Digital Libraries and the Practices of Scholarly Communication

TABLE OF CONTENTS

1. Project Summary
2. Introduction to the Project
3. Research Methods
4. Digital Materials and Disciplinary Practices
5. The Supply of Digital Library Services on University Campuses
6. Material Proficiency and Disciplinary Practices for Obtaining Paper and Digital Documents
7. Electronic Journals in Systems of Scholarly Communication
8. Digital Shift or Digital Drift?:Organizational Processes and the Transition to Electronic Publishing and Digital Libraries
9. Conclusions
10. Bibliography/References
Appendix I -- Faculty Interview Schedule
Appendix II -- Notes on Research Methods
Appendix III -- A Brief History of the Use of Gopher and WWW in 1993-1995.
Appendix IV -- Acknowledgements

---

1. SUMMARY

This document reports an empirical study of the ways that faculty and graduate students in several fields in 8 major US research universities use Digital Library (DL) services in the course of their routine work. In the Winter and Spring of 1995 we carried out field studies and interviewed 124 faculty and graduate students in four disciplines. In addition, we interviewed over two dozen academic administrators about their investments in campus networking, and paper and digital media (See Sectioin 3). This document reports some early analyses of our data. At this point we draw the following conclusions:

a. There is substantial variations in the use of digital libraries across disciplines (Section 4). Literary theorists, sociologists and molecular biologists make significant use of indexes and abstracting services that were mediated by publishers and provided by their campus libraries. In contrast, computer scientists do not use such indexing and abstracting services; they were also the primary researchers who extracted electronic texts from Internet-mediated sources. (Section 4).

b. Universities vary substantially in the extent to which they provide computer networking and Internet access for the faculty. Some major universities have spent considerable money and effort to wire all campus offices while others have (so far) emphasized the laboratory sciences. Network access alone is not a good predictor of DL use, since productive scholars also want (network) access to discipline-specific indexes, abstracts or text corpuses. (Section 5).

c. The "Principle of Proficiency" influences professional practices for using documentary materials. This principle leads many scholars to organize their own private collections of materials, and also leads the publicly accessible data on the WWW to be of little value to many sociologists and literary theorists (Section 6)

d. The Internet might be of greater interest to scholars if electronic journals become a major medium of scholarly communication. We have found that too few scholars perceive electronic journals to be legitimate means of communication for them to become major media soon (Section 7).

e. While this appears to be a period of major structural transformation in a shift from print media to digital media, the internal structure of universities and budgetary limits lead to local processes that are better characterized as organizational drift to digital media. (Section 8).

2. INTRODUCTION TO DIGITAL LIBRARIES AND SCHOLARLY COMMUNICATION

There is an enormous decentralized national investment supporting the development of Digital Library (DL) services for colleges and universities. Much of the excitement focuses upon the computerization of new corpuses, the expansion of emerging formats (such as electronic journals), and the development of new electronic distribution systems (e.g., WWW). But it is not automatically the case that faculty and students will effectively use sophisticated and intellectually rich DL services simply because they are available on their campuses.

This document reports an empirical study of the ways that faculty and graduate students in several fields in research universities use DL services in the course of their routine work. We characterized DL services in broadly to include resources that are anchored in the traditions of library automation (ie., Online Public Access Catalogs, digital abstracts, full text databases) and digital materials that are available through the Internet (ie., certain electronic journals, conference proceedings, technical reports).

The project was funded from October 1, 1994 through September 30, 1996. The main sources of data derive from field studies in eight research universities that we carried out in the Winter and Spring of 1995. Some results of this project have been published as conference papers, journal articles, and a Ph.D. dissertation (Appendix I). In the body of this report we explain the project's background, our empirical research methods, and some of our major findings.

2.1 PROJECT BACKGROUND

We conceptualized this project in the Fall of 1993 when there was little reliable information the working conditions and institutional and organizational practices that make DLs most usable and used by faculty, academic staff, and students. At that time there was tremendous excitement by information and computer scientists, some librarians, some scholars, and some politicians about the promise of DLs. These professionals -- as well as some foundation officials and some professionals in the publishing industry were curious and enthusiastic about the potentials of electronic publishing for speeding up and broadening access to materials in various media -- including print, data, pictures, and software.

It will be helpful to label this diverse group of professionals who have explored and also helped stimulate interest in the potentials of digital libraries. We use the term "DL movement" to refer to the professional movement to create digital libraries and popularize their value.

Sociologists have used the concept movement to refer to many different kinds of collective action. The most common term found in this literature is social movement, often used in a generic way to refer to movements in general. But sociologists also have written about professional movements, artistic movements, and scientific movements. What analyses of these movements share is a focus on the rise of organized, insurgent action to displace or overcome the status quo and establish a new way of life. Computerization movements are no different. Large-scale computerization projects are typically accompanied by political struggle and turmoil as the established structure is threatened and powerful actors fear being displaced. Historically, those who advocate radically new forms of computerization find themselves in the role of challengers of the status quo." (From, Iacono and Kling, 1996).

Professionals who organize "computerization movements" play key roles in stimulating interest in certain computer applications or modes of computer use -- such as personal computing, artificial intelligence, and instructional computing (see Kling and Iacono, 1988; Iacono and Kling, 1996).

DL movement participants had written enthusiastically about the promise of "electronic libraries" (Buckland, 1992) and electronic media to support scholarly communication (Okerson, 1991, 1992) for over 20 years, in a literature that is traced to a mid-1960's report by J.C.R. Licklider and Vanevar Bush's 1948 Atlantic Monthly article "As We May Think." Drabenscott and Burman's (1993) Analytical Review of the Library of the Future that was produced with support from the Council on Library Resources reflects the enthusiasm in this literature. Their report is a well organized compendium of generally optimistic quotations from earlier studies, but it is not a probing analytical examination of their repercussions for libraries or people who use them.

Through 1993, the majority of DL movement participants were information and library scientists (in diverse institutional locations), and their primary audiences were other movement participants, librarians, and some scholars. The DL movement was relatively small and specialized. The visions of DLs gained substantially greater public visibility as a byproduct of the Clinton-Gore administration's treating DLs as a major "application" of an enhanced National Information Infrastructure (White House, 1993).

The DL movement grew and began to jell in this period and focussed on the promise of DLs helping anyone access materials -- anytime and anywhere. For example, Vice President Al Gore gave several well-publicized talks in which he highlighted the possibility of a little girl in a small Tennessee town accessing the Library of Congress over the Internet. This image had immense symbolic power, even if though was problematic in many ways (ie., it undermined local booksellers and libraries, and focussed on a non-circulating library whose holdings required travel to Washington DC.).

The Internet was being opened for public access in a way that was unprecedented. Organizations, mostly academic, governmental and non-profit, were experimenting with Gopher services as ways to electronically publish various documents and make them accessible to diverse audiences. For example, the Clinton-Gore administration's Information Infrastructure Task Force reports were available on "the White House gopher." Universities experimented by putting materials such as course bulletins, faculty rosters and class schedules on campus Gophers. (These electronic documents helped us locate appropriate faculty informants when we were conducting this study). Some electronic journals, such as PostModern Culture and the Electronic Journal of Virtual Culture shifted to Gopher services (from LISTSERVs) for archiving and distribution. And some non-profit organizations began to issue their reports via Gophers in addition to printed forms.

In the six month period between January and June, 1994 that our proposal was under review at the US Department of Education, the DL world was punctuated by some major events. One set of events was a multimillion dollar "Digital Library Initiative" funded by the National Science Foundation, ARPA and NASA to grant six "awards of up to $1,200,000 a year for up to four years ... to conduct programs of research and to develop and test elements of a digital library on a significant scale in a distributed environment." This research initiative was primarily of interest to academic participants in the DL movement. But it served to legitimate the label, "Digital Library" (over competing labels, such as "electronic library") and give these substantial technologically-anchored DL projects notable visibility within a broader segment of the information and computer science research communities. Further, The DL Initiative helped to give a more substantial information and computer science research base to the DL movement and to amplify its intellectual momentum through a new conference series -- the annual Digital Library Conferences that are organized by the Association for Computing Machinery. The DL initiative also represented a major milestone in creating highly visible and technologically innovative DL projects that were not anchored in traditional libraries, traditional publishing houses, or even traditional materials, such as books and articles.

A second set of key DL events in Winter and Spring of 1994 cluster around the emergence of the World Wide Web as a major medium for documentary communication on the Internet. While the WWW was developed in 1989-90 at CERN as a medium for communication amongst international teams of physics research teams, it became a popular medium of communication by mid-1996. It is difficult to identify key markers of the rise of the WWW (and the demise of gopher) in the early 1990s. It's important to note that the earliest web browsers, such as Lynx, were text based and presented a viewing screen that was little more interesting than Gopher. (We briefly sketch some indicators of the shift from Gopher to WWW in 1993-1995 in Appendix B). In 1994 the WWW was actively popularized by the National Center for Supercomputing Activities, a developer of one of the first graphical browsers for the WWW. One of NCSA's promotional devices cleverly used WWW linking to publish a monthly list of "what's new" on the WWW. People using the WWW could see a burgeoning list of WWW sites, and quickly visit them by clicking on their links. This monthly list helped NCSA illustrate the diversity of possible WWW content -- including university labs, software support, commercial ventures, community libraries, and so on.

It is helpful to characterize the shifts in perceptions of DL services in the period in which we framed and conducted this study. This was a period of rapid change in the technological infrastructure that could support DL services. During our study, one molecular biologists asked us directly "How can you be sure that your results won't be out of date in six weeks?" Another colleague cautioned us:

"Both the available materials and the usability of the forms in which they are delivered will be changing drastically over the next five to ten years .... "library services" as viewed through today's libraries will not adequately capture the many "digital information services" being used in a less formal way, such as user-provided WWW pages, which will be integrated into the digital library as it develops."

One challenge was to develop a study whose results were sufficiently robust to be of interest to scholars, DL movement participants, librarians, and academic administrators during the next decade. This study pays much less attention to static behaviors (ie., that X% of faculty prefer to search their library's periodical shelf while Y% prefer to start with Internet sources) than to the social practices and organizational arrangements that tilt faculty towards periodical shelves and various DL services.

Our colleague understood our approach when he noted in a subsequent message:

"the important generalizations are not about the technology directions, but about models of human and organizational behavior. In fact, the studies can be extremely valuable to help guide the direction of the technology, when they aren't misinterpreted as blanket assessments of its value."

2.2 ACADEMICS' USE OF DL SERVICES

In order to have an empirically anchored study, we wanted to study the way that scholars and students in 1995 used DL services. The wide range of services that form today's academic DLs come from two related, but different professional worlds.

One set of facilities emerges from the world of publishers and libraries, and places skilled professionals in many intermediary roles to select valuable corpuses of materials, to index and organize them, and to help people find documents through them. Online Public Access Catalogs, abstracting services, and citation systems illustrate these facilities. Some of them have been in use for over 20 years, and these facilities of these sorts are widely used in academic libraries. While people who use these families of DL resources may sometimes be satisfied with a set of citations or abstracts, they usually use these computerized services to locate a paper book or article, which they often physically retrieve from a local library.

In addition, some university (and corporate) libraries subscribe to services that provide the complete text of articles from selected magazines, newspapers and journals. Companies such as Lexis/Nexis, KRI/Dialog and ABI/Inform license such services, but at prices that seem most affordable by relatively rich organizations -- such as research universities rather than comprehensive universities.

Another set of newer DL services that are more directly associated with the Internet come from other research worlds, especially computer science. These DL services, include (a) distribution systems such as WAIS and the World Wide Web; (b) electronic newsletters, such as those distributed through LISTSERVs; and (c) electronic journals. Universities usually place faculty, staff and students who use these Internet-based services in the roles of serving as their own librarians and computing consultants. They have to locate relevant sources (e.g., LISTSERVs, electronic journals, WWW sites), and find their own ways to download and perhaps print relevant files. Part of the reason for this shift in roles from personal service to self-service is that faculty and students may access some networked services from their offices and homes and on 7-day 24-hour schedules. Those who do, may seek access to DL resources at places and times where one does not traditionally find reference librarians and similar skilled help.

The effective value of DL services depends upon their practical usability by specific academic communities in their day to day working settings. Faculty and students need to learn about their contents, learn to use them, and have access to some complimentary computing resources (such as disk space and printers) to effectively utilize their output. University departments (or schools) vary in the extent to which they support DL services. Some administrators expect their faculty to rely primarily upon central library services or their own skills with DLs. Other administrators have developed departmental or school-specific DL support groups for their faculty and students. The local DL support groups also vary in the range of services they provide. At the extremes, some provide only core technical services while others provide consulting roles akin to research and reference librarians for Internet-based resources.

2.4 FROM DL SUPPLY TO DOCUMENTARY CONSUMPTION

We are interested in the possible value of DLs in supporting scholarship and education. The study reported here examines the organizational and working conditions under which university faculty and students use of a variety of DL services in their routine work, and the institutional and organizational practices that make them most useful and usable.

The primary researchers on this project were both keenly interested in the ways that faculty and students could effectively use DL services in support of research, teaching, study, and other activities. We used certain DL services in our own research, teaching and study. Professor Kling -- then on the faculty of the University of California, Irvine (UCI) -- began actively using the on-line abstracting services licensed by the University of California from his home and office in support of his research and teaching in the early 1990s. He began using Gopher in the Spring of 1993. He found these services to be of substantial value in his own work, and discussed them informally with his colleagues (in various fields) and PhD students. Through these informal conversations he learned that few UCI faculty were aware of the diverse text databases available through the University of California Melvyl system, or used them in support of their research or teaching. Dr. Lisa Covi, then a Ph.D. student at UCI, was also keenly interested in the use of DL services. She worked with library automation and the early versions of the Internet in previous staff positions at the Columbia University Computing Center and at EDUCOM.

This study was carried out by a research team that was enthusiastic about the potential value of DLs, but who also were aware of various practical limitations and complexities and the fact that they could be "under-utilized." Supply-side accounts of DLs don't help us to understand the actual ways that people care to use or ignore DLs. The NSF/ARPA/NASA DL Initiative and the earliest demonstrations of Gopher and WWW use were "supply side" activities. Without digital documents and ways to access them, digital libraries would be purely fictional. However, it is possible to develop digital corpuses that people do not know of, do not value, find cumbersome, or simply do not use.

We conceived of this study to complement the "supply-side" visions that were common in the DL literature and in the discussion of various professionals about the promise and limits of the Internet. Most supply-side characterizations of DL use treat the acquisition and use of electronic materials as relatively unproblematic. In the simplest scenarios, a person -- scholar, student or teacher -- has technical access to an appropriate computer network. The person is usually searching for a specific document or kind of document and is portrayed as finding it rather easily. That is, suitable documents are in the electronic corpus and are locatable via electronic searches that the person is skilled in doing.

From our studies of computer use in various institutional settings (e.g., Kling and Jewett, 1994), and our observations of scholars and students we believed that consumption of electronic documents was often more problematic. The person might not know that some items that they want are available on-line, might not have technical access to a suitable computer system, might not have the skills to search effectively to find the items, and might not be able to effectively read the items even when they find them.

2.5 SOME KEY QUESTIONS ABOUT ACADEMICS' DL USE:

To effectively deploy DL services, we sought reliable systematic answers to some questions where there was anecdotal evidence and ungrounded speculation

1. How accurately do faculty and students perceive the contents and formats of information that these services provide;

2. How much do faculty and students actually utilize these resources, and how do they fit their informational preferences;

3. Under what conditions do faculty and students prefer electronic information to be available in specific forms electronic for reformatting or restructure; paper for mobility or annotation;

4. To what extent do faculty and students use services where they have assistance from skilled help -- such as reference librarians or colleagues;

5. How do these patterns of preferences and usage vary with the different disciplinary traditions and working conditions;

We did not begin this study with a blank slate. We examined a body of research about the conditions under which people will use computerized information and network resources and it suggested some important hypotheses that we extended to DLs (e.g.,, Kling and Jewett, 1994):

(1) While usage patterns will vary between individuals, and between groups, some structural conditions will shape the effective usability of DL resources.

(2) Disciplinary patterns will influence the extent to which important scholarly materials are likely to appear in electronic form, and the extent to which people find incentives to conduct one-shot searches or continuous scans.

(3) People's abilities to effectively use DL resources can depend upon their access to complementary computing resources (e.g., disk space, printers, local data management software).

(4) Access to these complementary computing resources will depend upon the ways that campuses allocate computing resources to faculty and students and upon the amount of outside funding a person's academic unit.

(5) Faculty and students will often need human assistance to effectively use DL resources, especially for newer services and one-shot searches.

These hypotheses guided our study design. Our project built on these hypotheses through an empirical study of the ways that faculty and students use a variety of DL resources to support their scholarship, teaching, and study. Among the nations' college and university students, graduate students are most likely to use libraries systematically and intensively and they were the student informants in this study.

To keep the project's costs within feasible bounds, the research team proposed a multilayered study design. We examined the nature of DL services and faculty usage through short field studies at eight diverse universities. These were chosen from among those campuses that had reputations for providing a relatively rich array of DL resources to their own faculty and students.

At each university, we interviewed faculty and graduate students in four disciplines: a lab science (molecular biology); an artifact-based discipline (computer science); a social science (sociology); and a humanities discipline (literary theory). These fields were chosen because of the different types of materials different disciplines use for their work. They also vary in their relative wealth, and the extent to which faculty and graduate student offices (or laboratories) might be linked to campus computer networks.

We also proposed more intensive interview and survey studies of the faculty and graduate students on one university campus (UCI) which had a significant array of DL resources through the universities libraries and through campus-wide Internet connections. This portion of the study was dropped after we learned hat we were funded at less than 50% of our proposed budget.

We will discuss our research methods in substantial detail in the next section.

2.6 FROM DIGITAL LIBRARY USE TO SCHOLARLY COMMUNICATION

In proposing this study, and in the introduction to this point, we have focussed on the extent to which faculty and students use DL services. We learned that our proposal was approved for funding in July 1994 and refined our conceptualization and research methods for this study.

Our major conceptual shift -- to focus on the use of DL services as part of a process of scholarly communication -- came as a byproduct of our efforts to identify meaningful social contexts for faculty and student use of DL services. In our early conceptualizations, we focussed on "local social contexts" -- such as the department and campus since both of these influenced the kinds of DL services that would be available in campus libraries, via campus computer networks, and available in faculty offices. We focussed on disciplines insofar as DLs would be likely to provide relevant materials. For example, we expected that literary scholars -- who rely upon books as sources of texts for analysis and for scholarly commentary -- would find DLs much less useful than would scholars who rely upon a relatively recent corpus of research reports and journal articles.

This conceptualization of DL use based on contexts that are "close at hand" to the person is congruent with much of the research literature on "information seeking behavior" and on the use of DL services. For example, the final report of a major project (TULIP) that enabled nine major research universities to provide the full-text 43 Elsevier science and engineering journals to their faculty, staff, and students (Borghuis, et. al. 1996) emphasizes the role of proximate context (ie., campus computing resources) in explaining relatively low levels of faculty usage. While the report is careful to characterize campus network and computing resources in some detail, it mentions the roles of faculty scholarly worlds in one sentence:

"The searching behavior of faculty is more focussed (than graduate students), because the network of personal contacts, etc. indicates the relevant sources" Borghuis, et. al. 1996:71)

The scholarly worlds of faculty -- and communication within them -- are more than a network of personal contacts. Especially at major research universities, such as those that participated in the TULIP project, faculty careers hinge on communicating within scholarly worlds so that one is seen as an important contributor.

At these universities, which included M.I.T., and the Universities of California, Michigan, Tennessee and Washington, faculty must do more than "simply publish" for various career rewards. They must publish work of sufficient quality (and quantity) that they are viewed as strong scholars (at their career stage) by their peers. These rewards include advancements in rank, being tenured, salary increases, and access to scarce campus research resources such as space, graduate assistants, research funds, and periodic teaching reductions. Some of these career rewards, such as advancements in rank to Associate and Full Professor, and awarding tenure can entail stringent reviews that include written evaluations by one's peers at other universities. Other rewards, such as small departmental research and travel grants may be based on short proposals and less formal internal reviews.

In any of these cases, the faculty member's recent publishing record is scrutinized for signs of quality. People who want to be rewarded well have strong incentives to publish in the major journals or with the major presses in their fields. Since the major journal editorial boards and presses turn to trusted scholars who publish through them, scholars are incentivized to read the major journals in their fields to know the kinds of scholarship and research results that is seen as leading edge.

This short sketch condenses a myriad of variations into a compact formula. Fields differ in the degree of consensus among their participants about "the best journals" and "the best publishing houses" (for book-oriented fields). At the extremes, every molecular biologist whom we interviewed reported regularly reading Science and Nature. In contrast, the computer scientists did not have substantial consensus in their choices of journals, because the field is split into numerous specialties with its own (overlapping) constellation of journals and conferences.

Faculty at the major research universities focus on the publishing outlets where their peers are most likely to read their work. This simple rule also has exceptions, such as the scholar hungry for rapid publication who tries to publish in a second tier outlet where her work may be more rapidly accepted. In short, scholarly communities (sometimes called "invisible colleges" (Crane, 1972)) can be viewed as communities of authors, editors, reviewers, and readers who treat specific journals, conferences, and publishing houses as major channels of communication.

Invisible colleges play strong roles in shaping formal (publishing) communication channels for research-active faculty in research universities. In our study, we found few exceptions to this rule. The few exceptions were senior established faculty who were seeking new audiences beyond their invisible colleges. In these cases (primarily in the humanities), the faculty were most enthusiastic about publishing in outlets that helped them reach a new group of readers.

There certainly are precedents for examining academic library services in a context of scholarly communication (Cummings, et. al., 1992). The major issue in identifying contextual dimensions is to identify those that will actually influence human behavior. In Section 6 we will examine the ways that the pursuit of mastery and effective communication in a discipline influences the ways that faculty and graduate students use DL services.

We informally renamed this project as the Scholarly Communication and Information Technology (SCIT) project. In the study we paid attention to both "local departmental campus contexts" in structuring access to resources, and to the ways that faculty and graduate students sought specific resources as part of their scholarly communications.

3. RESEARCH METHODS

In order to study how researchers can more effectively use digital libraries in their routine work, we designed a study of both material use practices and material provision arrangements. The study of material use practices examined how faculty and advanced doctoral students make use of both paper and electronic materials in their routine research practices. The study of material provision arrangements examined sources of reported research materials, opportunities for obtaining research materials and provision policies for departmental, university-wide and regional collections.

We chose a research approach that elicited crucial details about routine work practices and local arrangements: an organizational field study with comparative cases. Because scholars are at the intersection of organizations (university departments) and transorganizational social systems (disciplines, invisible colleges), we needed to understand both components of their work life. Our study design provided comparisons between universities, work practices in different disciplines and research material provision in different departments.

We designed this project to collect our data in a series week-long field studies at eight diverse universities in four fields of research that used different modes of inquiry. We considered other methods such as surveys and telephone interviews, but face to face interviews and direct observation were the only options to collect this kind of data for several reasons. First, our informants were elites in the university research community. Their time was important to them so it was necessary to engage their interest as well as cooperation. Second, the phenomena we were studying was extremely complex. It would have been impossible to devise a questionnaire about the mix of materials they use or provide without interviewing them to understand what to ask them about. Third critical details would have been hard to see from afar. For instance, several times, at the end of the interview, we would learn something about the researcher, the university, or the informant*s professional life that altered our interpretation of their account. Fourth, most researchers and some providers did not completely understand the phenomenon that we were studying.

Many researchers were not familiar with the term "digital libraries" and some did not believe that their behavior was relevant to this study. Therefore face-to-face interviews helped to establish rapport, explain that we were interested in their use of paper materials as well as digital media, and engage the informants by discussing their scholarly work (a topic with which they were very familiar). Although faculty researchers were usually more aware of their research endeavors then the nuances of their material use practices, they could still report useful data in the context of projects or problems which consumed their worklife. Finally, this approach was necessary because we were not previously familiar with critical aspects of working in the chosen subspecialties and provision of particular types of materials. In the course of conducting interviews, we could, for instance, ask questions about the terminology researchers and providers were using to describe their work and verify initial patterns by actually asking other researchers about disciplinary practice in subsequent interviews.

We spent one week each per site and cross-checked different accounts between providers and scholars through interviews and exploration of the site. Most sites graciously provided us office space or laboratory access on their campuses from which we could coordinate our study during the week. Each day, we conducted four to seven interviews which lasted between 45 minutes to one hour. Interviews followed a semi-structured format customized for providers, faculty scholars or graduate students. During the course of the site visit we also collected use and provision documentation, and took notes on direct observations of use and provision behavior.

At each university, three faculty researchers were interviewed in each of four fields: a lab science (molecular biology); an artifact-based discipline (computer science); a social science (sociology); and a humanities discipline (literary theory). These fields were chosen because of the different types of materials different disciplines use for their work. The interviews focused on one or more exemplary research projects. When doctoral students were available in each field, they were also interviewed for this study. Doctoral students were included based on the resource discrimination rationale: doctoral students usually have access to different types of resources and sometimes have different work patterns than their advisors. In total, this study includes interviews with 96 faculty and 28 doctoral student researchers. This design allowed for cross-case comparisons between universities, between disciplines broadly, between departments narrowly and between multiple, varied resource levels.

We also conducted interviews with 23 key digital library infrastructure providers (several at each campus) such as a university librarian, a director of academic computing, a faculty senate library chair and other senior academic administrators such as a provost. These interviews focused on data about campus-wide patterns of material use, patterns across disciplines, and university investments in library and computer support. During visits to each campus, tours of library and computing facilities were documented to examine first-hand the resources and services reported in the interviews.

3.1 SELECTING EIGHT RESEARCH UNIVERSITIES AS SITES

Our methods to discover how resource arrangements influence researchers' abilities to effectively use digital libraries was heavily guided by the quality of data collected from each site. Therefore, site selection played a critical role in the design of this study. The approach to site selection was based on a standard social science strategy for selecting cases for comparative study (Danziger et al., 1982). We identified three characteristics of each university that could substantially influence the ability of faculty to obtain or use paper and electronic materials on their campuses.

These three characteristics helped us to select sites for the study: investments in library resources, university governance (public/private), and the degree of library decentralization. We treated these characteristics as dictomous, and selected four universities that would be relatively high and four taht would be relatively low in each characteristic (ie., relatively rich or relatively poor in library reources).

a. Library investments (dollars per faculty member). We wanted to have a mix of richer and poorer universities in our study. We believed that a study who sample included eight of, say, the top 20 richest research universities, would not readily generalize to a wider mix of institutions.

We selected library investments per faculty member to indicate the existing resources for local library collections of all formats. This is a somewhat problematic measure for a study of DLs, since it does not directly tap DL support. Unfortunately, we could not find any systematic quantitative sources of data about campus investments in DLs, or even DL infrastructure, such as the fraction of faculty offices connected to the campus computer networks.

Measures of campus library investments are available through the United States Department of Education's Integrated Postsecondary Education Data System (IPEDS). Data are also available for the 108 members of the Association of Research Libraries (ARL) (http://arl.cni.org). In 1991, the nation's research universities spent between $3,200 to $37,500 per faculty member per year. Even the more elite ARL libraries made widely varied library investments (ranging between $5,000 and $37,100 per faculty member in 1994, with a median expenditure around $10,000 per faculty member). Universities such as Oklahoma State, Temple, Howard, Case Western and the University of Florida exemplify ARL campuses that made relatively low investments in their campus libraries (less than $7000/faculty member in 1994). In contrast, UC Berkeley, Brown, Stanford, Princeton and Harvard invested over $20,000 per faculty member in their campus libraries in 1994.

In selecting sites, we examined other per capitized measures (ie., library investments per graduate student and for all students) to insure that we our sites were at the relatively high end or low end on each of these scales. We also considered the role of academic computing infrastructure in supporting alternative digital library infrastructure resources and services.

b. University governance (public or private): We suspected that private universities would be more likely than public universities to expect each academic unit to "pay its own way." We expected that departments in "grants poor" fields, such as humanities and social sciences would disproportionately weaker local computing support in private universities than in public universities.

c. Degree of centralization of library facilities (the number of branch libraries): We suspected that faculty (and graduate students) would use libraries if they were closer to their offices, and would have greater incentives to value DL services if paper materials were less proximate.

To chose eight sites, we gathered data on 76 U.S. research universities including the winners of grants for the digital library initiative, institutions that expressed interest in participating in our study, institutions who had high-profiles in the development of electronic collections and institutions that might offer useful comparisons. (Our data sources were primarily the 1990-91 data from United States Department of Education's Integrated Postsecondary Education Data System, and self-reported figures for data missing from those studies.)

This group of 76 universities included a mix of public and private universities of the 179 universities that the Carnegie Foundation for the Advancement of Teaching characterizes as Research I, Research II, and Doctoral I (Carnegie Foundation for the Advancement of Teaching, 1994). These three categories are based on the total amount of annual research funds spent by a university, the number of doctoral degrees awarded annually and the diversity of its productive doctoral programs.

When we tried to learn about active research programs in each of our four disciplines (see below), we found that the Doctoral I universities might not have active Ph.D. programs in each of the four that we selected. We subsequently had to narrow our sample to the 67 of the 126 universities that fall into Carnegie Research I and II classification (institutions awarding more than 50 doctorates annually and with federal funding over 15.5 million).

Our sample (Table #3.1) included four public universities and four private universities with annual library investments (per faculty member). Each university is assigned a pseudonym to help protect the anonymity of our informants. If we identify an informant as a sociologist who studies social networks at say, the University of Toronto, he would be readily identifiable (in this case, professor Barry Wellman). Since all of our sites are U.S. universities, Toronto is not one of our eight sites. The University Librarians, provosts, and Directors of Academic Computing would also be quickly identifiable if we disclosed the names of these universities.

When we quote our informants, we use the abbreviations in Table 3.1 and a simple code to indicate a specific informant. For example, our computer science informants at Harbor University will be denoted by codes HUCS1, HUCS2 ...; Harbor's literary theorists are denoted by the codes HULT1, HULT2 ... and so on for each university and discipline. Harbor's University's Librarian is coded as HUUL, and similarly for other universities in our study.

Table 3.1 indicates broad ranges for the relevant parameters (ie., investments, number of library branches), rather than the specific ranges for these eight universities. These eight universities also varied in the number of branch libraries they operated. Our final choices were also influenced by budgetary and practical considerations: we selected some university pairs in the same metropolitan area to reduce airfare and also to enable us to have a two week window for interviews instead of the one week that we allowed for each university. This choice had the unanticipated serendipitous effect of our coming to appreciate the nature of metropolitan academic library resource ecologies.
 

	PRIVATE		PUBLIC
	Centralized Libraries	Decentralized Libraries  (4-76 branches)	Centralized Libraries	Decentralized Libraries (4-76 branches)
High Library $ ($10K-$25K library $ annually per faculty member)	Harbor University (HU)	Mountaintop University (MU)	Forest State University (FSU)	Branch State University (BSU)
Lower Library $ ($3K-$10K library $ annually per faculty member)	Technical University (TU)	Revere University (RU)	Diamond State University (DSU)	River State University (RSU)

Table #3-1: University Sites for the SCIT Study

3.2 SELECTING FOUR DISCIPLINES

We chose four disciplines which would provide contrasts in material use practices. We selected disciplines that vary in their modes of inquiry, resource levels, research materials and the visibility of their presence in online-discussion groups and networked bibliographic databases: a lab science (molecular biology); an artifact-based discipline (computer science); a social science (sociology); and a humanities discipline (literary theory). We explicitly decided not to study professional fields such as medicine, law and business because of practical considerations (off-campus facilities, the predominance of professional experience in training).

In each case, we considered alternative disciplines, such as mechanical engineering as an artifact-based discipline, philosophy as a humanities discipline, economics or psychology in the social sciences, and chemistry as a lab science. Our basic familiarity with a field so that we could more quickly grasp the worlds in which these researchers do their work was also a consideration. We will briefly discuss the bases for each of our final choices.

a. Molecular Biology is well-funded, is based upon laboratory work and work is interdependent on shared materials and data. Molecular biology is also the focus of several studies in Computer Supported Cooperative Work about the Worm Community (Schatz, 1993, Star and Ruhleder, 1994) and was also one of the exemplars for a recent study of national collaboratories (Computer Science Telecommunications Board, 1993). Within the molecular biology community, however, we chose the drosophila (fruitfly) subspecialty because more drosophila researchers were available and we were aware of their use of FLYBASE electronic database, a system that was similar in some ways to the worm community system.

b. Sociology encompasses multiple paradigms of inquiry and some specialties use large data sets or mathematical analytical techniques. We focused on the social networks subspecialty since we suspected that its participants might be more likely to have some familiarity and use of computer networks.

c. Computer science was a locus of technological development of digital library infrastructure. Within computer science, we focused on computer networking researchers since they were the first specialty area to adopt computer networks in support of their own work.

d. Literary Theory is a text-based discipline that had a large stake in print materials and collections. We chose to focus on the subspecialty of comparative literature since we knew of some appropriate electronic collections available and there were some avant guard uses of electronic materials, such as electronic journals.

3.3. SELECTING FACULTY AND GRADUATE STUDENT INFORMANTS IN EACH DISCIPLINE

Before this study, we executed a small pilot study with a simplified design to examine conceptions of digital library use focusing on faculty and digital library infrastructure providers on two campuses (Covi and Kling, 1996). This study provided a means to test the protocol of the proposed research. The pilot study led us to refocus our data collection on a larger number of faculty researchers and decrease the number of digital library infrastructure informants.

We interviewed three faculty members in each of four departments. (See Appendix II for a copy of the faculty interview schedule). We initially tried to focus on arranging interviews with informants in particular research subspecialties to develop some commonalties within each discipline. When possible, we selected informants at different institutions who worked in similar research subspecialties to follow invisible college ties through referral, journal editorial boards, and conference program committees. This strategy also helped us verify whether researchers shared reported work practices across a particular subspecialty.

The faculty informants assisted us in contacting graduate students. We chose to include doctoral student informants in our study for several reasons. First they often are instrumental in producing research (their own or in conjunction with their advisor) and could provide us some perspective on the outlook for research as they consider embarking on careers in academia or industry. Also, some of our colleagues suggested to us that doctoral students would be more comfortable with the technology and therefore more likely to use it. Besides getting a generational difference perspective in graduate preparation (which later showed up in our interviews of faculty informants), the doctoral student informants had differing access to materials, technology and other resources than the faculty.

Table #3.2 characterizes our informants by their academic rank. We interviewed more graduate students in molecular biology and computer science because they were more likely to work on campus in labs and offices in their departments. There was a fairly good mix of informants by rank in the other disciplines. The overall mix of informants rank was somewhat representative of faculty in their departments in terms of the growth of their research subspecialty and promotion practices in the different disciplines.
 

	Graduate Students	Ass't  Professor	Assoc.  Professor	Professor	Total (per field)
Molecular Biology	10	6	11	7	34
Literary Theory	6	4	6	14	30
Sociology	4	3	9	12	28
Computer Science	8	8	6	10	32
Total (by rank)	28	21	32	43	124

Table #3.2: Profile of Faculty Informants by Rank and Discipline

3.4 INTERVIEWS WITH UNIVERSITY LIBRARIANS AND ACADEMIC COMPUTING ADMINISTRATORS

We primarily interviewed digital library infrastructure providers responsible for the key resource allocation decisions and supplement our understanding of what digital library infrastructure resources were available at each campus through collecting documentation at campus computer centers and libraries. In many cases, we also spent several hours at each campus utilizing their library and computer resources ourselves. Based on previous work and consultation with expert colleagues, we focused data collection around features of campus and departmental organization perceived to effect researchers' accessibility to paper and electronic materials (Kling, 1987; Kling and Jewett, 1994).

Table #3.3 indicates our key informants for an overview of the digital library infrastructure on each campus.
 

	DSU	MU	TU	RSU	RU	HU	BSU	FSU
University Librarian	X	X	X	X	X	X	X	X
Dir. Academic Computing	X	NA	X	X	X	X	X	NA
Chair, Faculty Library Committee			X	X	X	X	X	X
VP/Provost	X	X	X					X

Table #3.3 Informants about Campus Digital Library Infrastructure

In two cases, there was no comparable position for Academic Computing Director and we met instead with the Vice-President, Provost or Senior-level administrator who oversaw the academic computing domain. After the first two site visits, we added an interview with the faculty member at each campus who chaired the faculty library committee. The faculty library chair supplemented the accounts of the digital library infrastructure providers with opinions on resource provision from a faculty perspective. These interviews also provided insights into whether the needs faculty informants expressed were under consideration by providers.

For the interviews with the key digital library infrastructure providers, we used the following or a very similar variant of the following protocol for a semi-structured interview:

Background:

Professional/Educational Background

At what other institutions/universities has s/he worked?

Years at this university

Scope of Responsibility

Decision-making:

Is there a preference for electronic formats in allocation for scholarly resources?

What are the current budgetary patterns?

What is their relationship to neighboring universities'/communities' libraries or computing facilities?

What consortiums are important to the facility?

What are surprises about coming to this institution or with new senior administrators/presidents?

What are the key trade-offs?

Status of Services

What are the strengths and weaknesses of this Library/Computing Facility?

What is the faculty demand for computing/library services?

What are the constituencies for this demand?

Are there specific constituencies who prefer electronic or paper materials?

Is there pressure for expanding library or computing services?

What joint projects (if any) are there between the library and the computing center?

Requests for Materials

Organizational Charts, Vision Statements, Recent Reports or Assessments, IPEDS data, ARL report, Annual Reports, Budget Statements

3.5 OVERVIEW OF DATA COLLECTED

Several kinds of data were collected for this study. Each interview was tape-recorded except for meetings with infrastructure providers and several researchers who were uncomfortable with the recording. In addition to audio record, transcripts and field notes, we also made annotations on interview schedules to assist in analysis. When possible, we photographed the researchers in their offices as a memory aid to analysis and to compare the amount of paper and technology in different offices throughout the study.

At all sites, a site host (usually the university librarian or academic computer director) granted us access to library databases, facilities and Internet. We kept field notes on our direct observations of digital library and insights into reported difficulties or dilemmas with campus resources. If necessary, we consulted with the key digital library infrastructure providers for explanations of service trade-offs and continuing problems.

During interviews with the university researchers, we used a "grand tour" question (Spradley, 1979) to elicit material use patterns by asking researchers to select a manuscript they were working on in an advanced stage or had published and describe the sources of materials they used in its development. In most cases, we worked our ways sytematically through the bibliography -- asking how the person learned about and obtained copies of each item cited. Some of the researchers were able to provide to us a copy of their manuscript which aided transcription and analysis. We also took field notes on observations when informants demonstrated certain paper and electronic resources at hand they regularly.

We also collected materials from public information sources (brochures, newspapers, etc.), Academic Computer Centers and Libraries. Most universities' libraries and computer centers published resource guides, user documentation, directories, organizational charts, departmental and campus-wide studies, vision statements and statistical reports. We also collected some local publications that provided a sense of the cultural and intellectual character of the community.

3.6 DATA ANALYSIS

The main form of data analysis was theoretical evolution through grounded analysis of this data (Strauss, 1987).

We coded from transcripts of the richest faculty research interviews. Proceeding from the initial coding we discussed and wrote analytic memos about the match between the analysis and the hypotheses and other emergent themes. Our initial coding focused on ways that researchers' concern for, knowledge of, and use of paper and digital resources were influenced by disciplinary, campus and departmental resource arrangements.

Each chapter in this report examine different themes, and data were analyzed somewhat differently as we developed each of these major themes. We discuss some of the limitations of our research methods in Appendix III.

4. DIGITAL MATERIALS AND DISCIPLINARY PRACTICES

We selected four disciplines to serve as a focus of this study: a lab science (molecular biology); an artifact-based discipline (computer science); a social science (sociology); and a humanities discipline (literary theory). These fields were chosen because of the different types of materials different disciplines use for their work. They also vary in their relative wealth and access to networked computing resources.

It helps to give the reader a portrait of the typical ways that faculty in each of these disciplines integrate paper and digital documents into their research. We will discuss the disciplines in a sequence -- from computer science to sociology -- that parallels the extent to which most faculty seemed to make decreasing use of digital documents.

4.1 COMPUTER SCIENCE

Computer scientists study computers as artifacts and also examine some applications of computing to "real world" problems. This short description provides a general thumbnail sketch of the field:

Computer science is the study of computers--namely, their design (architecture) and their uses for computations, data processing, and systems control. Computer science includes engineering activities such as the design of computers and of the hardware and software that make up computer systems .... The major subdisciplines of computer science have traditionally been (1) architecture (including all levels of hardware design, as well as the integration of hardware and software components to form computer systems), (2) software( the programs, or sets of instructions, that tell a computer how to carry out tasks), here subdivided into software engineering, programming languages, operating systems, information systems and databases, artificial intelligence, and computer graphics, and (3) theory, which includes computational methods and numerical analysis on the one hand and data structures and algorithms on the other. (Britannica Online, 1996c).

Computer science had some identity problems as a discipline: computer scientists did not have a consensus on the balance between responsiveness to external interests for information technology and preservation of the integrity of the discipline as a "self-oriented science" (See Hartmanis and Lin, 1992). This controversy reflected the dual origins of computer science from engineering (as an applied discipline) and mathematics (a "pure science").

Our study focuses primarily on the more engineering-like research subspecialties. Specialization in computer science is primarily organized around professional organization umbrellas, such as the Institute of Electrical and Electronics Engineers(IEEE) or the Association for Computing Machinery (ACM) which have numerous special interest groups and conferences. There are other organization that sponsor publications and conference in computer science but they often correspond to specialized groups in the ACM and IEEE.

Computer Scientists produced artifacts and theoretical findings and published their analyses in scientific journals, conference proceedings, and a variety of reports. Their work built on existing findings and technology. But research production requires a new or creative contribution via theory, proof of concept, or application of computational methods. Computer scientists tended to work within major subdisciplines, such as those mentioned above, but sometimes worked in several subspecialties such as drawing on information theory to design artificial intelligent agents. Most computer scientists belonged to several special interest groups that supported conferences and journals in their subspecialties.

Their work was also organized in ways that reflected doctoral training, external funding, and the division of labor among research groups headed by faculty investigators. In general, computer science doctoral students were supported by research grants administered by their advisors. They also often pursued internship opportunities in industry at different points of their doctoral training. The computer science faculty usually had close ties to funding agents in industry and government. They worked on multi-year grants funded typically by the Advanced Projects Research Agency (ARPA), National Science Foundation, and corporate sponsors. Funding levels varied (some of our CS informants had little or no external funding). Our typical CS faculty informant had annual grants of several hundred thousand dollars per year and several had over one million dollars per year. Computer science faculty worked as managers of the funded work, supporting groups of doctoral students who designed and implemented the systems and published on specific aspects of the project for their dissertations.

One assistant professor contrasted the shift in his work to writing (using the LaTex text formatting program) to programming in the C language:

I spend a lot more of my time writing English than writing C code -- programming. I run LaTex a lot more than I run the C compiler. [TUCS3]

The inputs to work production in computer science in terms of personnel and equipment were not always shared between research groups. Computer science departments usually provided a small group of computer support professionals to install and maintain computer systems and networks for all projects in the department. Although many researchers no longer used shared computational resources such as mainframes and central file servers, almost all researchers had desktop workstations connected to local high-speed computer networks in their offices which required coordination and support. In addition, some computer science departments provided direct dial-up access to the local resources to alleviate difficulties in access departmental resources through the regular campus network. Some computer science researchers used shared facilities such as supercomputers, but predominantly the department provided electronic mail access and storage backup rather than the bulk of computational resources.

Computer scientists had a normative work pattern that influenced their demand for infrastructural resources. They favored a work day beginning late in the morning and extending into the wee hours of the night. Although some researchers worked more of a business day schedule due to class schedules, outside consulting or personal family demands, computer scientists demanded a high level of access to and reliability in their computational resources around the clock. For instance, an unexpected outage of electronic mail was perceived as disastrous. often computer scientists organized their work to maximize their level of control, access and reliability to needed resources. Some computer scientists kept their primary work machine at home maintaining acceptable connectivity to the campus network. In general, computer scientists worked in the location that provided the best access to computational resources.

Research outputs were primarily short articles which appeared in diverse forums, usually in IEEE, or ACM-sponsored scientific conferences and journals. Conferences in computer science were considered a serious form of publication and were counted in performance evaluation. The importance of producing viable contributions for externalfunders was one explanation. Computer scientists judged the quality of the conference based on its rejection rate: the higher percentage of papers rejected, the better the conference. A "strong conference" has a 20% acceptance rate.

(In talks about our study, computer scientists have been surprised to learn that molecular biologists accept almost 100% of the submitted conference papers for presentation. In contrast, biologists have been surprised that computer scientists treat conference proceedings as a serious form of publication.)

In addition, computer scientists often published project reports, professional articles and technical reports distributed by their computer science department which were typically not refereed nor counted heavily toward promotion and raises. However, these non-refereed publications did play a role in securing and soliciting external funding. Computer science publications were predominantly multi-authored (with the convention that the first author received the most credit for the work, unless the order was alphabetical).

Computer scientists also produced software systems as another type of research product. Software systems were often used by research groups for several years. Sometimes computer scientists shared their systems with other researchers in a similar subspecialty, but these systems were typically used for research problems rather than industrial-oriented applications. Sometimes a group of researchers in a subspecialty shared test suites or common approaches to test the robustness of the systems they produced. These outputs were also cumulatively compiled and used in multiple projects over time. Experimental computer scientists were expected to develop new systems every few years in order to be considered productive, but also because of the rapid pace of technological innovation .

The enthusiasm for creating new artifacts did not parallel the technological innovation for publishing research in new media. The case of electronic journal publishing is an useful contrast that supports the notion that material use practices will change more slowly than the availability of new artifacts. Few informants in any discipline used electronic journals to identify source material. One computer science informant had contributed an article to the Electronic Journal of Virtual Culture "as an experiment" but did not believe his colleagues would consider it recognized as a legitimate publishing outlet.

I actually published in one [electronic journal]... To tell you the truth I haven't looked at it since then. In fact, I don't know whether it is still around or not .... Since I haven't come up for tenure, I think that [whether his article will count for tenure] will be fully answered only at that point. I only did it because I was curious about the whole electronic publishing area. This was... sort of a chapter from my dissertation which was my opinion about an issue .... I deliberately went to an outlet which I found was not controlled by the hierarchy so I could get my ideas published and get them out of my head and move on to something else. [DSUCS3]

In general, computer scientists felt that journals distributed in paper format were more legitimate, since most prestigious journals in their research subspecialties were distributed in paper (Kling and Covi, 1995). However, there was one exception that illustrated the pervasiveness of these normative values. Artificial Intelligence researchers published in and regularly read the Journal of Artificial Intelligence Research (JAIR). However, JAIR differed from other forays into electronic journal publishing because its publisher (Morgan-Kaufman) sold each volume in paper bound form at the end of each year. Although it was published and distributed electronically free of charge, the papers published via JAIR were not visually identifiable as electronic journal articles. Instead of distribution in plain text, JAIR distributed articles in postscript format which closely resembled a photocopy from the bound volume. One informant explained his understanding of JAIR's value:

The major journal of AI is the AI Journal [AIJ - paper] and it has a 2-year backlog and it takes a year to get papers reviewed... by the time they come out they're not relevant any longer. So [JAIR-electronic ] was an attempt to do something [about this problem]... It's not uncommon to try to get the reviews back in 6 weeks. You can get a paper published within two-three months of writing it .... The idea [behind using an electronic format that prints like a print journal] is ... if your deans are going to say, "Is this an electronic journal?" ... you can show him that it's a real journal, that the people using it can read it [in a print journal format] .... I guess I'm not quite sure [if it's "better" to get into AIJ-paper]. My feeling is that actually the [JAIR-electronic ] is better. [AIJ] unfortunately has had the same editor for 20 years and he*s been focusing it towards a certain class of research which is becoming less and less relevant to my own work .... [FSUCS1]

In this account, the informant first depicted the value of JAIR as a faster way of publishing research before it gets out of date. This account supports arguments about the impending demise of paper journals due to their inflexibility, the availability of the article around the clock, and the advantages of online searching (Odlyzko, 1995). However, as the informant continued to explain his use, he placed JAIR in relation to the norms of work production. JAIR met a need in the subspecialty to have an additional publication outlet with a faster time to publish. In addition, he has marginalized the value of the older paper journal due to the content rather than the distribution mechanism.

Even computer scientists who were associate editors of electronic journals expressed some doubts about the benefits of electronic publication. This senior professor contrasts benefits with the central issue of quality in research publication .

I can't answer [how e-journals figure into merit and promotion decisions_] very definitely because I do not know what quality these electronic journals will be. The conviction which we have for the [e-journal for which he is an associate editor] is that it will be fiercely written and so I would have no objections [to having e-journal articles count for tenure] if the evidence is strong that these are not just a sloppy way of getting papers published, you know -half-baked ideas .... [MUCS2]

Work production in computer science, as in the other disciplines, thus hinged on evaluating work based on the social processes that determine quality. In order to create quality contributions, researchers drew upon norms of work production.

4.1.1 COMPUTER NETWORKS SUBSPECIALTY

Computer Networks research examines the architecture of computer communications. Here is a brief description of the major aspects of the research area:

An other important architectural area is the computer communications network, in which computers are linked together via cables over short distances to form local-area networks (LANs) or via telephone lines or satellite links to form wide-area networks (WANs). By the 1990s, worldwide communication became possible by internetworking, the interconnection of multiple networks by means of so-called gateways. Linking computers physically is easy; the challenge for computer scientists has been the development of protocols (i.e., standardized rules for the format and exchange of messages) to allow processes running on host computers to interpret the signals they receive and to engage in meaningful "conversations" in order to accomplish tasks on behalf of users. (Britannica Online, 1996d)

Much of the impetus for this work came from government and industry stakeholders in developing these protocols. Like other highly visible subspecialties in computer science, researchers were eligible for funding under large projects. This informant described a well-funded joint project to implement and test technologies such as the Asynchronous Transfer Mode (ATM):

The key product of our research is developed into protocols - network protocols and these protocols, we usually simulate, software simulation. We evaluate the analytic tools for simulation. And in some cases, we also implement. Now we have a couple of ARPA grants for which we are required to actually implement these protocols to see if they work .... We are now part of a testbed called [project]. It is an ATM testbed ... we're in the process of buying a switch right now and we will evaluate different congestion control schemes for ATM. For that we will have to be complying to standards. [BSUCS1]

In relation to other subspecialties in computer science, computer networks was one of the best funded and had the largest audience due to increased public interest in computer networking. The time frame for producing work in computer networks research was typically quite short in comparison with other disciplines. Here is an account of the genesis of a project from one researcher:

the workshop paper was presented in December, was written within the few months before that, I probably first started thinking seriously about it around a year ago or maybe a little more than a year ago. [TUCS3]

Although there was some competition in the computer networks subspecialty, there was less risk of loss for priority of discovery (as compared to molecular biology research). Like other computer science subspecialties, there were few penalties for not citing other relevant work. Researchers instead valued producing something important to address current problems in an appropriate time frame.

I guess I saw that this paper was coming out and from the title of the paper I knew it was related to what I was doing and I was sort of anxious - did somebody else just think of the same idea? But no, I wasn't scooped. What this other paper was, was sort of a pretty good improvement on the same basic way that the ten-year-old stuff had done. Interestingly unaware of the ten-year-old stuff as it turned out. [TUCS3]

Computer networks researchers typically organized themselves in project groups of doctoral students directed by a faculty researcher. They had regular meetings for their project group. The doctoral students shared group offices with separate desks and workstations whereas the faculty member usually worked in an office elsewhere in the department. Research groups usually shared meeting facilities, computer equipment and journals and conference (usually owned by the faculty director). In addition, doctoral students used departmental reading rooms and campus branch libraries to obtain paper materials.

The inputs for work production in computer networking often followed from needs in the telecommunications industry. Doctoral students in computer networks, like those in other computer science specialties often found ideas for dissertation projects in work related to projects their advisors were conducting. Computer science doctoral students would often work for industry for short periods of time, especially when their project was related to a current problem faced by industry.

Although several researchers in computer networks mentioned accessing bibliographies mounted on the WWW, they found the most relevant materials in conference proceedings. They attended and collected proceedings for the IEEE INFOCOMM, IEEE ICCC, ACM SIGCOMM, and the jointly sponsored IC3N conferences. Computer networks researchers belonged to the IEEE Communication Society as well as ACM's special interest group SIGCOMM. They typically read IEEE Networks, IEEE Transactions on Communications (TOC), Communications Magazine, IEEE/ACM Transactions on Networking (TON) as well as other subspecialty journals.

Students tended to use bibliographic databases, use net newsgroups and WWW more intensely than faculty researchers. They often served as "human" intelligent agents and filters to help faculty researchers identify lesser known literature or discussions. However, the primary role of the students was to produce the systems and analysis for the project. The faculty were generally pleased when the students found relevant references for the project, but faculty researchers tended to rely on current publications in conferences, journals and magazines, many of which they examined before publication through peer review.

The outputs of computer networks research appears in conference papers, journal articles and nonrefereed publications. The average length of computer networks publications were shorter than papers in most areas of sociology and literary theory but longer than the molecular biology papers. However, the production nonrefereed work distinguished computer science work production from the other disciplines.

4.2 MOLECULAR BIOLOGY

Molecular biology as a discipline is a more recent subspecialty which grew from the discovery of DNA in the 1940s. This description provides an overview of the topic of inquiry:

Molecular Biology is a field of science concerned with studying the chemical structures and processes of biological phenomena at the molecular level. Of growing importance since the 1940s, molecular biology developed out of the related fields of biochemistry, genetics, and biophysics. The discipline is particularly concerned with the study of proteins, nucleic acids, and enzymes -- i.e., the macromolecules that are essential to life processes. Molecular biology seeks to understand the three-dimensional structure of these macromolecules through such techniques as X-ray diffraction and electron microscopy. The discipline particularly seeks to understand the molecular basis of genetic processes; molecular biologists map the location of genes on specific chromosomes, associate these genes with particular characters of an organism, and use recombinant-DNA technology to isolate and modify specific genes. (Britannica Online, 1996a).

Work production in molecular biology consisted of producing unique research results of that were acceptable for publication in a set of molecular biology journals. Molecular biologists worked on biological phenomena building directly on preceding work. They established themselves in specialties through many years of specialized training after a bachelor's degree including positions as technicians, doctoral student researchers, postdoctoral researchers and sometimes experience in industrial labs. Molecular biologists identified themselves in a hierarchy by their doctoral advisor and sometimes spoke of being a member of a generation of study in a particular specialty. Lab directors usually served as doctoral advisors for the doctoral students working in their labs. Lab directors were faculty researchers who coordinated the activities of the lab and facilitated close cooperation within the lab and with collaborators in other labs. Contribution to work in molecular biology was predicated on priority of discovery. Together with the temporal nature of the their biological materials, researchers were very concerned with coordinating their work with collaborators and if possible, competitors.

Molecular biologists shared inputs to work production. As a lab science, molecular biologists worked in lab groups that shared the same facilities and related research pursuits. They trained in laboratories run by faculty researchers who served as principal investigators of multiple grants providing total support upwards of $100,000 each year.

Because these researchers worked with biological materials which change over time, their work centered around access to their lab. They often worked on weekends in their facilities to support their tasks. They needed reliable access to biological, paper and electronic materials in their laboratories to conduct their work. Labs shared biological materials through centers that distributed biological material such as stocks, strains or cultures. They shared documents usually through fax or postal mail. They used electronic mail frequently to order biological materials or request documents but not to exchange documents. Some researchers used electronic mail to send reviews of journal, conference or grant submissions.

The output of work production in molecular biology appeared in a fairly homogeneous set of scientific journals. One set of outputs were short articles published in scientific journals chronicling advances that researcher teams had achieved. Molecular biologists coauthored most work which reflected the contributions of a variety of participants who worked on different parts of the problem. Usually, the first author was the person who took the lead on the project and the last author was the faculty researcher that supervised the project.

Because molecular biology production builds so much on previous work, crediting previous discovery when reporting a new discovery plays an important role in work production. However, since journal articles are relatively short, researchers must make choices about how to credit previous work. An other distinguishing feature of publication in molecular biology was its typical frequency. Nature and Science were both weekly, Cell was published biweekly and Genes and Development was published monthly. Although the publication lag was much shorter than in other disciplines, molecular biologists were greatly concerned about coordinating publication with discovery.

An other set of outputs were contributions to centrally maintained databases such as GENBANK. Publishers and granting agencies required that molecular biologists submit sequencing data to GENBANK before they would accept an article for publication or renew a grant. This public data was often repackaged for use in other specialty databases. The sharing of gene sequences was very important to the flow of production among molecular biologists. The following example illustrates how a field grew from the discovery of a homology (a sequence match) between different genes different researchers were studying:

It's funny, for 5 years [we didn't have any competitors] not at all. And it turned out that this ... gene had homology to some other genes that people were studying. It wound up being very important and so now, it went from a field that had maybe 10 people working on it, to a field that has maybe has 100-150 people working on it which was both a blessing and a curse. It's a blessing in the sense that there's a lot more interest in it and in a way there's more money involved in it. But it's a curse in a sense that constant pressure to get something novel out. So yes, now there are many people working on it because of these genes. I told you it's what's called a transcription factor, it turns on other genes. It turns out that one of the those things it turns on is HIV- the AIDS virus. So this is a protein that controls the replication of the AIDS virus. So right away there's a lot of people interested. [RUMB1]

Conferences provided an other forum for the dissemination of research results. Researchers were required to submit an abstract to present at the conference, but acceptance was open since most researchers waited until research articles had been accepted for publication in a scientific journal before announcing their results. There were also smaller invited meetings called "Gordon Conferences" run by a group of senior researchers to encourage open communication between researchers. At Gordon conferences, researchers examined the direction of advances in the field rather than producing results from discussions at the conference. Researchers who lived in metropolitan areas often attended city or regional research seminars coordinated via electronic mail. Some researchers found these meetings to be convenient opportunities to arrange face-to-face meetings with our of state researchers giving visiting talks.

An other key aspect of work production in molecular biology was the way lab directors coordinated specialization of their lab's work. Because of the desire to coordinate publication with discovery, and the risk of an other lab's result eclipsing their work, lab directors spent a great deal of time (some reported 50%) communicating with collaborators (and sometimes competitors) on the telephone, over electronic mail and fax.

As a result, lab directors tended to know more about what other labs were doing than their staff members. Conferences (especially the Gordon conferences) supplemented the individual coordination laboratory directors maintained in order to produce contributions in their research subspecialty. Lab directors also coordinated collaborative projects with other labs. They shared biological material as well as paper and electronic materials.

Molecular biologists described their specializations in several ways. The most specific way was the particular problem they were working on such as identifying a structure, sequencing a particular gene, determining the function of a gene. More broadly, molecular biologists work in research subspecialties focusing on molecular aspects of model organisms. In this study the majority of molecular biologist informants were drosophila (fruit-fly) researchers. We also interviewed researchers in other model organism communities including one c.elegans (worm) researcher and several e.coli researchers. Although molecular biologists also identified themselves in terms of biochemistry, genetics, cell biology and microbiology, our analysis examines patterns within the model organism communities since that was the subspecialty to which researchers most often referred.

4.2.1 DROSOPHILA SUBSPECIALTY

Drosophila researchers as a research subspecialty in molecular biology, were an older subspecialty (than c. elegans), and had developed electronic resources for their work. This section provides details from this subspecialty that illustrate specific aspects of work production in molecular biology.

Drosophila melanogaster is a fruit fly, a little insect about 3mm long, of the kind that accumulates around spoiled fruit. It is also one of the most valuable of organisms in biological research, particularly in genetics and developmental biology. Drosophila has been used as a model organism for research for almost a century, and today, several thousand scientists are working on many different aspects of the fruit fly. (Manning, 1996)

Drosophila researchers study the fruit fly as a model organism.[1] They refer to their peers as "fly people" or the "fly community." Biologists have been working with drosophila for many years so as molecular biology has grown in importance, drosophila became a popular model organism. Compared to other model organism work, drosophila work was cheaper. A group at Berkeley was conducting the drosophila genome project to clone all the drosophila genome and then to sequence it. This would provide drosophila researchers a different starting point of focus on more experimental issues without having to sequence it first. This example illustrated how the fly community's work is expected to change as a result of a large scale efforts to provide an exhaustive infrastructure for work.

Currently, the inputs for work include the sharing of biological materials through stock centers and documentation of sequence information through GENBANK. Drosophila researchers frequently mentioned using an electronic resource called FLYBASE usually via the gopher client software (it had recently become available on the WWW). It contained contact information about the drosophila researchers themselves as well as the addresses of the stock centers. It was started in an effort to make a list of cloned DNA sequences available to all researchers and put the data in the field-defining "RedBook" (Lindsley and Zimm, 1992) online. FLYBASE now includes:

! A bibliography of over 82,000 drosophila citations

! An address book of over 5,000 drosophila researchers

! Information on more than 32,000 alleles of nearly 10,000 genes

! Descriptions of over 12,000 chromosomal aberrations

! Drosophila genetic map information

! Information on the functions of gene products

! Lists of stock center and private lab drosophila stocks

! A listing of over 9,000 nucleic and over 3,000 protein sequence accession numbers

! Lists of over 7,000 genomic clones

! Allied databases

! Berkeley drosophila Genome Project data

! European drosophila Genome Project data

! the bionet.drosophila archives

! Drosophila Images

! Wild type drosophila strains and chromosomes (Flybase, 1996)

Most of the drosophila researchers primarily used the FLYBASE front-end to locate other molecular biologists or order stocks. For gene sequencing, most researchers used other computational resources over the Internet (such as BLAST or GCG) to match DNA and protein sequences in central databases (such as GENBANK, EMBL, FLYBASE data sets).

In terms of disseminating their results, drosophila researchers published in a similar set of scientific journals. Most molecular biologists subscribed to Nature, Science, Cell, Development and sometimes Genes and Development, and Genetics. These major journals were also indexed in MEDLINE which was available at all universities in this study. In addition some researchers also used BIOSIS, a database with similar coverage of journals but was based on Biological Abstracts (produced by BIOSIS) rather than Index Medicus (produced by the National Library of Medicine). Both bibliographic databases provided a large percentage of abstracts in addition to citations.

An other means of disseminating work outputs were conferences. There was a large annual research conference which they call the "fly meeting."They also attended a variety of annual conferences and often regional seminars. They shared with other molecular biologists materials such as biological indices, supported electronic and postal mailing lists and paper journal publications. In addition, they, like other model organism researchers, had developed electronic resources for disseminating research results, reference resources and contact information unique to their subspecialty. In addition to personal electronic mail and regional seminar mailing lists, some researchers, particularly doctoral students, used BIONET newsgroups to discuss techniques, get help with problem solving or read announcements.

BIOSCI [BIONET] is a series of freely accessible electronic communication forums (i.e., electronic bulletin boards or "newsgroups") for use by biological scientists worldwide. No fees are charged for the service. The system is intended to promote communication between professionals in the biological sciences. All postings to the newsgroups should be made in that spirit. While the general public may"listen in" to the discussions, these newsgroups are intended primarily for communications between researchers. There are other forums on use net such as sci.bio for the asking and answering of biological questions from lay persons. (BIOSCI/BIONET, 1996).

Although they sometimes found these newsgroups useful, faculty researchers sometimes found them too repetitive and chatty.

Drosophila lab directors resembled the other molecular biology lab directors in terms of coordinating the work of the lab and defining their specializations. However, one informant described an example of openness on the part of a leader in drosophila work:

the field as a whole is cooperative - it has had to be and we were lucky that the huge movers and shakers like Jerry Rubin... did something .... In 1981, Jerry and Allen Spradler figured out how to transform drosophila...absolutely necessary biological research. And it's like harder than hell to do. They figured out how to do it .... So what they did is that they announced it at one of these annual drosophila meetings. And of course, you can image how one might announce this. It's like the coming of the holy grail, you can do it to get everyone on their knees ...What they did was say okay,...Here's a 3X5 card going out. Put your name and address if you want the recipe and the stocks and we'll send them to you. I mean they went out of their way to be open, to be cooperative to be as helpful as they could be. Because they said look! we live on the fact that the fly community has existed and has made these mutations and has done all this work. We couldn't have done what we did with out the fly community. We want to share and give back just you do and in so doing, of course, they set the tone which had already -they confirmed the tone. That this is a community that shares. The guy who started that was the guy who started flies and that's Thomas Hunt Morgan. He had a dictum to share views. {BSUMBS]

The drosophila subspecialty was a rather typical example of a model-organism molecular biology subspecialty. Although work in molecular biology cross-cuts model organism subspecialties, the nature of working with a particular organism was a good way to compare work patterns.

4.3 LITERARY THEORY

Literary theory, as a research discipline in this study, refers to the analysis of literature by drawing on theories of philosophy, linguistics, cultural studies and other fields in humanities and social sciences. A literary theory, as a particular analytic approach refers to one of a set of theories employed in textual analysis. The use of these terms can be confusing, even to people who work in the discipline. One definition equated literary theory with one particular theory: poststructuralism (Komar, 1994). On the other hand, one informant described his theoretical work as "literary theory of romanticism" which was, in fact, opposed to poststructuralism. Literary theorists often provided their own definitions when writing about "literary theory" and other related terms. In an introspective paper about literary criticism and theory, one scholar writes:

By 'literary discourse' I mean writing about literature; by 'theory' I mean general principles applied in literary discourse; and by 'criticism' I mean commentary on specific literary texts. (Harris, 1996)

The confusion over the use of the term literary theory was characteristic of work production in the discipline. Establishing meaning was a central activity of work production in literary theory. with respect to the ways the informants presented their work production, this study refers to them as scholars rather than researchers. This term sets them of from informants in the other three disciplines who predominantly characterized themselves as scientists: computer scientists, social scientists or biological scientists. Instead, literary theorists are humanists, part of the family of inquiry that includes philosophy, languages, classics and history. The term research was also problematic. For example, one informant took research to mean scouring the library (or even the libraries of the world) for critical materials:

It would probably be a misnomer to say I do research. I mean I do arguments and so I'd be more interested in, informed by what other critics have had to say about this argument. So I read a lot in philosophy, I read alot in social theory, but I don't do anything resembling research and there's never a point where I would need the standard week in the library to find arcane text .... I deal primarily with fairly well known primary texts. [DSULT3]

However, the term scholar and scholarship was problematic for other informants. For this informant "scholarship" connoted a historical tradition of an ascetic lifestyle of library work separated from worldly concerns.

I'm not really a scholar .... I guess I do have a feeling, especially the semester I spent in the Bibliotheque Nationale, it was sort of playing at being a scholar. Going to the library everyday and reading all these old 19th century books, and that was - I knew that was not really me... I was playing that role for a moment ....Critic versus scholar is usually is the way we used to speak of it. But I'm not really, not particularly nostalgic for that. [MULT1]

Noting these conflicts, we often refer to literary theorists as scholars rather than researchers.

Similar to sociology, the pluralistic nature of the discipline produced central disagreements about a common paradigm for inquiry. In several key debates in the study of literature, literary theory as a research discipline plays a central role. Jay Parini, a literary theorist, wrote in an opinion column in the Chronicle of Higher Education about one key debate:

Traditional scholars -- those who edit texts, write biographies, and provide close readings of poems and novels -- are upset by literary theory and its supposed downgrading of literature into "textuality." (Parini, 1995)

Although some informants relied on more "traditional approaches" in their work, they were all aware of the literary theory approach and provided contrasts between their techniques and those of "literary theory." Another conflict in the study of literature was between literary theory, creative writing, and criticism outside the university. For example, Mark Edmundson, a literary theorist at the University of Virginia, characterized this conflict as reminiscent of the classical debate between poets and philosophers about the nature and purpose of literature (Edmundson,1995).

These debates provide clues to the boundaries of what literary theorists consider their discipline. These definitions of work also circumscribe the generalizability of this studies findings to the "literary theory" approach to analyzing literature. There were several informants who worked in both literary theory and other approaches or disciplines. We will note differences in those informants' material use practices.

In the acrimony of some debates, some academics and administrators question the legitimacy of the study of literature. The tradition of literary studies dates back to ancient times but had declined in status over the years in the face of increasing emphasis on big science at most research universities. Sociology suffered similar problems with legitimacy. Reginald Gibbons, a poetry critic, remarks in his examination of criticism in the university:

If there is a crisis in literary criticism as it is practiced in the academy, it is not the struggle between opposing theoretical camps, but the question of whether any kind of literary criticism is of great value now (Gibbons, 1985)

The declining legitimacy of literary studies affected the infrastructural arrangements for work production. There were few sources of funding for their scholarship. Small grants from within universities and fellowships outside the university often paid for conference attendance, travel to remote libraries for study and sometimes even infrastructure for materials use. However, most literary theorists depended upon allocations from the department, school or university. They often used old computer equipment, slow computer network connections and scarce computer support assistance from their department offices. In several departments, informants did not even have access to high speed networks. When they did have access, it was usually because of some additional responsibility that spurred an installation :

Well, I think it started with this [university committee about technology] that I was on ....Then the chair says well of course we'll send memos to each other via e-mail and I was the only person in the humanities and I was the only person that was not connected to e-mail. And so... The chair ... had to get special permission for me to plug into the [campus network] because this building was not even wired ... other people [in the building ] had to use modems... so that I could use e-mail. So then I was probably a year or two ahead of colleagues in using e-mail .... A work study student gave me a ten-minute orientation. with that ten-minute orientation I was the department's instant expert. [MULT3]

This informant emphatically denied that she was knowledgeable about computers. In fact, although she was publishing a paper concerning theoretical issues in computer mediated communication, she confessed that the occasion for her increased interest and use of electronic networks and services was because of a family issue at home:

That actually came about because of a news report a couple of years ago with the earthquake. The LA earthquake. There were earthquakes in Los Angeles and there was major blizzards up east. And there was a news report that Compuserve's use jumped those couple of days - 70% and most of the place where it jumped was on kids whowere home from school and were sending messages to each other across the country saying what about the earthquake. And we thought, our children are missing out. [MULT3]

In general, literary theorists had low access to computer skill and depended upon the scarce help of shared computer support specialists. At several universities there was only one specialist assigned to all departments in the humanities. Therefore literary theorists primarily had to rely upon themselves to learn how to use digital resources effectively in their work. Many of their institutions offered Internet courses, but there was a common problem with these opportunities:

Unless it is directly relevant to what I'm doing at the moment, if I go to a 2-hour seminar on the world-wide web, it sounds interesting ....I did that at the very beginning of the term, back in January and now I've basically forgotten everything he's said. I've probably got a piece of paper that he handed out somewhere and I might be able to reconstruct it, but .... [MULT1]

This corroborates findings from other knowledge work settings about the importance of introducing new skills and technologies during a "window of opportunity" (Tyre and Orlikowski, 1994). The digital library infrastructural arrangements for work production in literary theory were often minimal and sometimes substandard with respect to other disciplines. Next, I explain the predominant characteristics of work practices of the scholars.

Work production in literary theory included a set of very individualized work practices that resulted in the production of essays, arguments, presented papers, articles, book chapters or books about literature. Work production depended upon discourse, contemplation, writing, and a theoretical orientation with which to interpret the work. Literary theory did not build directly upon previous results in the way that scientists depended upon each other 's findings. Instead, they organized themselves according to their common interests based on theoretical approaches, arguments, or materials. Most literary theorists did not directly share inputs to work production, even during graduate training. Unless they shared works that were used in special library collections, literary theorists tended to collect what they need, work individually and share ideas through written or oral discourse.

Literary theorists described their work flow in terms of projects. The occasions for initiating a project were diverse: invited papers, tangents from ongoing projects, or the discovery of a new material. often, the first book literary theorists wrote was a revised and extended version of the dissertation. Typically, one work leads to an other helping them produce several articles, papers and book chapters related areas:

I was invited to contribute a book chapter ... on the strength of [my] book. So the project's been through a number of phases. I've written two papers based on it: one that I gave at the Shakespeare association on [topic] in the 18th century, and one that I gave at the Society for the [related topic] on publishing at the Library of Congress which focused on a little more on [narrower topic] per se. Then I actually gave an other lecture where it's much closer to the final form on - at a conference on [topic] at [Neighboring State University]. And using all that material, I submitted one draft of the article. In November I did an other draft, a revised draft right around the 8th of December... [DSULT1]

In an other case, the occasion for an initial work was the organization of a group of literary theorists to comment on a particular work. These occasions sometimes triggered extended discourse with other scholars on the same project.

[Author] had this essay which the editorial board found very interesting -provocative and invited a number of people to respond to it. And the idea was to make a special issue then in which they would have this essay and have this essay responses from these people. [MULT3]

Book projects differed in many ways from the projects producing articles. One common approach to publishing a book was for a literary theorist to select several previously published essays written on different occasions united by one or more common themes between them. Typically books included original work in addition to previously published articles or extensions to previous papers. other occasions for pursuing new themes came from invited lectures and discussions in graduate seminars. This example illustrates how a literary theorist drew upon his teaching experience to revitalize a 10 year old book project:

While I was reading for that, I discovered in fact, ...that I wasn't doing that at all. I was off on, in fact, an other book which grew out of that which is the one I've now finished ....I just got back my original proposal and I've been teaching a course for the second time that goes back to the original proposal. And I thought, how did I ever get around to this?... That's not how I'm teaching it anymore. [RULT3]

In other cases, scholars initiated work upon discovery of new material: a critical obscure book that provided rich material for exploration, access to a newly cataloged collection in a scholar's own university library, or a deeper exploration of a previously known book. Here is an example (from field notes) of the way a novel sparked a book project:

The piece on [Author] was started 2 years ago when she encountered a novels he hadn't read before... This was [Author]'s last novel which was a miserable failure, but experimental and more daring than novels nowadays. She said that this novel "took over my life." It was so compelling that she taught a graduate seminar around value of fiction, especially as expressed in this book. They used this book " as a prism to read everything else." [HULT1]

Typically, the source texts about which literary theorists write tended to be older than the materials used in the other disciplines. They also used browsing to examine journals in the library and journals they owned. Literary theorists were not in competition with each other for credit but nevertheless took pride in discovering little-known unique or rare texts. Reading as an activity was more central to disciplinary practice than in the other disciplines. In addition, some literary theorists were beginning to explore themes having to do with theoretical views of discourse via and about internetworked technologies (e.g.,electronic mail, electronic journals). There was a burgeoning number of gopher and WWW sites publicizing and archiving discussion lists and journals for literary theory.

Literary theorists typically worked on projects (especially books) over a number of years. Two years was a short time frame for a book as compared to ten years or more for some projects. Although most literary theorists tended to work intensely on one paper or chapter at a time, many projects were ongoing. The balance between reading and writing was more critical in this discipline than typically in the other three of this study:

In practice, since you usually have deadlines of one sort or an other for a lecture or a conference paper that you agreed to give so you are sort of reading and writing at the same time. You're writing some part which may not be designated as a particular part of the project that is part of it nonetheless. A talk for a conference or an essay. So you are doing both. And certainly that's one reason one accepts those obligations, is to keep yourself, keep writing. Right, because it would be easy to just go on reading forever, with out writing. [MULT1]

The careers of literary theorists tended to be quite closely tied to classroom responsibilities. Doctoral students supported themselves largely on teaching assistantships and it was not uncommon for literary theory graduate students to teach every term (especially in compulsory writing programs for undergraduates). Some doctoral students even found positions in other areas of the university based on other skills. Finding academic positions had become highly competitive in literary theory and most students could not recruit for tenure-track positions until they had completed their doctorates. Faculty advisors carried a larger number of graduate students than the other disciplines but work with them less frequently for joint projects.

Although most work was single-authored, literary theorists connected their work to other scholars in several ways. Several informants had created their own groups of colleagues with whom they shared their work, sometimes at workshops, journal editorial board meetings or private electronic mail discussion lists. Some informants found new technologies helpful to same precious money and time in gathering resources and communicating with colleagues. For example, this literary theorist used electronic mail to supplement her interaction with a valued colleague.

I think I sent a draft off to a friend of mine [at Canadian University],... and he made some suggestions. We exchanged work because he is doing something for the same collection. So we had a conversation about, [common topic of interest] So we exchanged sources that way .... He read my dissertation actually and we did similar kinds of works and we wrote and we e-mailed and we're seeing each other times at conferences. [DSULT1]

The main professional organization to which the literary theorists belonged was the Modern Language Association (MLA). They met yearly in late December at a large scholarly conference that hosted paper sessions and professional services (workshops and employment resources) for their members. Several informants preferred smaller conferences and workshops to the MLA conference:

I actually tend to avoid MLA - I like the smaller conferences: the Shakespeare Association or more specialized conferences rather than MLA if I can manage it .... I average about 2 a year. [DSULT1]

Work practices varied by individual, but typically literary theorists studied and wrote at home, while visiting other institutions, during summers and while on sabbatical. This literary theorist provided a typical example of how isolation from library materials and colleagues often spurred interest in trying new types of materials and technologies to connect with other scholars.

Next year, I'm going to be... living in a village in [Europe] and the electronic world is likely to be much more important to me .... I know very little about the resources that the local university library and the [university] so my plan is not to worry about it too much. I'll be there with [ the author he is writing about] and such notes as I have on my laptop and huge hard disk - just working. But I am ...trying to rely on e-mail for contact with people... I could imagine in certain sense in which I want to talk. [MULT1]

In summary, literary theory, as a discipline, drew on different theories to analyze literature and had construction of meaning as a central activity of work production. The production of papers involved personal contemplation and individual writing. However, scholars connected their work with other through written and oral discourse.

4.3.1 COMPARATIVE LITERATURE SUBSPECIALTY

The comparative study of literature is concerned with the relationships between literature and other arts and fields of knowledge. Its traditional emphasis has been on the systematic comparison of literary works from more than one country. This comparison may be made in the framework of a literary genre, of a period in literary history, or of dominant themes and motifs; or it can be undertaken in the context of the mutual impact of two national cultures or entire civilizations. (Stevens, 1996)

Comparative Literature informants further identified their specialties by the kind of texts they used ( often read in other languages) or by their incorporation of cultural theory into literary interpretation and criticism. They regularly worked abroad, sometimes for part of the year. Some of them published in foreign language journals.

Comparative literature informants were much more likely to use foreign language libraries or collections than the literary theorists who primarily worked with English. Work production other wise resembled the other specialties in terms of flow, individual orientation and outputs.

I had an [] fellowship for a different, I mean it was a [Author] project, but it had a different title and a slightly different focus back in 1987-88, I guess ....But it's now different, it's now rather different - it's focus than what it was then. But still .... that was still a year that I spent, I spent that year at [Prestige University]. I had been doing a lot of reading in the library ....And the bibliotheque national e, in 1990, I spent a lot of time reading earlier criticism of [author], the 19th century criticism and the early 20th century that I've never tackled. [MULT1]

Comparative literature informants relished unique opportunities to work with colleagues in different countries. Even when they didn't work abroad as frequently as they would have liked, they maintained contact with colleagues via correspondence and increasingly electronic mail. Sometimes projects arose from this correspondence. In this example, a comparative literature scholar described how her experience using e-mail during a political event provided material for an article about themes of location :

This is an article which began as a response to a theoretical discussion on [subspecialty] literature and the postcolonial situation. And so there was an original article and ... a full length article .... that was a response to it. And when I wrote the article it was about a year ago exactly when the [political event happened]. And so it became kind of a meditation on topics of location and the location that it specifically used was the location of e-mail. That is, what does it mean when the location is the net rather than specific geographical space. And so I wrote this article using [political event] as the example, incorporating a lot of stuff that was coming over the e-mail network. [MULT2]

Opportunities for writing figured as key factors in work production. Access to people and materials was also important but the amount and quality of time shaped that access.

[Mountaintop]'s holdings are actually quite good. It's just somehow when I'm here, I don't spend enough time in the library .... So even [FSU], if I may say so... I was more productive there with a less good collection than here, but with more time to get at it. [MULT1]

Comparative literature scholars were also more likely to span multiple subspecialties than literary theorists in other specialties who worked with one theory, genre or time period. As was the case in other disciplines in this study, when people worked in various subspecialties, they often sought different kinds of materials. This example illustrates an occasion for finding source materials on the WWW:

As an idea, of course it floats around in European philosophical circles rather than American circles .... So the French have established on the [world wide] web culture and history [home pages] where there is considerable bibliographical resources as well as some useful information on current work in architecture or architectural history, current exhibitions in Europe. [RSULT3]

Since the subspecialty of comparative literature was predicated upon comparing materials from at least two countries, scholars needed to be able to identify and select materials from a variety of sources in different languages.

The publications of comparative literature scholars were not homogeneous. The choice to study Judaic texts, classics, Hispanic literature or German culture influenced the outlets where informants preferred to publish. However, there were some common journals that most comparative literature scholars subscribed to or browsed occasionally. In terms of length, number of references these journals most resemble those in sociology. Both these journals and those in (Sociology) usually come out quarterly or bimonthly rather than every month(Computer Science) or every week (Molecular Biology).

At the time of this study, several new journals had been started which were distributed in electronic format (e.g., E-Journal and the Journal of Postmodern Culture). However, informants were still reading and publishing in predominantly traditional print forums.

Book publishing played a greater role in comparative literature and literary theory as a means of establishing one's contribution.

4.4 SOCIOLOGY

Sociologists study societies and social-scale behavior in a wide variety of settings. The report their research in articles and books. Sociological research concerned topics which had broad interest outside the specialty area, though the approach and conceptualization was not always easily understandable by a lay person. The following excerpt describes the way this issue influences the character of the discipline:

It is evident that sociology has not achieved triumphs comparable to those of the several older and more heavily supported sciences .... The true situation appears to be that in some parts of the discipline...there has in fact taken place a slow but accelerating accumulation of organized and tested knowledge. In some other [parts of the discipline] the expansion of the volume of literature has not appeared to have had this property ....Bias, in more than one direction, is sometimes presumed to be a chronic affliction of sociology. This may arise in part from the fact that the subject matter of sociology is familiar and important in the daily life of everyone, so that there exist many opportunities for the abundant variations in philosophical outlook and individual preferences to appear as irrational bias (Britannica Online, 1996b)

This excerpt indicates that the broad interest in sociological inquiry undercuts its legitimacy. In sharp contrast to computer science where external interests provided support for research, sociological inquiry had been to a large extent absorbed into other research disciplines where funding is less focused on social behavior and more focused on social behavior in a particular context. Sociology also lacked a common core of knowledge.

As a multiparadigmatic discipline, sociology encompassed several approaches for inquiry. This study included informants who used experimental paradigms, simulations, modeling, statistical analysis, historical and qualitative analysis. Work in sociology tended to be project-oriented, focusing on theoretical development or analysis of different kinds of data. Sociologists spoke of "getting papers out of " projects. In this example, a sociologist described a shift in his work from secondary to primary data analysis.

I think part of [ the shift in his work from analyzing secondary data to collecting his own data] is that if you have the opportunity to interview these people, it gives you more information. But the problem with that is that you have to have some funding to do that. [FSUSOC1]

Grants helped sociologists purchase data sets, collect data, analyze data, collect supplementary materials or collaborate with distant colleagues. Sociologists in this study varied in terms of grant funding. The majority of the 24 informants were not currently funded by grants. Several had grants of $25,000-60,000/year. A few sociologists had grants over $100,000/year for large or multiple projects.

Many of the sociologists collected their own data, but some acquired datasets from public or private sources. Some sociologists created computer programs for analyzing data, or creating simulations. Others worked with computers much as they would have 20 years ago, using statistical programs to computationally analyze large data sets, though in recent years the resources they needed had become faster and more affordable. Some sociologists explored interaction in networked computing environments. Others used their computing resources primarily to prepare papers, communicate with collea