UNIVERSITY OF CALIFORNIA SAN DIEGO 3 1822 00808 1234 N0i ■■ >* ; rv"i >K *n W 3 UNIVERSITY OF CALIFORNIA SAN DIEGC 3 1822 00808 1234 2 H. h> /«. 7 U*CHZ ^LL^'il Cc/v^Xa C ft^lt-L*^?; 1 ' rA 1 gale bicentennial publications ON PRINCIPLES AND METHODS IN LATIN SYNTAX gale ^Bicentennial publications With the approval of the President and Fellows of Tale University, a series of volumes has been prepared by a number of the Professors and In- structors, to be issued in connection with the Bicentennial Anniversary, as a partial indica- tion of the character of the studies in which the University teachers are engaged. This series of volumes is respectfully dedicated to ST&e 45raDuate0 of ttje zamtaraitp ON PRINCIPLES AND METHODS IN LATIN SYNTAX BY E. P. MORRIS Professor of Latin in Yale University NEW YORK: CHARLES SCRIBNER'S SONS LONDON: EDWARD ARNOLD 1901 Copyright, 1901, By Yale University Published, September, iqoi UNIVERSITY PRESS • JOHN WILSON AND SON • CAMBRIDGE, U.S.A. PREFACE Two things only need to be said by way of preface to this book. In the first place, it makes no claim to the character of a systematic statement of the principles of syntax. It is a discussion of certain principles and of the methods of investigation to which these principles lead. In the second place, it deals primarily with Latin. If some of the chapters are equally applicable to the syntax of other languages, that is only because it is impossible to write of the fundamental questions of syntactical method without going beyond the phe- nomena of a single language. The illustrations are all from Latin, and nearly all from Plautus, many of them being taken, as are portions of some chapters, from articles of mine in the American Journal of Philology, to which reference is made at the proper place. My obligations to other writers on syntax and linguis- tics are indicated occasionally in the notes, but the character of the book does not call for a bibliography. I cannot refrain from expressing my regret that the second part of the first volume of Wundt's Volker- psychologie appeared too late for me to use it. I am under personal obligation to several of my col- leagues in Yale University: to Professor Duncan for a patient hearing of Chapter II, to Professor Sneath for helpful criticism, to Professor Ladd for sugges- tions acknowledged in the note on p. 145, and to my philological colleagues, Professors Goodell, Hopkins vii PREFACE and Lang, for much suggestion and encouragement. But my heaviest obligation and one which I scarcely know how to express sufficiently is to Professor Oertel. Many of the subjects in this book I have talked over with him repeatedly during the past ten years, seldom without enlightenment and quickening. It would not be possible for me now to discriminate in certain chap- ters between what I owe to his suggestion and criticism and what is my own, and I must content myself with this general and grateful acknowledgment. E. P. MORRIS. Yale University, July, 1901. VU1 CONTENTS [The numbers refer to pages.] Page I. Introductory and Historical 1-35 The limitations of this sketch, 1. The syntax of the middle of the century, 2. Lange, 3. Logical and de- scriptive syntax : Holtze, Draeger, Kuhner, 4. The influence of comparative philology : Cdrtius, 6. The psychological tendency : Delbruck, 10. The American school: Hale, Bennett, Elmer, 14. Dittmar, 18. The influence of morphology : Ziemer, 23. Of general lin- guistics : Probst, 24. Of semasiology : Ries, 26. The later descriptive and historical syntax : Blase, Schmalz, Wolfflin's Archiv, 27. Summary of tendencies, 30. II. The Grouping of Concepts 36-47 Wundt's Volkerpsychologie, 36, note. The steps in the arrangement of concepts, 38. The germ-concept, 39. The relations of concepts in a group, 40. The close of the process of analysis, 40. The concept-group after speech, 42. The mental process of the hearer, 43. Sub- stantive and transitive elements, 44. General comments, 45. III. The Means of Expressing Relations . . 48-62 The musical elements, 49. Inflection : the agglutina- tive theory, 51. The unsystematic character of inflection, 52. Word-forming suffixes, 54. Single words as means of expressing relation, 56. The grouping of words, 58. These means not invented for the expression of rela- tion, 59. Their early history comparatively unimpor- tant, 60. The interchange of different means, 61. IV. The Process of Adaptation. — Inflection 63-101 The nature of the problem, 65. The method of approach to it, 67. Inflection and word-meaning : word-formation, 68. Case-endings, modes, and tenses, 69. The kind of evidence to be sought, 74. The influence of secondary inflection, 75. The effect of the context, 79. Problems ix CONTENTS Page suggested, 89. Illustrations of the working of these influences : the optative subjunctive, 92. The potential subjunctive, 94. Summary, 100. V. The Expression of Relation by Single Words 102-112 Prepositions, 103. Conjunctions, 104. The subjunctive in the subordinate clause, 104. Deecke on qui, 107. Probst's method, 108. The difficulties of the problem, 109. The origin of conjunctions, 110. The shifting char- acter of relational function, 111. VI. Parataxis 113-149 The methods in use, 113. Three points of view, 115. The psychological foundation of parataxis, 115. Absolute independence between successive concept-groups impossi- ble, 115. Varying degrees of closeness in the relation, 117. Groups connected as wholes, 118. Similarity and contrast between groups, 118. The synthetic summary of a group, 119. The analysis of a member of a group, 120. The means which suggest the paratactic relation, 121. Contiguity, 121. Order, 124. Inflection, 127. Single words, 127. The resulting forms of sentence, 130. Cor- relative parataxis, 131. Defining parataxis, 132. Psycho- logical definitions of parataxis, 142. Formal definition, 147. The extent of the field, 147. Lines of work suggested, 148. VII. The Subordinating Conjunctions in Latin 150-182 Their origin, 150. Inferences from etymology, 151. From case-form, 152. The problems presented, 155. Stable and unstable elements of meaning, 155. The acquisition of subordinating force: quin, 159; ne, 160; ni, 166; quamuis, 167; simul, 168; modo, 168; et and atque, 169. Summary of different types, 171. Tables of relationship, 173, 175, 178. Classification of Latin subordinating con- junctions, 181. VIII. The Grouping of Words 183-196 Explanation of the term, 183. The distribution of group- meanings, 184. The germ-concept, 185. The expansion of the germ, 187. The accommodation of meanings, 190. The re-transfer of meanings to the group, 191. The for- mation of idioms, 192. Partial re-analysis, 193. The association of analyses, 194. Analogy between word- groups, 195. X CONTENTS Page IX. Form, Function and Classification . . 197-232 Syntactical form, 197. Form and content, 198. Syntacti- cal function, 199. Classification by function: its advan- tages, 204. Its defects, 206. The reasons for the continued use of functional classification, 213. Its connectiou with logic and metaphysics, 214. Its unsuitability for historical syntax, 218. Formal classification; its disadvantages, 225. Its advantages, 227. Its use in psychological inter- pretation, 228. Classification as a tool of investigation, 230. XI LATIN SYNTAX INTRODUCTORY AND HISTORICAL Intelligent scientific work demands an intelligent understanding of method, upon its theoretical as well as upon its practical side, and method is intelligible only through its history. It is proposed, therefore, in this introductory chapter to sketch briefly the methods employed during the last quarter of a century in syn- tactical work in Latin. The subject of itself imposes certain limitations which, to prevent misunderstanding, must be laid down at the outset. In the first place, the sketch will be confined to work in Latin. It is to be regretted that students of Latin syntax are not more familiar with the work in Germanic and Romance philology, where the influence of earlier systems has been less strongly felt and where originality of view and of method is easier; but it is apparently a fact that Latin syntax has not been influenced by the syntax of the spoken languages. To a considerable extent this appears to be true also of Greek work, but as the methods employed are in the main identical, the influence is more difficult to detect and at the same time less important. The methods of comparative syntax, however, must be to some extent i 1 LATIN SYNTAX included within the discussion. In the second place, this is a sketch of methods, not of results, and some contributions to Latin syntax which would deserve a large place in a complete history of Latin philology may be passed over in a sketch of the history of method. Nor will any attempt be made to pass judgment upon the merit of the works mentioned. Great discoveries have been made in poorly equipped laboratories, and, on the other hand, the excellence of the method em- ployed may be unnoticed because of the writer's imper- fect use of it or his ignorance of the facts. In the third place, I am not sufficiently acquainted in a prac- tical way with the controversies about the case-system to be willing to enter upon that field. This is, cer- tainly, a large omission, and I regret it the more because case-syntax seems to be in advance of mode -syntax in freeing itself from the dominant influences of the half- century. Yet it may, I think, be assumed that the general course of case-syntax has been the same as the course of investigation into the meaning of modal and temporal forms. Of the syntax of the middle of the century much has been written and its characteristics are well known. It was not a special science, working for its own ends, but like palaeography or text-criticism it was still in service to classical philology. Even in this field the amount of detailed work was still small, 1 and the range was narrow. The sub-title of Weissenborn's Syntax der Lateinischen Sprache (Eisenach, 1835) is fur die oberen Klassen gelehrter Sclmlen ; that is, it was in- 1 See Draeger's statement of the extent to which he was obliged to rely upon his own collections {Vorrede to Vol. I, pp. iv ff.) and compare the small number of syntactical works referred to in Ritschl's first edition of Plautus with the long list in the final edition of Goetz, Schoell and Loewe. 2 INTRODUCTORY AND HISTORICAL tended to be a descriptive and practical statement of the facts of usage. But in one respect the early syntax was theoretical, — in the philosophical or logical conceptions which formed the basis of its schemes of classification. 1 Much of this has now been swept aside, sometimes with an insufficient appreciation of its real meaning and of its lasting value, but in two directions it still influences our syntactical work. The first of these is in the classi- fication of subordinate clauses, where the logical or metaphysical categories of time, purpose, condition, etc., still prevail in most grammars, though they are not so largely used in the actual work of investigation. The second and perhaps more important influence is in the definition of modes and cases. It is of the essence of philosophy and logic to reduce all phenomena to system by definition, to find the single underlying truth about which all things are to be grouped. And therefore the chief object of a logical scheme of the modes was to discover the Grundbegriff, and the chief inheritance which we still preserve from the syntax of the eighteenth century and the first half of the nineteenth is the belief that the enigma of modal use, and indeed of syntax generally, is to be solved by some kind of definition, psychological if not metaphys- ical or logical. The content and basis of the definition have changed; the feeling that a definition of the modes is necessary has remained a dominant force in syntax up to the present time. The first significant break from these systems was made by Lange in the paper entitled Andeutungen iiber 1 On all this see the valuable programs of K. Koppin, Deitrarj zur Ent- wickelung und Wurdigung der Ideen iiber die Grundbedeutungen der grie- chischen Modi; I, Wismar, 1877; II, Stade, 1880. LATIN SYNTAX Ziel und Methode der syntalctischen Forschung. 1 Lange's prime object was to claim a place for syntax as a special science, with aims and methods of its own, and this object was so far attained that he has been recognized as the founder of modem historical syntax. But this is not the only merit of the paper. It touches upon nearly all the problems which have occupied the science since that time — the distinction between form and function, the relation of syntax to semasiology, para- taxis, the formal classification of the subordinate clause — and frequently suggests in single sentences most remarkable anticipations of the method and aim of later work. Its immediate influence, however, was not great, at least so far as appears in the printed work of the time. Holtze's Syntaxis Priscorum Scriptorum Latinorum, 1861-62, which was of course planned and largely com- pleted before Lange's paper appeared (et est hie labor . . . plus viginti annorum, Praef., p. v), still followed the older lines, and Draeger's greater work, superior as it was in logical precision and in detail, introduced no new principle. Holtze's selection of the early Latin as a special field was in fact a recognition of the desir- ability of historical treatment quite as distinct as was indicated by the word historische in Draeger's title. For historical, in the sense in which Draeger uses the word, is scarcely more than chronological ; that study of the sequence of causes and effects which is suggested to our minds by the phrase historical syntax was un- known to Draeger. In his frank and interesting Vorrede he compares himself to an entomologist or a botanist, that is, his work was like the classifying and 1 Printed in the Verhandlungen d. 13 ten Versammlung Deutscher Philo- logen, pp. 96 ff., Gottingen, 1853. 4 INTRODUCTORY AND HISTORICAL descriptive sciences, before the publication of "The Origin of Species." The influence of the book, the extent of which can be seen in the great number of doctor-dissertations which follow its system, is due to the extreme clearness and precision of the scheme of functional classification, rather than to any originality in the syntactical method. In general, Kuhner's Ausfuhrliche Grrammatik der Lateinischen Sprache belongs to the same school of thought, as it belongs to the same period of time, as Draeger's Syntax. There is the same functional classi- fication, the same elaboration in subdivision. But the fact that Kuhner's plan involves explanation of facts, as well as classification, necessitates definition, and the definitions reintroduce the logical conceptions of lan- guage from which Draeger, except in his system of classification, was more nearly free. Kuhner's defini- tion of the sentence — that dangerous point for all syntacticists — is logical, and his scheme of the modes, though he repudiates the philosophical categories (Vol. II, p. 126, Anm.\ is only partially psychological. Even where the point of view is correct, the practice of beginning each subject with definitions, of which the succeeding sections afford illustrations, leads to a priori statements which in their spirit and tendency belong to philosophical syntax. See, for example, the distinction between the dative and the ablative, II, 256, and between the ablative and the adverb, II, 257; these are not the result of induction, but are deduced from general definitions of the cases and the adverb; they are rather descriptions of what a logically precise lan- guage ought to be than statements of the actual usage of so irregular and hap-hazard a mass of phenomena as language presents. 5 LATIN SYNTAX Draeger and Kiihner may be taken as the last and best representatives of logical grammar. The traces of that school of syntax which still remain in our gram- mars do not indicate an active working of the older conceptions, but a passive survival, a traditional preser- vation of the ideals of the previous half-century. Meanwhile the main work of classical philology dur- ing the middle years of the century was in text-criticism, and the main work of philology in the narrower sense was in comparative philology and in morphology. The influence of this work was not greatly felt by Draeger, nor, in spite of the fact that his first volume is a com- pendium of Latin morphology, by Kiihner, perhaps because the original plan of both works dates back to a period when classical philologists were still somewhat suspicious of the newer science, but upon later methods in syntax the influence of comparative morphology has been very great. Georg Curtius may be taken as the representative of this influence, not so much for what he did as because it was he who more than any other philologist interpreted to classical scholars the work of comparative philology. The beautiful clearness of his system as taught in his lectures and in his writings especially fitted him for this office, and gave to his teachings an authority with classical scholars greater even than the authority which was conceded to him by other comparative philologists, great as that certainly was. It was, in particular, through his application of the theory of agglutination that he affected the method of investigation in syntax. For that theory taught that the inflected forms of the Indo-European languages were the result of the appending of elements once dis- tinct and having distinct meanings to stems which also had distinct meanings. The result of such composition 6 INTRODUCTORY AND HISTORICAL would be a form of distinct meaning, and where it was possible, as it appeared to be in the case of many verb- forms, to analyze a compound into its component parts, the original meaning of the inflected form could be known, and would be the proper starting-point of any syntactical or semasiological study of its historical uses. In some form, and with reference chiefly to the morpho- logical side, this theory is perhaps still the passively accepted belief of philologists, but it was held thirty years ago with a much stronger and more unquestion- ing conviction, and especially with more confidence in the explanation of meanings by the process of analysis. Reservations and scientific caution are less easily learned than general theories, and it is probably true that clas- sical scholars accepted the results of comparative phi- lology with an unjustified degree of confidence, and applied them more sweepingly than their authors would have ventured to do. There is, indeed, evidence that Curtius himself did not draw from the theory of agglu- tination the conclusions which were drawn by syntac- ticists, but sanction from without, from another science which is imperfectly understood and therefore the more respected, almost always carries undue weight. In this case the tendency was strengthened by the fact that the habit of definition, inherited from philosophical syntax, the predisposition to explain a case or a mode by some single word broad enough to cover all its uses, still remained after the views of language which gave rise to it had been discredited. To this predisposition com- parative philology seemed to give a scientific support. Definition by a process of analysis which determined the significant elements of an inflected form, and which thus determined the original meaning of the form itself, took the place of definition by philosophical categories. LATIN SYNTAX The basis of the definition was changed, but the habit of regarding the discovery of some single meaning, about which all other meanings and uses could be grouped, as the proper and sufficient explanation of an inflected form still remained and became the dominant influence upon syntactical method for many years. It is not unlikely, also, that the disposition to seek for the ultimate explanations of syntax in the primitive meanings of forms was strengthened by the general drift of the thought of the nineteenth century toward the study of origins. Some other branches of philology were distinctly affected by the methods of natural sci- ence, and the early use of the comparative method and its application to text-criticism show that philology shared, if indeed it may not be said to have started, the current of its time, just as it had shared the philosoph- ical tendencies of the eighteenth century. The influence of the views of Curtius is well illus- trated in Liibbert's first work, Der Conjunctiv Perfecti und das Futurum Exactum. This was published in 1867, before Draeger's Syntax or Kiihner's Grammar. The question which it discussed, the difference between fecero, fecerim and fazo, faxim, had been treated before by other scholars, especially by Madvig, and their ex- planations had involved a theory of the morphology of faxo t faxim; but in Liibbert the morphological argu- ment is not incidental, it is one of the two main sup- ports of his conclusion. In the order of the sections, those which deal with the form precede those which deal with the syntactical usage, and the conclusion of the whole book is plainly based upon the belief that faxo is aoristic in form and that this fact determines the fundamental meaning and therefore the later usage. It is true that in some of the details of the argu- INTRODUCTORY AND HISTORICAL ment Lubbert follows Bopp rather than Curtius (e.g., pp. 67-68), but in the general character of the argu- ment from formal analysis to Q-rundbedeutung and from this to usage, he is following the method of which Curtius is the most conspicuous exponent. Since Liib- bert's time it has become the standard and orthodox method to begin the syntactical treatment of an inflec- tional form with a morphological argument — which in many cases is in truth a necessity — and to make the meaning which is obtained by means of morphological analysis the foundation of all later usage. The supplanting of philosophical views of language by psychological conceptions has been a long process and is not yet complete. Koppin, in his program of 1880, pp. 3-12, shows that at least as early as the be- ginning of the century psychological definitions of the Greek modes were attempted, and during the first half of the century the gradual falling away from logical systems was due in part to the gradual rise of other views, as, for example, of the localistic theory of the cases (1831), which in spite of its use of semi-logical categories is in essence psychological. Kuhner's modal scheme, as has been said above, is partly psychological ; begehren and vorstellung are used together in describ- ing the subjunctive. Lubbert's first section is headed Psychologische Grrundanschauungen in der Sprache, though it is very brief and the psychology is scarcely distinguishable from philosophy. In the wider field of general linguistics the intimate relation between psy- chology and philology was of course recognized much earlier than in syntax; Steinthal's Grrammatik, Logih und Psychologie was published in 1855. But the sub- stitution of psychological fundamental meanings for philosophical in the field of the modes marks, never- 9 LATIN SYNTAX theless, the beginning of a new epoch in syntactical method. The book which brought about this change was Del- briick's Conjunctiv und Ojptativ, the first volume of the Syntaktische Forschungen, published in 1871. 1 It is not necessary to make a detailed statement of the con- tents of this well-known work. Of its two main theses the second, that all subordinating function is acquired, was the less original and has been the more widely accepted. The first proposition (pp. 11-30), that the earliest meaning of the subjunctive was Will, of the optative Wish, has been at the same time more impor- tant in its influence upon later work and more earnestly questioned. 2 The method used in establishing this proposition therefore demands special comment. 1. The field of inquiry in regard to the earliest mean- ing of the modes is narrowed by the exclusion of inter- rogative sentences on the ground that the question represents, psychologically, a check in the natural movement of the train of concepts, and of negative sentences on the ground that they are modifications of the declarative sentence. It is difficult not to see in these exclusions an unconscious survival of the logical 1 Delbriick's earlier work, Ablativ, Localis, Instrumental 'is, Berlin, 1867, I venture with some hesitation to leave out of this brief sketch, in spite of the fact that it is the basis of all later treatment of the Latin ablative. Its method is, in my judgment, superior in some respects to that of the Con- junctiv, in that the uses of the cases are interpreted in the light of the accompanying verbs and prepositions. But in part the method is the same ; usages are held to be connected when they have a common element in meaning. It is this part of the method which has apparently found most followers. 2 This was later modified by Delbriick so far as to make futurity a more prominent element, either because of the criticism of other scholars (see Greenough's review in The North American Review, 1871, CXIII, 415) or as a result of the author's own revision of his position. 10 INTRODUCTORY AND HISTORICAL view of language, which took the unemotional declara- tive sentence as the normal type. On the contrary, the unemotional narrative sentence is the later and, so to speak, the more artificial form, nor is the exclamatory and questioning attitude of mind any less natural and primitive on the psychological side. In following only the modal development of the non-interrogative, non- negative sentence, Delbriick ran the risk of omitting elements which entered into the primitive meaning and of reaching an incomplete and one-sided result. The exclusion of all persons except the first person singular is still more important. In Latin the subjunc- tive is used but rarely in the first singular in indepen- dent non-interrogative sentences. In Plautus the ratio in the present tense is about 1:12; excluding the form uelim, it is only about 1 : 35. The ratio may be much greater for Homer, but the probability that the exclu- sion of the second and third persons has distorted the result is still considerable. The reason given for this narrowing of the field is (p. 13) that the wish in the first singular involves only one person while the wish in the second or third singular involves at least two; we must therefore suppose that the earliest use of the optative is found in the first singular and may expect the same thing in the subjunctive. This is, so far as the optation is concerned, a pure assumption, for the attitude of mind involved in wishing that another person may come to harm is not more complex than that of wishing for one's own well-being, and, so far as the subjunctive (and the will) is concerned, it is a mistaken assump- tion. The situation in which one expresses his will in regard to his own action is comparatively rare and artificial, except when the will takes the form of deter- mination, which is usually expressed by a future. The 11 LATIN SYNTAX will, in the somewhat unfortunate sense in which that word is used in English by philologists, is most natu- rally felt and expressed in regard to the action of other persons than the speaker. 2. The object of the investigation was to find, not the most abstract term which would cover all the uses of the modes, but the earliest meaning, the primitive meaning. This is a question of chronology, of dates, or, where the evidence of actual usage is not accessible, of relative antiquity. Such a question is of course surrounded by immense difficulties and the solution can be at best only an approximation. But the greater the difficulties, the more distinctly must they be faced and the more clearly must we keep in mind the fact that the investigation is fundamentally chronological. It is at this point that Delbruck's method is least clear. There are no criteria of the relative age of different usages beyond the criteria implied in the exclusions noted above, with the accompanying reasons, and after the first few pages the question shifts from the position taken at the outset and becomes a question of the psy- chological analysis of certain usages. 3. As a basis for subdivision, after the main classi- fication by form of sentence and person and number of the verb, Delbruck selected the intensity of the expres- sion of will or the proportion of will to expectation and opinion, rejecting the attempt to classify by the content or object of desire, that is, by the meaning of the verb. This selection and rejection was the parting of the ways. It has had two consequences. In the first place, the sub-classes thus made are large and vague abstractions, — will, exhortation, command, obligation, wish, concession, futurity, — abstractions which cannot be defined with precision nor discussed without the danger 12 INTRODUCTORY AND HISTORICAL of confusion. The determination of the class to which a particular case belongs must be chiefly by translation, and though in Delbruck's hands the tendency to rely upon translation and to treat abstractions as realities is checked by scientific reserve and candor, in the hands of some of his followers it has been distinctly mislead- ing. In the second place, the grouping of expressions of desire according to the intensity of the desire brings together things which may be associated in a system, but are not associated in our actual psychological expe- rience. In experience we associate our desires in groups according to the thing desired. Hunting, fishing, sail- ing; reading, studying, thinking; eating, drinking, rest- ing, smoking; gardening, carpentering, tinkering: it is in such groups that our desires, whether in the form of advice or concession or exhortation to others or with reference to our own action, are associated in our minds. And it is these lines of association which give rise to analogies and assimilations of expression, and which therefore indicate the fruitful lines of syntactical inquiry. The turning aside from this field of study to the classification according to intensity of will and wish is the serious defect of the book. The criticisms which followed Delbruck's attempt to establish the Grundbegriff of the subjunctive and the optative need not be taken up here. They came from Lange, Ludwig, Bergaigne, and Koppin, but did not prevent a rather general acquiescence in Delbruck's results. The most elaborate of them, Bergaigne's De conjunctivi et optativi in indoeuropaeis Unguis informa- tione et vi antiquissima, Paris, 1877, has apparently attracted less attention than it deserves. It is, how- ever, mainly a discussion of questions of comparative philology, and must be passed over here with a general 13 LATIN SYNTAX commendation of its method to any who may be inter- ested in these questions. In America Delbriick's work has been accepted with enthusiasm, though Greenough, in the review mentioned above, questioned its results and proposed to substitute futurity as the common beginning of both modes. The method of Greenough's own pamphlet on the Analysis of the Latin Subjunctive, 1870, is not dissimilar to Delbriick's, but it was privately printed, and its bril- liant speculations and fruitful suggestions, though they doubtless prepared the way for the acceptance of Del- briick's more elaborate work, have not directly in- fluenced the course of Latin syntax in this country. 1 After Greenough three American scholars may be named as representing in different ways the prevalent school of Latin syntax in this country. 2 Hale's work is in The Cum Constructions, Part I., 1887, Part II., 1889 (German translation, Teubner, 1891), in The An- ticipatory Subjunctive in Greek and Latin, 1894, and in various articles. Bennett's work is in the Appendix to his Latin grammar, Boston, 1895, and in his Critique of Some Recent Subjunctive Theories (Cornell Stud- ies, IX, 1898), and more recently in The Stipulative Subjunctive in Latin (Transactions of the Amer. Philol. Assoc, XXXI, 223 ff.). Elmer's chief contri- butions to syntax appeared in the American Journal of Philology, XV, 2, 3, reprinted as a pamphlet en- titled The Latin Prohibitive, and in Vol. VI of the Cornell Studies (Studies in Latin Moods and 1 Indirectly, through the Allen and Greenough grammar, some of these suggestions have exerted a considerable influence. But for various rea- sons school and college grammars must be passed over in this sketch. 2 Et monere et moneri proprium est uerae amicitiae et alterum libere facere, non aspere, alterum patienter accipere, non repugnanter. 14 INTRODUCTORY AND HISTORICAL Tenses). I venture thus to place the work of these three scholars together, because the question is not of results, in regard to whicli they differ, but of method, and their method is in all essentials nearly enough iden- tical to justify general rather than individual descrip- tion and comment. It has been said above that the methods and results of the Conjunctiv und Optativ have met with general acceptance in America. This is so far true that it is proper to speak of the American work as a continuation of Delbruck's work and an application of his methods to Latin syntax. It is not necessary therefore to repeat the characterization of that method attempted above, but only to show what aspects of it have been espe- cially emphasized. The most important of these are the result of a failure to distinguish sharply between the work of comparative syntax and the work of Latin syntax. In consequence of this there appears in the work of American scholars an undue emphasis upon inferences as to the prehistoric stage and a tendency to make too large use of the methods of comparative sj r ntax ; that is, there is con- fusion both as to aim and as to method. The aim of comparative philology is the construction of hypotheses and the suggestion of possibilities which will be in harmony with the facts of the historic periods and will throw additional light upon the phenomena of the single language. This aim is primarily historical and direc- tive. The aim of Greek or Latin syntax is to study processes and to formulate laws; it is primarily psy- chological and linguistic, and only secondarily histor- ical. It is hazardous to attempt to interpret the aim and purpose of others, but it is difficult to resist the impression that the contribution which Latin syntax 15 LATIN SYNTAX may make to prehistoric syntax has occupied a place of undue importance in the work of American schol- ars. Accepting without reserve the categories of modal function which Delbriick has made for the Indo- European speech, the endeavor of syntacticists has been to discover survivals of these functions in Latin. Thus Hale has connected the future force of the sub- junctive in certain sentence-forms with the primitive future, and Elmer would establish a function of obligation reaching back to the Indo-European stage. The emphasis placed upon this side of the work — a perfectly legitimate side in itself — has brought about a tendency to regard such connection with the earlier period as the most important part of syntactical work and as the ultimate and sufficient explanation of Latin usage. The inevitable result is a with- drawing of interest from the proper work of Latin syntax. Of aims one must speak with some hesitation, but in regard to methods one may speak with more positive- ness. By the necessity of the case the student of pre- historic speech must depend upon inference, since he has no contemporary data and can look only for some- what general results. The nature of his problem com- pels him to run the risk of dealing with abstractions and with bare probabilities. But the student of Latin may and therefore should use a method which keeps closer to the facts. His work is one of observation and of accurate induction within narrow limits. The use of large functional classes, like the volitive, the poten- tial, the optative, as the tools of investigation, when more precise formal classification is made possible by the possession of abundant data, is a considerable defect in method — in this case the result, apparently, of a 16 INTRODUCTORY AND HISTORICAL too complete adoption of the methods of comparative syntax. 1 One other characteristic of the method of American syntax deserves special mention ; it has been, in Greek as well as in Latin, conspicuously systematic. There is doubtless some danger in the use of tables and graphic schemes of syntax, the danger that they may become traditional and may lead to the ignoring of the irregu- lar, the exceptional. Language is so hap- hazard, so complex, that the exceptional cases which do not fit into systems are the cases which deserve most attention and may afford suggestions for new discovery. But the dangers of formlessness and absence of system are still greater. Facts, if they are truly and fully appre- hended, will in the end always group themselves sys- tematically, and the emphasis which American syntax has laid upon system is a real contribution to the science. 1 These remarks are perhaps liable to misconstruction. I certainly do not mean to take the position of undervaluing, or even of criticising, so monumental a work as Delbriick's Comparative Syntax, either as to its method or its results. But the methods of comparative syntax are entirely inapplicable to the syntax of Latin or of Greek. It is the results that are of interest to the Latin scholar. And even with reference to these he must exercise some reserve, not only because the science is still somewhat young, and many of its results not yet a part of the accepted doctrine, but also because the classical scholar must take them, if he takes them at all, in a rather uncritical way. The linguistic equipment of most Latin scholars, to speak frankly, consists of a knowledge of Latin and Greek, a reading and speaking knowledge of English, German, French, and per- haps Italian, and a fading recollection of Sanskrit. The ability to read the examples in a number of languages is a poor foundation for critical judgment, and the only part of this equipment that is of much value is the Latin and Greek. It does not follow from this that we must forego entirely the enlightenment which comes from finding parallel phenomena in another language, or in many Indo-European languages, but only that we must face our limitations and do our work where we are competent to do it. Otherwise we shall run into parallel grammars of Greek and Latin and other scientific anachronisms. 2 17 LATIN SYNTAX Dittmar's Studien zur Lateinischen Moduslehre opens with an elaborate discussion of Hale's Cum, and in its general outline so much resembles that book that it may properly be regarded as a continuation of the same method, in spite of the fact that in some details of treatment and in its results it is a protest against the prevalent views. The main thesis is that the confu- sions of modal syntax can be removed if the subjunctive be regarded in all its uses, either in independent or in subordinate clauses, as the expression of an attitude of mind which is described by the adjective polemisch, while the indicative in like manner expresses the souverdn attitude. It is evident that this is a form of psychological Grundbegriff and that the book belongs therefore in the line of work in which the discovery of a fundamental meaning is the ultimate aim, but in many details the method employed is unusual. In the first place, it is not Dittmar's purpose to show that the polemic element underlying the subjunctive is its earliest meaning, though in a brief summary he ex- presses the opinion that this meaning was found in the Indo-European period, but rather to show that it is found in equal measure in all uses of the subjunctive from Ennius to Juvenal, substantially unchanged. He does not, therefore, deal largely in prehistoric specula- tion, but cites a considerable number of cases without much regard to chronology, and finds in each subjunc- tive a polemisch element, in each indicative something souveran. In the second place, the cases of the sub- junctive are all (except half a dozen repudiating ques- tions) taken from subordinate clauses. Whether this is because the polemic character of the subjunctive in leading clauses is regarded as self-evident or because Dittmar believes that the true character of the mode is 18 INTRODUCTORY AND HISTORICAL most apparent in the subordinate clause, is not stated. A third characteristic of the method is the arrangement of many of the cases, e. g., the cases of ^uj-clause, in groups by form, not by function. There are thirty-six such groups under qui, nineteen under cum. But this is not followed out systematically; there are many more than thirty-six formal varieties of ^wi-clause, and the three groups of w^-clauses are still more short of com- pleteness. But the beginning of a formal classilication is worthy of note. I have thus far followed the dominant school of syn- tax down to the present time. The characteristic of its method which connects the later work with the earlier in an unbroken line is the conception of a Grund- begriff as the goal of syntactical investigation. At the beginning of the century the fundamental meaning was sought in metaphysical or logical definition; under the influence of Bopp and Curtius and the comparative etymologists it was sought by analysis of inflected forms into their significant elements; under Delbriick's leadership it has been sought by psychological defini- tion. The method of the search has varied, the object sought has been essentially the same. In thus following down to the present time the lead- ing school of syntactical method I have passed by much work which was going on upon slightly different lines. The greater amount of work, especially the dissertation work, of this period has been along the lines laid down by Draeger. A good illustration of its character is Dahl's Die Lateinisehe Partikel VT, Kristiania, 1882. In his treatment of the modes Dahl shows the influence of Delbriick, but the general scheme of classification is functional, more elaborate than Draeger's or Kuhner's, as the narrower field permits, but essentially the same 19 LATIN SYNTAX in method. The book deserves mention in such a sketch as this merely as a reminder of the fact that no single school has completely occupied the ground; descriptive and functional syntax has held its own by the side of speculative and comparative methods. Meanwhile two movements had begun within the general field of philology which have already influenced the methods and even the aims of syntax and which are destined to influence them still more deeply in the future. These are the modern school of phonetic sci- ence, often called the Junggrammatisch school, and the science of general linguistics. The history of the neo-grammatical movement has been sketched in various places 1 and need not be at- tempted here. To an outsider the doctrine that pho- netic laws work without exception does not seem to be so much a fundamental principle as an incidental doc- trine, a step in the progress of the movement, raised into a somewhat factitious importance by the fact that it happened to become a point of attack and of defence. But it behooves one who looks at the question from the outside to speak with caution ; the doctrine is at least illustrative of the shift of the center of interest from morphology to phonology. The three steps have been etymology, morphology, phonology. In the second stage the laws of sound-change were studied in order to explain forms ; in the third stage forms were studied in order to discover the laws which had been at work in producing them. The effect of this shift of center upon syntactical method is indirect but strong. In the first place, it has to a considerable extent withdrawn interest and conviction from the earlier conception of 1 E. g. Delbriick, Einleitung in das Sprachstudium, especially pp. 54 ff., Ziemer, Junggrammatische Streifziige, pp. 1-29. 20 INTRODUCTORY AND HISTORIC A I. agglutination (cf. Delbruck, Einleitung, p. 55), and without actually proposing a substitute lias led to the suggestion of various hypotheses which are not easily reconciled with any theory of agglutination, that is, of composition of significant elements. The scientific support which the teaching of Curtius had given to the explanation of inflectional forms by definition, by a Griuidbegriff, whether psychological or metaphysical, thus falls away; the use of this term to describe the sphere of application of syntactical forms 1 is evidence of this. In the second place, the newer attitude toward forms and the laws of sound turns the attention from re- sults to processes, to laws. Syntax has been mainly oc- cupied with results, with the tabulation and description of the facts of sentence-structure; such attention as it has given to laws and processes has been vague and lacking in precision because it has been on too large a scale. The science of phonetics sets to the sister sci- ence an example of minute and patient observation, and one instance of the direct transfer of a method of study from phonetics to syntax will be mentioned below. The second great movement of recent years in phi- lology is the rise of the science of linguistics. It is the result of the work of many scholars 2 who have con- tributed to it from various sides, phonetic and psycho- logical, and as the movement of the Neugrammatiker is gathered together in Brugmann's G-rundriss, so the science of linguistics is summarized in Paul's Prin- ciples, der Sprachgeschichte. The close connection be- 1 Delbriick in Brugmann's Grundriss, III. I, p. 81 ; Brugmann, Indog. Forsch., V, p. 93, n. 2. 2 Among them an American may with justifiable pride name "Whitney of Yale. 21 LATIN SYNTAX tween the two movements is sufficiently attested by the fact that the lists of contributors to them consist largely of the same names. To a considerable degree the field covered is also the same, namely, the unconscious and automatic psychology of the production and reproduc- tion of articulate sounds. The aim and method, too, are similar. The science of linguistics is interested in processes, in the process of speech-learning and speech transmission, in the accumulated variations which re- sult in dialect, in the steps which separate the popular from the written language, and in all the forces and laws which bring about and control these processes. In these respects its methods would naturally exert an in- fluence upon syntactical method like the influence of the new phonology, but greater. Both sciences suggest also the inadequacy of older classifications of the phe- nomena of language or their uselessness for the solving of the newer questions. Phonology cuts across the parts of speech, taking its illustrations indifferently from nouns or adverbs or verbs, and linguistics finds proofs of the working of analogy alike in sounds or word-forms or word-meanings; it is in part this dis- regard of distinctions which have been regarded as fundamental that made these sciences at first so con- fusing to the ordinary classical philologist. But the example is one which the syntacticist may well lay to heart. These contributions to syntactical method are in- direct, in the way of suggestion merely ; but linguistics is concerned also with the conscious psychology of speech, with the train of thought which accompanies and is associated with utterance, and in this respect it approaches the field of syntax more closely. Between semantics and syntax it is not possible to draw a sharp 22 INTRODUCTORY AND HISTORICAL line, and the principles and methods of semantics as set forth, for example, in Brdal's Essai de Simawtigue or in Darmesteter's La Vie des Mots are immediately avail- able, with but slight change, for the uses of syntax. The influence of the newer phonetics, of linguistics in general, and of semantics may be illustrated by a consideration of the methods of three books published within the last twenty years. Ziemer's Junggrammatische Streifzilge (Colberg, 1883) first appeared as a program in 1879 under the title Das Psychologische Moment in der Bildung Syntaktischer Sprachformen. In the enlarged form a historical intro- duction gives a sketch of the neo-grammatical move- ment, and discusses the possible application of its methods to syntax. The body of the book treats of the psychological element in speech, discusses the nature of assimilation as it appears in syntactical forms, defines three kinds of syntactical assimilation, and illustrates them at length with many examples taken chiefly from the Latin. The book is thus a direct transfer of the methods and the aims of morphology to syntactical investigation ; the laws of assimilation under the influ- ence of analogy were discovered and worked out in the field of morphology, and Ziemer's purpose is to show that good results may be obtained by a like method in the field of syntax. But while the results are interest- ing and the explanation of many peculiar constructions is clearer than any previous explanation, the value of the book is in its method, in the fact that it does not aim primarily at rules for case-constructions or for the use of the subjunctive mode, but at the establishment of a law of speech which underlies case-construction and modal use alike. It directs the attention away from classifications and fixes it upon the working of 23 LATIN SYNTAX the mind of the speaker, suggesting thus a profounder syntax working in the sphere of causes. The influence of this original and significant book has not been as great as might have been expected. Its merits were recognized by other scholars in the neo- grammatical school, to whom the method, in its mor- phological applications, was already familiar, but it was perhaps too bold a departure from the ordinary method of syntax to meet with an immediate acceptance at the hands of classical scholars. Even Ziemer him- self in his Indogermanische Oomparation (1884) returned to the older method, analyzing the comparative forms into supposedly significant elements and deriving later uses from the fundamental meaning thus obtained, and no one has taken up the fruitful line of study sug- gested in the StreifziXge. For this method is capable of wide extension. Historical syntax consists in part in the tracing of relationships between different struc- tures, in the determining of the influence of one mode of expression upon another, either in the way of attrac- tion or of competition. But all this historical and genetic study of syntax is in danger of remaining vague and inconclusive, unless it is completed and fortified by the most minute and detailed analysis of the under- lying psychological resemblances and differences which are the cause of relationships and competitions. Such analysis may be employed, as Ziemer has employed it, in wider fields. Ziemer's Streifziige may properly be connected par- ticularly with the school of morphology to which he belongs, in spite of the fact that he often refers to Paul's Principien and uses, as is natural, the terms of general linguistics. In like manner Gutjahr-Probst, though he quotes and refers to the Neugrammatiker, 24 INTRODUCTORY AND HISTORICAL may be properly regarded as more distinctly an expo- nent of the influence of general linguistic science upon syntax. His Beitrcige zur Lateinischen G-rammatik appeared in three parts: the first (1883) dealing with the verb, the second (1883) with particles and conjunctions, the third (1888) with ut in Terence (und Verwcmdtes). It is in the second and the earlier portions of the third part that the characteristics appear which make it worth while to mention the book in this sketch. These sections deal with the history of the subordinating conjunctions and particles. Since the fact was first recognized (by Lange and, in more detail, by Delbriick) that the sub- ordinating function is an acquired function, little had been done in the way of precise study of the steps of this acquisition, at least in Latin. Kienitz' study of quin (1878) is cited by Probst in the bibliography pre- fixed to the second part, and O. Brugmann's Gebrauch des condicionalen Ni (1887) in the third part, but even now, though there are many useful contributions to our knowledge of conjunctions, there has been but little study of their early conjunctional history that goes beyond the very general principles laid down by Del- briick. The main contribution which Probst's work makes to syntactical method is, therefore, in its general attitude, in the author's perception of this large gap in our knowledge of the steps of the process of acquiring subordinating power. In his manner of approaching this question, also, Probst is, in general, guided by correct principles, for he emphasizes and to some degree illustrates the true linguistic ideas that, in the first place, a particle may start, so far as our knowledge goes, from a variety of applications and uses, not from some single fundamental meaning, and that the process 25 LATIN SYNTAX which is to be studied is a movement toward precision, not away from it. And, in the second place, he recog- nizes the general principle that a conjunction or particle acquires its meaning from the sentence, not the sen- tence from the conjunction. These two leading prin- ciples come from the science of linguistics, and it is their recognition which makes Probst a fair representa- tive of the influence of linguistics upon Latin syntax. As in the case of Ziemer's Streifzuge, the recognition which Probst's work has received has been chiefly from scholars of the neo-grammarian and linguistic schools, who looked primarily at the method. Students of Latin syntax, who were concerned chiefly with the results, have been much less favorable in their judgment. Some of the details of Probst's work will come up later for consideration ; but it must be said that it contains some surprising errors of fact and of inference, e. g., the statements that quod passed through an interrogative stage, that an acquired a special function in competi- tion with nonne, that enim was originally interrogative ; indeed, the whole theory of the interrogative sentence is incorrect. Such errors, however, should not prevent a recognition of the fact that in its fundamental prin- ciples the book will teach much which cannot be learned from the far more precise and careful work of Dahl or of Schnoor on ut in Plautus. The third book which deserves mention as illustrat- ing the influence of other branches of philology upon syntax is Ries's Was ist Syntax? Marburg, 1894, almost the only work of recent years except Koppin's programs, mentioned above, which deals at any length with ques- tions of the method and theory of syntax. Its main thesis is that, as single words are studied with refer- ence both to their form (morphology) and their mean- 26 INTRODUCTORY AND HISTORICAL ing (semasiology), so groupings of words should be studied with reference to their structure (formal syn- tax) and their meaning (functional syntax). The dis- tinction between form and function is a very old one, and its application to sentences was suggested, as so many other developments of modern syntax were sug- gested, in Lange's paper (1852). Ries, however, while giving full credit to Lange, has much enlarged the sub- ject and applied it with all the added light which the advance of semantics since Lange's time has thrown upon it. All this (the details are too well known to call for mention) constitutes a large and very direct contribution to syntactical method. So far as the syn- tax of inflected forms is concerned (case-construction and much of modal syntax), the laws which govern the change of meaning of single words apply almost with- out change, and the problem of the acquisition of sub- ordinating force is, quite strictly, a semantic problem. The treatment of ut in a lexicon should be essentially the same as its treatment in a scientific grammar. A larger modification is necessary in transferring the methods of semasiology to the syntax of groups of words ; meaning plays but a subordinate part in deter- mining the form of words, but in the grouping of words meaning is the shaping and controlling force. The three works last mentioned, embodying, respec- tively, the suggestions which syntacticists may receive from morphology, from general linguistics, and from semantics, may be said to be on the outskirts, or per- haps to be outposts, of syntactical work. The main current of work has kept somewhat closer to the lines of descriptive and statistical grammar, though it has been affected in varying degree by the influences of the other schools of philology. Of this solid and intel- 27 LATIN SYNTAX ligible work the articles and books of H. Blase may be taken as representative. They consist, beside the doctor-dissertation on conditional sentences in Plautus (1885), of the Gescldchte des Irrealis (1888), the Gre- schichte des Plusquamperfekts (1894) and various articles in Wolfflin's Archiv. The method employed in them is characterized by its close adherence to facts ; much of the work is descriptive and interpretive, with statis- tics in condensed tables and with careful observance of the local and stylistic peculiarities of the writer. An unusual amount of attention is given to late Latin and to the connection with the Romance syntax. The article on the futures and the perfect subjunctive in Wolfflin's Archiv, X, 3, starts from the state of things in certain Romance languages and traces this back to its origin in Latin. On the negative side the method is no less noteworthy. The PlusquamperfeJct and the Irrealis both begin with the material in Plautus, and in all the work there is a complete avoidance of speculation or even of the simplest inference in regard to prehistoric syntax. The whole question of the fundamental mean- ing of the subjunctive is ignored, with all the related questions in regard to a supposed Indo-European origin of this or that usage. The discussion in regard to absolute and relative time, the most extensive if not the most important discussion in modern syntax of the verb, is briefly dismissed in a few pages of the Plus- quamperfekt. In all this the connection of Blase with the work in the syntax of early Latin under the direc- tion of Studemund is apparent; his method is in general that of E. Becker and Richter and Bach in Stude- mund' s Studien. But while he thus ignores, appar- ently of deliberate choice, much of current work, he is not uninfluenced by the thought of the general science 28 INTRODUCTORY AND HISTORICAL of linguistics. The explanation, for example, of the Plautine/M eram (=fui) is by a Kombinations-Ausgleich- ung, according to the doctrine of Ziemer. The work of Schmalz in the revision of Krebs's Antibarbanis, and especially in the Latin Syntax of Vol. II of Midler's Handbuch is too well known to need description. In general, his method of treatment is like that of Blase, exhibiting the same tendency to avoid speculation and to keep close to facts. The pur- pose of the Antibarbarus is mainly stylistic, and the habit of close observation of stylistic peculiarities is carried over into the Syntax. The introduction, in- deed, dwells at somewhat disproportionate length upon the need of distinguishing between the spoken and the written Latin and upon the individual and local pecu- liarities of authors, so that in the third edition only a few lines are given to the explanation of the system which is followed in the book. The object of the Syn- tax, also, — to serve as a compendium or exhibition of the accepted results of syntactical science, — tends to exclude theory, and in many parts, where no reasonably satisfactory theory exists, only a bare presentation of facts is possible. The result of these limitations is that, valuable as the work has been to syntactical scholars, it makes but slight contribution to the theory or method of syntax. The only novelty is the substi- tution of a formal classification of subordinate clauses, by the introducing word, for the prevailing functional arrangement. This was proposed by Jolly in 1874. The foundation of the Archiv fur Lateinische Lexico- graphic unci G-rammatik in 1884, and the inclusion of syntax within its scope, has brought together many syntactical articles. They vary somewhat in method, but there is a large enough common element to justify 29 LATIN SYNTAX a general comment without mentioning individual work, even that of the editor, Wolfflin. 1. There is little comparative or prehistoric syntax, little speculation, little consideration of fundamental meanings, and the influence of the methods of mor- phology and general linguistics is not great. 2. But there are many articles on the border-land between syntax and semantics, especially the "Was heisst . . . ? " articles. 3. Much attention is paid to local varieties of Latin speech, particularly to African Latin, and to the pecu- liarities of late Latin and the connection with the Romance languages. 4. Many of the articles follow single constructions through all periods, making perpendicular sections, so to speak, instead of following a group of related con- structions horizontally through a single author or period. 5. The most noticeable characteristic is that which has been already noted in the work of Schmalz, the tendency to be satisfied, at least for the present, with recorded and unconnected observation. The journals of the natural sciences are full of such work, e. g., in chemistry or natural history, and the value of it is un- questionable. But it is also evident that it is mere material until it is organized by theory. The period which has been under survey in the pre- ceding pages, covering a little more than a quarter of a century, is not easily summarized, yet certain lines of historical connection run through it, as I have attempted to show, and serve as a basis for understanding its aims and methods. It has been, in part, a period of healthy variety. Perhaps the largest portion of it in bulk, if all doctor- 30 INTRODUCTORY AND HISTORICAL dissertations be included, is descriptive, and this por- tion varies considerably from the unconnected record of facts, such as may be found in the briefer notes in Wb'lfflin's Archiv, to the systematized and classified description in Dahl's ut or the papers in Studemund's Studien. Much of this description is given, of course, as a foundation for scientific induction, and is arranged with reference to that end. There has been also much functional study, the ultimate end of which is the more precise statement of a syntactical rule or the more exact determination of the means used to express a function like the causal or the conditional relation. Nearly all study of conditions has been functional. There has been also a considerable amount of speculative work, the result mainly of a desire to systematize the observed phenomena. But discussion of the principles of syntax has been almost wholly passed over. In this respect syntax has been remarkably conservative. This variety in aim and method is, I have said, a natural and healthy variety. There is no single aim for such a science as syntax, and there is therefore no single method which can properly claim superiority at all points. Aims vary in importance, and the methods appropriate to them also vary, but knowledge is many- sided, and all aspects of it are legitimate. Descriptive syntax has by no means covered the whole field; of such an important subject as the relative clause we have still but a fragmentary and incomplete description. And the methods of descriptive and statistical syntax are still inexact; it is not possible to describe species by the present method so that single cases of the species can be identified beyond dispute. Functional study, also, has a legitimate place ; it increases the precision of interpretation, and is often the best means of approach- 31 LATIN SYNTAX ing a mass of cases and of discovering formal differences. But it also offers many opportunities for improvement in method. Its categories are still too vague and sweep- ing. } And of speculation in its proper place, not as a substitute for knowledge but as directive of investiga- tion, there must always be need. It is perhaps the highest, as it is certainly the most difficult and the most attractive, exercise of the mind in scientific work. But it must be controlled by knowledge of what has already been attempted or accomplished, lest it fall into the error of repeating in slightly changed form hypoth- eses which have been already suggested. This danger is, of course, common to all kinds of investigation, but the necessary vagueness of speculation makes it pecu- liarly open to it. Through the variety, however, which has marked the period, there has run one dominant note. The power and brilliancy of Delbriick as an investigator, his im- mense knowledge and the clearness and persuasiveness of his presentation, which pointed him out as the natural co-worker of Brugmann on the Grundriss, have made him easily the first scholar of the period in syntax. Of either the results or the methods of his work in compara- tive syntax no one is competent to speak who is not him- self a comparative philologist, but Delbriick has in his Greek Syntax set the example of applying the same method to a single language, and other scholars have in like manner applied it to Latin. My reasons for believ- ing that the epoch which has been especially characterized by this method is approaching its conclusion have been 1 See, for instance, the programs of Imme, Die Fragesatze nach psycho- logischen Gesichtspunkten eingeteilt und erlautert (Cleve, 1879, 1881), and compare the fruitful results there reached with the ordinary careless three- fold division of interrogative sentences according to the answer expected. 32 INTRODUCTORY AND HISTORICAL suggested incidentally above, and may be put together here in a more formal way. 1. The method of explaining the meaning of an inflected form by means of a Qrundbegriff is an uncon- scious survival of the logical or metaphysical view of language, which made definition the ultimate explana- tion. It received scientific sanction from the etymo- logical system of Bopp and Curtius, which analyzed inflected forms into significant elements and suggested the substitution of a psychological for the metaphysical content of the definition. 2. It belongs to the earlier and looser period of the study of origins, when, in the first application of the comparative method to living organisms, biological sci- ence was sufficiently occupied with the tracing of rela- tionships between species by finding or inferring a common ancestor. That method of study has long ago given place in biology to more exact methods and other problems, as it has in morphology, and as it must in syntax. 3. The permanent value of the method — and it is great — lies in its introduction of psychological expla- nation and in the emphasis which it places upon the historical method. These are indirect and suggestive, but it is now some years since any direct result in the syntax of the modes has been attained which commands general assent. Its continued use in Latin has brought out many new names for certain groupings of subjunc- tive usage — anticipatory, Active, polemical, obligation, stipulation — and these have their value, which I would not underestimate, in bringing into clearer light the common elements in groups of usage, but they have not proved, and in the nature of the method cannot prove, historical relationships or contribute to the under- 3 33 LATIN SYNTAX standing of the problem of inflection. On the contrary, this loose genetic method tends to substitute vague phrases and vague references to an Indo-European origin for precise knowledge. The method is becom- ing barren of results. To these indications that we are at the end of a period may be added some others more general in char- acter. They are to be seen in the dissatisfaction of the general philological public with the vagueuess and lack of intelligibility of syntactical work; in the in- creasing inclination to turn back to general principles, as shown in Koppin's programs, in Ries's book, and in the first part of Ziemer's Streifziige ; in the reaching out after new methods, illustrated by Ziemer and Probst and Ries; and perhaps most distinctly in the falling back upon simpler methods, upon description and sta- tistics, in Wb'lfflin's Archiv and other work of the same school. These are indications which students of syntax are bound to note and to interpret. New epochs in science come as the result of the in- jection of new ideas into the thinking of scientific students. The revolution is most striking when it is brought about by some single leader or by some one book, but it is quite as often a gradual change, the effect of many causes working together. Such influ- ences are now at work upon syntax from the side of morphology and phonetics, of general linguistic science and of semantics. So far as one may venture to antici- pate the course of syntactical thought, they indicate that syntax is turning away from genetic problems and from grammatical rules and will be directed to the dis- covery and illustration of the psychological phenomena which underlie the grouping of words, and to the inves- tigation of the laws which have governed the process 34 INTRODUCTORY AND HISTORICAL of group -making. The method by which this research will be conducted will involve a return to more minute observation of the details of the process. Such a change as this may be hastened and made more secure and regular by a discussion of the funda- mental principles of syntax; it is in the hope of con- tributing to such a discussion that the following papers have been put together. 35 II THE GROUPING OF CONCEPTS The psychology of speech * may be divided into two parts: first, the psychology of sound-production, which has to do with the reflex and unconscious action of the organs of speech, including the memory of such action by which sounds are repeated and in which they are associated and, so to speak, preserved ; and, second, the psychology of the train of thought which consciously accompanies utterance and which speech is felt to ex- press. It is with the latter, of course, that syntax deals, though not with all of it. So far as words can be separated for scientific purposes from the sentences in which they are used, their meaning belongs to seman- tics. Syntax may claim as its special province 2 on the 1 The main points in this chapter have been anticipated by Wundt, Vdlkerpsychologie, I, 2, 234 ff., in pages which I read with mingled dismay and pleasure. But such unity as the papers in this book possess is depend- ent upon this chapter, and I have therefore left it substantially as it was written out some two years ago. Perhaps the fact that it is a venture of a philologist into the field of psychology will justify the repetition, as I had hoped that it would excuse the mistakes which I have no doubt made in an unfamiliar science. 2 I have not cared to discuss here or later the delimitation of syntax from semasiology, though the question of distinguishing the meaning of a word from its function as an inflectional form might come up below (Chap. IX). The matter has, indeed, been so clearly treated by Ries that it may be considered to be settled. 36 THE GROUPING OF CONCEPTS psychological side only those concepts which find ex- pression in and through the combination of words, and the fundamental question for syntax is the question as to the nature of these concepts. In order to under- stand them it is necessary first to trace the mental processes which precede, accompany and follow the utterance of a group of words. Both thought and utterance, going on at the same time, are so rapid as to seem almost automatic. So far as we give attention consciously to the process, it is ordinarily rather to the process of utterance, to the selection of words, than to the selection of concepts. A hindrance or obstacle to easy speech seems to be a difficulty in the finding of words rather than in the finding of ideas. But the fact that thought also calls for selection and arrangement is apparent enough upon a moment's reflection, especially when the thought is somewhat careful, as in writing upon a serious topic. In such a case we begin with the general subject, which lies in the mind in vague and general form. The first step is the directing of the attention upon some special aspect, which is one of many aspects sug- gested to the mind by association, and the separation of this from the rest. It is not, however, a complete separation, for the mind retains a sense of the relation of the special aspect to the general topic. Then, as a second step, either some other special aspect is brought before the attention, or the aspect first noticed is in its turn subjected to a like process. The mind again selects from the concepts which are suggested by asso- ciation one to which the attention is given, and this also is viewed apart from others and yet at the same time in relation to the more general concept of which it formed a part. This process of arrangement, by successive 37 LATIN SYNTAX analysis, will be carried far before the thought is suffi- ciently digested for expression in written words. In this elaborate preparation of thought for speech three steps may be distinguished. First, there is an analysis of the germ-concept, by means of the fixing of the attention upon some single one of the concepts sug- gested by association. Second, the suggested concept is not so much separated from the germ as viewed in its relation to the germ; the mind is always conscious of the relation, and it is because of the nature of the relation that the mind selects that particular concept rather than some other. Third, there is through the whole process a more or less conscious reference back to the original idea, a comparison of the analyzed group with the original unanalyzed germ in order to see how far the result of analysis has corresponded to the preconceived idea. These three steps go on simul- taneously, though in a long analysis the third process may be somewhat intermittent. In ordinary experience the arrangement of thought is of course less orderly than this. The association may be aided by accumulated notes which disturb the regu- larity of the process, or the subject may be one which requires little analysis, like a narrative of events already connected by the order of their occurrence. In the still more rapid processes of conversation, where the thought is constantly diverted by the suggestions of others, it is quite impossible to follow the action of the mind into details. The interaction of associations from within and from without, the variety in our modes of thinking under different conditions, and the differences which result from the subject of thought all result in great variations in the arrangement of ideas. But the essen- tials of the process, namely, the existence of the thought 38 THE GROUPING OF CONCEPTS as a whole, as a germ-concept, 1 and the immediate analysis of it, seem to remain the same, and may often be detected by observation of our own thought preced- ing speech. We may, for example, in listening to another person, believe that we detect an error in state- ment or a fallacy in reasoning. At the first instant this is entirely vague, scarcely more than would find expression in the exclamation "Wrong!" Then we may become conscious of the nature of the correction or the counter-argument in a like vague way, that is, as a whole, unanalyzed. If the circumstances make interruption permissible, an impulsive person will often break in while his own thought is as yet unanalyzed, and will find himself for an instant conscious of what he wishes to say and yet unable to say it. At such a moment one may detect in his own mind a kind of whirl of thought, almost a mental dizziness, as the swift process of analysis goes on. Then the thought begins to clear itself and to find expression in words. Or it may be, if interruption is not possible and we go on listening to the speaker, that the germ-concept is lost and, as the phrase is, we "forget what we were going to say." So when we see a child doing some- thing dangerous or otherwise undesirable, the impulse to interfere is at first no more than a willed negative, ne without a verb ; then the process of analysis begins, if clearness demands anything more than a prohibition like don't. The germ of a question is in like manner often to be detected before analysis, at first in the form of a mere desire for information excited by a suggestion from without, and then associated with the circum- stances or the speech which excited it. Many questions need but slight analysis; they consist of an interroga- 1 Gesammtvorstellung is the term used by Wundt. 39 LATIN SYNTAX tive sign and of a phrase which refers to the source that suggested the question. An interrogative sentence which contains more than this is something more than a question, involving also assertion or argument. A narrative, especially a short narrative, is remembered as a whole and may be so recalled by a single phrase, as "P's whistling story" or Livy's account of the de- struction of Alba Longa. Just so a picture may be remembered as a whole. The analysis which precedes the telling of a story is particularly simple and easy to follow because the association is one of mere succession and the effort of the mind in fixing upon the order of events is a familiar one. Even the briefest remarks or comments, uttered in the midst of conversation, may frequently be caught by careful observation at the in- stant before analysis, and we may detect the germ- concept and may be aware of some rapid process of thought which must precede speech, though it may be impossible to follow it in all its details. The process of analysis is dependent upon associa- tion. It is through association that the mind passes from the original germ-concept to the separate concepts. Among the concepts thus suggested the choice of the one upon which the attention shall settle is determined by its relation to the germ or its relation to the general course of thought. Each concept which is singled out from the germ for special attention stands, therefore, in a definite relation to the germ, a relation determined by the line of association. And because each is related to the original concept, each is also related to every other and a concept-group is thus constituted, bound together by a network of mutual relations. All the elements of such a group are held in consciousness at the same time, though the attention is not directed upon all at 40 THE GROUPING OF CONCEPTS once. The group as a whole is consciously felt, with its relation to the preceding group and its associations reaching out toward the groups which are to follow. By them it is a member of a still larger group of groups. The separate concepts in which the analysis ends are also held in mind, and the attention is directed upon them in succession. And the relations also are present in consciousness, though less clearly and distinctly, serving to direct the analysis and to retain the sense of its unity. If it be asked what brings the analysis to a close, the answer will be that speech may begin at almost any point in the process. The prohibition may be expressed by "Don't!" the question by "What?" or "Who?" The story may be introduced by " That reminds me," and in fact such phrases, which are attempts to speak before the analysis is complete, are very common in colloquial speech ; thus quid ? is used as an introductory question or quid again f quid faciam ? precedes a more detailed question. So a whole oration might be in an imperfect way expressed in a single sentence, " Catiline is danger- ous," "Archias deserves citizenship." But ordinarily the process will go on until the analysis is complete enough to exhibit all that, to our thinking, was in- volved in the original germ. The aim and end are the same, the satisfaction of the desire to express in its details the concept which was originally in mind. The process which I have been attempting to describe precedes speech. In its outline and in most of its details it must be completed before the words which are to suggest it to the hearer begin to be uttered. The effect of hurrying forward the words before the analysis is fairly complete is to make the sentence confused in its ending ; this is one of the most frequent causes of 41 LATIN SYNTAX confused and inaccurate sentence-structure. But when the analysis is completed, the fitting of sufficiently accurate words to the grouped concepts is almost auto- matic. Because thinking is so generally associated with words, the analysis is instinctively directed toward concepts which have been before associated with words. These are the natural ends toward which the analysis moves, and when the analysis reaches this point, the words are already suggested. The only thing neces- sary, therefore, during utterance is that the concepts, grouped by their relations, should pass in succession before the mind, or, more precisely, that the attention should be directed upon them in the succession which their grouping suggests. This operation is so nearly automatic that the conscious activity of the mind may at the same time call up the next group and perform upon it the necessary analysis. After a group of words is uttered, a reverse process, one of recombination, begins. As soon as the attention is withdrawn from the distinct concepts, they begin, as it were, to sink back into the unanalyzed condition. The sentence which is in process of utterance is held until it is finished, but the sentence which had been uttered just before is held in the mind less distinctly and the preceding sentence is still less clear. If it is recalled by an effort of memory, the words will prob- ably be called up first and the concepts will be recalled by means of the words, or the general group-concept will be recalled and analyzed a second time. In a long speech, lasting for several minutes, the speaker will remember what he has said only in the most general way. The thought will lie in his mind very much as it was before the analysis and utterance, except that the fact that the groups have been once analyzed will 42 THE GROUPING OF CONCEPTS render a second analysis easier. It is in this unana- lyzed or only partially analyzed form that any connected series of thoughts lies in the mind ready to be recalled. And as short sentences, prohibitions or brief questions, or short anecdotes or illustrations, may by a little self- examination be found to arise in the mind for an instant as wholes, so after they have been uttered in words they may for an instant be detected in the mind as wholes, undivided groups, accompanied by a sense of satisfac- tion, just as before utterance they were accompanied by a feeling of desire which called for satisfaction. The three successive stages are therefore these : first, the group-concept is analyzed into a group of concepts with their connecting relations; second, the group in its analyzed form is clearly held in mind while the associated words are uttered ; third, the group of con- cepts immediately begins to fade back into its unana- lyzed form. The action of the hearer's mind most nearly resembles the second and third of these stages. The uttered words excite in his mind the associated concepts with all their suggestions and implications of relation, and these he instantly begins to combine into a group, which, if the whole operation is skilfully performed, will be essentially the same as the unanalyzed group in the mind of the speaker. This is done so immediately and unconsciously that, if the attention of the hearer is fixed upon the thought, he will often be quite unaware of the analyzed elements and be conscious only of the result of the recombination, the group-concept. It is the power of performing this process of recombination rapidly and unconsciously which enables a practised reader to grasp whole sentences or even whole para- graphs at a glance; he dispenses with the laborious 43 LATIN SYNTAX analysis and by catching a word here and a word there is able to divine the group. Where the subject is unfamiliar or the language is foreign and only imper- fectly known, the reader or hearer must perform the recombination slowly and carefully. To some extent, also, the mind of a quick hearer or reader is at work upon the unexpressed thought, per- forming also, as does the mind of the speaker, an analy- sis ahead of the point of utterance and anticipating to some extent and tentatively the course of the speaker's thought. This enables the hearer to grasp the group of concepts, when it is suggested by words, to some extent as a whole, and still further lessens the amount of analysis necessary. The psychology of speech has not hitherto J occupied a large space in the standard works on psychology, but some confirmation of these views, which are the result of an attempt to understand the basis of syntactical expression, may be had from the extremely interesting chapter in James's Psychology, 2 entitled "The Stream of Thought." The fact is there stated and illustrated that two elements are present in the succession of con- cepts which follow each other in the mind. Of these the more obvious consists of the more definite and, so to speak, concrete concepts. When one attempts to arrest the stream of thought, as, for example, in answer to the question "What are you thinking about?" the attention is likely to be fixed upon some definite object, — a thing, a person, an institution, an event. This will be the case even when the definite object is really quite secondary, when the thought was really fixed upon a question of duty or expediency. But a closer 1 Before the appearance of Wundt's Volkerpsychologie. 2 The Principles of Psychology, I, chap, ix, pp. 224 ff. 44 THE GROUPING OF CONCEPTS self-examination shows that such definite concepts are always connected by relations which form a part, and an important part, of the stream of thought. To these two elements James gives the names substantive and transitive. It is evident that the substantive concepts correspond in general to the separate and distinct concepts into which a germ-concept is analyzed, and that the transitive elements, which are also themselves concepts, correspond to the relations which bind the concepts into a group. Speaking broadly, the substantive concepts, the sepa- rate concepts upon which the analysis comes to an end, are expressed in language by words, and the laws which govern their association with words make up the science of word-meanings. The science of word-combinations has to do with the meaning of the whole group and with the relations by which its parts are held together. The following papers will contain some illustrations of the ways in which these general characteristics of thought-structure are repeated in sentence-structure. Meanwhile some general observations are in place here, with reference to the emphasis which these facts may lead us to place upon certain general principles of syn- tactical investigation. 1. The unity of a concept-group is not something produced by the act of expressing it in words, nor is it in any way the result of the putting of words together. It is, on the contrary, antecedent to expression, and is an inherent element of thought. The various defini- tions of the sentence which imply that a sentence ex- presses the completion of an act of combination define the process from the wrong end ; the sentence expresses the result of an analysis, and ever} r thing in it which binds the words together is the sign of the original 45 LATIN SYNTAX unity, not the means of producing unity. This is true also of sentences or verbs in the paratactic structure (see Chap. VI); the relation is not created by putting the verbs together, but is the reason for combining them into one group. The principle of first importance in following the growth of syntactical expression is to recognize the fact that all that finds expression exists first in thought and must be felt with some degree of distinctness before it can find any kind of expression. All this justifies the use of the term semasiology and the application of semantic laws and methods to the association between the concepts of relation and the means of syntactical expression. But the semasiology of word-groups is more difficult than the semasiology of single words, both because the concepts are more shift- ing and evanescent, and because they are expressed by means so varied and complex as sentence-structure. 2. The importance of studying words in groups, which is often recognized, is further emphasized by what has been said in regard to the action of the mind of speaker and hearer. The isolation of a word for detailed study must be followed and corrected by a study of it in its true condition as part of a group. The fact that each concept exists for the sake of the group, that the speaker is endeavoring to express the group-concept and is using the analyzed group of con- cepts only for this purpose, makes it possible for him to use words with greater freedom ; meanings need not be precise, because they are limited by the other concepts of the group. This applies to inflectional forms as well as to words. The precision which they seem to have is often due to the rest of the group, and they can there- fore be properly interpreted only when the limitations which surround them are taken fully into account. 46 THE GROUPING OF CONCEPTS 3. The relations between the concepts are dependent upon the associations between them, and these, when they are not purely accidental, depend upon the nature of the concepts themselves. The relation of cause and effect can exist only between objects in the same sphere of causation, not, for instance, between a cloud and a house; purpose implies personality, a time-relation im- plies a time-word. In other words, the meaning of words has much to do with syntactical expression. There is sometimes an apparent disposition to separate these two means of expression, single words with their meaning being put into one science, while syntax is treated as if inflection were independent of word-mean- ing, and as if the syntactical forms were shells which could be filled with any content without altering their character. This is, of course, the result of the perfectly proper effort to isolate syntactical expression in order to study it without the disturbance of anything foreign to it. But it is quite certain that syntax can be under- stood only when it is studied in the closest association with word-meaning and that a large field of work is opening out in this direction. 47 Ill THE MEANS OF EXPRESSING RELATIONS The means employed in language for the expression of the relations between concepts are in part the same as the means at the command of language for express- ing the concepts themselves, and in linguistic discussions the two are often treated together without distinction. It is not, in fact, possible to draw a clear line of distinc- tion. In general, single words correspond to distinct substantive concepts and the study of their associa- tion with such concepts belongs to lexicography and semantics. But when a relation, as a result of frequent use, comes to be clearly and vividly felt, it has itself become a concept, apparently much as any concept is formed from percepts, and may then be expressed by a single word — a preposition or conjunction — the study of which belongs alike to semantics and to syntax. In- flected words also have both meaning and function and, just as in Latin the stem is never found without an inflectional ending, so the meaning and the function always go together and are inseparable. Even the parts of speech have to do partly with word-meaning and partly with syntactical function, since the differ- entiation is brought about within the sentence in the effort to express relation. The verb does not differ from the noun in meaning only, but also in use. The 48 THE MEANS OF EXPRESSING RELATIONS ordinary definition of the verb as a word which denotes action or state and of the noun as the name of a person or thing is evidently one-sided and defective, since a noun may denote action or existence and a verb-form may be the name of a thing. Another and truer dis- tinction is based upon the use to which words are put in combination, and the differentiation of parts of speech is a means of expressing at the same time substantive concepts and concepts of relation. No definite line, therefore, can be drawn between the means employed in language for suggesting concepts and those which are at the command of language for the expression of rela- tions. But it is nevertheless worth while to group together those characteristics of speech which have been more distinctly appropriated to the expression of relation, as a preliminary to the consideration of the process by which they have been adapted to such expression. The musical elements of speech have to do chiefly with emotion, not with the intellectual side of speech. But a change of tone, indicating primarily a change of emotional attitude, may also serve to suggest in a very general way the relation of that which follows to that which precedes the change. This is especially clear in the utterance of a parenthetic explanation or comment, where the lowered tone and perhaps quickened time aid in suggesting the parenthetic and unimportant char- acter of the thought. A change of tone will also accompany and partially express a change from argu- ment to narrative or the reverse, and may thus vaguely suggest even the nature of the relation between two groups of words. The tone in which a conditional clause is uttered differs from the tone of a causal clause. The pauses between groups of words are more directly 4 49 LATIN SYNTAX contributory to an understanding of the grouping of concepts. They are in origin physiological, that is, they are due to the necessity of refilling the lungs. If it were possible to conceive of thought as a perfectly homogeneous stream, as unvarying in quality as it is unbroken in its continuity, the pauses would occur at substantially equal intervals. But as thought is con- tinually varying in quality and concepts occur in groups, the pauses in speech have been accommodated to the suggestion of groupings and occur not regularly but at varying intervals, so as to mark the fact that one group of concepts is completed and another is about to begin. Further, the length of the pause may indicate the size and complexity of the group, especially in very deliberate speech when the pause is utilized by the speaker for the analysis of the next group. The slight pause at the end of a clause, such as is indicated in printed language by a comma, suggests in conjunction with the raised inflection of voice the conclusion of a small group which is part of a larger group ; the longer pause and falling inflection mark the close of a more fully completed group, and the end of a still larger group of groups is marked by a longer pause and a change of tone, as it is indicated in print by a new paragraph. The pauses thus mark groups and suggest slightly the relations between them, though they do not indi- cate the relations within the group, and they illustrate well, in an elementary way, the process of adaptation ; in their origin they have nothing to do with speech as a means of expression, but are the result of physiological conditions, yet they become one of the most funda- mental means of indicating the nature of the stream of thought. 50 THE MEANS OF EXPRESSING RELATIONS The three most direct means of expressing relations are inflection, single words, and groupings of words. A natural starting-point for considering the nature of inflection would be its origin and, if there were a satis- factory and generally accepted theory upon this impor- tant point, it would simplify some of the problems <.f syntax. But there is no such theory. The explana- tion of inflection as due entirely to a process of agglu- tination, once a part of the orthodox philology, is quite certainly no longer accepted without serious question as to its value. The grounds for this change of opinion it is not for a syritacticist to discuss, but in general the theory appears to lack a good psychological basis, to involve an appeal to laws and forces other than those which are in operation in historic periods and to be too sweeping in its conclusions. Composition of some kind must be supposed to have taken place in order to account for some verb forms — e. g., for fueram or the imperfect in -bam — but composition can scarcely ex- plain all the phenomena. At any rate, the acquiescence of morphologists in the agglutinative theory, so far as acquiescence exists, is dissatisfied and agnostic, the result mainly of the fact that no substitute has been proposed, and the attempts 2 that have been made within the last few years to explain Indo-European inflection look in other directions. If this view of current opin- ion seems too unhopeful, it must at the least be said that syntax has at present little to gain from the theories and speculations in regard to the general nature of in- 1 Bergaigne, de conjunctivi et optativi in Indoeuropaeis Unguis informatione et vi antiquissima ; E. W. Fay, Agglutination and Adaptation, Amer. Journ. of Philol., XV, 4, XVI, 1 ; Audonin, de la declinaison dans ies langues Indoeurope'ennes, Paris, 1899; Greenough, in Harvard Studies in Classical Philology, X, 1 . 51 LATIN SYNTAX flection. They are more likely to mislead than to fur- nish a substantial basis for syntactical work. In these circumstances it is all the more desirable to get as broad and clear a view as possible of the nature of inflection as it appears in historical times. Looked at broadly, the most striking characteristic of Latin inflection is that it is not a system, but is on the con- trary highly unsystematic. This statement may seem unnecessary, but the opposite opinion, held more or less consciously, underlies much of the recent syntactical work, especially in America. The impression of sys- tem comes, no doubt, from the way in which we learn the facts of inflection. For the purposes of teaching, the grammars very properly emphasize as much as pos- sible such measure of system as Latin inflection permits, producing at the beginning of one's acquaintance with Latin the impression of a series of graded forms and meanings covering most accurately and completely the whole range of expression. But it is obvious that this is a false impression and, so far as we retain it, we are building upon a wrong foundation. Neither the forms nor the meanings are systematic. The perfect stem is not one, but is an irregular mixture of at least four different stems: the reduplicated (cecidi), the length- ened (ueni), the s stem (dixi) and the -ui form (amaui). Of these a few verbs have two (parsi, peperci) and the rest have one or another termination for reasons which are apparently phonetic and have nothing to do with the meaning either of the verb or of the termination. The subjunctive has two formations (dicam, amem) with a mixture of optative forms, also without distinc- tion of meaning. Or, taking the tenses in order, the present stem has many variations in form, some of them significant, others apparently without meaning ; the im- 52 THE MEANS OF EXPRESSING RELATIONS perfect indicative is an Italic formation ; the imperfect subjunctive is entirely different and is of unknown origin; the future is either an Italic formation {-bo) or it is a modal form, different in the different persons (-am, -es). The personal endings are not more system- atic; the first and second singular may be connected with the pronouns, but the second and third plural are inexplicable. To these illustrations from verbal inflec- tion must be added the irregularities of pronouns and of the third declension of nouns, the immense variety of word-building suffixes and all the irregularities which have been brought together by Osthoff in his papei on Suppletivwesen. But it is not necessary to heap up evidence on this point. A glance at the facts of Latin morphology as they are presented in any full Latin grammar, or in Brugmann's Grundriss, or in Lindsay's "Latin Language," where large masses of facts which defy classification are brought together, furnishes con- vincing evidence that irregularity and absence of system are not merely occasional but are the fundamental char- acteristics of Latin form-building. It is the regularity that is unusual and exceptional. The same absence of system appears in the mean- ing and use of inflected words. Forms do not have single meanings but many and varied meanings, which do not combine into a system but overlap one another. The present tense of the indicative overlaps the future (quam mox seco ? quid ago f) and the past (historical present) ; the imperfect is often indistinguishable from the perfect and it has conative and inchoative uses which are really modal. The future is often used to express determination (ibo), thus confusing, as the im- perfect does, the distinction between mode and tense. The perfect, from at least four distinct sources, has two 53 LATIN SYNTAX distinct uses, which, however, do not coincide with any of its forms. Tables of tenses of the indicative, in which present, future, past combine with beginning, duration, completion into a symmetrical scheme are far from the facts of language. 1 In the subjunctive the tenses are so confused and overlaid by the modal force that it is a question whether they may properly be called tenses. The scheme of case-constructions given, e.g., in Draeger is so elaborate as to cover apparently all possible uses ; but it must be remembered that this is a presentation of all case -uses of all nouns and pro- nouns. The range of a single noun is quite limited. To take an example almost at random, tempus is not used in the accusative of the place to which, in the dative of possessor, of agent, in the ethical dative, in the genitive subjective or the genitive of value or after refert, in the ablative of place, of source, of man- ner, of accompaniment, of price or of agent, to mention only some of the more common constructions. With reference to meanings, it might be said that almost any noun is extremely defective in its case -uses, just as many nouns are defective in case-forms. The unsystematic character of inflection appears even more clearly in word-formation. The suffixes used in the building of words are so irregular that it is almost impossible to match forms with any system of meanings. For example, the terminations denoting action are given in a particularly careful grammar 2 as a, io, ia, i> a double expression of the same concept-group. So, i y sequere illos : ne morere, M. G., 1301 ; uide ne ties in expectatione : ne ilium animi excrucies, M. G., 1280; da . . . mi . . . ueniam : ignosce : iratii nesies, Amph., 924; emitte sodes : ne enices fame : sine ire pastum. Pen., 318. In some of these cases, where the change of ex- pression is greatest, a trace of purpose appears to under- lie the words; thus M. G., 1361, might easily be taken to be "follow, them in order not to delay them" or M. G., 1280, might be "don't keep her waiting lest you torment her." In spite of the grammatical indepen- dence, the relation between the sentences is felt, though not distinctly enough for certain identification. This is equally true where the preceding sentence is a statement. The statement gives the ground for the prohibition or the prohibition expresses the object, i. the purpose, of the statement. Thus noui: ne doceas, AuL, 241 ; non morabitur : proin tu ne quo abeas longius^ Men., 327 (the relation is in part expressed by proin) ; istie homo rabiosus habitus est in Alide : ne tu quod istic fabuletur auris inmittas tuas, Cap., 548. In some in- stances the ne sentence expresses the purpose of the speaker in making the previous statement, not the purpose of the act stated: dormio : ne occlamites, Cure, 183, which is either " I 'm asleep : don't make such a row " or " I tell you that I am asleep in order to induce you to stop your shouting." So uapulare ego te uehe- menter iubeo : ne me territes, Cure., 568, which differs from uapula : ne me territes only in having the verb of saying expressed. In Cure., 565, nil (agit) apud me quidem — ne facias testis — neque equidem debeo quia- 163 LATIN SYNTAX quam, the ne clause is a parenthetic insertion and shows more clearly its connection with clauses like ne erres, ne frustra sis, ne te mover, which are sometimes called par- enthetic clauses of purpose. Instances of this kind are frequent in conversational Latin. It is probably in part upon their occurrence that the statements quoted above (p. 116), that in parataxis two independent sentences come to be thought of as one, are founded, but the facts do not bear out that interpretation. The relation be- tween the two sentences is not expressed in any word, and therefore the two sentences may be regarded as in- dependent in the grammatical sense, but the relation be- tween the two concept-groups is just as real and was probably as strongly felt as if it had been expressed in some single word. It is suggested by the juxtaposition of the sentences and was doubtless felt in the tone and the length of the pause, though these can now only par- tially be recovered. In Amph., 924, the editors separate da . . mi . . ueniam, ignosce, irata ne sies by commas, but in Pers., 318, emitte sodes : ne enices fame : sine ire pastum, they use colons ; the length of the pause and the inflec- tion of the voice, however, cannot be very different in the two cases. These cases therefore illustrate the kind of sentence-relation in which there is no expression of the relation except by the musical elements of speech and by the mere contiguity, while the fact of the rela- tion itself is nevertheless perfectly certain. No one could use noui : ne doceas without relation between the two thoughts nor could a hearer easily fail to grasp the relation in a general way. The nature of the relation, however, is undefined, or at least is not defined in terms which discriminate between coordination or parataxis and subordination. It is sufficiently represented either by " I know and don't want you to teach me " or by " I 164 SUBORDINATING CONJUNCTIONS IN LATIN know, and I say this lest you should try to teach me " or by "I know; therefore (proin^) don'1 teach me." The attempt to dehne this relation psychological lv by deciding upon one or another of these ways of express- ing it would result merely in a forced and arbitrary selection which would not carry conviction to another interpreter and which would involve the neglect of other equally real though perhaps less evident elements. The relation is real, is felt by speaker and hearer, but it is not defined. It is not the result of some process of gradual melting into one, some coming to be thought as one ; that is, expressed in a rather inaccurate way, a de- scription of the process by which the sentences are brought together, but not in any way a description of the psychological situation. It is not possible with this kind of parataxis, any more than it is possible with defining parataxis, to draw a line between the paratactic structure and the full subordina- tion, and for the same reason, because it is impossible without the help of the spoken language to tell just when ne begins to be associated with the concept of re- lation. But illustrations may easily be found among clearly subordinate ne clauses, which are plainly like the independent sentences given above. Thus AuL, 340, si quid uti uoles, domo abs te adferto, ne operant perdas poscere, expresses in the imperative one view of the command, in the ne clause another ; compare i, sequere illos, ne morere, with intro abite atque haec cito celerate, ne mora quae sit, cocus quom ueniat (Pseud., 168). With Cure, 568, compare Poen., 1155, audin tu, patrue ? dico, ne dictum neges. Similar cases are Pud., 443, dabitur tibi aqua, ne nequiquam me ames ; Aid., 54, ocidos . . ., ecfodiam tibi, ne me obseruare possis ; Pud., 1013, at ego hinc offlectam nauem, ne quo abeas ; Cas., 394, nunc tu, . . ., 165 LATIN SYNTAX ne a me memores malitiose de hac re factum aut suspices, tibi permilto : tute sorti, " don't say or think that I 've cheated, for I leave the drawing of lots to you." These are all, of course, ne clauses and are to be so taken, but a slight change in the phrase or in the thought would put them back into ne sentences, independent of the leading clause so far as structure is concerned. The acquisition of the subordinating function by ni, through association with a conditional clause, has been followed in the early Latin by Oskar Brugmann. 1 Its history is quite different from that of ne ; it does not pass through the stage of defining parataxis, evolving its own leading clause out of itself, nor has it precisely the same kind of relation to its context that a prohibi- tion bears to a preceding imperative or indicative sen- tence. Between a ne clause and a preceding or following statement there is no resemblance in the form of the sentences ; the relation expresses itself in other ways. But the relation of the ni sentence to its context, before the conditional function is attached to ni, is one of correlation. The essential point in it is that the resemblance in structure is the result of the speaker's retaining the first member vividly in memory while he is uttering the second member and by this means induc- ing the hearer to recall the first member as he hears the second. The two concept-groups are thus set in con- trast with each other. This does not imply that the relation between them is necessarily that of protasis and apodosis ; it is an undefined relation which may be one of mere comparison as to quality (talis — qualis) or as to quantity (tantus — quantus) or degree {tarn — quam). In the case of ni it was in part at least helped toward precision by the subjunctive mode, which of itself sug- 1 Ueber den Gebrauch des Condicionalen Ni . . ., Leipzig, 1887. 166 SUBORDINATING CONJUNCTIONS IN LATIN gests supposition, concession, condition. The relation between the ni sentence and the corresponding and cor- related member was therefore existent without regard to the presence of ni and was imperfectly suggested. Its transfer to ni was due in part to the position of the negative before the verb and frequently at the head of the sentence, and in part to the fact that, in a condi- tion, the negative, if it belongs to the sentence as a whole and not to some single word in it, naturally asso- ciates itself with the conditional particle because that particle also goes with the clause as a whole. The negative particle negatives the conditional relation and therefore the two are associated together. The association of quamuis as a conjunction with a concessive sentence differs from that of ni in two re- spects. The relation between the sentences is not cor- relative and quamuis is not, as ni is, a necessary part of the sentence with which it is associated. As to the relation between the two parts of the sentence, it is expressed, when quamuis is not used, in the mode only, being in this one respect like the ni clauses with the subjunctive. Not that the subjunctive is a mode of subordination, but the use of the subjunctive is in many cases by its necessary relation to the context ex- pressive of a proviso or concession. But there is no suggestion in the form of sentence of any kind of cor- relation, and it is unsafe to suppose that sentences have been correlated unless the supposition is supported by a general likeness in structure between the two members. The other difference between ni and quamuis is that the former is necessary to the sentence in which it occurs ; without it the sentence would have an entirely different meaning. This is true also of quin and of ne. But quamuis is an addition to a sentence which is a fairly 167 LATIN SYNTAX adequate expression of the thought without quamuis. The addition of this word is only a further expression of an idea already contained by implication in the mode or in the mode with other accessory expressions. In this respect quamuis is like licet and its addition to the sentence is the result of expansion. It would be in- exact to call it defining parataxis, but it is of the same general nature, definition by expansion. Probably its verbal nature is not wholly lost until it begins to be confused with quamquam, and the verbal force would perhaps predispose it to a use so closely akin to defini- tion by the addition of a verb. Neither simul nor modo becomes wholly conjunctional and this fact perhaps accounts for the small amount of attention that has been given to their use. Neither appears to be in use alone as a conjunction in Plautus. Modo is associated with dura and simul with atque (ac) and each acquires its conjunctional force by the asso- ciation. In Plautus dura is already in free use and ac, atque is used not infrequently in correlation with other words, aeque, item, aliter. There is a difference, how- ever, in the causes which lead to the association. Modo is added to the dum clause as an adverb to express more fully the idea of limitation which is partly implied in dum, though not with sufficient distinct- ness. Then, just as the negative force of ni leads to its close association with the conditional relation and so with si in nisi, so the limiting meaning of modo belongs to the whole clause of proviso and especially to the conjunction which introduces the proviso. Thus it comes to be compounded with dum. On the other hand, simul is a necessary part of the sentence; it forms the necessary correlative to ac, atque, which is not used without some correlative. The composition 168 SUBORDINATING CONJUNCTIONS IX LATIN or close connection of timid with ac is therefore like the composition of post-quam, prius-quam, not like dummodo. The partial displacement of dum and dummodo by modo is due to the fact that dum, though it may be used with a clause of proviso, has other and more important uses. The element of proviso is more precisely expressed by modo, which thus becomes especially the bearer of this kind of relational concept. In like manner the particu- lar meaning of simul ac, which distinguishes it from other temporal conjunctions, is the element of simulta- neousness; this is not expressed with the same degree of clearness by cum or cum extemplo or postquam or ubi, and though the relative force, the subordinating func- tion, resides in ac rather than in simul, the more definite and as it were more noticeable element is that which simul supplies. The relative force therefore passes over to simul and it alone expresses both elements. These two words, then, illustrate the acquisition of the subordinating force by association with other words which already had that force. It is correct enough to say in general that conjunctions have acquired sub- ordinating force by passing through a paratactic stage, but it is worth while to note these exceptions to the general rule. The acquisition of subordinating force by a process different from any of those mentioned above is begun, though it is not carried to completion, in certain uses of atque and et. The latter is the most colorless repre- sentative of a purely coordinating conjunction and atque, in spite of the various demonstrative and strengthening uses, is also in the main a coordinating word. But with either conjunction the relation between the two clauses may be so varied by the content of one clause or the other that it approaches a subordinate relation. 169 LATIN SYNTAX This appears in the use of et or atque after words of likeness or unlikeness, aeque, par, pariter, idem, alius, of which there are many instances from Plautus down : pariter hoc atque alias res soles; germanus pariter animo et corpore ; par ratio cum Lucilio est ac mecum fuit ; aeque amicos et nosmet ipsos diligamus. In all such cases the word of resemblance or difference gives to the relation between the words or clause a shading which in more precise expression would call for the use of quam or some other distinctly subordinating conjunc- tion. The use of atque ut or ac si is similar in charac- ter, as is the use of et to connect two successive events or points of time, frequently supported by a negative or by uix in the first clause and by a difference in tense between the two verbs. This well-known use is espe- cially frequent in Vergil and in the poets after him. It is essentially of the same character as the use by a poet or an imaginative and emotional prose writer of common words in unusual connections, where they often give a peculiarly vivid effect. The movement of the emotional stream of thought is so strong that it is fol- lowed by the reader in masses, with long strides, with- out the need of precise expression ; the writer may thus vary his choice of words more freely and may for the moment give to words a meaning peculiar and almost foreign to their usual sense. It is thus that a co- ordinating conjunction may be used to express a tem- poral relation which in precise speech would require cum or some other conjunction of time. For the moment et becomes a subordinating temporal conjunc- tion or et or atque a subordinating conjunction of com- parison. The overwhelming preponderance of the coordinating use is sufficient to prevent this acquisition of the subordinating function from becoming perma- 170 SUBORDINATING CONJUNCTIONS IN LATIN nent; indeed, in many cases the order indicates that the writer had the ordinary meaning of et or utque in mind while he was nevertheless using it as a subordinating word. But for the instant and to a certain degree the association of these words with a subordinate relation is real, and it deserves notice because in this case also, as in the case of modo and sunul, there is no paratactic stage. The passage from coordination to subordina- tion is immediate, without intervening steps. There are no other Latin conjunctions whose early history can be followed with any considerable degree of certainty through the process of acquiring the subordi- nating function. Probably dum would come nearest to those mentioned above, 1 but there are only two instances, neither of them beyond question, of the correlation of dum — dum and it seems possible that some forms of the dum-cl&use are the result of defining parataxis. In the case of si there is abundant later evidence, but this par- ticle became the regular conditional conjunction long before the time of Plautus. The quis-con junctions and the relative pronoun are all so early that it is not likely that their passage from the interrogative to the relative use can ever be clearly known. So far as the attempt is made, however, to reconstruct this early history by inference, it must be done upon the basis of what is known directly. No other kind of parataxis and no other process of passing from adverb to conjunction should be employed in such a reconstruction than those kinds which can be actually followed with sufficient evidence. Briefly summarized they are as follows : — 1. Two sentences independent of each other in ex- pression may be closely related in thought and the 1 See Richardson, de dum particulae apud priscos scriptores Latinos usu, Leipzig, 1886. 171 LATIN SYNTAX relation may ultimately find expression by becoming attached to a particular word. This is the case with prohibitions following a statement, the relation becom- ing attached to ?ie. 2. Correlation may produce conjunctions, as in ni and doubtless in si and perhaps in dum. This may be the explanation of qui. 3. An element contained by implication in a sentence may be expanded into a leading clause, as is the case with many indirect questions and with nulla causa est quin. 4. The denning addition may itself become a con- junction — quamuis, licet. 5. The subordinating function may be acquired by association with another conjunction — modo, simul — or inherited from the relative pronoun — quod, quom. 6. Coordination may occasionally pass over directly into subordination — e£, atque. Some further light, beside that which may come from detailed study of the paratactic stage, is thrown upon the g'Mzs-conjunctions by a consideration of the direct- ness of their relationship to quis and by an attempt to distinguish between interrogative and indefinite uses. Of these conjunctions a small group — quod, quom, quoniam, quo (with comparatives in a subjunctive clause), quominus — are directly from qui and only indirectly from quis. This is proved most clearly by the absence of interrogative use, and by other evidence also. Thus quod can be plainly followed in Plautus from the relative pronoun in the accusative of compass and extent to the causal conjunction; the loss of definite case-relation and of reference to an antecedent may be traced in detail. 1 The connection of quom with qui is supported by the 1 See Ingersoll, The Latin Conjunction Quod, soon to be published. 172 SUBORDINATING CONJUNCTIONS IN LATIN resemblance between some of the cum constructions and corresponding relative clauses, 1 and quoniam is an exten- sion of quom. There are several different kinds of quo, doubtless of different case-forms, but quo as used in clauses of purpose with the subjunctive and a word in the comparative degree is certainly nothing but the ablative of degree of difference, used exactly as the pronoun qui is used with the subjunctive to express purpose. And quominus is a special variety of this quo, permanently compounded with a particular comparative. Their relationship to quis and to each other may be shown by a stemma : — quis quod qui quom quoniam quo quominus The assumption of the subordinating force took place in the stage between quis and qui; no question of this nature therefore can arise in connection with any of these conjunctions; they inherited the subordinating function from qui. The only question to be asked in regard to the origin of quod, quom or quo is as to the process by which they changed from case-forms of the relative pronoun to relative adverbs. This is a kind of question which, fundamentally, has nothing to do with subordination or with conjunctions. The process was essentially the same as that by which eo, the adverb 1 Hale, The Cum-constructions. 173 LATIN SYNTAX meaning "therefore," was formed from is, though it is complicated by the later acquisition of functions of cause and time. The history of quoniam and quominus is one step further removed from the question of acquir- ing subordinating force. With these conjunctions the process to be considered is that of the specialization of meaning of the conjunctions quom and quo by means of the strengthening of quom to quoniam (or its composi- tion with iam) and the composition of quo with a partic- ular comparative of negative meaning. These two conjunctions, therefore, are to be studied by methods entirely different from those which apply to quod, quom or quo. A somewhat similar table of relationships can be made out for quam and its compounds, though with more uncertainty at some points. The use of quam in inter- rogative sentences, direct and indirect, must be taken to indicate a direct connection with the interrogative quis, but it must also be acknowledged that exclamatory uses may have some connection with the indefinite meanings of quis. The compound quamuis seems, when taken alone, to be from an interrogative quam, but it cannot be separated from quiuis nor quiuis from quilibet. The indefinite force of quamuis and quiuis appears, however, to be due to the whole sentence, quam uis, rather than to quam alone. In the same way quamquam cannot be separated from quisquis ; the indefinite force of either is the result in part of the doubling of the stem, and it is scarcely possible, on the one hand, that quamquam is the result of a doubling of quam, uninfluenced by quisquis, or that it is, on the other hand, a direct derivative from quisquis, uninfluenced by the simple quam. There is less difficulty in regard to tamquam, which is evidently the result of correlation. It appears to be phonetically 174 SUBORDINATING CONJUNCTIONS IN LATIN impossible to derive qutisi from quam-si, but syntacti- cally it behaves as a compound of quam and si might be expected to behave. The connection of quando with the other compounds of quam is entirely uncertain. Omitting these two, the relationship of the more im- portant compounds of quam may be represented as follows : — qms quiuis quisquis quamuis (adv.) tamquam ?ttaro(than) quamquam quamuis (conj.) antequam priusquam postquam The only value of such a table is that it represents graphically the differences in the history of different conjunctions and especially the different points at which they acquire subordinating force. This process takes place in tamquam and quam " than " in the process of shift from the interrogative adverb quam by means of correlation with tarn or with a comparative, but in the case of quamuis the shift to a subordinating conjunc- tion occurs at a later point in its history and by an entirely different process, that of association with a concessive sentence. But antequam and postquam inherit the subordinating function from quam "than/' to which they stand in the same relation as that in which quominus stands to quo with comparatives. If the conjunctional (relative) force of these words be con- sidered apart from their special adverbial shading (con- 175 LATIN SYNTAX cessive, temporal), then quamuis acquires its special adverbial shading first and its conjunctional force later, in consequence of its adverbial force, while postquam acquires conjunctional force first and its special tem- poral force at a later stage and as a result of the use with a comparative word. It is evident that such differ- ences in history require a difference in method of treat- ment. Between quarmiis, with its complicated history, and quamquam, which is a simple derivative from quam by doubling, there is no connection close enough to justify their treatment together. It is now somewhat generally acknowledged that a classification of subordinate clauses by function, as causal, temporal, final, consecutive, etc., whatever may be its value for school-grammars, is of no value or is even misleading in scientific work on historical prin- ciples. The better mode of classification is by the introducing word, pronoun or conjunction. It is there- fore necessary to consider the best method of classifying the subordinating words. There appear to be three methods possible : first, by the case-form of the conjunction ; second, by the kind of paratactic process through which it has passed ; third, by the derivation of the conjunction. The first method was used by Schmalz in the first and second editions of his Syntax (Midler's ITandbuch, II, 2). It has two disadvantages : first, the uncertainty in regard to the case-forms of many of the conjunctions ; second, the fact that the character of the clause is only remotely influenced by the case of the introducing relative word. Thus si may be locative and quom instrumental (so Schmalz), but the clauses are unaffected by this fact, as the relative clause is unaffected in general by the case 176 SUBORDINATING CONJUNCTIONS IN LATIN of qui. This arrangement is apparently dropped in the third edition. 1 The second method of classification, by the kind of paratactic stage through which the sentence passed or, more broadly, by the kind of process through which the word acquired subordinating force, has some distinct advantages. It would bring all conjunctions of what- ever origin into a single scheme, not separating the quis- conjunctions from dion, si, ne and the others of various origin. Further, the character of the conjunction de- pends to a considerable extent upon the nature of its early stages ; these leave distinct traces, e. g., in the negative of the leading clause upon which quin depends, in the difference between ne clauses which are the result of defining parataxis and those which have come about in other ways, in the difference between si clauses after miror and the ordinary protasis. Also, one of the ob- jects in studying the subordinate clause is to understand the process by which it became subordinate, and a classi- fication of conjunctions by their paratactic uses would lead directly toward this end. On the other hand there are some disadvantages, equally distinct, connected with this method. While it is true that some of the pecu- liarities of conjunctions are the result of a peculiar par- atactic structure, other characteristics not less important are to be explained by the origin of the conjunction. All ne clauses are strongly colored by the negative char- acter of ne and this coloring is really more important than the distinction between ne clauses after verbs of special meaning (object clauses) and those after other 1 The same order is retained in the third edition, though the main divisions into accusative, locative, modal and ablative case-forms are not used, and the case forms of some conjunctions are differently given (e. g., quom instrumental, not locative). 12 177 LATIN SYNTAX verbs (purpose clauses). The meaning of licet has as much to do with its conjunctional use as the fact that it was prefixed to a concessive clause as a synthetic defini- tion. It must also be said that at present too little is known about the varieties of parataxis to furnish a solid basis for classification and it is possible that the para- tactic stage of many of the ^Mis-conjunctions and espe- cially of the relative pronoun qui must always remain obscure. .qum -qum t quiuis- quam ?- -quamuis- ' (adv.) -tamquam quam- (than) quis quisquis quamquam ^■quo-ad ? quoad W ? — — ut — quamuis (conj.) ( postquam J antequam | priusquam (abl.) quoniam quominus A third method of classifying conjunctions is by their origin. The disadvantage of this is that it is applicable or at least is valuable only with reference to the con- junctions derived from quis. Conjunctions from other sources have no common starting-point for genealogical 178 SUBORDINATING CONJUNCTIONS IN LATIN classification; they are derived from a noun (mo Jo), a verb (licet), a negative (ne, ni) or from some other source and have nothing in common. The (/wi's-conju na- tions, however, may be arranged in a table as the com- pounds of quam were arranged above and this method leads to a partial classification. This table is of course questionable at various points. The most important distinction suggested by it is the distinction between those conjunctions which are derived from quis with only one intermediate step and those which go through two or more middle stages. Wher- ever an interrogative form corresponding to the conjunc- tion is still in use, it may be taken to indicate that the conjunction is derived from quis through the interroga- tive adverb. This is the case with quin, quam, quoad and ut. But as the forms of the relative pronoun would in general be similar to those of quis, it is always pos- sible that the relative use of the conjunction, e. g., of quam, may have been influenced also by qui, or, in other words, that there may have been a quam from qui, as well as a quam from quis through the interrogative quam. This is very unlikely to have been the case with quin, because the connection between the interrogative and the conjunction is well-marked and the range of quin is narrow. It is more easily supposable in the case of quam, especially if qui acquired its subordinating force through correlation, as tamquam and quam with com- paratives did. On the other hand, quod and quom are never interrogative and the history of quod shows clearly that it comes from the relative ; this appears to be the case also with the ablative quo, though it is less certain. As to quia, it is a much older word than quod and is in Plautus the usual causal conjunction. It is found in a great number of cases beginning the answer to a ques- 179 LATIN SYNTAX tion with quid or some other word for why and this fact must be the starting-point of any theory as to the inter- mediate stage between quis and the conjunction. Nor does quia show in Plautus any trace of those uses through which quod passed from the accusative to the causal meaning ; the accusative of compass and extent is properly a singular. The few cases of quianam ? may perhaps indicate an interrogative use, but it is also pos- sible that the interrogative sense is here due to nam. For these reasons the middle step of quia must be re- garded as doubtful and the same must be said of the intermediate stage of qui. Some of the difficulties pre- sented by the ordinary explanation of the passage from quis to qui have been pointed out above (p. 107). The inclusion of ut in this table of <^is-conjunctions rests upon syntactical grounds. The phonetic difficulty of accounting for the loss of the initial qu sound is stated by Brugmann, Grrundr., II, 772. It is partly met by the parallel of ubi, unde, and si-cubi, ne-cubi, ali-cunde and by the Oscan puz, Umbrian puze. But the main reason for connecting ut, ubi, unde with quis is the fact that all their uses find in this way their most natural explanation. They are all found as in- terrogatives and go through the shift to the relative function precisely as they would do if they were derived from quis. To explain them in any other way is to do violence to a considerable mass of syn- tactical evidence. It is unfortunate that such a dilemma should present itself in the history of so important a conjunction. Upon the basis of this table the principal quis- conjunctions fall into three main classes according to the directness of their connection with quis and the point at which the subordinating function is acquired, 180 SUBORDINATING CONJUNCTIONS IN LATIN and each class is subdivided according to the nature of the shift of meaning : — 1. From a <^a's-adverb, by means of some kind of paratactic association. a. By defining parataxis — quin, some uses of »t with the subjunctive, and perhaps quia. b. Under the influence of an indefinite pronoun — quamuis, quamquam. c. By correlative parataxis — tamquam, quam (than), ut with the indicative, and perhaps qui, ubi, unde, quoad. 2. From quis indirectly, through a word having subordinating force. a. Through qui (the pronoun) — quod, quom and probably the ablative quo. b. Through quam (than) — postquam, antequam, priusquam. 3. From quis by two stages, through a specialized form of a relative conjunction. a. Through quom — quoniam. b. Through the ablative quo — quominus. This tabulation does not, of course, add anything to our knowledge of conjunctions; its object is to analyze the general problem into classes of minor problems. The problem in regard to class 1 is to discover the steps by which an interrogative adverb becomes a conjunction ; in class 2 it is to trace the passage from relative pronoun to conjunction or the acquisition of temporal meaning by composition with adverbs; in class 3 it is to follow the specialization and differentiation of a compound from a simple conjunction. These problems are essentially different and are to be approached in different ways. At the same time the tabulation shows the points at which our knowledge is insufficient. The distinction 181 LATIN SYNTAX between ut with the subjunctive, as the result of defin- ing parataxis, and ut with the indicative, as the result of correlative parataxis, is not made without some study of these uses, but it is assuredly not an established connection; it may be said to be important, if true, and the tabulation is meant to bring out its importance. So also the differences between quom and quoniam, be- tween postquam and quom, between quod and quia, are so great that the first step toward the solution of the problem is to face its nature. The last distinction, especially, between quod and quia, is often overlooked because of the resemblance in function. 182 VIII THE GROUPING OF WORDS The means of expressing the concepts of relation already described — the musical elements of speech, order, inflection and single words — express also the unity of the concept-group. But they do so indirectly and not completely. There is a unity of the group, a homogeneity and harmony running through the whole sentence, which is deeper than these means alone could convey. It is seen at its highest point in idiomatic word-groups, like ut ita dicam, quae cum ita si?it, in which the meaning of individual words is almost lost and the phrase is felt simply as a whole. Less distinctly it is to be seen in longer ut or cum clauses of purpose or cause, in which the meaning of individual words is felt somewhat distinctly and yet the meaning of the whole group, as an expression of purpose or cause, is also felt. Less obviously, but not less trul} r , a similar unity is to be found in every sentence and is a part of the total impression made by the word-group upon the hearer. It is this unification and fusion which the term grouping of words is meant to describe and which it is proposed to consider in this chapter. The process of analysis described above (Chap. II) does not result in a displacement of the germ-concept by the separate members of a concept-group ; on the con- trary, the germ of the thought is retained more or less 183 LATIN SYNTAX clearly in consciousness and there is a frequent, though perhaps intermittent, comparison of the result of analysis with the unanalyzed germ. Or it might be said that the consciousness of the germ guides and determines the analysis. The sense of the unity of the group is there- fore one of the prominent elements of the thought and is the source and cause of the impression of unity which the corresponding word-group makes. At the same time the process of analysis is distinguishing the various ele- ments in the meaning of the germ and, as it were, dis- tributing them for purposes of expression among the different members of the word-group. The process may be compared to expansion and distribution, or may be called a transfer of function from the word-group as a whole to the individual members of the group. It is antecedently probable that both aspects of this process, both the unity and the distribution of meaning, will be to some measure reflected in the growth of sentence -forms. Actual observation in this direction in the Indo-European family of languages cannot be appealed to nor has the observation of other languages still in the primitive stage been sufficiently full and exact to justify the positive statement that the primitive sen- tence was short and simple. In the most general way it is, however, probable that the growth of the sentence also, like the analysis of the concept-group, has been in the way of expansion from a germ and of distribution among an increasing number of words of functions which were once expressed by few words. It is not to be ex- pected that, after the transfer of group-meanings to single words or to small groups has been going on for so long a time, it should now be possible in every sentence to distinguish the germ from the later accretions or to follow in all its details the process of expansion and dis- 184 THE GROUPING OF WORDS tribution, but within narrow limits the process is still going on in historical periods and may be detected in certain kinds of sentence. In some few types of sentence, especially those which are strongly emotional and in which, for this reason, there is little information or reasoning, it is often possi- ble to detect the germ, to fix upon the center of gravity. It has already been said that prohibitions are for the most part only willed negatives. The essential thing in them is ne and often the verb is a word of the most gen- eral meaning, serving in truth no other purpose than thai of carrying the subjunctive or imperative ending. Or else it is a repetition from a previous sentence, either the same word precisely repeated or a verb of similar mean- ing. The circumstances or the context have already defined the action prohibited, before the verb is uttered. In like manner the central idea of many questions may be seen to lie in some one word or in some two or three words. Questions with non or nonne, especially in conversation, where the particles are less stereotyped, are usually only repetitions in a different form of what has been previously said or implied. If the speaker desires no change of phrase, the single word non, uttered probably with rising inflection, suffices without addition. In all these cases — which have abundant parallels in modern languages — the germ of the thought is the mere questioning of the previously expressed or implied statement. With slight differences this is true of many ^m's-questions. Either the verb is an empty form (quid faciam ? quid agam ?) or it is a repetition (redde. || quid reddam?), either a precise repetition or a repetition with variation of phrasing. And very frequently the inter- rogative word (quid? cur? quippini? quomodo? quamo- brem?) is all that is required, the rest of the thought 185 LATIN SYNTAX being left unexpressed because it is so easily supplied. The germ of most exclamations is really the tone, the words being merely repeated with more or less amplifi- cation, as in Plaut. Aul., 783 f., Is me nunc renuntiare repudium iussit tibi. || repudium rebus paratis exornatis nuptiis ! This is really in three parts, repudium — rebus paratis — exornatis nuptiis, which indicate the expan- sion of the thought as the speaker gradually perceives the different aspects suggested by the word repudium. In the more unemotional sentences of connected writ- ing it is less easy to detect the germ of the thought, but something may be done by distinguishing what is new in the sentence from what is repeated or carried over from the previous thought. The latter element is in almost all connected writing larger than would be expected by one who has not analyzed the relation of sentences. In writings upon abstract subjects, as in Cicero's or Seneca's philosophical works, each sentence carries forward the thought a little way, but each sen- tence also reaches back into the previous thought, in order to bring the known into relation with the new ele- ments which the sentence contains. Evidently it is the new elements which contain the germ of the new con- cept-group. The following sentences from a book- review illustrate this point : " It is impossible to give any detailed account of the author's position on these sub- jects. He assumes throughout a scientific rather than a polemical attitude towards the various forms of occult- ism." In the second sentence throughout is merely the opposite of detailed, while attitude is the same as position and towards . . . occultism repeats on these subjects. The new element is in the word scientific ; this is the germ of the thought and the whole sentence might have been briefly though inadequately expressed by saying 186 THE GROUPING OF WORDS "But in general (= throughout) it is scientific." The words rather than polemical are added as a definition of scientific. The rest of the sentence is mere syntactical filling-in, and it may be noted in passing that syntactical structure has nothing to do with the distinction between the germ and the amplifications or repetitions. The same mingling of repetitions and new elements may be illustrated from almost any connected passage in Cicero's philosophical works, e. g., from the Cato Major, §67: quamquam quis est tarn stultus, quamuis sit adulescens, cui sit exploration, se ad uespcrum esse uicturum f (The next sentence adds "it is worse than that," and the germ is in plures.) Quin etiam aetas ilia multo plures, quam nostra, mortis casus habet. (The following sen- tence merely specifies ; the germs are perhaps in the three verbs, which are variations of one idea.) Facilius in morbos incidunt adulescentes, grauius aegrotant, tristius curantur. Even when the germ of the thought can be discovered and located, it is still difficult to trace through a long sentence the process by which it is ex- panded and its meaning is distributed. But on a smaller scale the process may be followed in the formation of minor groups within the sentence. One of the most striking and instructive illustrations is afforded by defin- ing parataxis, especially the cases in which the defined verb is in the subjunctive, which have been given in some detail above (Chap. VI, p. 132 ff.). In all these cases the verb in the subjunctive is the germ. It con- tains in an unanalyzed form a number of different ele- ments of meaning; all of them or many of them are suggested at once by the form of the verb. But among them some single one is in a particular case more prom- inent than others. Upon this the attention of the speaker is directed and it is thus made so prominent as 187 LATIN SYNTAX to call for fuller expression by means of another word. Thus the first singular of the present suggests propriety or obligation and this element finds expression in nunc adeam optumumst. The expression of feeling in "I desire to love you " involves a request for permission ; te amem expands into sine te amem. Further details have already been given and need not be repeated. It is the very remarkable vagueness of the subjunctive taken by itself, the fact that it serves such a bewildering variety of purposes, that calls for such a variety also of definition. Every element of meaning thus repeated in the defining word was before contained by implication in the subjunctive form. And it is to be noticed that here also the syntactical center and the center of meaning do not correspond; the clause of secondary meaning be- comes the leading clause in the syntactical structure. The same process of distribution of function underlies the use of prepositions with case-forms. The case -form carried implications of meaning, some of which were prominent, others latent. As one or another of these elements became prominent it called for more definite expression and to this end adverbs were prefixed to the case -forms, as verbs were prefixed to the subjunctive forms. Thus contra, ad, ante, ob, all express with some degree of precision and with accompanying additions the idea of direction or limit of motion which exists also in the accusative, and ex, ab, de, repeat and define the ablative. The prepositions thus afford a definition of the cases, as the prefixed verbs do of the subjunctive, and by the fact that a definition is needed they prove the vagueness of the bare case-forms. In the process of definition by expansion and by dis- tribution of function a distinction is to be made between the motive which leads to expansion and the condition 188 THE GROUPING OF WORDS which permits it. The motive is the desire for more precise expression, aroused by the directing of the attention upon elements of meaning which do not find sufficient expression in the mode or the case-form. The condition which permits the distribution of the meaning between the two words is that the added word shall in part repeat the meaning of the word to which it is added. There must be a common element of meaning, since the object of the expansion is definition; as the two ele- ments of meaning were at first combined in the meaning of one word, they must have been congruous and har- monious. This explains the unity of meaning which still exists after the function has been divided between the two words, fac sciam, uolo abeas, tube ueniat; though the meaning is distributed, the two elements of it are still closely related and their unity is still felt. This is the reason why they are usually in close juxtaposition in the sentence. They constitute a word-group, made by a process of analysis and expansion and bound together in the closest unity. It appears to be a justifiable inference that it is through some such process as this, much more complicated and doubtless involving steps not represented in these simpler illustrations, that harmony and unity are preserved throughout a word-group. These illustrations have to do with the distribution of the function of an inflected form, but an uninflected word may in like manner con- tain elements of meaning which call for definition or for fuller expression than a single word can give. In such a case the most obvious aspect of the concept finds ex- pression in one word and another word is added to limit it more precisely. For the analysis of a concept does not go on equally in all directions at once. Depending as it does upon suggestion by association, it follows first 189 LATIN SYNTAX the line of strongest association and that element which comes first into consciousness is the one which first finds expression in a word. It is not' necessarily the first in order in the sentence, when the sentence is finally ready for utterance. Very frequently one is conscious of the fact that he has fixed upon some middle or later portion of his sentence and is obliged to accommodate the earlier parts to it. But the analysis reaches the point of dis- tinctness, the point where the concepts suggest words, earlier in some one direction than in others. The word thus suggested may not be retained ; the further analysis may render it necessary to reject it and select a different one; but usually it becomes the fixed center of the ex- pression and about it all other words are grouped. The distribution of the remainder of the meaning is largely determined by the selection already made and the mean- ings of all other words must, directly or indirectly, accommodate themselves to the meaning already ex- pressed. Viewed in this way the formation of a word- group is a process of accommodation of meanings, in which each word, as it is selected, forms part of a frame- work to which the other words must be adjusted. It is this adjustment which insures the harmony of the whole and thereby reflects the unity of the concept-group. Adjustment or accommodation involves partial loss of meaning. Of the various meanings associated with a word some will be appropriate to a particular setting, while others will be inconsistent with it. The limita- tions which the context places upon the meaning of in- flectional forms has been noted above; limitations entirely similar are placed also upon the meanings of all words by their use in combination with others. It is because of this that it is possible to use without confu- sion words which have many and quite different mean- 190 THE GROUPING OF WORDS ings. In a particular setting only the meaning which is consistent with that setting is suggested to the hearer; the other incongruous meanings do not occur to him at all or are immediately excluded by the surroundings. The meaning of the word is, in a general way, inferred from the rest of the group, as it would be possible to infer the general tone of the group from the precise meaning of the word. This subject belongs in strictness as much to seman- tics as to syntax. To the syntacticist its chief interest is in its bearing upon the function of inflections and in the suggestions it affords in regard to the transfer of mean- ing from the group to single words. To determine where and how group-functions have become attached to single words or to inflectional forms, and still more to perceive that in many cases the transfer is incomplete, is to take a considerable step toward the understanding of the whole process of grouping words. The transfer of group-meanings and functions to single words is matched by a process exactly the reverse, the re-transfer of the meanings of single words back to the group. This is seen in its most complete form in idiomatic phrases, ut ita dicam, quae cum ita sint, "so to speak," "for that matter." The process which ends in such phrases begins like any other utterance with an analysis of a complex concept and the fitting of suitable words, each with its own meaning, to the resulting members. But if the concept-group is one which frequently recurs in thought and if the words are adequate expressions of it, it easily and surely comes about that the process of analysis is at first partially avoided and then wholly or largely omitted. The consciousness of the whole is 191 LATIN SYNTAX greater than the consciousness of the parts ; the parts are only a means, at the best, of expressing the whole, the germ of the group. Association is therefore set up directly between the group-concept, and the whole phrase, the word-group. The difficult process of analysis is thus avoided; for the analysis, rapid and automatic as it becomes, is at first and when it is attempted upon unfamiliar material difficult and slow. But when both the thought and the words are familiar, analysis is no longer necessary. As the thought, after it has been analyzed, sinks back at once into the un- analyzed condition and is remembered only as a series of connected groups, so a familiar concept-group is for a moment the object of attention and is matched by the associated word-group without the necessity of analysis. The result of this direct association of word-group with concept-group is that the members of the concept-group are not brought forward into consciousness at all and the individual words therefore lose in large measure their separate meaning. To complete the process and produce a true idiom it is only necessary that the phrase as a whole shall be in use so long that the original analy- sis shall be forgotten and perhaps that some of the words in the phrase shall have changed their ordinary meaning, so that they would no longer be combined in the same way or the same words would no longer be selected, if the concept-group should again be analyzed. A step further is taken when, by shift of accent or other change, such a word -group suffers phonetic decay and its parts are compounded so that they are no longer felt as sepa- rate words. When an idiomatic word-group has reached the point of suffering phonetic decay, re-analysis, except in the scientific sense, is impossible ; it is difficult and unlikely 192 THE GROUPING OF WORDS to occur, even though the words remain phonetically distinct, if an important word in the group has in process of time considerably shifted its ordinary mean- ing. But in the case of many idioms a partial re- analysis is not infrequent. It occurs when the speaker happens to give special attention to the concept, either because it is of special importance to the thought or because he desires to avoid a trite form of expression. Thus ut it a dicam may become ut sic < lis* rim or ut hoc uerbo utar; thus ut seias, ut tu scias, ut tu sis sciens, ut scire possis, ut tu meam sententiam noscere possis (all from Plautus) represent various re-analyses and expan- sions of a single concept. It is not necessary to supp< >se that one of these must have been the original from which the others have come, or that a distinction as to age can be made between quid ego nunc faciam? and quid ego nunc agam? They represent two forms of the same analysis. The likeness between different analyses of the same group or between the analyses of but slightly differing groups is not to any great extent the result of direct influence of one word-group upon another, but of in- direct association through the concept-groups. After a group has been analyzed, there remains in the linguistic memory a remembrance of the way in which the anal) sis proceeded. With repetition the analysis becomes habit- ual and is more or less permanently associated with the particular concept. The recurrence of the concept brings up also the particular analysis and the memory is, as it were, stored with such schemes or forms of analy- sis. There are thus in the mind three connected memo- ries, the concept itself, the form of analysis and the word-group. These are all associated together, but they are also associated with other memories of the same 13 193 LATIN SYNTAX kind, concepts with other like concepts, analyses with like analyses and word-groups with like word-groups. But of these series of associations the direct association of word-group with word -group is the weakest, that of analysis with analysis is stronger and that of concept with concept is the closest of all; the sense of unity, upon which the association depends, is clearest in the concept. The connection between sentence-forms like the variations of ut scias quoted above is therefore not simply a direct connection; there is also a connection, much stronger and more important, through the concepts represented by the similar sentence-forms. There is a limit, however, to the formation of such associative series. As the analysis becomes more com- plex, the resemblances are less obvious because, as has been said, the sense of unity is less clear and it is more difficult to grasp the group as a whole. There is a unity, it is true, in each sentence, however long it may be, but long periodic sentences belong to written, that is, to artificial, language. They represent the utmost pos- sible extension of concept-grouping and of word-combi- nation and they exercise but little influence upon the life and movement of language. A greater degree of unification is found in subordinate clauses, even when they are long. A consciousness of the group-concept is preserved, e. g., through a long ut or cum clause, so that it is felt through all its length to be an expression of purpose or time and its relation to the leading clause is not lost sight of. But it is of course true that no long idioms are formed. The longer the phrase, the greater is the difficulty of grasping the group-concept without analysis and the less is the probability that the concept as a whole will become associated with the word-group as a whole. No precise limit can be set in such a mat- 194 THE GROUPING OF WORDS ter, but it is true as a general principle that the memory of sentence-forms, the feeling of similarity in analyses and concepts, by which they are bound into series, are strongest with short sentences or phrases and grow rapidly weaker as the groups increase in length and complexity. The tracing of the influence of one construction upon 1 another, which is the office of historical syntax, may thus follow two or three different lines. A word will influence another word directly and an inflected form of • a noun or verb may lead by analogy to the use of a simi- lar form of another noun or verb. But the making of one inflectional form after the analogy of another and with the same meaning can take place freely only when the meaning has become somewhat firmly associated with the form, that is, when the shift of function from the group to the inflectional form is largely complete. As long as the function requires for its full expression that the inflected form shall be accompanied by other words, as the preposition ah must go with the ablative of the agent, it remains in part a group-function. And so long as it remains a function of the group, the influ- ence upon other forms of expression, the lines of ana- logical influence, must be indirect, through the similarity of the concepts. This has been already alluded to with reference to the potential subjunctive. This kind of shading of a statement is almost never firmly attached to the subjunctive form alone ; it requires an interroga- tive, a comparative, a protasis or some other accompani- ment. There can therefore be no association between potential subjunctives except through the group-con- cepts ; when these are unlike there is no association. Thus tu fortasse me putes indulsisse amori meo cannot be associated with non quiuis . . . describat uulnera Parthi 195 LATIN SYNTAX nor either of them with nimis nili tibicen stem. There is a single element of likeness in all these phrases, but it is not sufficient to lead to the association of the group- concepts or of the forms of analysis, which are entirely dissimilar. In the sphere of word-groups the working of analogy must not be lightly assumed nor widely ex- tended without careful consideration. Its place is in short sentences and phrases, where the similarities of concepts can be observed and finally reduced to system. 196 IX FORM, FUNCTION AND CLASSIFICATION The distinction between form and function in syntax was touched upon in a few words by Lange in his paper of 1852 and has more recently been discussed with much fulness and with much resultant clearing up of the sub- ject by J. Ries in Was ist Syntax? It is a distinction of fundamental importance in its bearing upon the methods of syntactical investigation. Syntactical form includes, in general, all those ele- ments of language which serve in any way to bind words together and to express the relation of concepts. Some of these elements have been considered in detail above. A complete formal description of a particular word- group calls for the noting of the following particu- lars : — 1. Pauses, time, tone, sentence-accent, especially in spoken language and, so far as it is possible, in written language also. 2. The inflectional form of words, including the per- son, number and voice of verbs, as well as the mode and tense, and the number of nouns as well as the case. 3. The order of clauses with reference to one another and of words within a clause. If the immediate bearing of this upon the expression of relation is not evident, as it is not, there is the greater need of observation and record. 197 LATIN SYNTAX 4. The general sentence-structure, as distinguished from order. This includes the presence or absence of a dependent infinitive, of an ablative absolute, of modify- ing clauses. For the recording and exhibiting of these particulars of form the use of symbols has been suggested, in order the better to represent sentence-form without including word-meaning, but no scheme of much complexity has ever been used and it is doubtful whether such schemes could exhibit in detail the varieties of structure. 5. The kind of sentence, including a distinction be- tween subordinate and leading clauses, between inter- rogative and non-interrogative sentences and between those which contain a negative and those which do not. 6. All single words expressive of relation and all modifying words, adverbs, adjectives, pronouns and par- ticles. 7. The word-meaning of all inflected words, e. g., in the case of verbs, whether the verb is one of will or de- sire, of saying or thinking, of effort, of precaution, of fear, etc., whether it denotes attempted action or pro- gressive and continued action or a completed state ; in respect to nouns, whether they are locative, instrumental, temporal, causal, whether they denote a person or a thing, whether they are abstract or concrete, and similar characteristics of the stem-meaning. In some of these particulars, especially in the last two, the line which divides form from meaning is crossed. There are, in fact, three points of view which might be taken into account in looking at word-groups, the form, the content 1 and the function. By the content of a single inflected word would be meant the stem-meaning, while 1 This distinction is suggested by my colleague Professor Oertel, to whom I am already under many obligations. 198 FORM, FUNCTION' AND CLASSIFICATION the meaning of the case-form would be the function; the content of a clause would be the thing said in that clause, the function would be its part in the course of thought, as an assertion, a question, an expression of purpose, of time. But for syntactical purposes, the phonetic form of a word or a word-group is unimportant except as a means of determining and describing more accurately its content, the meaning of its several parts. Syntax is concerned directly only with meanings ; its province is to find the laws which govern the shifts of meaning which accompany inflectional change (a prov- ince which it shares with semantics) or to follow the process by which a word-group acquires a meaning be- yond the meaning of its several parts. For this purpose phonetic form is of only secondary importance and the important distinction is that which separates syntactical function from word-meaning. The term syntactical form is therefore used as above, to cover all that is not function. The word function has probably come over into syntax from physiology. It involves, as all such transferred terms do, a figure which is not strictly accurate in all its details. In the physiological use function is the peculiar or appropriate action of an organ or the capac- ity for such action. But words and sentences are not organs ; they are articulate sounds uttered in connection with a train of thought as a means of exciting a similar train of thought in the mind of another person. In this operation there is nothing organic and nothing, in the strict sense of the word, functional. There is a certain train of thought suggested and this is called the mean- ing of the sentence. Where then is the function ? To answer this question it is necessary to revert to the nature of the concept-group and to recall again the 199 LATIN SYNTAX fact that it includes two kinds of concepts, differing somewhat in their nature and in the clearness with which they are felt in consciousness. The substantive concepts (p. 45) are the more distinct, they correspond more nearly to objects perceived by the senses and they find earlier and easier expression in language. It is the substantive concepts which constitute word-meanings. The concepts of relation, the transitive concepts, are less vividly felt or the conscious attention is less easily directed upon them and they have been later in finding definite expression in language. It is of the expression of these concepts, of the expression of relations, that the word function is used. The distinction between word-meaning and function is not, however, made quite clear by saying that the substantive concepts are the meanings of words and the transitive elements are functions, for there has been in language a constant transfer of words from the ex- pression of substantive concepts to the expression of relation. This is true of all conjunctions, so far as their history is known, and of prepositions. In some earlier use they were adverbs or pronouns or verbs and by a gradual shift they have lost their "meaning" and have assumed, wholly or in part, a " function." The word- meaning of et or sed is a relation, which is felt with almost the same definiteness as the meaning of uirtus or uis or consul. In the case of such complete conjunctions word-meaning and function are identical. But other conjunctions or prepositions which still retain in part their original use like modo, licet, supra, prope, may be said to have at one time word-meaning, at another time function, or, more accurately, to have both together, since the substantive concept and the transitive concept have elements in common. In all these cases, where 200 FORM, FUNCTION AND CLASSIFICATION single words have been transferred to the expression of relation the use of the term function is somewhat inexact. The employment of the term with reference to in- flected words seems at first sight more precise ; the meaning of mensd is " table," the function is, e. //., loca- tive. But this assumes that the relation is and has always been expressed by the termination, an assump- tion which is not borne out by the facts. It is rather to be assumed that the word-meaning is the determining element in making die an expression of temporal relation, humi a locative and gladio an instrumental. The func- tion is in part, certainly, limited and fixed by the mean- ing of the stem and, in so far as this is the case, the separation of word-meaning and function is impossible. With reference, therefore, to some inflected words — and probably to the greater number — the term function is liable to lead to confusion of thought and can safely be used only in a general way. The function of a word-group is the expression of the relation of its concept-group to another concept-group. In part, this may depend upon a word which introduces or " governs " the group, a conjunction or a preposition, and to this extent the function of the group is dependent upon the function or the meaning of one of its members. Thus cum or si determines the function of the clause which it introduces. But in part the function of a clause is conditioned upon its being grasped as a whole, that is, upon the clearness with which the unanalyzed whole, the group-concept, finds expression. For relation cannot be felt except when the unity of each of the related concepts is felt. Through these two means, the introducing word and the unification of the group, the function may be expressed and may be 201 LATIN SYNTAX quite accurately distinguished from the content or meaning. It follows from this that the function, whether of a single word, of an inflected form or of a word-group, is in itself but vaguely defined. Such definition as it has comes from word-meaning; the function of a conjunc- tion is most clear, because function and meaning are identical ; the function of a case-form is less definite and that of a word-group is still less defined, unless it gets definition from the meaning of an introducing word. The relation of two clauses in parataxis illustrates the vagueness of function, when it is suggested merely by the unified word-groups without the aid of the meaning of a conjunction. Lacking the definition which is given by the meaning of the stem, an ablative may suggest cause, manner and means, all at once, and a clause may perform both final and consecutive functions or may be at the same time a clause of time and of condition. It must be said in general of function that its clearness is in inverse ratio to the degree to which it can be sepa- rated from word-meaning. The classification of syntactical material, while it has usually been based either upon function or upon form, has varied somewhat according to the object aimed at. For pedagogical purposes the usual arrangement is functional. This is partly a matter of tradition, now apparently somewhat weakened, and partly because a grammar does not so much present the material of syntax as its doctrines. A grammar is made up in large part of rules to serve for the interpretation of a foreign language or, in the older Latin grammars, to teach the pupil also how to write Latin. For these pur- poses classification is a means of presentation, not a tool 202 FORM, FUNCTION AND CLASSIFICATION of investigation, and the question what method of clas- sification is best adapted to this purpose is one to be answered by the experience of teachers. A functional arrangement, if it is clearly stilted, has a certain unity of system and permits easier parallels between different languages, especially between Greek and Latin. There is, however, a tendency at present in school grammars to substitute a formal arrangement and it is to be hoped that this method may be fully tested in the class-room. There is a considerable amount of work upon single authors which is usually regarded as syntactical, but which is quite as distinctly stylistic in its object. It is found chiefly in doctor-dissertations upon single cases or tenses or prepositions or conjunctions, as used by one author or a group of authors. On a larger scale Tacitus has been thus treated by Draeger, Nepos by Lupus and Livy by Kuhnast and by Riemann. These works, however, are not greatly concerned with syntac- tical theory or law ; they afford excellent material for theory, but their real object is the study of an author's style by comparing it in detail either with general usage or with the style of same other author. The best system of arrangement is therefore one which most facilitates comparison with other similar collec- tions and for this purpose some familiar arrangement is convenient. Draeger's scheme is most frequently fol- lowed in Latin and no doubt the impulse to this kind of work was given chiefly by his Historical Syntax. This fixed scheme is better adapted to such purposes than any novelt}^, even though the latter might have scientific advantages. Classification for scientific purposes, as a tool for the investigation of a mass of syntactical material and a means of discovering the laws of syntax, is of such 203 LATIN SYNTAX importance that the two systems in use, the functional and the formal, deserve deliberate consideration with reference to the advantages and the disadvantages of each. The advantages of classification by function lie chiefly in its systematic character. It presents the facts in a logi- cal scheme, showing their relation to each other and giv- ing to each function its proper place in the whole scheme of expression. Thus a functional presentation of the tenses shows the expressions for past, present and future, for action beginning, continued, complete ; clauses divide into independent and subordinate, and subordinate clauses into temporal, causal, conditional, etc., or into substantive, adjective, and adverbial. There is a com- pleteness and an ideal simplicity and clearness in such a scheme as is given, for example, in Draeger's table of contents to Vol. II, which is not to be undervalued. In the study of language the mass of facts is so enormous and the difficulty of reducing them to order is so great that the danger of over-theorizing may well be risked for the sake of arriving at some kind of system. A second advantage of functional classification is the directness with which it leads toward what must be recognized as the ultimate end of all linguistic study. In the discredit into which logical grammar has fallen, the fact is sometimes overlooked that the admeasure- ment of language by logical standards is a perfectly legitimate and indeed a necessary step to the complete understanding of its nature. The logical standards supply a final test of the degree to which language suc- ceeds in fulfilling its mission. Or perhaps it is more correct to say that the comparison of language with the ideal standards of logic affords a measure of the differ- ence between adequate expression — which is all that 204 FORM, FUNCTION AND CLASSIFICATION language attains — and absolutely precise expression. To apply this test is to use one of the means of under- standing the limitations and the nature of Language. This is the reason why the results of functional syntax seem to have so great a degree of finality. Winn a study of a certain form of clause results in the con- clusion that it is a clause of purpose, this result seems to be the end of all question in regard to the nature of the clause. A third direction in which functional observation and classification may be used to advantage is as a means to the discovery of formal peculiarities. In working over a considerable mass of syntactical material — two or three hundred cases — it will at times be difficult to discern variations in form upon which a formal subdivision can be based. Formally the cases look all alike or else the}' present a bewildering variety of form in which it is impossible to find a clue to the significant variations. In this situation it is sometimes easier to begin with the observation of functional differences and subdivisions in the confident expectation that this method of pro- cedure will lead in the end to the detection also of formal differences. The advantages of functional syntax, it will be seen, are chiefly the result of its systematic and logical char- acter and appear almost wholly in the presentation of results. Its defects are most apparent and most serious when it is regarded as a tool of the scientific investi- gator, as a means to the establishment of historical sequence and the discovery of laws of development. The defects of functional classification must be stated at greater length, since it is the system in common use and its deficiencies are hidden under the sanction of tradition. 205 LATIN SYNTAX 1. Classifications based upon function are vague and consequently unstable, that is, they admit of wide diver- gence of opinion in regard to the proper class to which a particular instance should be assigned. Thus differ- ent commentators upon a passage hesitate between the dative of advantage and the ethical dative, between ablative of manner and ablative of means, between ablative of means and instrumental ablative, even when there is no question in regard to the interpretation of the passage. In the classification and naming of modal uses there is even more divergence of opinion. No two scholars could take a thousand instances of the subjunc- tive at random and divide them between the two ordi- nary classes of volitive and potential and expect the results to agree. The system in use fails to meet the most elementary requirement of a scheme of classifi- cation, namely, that it should be clear and stable and that under it species and varieties should be identified with reasonable exactness. This failure is not due to lack of care or ingenuity in the application of the system, but to the radical defect of vagueness in the characteristics of the classes. Func- tions are not divided from each other by sharp lines, but by neutral belts. The simplest kind of functional divi- sion, like that between coordination and subordination, is not precise ; the subordinating function is a relation which varies almost infinitely in closeness and strength, from the fullest incorporation of one clause into another to relations which are so near to pure coordination that they are expressed by coordinating conjunctions. So purpose and result run together and inference, mo- tive and efficient cause, as colors shade into each other in the spectrum. The functional characteristics of words and word-groups are most certainly of the very highest 206 FORM, FUNCTION AND CLASSIFICATION importance ; they are indeed more important to the stu- dent of syntax than any other characteristics ; but they are an unstable foundation for classification. A system which uses function for its measuring-rod will always be shifting and uncertain. 2. A partial recognition of the above-mentioned defect has led to some attempts to remedy it. It has been thought that the difficulty lies in a lack of precision in the definitions, and that it might be removed by the use of sharper definitions. But this is an error. The variety and instability lie in the functions themselves, not in the definitions. This is, on the whole and in the long run, so plain that individual and clear-cut distinc- tions between functions, e. g., between relative clauses of characteristic and relative clauses of result, are rarely acceptable to any one but the maker of them, because the precision is attained by the omission of some char- acteristics. Precise definition of function has the fur- ther disadvantage of leading to forced interpretation and to interpretation by translation. Being founded upon some one or two characteristics of the function, and, naturally, upon prominent characteristics, it is almost inevitable that the maker of the definition should see those characteristics and only those wher- ever he looks. Thus all subjunctives come to look polemical or all Active, all potentials are translated by would, all result clauses contain a potential. But the difficulty lies in the system, not in the definitions. They are as precise as definitions of large functions ought to be. 3. One of the most serious defects in the systems of functional classification at present in use is the great extent of the classes and the lack of subdivisions. Thus purpose is treated as a single function, without 207 LATIN SYNTAX subdivision except that between positive and negative ; few grammars subdivide the function of cause ; the sub-classes of conditional sentences are few in number and are really more formal than functional. With terms so general as these the investigator cannot do precise work. Careful observation is discouraged, because no careful observation is necessary, and the whole syntac- tical study ends with the placing of a particular in- stance under one of these broad classes, no attention being paid to the minute differences in the understand- ing of which lies the possibility of all discovery. To some extent this defect is being remedied. Quite recent American work has shown that functional syn- tacticists feel the need of more subdivision, e. g., of the characterizing qui clause, and some years ago the pro- grams of Imme * gave an illustration of the excellent results which could be reached by a functional subdi- vision of the interrogative sentence, as compared with the older triple division. The most brilliant example of functional classification carried out systematically into the extreme of subdivision is that afforded by the programs of Hentze, mentioned above (pp. 113 f.), on parataxis in Homer. Functional treatment of this kind, so free from sweeping terms, so detailed and precise, does not yield in effectiveness to any formal classifica- tion. A glance at the pamphlets will show the reason. Each functional subdivision is also described in formal terms ; the mode and tense, the pronominal forms, the repetitions of words from a previous clause, the order of clauses, in short, all the formal elements are noted quite as carefully as is the function, and the functional classi- fication is in fact a formal classification also. One may hazard the guess that in working it out the author really l See p. 32. 208 FORM, FUNCTION AND CLASSIFICATION made use first of the formal distinctions, though for Ins purpose he finally placed the functional characteristics in the foreground. To some considerable degree the same thing may be said of Imme's programs. The truth is that an ideal classification is both formal and functional. When the utmost had been done in the direction of subdivision, so that the most minute pecu- liarities had been made to serve as a basis for sub-classes, the result would be to bring the two systems into har- mony. Each species would have its carefully defined form and its correspondingly restricted function. But it woidd still be true that to reach this ideal correspond- ence and harmony the path of formal observation and classification is both surer and easier. 4. Connected with the use of large functional classes and made more dangerous by it is the tendency to rea- son about functions as if they were entities, as if they influenced other functions or were influenced by them or were derived from similar functions or became the source of other derived functions. This is particularly frequent in modal syntax, and begins in fact with the treatment of the subjunctive mode itself as an existing reality instead of a general term applied to a number of related forms and functions. The most striking — and, it may fairly be said, astonishing — illustration of this is in the treatment of what is called " the potential.*' Here are a dozen or more of different word-groups hav- ing in common the presence of a subjunctive form and having a common functional element, the expression of an opinion in regard to an imagined or ideal or contin- gent act. But in syntactical work " the potential "' is said to be derived from something or something else is said to be derived from it, as if it were a distinct entity instead of being a name for an abstraction. The result 14 209 LATIN SYNTAX is that, in the first place, the other members of the word- group are wholly or largely overlooked and, in the second place, the distinctive elements in the several functions are forgotten and attention is directed solely upon the common element. Thus the individuality of word-groups and of their functions is lost sight of and the whole pro- cess of reasoning becomes a kind of ingenious syntacti- cal algebra, in which symbols have displaced realities. 5. This leads to still another tendency which, if not inherent in the functional method, is at least commonly associated with it, namely, the tendency to attach the function of a word-group to some single member of the group, usually an inflected word or a conjunction. Thus it is common in grammars and in syntactical work to speak of the deliberative subjunctive and to say that it is derived from the volitive (or the potential) and is the source from which other "subjunctives," e. g., the subjunctive in indirect questions, are derived. But the function belongs to the whole word-group. In the typical form quid ego nunc faeiam? each word con- tributes to the total meaning, though the exact force of ego is less clear than that of the others. If either ego or nunc is omitted, as is sometimes the -case, the func- tion is less distinct; if both are omitted, the question is not necessarily dubitative. The subjunctive form also contributes to the expression of the function of the group, though it is not essential, since the same function is occasionally expressed by sentences with the present indicative. But deliberation cannot be expressed by any one of the four words alone and is not therefore a function of any one of them alone. There is no such thing as the deliberative or dubitative subjunctive; to use the term is to attribute the function of the word- group to a single member of the group. 210 FORM, FUNCTION AND CLASSIFICATION Another and more striking illustration of this ten- dency is given in Hale's Anticipatory Subjunctive. 1 The subjunctive is used occasionally in subordinate clauses after ueniet aetas cum, ueniet iterwm (dies) qui, ueniet aliquod tempus quod ; an example in full is Alter erit turn Tij>hys, el altera quae uehat Argo Delectos heroas. Verg. Eel. 4, 34. Futurity is in all these cases plainly a function of the word-group, of the whole sentence. So far as it is distinctly attached to any words, it is to the verb in the future indicative, which in nearly all cases stands in an emphatic position, and to the noun of time and the adverb (turn, iterum) or some other word (alter, altera, lustris labentibus). These are the members of the group which contribute the strong future function. Into such a group a subjunctive form may be inserted, because there is in many subjunctive uses an implication of futurity; this is the condition, the common element, which permits the use of a subjunctive form in a group strongly marked with future meaning. But the reason for selecting a subjunctive rather than a future indica- tive is quite different; it is to express more clearly the element of destiny, of purpose (in the broad sense), of fate. It is not for the purpose of expressing more fully the future idea, already epiite sufficiently expressed, that the subjunctive is used, but on the contrary for the purpose of differentiating the whole concept-group from simple futurity and adding another slightly different element to the total meaning. Hale's method here seems to me to involve the error of attributing a function of the whole sentence to a single member of it, with the 1 Chicago, 1894, pp. 20, 21. 211 LATIN SYNTAX further consequence of erecting an occasional function of a group into a permanent function of a mode. Into this error, if it be an error, he is led by the fact that he is using a functional classification, to the neglect of for- mal characteristics. 6. Classification by function brings together objects which belong apart and separates objects which belong together. This is to some extent true of all classification. A division of animal life by habitat, into land animals and water animals, for instance, brings together warm- blooded and cold-blooded animals. A classification by bony structure neglects color, habits of life, edible qual- ities, etc. This is a matter of course ; no basis of classi- fication serves all purposes. Every permanent system of classification is to some extent conventional. But there are degrees of suitability and usefulness. In the division of subordinate clauses motive, efficient cause and inference are thrown together and clauses intro- duced by quod, quia, quoniam, cum and quando, taking sometimes the indicative, sometimes the subjunctive, are treated as if they were all alike. Under purpose we put together clauses with id, ne, quo, quin (and some- times quominus*) and we might include qui, the gerund or the gerundive and the supine. The list of conjunc- tions brought together under the vague term concession is especially remarkable: quamquam, quamlibet, quam- uis, quantum uis, ut, ne, cum, licet, etsi, tametsi, etiamsi; this is, from the functional point of view and for the purposes of a grammar, a perfectly correct list, but from the historical point of view and for scientific purposes it is a mere jumble of incongruities. On the other hand functional classification separates things which ought to be kept together. The si-clause 212 FORM, FUNCTION AND CLASSIFICATION of condition, the si-clause of concession, viiror %% and si quidem are to be found under four different functions. So the uses of ut are scattered and metuo ne is in one place, abeo ne in another and ne clauses of result in a third, if they are not entirely ignored on the functional ground that all result clauses are potential and thai no potential takes ne. But the history of the n-clause, of the Mi-clause or of the ne-clause is one connected history and a classification which in effect denies this continuity is a mistaken and harmful system. If it be asked how a system so defective comes to be employed by investigators, the answer is to be found, first, in the real value of functional classification for cer- tain purposes and, second, in the history of syntactical method. The real and lasting value of the functional treatment of sentences has been in part pointed out; it lends itself admirably to a clear and logical exhibition of the facts of thought-relations. But beside this it has been a most useful means of teaching exact and critical translation, either into Latin or from Latin into the vernacular. At a time when "natural" methods are so strongly urged and so widely used as they are at present, it is worth while to dwell for a moment upon the other side. The necessary middle step in translation is a thorough com- prehension of the thought, a comprehension which goes deeper than the forms in which thought expresses itself in any one language. Without this, translation will not be flexible or idiomatic ; it will be tinged with the color of the language from which the translation is made. The failure to secure such a thorough comprehension is one cause of the stiff and mechanical translations to which . teachers are obliged to listen in the class-room, 213 LATIN SYNTAX as it is the source of much that is unclassical in modern Latin. Now to this kind of conscious and elaborate comprehension of the underlying thought nothing con- tributes more directly than the functional analysis of the sentence. When a thorough logical analysis is based upon a knowledge of the meaning of words and is fol- lowed by a synthesis, a re-composition of the thought, it results in an understanding of the sentence such as no "natural" method can give, an understanding which is the solid foundation for criticism, for interpretation and for translation, and which is in itself a valuable means of mental training. It is in part the recognition of this kind of value which has led to the retention of func- tional classification in grammars and in the school-room. And the fact that almost all syntacticists are also teachers has led to the use of the same system in inves- tigation, where its advantages disappear and its defects become prominent. A second cause which has led to the retention of functional classification in the science of syntax is to be found in the history of syntactical method. It would be incorrect to say that this system is purely a logical sys- tem, but it is correct to say that it is the product of logical and metaphysical views of language. The func- tions upon which the classes are based — time, place, cause, purpose, etc. — have not been selected and named for the purpose for which they are now used. Some of them are categories of thought established by Aristotle and employed in logical systems down to the end of the eighteenth century, others are terms which acquired their meaning and importance in metaphysics, and they have come over into syntax and held their place there primarily as a result of the influence of logic and meta- physics upon all thought and especially upon Latin 214 FORM, FUNCTION AND CLASSIFICATION grammar. This is the explanation of some of the < nidi- ties of our functional classification. It is for this reason that cum clauses with the subjunctive expressing an inference and quod clauses with the indicative express- ing a motive are classed together under the general head of cause; cause was an important logical and meta- physical category, while inference and motive were not. Our syntax has therefore to this day disregarded in clas- sification both the formal and the psychological differ- ences between the cum clause and the quod clause under the influence of the metaphysical tradition. In the same way we have a functional distinction between the abla- tive of time and the ablative of place, though in truth the functions, so far as they can be separated from the stem-meaning, are identical and the only difference in meaning is in the words themselves. It is a curious fact that, in spite of its generally logical character, the scheme should make no separate class for clauses of place ; this is of course because such clauses call for no special "rules," as the clauses of time do, and is an illustration of the influence of teaching upon science. All this does not mean that the functional system is now properly to be called a logical or metaphysical sys- tem nor that it has been uninfluenced by psychological and formal considerations, but only that it is an inheri- tance from logic and metaphysics and that a large part of its hold upon syntax is in fact the tenacious hold of the traditional. It is antecedently probable that a system which has been unconsciously shaped by various and incompatible influences will be less suited to entirely new uses than a system which is the product of the newer aims or is deliberately devised to meet the newer demands. But it is a fact, which should receive the fullest recognition, 215 LATIN SYNTAX that functional classification has been in part accommo- dated to the changes which have taken place in the aims and consequently in the methods of syntax. But it is only a partial accommodation. These changes are the result of the rise of syntax to the rank of an independent science. The service which it may render to interpre- tation and translation and its value as an educational tool are not, from this point of view, considerations of the first importance; it must seek its own ends by methods of its own. These ends are at present psycho- logical interpretation and historical explanation, and to neither of these is a functional system adapted. It will no doubt be said by scholars who use a func- tional method that the content of the terms for functions has changed and that purpose, cause, time, place are now employed in syntax with a psychological meaning. This is true, but it is not quite the whole truth. The difference between logic, metaphysics and psychology is in the point of view; the objects studied, the material of these sciences, are the same, namely, the soul and the laws of its working. This subject-matter logic exam- ines with reference to the forms of thinking, and meta- physics with reference to the laws of being; the object of psychology is not so much the testing of spiritual and mental activity by reference to any laws as the understanding of the actual process in its details. It may be said that logic and metaphysics are more dis- tinctly classifying and testing sciences, while psychology is more interested in the understanding of activities. In logic the end sought is the bringing of a certain kind of mental activity under a certain category; in psychol- ogy the end sought is the detailed knowledge of the mental operation, and to this end classification, that is, comparison with similar mental operations, is only a 216 FORM, FUNCTION AND CLASSIFICATION means, a mere step in the scientific treatment, a con- venience, not an end. Now in this respect func- tional classification more nearly resembles logic than psychology. The placing of a certain clause in the category of purpose or of cause is the end, the object aimed at. About such questions, either the number and naming of the categories to be employed, as in regard to conditional sentences, or the placing of a par- ticular clause in this or that category, the discussions of functional syntacticists center. The inevitable result of over-attention to classification is a diversion of atten- tion from details. It has been, to take an example, beyond question a gain to have will and wish brought forward as the fundamental ideas of the Latin subjunc- tive and a further gain to have the potential meaning regarded as another distinct function. These lend themselves to psychological treatment somewhat more easily than the older meanings of subordination or sub- jectivity, though the fact is sometimes overlooked that even these terms may have a psychological content. But the discussions looking toward the placing of the various subjunctive sentences in one or another of these classes have contributed very little to psychological inter- pretation. At its worst it has consisted in translating each subjunctive by some English auxiliary verb and on this basis settling its classification; at its best it consists in finding in a particular case the element of will or wish or of the potential and on this ground assigning it to its appropriate class. When it happens that two of these elements are present at once, discussion follows, a fruitless discussion because the shield is golden on one side and silver on the other. It is, indeed, quite inevi- table that, starting with these very general and imper- fectly distinguished elements of meaning, the student 217 LATIN SYNTAX should be satisfied with finding one or another of them in a particular form and should regard his task as ended. But the defect of the method is that the meaning of the form may contain many other elements beside those of will and wish, the omission of which leaves the psycho- logical interpretation incomplete and one-sided. The mere determination of the presence of a single element, even if it be the dominant element, contributes but little to an understanding of the working of the mind as ex- hibited in modal forms. The truth is that, while the content of the terms of functional classification have been in some measure adjusted to psychological treat- ment, the actual process of classification shows plain traces, as the terms themselves do, of the logical origin of the whole system. The movement of syntax toward psychological interpretation will require either a much greater adjustment of the functional system or a com- plete abandonment of it as the principal tool of scientific discovery. The deficiencies of functional classification as a means of determining historical sequence and the derivation of one function from another are partly incidental, having to do with the steps of the reasoning, partly inherent in the system. Taken together they seem to the writer to make historical investigation by means of functional treatment almost an impossible combination of terms. This statement will no doubt appear to functional syn- tacticists either entirely wrong or greatly exaggerated. The grounds upon which it is made are therefore given in some detail. Either of the two sides of language may be made the subject of investigation with reference to its history; there may be a history of meaning or a history of form, or, to bring the matter to a more practical point, we 218 FORM, FUNCTION AND CLASSIFICATION may trace the history of a meaning as it appears in a succession of different forms (words or word-groups) <>r we may trace the history of a form as it expresses a succession of different meanings. Either procedure is possible, whether with .single words and their meanings or with word-groups. In semasiology there appears to have been some hesitation between the two methods; at least the earlier definitions seem to suggest that by semasiology was meant the study of the various forms through which a single idea had been expressed. Tims it would be possible to begin with the concept horse and to trace its expression in successive words or with the concept to say, to speak, and follow down the different aspects of this concept as reflected in the many verbs which have carried the elements of it. But the diffi- culties of this kind of investigation are evident; it would require great precision in definition, great knowledge of the actual meaning of words and a very wide range of information. It would be in fact possible only after the foundation had been laid by a great num- ber of separate studies, each dealing with a single verb — with orare, dieere, loqui, narrare, etc. That is, it would be a possible and an interesting manner of pre- senting the results of other studies, but as a method of investigation it is practically impossible. Historical semasiology has therefore chosen the other course; it begins with the word, the form, and follows the history of the word out into its varied and historically connected meanings. It has been remarked above that syntax, which is the semasiology of word-forms and word-groups, has much to learn from the semasiology of words, and this is one of the lessons. For if the history of word- meanings is too difficult to be attacked directly and must be approached through the lexical method, much 219 LATIN SYNTAX more is the history of concept-groups and concept- relations too difficult to be followed directly. To start with the concept of cause, with the causal function, and follow it into all the forms in which it finds expression would carry the investigator into word-meaning, case- syntax, the syntax of prepositional phrases, the syntax of more than one subordinate clause and the syntax of the paratactic relation. In the same way the function of purpose is expressed by single words, by preposi- tional phrases, by participles, by gerunds and gerun- dives, by the supine, by several different clauses and by parataxis. To follow the history of the expression of either of these functions in Latin would be an extremely difficult task, in fact, an impossible task until the his- tory of each of the forms of expression had been sepa- rately worked out, that is, until formal syntax was complete. As a matter of fact no such elaborate and systematic treatment of the historical expressions of function is ever attempted. The method in actual use is a compro- mise in which formal and functional classes are used without distinction. The ablative of cause and the ablative after utor are placed side by side ; the dative of advantage (functional) and the dative after similis (formal) are treated alike; the functions of cause or purpose are not followed out into all their expressions, by case, participle, etc., but are studied only in the clauses of cause and purpose, with cum, quod, qui, ut, etc. This compromise in the system results in a loss of breadth of view and conscious working toward a dis- tinct end without the compensating advantages that would follow a frank abandonment of functional treat- ment. In general, the directions in which it would be said to 220 FORM, FUNCTION AND CLASSIFICATION have contributed to historical syntax are, first, in the establishing or suggesting of a historical connection be- tween some of the cases and their Indo-EuTopeaD prede- cessors; second, in making a similar connection between certain subjunctive functions and the corresponding original meanings; third, in connecting the uses of the subjunctive in certain subordinate clauses with corre- sponding uses in independent sentences. These are really reducible to two, connections with Indo-Eurdpean functions and connections between subordinate clauses and independent sentences. As to the first, it is not in point here to repeat what has been said above (pp. 32 f .) as to its speculative character ; the question is rather as to the method used in suggesting or establishing histor- ical connection. That method is based upon resem- blance in function. Its fundamental principle, acted upon but not formulated, is that a sufficient degree of resemblance between an earlier and a later function of an inflected form indicates a historical connection be- tween the two functions. In this way a resemblance between the ablative of cause and the early ablative proper is held to show that the ablative of cause is de- scended from the Indo-European ablative and the abla- tive absolute from the locative. In the same way uses of the subjunctive, named t>y their function volitive, anticipatory, optative, potential, are regarded as de- scended from Indo-European functions, and uses of the subjunctive in subordinate clauses, of purpose, result, cause, time, are connected with one or another function of the mode in independent sentences. A brief sketch of the method which he follows is given by Hale J in a discussion of the origin of the dignus qui clause. It consists in asking the question " Will the Volitive idea 1 In a periodical called The Latin Leaflet, Brooklyn, April 22, 1901. 221 LATIN SYNTAX [or the Anticipatory or the Potential] supply a natural and easy starting-point ? " and answering it by saying, " It seemed to me that it did ... I therefore . . . placed the construction under the head of the Volitive." Hale explains further that he changed his result by asking the question in regard to the " Subjunctive of Obligation or Propriety," which seemed to offer a still easier starting- point, but the method is the same. It is nothing more than seeking for a resemblance as close as possible be- tween two uses of the subjunctive, in independent sen- tences and in a subordinate clause, and holding that such resemblance indicates historical connection. This method is probably the one followed in classifying abla- tives as derived from the ablative proper, the instru- mental or the locative ; the earlier functions are defined, though necessarily somewhat vaguely, by comparative syntax and the later Latin uses are classified according to their more or less close resemblance, the ablative of cause as an ablative proper, manner as instrumental, the ablative absolute as locative, and so on. This method is liable to several incidental errors, the possibility of which lessens the value of the result. In the first place, there may be close resemblance in func- tion where it is clear that there is no historic derivation. The genitive after similis is scarcely distinguishable from the dative and the ablative of characteristic or quality very closely resembles the genitive of characteristic, yet we do not say that one is descended from the other in the sense in which the ablative absolute is said to be de- scended from the locative. A prohibition with the im- perative is often indistinguishable from one with the present subjunctive, and functions of the present indica- tive and the present subjunctive, of the present, the imperfect and the perfect indicative, of the perfect and 222 FORM, FUNCTION AND CLASSIFICATION the pluperfect, overlap one another frequently. It is, indeed, a law of language that meanings of winds and functions of forms extend themselves until they are met by the meanings and functions of other forms on a neu- tral ground where either may be used. The existence of such doublets is one of the common facts of language and the close resemblance of meanings derived from widely distinct sources occasions no surprise. In the second place, even when similarity of form makes a historical connection of some kind probable or certain, the determination of the actual line of connec- tion cannot be inferred from the apparent closeness of the resemblance, by finding the " natural and easy " line. Here again syntax may learn from semasiology that meaning is often carried from word to word by the most incidental associations, in the most unexpected and zig-zag lines. The older et} y mology would afford illus- trations of the mistakes which are likely to result through inferring historical connection from resemblance in meaning. The meaning of a word or the function of a form is not single but composite ; it contains many elements, and any one of these may be the middle step through which two somewhat different meanings are connected. Moreover, the determination of the domi- nant element in a function is in most cases an impossible task ; the resemblance between the ablative absolute and the ablative, locative or instrumental depends altogether upon the selection of the typical instances. These liabilities to error are, as has been said, rather incidental to the method of tracing historical connection by function than inherent in it. But the conception of the process by which function is transmitted and shifted is such as to involve more fundamental errors, errors of principle, not of practice. 223 LATIN SYNTAX The function of an inflected form is the word-meaning so modified as to express the relation of the concept of the word to other concepts. It exists only in associa- tion with an inflected form, not as something apart by itself, having an independent existence. If the inflected form and the modified meaning continue in use without change, maintaining their association, there is no trans- mission of function. Transmission and extension of function is the carrying over from one word to another of the inflectional variation and the associated modifica- tion of meaning. The necessary condition of this is an association of some kind between the two words, an association which may come from a common element in the meaning of the two or, what comes to the same thing in the end, from their frequent use together. By virtue of this association the second word is varied in form and modified in meaning as the first word had been. In other words, function passes from word to word as inflectional variations are extended, by analogy resting upon association. And shift of function takes place in the same process. The second word is not identical in meaning with the first, but slightly different, and the function is therefore slightly changed. If we suppose that the ablative form passed from gladio with an instru- mental meaning to poculum, it could not carry the instru- mental meaning unchanged. The meaning of poculum, the different nature of the object, would in some connec- tions suggest a locative meaning for poculo, and if from poculum the ending and the function passed over to some other word whose meaning still more distinctly sug- gested the idea of in, the function would be still further modified and shifted. So also with word-groups ; the transmission of function from one group to another is conditioned upon an association between the groups, in 224 FORM, FUNCTION AND CLASSIFICATION use or in form or in the meaning of parts of each. In either case the function plays the passive rOle; it is transmitted, is modified; but the necessary condition lies in the associations of form and word-meaning. Now the defect of the method under discussion is that it follows the passive line of connection. If there were sufficient data and the facts could be set forth in proper order, the functional line would consist of a series oi gradually changed meanings, of the ablative for exam- ple, extending from the meaning separation or source to the meaning cause. The other line would consist of a series of words connected by associations of different kinds and having meanings varying more or less. The first series would show the effects only ; the steps of the process and the working of causes would lie wholly in the series of associations. The following of the first series is a barren process, having as its result only the facts of connection, like a chapter of genealogy from the Book of Genesis. The following of the second series is a study in processes and causes. The first is historical only in the sense that a succession of events is history ; it is historical syntax with almost all that is instructive and interesting left out. Classification by syntactical form is in general the direct opposite of functional classification. Its advan- tages and defects have therefore been suggested by con- trast in discussing functional treatment and may now be the more briefly described. Its disadvantages are chiefly in the presentation of results and in its use in teaching. A presentation of the forms of a particular construction, if it is full enough to gain the advantages which the system offers, will be very elaborate and complex. A formal exhibition of the is 225 LATIN SYNTAX subjunctive in independent sentences will divide the cases according to the presence or absence of the nega- tive and according to the kind of negative (ne, non, nullus, etc.). Each class might then be subdivided into purely independent and paratactic. The next subdivi- sion would be by the kind of sentence, interrogative or not, then by tense, by person, by number, by voice, and by the meaning of the verb itself. The presenta- tion of the si-clause in Lane's Grammar, §§ 2025-2088, which takes into account only the mode and tense of the two verbs, gives more than fifty classes, not includ- ing some special uses like miror si. As compared with the ordinary classification into four or six forms, this is extremely elaborate. Yet it does not make a subdivi- sion by person or by verb-meaning or by the presence or absence of adverbs, negative or other, all of which have a distinct bearing upon the expression of the relation of protasis and apodosis. A complete and detailed presen- tation of all the forms of the conditional sentence would be intolerably long and complex as a means of making a student acquainted with the facts. A further disadvantage, more or less common indeed to all schemes of presentation, is the separation of things which belong together. If the main classes are based upon mode and tense, then the protases containing a verb of thinking or of saying will be scattered about in different places in the system. This is of no importance to the investigator, but it is confusing in a school grammar. Though a formal, classification results finally in a study of functions and relations, it does not contribute so immediately to interpretation and critical translation as a purely functional arrangement does. This is no slight defect and must be set down as a large item to the credit 226 FORM, FUNCTION AND CLASSIFICATION of the functional system. It carries with it all that is said above in regard to the value of logical analy- sis of sentences as a means of instruction and mental training. The advantages of detailed formal classification consist in the main in the avoidance of those defects which are inherent in or incidental to a functional scheme. It lessens the liability to fall into forced interpretation and translation, because the attention is fixed upon the form in making divisions and classes and the determin of meaning is the last step. It removes entirely the danger of large classes, of the symbolic use of functional names and of • the undue prominence of nomenclature. It distinguishes with much precision between group- function and the function of an inflected form. It is especially superior to functional classification in preci- sion and in the statement of the details by which a par- ticular species maybe identified. Thus the twenty or thirty formal categories of the qui clause made by Ditt- mar (see p. 19), though they are far from exhaustive, are immensely superior to any of the functional schemes. e. g., to that in Bennett's grammar, appendix, or to that in Hale's Cum, as a means of identification. It is possible to write of a Gains primus est qui clause with the expectation that the reader will know what the object under discussion is, but a "Determining [Rela- tive] Clause of the Developed Type" is a kind of thing about which two scholars might write at some length only to find in the end that they had been talking of two different objects. This superiority of formal classifica- tion is greatest with reference to word-groups; in case- syntax it is much less marked, since a formal classifica- tion of ablatives or datives would depend in large measure upon the meaning of the noun. The nature 227 LATIN SYNTAX of the ablative poculo depends upon the meaning of the word, whether it is an instrument for holding wine ready for drinking or the vessel in which the wine is spark- ling. But this, on the other hand, is determined with much precision by the context; when it still remains vague, the very vagueness is a fact of language, as much to be noted as any other. The most important point of contrast between the two systems is that which relates to the two chief character- istics of the philological research of the present day, psychological interpretation and historical explanation. The central requirement for reaching either of these ends is minuteness of observation. Upon the small de- tails of language both the psychology and the history of speech depend. The formal treatment of the qui clause, for example, would depend in part upon the form of the clause itself, the mode, person and voice and to some extent the tense and number of the verb, as well as upon the other significant words in the clause. But it would depend quite as much upon the antecedent and observa- tion in this direction would not be confined to the pres- ence or absence of demonstratives or a certain number of adjectives, multus, unus, dignus, a superlative, but should be carried into minute subdivisions based upon the meaning of the noun. Proper nouns should be dis- tinguished from common, names of persons from names of things, abstracts from concretes, and so on. Dis- tinctions even more minute may be made with good results. Thus in Plautus the qui clause after the name of a character in the play is almost always colorless, the subordinating function is slightly felt and suggestions of cause or contrast or purpose are rare. But the names of gods or mythological heroes are followed by qui clauses which even with the indicative show traces of 228 FORM, FUNCTION AND CLASSIFICATION group-function. The name of a god in an appeal Lb fol- lowed by a qui clause giving the reason for die appeal, as in AuL, 396, Apollo, subueni . . ..- qui in re tali aliis iam subuenisti antidhac (the text is somewhat corrupt). The mention of a god leads to the addition of his special characteristic (Mercurius, Touts qui nuntius perhibetur, numquam aeque patrl Suo nuntium lepidum attulit, quam ego nunc meae erae nuntiabo, Stick., 274) or the mention of a mythological character is followed by a 7/// clause mentioning his most famous deed (Nam isti quidem herele orationi Oedipo Ojmst coniectore, qui Sphingi interpret fuit, Poem., 443 f.). But as the reason for introducing the name of a god or hero is usually connected with his attributes or deeds, these clauses almost invariably have a causal-adversative force: "Mercury, though he carries messages to Jove, never had a finer piece of news;" "this speech needs an Oedipus, for he could read any riddle." Similar minute differences in tin 1 function of the qui clause may be detected by dividing names of things from names of persons and concretes from abstracts. Taken by themselves they are of slight account, but when a number of them arc put together, they make a distinct and solid contribution to the psy- chology of the qui clause. . Similar variations may be noted in the ne clause by subdividing according to the meaning of the leading verb. The history of the process by which the si clause has come to express various shades of probability or possibility or unreality will never be unravelled by the functional classification now in vogue; this extremely interesting subject has lain for years untouched because of the barrenness of the method of treatment. But a careful formal study 1 will un- 1 Such as that begun by H. C. Nutting in the Amer. Jouru. of Philol., XXI, 3 (1900), No. 83, pp. 260 ff. 229 LATIN SYNTAX doubtedly give results of interest both to the psychology and to the history of the conditional sentence. The advantages of a classification by form, including under that head word-meaning, may be summed up in the statement that this kind of syntactical treatment rests solidly on facts and leads directly to processes and laws. Syntax is exposed to two dangers or, it may be said, is at present suffering from two evils. The first is the evil of theoretical speculation, dealing with periods where no facts are obtainable or, within historical periods, too far removed from the facts. The second is the evil of unconnected and meaningless accumulation of facts. The corrective of the first is more minute observation, the corrective of the second is more atten- tion to the meaning of facts. These two are really one, for they combine upon the single point of systematic accumulation of facts interpreted by many-sided obser- vation and study. For the accumulated observations of the meaning of facts, the patient interpretation of the phenomena within a narrow range, grows rapidly toward the understanding of larger fields, and many small laws put together make large laws. Speculation as to Indo- European origins is more attractive and more brilliant, and the mere recording of facts for others to interpret is easier, but neither is so fruitful as work which is at once more minute and more systematic. There has been much discussion, especially with reference to the natural sciences, of the best methods of classification. The points at issue have to do mainly with classification as a means of presenting scientific knowledge in intelligible form or, not infrequently, with a classification which shall serve the purposes both of presentation and of investigation. The difficulty of finding such a system comes from the attempt to com- 230 FORM, FUNCTION AND CLASSIFICATION bine the two different purposes. In syntax it is quite as great as in the natural sciences, but it is a difficulty which the scientific investigator is not called upon to face. He may properly re-aid classification as a mere convenience, a tool to lie used lor a certain end and then put aside. From this point of view the permanence of the classes is of no account nor is it necessary t<> con- sider whether the characteristics upon which the classes are founded are primary or secondary. Any number "I different classifications may he made for different pur- poses. Thus the ne clauses may be classified by the tense of the subjunctive, to find out whether this affects in any way the function of the clause. Then the same facts may be re-classified with reference to the person of the verb in order to determine whether the second person shows stronger evidence of connection with the prohibition than appears in the third person. A classi- fication according to the meaning of the verb of the //<• clause will bring out the peculiarities of expectation and of thinking wrongly, like ne erres, ne frnstr