i und ti, sondern ya und va als Radicaltheil erscheint j woraus zu schliessen, dass, als diese Determinationen vollzogen wurden, noch ya und va gesprochen wurde." In these views tbe student of Gothic, no less than the student of Sanskrit, will find it easy to concur. Anyhow, the vowels i and u, in their origin and history, cannot (so intimate is the relationship) be considered apart from the semivowels i/ and zv. And that the latter sounds should be eliminated from the " primeval phonetic sys- tem " (See the preceding note) seems to me clear, both for other reasons, and especially from their history in the Greek dialects. For the multi- plicity and diversity — as well as, in some cases, the violent nature — of the methods by which those dialects, when located by themselves, proceeded to get rid of both the semivowels, plainly suggest that these sounds could never have been indigenous to the (prospectively) Hellenic subsection of the Holethnos, and that they never took firm root therein : their fluctu- ating value and strange affinities mark them as foreigners. If, then, the sounds in question were once wanting in any part of the Holethnos, it follows, of course, that they must likewise, at a still earlier period, have been vv-anting in all the other parts ; and that both they and their related vowels, although primitive, were not primeval. § 64(c).] GRTMM^SLAW: A STUDY. 171 with those representing the phonetic combinations which gave off the Aryan and the European k respectively ; as, for example, in the Aryan c.ad- (from hyad-), " cec^-ere/' and the European hwa-, " who." May not therefore these affections of k fall, as particular cases, within the much broader stream of vowel-change, whereby i and u were evolved from a ? It seems to me that they may ; nor will the apparent difficulties in the way of such a connexion very obstinately resist re- moval. [b) The first of these difficulties arises out of the widely different destinies of the combinations kya and kwa, accord- ing as the palatalism and gutturalism represented by y and w attack the consonant or the vowel. In the one case we arrive at ga and qua- (or pa-), in the other at ki- and ku-. It is clear, therefore, that the original phonetic values of the combinations, notwithstanding their identity to the eye, were not precisely the same in the two cases ; and the dif- ference apparently depends on the side to which the semi- vowels incline. In the one case their inclination is towards the consonant, and may be represented hjky + a and kw + a; in the other, their inclination is towards the vowel, and may be represented by k + ya and k-\-wa. Hence, to permit of a common origin, we must suppose, in accordance with § 34 and § 35 notej that some slight difference of pronunciation originally " shunted " similar pairs of combinations of sounds on to two different lines of descent. But a second and more serious difficulty arises apparently out of my own hypothesis respecting the impure A;-sounds. For, by this hypothesis, the Palatal affection (ky) is Aryan only, and not European, and the Labial affection {Jew) is European only, and not Aryan ; whereas, of the corresponding vocalic affections, the palatal {ya, i) is as much European as Aryan, and the labial {wa, u) as much Aryan as European ; and they are so, in each case, no less after h than after any other consonant. (c) But if, in philology, we may claim a modest share in the privilege (so largely assumed by the biologist, the geolo- gist, and others) of drawing upon past time to an indefinite amount, we may perhaps find in the different dialectic condi- 173 grimm's law: a study. [§ 64 {d). tions of the Holetlinos at different periods an explanation whicli, if it does not necessarily connect the consonantal with the vocalic affection s^ wiU at least remove their apparent hostility. Treating first (this claim being admitted) the latter of the foregoing objections^, we have to admit that, as we have already (§61 (e)) assigned the Palatal affection, in the case of k, to the (prospective) Aryans, and the Labial affection to the (prospective) Europeans, we are bound by consistency to make a similar distribution of the corresponding vocalic affections also ; i. e., ya (afterwards i) must be supposed to have been originally Aryan, and wa (afterwards u) originally European. Hence each of these incipient dialects must be supposed to have adopted, or have attempted to adopt, the vowel-variation of the other, and that, too, in precisely the same instances, and not (as happened in respect o£ the con- sonantal affections) in different and diagonally related portions of their original and common vowel-series. In accordance therefore with § 28 (c, init.), we must further suppose that these two primary dialects of the Holethnos had not yet loosened out into that condition in which Mutual Dissimilation becomes possible — in other words, that the era o£ the vocalic affections must in its turn have preceded, wholly or in part, the era of the corresponding consonantal affections, jufet as the latter era preceded that of the changes represented by Grimm's Law. (d) This relative chronology of the two smaller movements (or, rather, as we may now perhaps say, two successive stages of the same movement) may obviate the former of the fore- going objections likewise, by suggesting why the earlier stage was vocalic only ; for it is highly probable that the consonants, and k above all, were originally uttered with a force (§§ 24, 25) which in later times was no longer employed, but which would then suffice to repel any debilitating accretion and to confine it to the conjoined and more impressible vowel. By the time, however, that vowel-change had proceeded to the length we find it did, the consonants would have become more open to attack ; pronunciation would perhaps have become § 64 (e).] grimm's law : a. study. 173 more agile ^> so that contact of the consonant with any accre- tion would be closer ; the dialects^ too, would have diverged more widely from each other ; and as palatism became more and more virulent on the Aryan side_, leaving no further room for a second and non-indigenous affection, the debilitating movement would at length pass into the consonantal stage already considered. (e) Nor are there wanting phonetic symptoms consistent with such mutual interchange of vowel -variation on the part of the two dialects. For what may be expected to happen in such a case is (§28 {d) and § 35 note) that the non-indigenous sound will not be exactly reproduced, or not produced with its native affinities, in the adopting dialect. And this cer- tainly seems to hold good in reference to the value and affini- ties of the labial affection on the Aryan side, as compared with its value and affinities on the European side. For where the Aryan w did not actually fuse with the following vowel into u, we still find it repelled so far from the kj and existing so independently of it, as actually to leave it free to fall a victim to the palatal affection likewise {e.g., Skt qvan-=Kvv-, ^^dog,^^ qvi-='' que-o,'' &c.). Besides, the two main Aryan dialects seem to have slightly differed, in respect of the value assigned to w, not only from the European but also from each other. Both perhaps gave it a too nearly consonantal value ; for in both, while it left k independent, it often likewise refused to amalgamate with the conjoined a. But this is more especially true of the Zend, in which the w actually passed into jo [qpan-, qpi-, he). And thus a clear distinction between the vocalic and the consonantal Labial affection is indicated ; for the former, although adopted on the Aryan side, is always repelled, as it were, from the k which precedes it, while the latter, on 1 This idea is Baudry's, to whose elegant (but, alas ! unfinished) sketch it is a pleasure to be indebted. With great ingenuity and probability he would trace the ancient metrical quantity by position to an earlier diffi- culty of pronunciation -, whereas, in modern languages, " la voix a, pour ainsi dire, fait son education, etl'articulation, devenue plus agile, n'eprouve plus aucune -peine a prononcer d'un seul coup deux ou plusieurs con- sonnes " (p. 13). See also his explanation of Vowel-weakening in Latin on p. 39. 17'4 grimm's law : a study. [§ 64 (/). the same side^ is entirely wanting^. On the European side, however, (where the sound was, so to say, more at home,) the treatment of the vocal affection is, as might be expected, often identical with that of the consonantal affection (e.^., theLat. '^cun-ire^' but " in-g-wm-are '' = Skt kun- ; '' que-o '' = qvi ; "equU'S" and LTTirofs^aqva} &o.) ; and so reversely (e. ^., the Lat. cum for quum, cutero- for quotero- {=7roT€po-) ; A-S. cum-aUj H.G. kom-men, Eng. come, for quim- or quam- ; Grk f^vv'7) for ^yFav-a ;' and many more) . Hence both Pick, who maintains a distinction between what he calls "radicaV^ {i. e., I suppose, our con- vocal) w, and Havet, who objects to Fick^s view on the subject, are both apparently half right. The former seems to have fixed his eye chiefly on the Aryan, the latter on the European treatment of w; as a vocalic affection. The former, while distinguishing between the phonetic value of the vocalic and that of the consonantal affection, appears completely to disconnect the two ; the latter, while maintaining their general sameness, appears to overlook their real historic and dialectic differences^. (/) In sum, then, the conjectures of the present section amount to this : — that while the Tenues were still the only mutes, and a the only vowel, a phonetic (palatal) accretion attacked the latter in one of two incipient dialects (the pro- spectively Aryan), and a second (labial) accretion attacked it in the other (the prospectively European), — these two accre- tions, or at least one of them, radiating perhaps from an accretion of intermediate value {u) intended to be a repro- duction of the other, and both being the earliest result of slovenly or drawling pronunciation ; that in the condition of close dialectic cohesion which still obtained, both of the enfeebled vowels became current throughout the Holethnos, first in their composite, and afterwards generally in their simpler forms {i and u), and were applied by the linguistic ^ There is a manifest consistency between these two facts. * It is but fair, however, to M. Havet to state that, in the article just now examined, he has noted " the origin of the groups kw and cw in the Aryan " as a question for future investigation. Whether he has since taken up the subject I am unaware. [§ 64 (/). GRIMM^S LAW : A STUDY. 175 sense to the extension of its powers of expression ; that after the one primitive vowel had been in this way extensively debilitated in two different directions, the same palatal affection (which seems to have been by much the more virulent) at length on the Aryan side attacked the guttural Tenuis; and that, the relationship of the two diverging dialects being by this time much less intimate, the conso- nantal debilitation evoked, first, Phonetic Resistance on the European side, and then Heflex Dissimilation, and all those attendant phenomena which have already been sufficiently discussed in 6& 55-62. UNJVEUSITY OJ^ CALIFORNIA / 176 APPENDIX A. APPENDICES. A. On the Affinity of R in English for the open A-sound, {a) At the Meeting of American Philologists^ held at Hartford, Connecticut, in the summer of 1874^, there was read a Paper, contributed by Professor Whitney (of which I have unfortunately seen only a very meagre sketch), giving the results of a statistical examination of all the sounds, vowel and consonantal, employed in the English language as at present spoken. The object of the Professor was to determine the relative frequency with which the several sounds recur ; and out of these sounds he seems to have selected for special treatment that of the old open vowel a (=Engl. ah). This sound formed the main phonetic characteristic, and once probably the only vowel ^, of the primitive I-E. language. Towards the epoch of the Separation, it still constituted nearly 30 per cent, of all the sounds^; but in modern English it has so nearly vanished that the Professor, in his own pronunciation, could discover only 56 of such a's in 10,000 sounds, or less than -y% per cent. ; while, owing to the preva- lent *' thinning ^^ or debilitation of that a before s and n {bask, chance, &c.), he was of opinion that in the popular speech of the United States the open a formed only some -rV P^r cent, of all the sounds ; i. e., out of every 150 primitive a-sounds, 149 have disappeared. {b) There is one point, however, of which the Professor does not seem to have treated, but which has an important ' While I was writing this Appendix, which was originally intended to form part of an extended investigation into the general principles referred to in the note to § 35 (a). =^ See §§6 and 64. ^ That is, of vowels and consonants together. Of the vowel-sounds alone, the long and short a are reckoned by Dr. E. Forstemann (Gesch. i. 21) to have constituted 76 per cent. ON THE AFFINITY OF R FOR THE OPEN -4-SOUND. 177 bearing on some parts of the foregoing treatise — namely, the influence of the sound represented by our r, not only in protecting an open «-sound coupled with it, but also in '*^ raising^' thereto other and feebler vowel-sounds, par- ticularly that of the dull e. This power of r (which, after all, is only one manifestation of a much wider influence ^) will, I fear, in those cases where it has hitherto afi'ected pronunciation only, be ultimately counteracted by the pertinacity with which our written and printed language maintains the weaker vowel. Thus, nearly everybody says c/ar^, s?irgeant, Ho^rtford, Bork- shire, [Lord) DsLrby, and astarn ; but we have to write and print clerk, sergeant, Hertford, &c. ; and where orthography is fixed and persistent, pronunciation, in the long run, can scarcely avoid becoming assimilated thereto. But where individual volition has had opportunities of making itself felt, — as in the case of proper names and technical terms — there the printing-office and dictionary are set at defiance ; so that we have Mr. Clark, Mr. Ds^rbishire, Mr.Jo^rman, Mr. Sergeant, Idirboard, st2^rboard, and others. (c) To observe this influence at work most vigorously, we must turn to the language of the people. Here many such modified forms have long been current as so-called vulgarisms; e. g., Iwning, 'dirnings, sartin ( = certain), consstrn ( = concern), varmint (= vermin), sarpent, 'varsity (= university), and '^ sarve him right. ^^ Two or three of these have crept into the slang of better- educated people ; but they will hardly creep any higher in our generation, if ever. In some of our provincial dialects we find this effect of r to be quite a cha- racteristic. The following words are from a humorous tale in theNorfolk dialect : — charch, consarned, Jarmins (Germans), lamed, matarnal, quarlsome, sarch, sarmon, sarpent, sarvice, 'tarmined, taming, and warld; where it will be seen that the change is determined by the sound and not by the sign of the vowel preceding the r (search, tuvnmg, world, &c.) . ^ See tlie section on Consonant Influence in Mr. H. Sweet's excellent " History of English Sounds," which, unfortunately, did not come into my hands until long after this Appendix was written. " The most marked influence is that exercised by the r. So strong is it, indeed, that in the present English hardly any vowel has the same sound before r as before other consonants " (p. 67). N 178 APPENDIX A. (d) Fortunately for tlie preservation of the open a-sound among us, many such ^^ raised '^ vowels had already been generally accepted by the standard dialect before the days of Dictionary-makers. In some of these the immediate parent vowel was eo ; as in carve, darling j form, far, farthing, hart, heart, hearth, smart, star, starve (=A-S. ceorfan, deorling, &c.). Of these, however, there were probably in very old times duplicate or dialectic forms involving ea instead of eo ; indeed, such a variation is a known characteristic of the ancient Kentish and Northumbrian dialects. But other words certainly made their first appearance in our language with a close vowel — as garland, marvel, parson, parsley, tarnish, varnish. In the case both of the foregoing classes of words and of many more individual examples (as bark, barley, bam, dark, hark, hearken, harbour, harbinger, yard, tar, mar), if our orthography had been fixed, say, by Chaucer, we should still be writing e (fer, hert, smert, &c.). It should be added that here and there, before rr, a ^*^raised ^^ sound appears as a short or as a faucal a; e.g., Harry, for Herry, by assimilation from Hewry^ {cf. arrant, tarrier, popular pronunciations of errand, terrier) ; quarrel ^ (Lat. querela) . In garner, garnet, there is metathesis (cf. granary, granite) . In the so-called vulgarism lark ("to have a lark, to go a-larking '') , the long a returns the compliment by attracting to itself an r. There can be no doubt that this is only a distortion of the old lac-an, Icec-an, which still lingers in Yorkshire in the form laik^, "to play/' The apparent identity of the word with the name of the bird (from which Webster derives it^) has led to a curious extension of the vulgarism ; so that we sometimes hear of people being ^ And not, surely, "a rough imitation of the sound of the French Henri," as Mr. Earle supposes (PhiloL, p. 344). ^ This -word (in which qu^kiv) really belongs to a group (tvarm, swarm, warp, &c.) in which the affinity of r for a is overpowered by the stronger affinity of w for d, — closely resembling that of v ixiS) for o in Latin. The relative potency of such affinities offers scope. for much deli- cate and discriminating investigation. « Identical in form, and nearly so in sound and meaning, with the Goth, laik-an. * Thus indicating a disadvantage under which lexicographers must labour who attempt to account for the constituents of our vocabulary without the opportunity of studying our dialects on their native soil. ON THE AFFINITY OP B FOR THE OPEN -4-SOUND. 179 ^' out SL-skyl^irkmg'' ! The sound of / also (which was originally in most cases derived from r) exhibits with us a similar in- fluence, although in much fewer examples — its more usual aflfinity being for the faucal a. Thus it raises a close sound in B^lms and perhaps behalf (where, according to some etymo- logists, the I itself is an accretion; cf. behoof) j and it protects the open sound in balm, calf calm, half, psalm, qualm, salve. In almond the I is again an accretion {cf. Fr. amande) . On the whole, we have here a clear case of counteraction and even reversal of the general tendency to phonetic decay (§ 8 (6)) by means of the action of one sound upon another in close contact with it. This special affinity of r for a, however, common as it is with us, is by no means universal : and a comparison therewith of its affinities in many other languages suggests that the English tendency is due to one of those slight differences of physiological formation referred to in the note to § 35 (a). {e) To obtain something like a definite valuation of this influence of r, I had already entered upon a statistical exami- nation of a similar character to Prof. Whitney^s ; but of this I shall here quote only so much as bears upon the relative frequency with which the several values of a, apart from other sounds, recur in our literary language. For examples of styles as different as could well be found — the loftiest poetical and the humblest colloquial — I selected the First Book of Milton^s '^ Paradise Lost ■" and Chapter xiv. in Dickens^s '^ David Copperfield.^' The former (omitting many foreign names) furnished about 1600 a^s of all sorts, and the latter about 1700. The proportions oi the several a-sounds making up these totals were Milton. ' Dickens. (i)^ a as in fate, mare . 27*9 per cent. 24*1 per ce^t. (ii) a „ all, what . 9*4 „ 8*2 „ (iii) a „ can, royal . 52*7 „ 59*4 „ (iv) a „ father . . 10*0 „ (nearly) 8*3^ „ (nearly). ^ In each of the first three classes, as we are here not much concerned with them, two or three shades, as it were, of sound are grouped under one symbol. ^ The word aunt occms about 100 times in the chapter examined. Assuming that it occupies the places of other a-words (of which, on the N^ 180 APPENDIX A. The difference of the styles^ and Milton's unrivalled command of all the phonetic resources of our language,, being considered, the smallness of the difference between the two series of pro- portionate values is very remarkable. The short and dull a, as might be expected, is, in Dickens's column as compared with Milton's, swollen to some extent at the expense of all the others ; but this may be almost entirely accounted for by the incessant recurrence of the conjunction and. (/) Pursuing the comparison of the sounds under class (iv) a little further, I found that in Milton about 158 recurrent sounds are distributed over 54 words, while in Dickens about 140 sounds are distributed over 50 words ; and out of these 54 and 50 words, respectively, 22 (representing no less than 98 recurrent sounds in Milton, and 69 in Dickens) are common to both. Again, if the same words be divided with reference to the connexion of a with r, there appear, of the 158 a-sounds in Milton, 88 ^^ protected " (of which 19 have been ''^ raised") and 70 unprotected, and in Dickens 69 protected (of which 12 have been raised) and 71 unprotected ; that is to say, some- where about half the open a-sounds are now protected by r. One may therefore venture to predict that this protected series represents the minimum limit to which the number of such sounds may some day fall, and that unless (as, indeed, is by no means impossible) some change in the organic or physio- logical value of r should gradually take place (§ 35, note) , the prospective debilitation of a will stop when it reaches that limit, if not before. Thus it will at some future time be a curious fact that the preservation of nearly all the open a-sounds in our language will have been due to the extraneous influence of r. It is worth while to remark in conclusion that whereas in the part of Dickens examined the verb are occurs a dozen times, in the part of Milton it occurs but once. For nearly up to Milton's time, as is well known, the whole of the present tense in which are is now found was furnished by the form be^ . We do our best, however, to neutralize this Miltonic scale, 10 per cent, should be a-words), I have reckoned it in only 10 times. If the word were rejected altogether, there would be left in this class a percentage of only 7-8 under Dickens. ^ '* If thou beest he. . . ." (verse 84). So always in the older wi'iters ; e, g.y Sir John Cheke : — << They he faithful at this day, when ye he faith- APPENDIX B. 181 advantage by eliding the initial a in reading and speaking {we're, you We, they We), unless the verb is emphatic or inter- rogative. Nevertheless,, on the whole, there is from this source a considerable phonetic gain. B. On some current Opinions respecting the Indo-European Aspirates. (a) The brief description of these sounds inserted in § 6 was adopted after a careful examination of all the evidence I could collect respecting them; but it is, of course, almost im- possible now to do more than determine certain limits within which their original phonetic value probably lay. There are, indeed, philologists who pronounce upon the nature of these sounds as decisively as if they had themselves been members of the Holethnos ; but there are others who, while accepting the Skt Aspirates as the best and indeed the only genuine re- presentatives of the primitive sounds,;speak of the latter with a hesitation justified by the obscurity which, after all, rests upon their origin. It is certainly becoming to exhibit such hesitation as to a point on which the old Hindoo Grammarians themselves are not quite so clear as one could wish^ I have already tried to show, however (§§ 33, 33), what seems to me reason for questioning the views most prevalent both as to the dialectic parentage of the Aspirates and as to the tendency of these sounds to descend to Medise. This neces- sarily shifts the ground on which their original nature has to be determined. Nevertheless, on one point we ought all to agree ; viz., that, whether the Media or the Spirant be the goal to which the Aspirate naturally tends, the first stage of Aspiration, although the memory of it may have perished, and its reproduction be impossible, can scarcely have been anything else than an incipient debilitation. {b) But it is just here that the oracles utter dark sayings. With Grimm ^, the Media is first *^^ thinned ^^ to the Tenuis and then ^^ thickened ■"' to the Aspirate. Curtius's opinion is less, not only to the king, whose subj ects ye be, but also to your lords whose tenants ye &e." ^ Max Miiller : Lectures, ii. 104, 16-5. ^ Gesch., p. 416. 182 APPENDIX B. decided that the Aspirate was an '^ AfFection of the Tenuis" ^ Grassmann ^ also considers the Hard Aspirate to have been directly derived from the Tennis. Schleicher^ (advancing halfway to the conclusion of § 25 ante) would strike out the Aspirate altogether from the sounds of the Ursprache ; and so would Prof. Max Miiller*. The latter allows^ indeed, that the '^ Aspiration of the Hard ^' was " the beginning of a pho- netic infection ^^j but his ^^ infection ^' apparently means ^^invigoration^^; for not far off we find the Professor figura- tively describing the Aspirate as ^' the boldest of the bold ^\ and representing it as the product of a ^^ raised"'^ Tenuis. If this really means that H is to be considered the weaker and A the stronger sound, it will, of course, greatly extend the ground of opposition between the Chronological Hypo- thesis of Grimm's Law and the Principle of Least Effort. But I can hardly think that such a view of Aspiration is the correct one. For if we bear in mind the ultimate forms to which the Aspirates are known to tend — /, v, )?, t5, and h ; which last (as in modern Italian and practically in French) has sometimes completely vanished, — the description of the earliest form of A (whatever it may have been) as a ^''raising'' of H, the purest and firmest of the mutes, will sound in our ears pretty much as if one should declare that the earlier stages of a disease ending in exhaustion and death are but the signs of a more robust health. In point of fact, the two termini of the sounds in question seem to me to furnish the only, or at any rate the most trustworthy, evidence on which to base a conclusion as to their former phonetic value and relative strength. For if they started from the complete contact of the Tenuis, and if they have descended to more or less open and unimpeded breathings, the Uniformity of Nature and the Principle of Transition require us to believe that the characteristics of all their intermediate stages must have lain between these two extremes. To assert the contrary would be to give to the line of phonetic development precisely opposite directions in different parts of its course. ^ Grk Et., 8rd ed., p. 437 ; but his hypothesis (§ 12 ante) really makes the primitive Aspirate a twofold affection — by softening as well as by aspiration. ' Kuhn's Z-S., xii. 81, &c. » Oomp., p. 11. * Lectures, ii. 222. ON SOME OPINIONS RESPECTING THE ASPIRATES. 183 (c) With respect to the almost universal assumption that the old Sanskrit Aspirates were the exact representatives of the primitive Aspirates^ I shall here add little to what has been already said in the text (§ 33, et alibi) . If that assumption were indisputably true, several considerations suggest that the prevalent deductions therefrom should be received with caution and tested with care. Our first business, of course, would then be to make sure that we had actually got hold of the most ancient Sanskrit Aspirates. Now the only means to this end are furnished by the descriptions of the old gram- marians and by the pronunciation of the modern Brahmins. As to the former, even if those descriptions were in some points less obscure than they are, it would still be scarcely possible for us moderns and foreigners (§ 35 note) to recon- struct therefrom the ej?act sounds described ; and the more elaborate the descriptions, the less, perhaps, our chances of success. If we turn to the latter, there arises the question how far the modern Indian pronunciation correctly represents the ancient sounds. Considering the general liability to change of the A spirates in cognate tongues, one would rather not assert that even this representation is exact. It may be granted that the mode of transmission of Sanskrit has been peculiarly favourable to the preservation of its phonetic sys- tem. But, without stopping to inquire whether a quasi-SLYti' ficial tradition of sounds may not be subject to dangers of its own, it seems to me scarcely possible but that the changes of vernacular in India during thousands of years should have affected the training of the Vocal apparatus of the natives. Brahmins included, and, in spite of and unperceived by themselves, should have exerted a sort of reflex perturbing action upon the sounds of the traditional language. (d) But, once more, the phonetic values currently assigned in Europe to the existing Skt Aspirates are generally deter- mined for us by our fellow-Europeans, and only indirectly by the Hindoos. The effect of this circumstance also must be taken into account when great exactitude is in question. In the case of ordinary men, their own belief that when speaking another language they exactly reproduce its sounds (where these differ to any extent from the sounds of their own lan- guage) is often an amiable delusion; and not seldom their ear 184 APPENDIX B. altogether fails to appraise some sound or other at its proper phonetic valued A closely related mistake is to suppose that sounds which do not occur in our native tongue, and which we therefore (especially in mature or advanced life) find diffi- cult to acquire at all, are in themselves essentially difficult. One's opinions on such points are_, malgre soij moulded by one's sensations; for the practical ease and difficulty of sounds depends greatly on the previous education of the vocal organs ; and where we too readily employ the terms ^^ easy '' and ^^ difficult " we should rather say ^^ familiar " and ^^ un- familiar''. The great Sanskritists of Europe tower, indeed, above ordinary men; but they would probably themselves be very backward in claiming exemption from the inevitable operation of physical laws. (e) To be brief, the points on which, as it seems to me, the current views must leave their adherents in doubt are : — (1) whether the ancient Skt Aspirates may not themselves have differed from the earliest] I-E. Aspirates (§ 33 ante) ; (2) whether the Aspirates of the modern Hindoos again may not in turn differ slightly from the ancient Skt Aspirates ; and (3) whether the pronunciation of these modern Aspirates and the estimate of their comparative difficulty by Europeans do not almost necessarily differ from those of the Hindoos them- selves. However minute these successive differences may be, they suffice individually (and still more collectively) to affect the judgment when great phonetic exactitude is required, as, for example, when we are considering the somewhat delicate questions : — whether Aspiration was, at its incep- tion, a debilitating affection or not; and whether or not the breathing, from which the Aspirates get their designation, was then an inseparable element in their composition, as it still is in that of their descendants the Spirants. I cannot now enter into the further inquiry whether, and how far, the accumulation of minute erroi^, like those suggested above, may have influenced the choice, or at least the general ac- ceptance, of the composite symbols [kh, &c.) by which in Europe the modem Indian sounds are represented. But as to the converse question, how far the prevalence of these ^ See the amusing case of hactshtasch (bakhshish) quoted by Max Miiller, Lectures, ii. 180. APPENDIX C. 185 symbols may influence men-'s ideas of the sounds^ I remark^ in conclusion^ that^ by their very composition^ they seem to give, to each of the two delicate phonetic questions just stated, an answer which I can scarcely believe to be, in either case, correct.* C. On the Extension of Language by Phonetic Variation. {Opinions and Illustrations.) (a) The process referred to in § 17 {d) and formulated in §§ 28-30, is too well known to need further theoretical treat- ment. I only propose, in this Appendix, to fortify my posi- tion by the authority of two or three writers whose opinions will hardly be gainsaid, and by a few familiar examples that may suffice to exhibit clearly the leading characteristics of the process. Bopp remarks: — ^' Die Spaltung einer Form in verschiedene mit grosserem oder geringerem Unterschied in der Bedeu- tung, ist in der Sprachgeschichte nichts Seltenes''^ (V. G., 3rd ed., i. 33) . And again: — ^' Hierbei hatte man zu beriicksichti- gen dass in der Sprachgeschichte der Fall nicht selten vor- kommt dass eine und dieselbe Form sich im Laufe der Zeit in verschiedene zerspaltet ; und dann die verschiedenen For men vom Geist der Sprache zu verschiedenen Zwecken benutzt werden'' (Id., ii. 391). Prof. Key is particularly felicitous in his illustrations :— ^ As an illustration of the dangers which beset us when we attempt to represent to the eye the sounds (sometimes even the simplest sounds) of an alien tongue, with which we suppose ourselves perfectly familiar, and of the mistakes which we may thus propagate, I cite the following fact : — A French "Professor of English" recently constructed an elaborate system of symbols for conveying to his countrymen correct ideas of our English sounds ; and among them I find these two — As and Mz. When the reader is tired of guessing what these can mean, he may like to be in- formed that they are intended to denote the two values of " le th anglais" ! The ingenious inventor certainly does his best to neutralize the appalling effect of his symbols by an elaborate verbal description of our Dental Spirants. Nevertheless what Frenchman, beholding those combinations, could possibly conjecture that these spirants are the simple, feeble, and to us easy sounds they are ? 186 APPENDIX C. ''When a word has established itself in several dialectic varieties of form it is a great convenience to distribute any varieties of meaning which may belong to the parent word between them j and thus a dissolution of partnership, as it wercj takes place, each dialectic variety commencing business on its own account with its own separate stock /^ ("Philol. Essays/^ Essay I.) Again, Tick : — '' Auch ist ja bekannt mit welchem Heiss- hunger sich die Sprachen auf neu hervortretende Lautspal- tungen werfen, um dieselben als Trager von Bedeutungs- differenzen sich nutzbar zu machen." ('' Spracheinheit,^' p. 202. See also p. 210 for some valuable illustrations.) Or, as Dr. E. Forstemann says, '' Ein Wort hat eine so grosse Begriffssphare dass dieselbe fiir die Sprache zu lastig wird ; da pflegt ihm denn ein neues Wort einen Theil dieser Sphare abzunehmen." (Gesch., p. 224.) So too Whitney (Jolly), p. 175 : — ''Nicht selten tritt der Fall ein, dass einer neuen Schattirung der Wortbedeutung eine Differenzirung der Lautform zu Hiilfe kommt^'; and numerous examples are quoted from the German [e. g., nah and nach ; konnen and kennen ; Geist, GaSj and Gischt ; opfern and opferiren; &c.). {b) English abounds in such cases, the invasions and powerful phonetic influences to which it has been exposed having been particularly favourable to their production. Many of these have diverged phonetically and differentiated their meanings almost, as it were, under our eyes. Here, for example, are some variations by simple vowel-change : — A b«nd, a bond ; borne, born, b«irn ; cheap, chap-man, chop, chaffer; to dab, to dawb ; human, humane; km, km-d ( = genus), km-d (=benignus) ; mood, mode; person, parson; piety, pity; a plait, a pleat, and (O.E.) plight; rise, raise, rowse ; scar, score, shear, share, shire, shore ; to sit, set, seat ; a stick, stake, stock ; throwgh, thorough ; i^ight, whit, au^ht ; &c., &c. The varieties produced by modification of the con- sonant are perhaps still more numerous. In the case of some, indeed, the logical variation is but slight, as in dviW, Mrill ; oweJ, OMght; carl, cAurl; cawon, cawwon. In others the least phonetic variation is made to indicate, for example, different parts of speech, as, advice, and to advice ; belie/, and EXTENSION OF LANGUAGE BY PHONETIC VARIATION. 187 to believe; glo55, gloze; lose, loss, loo^e; to nse {s=z), and use. The same result may even be attained by merely shifting the accent^ as in august', and August ; per'fume, and to per- fume. Some of the consonantal variations in English very closely resemble the debilitations to which I have attempted (§ 43) to reduce the so-called Lautverschiebung ; e. g., dike, di^, di^c^ ; deck, t?iSLtch ; wake, [wa^ ?] , watch ; stick, stitch ; hsmg, hin^e; milk, mik^; dra^, dvedge, dindge; reac?-y, ratA-er. In the transition from ancient and irregular inflexions to more modern and regular ones, some of the former, instead o£ being allowed to die out, are seized and diverted to a cog- nate use. This is particularly the case with the old passive participles in ~en. Thus, bounden, drunken, (for)lorn, molten, shaven, shorn, shrunken, stricken, (un)washen, are now gene- rally used as simple adjectives. But the most remarkable examples of variation are furnished by those words which have radiated, with a moral significance, from proper names. Thus, ^^ brummagem,''' a popular pronunciation of Birmingham, is or was often used, for a well-known reason, as an adjective =1 ^'spurious'' {e.g., of jewellery, &c.)j so, too, from Bethlehem (Hospital) we get ^^ Bedlam '' as applied to any scene of uproar and confusion, and from Magdalen " maudlin.'^ (c) These are only a few of the results of what one might almost call the instinctive artifices resorted to by the Lin- guistic Sense, in order to bring or keep its powers of expres- sion well abreast of the incessant additions to our ideas. In other languages we meet with expedients that seem (although they really are not) still more artificial. Such, for example, in French, is the variation of meaning attributed to certain adjectives according to their position before or after their nouns, and the assignment of dificrent genders even to nouns of one and the same origin {aigle, manche, pendule, &c.) ; in which latter case the means of distinction have to be supplied by the accompanying article. Somewhat similar to the former of these " artifices '^ is the English practice in reference to what are called ^' prepositional '^ verbs. Thus a country is over-run by the enemy ; but a person is run-over by a wagon ; a cup, pail, &c., is up -set ; but a mast or monument is set-up ; a swift horse out-runs a slow one ; but water runs-out of a vessel ; and so on. 188 APPENDIX C. [d) But perhaps tLe finest field in whicli the Dififerentiating Process may be observed at work, at least in modern lan- guages, and preeminently in English, is the desynonymizing of words once identical in meaning. The materials on which it here operates are not, indeed, supplied by dialectic varia- tions of sound; but the nature of the process and its logical effects are precisely the same. For closely related dialects we have now to substitute distinct languages, or dialects re- motely akin, which extraneous causes have brought together, and which contribute to the resultant language two or three totally distinct expressions for one and the same general idea. Many of these duplicate and triplicate forms of course die out, the corresponding ideas being sterile (§ 30). But those that are retained gradually become assigned to diverging varieties of meaning, in just the same way as the mere phonetic by-forms quoted above. Hence the description of the action of language in respect of these so-called synonyms is couched in almost the very same terms as those employed to describe variation by Phonetic Differentiation. ^^ All languages tend to clear themselves of synonyms as intellectual culture advances, the superfluous words being taken up and appro- priated by new shades and combinations of thought evolved in the progress of society ^^^. This division of the subject, however, has already been sufficiently elaborated by others^. I only refer to it with the view of strengthening the evidence for the operation of the vigorous linguistic process under con- sideration. It merely remains to argue, in conclusion, that if the phonetic process is so active in our own era, among mature languages, whose vocabularies are already so copious, and which have availed themselves so largely of one another^s resources, surely the same process would be Yery much more active in the infancy of a language, whose voca- bulary was yet, in great part, to construct, and which had to find all its resources within itself. ^ De Quincey : quoted by Archbishop Trench in his Lectures " On the Study of Words," Lect. v., note. ^ See the admirable Lecture just referred to ; or Lect. xxvi. (Smith's edition) of Marsh's " Lectures on the English Language." APPENDIX D. 189 D. 071 Mr. Sweet's Scheme of the ^' Lautverschiebung.^'^ [a) It is a misfortune for the subject I liave been attempt- ing to handle, that Mr. Sweet has touched upon it merely, as it were, by the way, and in subordination to an individual doctrine which, if true, ought contrariwise to be shown to follow from a correct rationale of Grimm's Law. The doc- trine referred to is, in its author's own words, that of the *' Vocal character of the Thorn'' — a phrase which, expanded, means that, among the Low Germans in general and the Anglo-Saxons in particular, there was originally but one Den- tal Spirant, and that that one was the Soft or Voiced Spirant [dh = th in thou) . It is a necessary extension of this doctrine that the '^ vocality " attributed to the Dental Spirant must likewise be predicated of the Labial and Guttural Spirants, as well as of certain pairs of Sibilants ; but the bulk of Mr. Sweet's remarks have reference to the Dental Spirant only. The arguments in favour of the unity of this Spirant are : — that in Gothic it is always represented by a single character ; that the earliest Old-English (i.e., A-S.) MSS. severally em- ploy one or other of the two characters ()? or ^) throughout, and not both; and that in H.G., both old and modern, the L.G. Dental Spirant has the single representative, d. The arguments in favour of its vocality are : — that the Latin di- graph th, which appears in some of the oldest remains of our language, and which should represent a hard or breath sound, was speedily abandoned as if unsuitable ; that the symbols J? and ^ seem to be both derived from the letter D {d) ; that the L.G. Spirant is represented by c? in H.G.; ^^and, in some of the oldest documents verging towards L.G.,^' by dh in all positions. These are all matters of fact. But to derive the L.G. Soft Spirant directly from the CI. Tenuis (as is required by the Historical or Chronological Hypothesis of Grimm's Law, which the author adopts) would be a departure from the Principle of Transition^; for there would manifestly be ^ Appendix I. to "Gregory's Pastoral Care" (edited for tlie Early English Text Society) ; ia connexion with which should be read the " History of English Sounds," issued by the English Dialect Society. ^ See ante, § 54 (/), note. 190 APPENDIX D. both loss of energy and breach of contact^ at one and the same time. Hence, considering that 8 in modern Greek, d medial and final in Danish and Old Norse, and sometimes (apparently) d medial in English {e. g.^ fac?er, moc^er^) have passed into Soft Spirants, Mr. Sweet feels justified in assuming '' an earlier stage of the Teutonic languages, in which the old Aryan t was changed into d) whence the later {dfi) arose from imperfect stopping/' Applying therefore {note to § 20 (a) ante) the process which I have designated Cross Compensation, he constructs the following scheme of Verschiebung for the dentals : — (A) Old Aryan t d dH (^oft Aspirate). Oldest Teutonic d dH (Soft Aspirate). Oldest Low German . (/^g^frant) ^ Oldest High German, d .jT (Hard , . *^ Aspirate) "^^ (5) It might seem to be intended that the O.H.G. system should be derived at once from the Oldest Teutonic ; to which it clearly holds the same general relationship as the Oldest L.G. system. But I have authority for stating that we are to supply a stage intermediate to the third and fourth of the foregoing systems ; so that the O.H.G. is supposed to be de- rived from the O.L.G. system by two stages, in just the same way as the latter from the ^' Old Aryan ''; only, by a slight variation of order, d in O.L.G. is first to remain stationary, while dh and t are to exchange places. Hence, with the ne- cessary extensions to the gutturals and labials, and using our own notation (§20 (a)), we arrive at the following general scheme : — (B.) S, = H A S, (%J) = S A H, 2„ = A s n,y (2 J = H S A, 2, = S H A. See § 6. ' See the provision made for these words in § 49, note. lautverschiebung/^ 191 (c) This scheme certainly represents a bold and original extension of the Historical or Chronological Hypothesis of Grimm^s Law; and if that hypothesis were demonstrably true so far as it goes^ every honest inquirer would hail this extension of it as a closer approximation to the ideally per- fect truth of which we are all in search. But^ as things stand_, there is reason to fear that any extension of the said hypothesis may only multiply some or other of those doubtful characteristics to which I have been unfortunately compelled to direct attention. From my point of view such fear seems to be justified in the case before us ; and so_, as the scheme appears only to have been put forward tentatively^ as it were,, and in order to support a doctrine which is itself open to vigorous controversy, I need hardly perhaps apologize for hesitating to believe that its claims to acceptance have been made good. I defend my hesitation by a reference to the objects I have striven to attain throughout the foregoing in- quiry, viz. : — to harmonize Grimm^s Law with the Principle of Least Effort ; to reduce the phonetic movements it really represents under known linguistic processes ; to show that the functional relationship of 'Z^, % , and S^ follows necessarily from the conditions of their evolution ; to explode the idea that the German Mutes have been of a more fluctuating nature than those of other Indo-European peoples ; and to construct a theory that should embrace and explain all the phenomena represented by the Law, and that should, if pos- sible, reduce its apparent complexity to simplicity, and not render it still more complex. {d) With these objects the acceptance of the scheme now before us is quite incompatible. If it is an objection, for example, to the Chronological Hypothesis pure and simple that it twice appears to offend against the Principle of Least Effort, much more will it be an objection to this extension that it appears to offend four times over^ The ^'Cyclo- functionaP^ relationship indeed (§§ 2, 3) of the three Mute- systems does not now offer much difficulty ; but that is only because it is completely destroyed ; that is to say, S„ no ^ It may be said that tlie application of Cross Compensation to the scheme removes this objection; but I doubt (§ 20 (c)) whether the other features of the scheme justify such application. 192 APPENDIX D. longer appears to be derived from (S^.,) by the same amount and kind of phonetic change as (S^) from ^^ ; and so of the rest. Again_, as to the nature of the mutes, instead of sta- bility we find perpetual flux. Let us trace, for example, the supposed metamorphoses through which the primitive Tenuis^ must pass before it reaches the final H.G-. stage, and let us seriously ask ourselves whether it is possible that the very bones of articulate speech should be reduced to such a pulpy and fluctuating condition ; or is it like the economy of nature that the framework of language should be repeatedly broken up and rearranged with the feeble result of bringing its vari- ous parts again and again into one and the same position? If we contrast this supposed transmutation or repeated revo- lution with the acknowledged quietude and stability of the liquids and vowels, the more impressible elements of speech*, we must, I think, hesitate to believe in it except upon a cogency of evidence amounting to the clearest demonstration. {e) But even this amount of change does not account for all the phenomena of the case. There are, in the first place, some anomalies of phonetic generation, so to say, yet to be explained. Not to mention the various ways in which every sound seems capable (Scheme (B)) of directly generating every other, and indirectly itself too, we observe in the author's own scheme (A) that the Aspirate exhibits in the successive sys- tems, widely different phonetic powers or qualities. It is, in the first two, a Soft Aspirate, in the next a Soft Spirant, in the last a Hard (but spurious) Aspirate ; and we are to imagine not only that each of these different phonetic quantities may directly generate or be generated by either of the others, but also that they may stand as compensatory equivalents to one anothel". Against the passage of the Soft Aspirate to the Soft Spirant (which is the natural line of debilitation) nothing, indeed, can be urged — if, at least, the O.L.Gr. system (Scheme (A) ) be supposed to have been cut off from the imaginary Old Teutonic some length of time previous to the evolution of the next following system. But that the O.H.G., in merely in- ^ Scheme B above ; where it will be seen that H is supposed first to have become S, then A, then H again, then S again. ^ The reader will have noticed that all the examples of Cross Com- pensation quoted in § 18 are furnished by the feebler sounds. 193 terchanging the Spirant (which it already possessed) with the Tenuis, should also have altered its nature and increased its strength,, is contrary alike to the Principle of Least Effort and to what we observe as taking place in Cross Compensa- tion. Attention has already, however, been directed (§ 20(c)) both to this last point, and likewise to the disappearance of some of the systems which must once have stood as the fixed and standard systems, and against which some of the fluctuating systems must have differentiated themselves^. Leaving these points, therefore, I only remark further that in the labial and guttural series, the earliest extant form of ^^, as we have seen (§§16 (c), 52 (e)), gives us — not p, ph, and B ; k, ch, and G, at which we ought to arrive by this scheme, but — ^, ph, and generally F; h, ch, and generally H. Of this phenomenon the scheme provides no explanation. By what intermediate stage, or by what mode of compensation con- sonant with the Chronological Hypothesis, these odd Spirants are made to occur in these places, it is not easy to see ; so that, in spite of its complexity, this scheme, after all, leaves us in ^ In doubting (§20 (c)) the possibility of a complete absorption of the standard by tlie fluctuating dialect, so that no trace of the former shall anywhere be found, I would not be understood to assert (what would be absurdly wrong) that the latter can make no impression of any kind upon the former. Stray examples of the process of Cross Compensation (like irregularities of Aspiration (§ 18 (d)) may, of course, in rude times, easily force their way into a more stable dialect. It is in this way, perhaps, that a short but striking list of exceptions in French derivation may be ex- plained. Everybody knows that the Latin internal h becomes v in French, that initial v remains r, and internal v disappears ; the examples of which changes are very numerous. Yet in some half-dozen words (See Brachet, Diet., s. V. hachelier, and In trod., p. 93, note) v is unaccountably strength- ened to b. M. Brachet (ubi sup.) asserts, indeed, that such exceptions '^ s'expliquent aisement''^ but he judiciously abstains from any "expli- cation" beyond vague generalities. If we suppose, however, that, in the early centuries of our era, while b and v still maintained their original places in the purer Latin of the upper classes, the debilitation of the former to v in the commingled popular dialect evoked a compensatory movement which often raised v to b, and that some of these &'s gra- dually worked their way upwards, we shall have an explanation which is rigidly precise. M. Brachet involuntarily supports this view by pro- ducing evidence to show both that these exceptional &'s were in early times much more numerous than now, and also that they were then coDsidered vulgarisms. . o 194 APPENDIX D. bewilderment. If_, then, the scheme, like the hypothesis to which it is attached, is really untenable, the rejection of the doctrine which the scheme is intended to explain may be thought necessarily to follow. But as the independent argu- ments by which it is attempted to establish that doctrine come into contact at several points with my own line of reasoning in the foregoing treatise, I will devote a few lines to a separate examination of them. (/) As to the original unity of the Dental Spirant (or of an older sound of which it is a modification) there can be little doubt : so much is requisite in order that the German Mute-system may be made to conform to the general I-E. type. But to how late a date that unity may have continued — or, what is the same thing, at how early a date a second and weaker Spirant began to be given off — nobody knows. The fact that a single sign is used in Gothic, and either \ or "S in the earlier A-S. MSS., does not, of course, prove that a splitting of the sound had not already commenced. Nay, if we suppose the evolution of the younger sound to have com- menced before the Germans had begun to use any one of the symbols which have come down to us, and that they appre- ciated the difference between the two Spirants, they would still be likely to think it a piece of wise economy to make one si^n do duty for two sounds so closely alike, and of unnecessary refinement to invent a second. Even so they would have shown themselves very superior to their English descendants, who, having two most convenient signs ready to hand, have neglected to keep either of them, and have sub- stituted for both the one clumsy and inappropriate Latin digraph th. If, however (as it will be safer to assume), the older sign ()?) alone was appropriated to the Spirant before the evolution of the younger variety of the sound had made any or much progress, or while (as is usual in the case of linguistic changes) the old and new varieties were still used indifl^erehtly, that one sign would continue to be used for both sounds as a matter of course ; just as, in modern English, the same written vowel continues, for merely historical reasons, to represent three or four distinct vowel-sounds, even when these sounds, phonetically considered, have other and more appropriate symbols. MR. sweet's scheme OF THE ^' LAUTVERSCHIEBUNG.^' 195 {g) But the relative age of the younger sound and the older symbol is a point of only secondary importance in comparison with the relative age of tlie two sounds. It is agreed that the Dental Spirant was originally one, and that in later times a younger variety was evolved from that one. It follows^, by the Principle of Least Effort^ that the original sound must have been the stronger or more difficult_, and the younger the weaker or easier, — and conversely in each case ; for, with respect to the Dental Spirants, we can now perhaps only determine the age by the strength, and not the strength by the age. But which is the stronger ? To this question, till recently, there would probably have been but one answer. It was, in fact, answered in advance by the very epithets '' hard " and '^ soft/' by which we distinguish the th of ^' thing'' (for example) from the th of ^^ thou'' and ^^ thee." Mr. Sweet, however, as we have seen, propounds an opinion hereupon which (as he justly apprehends) is scarcely of a kind to meet with acceptance among philologists. It is none other than that what is called the Hard or Breath sound is really the weaker, and requires less effort to produce than the Soft or Vocal Spirant, owing to the contributory action of the chordae vocales during the utterance of the latter. Now, if this were a difference to be settled by weight of opi- nion, I should not risk mine against the one just stated. But it is not : it is a question of facts. And it is a fact, I believe, in the first place, that the Soft Spirant may be made as voiceless as the Hard ; in truth, it is by testing the whispered sounds one against the other that we can best appraise their relative strength. This, however, is by no means essential ; for, in the next place, I believe it is another fact that the action of the chordae in the production of voiced sounds, if perceptible at all, is less so than that of any of the other organs or muscles concerned in the process, and, where two sounds would otherwise be of exactly equal difficulty, that action, if called for by one of them, will scarcely cause this one to preponderate in difficulty. The part they play in speech may almost be compared to the part played by the iris of the eye in vision; they act, at least within certain limits, by almost involuntary mechanical adjustment to influences external to o2 196 APPENDIX D. themselves. The effort of speaking is rather to be measured by the degree to which it calls into action the larger muscles, closely connected, indeed, with the vocal and respiratory passage, but yet not themselves strictly " organs of speech/' Thus the ^^back'^ vowels are stronger than corresponding '^ front ^' ones, principally because of the amount of work which the distinct production of them entails upon the max- illary muscles. The calls upon the respiratory apparatus are pretty nearly the same for both ; but the former require a considerable hiatus, while for the latter the mouth may be nearly closed. So, in the case before us, it seems to me yet another fact that the diaphragm, and even the abdominal muscles, are, in producing a clearly-marked Hard Spirant, exerted with a vehemence which, in producing the corre- sponding Soft one, we cannot equal if we would. In passing from the former to the latter, it is the very relaxation of the muscular tension which leaves the chordce free to \4brate ; and in changing, reversely, from the ^^ Soft " or " Voiced " Spirants (v, z, zh, "S) to the related " Hard '' or " Voiceless '^ ones (/, 5, sh, \), the suppression of the vibration will be distinctly felt to be due to the additional vehemence of the general effort. But even the action of the smaller muscles must not be left out of the account. In the production of the Hard or Voiceless Dental and Labial Spirants the pres- sure of the tongue and lip respectively against the teeth is firmer, and the rima through which the breath has to be forced is narrower, than in the case of the corresponding Soft or Voiced Spirants. Indeed the greater resistance thus offered to the passage of the breath is the necessary counter- part of the greater sum-total of force employed in its expul- sion. All this is a matter of experiment which the decided difference between the two related Spirants forming each pair allows every man satisfactorily to try for himself. I am afraid therefore we have no choice but to revert to the cur- rent opinion that the Hard or Voiceless Dental Spirant is the stronger, and ipso facto the older, of the two, and to inter- pret accordingly all the subordinate evidence adduced by Mr. Sweet. But perhaps even this evidence should hardly be passed over in silence, lest it should be thought to offer MR. SWEET^S SCHEME OF THE '^^ LAUTVERSCHIEBUNG/^ 197 independent support to the opinion I am controverting^ and to be irreconcilable with the opposite opinion ; which is by- no means the case. [h) It is manifest that the question under debate, which- ever way it is settled_, cannot be settled for one pair of Spi- rants (as ]? and S) apart from the other pairs (as / and v, s and 2'), the members of which are mutually related in precisely the same way. Indeed Mr. Sweet himself proposes to embrace the labials, at any rate, under the novel doctrine he lays down. For support he refers to the history of the u ( = v) of O.H.G._, which V is now =f, and to the prevalence of v for / in modem Dutch. The former circumstance has already been embodied in its place in the general theory of this book (§ 54 (/)). The value of the latter depends on the answer to the question whether the Dutch pronunciation, which cannot be traced back for more than three or four centuries, is to be taken to represent the ^' Oldest Low German '' in preference to that of an overwhelming majority of the cognate peoples. The affirmative could only be justified by satisfactory independent evidence. On the other hand, over and above the physiolo- gical argument of the preceding paragraph, a great deal of evidence for the opposite view might be collected by any one who had leisure for the task. There is, for example, the argu- ment from '^protection^^ and ^^ assimilation^' (e.^., thrit;e but thri/t, give but gi/t, &c. ; with which compare /ived, shoYed, moved, &c. ; compare also kicJcSj smi^e^, whips, Sec, with wags, huilds, robs, — i. e. wagz, &c. &c.) ; from which we may infer that f:v, or s : z : : k : g : : t : d, &c. Similarly, in Latin, we have scrihere but scrii^si, nvhere but nw^^i, &c., where s is itself the assimilating agent : compare also the insertion of p in hiemps, sumpsi, dempsi, &c. Reversely, when a protecting Tenuis has disappeared, a Spirant that changes at all changes from voiceless to voiced. Thus in our incessantly used " is " the 2'-sound is demonstrably younger than the 5-sound which the word must have had when it ended with t (Goth, ist, Lat. est, Grk. earl, Skt. asti, &c.). Add the case of the soft sibilants in our abstract nouns in -ion derived from the Latin {e.g., '^'^ invasion,'-' "derision,'" '^'^ prison,-'' &c.), where the ^'s were originally evolved directly from t. With respect to s and z, indeed, no evidence can be clearer than that furnished by 198 APPENDIX D. the Gothic,, where the connective and relative enclitics begin- ning with a vowel^ as well as case-suffixes^ [-uh, -ei, -is, -a, &C.J often coalesce with cases and stems ending in s, with the general eflPect of turning that s into z. Thus hvas + uh gives hvazuh ; J7is + ei gives ]>izei ; Moses, + is gives Mosezis ; ans + a gives anza ; and so on, in instances too numerous to quote. Two things are thus clearly indicated — (1) the relative age of the two Spirants, and (2) the agency (bivocalism) to which the softening of 5 to 2* is due. If, therefore, returning to the Labials, we find a considerable number of words (like ele\en, tivehej gioYe, hease, sto\e, over, heawen) involving a t; in modern English, but only found with / in the older L.G. dialects, we may be sure that the apparent history of the change is the real one, and agrees with the physiological cha- racter of the two sounds — that the voiceless /, like the voice- less s, is the older and stronger; and the voiced v, like the voiced z, is the younger and weaker. And what is true of these must be true of the Dental Spirants likewise. (i) Such apparently obvious conclusions from no scanty evidence Mr. Sweet^s general doctrine compels him of course to repudiate. By this doctrine our Hard or Voiceless Dental Spirants are descendants of older Soft or Voiced Spirants. Hence the supposed " mystery ^^ of the Soft or Voiced initials in some of our commonest words {/^ee, thou, then, there, &c.) is ^^ solved '^ by assuming that '' these words are simply archaisms — remnants of an older stage of pronunciation preserved unchanged by the frequency of their occurrence." But, even granting the " mystery,^^ surely this " solution,^^ from whichever point of view it is regarded, is open to sus- picion. For, in the case of an indigenous debilitation, such as is here in question, it is the very words in most fre- quent use that ought to be soonest and most completely affected' ; so that if the initial Spirants just referred to were originally Voiced (as they now are) , and the natural direction ^ If the case were one of phonetic attack from without, where one dialect is called upon to resist the infection of another in contact with it, the solution above challenged might then hold good. Such resistance (" Retention of the Tenuis ") has therefore been properly assumed as the basis for an explanation of the Exceptions to Grimm's Law (§§ 48 (c) and 50 (a)). 199 of change ran towards the Hard or Voiceless Spirant^ the words involving them should have been among the first to be drawn in by the movement, and should now exhibit the Hard Spirant — which they do not. If, on the other hand, those Spirants were originally Hard or Voiceless, and debilitation took the direction currently believed, then the words in question actually do exhibit the very change that they ought. On the prevalent (and, I believe, correct) doctrine, therefore, everything is in order, and there is no mystery in the case ; for the softening of the Dental Spirant must be supposed to have proceeded in the same direction, and in a parallel line, with that of the other Spirants, only perhaps more rapidly and much further. Among all of them the softening process clearly originated, as might be expected, in the interior or bivocalized position ; which fact alone indicates distinctly enough the nature of the change (§ 34, last notej and Appen- dix G) ; but in the case of the Dentals the debilitation not only reduced under its sway all the Spirants so situated, but even attacked some of the initial ones. [k) As the view I am defending represents beyond ques- tion the order of nature, there is no real need to enforce it further. Nevertheless, as Mr. Sweet^s additional arguments raise one or two other points of interest, I may perhaps be allowed to suggest the way in which the theory of the fore- going treatise requires those points to be dealt with. As to the fact that H.G. c? represents L.G. ]?, it is a vital part of our theory that these two sounds were independent debili- tations of the primeval Tenuis / (§§ 26, 31, 36, et passim) ; so that the strength of one of them gives no indication whatever of the strength of the other. Again : the fact that dh is written for d in some of the oldest remains of H.G. admits of two explanations : either dh was simply =d, just as, some- what later, th was in H.G. (as it still is) =^ ; or, if dh really represented a Voiced Spirant, then that Spirant occupied precisely the same position between the ideally correct H.G. d and L.G. )? which v at the very same time occupied between the ideally correct H.G. Z> and L.G. / (§ 54 (/)) ; that is to say, it represents a partially executed assimilation of some of the H.G. subdialects to the L.G. in respect of the Dental as well as of the Labial Spirant; which movement, however, as regards 200 APPENDIX D. the Dental, was unable to advance^ or even to maintain its ground. Once more : that the Latin digraph th should have been dropped or expelled from A-S. is just what we should expect j for, in the first place,, there was already at least one native symbol ()?) (probably there was also a second (S)) actually in possession. In the next place the A-S. Alphabet was based — not directly on the Roman Alphabet, but on the Old Irish variety of it^ The former, although used by the missionaries from the Continent, failed to establish even its own simple symbols (where these diflPered from the Irish), and was therefore still less likely to succeed with its com- pound ones. But, thirdly, the diagraph th, even on its native soil, never, so far as we know, represented a Spirant, and scarcely even the Greek form of the Aspirate (§ 35, note), except perhaps to the Roman doctij by or for whom it was invented. In all the Romance languages^, as is well known, it is pronounced and mostly written as / ; so that the com- bination was probably unsuitable for the Spirant, not so much on account of the hardness or energy, as of the too complete contact, assigned to it. From the fact that, in spite of all these drawbacks, attempts were actually made to em- ploy the digraph at all, I should draw just the opposite conclusion to that which Mr. Sweet draws from the fact that the attempt was unsuccessful. (/) The only point of interest now remaining to be con- sidered is that arising out of the shape of the symbols )> and "S — or, really, as I shall presently show, of the older J? only. Mr. Sweet adopts the supposition^ that this symbol was formed from the Latin D by producing the perpendicular stroke in both directions. Now D represents a soft sound, so therefore (for such is the assumption) did \. But the danger of basing an argument on such an assumption will be manifest when we remember that the wen (p) approaches still more nearly to D in shape, the upright stroke being produced only one way. In point of fact, no inferences would generally be more fallacious than such as should propose to determine, from 1 Latham, Introd. to Johnson's Diet., p. Ixxiv j Earle, " Philol. of the Engl. Tongue," pp. 103, 109. 2 Diez, Grammatik, i. 226. .From Mr. Vigfiisson, Icelandic Diet., suh lit. J>. . lautverschiebung/' 201 resemblances among symbols,, either the relative phonetic powers of the sounds they stand for, or indeed the origin of the symbols themselves. On such grounds the powers of our O and Q, F and E^ P and R should be nearly the same, and those of T and D, S and Z, F and V, K and G should be widely dissimilar ; while it might be inferred that the latter member of each of the three first-named pairs of symbols was manufactured out of the former. But even where such an inference would be apparently correct it might be historically false — as in the case of P and B ; for here it is certain that the former came — not from the latter (nor this from that), but — • from the Greek IT by first shortening (F) and then bending up (P) the right leg^. The only case in the Roman Alphabet and its descendants where the inference would be (under certain conditions) correct is that of C and G ; of which the latter was originally a mere graphic variation of the former, although subsequently the former, oddly enough, acquired the power of, and at last ousted, the early Roman K. Yet this case, which (like others to be presently noticed) is one of spontaneous graphic radiation, as it were, within one and the same language, can hardly be quoted in support of the intentional manufacture of a new sign, for the purposes of one language, out of the symbols of a different language. Besides, the argument based on the supposed derivation of ]> from D may be made to cut two ways. For the object of connecting those symbols is to suggest that the sounds they stand for have the common quality of softness. But the Soft sound represented, it is said, by )? became the parent of its corresponding Hard : why should it not be maintained there- fore that the Soft Dental Mute was also the parent of its corresponding Hard ? and so, of course, with respect to the Soft Guttural and Labial. Combining this extension with that in h {ad fin.) above, we should, in truth, give the doctrine in debate its widest scope; but, in so doing, we should both fight against patent facts, and overturn what have hitherto been considered the very foundations of Phonology. (m) But if the supposed connexion of )? with D were as certain as it is uncertain, Mr. Sweet^s main doctrine would ^ See Ritschl, "Zur Gescli. des lat. Alphabets/' in the '^Rhein. Mus. fiir Philol./' 1869 ; or the first chapter of Corssen's " Aiissprache." 202 APPENDIX D. not necessarily be any the more probable. For,, as was just now said (paragraph (/) above) , we cannot be sure that there were not two varieties of the Spirant at any given epoch, because there was, at the same epoch, but one sign, any more than we can assert that there were two varieties at a later period because there were then two signs. Let us for a moment suppose (what cannot be disproved) that both varie- ties of the Dental Spirant were in existence before the appro- priation to that sound of the older sign; then, if only one sign was to be constructed for both varieties, aud if that sign had to be based either on t or on d, the chances are clearly as much in favour of the selection of one of these letters as of the selection of the other; that is, it is just as likely that the Soft Spirant should have led to the selection of the Media, as that the Hard Spirant should have led to the selection of the Tenuis ; so that the actual choice of d would, on this sup- position, by no means involve the non-existence or juniority of the Hard Spirant. I believe, however, and indeed shall presently urge, both that there was, to a comparatively late period, only a single Dental Spirant, and also that the two signs (I? and ^) originally represented that single sound. Further, I should even be glad if the probabilities in favour of a con- nexion between J? and D were much stronger than they are ; for I should then claim that connexion both as supporting and as being explained by the provisions of my own theory. But I should refer the connexion, not to a phonetic, but to a dialectic cause, and should look for its explanation in a state of things to which I have proposed to trace many other phenomena, viz., simple dialectic mixture. Where some of the L.G. tribes and sections of tribes used the correctly verschoben Spirant, others, in contact and even commingled with them, employed the Media (§§ 35 (A), 49 [e) note, et alibi), accompanying in this point the H.G. tribes, with which also many of them were, by our hypothesis, in con- tact (§ 53). Under these conditions an intimate relationship and approximate equiv^alence of the two sounds might in- stinctively have been taken for granted, and efforts might have been made to denote this relationship by constructing a symbol for the Spirant which should closely resemble the symbol for the Media. But any connexion between the MR. sweet's scheme OF THE ^'^ LAUTVERSCHIEBUNg/' 203 symbols is so doubtful that I feel precluded from basing any conclusion thereupon. If I were compelled to find an imme- diate parent for the symbol ]>, I would rather turn to the Gothic symbol (\jj which, when made swiftly by hand^ would easily pass into the former by the gradual extension of one side of the curve at the expense of the other. {n) The younger symbol (^) may now be dismissed in few words. Its value as a piece of evidence depends entirely on the supposition that it was independently constructed out of d. Even then its value is small ; for in its construction the precedent set by the older sign would naturally be fol- lowed. But I have little hesitation in totally rejecting that supposition. A life-long familiarity with the vagaries of handwriting convinces me (so far as the case allows of con- viction) that the younger sign is not an independent creation at all, but simply a graphic variety of the older one, result- ing from the gradual accumulation (perhaps the work of centuries) of minute diflPerences of formation, of which the more marked stages may be represented thus : — jff p ^ i^ '^ From this point of view the resemblance between c? and ^ becomes a curious accident. It is manifest, howevei% that while d (A-S. b) does not, ]> does, account for both the parts of 'S. The cross line, in particular, which is such a striking feature of the latter, receives a perfectly rational explanation, instead of seeming to be a purely arbitrary addition^. Assu- ming, then, as we now may, that the sound, after the evolution of the younger symbol, was still felt to be a single one, we need not think it more remarkable that one old scribe should ^ The extremes of this series differ much less than many other pairs and sets of symbols in daily use among ourselves, the members of which nevertheless are certainly derived one from another. Compare C and ^, ^(^, T, 1 and V? O^ and <^, ^ and /, &c. 5 compare also the quaint but elegant forms of the modern German script with their ancient prototypes. ^ Compare the rudimentary cross stroke of the ordinary script Jp with the perpendicular stroke of 4, which it represents. 204 APPENDIX D. always write )? and another ^, than that one of our own corre- spondents should always write /^ Z> Jd, and another ^r^ dd, and the like; nor yet more remarkable that in other old MSS. \ and 'S are written indiscriminately, than that now-a- days Z and T, /and /. &c. are^ as often as not, used indiscri- minately by one and the same writer'. What is remarkable is the way in which the language of signs instinctively follows in its methods the language o£ sounds. For just as (§§17, 28-30, and Appendix C) varieties of sound pair off with varieties of meaning, so (only more rarely) do varieties of symbol pair off with varieties of sound. Thus the letters / and J, as well as V and V, were, till quite recently, as Mr. Sweet reminds us, used (like the old Latin C and G) without distinction of value ; and to each member of each of these pairs has been assigned a separate function almost in our own day. Precisely the same seems to have been the case with \ and "S. Used impartially and indiscriminately as long as the Dental Spirant was, or was felt to be, one, they gradually, after the younger sound was clearly marked off from the older, became assigned, the former generally to the Hard Spirant and the initial position, and the latter generally to the Soft Spirant and the interior position. ^ Mr. Helfenstein (C. G., p. 126) seems to think the practice of the old scribes in this respect very reprehensible, and still more so that of those editors of their MSS. who imitate their "lawless" course. It is to be hoped, however, that editors will continue to be guilty of the same crime, and not virtually assume that the old copyists did not know what they were about. As to the point of phonetics discussed in this Appendix, it is manifest that such imaginary " lawlessness " is just what would give to any MS. its value as a piece of evidence ; for it would mean, if it meant any- thing, that the distinction between the Soft and Hard Spirant either had not yet arisen or was not yet perceived, and also that there was attributed to the two symbols precisely the same phonetic value. The Teutonic scholars whose uniform distribution of > to the initial and % to the interior position (in conformity with the later state of our sounds) accords with Mr. Helfenstein's views, would have been, I imagine, tliv6-. In the next place Grk and Skt, though by this time long sepa- rated, proceeded, we are told, to deal with the root in precisely the same way ; i. e., they both deaspirated the initial, making in the one case ttvO-, and in the other bud/t- ; whQe the Latin proceeded to deaspirate the final, and to thin down the initial into a spirant. Three doubtful points, therefore, present them- selves at the outset : — ^first, the apparent disregard, or rather reversal, and that often repeated, of the Principle of Least Effort {bb and dh becoming and 6 in Grk, ^ becoming tt, and bh in Skt becoming b) ; second, the multiplicity of forms supposed to have been assumed by the primitive Aspirate ; and third, the accidental coincidence of action, on the one hand, v0-) ^. But the Grk and Skt sub- * I am here looking at this diLster of roots in the light in which Grassmann himself places them. In the note to § 49 («) I have inten- tionally tried to show, with reference to the root bhuff- (4>^fY-), which is one of his examples, that a widely diB&ront explanation may, in that and perhaps other instance, be applicable. ^ Unless it is forbidden to apply to European langoages modes of reasoning which are Intimate as applied to the langoages of Pdynesia. See the method by which Max Miiller (Lectures, iL 138) restores Sacaiki as an older form of Hawaii. * I should fadl back upon Grassmann's reiterated deiiTation of ihe pri- mitive Hard Aspirate from the Tenuis (See Appendix B) as a Taluable support to the mode of generation assumed in the pres^it tzeatiae, were it not that he makes that rery Afpiiate the parent in turn of the Tenuis in p2 212 APPENDIX F. dialects, at any rate,, did not ; and whatever may have been the cause of their objection to bi- aspirate roots, we must, by our hypothesis, suppose (what, in fact, is more likely) that it operated to prevent the formation of such roots, and not (what is less likely) to alter them after they had been in use for an unlimited time ; that is to say, those subdialects (in accordance with the general principle laid down in §§ 48,49) maintained a Tenuis which should have been debilitated to an Aspirate. This Tenuis therefore remained identical with a Tenuis in S^, and (what was ultimately the same thing) cor- responded to a Media in 2 (Scheme (\/r) of § 49 (e)) . The resulting root {puth-, irvO-) must then be supposed to have remained permanent in Greek [iTvd-firjv) , while in Sanskrit the Aspirate, which was originally imperfect or was subse- quently softened (§§ 33, 34, 45, wo^e), ultimately, by assimi- lation, dragged down the initial Tenuis to a Media ; in which movement it partially affected the still commingled Greek {cf. 0v6-fi6<; and ^vcr-ao^j i.e., 0vd-jo<;), so that both acquired, as regards the initial, a resemblance to Low German. (c) The Latin subdialect, indeed, seems to have striven in its imperfect way to differentiate the root completely ; and this fact, so long as it was assumed to be incontrovertible that aU the CI. dialects varied their roots from a common parent form, was an argument in favour of an original two- fold aspiration on the part of the other dialects. For on such an assumption the initial / in /undus necessarily implied an Aspirate in the once common form, while the internal d was quite consistent with the existence of the second Aspirate actually exhibited by Greek and Sanskrit. But it is provided by our hypothesis that the subdialects of a leading dialect, although agreeing in the main, may not in every case have carried out their differentiation to the same degree of per- fection. We may suppose therefore that the Latin subdialect endeavoured to accompany the other CI. subdialects in aspi- German, Lithuanian, Slavonic, and Celtic. The CI. Soft Aspirate, too, is supposed to beget both the Greek Hard Aspirate and also the L.G. Media, which latter again is supposed to beget the H.G. Tenuis. Thus we are here (as in the Schemes of Appendix D) in a whirligig, where everything becomes directly the parent of everything else, and indirectly of itself too. rating {more suo) the primitive internal Tenuis ; but that the resulting pseudo-aspirate speedily assumed the Medial form^. As the resulting root pud-j however_, would still differ but slightly from the H.G. put- {pot-) and from the L.G. bud- {bod-), the Latin may further be supposed to have made an independent effort to dissimilate its initial Tenuis in con- formity with the normal relationship between 2_^, 2 , and 2^. Thus we arrive at the following dialectic series :: — H.G. . . [put-'], pot- ; L-G [put-, pot-], bod-; Greek . . [put-] , irvd- and I3v6- ; Sanskrit., [put-], puth-, pudh-, budh- Latin .... [put-, puth- ?] , pud-, phud-, fyd--^ia:&fund,- ; ^ - . ^ ^' f J where it will be seen that the history of the root in. each of.-' the CI. dialects is now in strict harmony with the Principle of Least Effort. Under the H.G. its ideally correct form only is entered. The correct L.G. form was actually preserved by the Old Saxon, which _, like the Latin,, weakened both consonants ; while the Old Norse and Anglo-Saxon, like the Greek, weakened only one. We are not, however, to suppose (what, indeed, is contrary to fact) that the O.N. and A-S. objected to roots of the form {s + vowel + s|, and that therefore, after evolving such roots, they ^^ raised '' one S to H. At this point the hypothesis of the foregoing treatise steps in with absolute impartiality; for just as it requires that the initial tt of irvO- (instead of being supposed to pass through several changes with the absurd result of finding itself at last just where it was at first) should be considered as the primeval Tenuis preserved from any change at all, so it requires that in the more prevalent L.G. form bot- the final t should be similarly considered as another such un- changed primeval Tenuis. But after all there is considerable dialectic irregularity in this example^ : where the L.G. dia- 1 In conformity with the conjecture of Appendix G, the Media in the Latin root might be a direct debilitation from t; but until that conjecture has been discussed I leave the above paragraph as originally written. 2 A more regular example (so far as it goes) is furnished by the root pdk-, "arm," "branch." This also was retained by the H.G., and m actually extant in the forms puac, puoc. The corresponding CI. root 214 APPENDIX G. lects go astray (if they really do so), it is in quite a different direction from the Greek and Sanskrit. (/) The force instigating to the violent changes assumed by Grassmann in the CI. forms is supposed by him to have been euphony simply. But the hypothesis just substituted involves the absence of change. Hence the agency concerned must have been preventive rather than effective. Possibly the ultimate cause is to be found in our oft-repeated proposition, that so-called " substitution^^ was due to the contemporaneous and reciprocal action of commingled dialects. To say that dialects were commingled means_, in other words_, that the people speaking them remained in constant contact and com- munication. Hence it would manifestly be essential for one section, while yielding in general to the principle of Dissimi- lation, to resist it whenever there arose a necessity for pre- serving words from being deformed past recognition by other sections. We have (or fancied we had) already discovered the operation of a similar counteracting necessity in a less com- plex case of sound-change (§ 39 (c)), and that the necessity should have operated in a more complex case seems even more likely. We do not, indeed, find among the ancient dialects any striking objection to roots of the form { S + vowel + S } ; but from all that we have conjectured respecting the earliest forms of the Aspirate, that sound may have seemed to the CI. tribes to be a wider departure than the Media from the parent Tenuis ; so that an instinctive objection to acquire roots of the form j A + vowel-}-A j is, to say the least, not impossible. G. On the interior Me dice for Aspirates in Latin. [a) With reference to the subject of § 35, further thought emboldens me to hazard a conjecture which, when writing should have been of the form phdkh- {<^ax-) ; but no trace of the initial Aspirate is discoverable (Skt bdhu — Zend hdzu = Gk nfjxv- : the Latin is unfortunately wanting). The L.G. dialects preserve their correct forms : O.N. b6(/r, A-S. buff (our bough and elbotc), to which the M. and N.H.G. forms (buoff- and buff-) are assimilated (§ 63). ON THE INTERIOR MEDIAE FOR ASPIRATiE IN LATIN. 215 that section^ I was influenced by some opinions of previous writers to suppress. It depends immediately upon the main theory of this treatise,, especially §§ 24-26^ and amounts, in sum, to this — that previously to the spread of Aspiration through the CI. dialects (§ 45) the Latin branch of the Ita- lian subdialect might, independently of the other branches, have already so far yielded to the powerful influence of bivo- calism (p. 72, note 2), or of conjoined liquids and nasals, as to have reduced to Mediae some of those interior Tenues which in cognate CI. dialects were still preserved as such, but which afterwards fell victims there to the new infection of Aspira- tion. Such softening has, it is true, been assigned (§ 35) to the H.G. dialect as its special characteristic, when viewed in relation to the symmetrical distribution of the mutes among the three great dialects; but it would be absurd to assert that so natural and universal a species of debilitation could not previously have prevailed sporadically, and perhaps over considerable areas outside those in which it was evoked by Dissimilation. In accordance with § 46 [a) any such inte- rior Mediae as our conjecture postulates would necessarily be protected from subsequent Aspiration ; and, as to their number, there needed to be only just so many of them as might suffice to set the fashion, so to say, of correspondence with the interior Aspirates (or pseudo-aspirates, § 33) of the surrounding or commingled dialects and subdialects. Other interior Tenues would, by analogy (§21 ad fin,), generally adjust themselves to that fashion ; and yet an exceptional instance or two would still be possible, in which, from special causes, an internal Tenuis might be drawn into the powerful Aspirating movement going on all around. Hence the h in mihi, trahoj veho. {b) It will, of course, be at once urged that any such ten- dency to softening ought not to have confined itself to those interior Tenues only which would otherwise have become Aspirates, but that it ought also to have affected some of those interior Tenues which should have remained Tenues. Pro- bably it did so. At any rate, numerous individual examples of that form of debilitation are to be found in extant Latin. Thus (1) ^ occurs for interior k in viginti, triginta, negotium, Saguntum, patina, vagire, mugire, and even initially in guber^ 216- APPENDIX G. nator, ^ummi, gloria^, and others; (2) d for interior t occurs only in the neighbourhood of liquids and nasals, as in quad- ratuSj mendiax, &c., or in terminations like -idiUS', but as a final in quod., id, illudij &c. ; also in the whole series of archaic ablatives (gnaivodi, sentential, &c.) and archaic imperatives (estod, facitod, &c.) ; in all of which the primitive t was once probably bivocalized (Appendix EJ ^ ; {S) b for p, both inte- rior and initial, occurs in ah (primitive apa^airo), oh (prim. api = iirl), suh [s + prim, upa = vtto), hiho, huxus, ^urrus (in Ennius, for Vyrrhus), and others. A considerable number of instances of softening in. proximity to liquids and nasals are omitted, although they really bear as forcibly on the point in debate as those that are quoted^. (c) It will further be urged that, if our conjecture were true, the interior b, which in Latin often represents the CI, Dental Aspirate, ought to have originally appeared as d. But here, again, facts are so far in our favour, that several words belonging to the same series as the words in which that b occurs, actually do, and for aught we know always did, in Latin, as such, exhibit an interior d; as mediuSj cedes, arduus, fundus, fides, vidua, abdere, credere, and others. If, however, all the interior Mediae answ^ering to Dental Aspirates were once c?'s, by what agency, it may be asked, could their trans- mutation into 6's have been effected ? It will perhaps pre- sently appear that even this difficulty is far from insuperable. I must first remark that, supposing the CI. Aspirates to have been of the nature generally assigned to them (§ 12 and Appendix B), the difficulty just stated seems to me but microscopically greater (if greater at all) than that of getting from dh or th [ = 6)iQ bhov ph (= ^) ; through which transi- tion the interior b is, by many leading philologists (in con- formity with Curtius's Theory of Aspiration (§ 12)), supposed to have been reached. The difference between the dental and labial seems to be, in each case, too wide and too sharply ^ These and similar examples serve for an a fortiori argument. ^ For cases of initial d=^t I must refer the reader to Corssen, " Beitrage " &c., pp. 83 et seqq. '^ For other examples of the way in which the tendency to softening continued to manifest itself throughout the whole history of the Latin tongue, see Corssen, " Aussprache/' &c., pp. 77 et seqq. ON THE INTERIOR MEDI^ FOR ASPIRATjE IN LATIN. 217 marked to allow of an immediate passage from tlie former to the latter by way of spontaneous evolution ^ {d) Leaving others, however, to bridge over the gap be- tween dh and bh as they best can, I propose to remove our suggested change from d to b altogether out of the range of spontaneous evolution. As in the older stage of the movement {a J above) , so here, I would fall back on the action of Analo- gical Adjustment evoked by the external pressure of dialects in contact (§21 and § 60 (t/)). For the majority, probably the vast majority, of the Old Italians reduced their variety of the Dental Aspirate, in the interior as well as the- initial position, at first, no doubt, to a series of Dental Spirants ( = }>), and then to a series of /^s in no way distinguishable from the /'s derived from the Labial Aspirate. If we sup- pose this final phonetic stage to have been reached before the Italians loosened out into their subsequent divisions, the members of the rudimentary or potential Latin division^ ^ Corssen, Baudry, and others suppose the CI. Labial Asr^vcoie, among all the Italians, to have been first reduced to /in every position, and then, in Latin, closed up to b within words. Their views as to the Dental and Guttural Aspirates are not easy to make out. Ascoli, in a series of vigor- ous articles (Kuhn's Z-S., xvii. and xviii.), lays down a scheme of evolu- tion that certainly has the merit of consistency, although the consistency appertains to the most doubtful point ; for he would reduce all the Aspi- rates to Spirants in all positions, and the X^ental Spirant (J>) in some cases to/, and then evolve from them the internal Mediae of Latin. Now there can be no doubt that the initial Spirants of Latin, and the initial and interior Spirants of the other Italian dialects, descended immediately from the Italian variety of the 01. Aspirate (§ 33, supra) ; also, that where / represents the Dental Aspirate it must have passed through the stage represented by \> ; for these two sounds are formed in close proximity, and the actual transition from > to / is demonstrably frequent. But for the closing up of h, J>, and /into (a change common among lispers in this country), and then that \> became / as before. 218 APPENDIX G. among whom the interior d still maintained its place, would gradually come to experience an uneasy sensation — to ac- quire a more or less dim perception — that their own internal Medise were not in continuous harmony with the single and uniform series of /'s of the bulk of their fellow-Italians. Where, indeed, the latter said tiie, ife, sife, puie, Safinim, alfa, AlBus, or the like, they themselves said tihei, ihei, sihei, quohei, Sahiniom, alba, Alhius, &c. &c. ; and thus, in a large number of instances, perpetuated the correct and regular cor- respondence originally established ; but in a smaller number of cases, as where the latter now said meiia, ruiro, uier, veriim, pruia-iarrij -fas, -fat, -ians^ or the like, (for the older me^ia, ru\}ro, &c.,) they themselves continued to say media, rudro, uder, verdum, proba-dam, -das, -dat, -danP, &c. &c. ; and here, although they were really in the right, (a fact, however, for many long ages forgotten,) a feeling to the opposite effect would be evoked by the preponderating influence of the dia- lects in contact. It is therefore just what might be expected, that they should proceed to adjust the latter series of ex- amples to the pattern set by the former, and thus endeavour to establish that continuous phonetic harmony which they felt to be wanting : that is to say, these internal d's would, at first occasionally and irregularly, and then permanently, be in this way transformed one by one directly to b's^. But if we sup- pose that, before the permanent transformation of the whole series of d^s was completed, any sufficient cause, such as the loosening out of the Italians into detached or almost detached divisions, arrested (as it certainly would) the influence of the other dialects upon the Latin, then we ought to find a part of ^ These forms are mostly conjectural, except as regards the/. * I adopt the keen-eyed view of Scherer (" Ziir Gesch." &c., p. 202), which connects the Lat. imperfect suffix with the CI. root d?ia = lj.G. da = our " do ", and thus links on the Latin (as regards the evolution of this grammatical form) to the German and Lithuanian dialects. (See Schlei- cher, Comp., p. 795.) 3 It would be quite consistent with the above hypothesis that the transmutation of the Latin c?'s to 6's should have followed close upon, and have advanced pari passu (so far as it did advance) with, the descent of the non-Latin t>'s to /'s. In this case the extiint Latin d's coincide with a part of the series of non- Latin Ks, which had not yet descended to /'s when the dialects separated. ON THE INTERIOR MEDI^ FOR ASPTRAT.E IN LATIN. 219 that series still remaining as d^s ; and this is exactly what we do find. (See the examples in c, above.) {e) I shall not now stay to elaborate the foregoing outline of a hypothesis. It will be obvious^ however, that our funda- mental conjecture {a, above) may easily be shown to have an important bearing in many other directions than that in which it has been here applied — as_, for example, upon the origin of similar Mediae in Greek, in Li- SI., and in various L.G. dialects. In reference to the latter two especially it might suggest some modifications of, or additions to, the suggestions thrown out in §§ 46, 49 ; and I regret that advantage was not there taken of it. Phonetically, its claims to consideration are : — (1) that it reduces the total amount of sound-change to a minimum; (2) that all the changes are brought under the Principle of Least Effort; and (3) that, where the line of phonetic descent is oblique instead of direct {i. e., from d to b), it points out an agency potent enough to wrench the line aside. 220 ADDENDA ET CORRIGENDA. Page 11, line 11. Of the foui* symbols H, S. — A, A', tlie first three are wanting in some copies, having dropped out at press. 20, note, add: '' So also are our numerous constructions with of (id., pp. 481-2)." 26-28. The general fact here established was noticed by Lord Bacon 250 years ago: — ''Linguae quae ex Gothis fluxere aspiratis gaudent " (De Aug. Sci. vi. 1) ; but, of course, his " Goths " do not include the Mceso-Goths. (I am indebted for this interesting " reference to Dr. Farrar's " Families of Speech.") 27, line 5 of note 2. After " woidd" imderstand : " on the Historical Hypothesis." 40, last line but 6, and page 48, line 9. The statement as to the exact equivalence of the interchanged sounds should be so far qualified as to admit of subsequent natiu-al debilitation in such sounds after the separation of one dialect from the other. (See Appendix D (e).) 41, line 12, for H.G. read L.G. 47, line 12. With reference to the '' fact of some interest " a friend remarks, " I should think everybody has noticed this." 51, note 5. It had escaped me for a moment that Max Miiller himself, in another place (Lectures, ii. 137), demonstrates very forcibly the present corrupted condition of modern Hawaian. 66, line 14 of 7iote. So, too, thinks M. Hovelacque in his recently published work " La Linguistique " (Introd.) : — " Nous ne chercherons pas a ^viter Texamen de la question de I'origine du langage L '^carter sous pretext equ'il faut proscrire toute recherche des ' origines premieres,' c'est admettre la possibility meme de ces causes premieres, dont les math^matiques et la chimie ont fait justice." But I do not see how H.'s transference of the question to anthropology (or rather anthropoidology) can possibly yield any satisfactory results. 68, line 10. The " trebling " of expressive capacity, here spoken of, must be referred to monoliteral roots, or, rather, roots originally involving a single tenuis (e. g., kar on p. 65). In the case of roots which were originally of the form {tenuis -^vowel-^tenuis^ the possible increase of capacity would manifestly be ninefold ; for each variety of initial (tenuis, media, and aspii'ata) might bo combined with each variety of final. 61, line 2 of formula (^), the last D should be D'. 64, line 1 of § 31, S is wanting in some copies. ADDENDA ET CORRIGENDA. 221 Page 71, line 14, "rightly enough," &c. I mean (as § 33 showed) that Hard Aspirates may have existed in one dialect of the Holethnos and Soft Aspirates in another. 72, § 35 ', and pages 95, 96. In reference to the questions here treated, Appendix G should be read. 73, note. A short but valuable illustration of the subject-matter of this note has recently appeared (where philological information would hardly be looked for) in the Blue-book for 1875-6, issued by the Committee of Council on Education. A paragraph in Mr. Rhys's Report on the Schools of Denbighshire and Flintshire gives some interesting examples of the contortions that English sounds undergo in Welsh mouths. See also an excellent article on that Report in the "Saturday Review" of Sept. 16, 1876. 75, note 2. The internal h of mihi should have been spoken of as " solitary " among the Latin representatives of the Labial Aspi- rate only, a qualification which was inadvertently omitted. Internal h represents a Guttural Aspirate in '* trahere " and "veherc." (See Appendix G {a)). In line 4 of same note, for " Curtius," substitute : " Schleicher, Comp., p. 243." 78, with note, and page 92. So M. Hovelacque : — " Xous pouvons penser, sans crainte d'erreur, qu'avant leurs migrations, ces populations occupaient un territoire assez vaste. En ces larges limites la langue commune indo-europeenne ne devait-elle point se modifier, s'alterer, se corrompre de fa9on diiferente dans les diiferentes tribus etablies sur ce territoire ? Nous pensons qu'il n'en pouvait etre autrement." — " La Linguistique," p. 339. See also the excellent fourth and fifth Lectures of Prof. Whitney. 80, subsection (&). The general similarity of linguistic conditions would lead us to expect that the interchange here treated of might be discovered, at least on a small scale, in the popular dialect of Normandy. Accordingly, " Le v devient w assez . frequemment: 'je m'en ?^ais' (rais) 'Cha?yois' (nom propre, Chaurois)," &c. See " Histoire et Glossaire du Normand," par Ed. Le Hericher, vol. i. p. 32. Where German words that once began with lu are preserved, the w has generally become v, as vatre, ■vinche, = w?ater, zmnch. (Id., iii. pp. 51, 52 of Appendix.) I have not been able to discover any trace of such interchange in the Norman-French transplanted to this country. 97, note. In line 2, read " Sprachemheit ;" and in line 9, /or Salv. read Slav. 106, note, bottom line, after -ths, add : " which, however, is not really an exception (See Schleicher, Comp., p. 316)." 114, line 6, after " reduplication," add: " on the pattern of pa-^aver, cu-ciimis, cin-cinnus, su-surrus, &c." ; line 7, " Aedic " is a mis- print for " ^.olic." 118, par. (v.), line 13, after "feder," add : "and on the pattern of our tDiggle-waggle, tittle-tattle, riff-raf, &c." 222 ADDENDA ET CORRIGENDA. Page 121, line 10, for bhaksjam read bhak.s7) examined, 194-202; ori- gin of the symbol, 203, 204. Ego (eyw), &c., original form and later varieties of, 66, 100, 167, 207, 208. Ellis, Mr. A. J., on the sounds of H.G-., 74, 75. 'Ejue, 208. English, disintegration of, 15, 16, 79, 80 ; mixture in grammar and vocabulary of, 19-21 ; sounds of, as pronounced by Grermans, 75 ; aiRnity of r in, for a{=ah), 176-180; examples of pho- netic variation in, 186, 187 ; clumsy symbol for Dental Spirants in, 194. Ennius quoted, 216. Ems, 208. European dialects, the affected gutturals of the, 135-169 ; origin of the labial affection of k in, 153-155, 172. Exceptions to Grimm's Law, Lottner's treatment of, 24, 99 ; Grassmann's, 99, 209-214 ; Forstemann's, 99, 101 ; dis- tribution of, among all the Mute-sys- tems concerned, 102 ; classification of, 102-104 ; reduction of, under a single general principle, 104-108. F, origin of sound of, 11, 68, 182 ; repre- sents all the aspirates in Latin, 69, 70, 217 ; agreement of the H.Gr. with the L.G-. /, 32, 125, 126, 193 ; an older and stronger sound than the related V, 197, 198. Fick, Dr. A., quoted, 9, 10, 27, 64, 66, 97, 100, 115, 137, 140, 142, 155, 163-170, 174, 205, 207 ; on,the multiplication of roots and words by phonetic vari- ation, 66, 67, 186 ; his hypothesis respecting the primitive Jc, 138-140 ; objections to that hypothesis, 140- 144 ; his views on the evolution of i and u from a, 170; on " radical" w, 174 ; his researches into the primordia of speech, 205. Finns, supposed contact of the Old Ger- mans with the, 17-21 ; conjectural similarity of their language to the primitive I-E., 56. ^idocpBeBeXd, 118. Foreiga sounds, tendency to alter the value of, 19, 62, 69, 73, 91, 113, 151, 153, 155, 173, 183, 221 ; difficulty of appraising the value of, 184. Forstemann, Dr. E., quoted, 17, 26, 27, 33, 56, 78, 97, 99, 101, 106, 111, 134, 176, 186 ; his form of the " contact" hypothesis, 17-21 ; his statistics of dis- tribution of I-E. Mutes, 27 ; his treat- ment of the Exceptions to Grimm's Law, 101 ; his chronology of the sup- posed Verschiebungen, 111 ; on pho- netic variation, 186. French, effects of bivocalism in, 72 ; treatment of the sound of qu in, 146, 147 ; artifices of, for varying meaning, 187 ; interchange of v and h in, 193 ; common origin of pronouns and articles in, 208. " Function," application of the term, to the relationship between the principal I-E. Mute-systems, 4-6 ; that relation- ship not due to accident, 21 ; but to the way in which those systems were evolved from a simpler system, 53-55, 77 ; commixture of dialects necessary to establishment of such relationship, 84, 88, 89,127, ef passim. Fund-ere, 210-213. G, the symbol, derived from o, 201 ; g for k in Latin, 215. G, the determinative, a byform of k, 66 ; accretion of, before w in Romance lan- guages, 146, 147 ; labial and sibilant affections of the Cl. g, 161-167. Gan- (= yev- &c.) and gan- { = gna-^ yvM-, &c.), 166. Gar- {See Kar-). Genealogy of the I-E. languages, 137. Geographical distribution, conjectural, of the leading tribes of the Holethnos, 96, 157. Tepyepos, 163. German, Germans: form of H. and L.G. mute-systems, 3-5 ; characteristics of the German dialects, 15, 16, 19-21 ; overwhelming share of Aspiration in L.G., 26-28, 68 ; agreement of H. and L.G. in the use of h and /, 30, 32, 125, 126, 131, 193 ; German mutes not more fluctuating than tl^ose of other I-E. Q 226 INDEX. tribes, 35, 43, 50, 98, 191, 192 ; origin of the aspirates in L.G., 68 ; and of the medi« in H.G-,, 76 ; H.Gr. sounds as pronounced by Eaglishmen, 74, 75 ; influence of the H.Q-. on the L.G-. mute-system, 76, 105-107 ; and of the L.Q. ou tne H.(>. sj-stem, 130 ; anti- quity of the two systems, 110-122 ; the H.G-. system once completely evolved, 123-125 ; subsequent assimilation of the H.G. to the L.G. system, 126, 127 ; relative stratification of H. and L.G., 127-132 ; probable reason for the sud- den appearance of H.G., 129; com- mixture of H. and L. Germans, and its linguistic effects, 130-132 ; pro- bable passage of ^ to A and of 6 to /in H.G., 131, 193; the force of the O.H.G. tenues (= L.G. and N.H.G. media;) Btill preserved in pronunciation, 132 : dh for d and th for t in H.G., 199. Gh, the aspirate {^8ee Kh). Criv- (= kwik-, gwiw-, &c.), phenomena exhibited by the root, 164-166. Gothic, purity of the vowels in, 16 ; dis- tribution of mutes in, 27 ; relation of, to the so-called H.G., 128 ; disappear- ance of, 129 ; the dental spirant in, 194, 203 ; passage of s to ^' in, 198. Grabh-, the root, 105, 110. Graeco-Italian, phonetic system of, the system of greatest resistance, 159. {See also Greek.) Grammatical structure, effects of lin- guistic mixture upon, 19. Graphic yariation, examples of, 201- 204. Grassmann quoted, 14, 22, 24, 93, 99, 100, 149, 182 ; his attempt to explain supposed sound - changes by moral qualities, 14 ; his treatment of some Exceptions to Grimm's Law examined, 209-214. Greek, aspirates of, 23, 70, 72, 93 ; distri- bution of mutes in, 27; transliteration of Latin words in, 113, 114; trans- position in, 121 ; labialism of k in, 155; fondness of, for reduplication, 166 ; treatment of w and y in, 170. Greeks, propensity of, to Grtecize foreign words, 113, 117, 118. Grimm quoted, 9, 12, 13, 22, 111, 122, 124, 132, 181, 207 ; his connexion of supposed sound-strengthening with moral qualities of the Germans, 13 ; his chronology of the supposed Ver- schiehungen, 111, 121-134 ; his inter- pretation of the Dacian words in Dioscorides examined, 112-121 ; his account of aspiration, 181. {See also Grimm's Law.) Grimm's Law, importance of, 1 ; satis- factory theory of, still required, 2 ; formulation of, 2-6 ; Mute-systems principally involved in, 3, and their symmetrical relationship, 4, 5 ; histo- rical or chronological hypothesis of, 6, 7 ; various forms of that hypothesis examined, 7-48 ( and tS in A-S. and modern Eng- hsh, J89, 194, 200 ; for t in H.G., 199. Theories, test of the value of, 25. Thorn {\>, th), origin of sound of, 11, 68 ; ^ Mr. Sweet's hypothesis respecting, ex- amined, 189-204 ; the sound stronger than that of edh («), 195-199 ; the symbols probably not derived from D, 200-204. {See also Dental Spirants.) Todhunter, Mr. I., quoted, 4. ToyX/3?;\a, 117. Touratrrpa, 118. Transition, the Principle of, stated and applied to the German v (=/), 131 ; to the evolution of the affected it- sounds, 153 ; to the history of the aspirates, 182 ; and to the evolution of the dental spirants, 189. Trench, Archbishop, referred to, 188. JJ, the vowel, derived from a, 9, 170 ; its fluctuating value in Europe, 153, 222. Ulhandus, 114. Umbrian, labialism of s in, 217. Uniformity of Nature, importance of belief in, 15, 33, 34, 67, 124 ; applica- tion of the doctrine to the history of the affections of k, 141, and of the aspirates, 182. Unterscheidungstrieb, Curtius's attempt to explain substitution by, 24 ; special defects of the explanation, 30, 32 ; extraordinary action of the supposed process, 32-35 ; difi'erence between it and the genuine dissimilating senti- ment, 48. Fin H.G. with the power of / expla- nation of the phenomenon, 131, 197 ; derivation of v from w, 146, and from / 197, 198. Fand b, interchange of, in French, 193. Fand to, interchange of, in Cockney dialect, 39, 40, 79-84, 134, — and in dialect of Normandy, 221. Variation ( See Phonetic Variation and Graphic Variation). Veh- {=Pox- &c.), 169, 215, 221. Verschmelzung, Verwisckung, linguistic processes imagined by Dr. Fick, 140. Vigfusson, Mr. G., referred to, 200. INDEX. 231 Viv-ere {See Giv-). Vocabulary, effects of linguistic mixture upon, 21. Vocalic affections of a, 170 ; their chro- nology in relation to that of the similar affections of k, 172. Vowels, purity of, among the oldest Grermans, 16, 17 ; derangement of, in English, 16, 194; "prosthetic" and " connecting," possible origin of some, 208, 209. {See also A, I, and U.) W, treatment of, in France, 146 ; the sound of, not primeval, 169, 170 ; affinity of, in English, for the faucal a, 178. {See also V.) Wa the immediate parent of u, 170. Wales, North, treatment of English sounds in, 221, 222. Weakening, phonetic ( -See Least Effort). Webster, an error of, 178. Whitney, Prof., referred to, 221 ; inves- tigation into English sounds bv, 176. Whitney (Jolly) quoted, 2, 8, 22, 25, 122, 129, 186. '• Word," the term, hovr used, 60. Y, the sound of, not primeval, 169, 170. Ya the immediate parent of i, 170. Z, the sound of, younger than that of s, 197, 198. Zeitsehrift, Kuhn's, quoted or referred to, 14, 22, 24, 26, 28, 32, 35, 71, 93, 99, 113, 114, 149, 182, 209, 217. Zend, aspirates in, how represented, 69 ; treatment of w in, 173. ZouooTiy, 115. THE EI^D. PRINTED BY TAYIOE AND FRANCIS, RED LION COURT, FLEET STREET. !i' |B„15l»Hb»< LIBRARY BERKELEV ^''elast date stamped below '^'^ 20 ,948 300rt5lBH l60cV51LU, KU D LD 22%[2W UliYl'^'GB-SPM V m mW MkWtTwimJB^ ^n m ^^ J^^ Wm jOjTjl ^T / T f^' ^HMJl m C3f^8 mi'' 4ii ^Msn 7^ ' W^^^v^lij r \wnl^^ iJmK kIu Iwl r V ^^^^^^^^^ mi ^&!j