College and Research Libraries ARL Academic Science and Technology Libraries: Report of a Survey Julie M. Hurd The ARL includes among its United States and Canadian member li- braries many of North America's largest research collections in science and technology. These libraries often serve as models or benchmarks for other institutions with respect to collection development and man- agement, and the provision of information services. Science and tech- nology librarians have used survey techniques to gather data on ARL science and technology collections for nearly ten years. This report pro- vides findings from a survey of ARL academic science and technology libraries conducted during 1993 and 1994, and updates three earlier surveys . Seventy-five ARL academic libraries returned questionnaires for a response rate of 69 percent. This article describes survey findings on organizational structures, collections, expenditures, and services. Some comparisons are made to earlier surveys. fliii!!!ii!~ he Science and Technology Sec- tion (STS) of ACRL represents librarians and specialists in the fields of science and technol- ogy. In 1984 the section established an ad hoc task force to collect statistics on stand- alone science and technology libraries. One purpose of the original project was to compile comparative data that science librarians could use in drafting propos- als and supporting funding decisions. At that time, there was no organized effort to collect data on science and technology libraries and collections. Association of Research Libraries (ARL) libraries pub- lish descriptive statistics, but these data do not break down the subject areas that are of interest to science librarians. The original task force has been continued by the ACRL STS Committee on Compari- son of Science and Technology Libraries (the Comparison Committee). The Com- parison Committee is charged with col- lecting, analyzing, and distributing com- parative data on North American aca- demic science· and technology libraries. To fulfill its charge, the committee has conducted various surveys of science and technology libraries; the results of several of these have been published in library literature. 1 Background: Earlier Surveys The Comparison Committee has con- ducted surveys for ten years. 2 The earli- est efforts focused on collecting data from Julie Hurd is a Science Librarian in the University Library at the University of Illinois at Chicago, and chaired the Committee on Comparison of Science and Technology Libraries from 1992 to '1995. 144 ARL Academic Science and Technology Libraries 145 the strongest central- ized science collec- tions, whether housed in stand-alone or free- standing facilities or existing as separate divisions in larger li- braries. Subsequent surveys added librar- ies with other variant forms of organization to encompass the di- versity of science and technology collections that might be exclud- ed by too narrow defi- nitions. TABLEt Comparison Committee Surveys Survey Survey Population Return 1. (1984-1985) 40 ARL stand-alone libraries 24 (60%) 2. ( 1986-1987) 145libraries (ARL + ACRL 45 (31 %) members with > 1 million volumes) Definition Survey' 107 ARL libraries 91 (85%) (1987-1988) Historical Survey2 118libraries (ARL +) 97 (82%) 3. (1988-1989) 148 libraries, as in 1986-1987 65 (44%) 4. (1992-1993) 118 ARL libraries 75 (69%) 'The Definition Survey examined existing physical and administrative structures in science and technology libraries. The Comparison Committee based its 2The Historical Survey traced the evolution of physical and administra- tive structures in science and technology libraries from the 1940s through the 1980s. first survey instru- '--------------------------' ment on the formulation used for the Annual Statistics of Medical School Librar- ies in the United States and Canada (com- piled by the Houston Academy of Medi- cine). The committee then revised the for- mulation to make it suitable for a science and technology library population. Over the years, Comparison Committee sur- veys have consisted of many of the same questions, with changes, additions, or de- letions made to reflect differing focus in the data gathering and new develop- ments in the profession. Almost everyone involved in the effort realized that major science and technology collections exhibit great variety in every feature, and design- ing a survey instrument that every re- spondent can complete in its entirety proved difficult. The Comparison Com- mittee was able to assemble a body of data that now spans almost ten years and may be used to track trends and new devel- opments. The ARL member list provided a start- ing point for defining each survey popu- lation; additionally, strong science and technology libraries that were not ARL members were sometimes included. Each survey conducted by the committee que- ried many of the same libraries, but it should be recognized that variations in the populations surveyed do exist and this must be taken into account in any lon- gitudinal comparisons that are made.. us- ing committee data. Three previous general surveys mod- eled on the above-mentioned survey of medical libraries have been carried out. These are referred to here as Survey 1, Survey 2, and Survey 3; the survey re- ported in this article is the fourth in this series. In addition, the Comparison Com- mittee conducted two more specialized surveys, both directed toward describing physical and administrative structures. The Definition Survey defined and de- scribed organizational and administrative structures prevalent at the time research- ers conducted it. The Historical Survey traced the evolution of physical and ad- ministrative structures from the 1940s through the 1980s. Table 1 defines the survey populations and shows response rates for committee surveys. Methodology Before undertaking the current survey, the Comparison Committee formed a subcommittee to contact former commit- tee members who worked on the 1988- 146 College & Research Libraries 89 survey and who might be able to iden- tify problems with the instrument that led to low return rates. The responses re- ceived informed the design and method of distribution of the current survey. For example, some institutions were unable to respond because they did not keep separate science and technology statistics. For these institutions the committee could expect few data. In other institutions, the Institutions had the option of completing one survey for all their science and technology collections, or individual surveys for separable collections for which they gathered statistics. questionnaires apparently never reached the individuals best suited to provide the data requested. Still other respondents perceived the survey to be too long and detailed. After discussing the findings of this subcommittee, the Comparison Commit- tee decided to define the current survey population precisely-all ARL libraries. (At the time of distribution of the ques- tionnaires, there were 119 ARL libraries, of which 108 were academic libraries.) Recognizing that even in this group there was great diversity, the committee made an effort to request some data that all re- spondents would have in common by vir- tue of their participation in the annual ARL member survey. Committee mem- bers would be in personal contact with respondents and would encourage the return of the survey even if not all infor- mation was available. In this way, the committee hoped to elicit some data from each institution-at the very least suffi- cient details to be able to characterize the type of organizational strucn,Ire for each. Institutions had the option of completing one survey for all their science and tech- nology collections, or individual surveys for separable collections for which they gathered statistics. March 1996 The committee eliminated questions for which few institutions collected data from the survey instrument. As there were many new technologies employed in libraries, new questions were written to elicit information on their uses in the provision of science and technology in- formation services. For example, the com- mittee added questions on local network- ing and use of the Internet. The final ques- tionnaire, including definitions and ex- planatory material, was thirteen pages long, counting a detachable page request- ing salary information that could be op- tional or returned separately to preserve confidentiality. (The survey instrument is not reproduced here, but a copy is avail- able from the author upon request.) Committee members each contacted a subset of the population to identify the individual in each library to whom the questionnaire should be directed. Follow- ing the late summer I autumn 1993 mail- ing of the survey, committee members stayed in communication with the re- spondents to offer encouragement and answer questions. By spring 1994, most libraries that were able to participate had returned the questionnaires and data analysis began. The committee chair made a preliminary report of the survey findings at the STS Forum on Science and Technology Library Research during the ALA annual conference in June 1994. During late summer 1994, committee members made reminder calls to librar- ies that had indicated their willingness to participate but had not yet returned the survey. By October 1,1994,75 institutions had returned 152 responses for a 69.44 per- cent return rate. Several institutions had indicated that they would not be able to participate. For those libraries, the com- mittee contact persons made an effort to learn about their organizational structure. In a number of cases, nonresponding li- braries totally integrated collections with no separable science and technology sta- tistics. ARL Academic Science and Technology Libraries 147 The author used Microsoft Excel for the Macintosh to enter returned survey data into a spreadsheet. The data were distrib- uted in both disk and paper formats as raw numbers to committee members as- sisting in data analysis. With this survey, the committee deviated from the past practice of distributing raw data to par- ticipants. The survey offered respondents the option of choosing to share their re- sponses, with their institutions identified, or maintaining confidentiality. Enough respondents chose confidentiality to sug- gest that only aggregate figures should be reported for this survey. Hence, no re- spondents will be identified by name in any of the committee's reports. Organizational Structure The Definition Survey and the Historical Survey carried out by the committee sev- eral years ago provided insights into the physical structures found in major science and technology libraries. These two sur- veys offer a snapshot of the structures found at that time and a retrospective perspective. Although the population varied somewhat from the current survey, earlier surveys included many of the same libraries. A comparison of the list of libraries responding to the Definition Survey with this survey found that sev- enty-two libraries were in both groups for an overlap of 96 percent.3 This high de- gree of overlap with the populations stud- ied earlier should permit tracking of trends in library organization. The arguments for and against central- ization of academic collections are sum- marized by Leon Shkolnik, who con- cludes that "Clearly, the trend in academic libraries is toward greater centraliza- tion .... " 4 Comparison Committee sur- veys document this centralizing tendency for the physical resources in science and engineering. Holdouts to the trend ap- peared to be some of the oldest and larg- est libraries, mostly located in the North- east. Factors that appear to contribute to this centralizing trend include economic constraints (especially costs for duplicate journal subscriptions), security consider- ations, space pressures, and the increas- ing interdisciplinarity of research. Con- versely, mitigating factors include ad- vances in automation that support decen- tralized data input and use, electronic for- mats, new delivery options, and faculty preferences for small, conveniently lo- cated collections. These and other influ- ences are discussed by Shkolnik. The current survey updates the His- torical and Definition surveys and offers details on the variety of physical arrange- ments found in ARL academic science and technology libraries. Because the committee's explorations into the reasons for nonresponse pointed to a need for more options in describing physical or- ganization of collections, some new cat- egories were created. This survey em- ployed the following definitions of physi- cal organization in analyzing data: • stand-alone: a separate, multidisci- plinary science and technology library housed in its own building; • main divisional: a science and tech- nology division, with separate statistics, housed in a main library; • multisubject departmental: a divi- sional library with more than three dis- crete subjects in its collection housed with science and technology departments; • subject departmental: collections com- posed of one or two subjects housed in a department (note: some closely related combinations such as mathematics and sta- tistics were treated as a single subject); • decentralized: a library system with three or more subject departmental librar- ies and no major multisubject collections; • integrated: a science and technology collection nonseparable from an entire li- brary collection with no separate statistics; • hybrid: various combinations of di- visional, departmental, and integrated collections not fitting into the above cat- egories (i.e., "other"). Libraries classed as stand-alone, main divisional, and integrated sometimes re- 148 College & Research Libraries March 1996 FIGURE 1 Organizational Structures: A Continuum decentralized libraries 3+ branches mullti-subject departmental libraries main-divisional libraries distributed collections based near users centralized models, possibly with 1 or 2 subsidiary branches no discrete sci/tech collection no separable statistics ARL Academic Science/Technology Libraries ported subsidiary branches, which will be examined in more detail in the following data. The categories used to describe the physical structures reported might be A total of fifty institutions, more than 63 percent of those whose organization is known, reported the presence of centralized science and technology collections, whether in stand-alone, main divisional, or multisubject departmental libraries. viewed as a continuum ranging from a very decentralized arrangement with many departmental branches dis- persed about a campus to a fully inte- grated, single institutional collection. Figure 1 displays that continuum of struc- tures. Using the categories defined above, the author coded returned questionnaires and entered data into a master spread- sheet; information supplied by the re- spondents permitted a categorization ac- cording to type. In cases where returns represented departmental collections, the author also assigned a category for the institution as a whole. Organizational in- formation for nonrespondents was in- cluded, if available. A sort by type for sev- enty-nine institutions is displayed in fig- ure 2. Different shadings distinguish li- braries with and without branches. A to- tal of fifty institutions, more than 63 per- cent of those whose organization is known, reported the presence of central- ized science and technology collections, L------------------------------------------------------------------------ - ----- ARL Academic Science and Technology Libraries 149 FIGURE2 ARL Academic Sciffech Libraries: Organizational Types 35 30 EJ branches ~ no branches 25 20 15 10 decentralized hybrid integrated whether in stand-alone, main divisional, or multisubject departmental libraries. Library size is one characteristic likely to influence physical and organizational structure, as demonstrated in the analy- sis reported in the Definition Survey. ARL rankings provide an index that measures relative size of university libraries taking into account the number of volumes held, the number added during the previous fiscal year, the number of current serials, total operating expenditures, and the size of professional and nonprofessional staff main division multi·dept stand-alone ies in seventy-nine institutions. Several institutions reported the presence of more than one library in a particular category, whereas others reported the existence of libraries in two or more categories. The mean ARL rank for nonrespond- ing libraries was 61.4; smaller institutions are underrepresented in this survey. Ear- lier observations that the largest institu- tions are more likely to maintain decen- tralized collections are corroborated by this study. A mean ARL rank of 33.4 char- acterizes the decentralized science and (excluding student em- .--------------------------, ployees). Because this index is both generally respected and widely TABLE2 ARL Rankings used, the author se- lecteditasthemeasure _C_a_te~g~o~ry~---------------------------------------Mean ARL Rank #of libraries of size for this analysis. The author entered ARL ranks for libraries into the spreadsheets and computed mean ARL rankings for each of the organizational categories and for those libraries not respond- Decentralized libraries Stand-alone libraries Main divisional libraries Hybrid libraries Multisubject departmental libraries Integrated libraries All respondents All nonrespondents 33.4 50.2 51.5 62.3 62.8 71 52.7 61.4 24 31 13 9 14 12 79* 29 ing to the survey.s Table *Some institutions reported multiple libraries in a category or 2 summarizes the data libraries in more than one category; "all respondents" includes all for 103 separate librar- .____i_ns_t_itu_t_io_n_s_s_up_p_I_yi_n_g_an_y __ d_at_a_o_n_o_r_ga_n_iz_a_ti_o_na_l_s_tr_u_ct_u_re ________ __, 150 College & Research Libraries March 1996 TABLE 3 Characteristics of Ty~es Type #of Size (sq. ft.) Seats Volumes Active Total libraries Serials Staff Departmental* 161 6,571 77 62,692 740 4.78 Multisubject 12 18,304 251 176,323 1,843 17.51 Main divisional 13 25 ,754 250 223,175 2,343 10.32 Stand-alone 30 68 ,322 586 372,260 4,040 29.01 *Averages for all subjects; see table 4 for a breakdown by subject of departmental library. technology collections as associated with the larger libraries. Conversely, stand- alone libraries are found in institutions with a mean ARL rank of 50.2, much closer to the median rank. This study did not collect data that would provide an ex- planation for this observation, but it ap- pears likely that more factors than library size alone will be determinants for a stand-alone library. The Historical Survey identified an increase in the number of stand-alone libraries, and a correspond- ing decline in decentralized collections, over the period studied. Numerous fac- tors might dictate whether a particular institution would choose to build a stand- alone library. For example, the availabil- ity of funding for new construction and the relative strength and number of sci- ence programs on campus could be pos- sible influences, as might the values and vision of campus decision makers. Addi- tional exploration of such issues is beyond the scope of this survey and is left to fu- ture investigators. Characteristics of the types of units are summarized in table 3. Mean size, seat- ing capacity, collection size, and total staff- ing (FTEs, including hourly student em- ployees) are shown for each group. De- FIGURE3 physics mathematics geology engineering computer sci chemistry biology astronomy agriculture Disciplines of Departmental Libraries (179 branches in 51 institutions) 10 15 20 25 30 35 ARL Academic Science and Technology Libraries 151 TABLE4 De~artmental Libraries Discipline Size (sq. ft.) Seats Volumes Biology 8,286 64 65,283 Chemistry 5,268 72 43,970 Engineering 10,739 130 103,108 Geology 6,033 67 71,605 Mathematics 4,543 62 49 ,353 Physics 5,342 75 58,181 partmental collections are the smallest, as expected, and stand-alone collections are the largest. Departmental Libraries Overall, fifty-one institutions reported the presence of one or more departmental li- braries, and many questionnaires pro- vided considerable detail on the nature of those 179 branch· collections. The au- thor sorted data for branches by subject. The disciplines most frequently served by departmental libraries are displayed in figure 3. Shading on the bar graph differentiates single- and shared-discipline collections. Each subject includes any shared-disci- pline collections, so the total represented by summing all bars is greater than the number of branch libraries reporting. This was done to provide a picture of the rela- tive prevalence of branch libraries in the various science disciplines. For example, mathematics departments are most likely to be served by a departmental library, whereas engineering units are least likely to share their libraries with other disci- plines. Respondents reported numerous combinations, and some of these certainly reflect rational pairings (such as physics Active Serials Total Users Total Staff 744 923 4.18 296 551 3.66 1,115 2,803 6.81 1,046 375 5.31 624 670 4.33 502 563 4.43 and mathematics) whereas others may merely result because two departments share a building. The most frequently named disciplines of departmental libraries are, in rank or- der, mathematics, engineering and chem- istry, physics, biology, and geology. The .Definition Survey-ranked list is quite similar: mathematics, chemistry, engi- neering, physics, geology, and biology. Survey data are not likely to be capable of explaining this order, but some specu- lation is possible based on the nature of the disciplines' literatures and the infor- mation-seeking habits of the various sci- entists. Traditions within a discipline may be another possible influence. Shkolnik, and many of the authors whom he cites, identify and analyze influential factors that contribute to the presence of depart- mental libraries. Table 4 shows the characteristics of the most frequently . encountered discipline departmental libraries. Engineering librar- ies appear to be the largest, but geology collections have nearly as many active se- rial subscriptions, although they serve far fewer total users. ("Total users" include faculty, graduate and undergraduate stu- dents, and other affiliated academic and TABLE 5 Com~arisons: Decentralized versus Stand-Alone Libraries Library Type(#) Decentralized (6) Stand-alone (21) Total Materials $1 ,731,631 $1 ,531,580 Total Users 6,372 6,326 Cost/User $271.76 $242.11 Total Staff 32.11 29.01 154 College & Research Libraries March 1996 TABLE6 Expenditures: Departmental Libraries Discipline Monographs Serials Biology $61,213 $229,412 Chemistry $25,381 $307,018 Engineering $84,634 $307,881 Geology $29,017 $128,964 Mathematics $30,100 $136,464 Physics $34,458 $313,457 professional staff for whom a departmen- tal collection is their primary library.) Some of the variations observed may be ex- plained by the relative sizes of the subject literatures, particularly the journal litera- tures. Other differences may reflect aca- demic program strengths and sizes, as well as availability of funding on a particular campus. Financial Support: Materials Expenditures Survey respondents supplied information on materials expenditures for the library unit they were describing. This informa- tion was then broken down into catego- ries for monographs, serials, and elec- tronic resources. For some organizational types, few data were reported. For ex- ample, integrated collections, by defini- Institutions with decentralized collections spent, on average, $1,584,970 on serials; stand-alone libraries reported spending $1,247,031. tion, were not able to report expenditures for science materials. Data were available, and means are reported in table 5, for stand-alone and decentralized libraries. Data for subject departmental libraries are displayed in table 6.6 Values supplied by Canadian libraries were converted to U.S. dollars at the then-prevailing exchange rate of $1 US= $0.74 Canadian. Not all libraries participating in the survey col- Total % Serialsffotal Cost/User $305,758 0.81 $331 $333,940 0.92 $606 $394,958 0.80 $141 $158,763 0.82 $423 $166,559 0.84 $248 $356,429 0.90 $633 lected all the requested data, but enough did respond that differences are evident. The survey requested information on expenditures for electronic resources, but, in fact, only a small percentage of librar- ies had such separate data for the survey period. Commentary provided by several survey respondents indicated that some institutions were moving toward allocat- ing a separate line in the materials bud- get, whereas others reported funding electronic resources from different bud- get lines or through special allocations and grants. Funding and accounting for electronic resources appears to be an area where change is in process. Subsequent surveys may be able to document this de- velopment. Table 5 provides comparative data on decentralized versus stand-alone collec- tions. These data should be interpreted cautiously because very few decentral- ized systems were able to provide details on aggregate expenditures for all their de- partmental libraries. Some of the very largest decentralized systems with nu- merous branch libraries did not report. Consequently, the averages shown may be lower than would be the case if they had included all libraries. (Many more li- braries provided data for individual de- partmental collections, as summarized in table 6.) Despite the limitations noted above, the data support the conclusion that de- centralized collections are more costly to maintain than stand-alone libraries. In- stitutions with decentralized collections ARL Academic Science and Technology Libraries 155 spent, on average, $1,584,970 on serials; stand-alone libraries reported spending $1,247,031. Duplication of journal sub- scriptions is a feature of decentralized collections, but the larger expenditures also may be related, in part, to library size overall as measured by ARL rankings. Conversely, stand-alone libraries reported higher monographic budgets than the participating decentralized libraries. Data here are insufficient to justify any more than speculation in explaining this phe- nomenon. For example, the presence of a stand-alone science library may indicate the high visibility and overall relative importance of science programs on a par- ticular campus, which would be mani- fested in a high level of commitment to develop the science monographic collec- tion. In addition, fewer duplicate serial titles could easily be reflected in larger monographic budgets. The cost/user fig- ure displayed in table 5 is based only on reported materials expenditures (mono- graphs, serials, and electronic); staffing costs are not included. Total staffing for each type of system also is reported and would represent an additional cost fac- tor to be considered in comparisons of the two organizational types. Table 6 summarizes the data reported for subject departmental libraries and confirms the well-known fact that scien- tific journals comprise a major portion of the materials budget in each of these dis- ciplines. The data are displayed as a bar graph in figure 4. The cost/user data for departmental libraries are based only on materials costs, as in table 5. For the disciplines represented in fig- ure 4 and table 6, the percentage of the materials budget spent on serials ranges from a low of 0.80 for engineering to a high of 0.92 for chemistry. Librarians who develop and manage science collections probably will not be surprised by these figures. In fact, the figures correlate closely with use patterns for types of materials as summarized by Robin Devin in a paper first presented at the 1988 Charleston Conference.7 Devin identified more than fifty studies on the character- istics of literature use in various disci- plines and summarized data from many FIGURE4 $350 ,000 $300 ,000 $250 ,000 $200,000 $150 ,000 $100 ,000 $50 ,000 $0 Departmental Libraries: Materials Expenditures Biology Chemistry Engineering Geology ~ Monographs 81l Serials Mathematics Physics 156 College & Research Libraries citation studies to compile a table of the percentage of citations to serials in sub- ject literatures. The percentage of materi- als budgets spent on serials in ARL de- partmental libraries correlates positively with the relative importance of serials as measured by citations in published litera- ture (correlation coefficient = 0.82). The percentage of the budget spent on serials also can be expected to reflect serial costs in a field. To test this hypothesis, the au- thor consulted Library Journal's 1993 pe- riodical price index to obtain an average In the period that elapsed since Survey 3, dramatic changes took place, with a proliferation of CD- ROM products. cost per title for each science discipline. 8 There is a very strong positive correlation of 0.94 between the percentage of the bud- get spent on serials and the average cost for serials in that discipline. The cost per user computed for each discipline ranges from a low of $141 for engineering to a high of $633 for physics. The cost per user can be explained by a combination of factors including typical serials costs in a discipline, the number of serials needed to support research in that discipline, and the size of the primary user population. Engineering libraries frequently serve large undergraduate programs as well as sizable graduate pro- grams, contributing to a lower cost per user than in disciplines such as geology where the enrollments are typically much lower. In both chemistry and physics, to- tal user populations are comparable and of moderate size, and journal costs are among the highest of aJl disciplines in an academic library. Physics and chemistry collections have the highest cost per user, although a mitigating circumstance, es- pecially in chemistry, may be the many uncounted users for whom the library serves as an important secondary collec- tion. For example, the increasing impor- March 1996 tance of interdisciplinary research results in departmental collections being used by researchers in other fields than the pri- mary subject. A chemistry library also may be important to scientists working in materials science, pharmacy, medicine, and many other disciplines. These sec- ondary users are more difficult to iden- tify and count, but they certainly benefit from collections that were developed originally to serve a particular discipline. Survey respondents reported serials cancellations and additions. 9 Forty-two percent of the respondents (32 libraries) reported qmceling serials during the pe- riod covered by the survey for a total of 5,525 cancellations. The average number of cancellations per library was 172, rang- ing from a high of 500 to a low of four among institutions canceling. Forty-one percent of the respondents (31 libraries) reported adding subscriptions for a total of 1,788 new subscriptions entered. The average number of additions among those institutions reporting was 57, rang- ing from a high of 612 new titles at one institution to a low of only one. Many in- stitutions adding new subscriptions (14 out of 31) added fewer than twenty-five titles. Only twenty institutions were able to add titles without cutting. The net loss TABLE7 Top Ten CD-ROM Products Rank Product Number Held 1 MathSci 44 2 Compendex/ 37 Engineering Index 3/4 Biological Abstracts/ 35 BIOS IS GeoRef 35 5 Science Citation Index 34 6 MEDLINE 33 7 IN SPEC 32 8 AGRICOLA 25 9 Applied Science & 22 Technology Index 10 Computer Select 21 ARL Academic Science and Technology Libraries 157 of subscriptions in this population was 3,737 titles for the reporting period. As- signing a dollar value to this very diverse set of titles would be difficult to do pre- cisely with available data, but the 1993 periodical price index does allow a range to be estimated. The price index includes cost information by broad subject catego- ries. For Science Citation Index titles, the average cost per title during 1993 was $345.99 for domestic publications and $658.78 for foreign publications. 10 It is unlikely that all canceled subscriptions were exclusively either category, but us- ing these values as extremes allows esti- mation of an upper bound of $2,461,860 and a lower bound of $1,292,964 in losses to publishers of scientific and technical journals through ARL cancellations alone! Electronic Resources This survey represents the first effort to collect detailed information on electronic resources. Survey 3 had only requested a figure on CD-ROM expenditures and identification of any locally mounted da- tabases. In the period that elapsed since Survey 3, dramatic changes took place, with a proliferation of CD-ROM products. Some institutions reported extensive tape loading of purchased or leased databases. New sections were written for the ques- tionnaire to elicit details on some of these types of activities. The survey included questions about database availability, whether as CD-ROM-based products on stand-alone or networked workstations or as tape-loaded databases on an insti- tutional mainframe. 11 Table 7lists the most frequently held CD-ROM reference prod- ucts, whether used over a network or on a stand-alone workstation. Table 8 shows the databases most frequently reported as locally mounted on an institutional com- puter. Table 9 combines the databases that are available in respondents' libraries, whether as CD-ROMs or locally mounted files. MEDLINE leads the list by a consider- able amount and was available in seventy TABLES To{! Ten Locall_y Mounted Databases Rank Database Institutions 1 MEDLINE 37 2 Psyclnfo/PsycLit 34 3 Expanded Academic 32 Index/MAGS 4 Current Contents 27 5 ERIC 26 6 ABI/Inform 18 7 News/National 16 Newspaper Index 8/9 Business Index 15 Periodicals Abstracts 15 10 Applied Science & 14 Technology Index General Science Index 14 CARL Uncover 14 libraries. Comparing the tables reveals details such as the fact that MathSci ap- pears more frequently as a CD-ROM product than a local database, possibly re- flecting its targeted appeal to a well-de- fined audience, primarily mathemati- cians. Other discipline-based CD-ROM products include Compendex, GeoRef, and BIOSIS. Although not analyzed for Rank 1 2 3 4/5 617 8 9 10 TABLE9 Top Ten Databases (OPAC or CD-ROM) Database Institutions MEDLINE 70 Psyclnfo/PsycLit 53 MathSci 47 IN SPEC 45 Biological Abstracts/ 45 BIOS IS ERIC 44 Compendex/ 44 Engineering Index GeoRef 40 Applied Science & 36 Technology Index Science Citation Index 34 158 College & Research Libraries TABLElO Salaries* in ARL Academic Science Libraries (N=63) Position Assistant Director (n=13) Head (non-AUL) (n=36) Public Services (n=52) Technical Services (n=9) Entry Level (n=59) Mean $57 ,765 $47 ,694 $35 ,764 $38,391 $25 ,814 Range $41,052-$76,900 $33,000-$67,061 $25,600-$52,200 $31 ,212-$51 ,000 $21 ,500-$34,000 March 1996 change used by the ARLin compiling their annual salary survey for the comparable time period .B Not all respondents provided salaries in each cat- egory requested; thus, the number of libraries reporting in each rank is indicated in the table. *11 respondents reported in Canadian dollars; these were converted to The total library U.S. dollars at the rate of $1 US= $.80 Canadian. L..__ _____________________ ___J budgets for the respon- this article, survey data exist that can be used to test whether specialized CD-ROM products are more likely to be found in subject departmental libraries than in multidisciplinary collections. A number of the mainframe databases are multidisciplinary in nature, such as Current Contents, newspaper indexes, and encyclopedias. Others are likely to be of interdisciplinary interest (Psyclnfo, ERIC) or used by larger groups scattered across a campus (MEDLINE). Of course, the availability of a database as a CD-ROM or a tape-loadable product also is a de- termining factor, as is the relative cost of the two formats. The survey included a question about provision of access to the Internet, an area where activity seemed to be increasing dramatically as this questionnaire was written. By the time the surveys were re- turned, the change appeared complete: virtually all respondents stated that their libraries offer access to the Internet. Salaries The questionnaire also requested salary information on a separable page. Some respondents chose to include that with the rest of the questionnaire; others re- turned the salary page under separate cover or not at all. 12 The salaries shown in table 10 represent data submitted by sixty-three institutions. Eleven libraries reported in Canadian dollars; these were converted to U.S. dollars at the rate of ex- dents range from just under $4 million to more than $35 mil- lion, with an average of about $12.5 mil- lion. ARL statistics report salaries to be, on average, 51 percent of a total library budget; that also is the percentage re- ported by these respondents. Science librarians often speculate about whether salaries are better for sci- ence specialists than for those in other subject areas. ARL salary statistics tabu- late salary data for law and medical li- brarians separately, but provide no details on science/ engineering librarians or any other specialists broken down by subject area. ARL data for 1993 report an aver- age salary for a ''branch head" as $46,838, and this might be compared to salary data reported here for "head" of a science li- brary as $47,694. The difference suggests some support for the hypothesis that sci- ence subject specialists earn higher sala- ries, but would need to be tested more systematically with data gathered in a fashion that allows direct comparisons across identically defined categories. Directions for Further Research The author did not analyze all the data collected for Survey 4 in this article. Ad- ditional analyses may result in further publication, and at least one project is under way describing staffing patterns in science and technology libraries. Marilyn VonSeggern of Washington State Univer- sity and Donna Cromer of the University of New Mexico are analyzing this portion ARL Academic Science and Technology Libraries 159 of the data and are preparing a separate article that will be submitted for publica- tion upon completion. For that reason, this analysis reports only very general staffing data. The Comparison Committee, under its present charge, is directed to collect sta- tistical data of this type on an ongoing basis. Future surveys that include simi- lar questions to those that were used here should continue to build a longitudinal collection of statistics supporting identi- fication of trends and documenting sig- nificant changes. Given the many devel- opments in research libraries, it is likely that Survey 5 will add new questions, es- pecially in areas that relate to electronic resources and applications of information technology. For example, electronic jour- nals are increasing in number and mode of distribution, especially in science li- braries. Documenting their use and im- pact on library operations might be an ad- ditional area of exploration for the next survey. As libraries were contacted for this survey, there was anecdotal evidence that some were reorganizing and flattening administrative hierarchies or trying new management techniques such as total qual- ity management (TQM) or reen~eering. Exploring how science libraries are partici- pating in these developments might pro- vide a focus for a future investigation. In addition, there will no doubt be other ar- eas, as yet unknown, that will result in the collection of new statistics descriptive of the diversity found in science and technol- ogy libraries. Note: The author benefited from significant input from committee members who assisted with data analysis: Andrea Duda, University of California-Santa Barbara; Bonnie Osif, Pennsylvania State University; Nancy Simons, University of Arizona; and Terry Wittig, Carnegie Mellon University. All the committee members contacted a portion of the survey population, answered questions from respondents, and provided follow-up to en- sure maximum return of questionnaires. This is a joint project by a true "working" commit- tee. The Comparison Committee expresses ap- preciation to all those ARL librarians who took time from busy schedules to provide data and thoughtful responses to this survey. Without their input,librarians would know much less about science and technology libraries in ARL. Notes 1. Survey 1 is described in Emerson Hilker, "Statistical Data for Stand-Alone Science/Engi- neering Libraries in the United States and Canada 1984/1985," Science & Technology Libraries 8 (fall 1987): 89-127 (contains tabulated data from all respondents). Survey 2 is described in Emerson Hilker, "Survey of Academic Science/Technology Libraries," College & Research Librar- ies News 88 (June 1988): 375-76 (a full data set was distributed to respondents in tabulated form). The Definition Survey is described in Elaine Brekke, Kimberly Douglas, and Elizabeth Roberts, "Academic Science and Technology Libraries: Facilities and Administration," Science & Technol- ogy Libraries 11, no. 3 (June 1988): 107-16. The Historical Survey is described in Elizabeth P. Roberts, Elaine Brekke, and Kimberly Douglas, "Physical Structure and Administration of Sci- ence and Technology Libraries: An Historical Survey," Science & Technology Libraries 11, no. 3 (spring 1991): 91 _:105. 2. Carole S. Armstrong, "Historical Background of the STS Committee on Comparison of Science and Technology Libraries," Report to the STS Committee on Comparison of Science & Tech- nology Libraries, ALA Annual Meeting, June 23, 1990. 3. Kimberly Douglas, private communication. 4. Leon Shkolnik, "The Continuing Debate over Academic Branch Libraries," College & Re- search Libraries 52, no. 4 (July 1991): 343-51. 5. ARL Statistics., 1992- 1993 (Washington, D.C.: ARL, 1994). (ARL ranks for the period corre- sponding most closely to the survey data were utilized.) See also URL = http: I I www.lib.virginia.edu/arlstats/ where data are available graphically and as files for ftp. 6. Bonnie Osif, Pennsylvania State University, carried out the analysis of expenditures for collections. 160 College & Research Libraries March 1996 7. Robin B. Devin, "Who's Using What?" Library Acquisitions: Practice & Theory 13 (1989): 167-70. . 8. Lee Ketcham and Kathleen Born, "The Art of Projecting: The Cost of Keeping Periodicals," Library Journal118 (Apr. 15, 1993): 42-48. 9. Terry Wittig, Carnegie Mellon University, compiled and analyzed responses on serials can- cellations. 10. Ketcham and Born, "The Art of Projecting," 45. 11. Andrea Duda, University of California-Santa Barbara, compiled data on CD-ROM and mainframe tape-loaded databases. 12. Nancy Simons, University of Arizona, analyzed the salary data. 13. ARL Annual Salary Survey, 1993 (Washington, D.C.: ARL, 1994). Available as an ERIC document: ED 359 955. 1> National Conference April11-14, 1997 Keep up with the latest on ACRL's conference plans! Point your web browser to: http://www.ala.org/acrl.html Association of College & Research Libraries • acrl@ala.org • (312) 280-2519