Grogg.p65 228 College & Research Libraries May 2002 228 Full-text Linking: Affiliated versus Nonaffiliated Access in a Free Database Jill E. Grogg, Debra K. Andreadis, and Rachel A. Kirk Jill E. Grogg is an Assistant Professor and Instruction Services Librarian in the Mississippi State Univer- sity Libraries; e-mail: jgrogg@library.msstate.edu. Debra K. Andreadis is the Science Liaison/Reference Librarian in the William Howard Doane Library at Denison University; e-mail: andreadisd@denison.edu. Rachel A. Kirk is a Graduate Student in the School of Information Sciences at the University of Tennessee; e-mail: rkirk1@utk.edu. This article presents a comparison of access to full-text articles from a free bibliographic database (PubSCIENCE) for affiliated and unaffiliated users. The authors found that affiliated users had access to more full- text articles than unaffiliated users had, but the maximum level of full- text access achieved directly through PubSCIENCE was 45 percent. Both affiliated and unaffiliated users could increase their level of access to full-text content by using alternative avenues for access above and beyond use of the database linkage. However, such additional search- ing requires a higher level of information literacy on the user’s part. ick O’Leary has noted that “One of the Web’s greatest ac- complishments has been to re- inforce the notion that all in- formation ought to be free.”1 So if everything is free on the World Wide Web, why should universities spend millions of dollars on bibliographic and full-text databases for their libraries? The answer often offered by university libraries is deceptively simple. High-quality peer- reviewed materials are not always avail- able for free on the Web, and university libraries provide access to journals, con- ference proceedings, and other peer-re- viewed materials through costly agree- ments with journal publishers and bibliographic and full-text database pro- viders. Mary Ellen Bates has warned: “There ain’t no such thing as free lunch.” Bates has insisted that free lunch is “one- stop access to the archives of a number of periodicals, support for complex search- ing, and no charge for access, searching, or downloading any available informa- tion.”2 University libraries provide access to full text via a variety of media, including print and electronic. However, such ac- cess comes with the caveat that the uni- versity, first and foremost, provides ac- cess for its own community of students, faculty, and staff. Although anyone may physically enter the library and use its materials, not everyone may enjoy elec- tronic access to bibliographic and full-text databases or to the full text of journal ar- ticles via the desktop. Only those officially affiliated with the university have desk- top access to the electronic full text of jour- nals, to the electronic bibliographic records, and to the entire spectrum of the university library’s digital collection. Us- ers have made clear that they want the Full-text Linking 229 convenience of electronic access to full text. Diane J. Hoffman has stated plainly: “Scholars and researchers expect seam- less access to network-accessible materi- als.”3 The unaffiliated user must explore other options for consistent desktop access to the full text of journal articles. Options avail- able include pay-per-view, full or partial subscription, or, in rare cases, free avail- ability. Fortunately, free full text is becom- ing more common. New services such as FindArticles.com (the result of a unique partnership between The Gale Group and LookSmart) not only provide free article search services but also deliver the full text for no charge.4 Other free full-text options include PubMed Central and BioMed Cen- tral. Although more options for free text are becoming available, for the affiliated, the university library remains a crucial subsidizer and information provider for students, faculty, and staff. Imagine if stu- dents or faculty were forced to pay for each article they needed for a project or a class. Affiliation affords users the convenience of electronic access to full text, presumably greater access than unaffiliated users en- joy. Logically, the percentage of full text to which a user has desktop access should be greater for the affiliated user than for the unaffiliated. For the affiliated, access to the full text is available through a vari- ety of sources: aggregators with internal links to a digital collection of full text (e.g., Ovid or OCLC); publishers’ Web sites without a secondary search service (e.g., ElsevierScience’s ScienceDirect or Springer-Verlag’s LINK); the university library’s Online Public Access Catalog (OPAC); or free bibliographic databases with external links to publishers’ Web sites (e.g., Ask ERIC, PubMed, or PubSCIENCE). The authors of this article are concerned with the latter of these find- ing aids—free bibliographic databases with external links to full text. Specifically, the authors used PubSCIENCE to exam- ine the percentage of electronic full text to which a university-affiliated graduate student has access versus the percentage of electronic full text to which his or her postgraduate, professional counterpart has access. In other words, for a series of given queries and retrieved relevant documents, the authors examined the percentage of full-text access for affiliated versus nonaffiliated users. For testing purposes, ecology or environmental stud- ies was chosen as a subject area, and the study was conducted at the University of Tennessee (UT), a typical Association of Research Libraries (ARL) institution. Context of the Case Study One route of employment for postgradu- ate ecology or environmental studies is in the area of environmental consulting. Consulting can be done in a firm or as an individual in a freelance situation. Envi- ronmental consulting firms in the Knox- ville-Oak Ridge, Tennessee, area vary in size from one to 260 employees, but the large firms are the exception. The median number of employees is four for the fifty- nine Knoxville-Oak Ridge firms listed in Dun & Bradstreet.5 Because the firms are relatively small, the high cost to subscribe to or lease large full-text databases or the journals themselves are not viable op- tions. Nor do these firms usually have the funds to lease or subscribe to a biblio- graphic database. Therefore, the free availability of a bibliographic database, specifically PubSCIENCE, should be an attractive alternative. PubSCIENCE, according to the Web site, “provides users the capability to search across a large compendium of peer-reviewed journal literature with a focus on the physical sciences and other disciplines of concern to the Department of Energy (DOE).”6 PubSCIENCE is a bib- liographic database service developed by the DOE Office of Scientific and Techni- cal Information (OSTI) and closely mod- eled after PubMed. The journal citations within the database are contributed by participating publishers, information in- termediaries, and the DOE database maintained by OSTI. Likewise, participat- ing publishers provide links within a ci- tation to the full text of journal articles. 230 College & Research Libraries May 2002 Such links move the user out of the PubSCIENCE environment to the journal publisher’s Web site. Bette Brunelle notes that this type of full-text linking is a “nice fit” because: the publisher model [for full text] lacks a way to make the body of lit- erature searchable—and the distrib- uted model provides it. A large userbase is already comfortable with using bibliographic databases as the gateway to the literature.7 If the affiliated user ’s university or employer has a subscription to the elec- tronic version of the journal, the access to the full text should be seamless. The un- affiliated user, unless he or she maintains a subscription, must explore other op- tions, such as pay-per-view, document delivery, or using another system to find free full-text access.8 As mentioned previously, a user affili- ated with UT should have seamless ac- cess if the UT libraries carry a subscrip- tion to the electronic version of the jour- nal. UT is a Carnegie class I institution, and its library system is an ARL member. According to the 2000 edition of The Carnegie Classification of Institutions of Higher Education, UT is a Doctoral/Re- search University—Extensive. This clas- sification is based on UT’s degree-grant- ing activities from 1995–1996 through 1997–1998.9 According to ARL statistics, UT awarded 248 Ph.D.’s in 1999, ranking 52nd out of the 111 ARL institutions. The degrees were awarded across forty-five fields, ranking UT 68th out of the 111 ARL institutions.10 Of the 111 ARL members, UT is ranked 53rd for total current seri- als, with 23,094. This total represents the combination of current serials purchased and current serials not purchased.11 The ARL Supplementary Statistics 1998– 99 indicates that the average dollars spent for electronic serials was $639,281. UT re- ported spending $449,695, an expenditure closer to the 1997–1998 average of $449,971. For purposes of comparison, following are expenditures for electronic serials from the University of Kentucky (UK), Louisiana State University (LSU), and the University of South Carolina (USC). For 1998–1999, UK spent $724,098, LSU spent $221,017, and USC spent $582,160.12 These three institutions were chosen for comparison because they are UT’s competing peer institutions and are similar in size. Taking into consideration individual and institutional policies, de- cisions, and budgets, the ARL statistics for 1998–1999 indicate that the UT library system is a relatively average ARL insti- tution with a generalizable environment. Review of Literature Greg R. Notess has provided an under- standing of the origins of PubSCIENCE and its relationship to PubMed. He clari- fied the intended scope of PubSCIENCE and explained that the service does not cover all the physical sciences but, rather, only the portion considered of interest to the energy community. His article does not address the main concern of the au- thors of this article, which is the availabil- ity of full text based on access methods (affiliated or unaffiliated).13 Peter Jacso also examined PubSCIENCE and com- pared its competitive value to fee-based databases such as EnergyLine and the DOE Energy Science and Technology Database (ETSD). Jacso specifically exam- ined how PubSCIENCE is both more and less than ETSD. PubSCIENCE provides less than ETSD in its searching capabili- ties but offers more because PubSCIENCE includes a sister database, the DOE Infor- mation Bridge. The Information Bridge offers free access to the full text (PDF for- mat) of DOE reports and gray literature. Jacso concluded that “PubSCIENCE is an excellent resource for those who cannot afford the fee-based databases” and that “it can easily become the preferred source also for those who currently subscribe to one or more of the commercial data- bases.”14 Neither Notess nor Jacso explored the advantages and disadvantages of the full- text links to publisher Web sites within PubSCIENCE and neglected to discuss Full-text Linking 231 the implications of these links for affili- ated and unaffiliated users. Robert E. Donohue has offered a comprehensive review of PubSCIENCE and has de- scribed planned enhancements for the database, including a migration to an ORACLE-based system; and Walter L. Warnick has outlined the history of PubSCIENCE.15,16 Like Notess and Jasco, however, neither Donohue nor Warnick commented on access for unaffiliated ver- sus affiliated users. Laura Gor don- Murnane did compare fee and free alter- natives for federal government search tools (FedWorld, usgovsearch.com, Google/Unclesam, and GovBot), con- cluding that the “best bet is to use mul- tiple engines, fee and free, to achieve bet- ter search results,” but for affiliated-ver- sus-unaffiliated access, significant oppor- tunities for research remain to be ex- plored.17 Brunelle discussed the growing de- mand for full-text database services. She began by identifying the potential of full- text linking and then presented three models of how this access is provided as well as the strengths and weakness of each scenario. Brunelle presented the publisher-supplied full-text concept, the aggregator-supplied full-text model, and the distributed “linked” full text.18 Lawrence Krumenaker advised research- ers on how to navigate between the dif- ferent scenarios to access needed docu- ments and overcome obstacles posed by the current publishing environment.19 Laurie A. Preston, Corinne M. Ebbs, and Judy Luther tested the value of full-text access through a sampling of journal is- sues available online and provided a model for evaluating the true functional- ity of full-text databases.20 Walt Crawford has discussed the disadvantages of rely- ing exclusively on one access method.21 Jacso examined one- and bi-directional linking and the accompanying partnerships between publishers and companies.22 He also examined new models of access to full text, including Northern Light, Contentville, Electric Library, and FindArticles.com. These services offer the bibliographic records for free and employ several models of providing the full text. Northern Light and Contentville have a pay-per-view option for full text, but as Jacso noted, the price is “only a fraction of what you would pay to traditional docu- ment delivery services.”23 FindArticles.com provides the full text from 300 journals for free, and the Electric Library offers unlim- ited full-text access in exchange for a monthly fee, less than $10 according to Jacso. Ann Okerson has offered a discus- sion about the history and future of e-jour- nal linking.24 Based on the review of the literature, there appears to be no current research that compares the full-text linking feature of free bibliographic databases, such as PubSCIENCE, by access method. How- ever, enough work has been done in the field of database comparison to under- score the significance of this area of re- search, and tested, widely accepted meth- odologies exist to use in the endeavor. Methodology: Phase One The authors of this article developed ten questions in the field of ecology by inter- viewing researchers at two environmen- tal consulting firms about the kinds of projects on which they worked (table 1). From examples of projects, one of the au- thors, who has advanced subject knowl- edge of ecology, crafted questions. The questions were designed to be broad enough to assume major journal cover- age and narrow enough to provide a ba- sis for evaluating relevance and precision. Identical literature search topics were run simultaneously in PubSCIENCE, one search as an affiliated user with a UT ac- count and one as an unaffiliated user with an America Online (AOL) subscription account. The searches were performed at the same time on two adjacent comput- ers. In PubSCIENCE, the search was re- stricted to articles available in full text so as to eliminate any purely bibliographic results. When search results were re- trieved for each of the ten queries, the af- filiated and unaffiliated searchers fol- 232 College & Research Libraries May 2002 lowed the full-text links from the biblio- graphic records to the publishers’ Web sites. The authors then noted access to full text for the affiliated (UT) and unaffiliated (AOL) approaches. After all ten searches were performed and all links to full text explored, the total number of full-text ar- ticles available for the affiliated-versus- unaffiliated user was examined and tested for significance with a chi square contingency test. The subject expert then evaluated the ten result sets. Excluding articles to which the affiliated and unaffiliated searchers had full-text access through a link in PubSCIENCE, she chose the most relevant article from each set, thus producing a list of ten relevant articles. Using this list, the authors attempted to find the articles’ full text via avenues other than PubSCIENCE. For the affiliated user, the authors searched: • UT’s Online Public Access Catalog (OPAC) • UT’s e-journal list (electronic jour- nals to which the UT subscribes either in- dividually or as a package from specific publishers) • A document delivery service (Carl UnCover) • Several commercial full-text data- bases (Lexis-Nexis, ProQuest, JSTOR, TABLE 1 PubSCIENCE Questions and Search Strategies Question 1. What are the effects of water warming by hydroelectric plants on fish? 2. What are the deleterious effects on PCBs on wildlife? 3. Does mercury contamination affect amphibian development? 4. How does clear-cutting impact small mammal populations? 5. Does wastewater treatment increase forest productivity? 6. Are reconstructed or man-made wetlands used by wildlife as readily as natural wetlands? 7. What toxins in fish pose the greatest hazards to human health? 8. Is drinking water near superfund sites a health hazard? 9. Can microbes be genetically engi- neered to remove heavy metals from streams? 10. Can wastewater be effectively filtered to produce drinking water by biologi- cal rather than chemical means? Strategy (hydroelectr* OR power plant*) AND fish* AND warm* (pcb* OR polychlorinated biphenyl*) AND (wildlife OR waterfowl OR wild animal*) mercur* AND (amphibia* OR frog* OR salamander* OR toad*) AND (develop* OR metamorphos*) (clearcut* OR clear-cut* OR clear cut*) AND small AND mammal (wastewater OR sewage) AND forest* AND treatment (reconstruct* OR construct* OR man- made) AND (wetland* OR marsh*) fish* AND toxi* AND human AND consum* water* AND superfund microb* AND genetic* AND engineer* AND (heavy metal* OR lead OR mercur*) (wastewater OR sewage) AND drink* AND biolog* Note: Operators must be in all caps to be understood by system. Full-text Linking 233 Dow Jones Interactive, Gale Group’s InfoTrac, and ScienceDirect) For the unaffiliated user, the authors searched free Web sources and noted prices where available: • HighWire Press (http://highwire.stanford.edu) • Google (http://www.google.com) • Northern Light Special Collection (http://www.northernlight.com/ docs/specoll_help_overview.html) • FindArticles.com (http://www.findarticles.com) In Google, the authors searched us- ing article title, and the results retrieved were few enough that all results were examined. The authors searched using journal title in HighWire, Northern Light, and FindArticles.com. Methodology: Phase Two Because the percentages of returned ar- ticles from phase one were surprisingly low (with only 12% return for the affili- ated searcher and 3% for the unaffili- ated), the authors considered the possi- bility that the research topic, though certainly a valid field of inquiry, may have been overly specific for this data- base, thereby skewing the results. There- fore, the scope of the searches was ex- panded and the authors used a second approach for searching the PubSCIENCE database. This approach assumed that if one were to examine each publisher rep- resented in the PubSCIENCE database by selecting two of its journals and at- tempting to retrieve a full-text article from each, the results would better rep- resent the percentage of full-text articles available via the PubSCIENCE database. A single, ubiquitous term was used (“state”) to search each publisher subset of the database. If only one journal title was returned, it was checked against the list of titles present from that publisher. In all cases, a single title returned was the only title provided by the publisher in question. For the publishers that had multiples titles present in PubSCIENCE, the authors looked at two publications from each. Again, the titles were ex- plored simultaneously on an affiliated computer (UT) and an unaffiliated (AOL) computer. The total number of links followed in this manner was fifty- eight from a total of thirty-three publish- ers, with eight publishers providing a single journal. The authors repeated the tracking of unavailable works as described in phase one for the affiliated and unaffiliated user. The same sources were searched as pre- viously, with the exception of the change from the Carl UnCover service to Ingenta.com, which had purchased UnCover in the interim. All thirty-two articles that were not retrieved by the af- filiated user were pursued in this man- ner. Results: Phase One As shown in table 2, affiliated users had better access than unaffiliated users. To test to see whether this difference was sig- nificant, the authors used a chi-square contingency test and, as table 3 shows, the difference was highly significant (p = 0.002). Still, very few of the articles re- turned were available in full text for ei- ther the affiliated or the unaffiliated user. The full text of some of the articles re- trieved should have been available to af- TABLE 2 Number of Returns Accessible for Each of the Ten Questions Question Affiliated Unaffiliated (UT) (AOL) 1 0/5 0/5 2 5/27 0/27 3 0/6 0/6 4 2/31 2/31 5 0/18 0/18 6 2/22 0/22 7 7/30 1/30 8 2/13 1/13 9 2/5 1/5 10 1/12 0/12 Total 21/169 5/169 234 College & Research Libraries May 2002 filiated or unaffiliated users, but the links from PubSCIENCE were not working properly. Six articles of the 169 had broken links. One article was posted in the wrong font so that the title appeared in Greek characters, even though it was an English-lan- guage title. Many of the retrieved ar- ticles were from the same publisher— NRC from Canada. NRC has changed its URL, and PubSCIENCE does not reflect this change. When NRC titles are accessed, the user is redirected to the home page of the publisher where it is then necessary to ferret out the journal and article of interest. The authors then sought to find the most relevant articles (one from each of the ten sets) through avenues other than links from PubSCIENCE. For the affiliated user, two of ten are available online with- out a fee. One is available through ProQuest; the other is available through the UT e-journals page. Of these two ar- ticles, only the one found in a journal listed on the UT e-journals page is tagged in the OPAC as being available in an online version. The journal containing the article available via ProQuest is not tagged in the OPAC, even though UT has a print subscription. None of the articles is available via any of the visited Web avenues. Going beyond electronic full-text op- tions, two additional titles were available via UnCover for the affiliated user at no charge (although these titles are held by UT in print, they are not listed as such in the UnCover database). For the affiliated user, then, four of the ten titles are avail- able at no charge without requiring a trip to the library. One title is available via UnCover for a fee of $19.50, which UT would not subsidize because the title is listed in UnCover as being owned in print. Another three titles are available from the publisher for $20.00 each. Thus, for a total charge of $79.50, eight of the ten articles could be recovered via the computer or fax machine. For the unaffiliated user, the only methods of access were ordering directly from the publisher for a fee and UnCover for a fee. Four were available via UnCover (three at $14.50 each and one at $19.50), and three additional titles were available directly from the publisher for $20 each. The unaffiliated user could receive seven of the ten articles for a total charge of $123. For either the affiliated or unaffiliated user, five of the titles are available in print form in the UT library. All of these also are available by one of the other previ- ously mentioned methods for UT users, although two if the titles would incur charges. For the unaffiliated user, one additional print title is accessible that is not available via electronic means. Table 4 summarizes the findings for the affili- ated and unaffiliated user for both fee and no fee access via avenues other than links from PubSCIENCE. Results: Phase Two The suspicion that the subject specificity (ecology) employed in phase one skewed the results proved to be true. In phase two, the authors examined a more sys- tematic cross section of the database and found that the percentage of access to full- text articles increased in both affiliated and unaffiliated users. As table 5 shows, affiliated users still had better access than unaffiliated users (p = 0.002). Overall, the percentages of accessible ar- ticles were greater when all publishers TABLE 3 Chi-square Contingency Table for Comparing Access for Affiliated and Unaffiliated Users From Phase One Affiliation # Articles # Articles Not Accessible Accessible UT 21 148 AOL 5 164 Total 21/169 5/169 The lack of additional access via the Web or low-cost pay-per-view services such as Northern Light was very disappointing. Full-text Linking 235 were sampled in this set of results (45% for affiliated users; 14% for unafffiliated users). The ratio of articles retrieved by unaffiliated users versus articles retrieved by affiliated users shifted from 24 to 31 per- cent. Given the broad selection of publish- ers sampled and the fact that electronic access is usually determined at the pub- lisher level as opposed to the individual journal level, 31 percent appears to be a reasonable estimate for the percentage of electronic access attained by unaffiliated users compared to affiliated users. Problems also arose in this set with broken links and links that led the user to the publisher ’s Web site or journal Web site rather than to the actual article of in- terest. This occurred in seven of the fifty- eight articles tested. The authors again sought to find the articles not listed as available via the cata- log link through other avenues (table 6). For the affiliated user, eight of thirty-two not available via links in PubSCIENCE are available online without a fee. One is available via the UT e-journal list, three through ScienceDirect, one through InfoTrack, one preprint is available via the Web, and two others are available directly from the Web (one found via Google and one free with registration at the journal publisher ’s page). Of these eight articles, only the one found in a journal listed on the UT e-journals page is tagged in the OPAC as being available in an online version. Four ad- ditional journals are tagged as being available in the OPAC, two from Dow Jones and two from ProQuest, but none of the ar- ticles needed was available from these sources. Six additional titles were available via Ingenta.com for the affiliated user with no fee (the print version is not avail- able from UT and the charge for each article is less than UT’s maximum allowable subsidy of $35). This means that the af- filiated user can view fourteen of the thirty-two titles for no charge and without requiring a trip to the library. Six additional titles are available via Ingenta.com for a total cost of $132.37, which UT will not pay because it owns the titles in print form. Four additional titles are available from the publisher for a combined fee of $151. Thus, for a total charge of $283.37, twenty-four of the thirty-two articles could be recovered via the computer or fax machine. For the unaffiliated user, the three pre- viously identified articles were available free via the Web (preprint, Google, regis- tration at the journal publisher’s page). The only other successful means of access was to order directly from the publisher for a fee and from Ingenta.com for a fee. Fifteen were available from Ingenta.com for a to- tal cost of $355.16, and four additional titles were available directly from the publisher for a total cost of $151. The unaffiliated user could receive twenty-two of the thirty-two articles (articles originally not available through PubSCIENCE for the affiliated user) for a total charge of $506.16. TABLE 4 Fee and No Fee Access for Affiliated and Unaffiliated Users from Phase One Access Means Affiliated Unaffiliated (UT) (AOL) Full-text databases 1 0 UT e-journals page 1 0 Free Web 0 0 UnCover (no fee) 2 0 UnCover (fee) 1 4 Publisher (fee) 3 3 Total 8 7 TABLE 5 Chi-square Contingency Table for Comparing Access for Affiliated and Unaffiliated Users From Phase Two Affiliation # Articles # Articles Not Accessible Accessible UT 26 32 AOL 8 50 236 College & Research Libraries May 2002 TABLE 6 Fee and No Fee Access for Affiliated and Unaffiliated Users from Phase Two Access Means Affiliated Unaffiliated (UT) (AOL) Full-text databases 4 0 UT e-journals page 1 0 Free Web 3 3 Ingenta.com (no fee) 6 0 Ingenta.com (fee) 6 15 Publisher (fee) 4 4 Total 24 22 Fourteen of the thirty-two titles are available in print form within the library. Four of them are unavailable in any other format mentioned previously for the af- filiated user. The six titles available for a fee from Ingenta.com are free in print form within the library. Discussion Although PubSCIENCE is a good re- source for the unaffiliated user to gain access to bibliographic information on scientific literature, it is not particularly useful for retrieving full text. Part of the problem was created by broken or out- of-date links. This also affected affiliated users because some of the broken links pointed to journals to which UT has elec- tronic subscriptions. Although these links should take users directly to the full text of the articles, they, instead, send users to the front page of the publisher, thus forc- ing the user to search for the article of in- terest. Although affiliated users could access more articles in full text (tables 3 and 5), the authors were surprised to find how few articles were available in full-text form (12% to 45%), even to the affiliated user, directly through PubSCIENCE. However, access is even more limited for the unaffiliated user: Electronic access to the full text without charge was found for only 3 to 14 percent of the articles. When search efforts were restricted to articles unavailable by direct means through PubSCIENCE, free electronic ac- cess increased for the affiliated user (20% and 44% retrieval for each phase, respectively). However, this process in- volved a lot of additional searching and still yielded a surprisingly low return for the effort. The outlook for the un- affiliated user is even bleaker. The lack of additional access via the Web or low-cost pay- per-view services such as Northern Light was very dis- appointing. There were other options for retrieving the articles to the desktop or office without having to make a trip to the library. These included paying for a copy directly from the publisher or using a document delivery service (UnCover/ Ingenta.com). These options greatly in- creased the percentage of articles that could be accessed (up to 80% for the af- filiated user and 70% for the unaffiliated user in phase one; 75% for affiliated and 69% for unaffiliated in phase two), but the coverage of these services overlaps some- what. Therefore, the user would have to run searches in multiple databases to find the best price available for office deliv- ery. The total price to recover the articles may be prohibitive as well. There are com- plications even for the affiliated user with document delivery services because in phase one the UnCover documents for UT-affiliated users includes one title that is actually held by UT. This would put UT in the position of paying for access to a volume that it holds in its collection or of informing the user that he or she must pay for the document delivery service charge after the fact. The lack of connec- tion from the OPAC to alternative deliv- ery options also is problematic. Having to search in all the various places that full- text access might be found is time-con- suming and requires knowledge that the average user may not have. Conclusion Research in the first phase was skewed by the fact that only a limited set of jour- Full-text Linking 237 nals in PubSCIENCE had relevant articles for the authors’ topic. Because the sub- ject area was not well represented by jour- nals within PubSCIENCE, the results were biased by one particular publisher ’s journals that came up repeatedly. The overall results when including all pub- lishers were more encouraging, but fewer than half of all the articles were available directly from the connections made from PubSCIENCE. This means that an affili- ated user, as represented by student en- rolled at the University of Tennessee, who searches the PubSCIENCE database can obtain direct and free access to 12 to 45 percent of the articles that meet the crite- ria of a search strategy. The unaffiliated user, unfortunately, will be only 24 to 31 percent as successful as the affiliated user. Exclusive reliance on PubSCIENCE for full-text articles results in a greater than 50 percent chance that an affiliated user will be unable to retrieve an article for free. For unaffiliated users, only a 3 to 12 percent probability of direct and free full- text access exists. On the whole, PubSCIENCE is a good bibliographic resource for all users and also may be a good resource for full-text access, depending on the number of jour- nals to which an affiliated user’s organi- zation subscribes. However, this access comes with a caveat: Even if the user ’s organization has a subscription to a spe- cific journal, access may take the form of other online full-text services or only a print subscription. In addition, because the links in PubSCIENCE point to pub- lishers’ Web sites, access to full text may be denied. Further, because there are so many options for full-text delivery, aca- demic institutions are faced with linking OPAC records for journals and integrat- ing records for their subscribed biblio- graphic databases, subscribed full-text databases, and direct subscriptions to publisher-supplied full text, among other options. To return to the question posed at the beginning of this article, the brief foray into the unaffiliated world proves that all information is not free via the Web. As a matter of fact, very little high-quality scholarly information is available for free via the Web. Perhaps Bates is correct, “There ain’t no such thing as free lunch.”25 Notes 1. Mick O’Leary, “Gale Goes Free,” Econtent 23, no. 6 (Dec. 2000): 71–73. 2. Mary Ellen Bates, “Tanstaafl: In Search of the Free Lunch and No-Cost/Low-Cost Full- text Archives,” Searcher 8, no. 6 (June 2000): 55–59. 3. Diane J. Hoffman, “Think Links: Full-text Linking Projects,” Online 25, no. 1 (Jan./Feb. 2001): 40–46. 4. Peter Jacso, “FindArticles.com Embraces Free Content Trend,” Information Today 17, no. 10 (Nov. 2000): 38–40; O’Leary, “Gale Goes Free.” 5. Dun & Bradstreet. 2001. Dun’s Market Identifiers. Databases online. Available via DialogClassic from http://www.dialogclassic.com/. (Accessed 17 February 2001.) 6. Department of Energy Office of Scientific and Technical Information. 2001. PubSCIENCE. Available from http://pubsci.osti.gov. (Accessed 15 February 2001.) 7. Bette Brunelle, “Quieting the Crowd: The Clamour for Full Text,” Online & CD-ROM Re- view 23, no. 5 (Oct. 1999): 297–302. 8. Jill E. Grogg and Carol Tenopir, “Linking to Full Text in Scholarly Journals: Here a Link, There a Link, Everywhere a Link.” Searcher 8, no. 10 (Nov./Dec. 2000): 36–42. 9. The Carnegie Classification. 2000. Available from http://www.carnegiefoundation.org/Clas- sification/. (Accessed 16 February 2001.) 10. Ibid. 11. Association of Research Libraries. 2000. “Association of Research Libraries Statistics.” Avail- able from: http://fisher.lib.virginia.edu/arl/index.html. (Accessed 16 February 2001.) 12. Martha Kyrillidou, comp. and ed., ARL Supplementary Statistics 1998–99 (Washington, D.C.: Association of Research Libraries, 2000), available from http://www.arl.org/stats/sup/ sup99.pdf. (Accessed 16 February 2001.) 13. Greg R. Notess, “PubSCIENCE: Evolution or Devolution,” Econtent 23, no. 1 (Feb. 2000): 64–66. 238 College & Research Libraries May 2002 14. Peter Jacso, “Energy Science and Technology,” Link Up 17, no. 3 (May/June 2000): 3, 8. 15. Robert E. Donohue, “PubSCIENCE: Accessing Scientific and Technical Journal Informa- tion at the Desktop,” Serials Review 27, no. 1 (2001): 18–25. 16. Walter L. Warnick, “PubSCIENCE: Hyperlinking to the Physical Sciences,” Learned Pub- lishing 14, no. 1 (Jan. 2001): 65–67. 17. Laura Gordon-Murnane, “Government Search Tools: Evaluating Fee and Free Search Al- ternatives,” Searcher 7, no. 8 (Sept. 1999): 66–76. 18. Brunelle, “Quieting the Crowd.” 19. Lawrence Krumenaker, “A Dillar, a Dollar—Where’s That Online Article, Scholar?” Searcher 7, no. 8 (Sept. 1999): 36–40. 20. Laurie A. Preston, Corinne M. Ebbs, and Judy Luther, “Full-text Access Evaluation: Are We Getting the Real Thing? At James Madison University, Report of a Workshop at the 1997 NASIG Conference,” Serials Librarian 34, no. 3/4 (1998): 301–5. 21. Walt Crawford, “Here’s the Content—Where’s the Context?; What Vendors Won’t Tell You about Full-Text,” American Libraries 31, no. 3 (Mar. 2000): 50–52. 22. Jacso, “Partnerships on the Web,” Computers in Libraries 21, no. 1 (Jan. 2001): 72–75. 23. ———, “Finding Free Databases to Replace Fee-based Information Services,” Computers in Libraries 20, no. 10 (Nov./Dec. 2000): 61–62. 24. Ann Okerson, “Are We There Yet? Online E-resources Ten Years After,” Library Trends 48, no. 4 (spring 2000): 671–93. 25. Bates, “Tanstaafl.”