Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
6 The Research Support System
:~: >~ ~ ~ at: : a:: a:: : a:: ~ :~: a:: : ,: :: : ~~!~133~l~ #~ /~s~s5~si~.. <7 |~ ~~> lSl?~S ~ LIZ? ~S~S~S~S~S~S~S~E~:~#SSSIS~Si, ... :,.,. .,: ~ S:S:S S.lS~lSlSISISIS~S~S~l ~S~ {~< <~SSSls>~) ~~...i ~ i A s S ~ S S S ~ s S i S [" " ~>~i#~s~Ls~s~s~s~ss~.) is.s66s.s.s.s.s.s.s~.< . #S~Sl~(S~: S? MASSES? ~'~S.S::'S'S:S:S'S'S' . ........ ' AS." .~''''~'~'~S~SISiS SSS>ISiS~#s>~S~s~s~{s {s~s~6 s A :ii~:SS;, ,.,._ {~!~ t~:s~^ s~sis~sis~s~s~ssSsssss? ~~< ~:~s<~:>~ (~> :s:~s/~ .is #.s.s.s<~;~1 S:i:S:S'S'S'S:S:~S~S~{S~
6 The Research Support S stem Previous chapters of this report focused on the substance of the behavioral and social sciences: What are the major questions and ideas that drive research and give shape to the fields? What advances are occurring in the methods by which new knowledge is discovered and validated? But some conditions affect opportunities for substantive progress across all of the topics, and so in this chapter we step back and consider the institutional context in which the work is conducted" the research support system. The elements of the system can be characterized by the resources on which research depends: · Human resources. People become researchers through programs of teaching, training, and professional certification in academic departments, affiliated or independent institutes, and professional associations. · Technological resources. These resources depend on programs of tech- nical support, maintenance, and procurement to ensure that researchers have high-quality, fully operational laboratories, field instruments, com- puters, specialized facilities, communications systems to facilitate knowl- edge and collaboration, and supplies. · Data resources. Of particular concern here are large-scale data sets, in- cluding those that are part of the federal statistical system; specially col- lected research data; and research-relevant record systems kept by other 203
204 / The Behavioral and Social Sciences private or public organizations principally for administrative, management, or intelligence purposes. Funding resources. The availability of public and private funds is deter- mined by the financial commitments of funding organizations and the procedures used by them to solicit, review, monitor, and coordinate ex- penditures through intramural and extramural budget allocations, grants, and contracts. These four resource categories provide a useful organizing framework, al- though the institutions that support research cannot be divided neatly by cat- egory. For example, colleges and universities consider teaching a primary re- sponsibility, but they also take responsibility for maintaining technological resources needed for research. Many academic institutions also maintain data archives, central computer facilities, or specialized research organizations, and many allocate funds to support faculty and student research. Similarly, the National Science Foundation is known principally as a research funding agency, but its Division of Science Resource Studies is a major collector, analyst, and source of data on scientific and engineering personnel, facilities, and expend- itures, covering federal and state agencies, other countries, and the private sector. And the National Institute of Mental Health not only funds a substantial amount of extramural research, but also houses some of the world's top labo- ratory facilities and scientific talent. This overlay of roles applies even to many rather specialized behavioral and social sciences institutions. For example, the Inter-university Consortium for Political and Social Research, headquartered at the Institute for Social Research at the University of Michigan but with nearly 300 member institutions, is a central archive for machine-readable research data, a center for training in quantitative techniques of analysis, and a source of technical computing as- sistance and software development. In short, institutions often play more than one role in supporting research. There are a number of opportunities for improving the research support system of the behavioral and social sciences. Each section in this chapter begins with an overview of the resource situation and then provides recommendations for change. The recommendations are not addressed to any one institution; they call for the cooperative and imaginative efforts of several. Even when a single focus seems apparent such as changes in the activities of government funding agencies- change is not simply a matter of deciding internally on new policy initiatives. Funding agencies rely very heavily on the research commu- nity to generate proposals for research and to provide critical technical evalu- ations, priority ratings, and program guidance. Those evaluations play a large role in determining what research is supported, at what level of effort and for what length of time, and how funds are allocated among individual projects, multipurpose equipment, collective data resources, and investments in human
The Research Support System / 20S capital. Moreover, government funding agencies need to justify their programs and budgets to policy makers in Congress and the executive branch, who are also responsive to the research community. Consequently, the analysis and recommendations in this chapter are as important to researchers as to admin- istrators, and addressed as much to universities and other research institutions as to foundations and other sponsors. HUMAN RESOURCES Measured by the number of PhDs (and equivalent research doctorates), the behavioral and social sciences are large. About 120,000 people in the United States hold doctoral degrees, and about 6,000 new PhDs are granted annually in these fields. These fields generally include anthropology, economics, geog- raphy, linguistics, political science, psychology, sociology, statistics, and closely related fields such as criminology and international relations. The numbers do not include the fields of history about S00 doctorates in 1986 or education, in which the subfields of educational psychology, educational statistics, and educational testing granted about S00 EdDs in 1986. This numerical strength does not translate into proportionate strength in research. Although behavioral and social scientists constitute 30 percent of all science and engineering doctorates, they constitute only 13 percent of those whose primary activity is research and development. On university and college campuses, behavioral and social scientists comprise 22 percent of all the full- time equivalent scientists and engineers, but only 8 percent of the 65,000 full- time equivalent positions devoted to research and development. This last per- centage is lower than it was a decade ago, in approximate proportion to a reduction in federal research support. But even at its highest point, campus strength in behavioral and social sciences research was well short of what one might expect on the basis of the numbers of trained personnel available. An explanation and a prescription for this persistent difference must begin by taking into account the balance between general education and research train- ing in the behavioral and social sciences in universities and colleges. Colleges and Graduate Schools Undergraduate Education The current structure of curricula, beginning in secondary school, leads to teaching demands that strongly affect the research productivity of talented young academic scientists in the behavioral and social sciences. We recommend that senior decision makers in universities and elsewhere support long-term programmatic and proce- dural changes to upgrade the level of behavioral and social sciences curricula in secondary schools and in colleges.
206 / The Behavioral and Social Sciences A very large number of college students take courses and major in the be- havioral and social sciences; about 110,000 bachelor's degrees are granted annually to behavioral and social sciences majors. Although substantially less than the peak of 14S,000 in 1974, the number is still approximately equal to the number of bachelor's degrees granted annually to physical, mathematical, and life sciences majors combined. The relative popularity of behavioral and social sciences undergraduate courses results in heavy teaching demands in these fields: more than one-fourth of doctoral psychologists and one-half of all other social scientists report that teaching, largely undergraduate instruction, · . . . . IS to heir prlnclpa . activity. Most students in introductory, lower-division courses in the behavioral and social sciences have had no opportunity to take precollege courses in these disciplines. They have gained a certain amount of knowledge in less systema- tized or differently organized courses in grade school social studies and in high school civics, economics, world geography, U.S. and world history, health and safety, and so on. But such courses seldom convey a sense of the scientific nature of inquiry, teach the theoretical foundations of knowledge, or provide basic lessons In research methods. This situation contrasts sharply with the experience of introductory students in such scientific herds as biology or physics: first, as a college requirement, many of them have taken specific disciplinary courses at the high school level; second, those courses include lessons in basic theory and hands-on experience with laboratory research methods. One consequence of this sharp difference is that high school students with talents and interests suitable for formal scientific or technical training are less likely to be aware of opportunities to pursue such intellectually challenging careers in the behavioral and social sciences. And in fact, students who major in the behavioral and social sciences are much less likely than students majoring in other scientific fields to continue on to graduate school. In addition, students in lower-division college courses in the behavioral and social sciences have usually done little or no work in mathematical statis- tics, decision theory, networks, or other mathematical subjects particularly relevant to the behavioral and social sciences. Past efforts in science education, sponsored by the National Science Foun- dation among others, have had much to do with upgrading high school cur- ricula in mathematics and the physical and life sciences. Virtually no such efforts have been undertaken at the federal level to upgrade precollege instruc- tion in the behavioral and social sciences, though some fledgling attempts have been made at the state level. Lower-division instruction in the behavioral and social sciences is, therefore, necessarily different in character from that in the physical and life sciences. In the behavioral and social sciences, instructors must first focus on defining the basic subject matter, recasting students' earlier, ad hoc knowledge into systems of disciplinary thought, and giving them rudimentary training in relevant re-
The Research Support System / 207 search methods. In the natural sciences, in contrast, introductory coupes can Ecus on deepening studeno' understanding of previously introduced theorem ica1 ideas and expanding their methodological skips. OveraH, greater demand ~ placed on behavioral and social sciences pro~s- . . .. . sors In unlverslues anc co flexes to prepare stuc ants or nonresearch careers. An appearance of similarhy in teaching loads (as measured, say, by number of students between bculdes oRen masks an important underlying difference: an imbalance between the number of students in lower- and upper-div~ion courses. At many universities, the physical and hk sciences facula teach red atively Ever courses, and most of the ones taught are either large introductory super or small laboratories or seminars. In the behavioral and social sciences, the lo~er-d~sion survey courses are smaller and upper-division seminars and practicum coupes are larger. The substanth1 difference between the numbers of students in lower- and upper-division courses avows intense professorial attention to students at the upper lever Instruction far physical and h~ science students at this level is oRen aimed at preparing them far scientific or technical careers. This contrast at upper undergraduate levels between the bculty-inten- shy and research orientation of natural science education and the broader intellectual net that ~ cast to serve the needs of less research-oriented under- graduate moon in the behavioral and social sciences may make the latter less attractive to many intellectually talented undergraduates. The Matures of the U.S. educational system that create these differences are not likely to change rapidly. Hence, the Caching burden on young behavior and social sconces professors can be expected to continue. Several of our recommendations are designed primardy to insulate more of the best young academic scientists Tom the systematic pressures that inhibit high-quali~, productive research. But in the long tea the upgrading of curriculum at the secondary school and undergraduate levee would pay large dividends in the quaff and quantity of research. Gr~Ju~fe [Juc~hon The research ~orfunides idenf~ed in [his rigor' ~re [hre"~ed ~ [~e 1~' ones since [~e miJ-l9~s in brow and of her Corer grow mark, es- peci~I~ "' rese"~-orienfeJ universides. ~ resummon f~1 ~ t~ accused and ricers ~rdruI"~d Oaf ~ "~erf"~en imme~i~fe~ to "Afro' Brewer pro- portions ~ rese~rc~-orienfeJ sfuJe"[s into gr"Ju"te programs in f~e Id and florid sciences "~] [o intense and ~~e [~e "verge IcveI ~ research Clinic in Ifs swami ~ rerommen~ $~D minion i" ne~nJi~r pre- orfor"I~s~s and ~"ini~g~n[~ This "moun'r~esenfs rIose to one~urf~ t~e nag human resource i"~esfments recommended in finis roars The most serious concern of graduate departments in the beha~ora1 and social sciences, especially those in whim training far research takes high prior-
208 / The Behavioral and Social Sciences TABLE 6-1 Support for Graduate Students in the Behavioral and Social Sciences at Doctorate-Granting Universities, 1975 and 1985 Number of Students Kind of Support 1975 1985 Research assistantships All sources 7,173 8,102 Federal sources 2,441 1,962 Fellowships and traineeships All sources 13,5S9 10,124 Federal sources 4,595 2,313 SOURCE: Data from National Science Foundation, Science Indicators: The 1985 Report, NSB85- 1 and Academic Science/Eng~neering: Graduate Enrollment and Support, Fall 1985, SRS87-D5. ity, is the increasing difficulty of recruiting and retaining talented graduate students with a commitment to research. Overall, graduate enrollments and the number of doctorates completed in the behavioral and social sciences have been fairly constant over the past decade, but the number of doctorates com- pleted in departments ranking in the top quartile of scientific quality (as in- dexed in a 1982 study by the Conference Board of Associated Research Coun- cils) declined by 17 percent between 1973 and 1983; this decline is more than twice the decline in the number of doctorates granted by the top quartile of all other science and engineering departments. While there are many reasons for the decrease in completed doctorates at top research departments, one major contribution is undoubtedly the retrench- ment of federal support (see Table 6-19. Between 1975 and 1985, federal support for graduate research assistants, fellows, and trainees decreased sig- nificantly. The decrease in support for assistants was more than compensated for by increases from academic or other sources (though the average value of stipends may have fallen), but this was not the case for fellows and trainees. Although federal support for them decreased substantially across all the fields of science, it was steepest, 2,250 or 49 percent of all positions, in the behavioral and social sciences the fields that were much more dependent than other sciences on that support. An additional 1,000 nonfederally funded graduate positions disappeared during the same years. The decline was especially evident in the most prominent research departments. For example, between 1975 and 1982, the total number of full-time graduate students receiving any kind of federal support in the top quartile of behavioral and social sciences departments declined by 53 percent. In other scientific fields, the comparable number in- creased by 15 percent. To support their graduate training, behavioral and social sciences graduate students now rely heavily on their own earnings, spouses' earnings, and loans (see Table 6-21. Moreover, the assistance that is available is skewed toward providing teaching assistants for lower-division instruction. From the perspec-
The Research Support System / 209 rive of research training, the most important point is that since there is limited financial assistance for actual research, the proportion of time devoted to de- veloping research skills and interests is similarly limited. As a result, completion of graduate training in the behavioral and social sciences is relatively slow, about 2 years longer than in natural sciences. There are also more part-time students, attrition is higher, and new PhDs are generally older. If present patterns of support continue, it will be difficult to sustain the base of talented and committed young scientists needed to keep the national research enterprise competitive and enable it to exploit the scientific opportunities dis- cussed in this report. It is especially important that high-quality graduate fel- lowships, traineeships, and assistantships once again become readily available in conjunction with faculty research and advanced training. Only by paying serious attention to the financial requirements of graduate training will the behavioral and social sciences stop losing potential research talent to clinical, business, legal, and other kinds of career training. In the short run, the most significant improvements will be achieved by increased support for predoctoral research fellowships and training centers. Support of graduate training in the form of national competitive fellowships has two functions: it provides incentives for individuals to enter the fields, and it provides information to undergraduate students about the kinds of training needed to enter those fields. Support of fellowship programs for graduate study in the behavioral and social sciences should be increased with the specific intent of encouraging more research-oriented students and those with rigorous undergraduate backgrounds including those with majors in the natural sci- ences or formal methodological disciplines, such as mathematics and logic, computer science, and statistics to undertake such study. The quality of graduate training can also be influenced significantly through the support of training programs that require students to go beyond the stand- TABLE 6-2 Sources of Support for Graduate Studies Reported by Doctoral Recipients in 1986, by Field (percentage) Source of Life Physical Behavioraland Support Sciences Sciences Engineering Social Sciences Own earnings 46 34 38 68 Spousal earnings 31 21 17 33 Loans 33 20 14 59 Teaching assistantships 40 70 43 57 Research assistantships 54 74 72 38 NOTE: Columns add to more than 100 percent due to multiple sources of support. SOURCE: Data from Summary Report 1986: Doctorate Recipients From United States Universities. Of lice of Scientific and Engineering Personnel, National Research Council. National Academy Press, 1987.
210 / The Behavioral and Social Sciences ard requirements of their disciplines, either across the fields of the behavioral and social sciences or in methodological disciplines. As noted elsewhere, uni- versity departments are organized primarily along disciplinary lines, and almost all graduate training takes place within disciplines. One of the costs of such departmental structures is that interdisciplinary research and training is overtly or covertly discouraged. The burden of proof as to the value of pursuing such interests falls on those who want to cross departmental boundaries. We believe that improvements in training and advances in research can be facilitated by crossing those boundaries. Universities, colleges, and support agencies should seek ways to reinforce activities that point toward promising areas of interdis- ciplinary and collaborative research. Graduate students should be encouraged to span departmental programs and participate in multidisciplinary research on campuses as freely as they undertake disciplinary activities. Postdoctoral Training and Collaboration Despite graduate-level financial support problems, a substantial number of students do eventually complete PhDs in the behavioral and social sciences. But it is neither easy nor typical for these new PhDs to enter research careers. In comparison with other fields of science, there are few entry-level behavioral and social sciences research positions available. In 1986, for the 5,700 new life science PhDs, there were about 3,750 postdoctoral research positions (fellow- ships, associateships) and entry-level jobs primarily devoted to research and development, about 66 prime research openings per 100 doctorates For the 8,200 new physical and engineering sciences PhDs in 1986, there were about 4,850 postdoctoral research positions and research and development jobs, about S9 openings per 100 doctorates. For the 5,8S0 new behavioral and social sciences PhDs in 1986, there were only about 1,600 such fellowships and jobs, about 27 openings per 100 doctorates. Adding jobs in which research and development is a secondary activity, the ratios of openings to new PhDs are 73, 68, and 43 per 100 for life science, physical and engineering sciences, and behavioral and social sciences, respectively. These differences parallel substantial differences in academic culture among new faculty at universities and colleges. As explained above, the teaching loads of assistant professors in the behavioral and social sciences are often heavier than in other scientific fields. Universities generally do not provide research funds or facilities for behavioral and social sciences faculty as routinely as they do for other new sciences faculty. Except in one or two fields, generally lower salaries put financial pressure on new faculty to supplement their standard 9- month teaching base pay. As a result, unless they can Earner immediate grant support, "spare" time must be devoted to summer teaching or similar employ- ment instead of to research. Thus, the most crippling constraint on research for most new (and mid- , ~ ~
The Research Support System / 211 career) scientists, for professors as we 11 as for those in other jobs, is a lack of time, and time is critical for research. Time must be found for writing proposals, organizing laboratory or data resources, thinking through experiments or an- alytical strategies, reading the relevant literatures, recruiting and training as- sistants, consulting with colleagues, and eventually, writing up results, inter- pretations, and theories. There is no single model of support to provide more time for research: released time during one or more academic years, summer support, and full-time leaves of absence for periods of weeks, months, or several years may all be appropriate in different instances. Other sources of support for research are advanced workshops away from a person's home institution, regional and national research centers, and traditional postdoctoral fellowships. The central point is that new PhDs as well as those early in their careers need to have access to a range of postdoctoral research opportunities that are not necessarily tied to specific grants for fully described projects. These op- portunities must be compatible with the career needs and research interest of young scientists. An increase in financial resources is needed, but an expansion of imaginative design of research possibilities is nearly as important as increas- ing the level of funds. Postdoctoral Fellowships and Traineeships We recommend an increase in the number of postdoctoral fellowships in the be- havioral and social sciences, to bring the availability of postdoctoral opportunities more closely in line with research needs and opportunities. Special attention should be given to those research areas for which advanced training in more than one discipline is essential. We recommend an aggregate annual increase of $18 million for postdoctoral support, which should be divided so that there is a balance between new training grants to institutions, individualfellowships at the junior level, and more advanced fellowships. In addition, the timing of award decisions should be changed: awards to potentialfellows, especially at the entry level, should be male early in the academic year to maximize their attractiveness. One of the most successful devices to increase training, accumulate experi- ence, and strengthen a scientist's research productivity is the position of the postdoctoral fellow or trainee. (As a rule, postdoctoral trainees are selected by the administrators of multiyear institutional training grants, and fellows are selected by direct application to the funding agency.) Fellowships are thus usually portable attached to an individual, not to a particular training pro- gram. Such a position may be for as long as three years, and in some cases leads to a second advanced degree. Present arrangements for strong postdoctoral programs in the behavioral and social sciences are unsatisfactory in at least two ways. The most obvious is that funds for postdoctoral training in the behavioral and social sciences are too small. Second, the timing of the cycle for application, evaluation, and award
212 / The Behavioral and Social Sconces of such Bullwhips ~ not always conducive to attracting the best candidates. Most colleges and universities extend oars for new instructors and assistant professors between Manual and ~arch: postdoctoral awards are t~icaDy oP bred later in the year. Many promising candidates therefore accept primal teaching positions that they might not have uken ~ an at~acOve postdoctoral fellowship had been available. ~ a result the candidate pool for fellowships reduced in numbed sad quaff. 4JvonceJ Training lnsf~ute~ ~ recommend nag Corer s~orf-~ur~fion "Jv~nced Minim insh'"les, gen- ~~1~ InBox' 3 '~ ~ Backs in t~e summer months. pundit Fancies should relict [heir rer~dvi~ [o such pr~osul~ which should 6e Rich co-~thive~ with c"~h other and sum individ""l~nves'~or pr~os~ ~ recommend initiation grimy 33 grooms per summer ~3 p~r~ic~"n~s catch, "' an shim ~r~e "nnu"1 cost ~ $7 million During the 1950s and I960s, numerous advanced gaining institutes (3 to 6 weeks in dufation) mere inidated: many centered on the teaching and re- Enement of mathemadcal and statical techniques useful in the behavioral and social sciences. Those who parOcipated~oth as teachers and students Fund these programs to be unusually successful. New methods mere brought much more rapidly into use, new lines of research developed, important results mere published, and people who mere otherwise isolated Tom pars of their or or closely related research herds mere brought electively and efRciently up-to-date because of the close interaction that takes place in such a sewing. Most of these acidities mere discontinued, not because the need far them diminished but because of changes in the mode of Ending. During the era in which they Rourished, special research Rinds mere available to institutions far such institutes. E~luadon of specific course proposal was conducted by the grantee institution, which had overaH responsibility far hosting and providing organizational support far the acidities, which mere usually held in the sum- mer. Thus, with a modest emu, it was possible far a group of researchers to develop a proposal far an institute and have ~ evaluated promptly and in co~edtion only with similar actuates. When the special hands mere te~i- nated, these proposals had to compete directly Ash all other research pro- posals. As a result researched had to Urge tar more elaborate justifications to confirm to standard invesUgator~e proposal fists and to visit longer fir decisions through the centralized review system. They mere open disappointed: when proposal evaluators confront direct choices between individual~nvesd- gstor projects sad those that involve collective and necessary more complex acting, the Aver are usually favored. In this instance, a ship in the structure of Ending led to the virtual abandonment of a very useful set of activities in needy all behavioral sod social sciences research cress.
The Research Support System / 213 The value of advanced training institutes justifies their resumption. There are many examples of areas in which training institutes would be very valuable: innovations in statistical methodology, some of which are powerful but not widely used; theories and applications of measurement and scaling; theoretical linkages across disciplinary boundaries, such as the interface between the neu- rosciences and cognitive sciences, decision making under risk and uncertainty, and the history and sociology of modern science (including the behavioral and social sciences); demographic and sociological aspects-of the life cycle; new biological assay techniques; and advances in computer simulation models and their applications. Collaboration and Communication We recommend a special initiative to facilitate greater communication and collab- oration among dispersed scholars working on substantively related research. We recommend that 45 new workshop programs be initiated, at an average cost of about $200,000 for a total of about $9 million per year in new funds. Moreover, research proposals that include requests to fund collaborative efforts, if recom- mended for award through competitive evaluation, should be provided sufficient funds to ensure that their collaborative features are retained. Behavioral and social sciences research is characterized by decentralization both within universities and geographically across the country. Most academic departments, seeking intellectual breadth and balance to ensure adequacy in the undergraduate and graduate curriculum, have only one or a small number of faculty members with expertise in any given specialized line of inquiry. To achieve a localized intellectual critical mass is therefore difficult. One way to do so is to create research centers that can facilitate communication and col- laboration among scientists, which is discussed later in this chapter. Yet less expensive and more flexible means than organized research centers are often sufficient to strengthen communication and collaboration among investigators who need to regularly exchange research ideas and methodological knowledge with colleagues located at other institutions. The need for exchange calls for a well-developed, effective system of national and cross-national research networks to bring people together regularly or on occasion, for periods of 1 to 6 weeks. A few such arrangements do exist in some areas of the behavioral and social sciences, but these hardly constitute a system. Travel grant funds that in the past facilitated such collaboration in one or another field for some period have been drastically reduced in recent years. When the costs for a proposal are scrutinized, travel funds are often the first point of attack in trying to reduce costs. This practice of treating travel funds as expendable is counterproductive to advancing needed collaborative research. Other mechanisms for interdisciplinary communication, advanced training, and collaboration, such as specialized institutes, conferences, and individual
214 / The Behavioral and Social Sciences travel grants, have also come under severe budgetary pressure. Continued progress at many research frontiers stands in need of new and renewed com- mitment to the promotion of intradisciplinary and interdisciplinary commu- nication, training, and collaboration beyond the confines of home departments. The need for collaborative working arrangements has several aspects: pro- grams of several investigators working together on a joint project; workers on mutually relevant but not directly related projects keeping closely apprised of and contributing collegially to each other's work; research scientists in different countries coordinating, comparing, and advancing their theoretical and em- pirical work. The mechanism of intensive research workshops, in which groups of active investigators gather, usually for 1 to 2 weeks during the academic year and 2 to 4 weeks in the summer, has been very fruitful in the few behavioral and social sciences areas in which it has been used. Such programs should be intended to run for a minimum of 5 years and to be open to competitive renewal. TECHNOLOGICAL RESOURCES The behavioral and social sciences have long made use of certain kinds of scientific and technical equipment, including computers, behavioral and neu- ropsychological sensors, and sound and picture recording and display devices. However, a substantial part of this technological force has been weakened in the past decade. The inability to meet material and technological needs, par- ticularly to replace old or obsolete equipment, is now a major bottleneck on the rate of scientific progress in some research areas. Particularly notable in all areas is the scarcity of powerful workstations and superminicomputers, which greatly exceed the capacities of personal computers but can readily be used as desktop instruments for departments or medium-sized research groups. Recent figures regarding technological renewal in the behavioral and social sciences provide little encouragement (see Table 6-3~. During fiscal 1981- 1985, federal expenditures for research equipment in psychology increased 39 percent and nonfederal expenditures in psychology increased 47 percent. But these increases were below the overall 58 percent increase for all the sciences; moreover, expenditures for social science research equipment showed no net increase. Altogether, only 2.7 percent of the total $655 million spent for re- search equipment on campuses in fiscal 198S was for psychology and the social sciences, down from 3.7 percent in 1981. There is a persisting view that behavioral and social sciences research can operate as a virtually equipment-free enterprise, a view that is completely out of date for research in many areas. Some examples of equipment needs dis- cussed in prior chapters include special video laboratories for the experimental study of human communication, interaction, and economic exchange; neu-
The Research Support System / 215 TABLE 6-3 Expenditures for Research Equipment at Universities and Colleges ($ million:) Science end Year Total Federal Nonfederal Psychology 1981 6.0 4.4 1.6 1982 5.7 4.2 1.5 1983 6.6 4.8 1.9 1984 7.0 5.0 2.0 1985 8.5 6.2 2.3 Social sciences 1981 9.2 4.0 5.2 1982 7.1 2.9 4.2 1983 9.0 2.9 6.0 1984 12.7 3.5 9.2 1985 9.3 4.2 5.1 All sciences 1981 414.0 265.0 148.7 1984 518.1 335.5 182.6 1985 654.8 427.2 227.7 SOURCE: Data from National Science Foundation, Science Indicators: The 1985 Report, NSB 85-1, and Academic Science/Engineering: R&D Funds' Fiscal Year 1985. reimaging technology to identify structural features and metabolically active regions in the living brain; and equipment to analyze satellite sensing of geo- graphical features. There has been a tendency to neglect these and other tech- nological needs of the behavioral and social sciences, even to the point of sometimes not including them from in national surveys of equipment resource needs in scientific facilities. Detailed analytical assessments of equipment needs in the behavioral and social sciences and year-to-year planning to meet the needs discerned are urgently needed. Computers We recommend $22 million annually in new funds to purchase and support im- proved computational technology. This amount is almost one-half of the $51 million total recommended in this reportfor technological resources. Centralized assistance in preparing, documenting, and providing software suited to behavioral and social sciences problems is needed; more particularly, programmers working in or trained at centralized computer facilities must tee familiar with such special software li- braries, needs, and problems. Access to advanced computingiacilities by behavioral and social scientists should be improved. Behavioral and social scientists need to be involved directly in the planning, assessment, and development of new super- computerfacilities. The behavioral and social sciences have a long history of using electronic computational devices for experimental control, data collection, data analysis,
216 / The Behavioral and Social Sciences and designing and testing models of social and individual behavior. Behavioral and social sciences researchers have also had leading roles in the design of user-friendly computer systems and special computer tools to aid human- computer interaction. The shift from mainframe computers to powerful desk- top machines, including new generations of very powerful workstations, has opened entirely new possibilities for rapid, real-time data analysis, sophisti- cated model building, and networking. The possibilities opened by electronic communication networks have barely been tapped and will require much greater future investments in suitable technology. Although funding for the purchase of computing hardware or the development or purchase of software tailored for behavioral and social sciences is growing, the need far exceeds the available funds. The currently available computer hardware and software are already inad- equate to meet the analytic, management, and retrieval needs of some of the larger survey data collections. As these data collections continue to grow, the adequacy of current computer facilities will be still further strained. The de- velopment of supercomputers is a possible solution to meeting the require- ments of users of the very largest data bases, but it is by no means certain. As development is currently planned, supercomputers will have the capacity to carry out very large numbers of calculations at very high speeds, which have great advantages in meeting the mathematical needs of complex iterative mod- eling, but they may not have capacities to manipulate large quantities of data efficiently (input-output capacities) and thus to handle analyses of large-scale data bases. Behavioral and social sciences research is straining the limits of present computational resources in several other areas, such as: the theoretical simu- lation of complex systems of simultaneous processes, for example, computa- tional models of the visual system, brain functioning at both the cognitive and neurophysiological levels, or human-computer interactions; and the empirical estimation, from even moderately large data sets, of the coefficients of complex systems of equations used in psychometric and test theory, scaling procedures, and models of national and international economies. Statistical covariance structure models, log-linear models of categorical data, nonmetric multidi- mensional scaling, and special clustering and classification models are suffering from computational overload with present facilities. Greater use of advanced workstations and supercomputers would also encourage better analyses of nonnormal multivariate distributions in nonexperimental data by such pow- erful techniques as structural equations, canonical correlation, principal com- ponents, and discriminant functions. Use of these techniques for nonnormal data requires computations several magnitudes more extensive than those needed for normal models. Currently, the natural scale of many scientific problems must be reduced to fit available computing machinery, in one of two ways. The investigator can
The Research Support System / 217 restrict the analysis to a smaller or less precisely specified aspect of the system under study, but thereby lose power to discriminate between alternative the- ories. Or the investigator can subdivide the problem into pieces that can be separately carried out on a conventional computer, but thereby lose the ability to optimize an overall solution or work with any rapidity. A better solution would be possible if the investigator could gain access to Class VI supercom- puters, such as the Cray-1 and Cyber 20S, or to use the capabilities of con- nectionist (non-von Neumann) architectures, such as the new large-scale par- allel-processing computers. Parallel architectures appear to be valuable for many applications in cognitive science, including connectionist models, com- putational linguistics, and the study of problem solving and theorem proving, and in macroeconomics for the simulation and analysis of multicountry and global dynamic models. An issue within universities is the allocation of mainframe computer costs. Decisions on how costs are allocated for central processor time, blocks of memory, and input/output channels can significantly affect computer use. For example, many university computer centers have high charges for memory storage and input/output channel use, which means that users whose work involves transformation or combination of large data bases pay far more to run their jobs than users with small data bases but using complex equation systems or iterative algorithms that require a great deal of straight computation. Yet there is little difference between the two jobs in terms of incremental cost to the facility. With grant sizes decreasing and computer costs increasing, re- searchers whose needs happen to be high on the wrong index of mainframe user-cost accounting may have inadequate resources to do high-quality work. The solution to this problem is to revise computer user-cost systems that create inadvertent cross-subsidies. Visual perception research has an especially rich history of advances corre- lated with progress in computer graphic capabilities. Important research ques- tions can best be addressed with experiments involving the rapid generation of medium- to high-resolution naturalistic images. Prerequisites to effective computer use in visual perception include wide diffusion of new software resources, appropriate local hardware (such as wide bandwidth communica- tion devices, graphics workstations, and image processors), and collaboration between visual science and computer graphics specialists. In addition, any field in which new computer models or methods are being developed needs rapid, cost-efficient ways to transmit and test programs and procedures among dif- ferent sites. Sharing of programs is crucial in order to permit scientists to experiment with each other's theoretical tools; appropriate networks are essen- tial. In summary, the importance of advanced scientific computing to progress on many behavioral and social science research problems needs to be recog- nized. Support must be available for use of Class VI machines and other ar-
218 / The Behavioral and Social Sciences chitectures, symbolic and parallel-processing capabilities, high-capacity com- munication channels, scientific computing networks, and local needs including new workstations and software development. Neuroimaging Devices We recommend an initiative to ensure access to neuroimaging devices for nonclin- ical researchers studying language, memory, or motivational phenomena. We rec- ommend that $4 million be provided annuallyfor this purpose. Some of the neurolinguists, neurobiologists, and psychologists conducting work on language, memory, and motivation require access to the new gener- ation of relatively noninvasive neuroimaging devices, such as computerized tomography (CT), magnetic resonance (NMR or MRI), and single-photon and positron-emission tomography (SPET and PET). Since these devices are now usually installed in clinical medical settings, gaining access to them by non- clinical or nonmedical researchers is difficult. Yet such devices are far too expensive for nonclinical investigators to acquire and operate strictly for re- search purposes. Animal Care We recommend an increase of $5 million annually to improvefacilitiesfor animals used in behavioral and social sciences research. For researchers using animals, it is essential that funds be available to ensure that the animals receive proper and humane care. Mandatory standards for animal care have risen greatly in recent years. Investigators who use animals in research have generally applauded the change, not only for ethical reasons, but also for practical ones; most research projects benefit from the increased health, longevity, and well-being of their animal subjects. Improvements in facilities for animals are very expensive, however, and little provision has been made to enable researchers to afford the costs of meeting the newly mandated standards. DATA RESOURCES A number of large-scale funded data bases serve behavioral and social sci- ences research. Some of these data bases entail costs comparable to capital pieces of scientific equipment. For example, the 1980 decennial census of population and housing in the United States cost more than $1.1 billion to plan, collect, and process. Of course, most of the cost of the census is to meet constitutional and statutory requirements for information about all of the peo- ple and places (not just representative samples) in the United States. The re-
The Research Support System / 219 search uses of the decennial census, which are very important, are virtually free riders on this massive exercise. (It should be noted that the research benefits of the decennial census are not captured until the census is published. It took up to 6 years after the 1980 census was taken to publish or release many of the detailed statistical tables and data tapes that are of use to researchers. Planning for the 1990 census should stress issuance of research data within 2 to 4 years.) A more typical large-scale sample survey designed to test scientific theories or to illuminate certain empirical processes costs far less than $1 billion. The World Fertility Survey, for example, developed data on family size, family planning, and fertility rates in 62 countries at a cost of roughly $S0 million during 13 years. The continuing Panel Study on Income Dynamics, covering the changing fortunes since 1968 of (originally about S,000) American families, costs about $2 million a year to update, analyze, and make available to re- searchers. There are three matters of particular importance with regard to data re- sources. The first matter is the implementation of appropriate criteria in gen- erating new large-scale research data collections, especially longitudinal ones. The second concern is the maintenance and enhancement of research access to governmental microdata files and record systems. A final important set of data resources that need attention are the records of local governments and large corporations. Large-Scale Data Bases We recommend a two-trac1: review process for all large-scale data collection pro- posals, including the $40 million in new collections recommended in this report: a technical review of the scientific significance of any such study, carried out in the same context as investigator-initiated proposals of smaller scale, but without direct reference to budgetary concerns; and, in parallel, an evaluation in terms of the special design features needed, for which there should be distinct review criteria, uniformfor all large-scale collections. One of the trends in the behavioral and social sciences during the past several decades has been a spectacular increase in the size of many data bases used in research. Large-scale longitudinal studies in particular have facilitated major advances in knowledge concerning psychological development (Chapter 2), the political behavior of electorates and legislatures (Chapter 3), and population dynamics (Chapter 45. In the future, new large-scale survey instruments will be needed to promote research advances in two major areas, human devel- opment (including cognitive, emotional, social, and health-behavioral factors) and the organization of work. The foreseeable need is for more, larger, and more complex data collections, including the development and maintenance of technical and administrative support facilities.
220 / The Behavioral and Social Sciences It is sometimes argued that large-scale data bases are involved in a "big science versus little science" contest, but the human scale of most analytic projects utilizing large-scale data collections is not different from other kinds of projects: it is virtually always a matter of individuals or small groups of close collaborators. The question is not whether one big team will displace many little teams, but how many small analytic teams can rely on any single, large data collection. The relevant competition for resources is really between the researchers who can share a data base and those whose research cannot benefit from such sharing. Large-scale investments present unusual problems in deciding which proj- ects to support. What mixture of data sets with how much replication among them will best serve researchers in different fields? Under what circumstances is one comprehensive large-scale effort preferable to several smaller projects? When do the particular scientific advantages of panel studies relative, for example, to repeated cross-sectional surveys or matched-comparison samples justify the particular kind of long-term commitment attached to them? Under what circumstances can randomized field experiments be effectively coupled with longitudinal surveys? Foremost among the difficulties is that the value of data cannot be assessed separately from an assessment of the questions different researchers wish to answer; the skills and analytic competencies of the principal investigators; and technical matters of optimal sample size, sample population, adequacy of the survey instruments to be used, and other methodological Issues. Prospective or longitudinal panel studies, which collect research data on a substantial sample of respondents over a period of years, provide a particularly valuable basis for testing causal connections for example, in the development and course of criminal careers (Chapter 2) that are difficult to find or confirm in short-term or cross-sectional studies. But longitudinal studies cannot be undertaken lightly; they require a long-term commitment, sometimes up to 30 years or more, to ensure that their advantages are fully exploited. Before such a study begins, significant users must conclude that it is of high potential value, a research organization must exist and be committed to keeping the study functioning, and funding sources must be committed to long-term support. Such studies are especially sensitive to interruption. Postponing or ceasing operations for a few years can ruin the fundamental scientific rationale for the existence of a longitudinal study: the systematic and continuous accumulation of useful data over time. Therefore, once such a study is established, it should be discontinued only after determination on the basis of specific criteria and a well-defined review that its maintenance is no longer the best way to commit long-term resources. Two criteria should be key in such a determination: Has the field moved sufficiently beyond the questions the data set was designed to answer? Have sample loss and inadequate replacement led to excessive drift from statistical representativeness?
The Research Support System / 221 Behavioral and social scientists have developed a series of criteria to evaluate new and continuing proposals for longitudinal surveys and other large~scale, long-term commitments. One set of criteria concerns the survey's methodo- logical contributions, as discussed in Chapter 5. A second set is the substantive and theoretical importance of the research questions that the study is designed to address. Other criteria derive from seeing how large-scale data collections have fostered the development of more effective research organizations. The most important institutional innovations are new forms of data dissemination and new modes of scientific decision making (or intellectual governance) within the research community. National and local data libraries designed to disseminate large-scale survey data for research purposes grew rapidly in the 1960s and 1970s. Their initial role was to serve as data archives, creating computerized files of data such as successive public opinion polls, county-level election returns, census infor- mation, and congressional roll-call votes. Present archiving initiatives look toward providing effective remote access as well as automated indexing, cata- loguing, and search capabilities to aid researchers in identifying data for specific research interests. These facilities also encourage behavioral and social scientists to combine data and conduct temporal and geographical comparisons that go beyond any one data collection. A major constraint on making archived data easily and quickly available is cost. The development of computer networks and the capacity for remote access is very promising on this score, and special-purpose capacities can be developed where needed (such as the University of Wisconsin SIPP-Access Center, to facilitate research using the Survey of Income and Program Participation con- ducted by the Bureau of the Census). Access has also been facilitated by re- quiring grantees and contractors to generate well-documented and readily ac- cessible public-use tapes and by providing resources for them to screen and respond to requests for additional data or documentation. But for the most part, new data collection efforts are likely to depend heavily on existing archives that are proficient in providing documentation, formatting and organizing data files, responding to diverse technical specifications, and using the expertise in data management that is essential for distributing data to researchers in different institutions with different data-processing and computational facilities. Important as it is, the archive function occurs at a late stage of the data collection process. Consequently, explicit arrangements have begun to be de- veloped for taking potential research users' interests and needs into account much earlier in the design and operation of large-scale longitudinal studies. These arrangements have included appointing a national board of overseers to these projects as trustees for the relevant communities of users. Such boards have been appointed for each of the three largest research data collections that are national in scope and support but conducted outside the federal govern- ment: the National Election Studies, the Panel Study of Income Dynamics (both
222 / The Behavioral and Social Sciences at Michigan's Institute for Social Research), and the General Social Survey (housed at the National Opinion Research Center at the University of Chicago). These boards of overseers have actively sought the participation of many re- searchers in user conferences, pilot studies, and ongoing design. More signif- icant are the new capacities and responsibilities developed by the overseers, principal investigators, host institutions, and permanent professional research staff in day-to-day charge of these data collection efforts. Host institutions, while benefiting from identification with major data col- lections, also accept organizational responsibility for maintaining intellectual leadership, appropriate facilities, and administrative and technical personnel. This responsibility is especially important in the case of longitudinal studies, as principal investigators retire, change career interests, or otherwise reduce their commitments and as significant changes in study content, measurement technique, research design, or collection methods accentuate the importance of staff methodological and technical expertise. Such projects thus engage their host institutions, far more deeply than do most smaller-scale projects, in the health of the research enterprise at a national level. To meet these national responsibilities, the institutional mechanisms of data archiving and collective intellectual governance of large-scale data collections should be further expanded. First, they should try to encourage simultaneous independent analyses of all data that have a bearing not only on important scientific questions, but also on matters of public policy. Independent inspec- tions, analyses, and criticisms of important new data are needed to inform public debate. The value of these procedures is evident in previous work with surveys of educational achievement, studies of income maintenance schemes, and experiments with alternative patterns of police patrol. Institutional mechanisms should facilitate collection of data in a manner that permits easy linkage to other data (see below, "Research Access to Government Data"~. The maintenance of an ongoing survey should also be viewed as a core activity to which independently designed but fully compatible special-purpose studies may be added on for limited durations, for example, for one or two waves of a longitudinal survey. Such studies could be methodological or sub- stantive (see Chapter S). Finally, periodic evaluations should be made of data- sharing strategies, accession and acquisition costs, and the scientific and prac- tical results of the data collection. These evaluations should be published: they can serve as a basis not only for evaluating the projects themselves, but also for advances in knowledge about how to evaluate and improve all such large- scale data collections. Research Access to Government Data We recommend that federal statistical agencies and programs work toward more access to federal data bases by behavioral and social sciences researchers. Risks of disclosure need to be balanced against the prospective research benefits (including
The Research Support System / 223 Me pr"~tic"1 ire ~~di~ 'h~' ~~ resultpom "~s '~ the dam. nags 'h", past research "guess to maw priviI~d, co-~en~i~t and Enormous " sags in the awed awes has no' produced ~ sitar c=e ~ harm involving ~ breach ~ the anosmia, age, or privet ~ "~ re~onJen~ ~ recommend $8 minion annuls ~ series Anew iniN"tives in rcse"rc~ "cress fo6~er"1 iscusseJ in curlier chafers ~ this r~orO. Much research in the behaviomI and social sciences benefits Tom data col- lected, processed, and made accessible by the faders statistical system. Re- search on labor markets social mobility, public finance, criminal justice, and other areas relies very heavily on federal surveys and sdminist~dve statistics. These data are not usually collected far the sole or even principal purpose of advancing scientific kno~Iedge, but Esther to develop information relevant to the missions of the agencies coHechng them. Yet the products of federal data systems are oRen vital to advancing basic research. In the absence of these statistical records, the sheer effort needed to design and collect this kind of information de nova would drastically increase the direct costs of research. In many cases, potentially valuable data are not svailshle in Ally usable An outstanding instance ~ decennial census cats. The Census Bureau did not begin producing public-use microdsts sample ales untie 1960, arid so immensely important in~rmadon on the massive social changes that had uken place in this county in this century sass not svailshle. A large, joint govern- ment-scademic project involving many university scientists has resulted to date in the production of publ~-use sample hiss for the 1900, 1910, 1940, and 1950 censuses. This new data resource greatly increases the informational psyoF Tom these very large national investments in censuses. These data are already proving important to the development of model of populadon change . . . unc fir vs~lug social snc . economic clrcumsunces. The National Sconce Foundation and other agencies have been prominent in the effort to enhance the research usefulness of federal stst~tical data and to bchitste access to them. To help make national and inte~stionsl data more adequate sc~ntiRcaLy, the Nsdonsl Sconce Foundation and the [ensue Bureau are providing support to an American Stst~tical Association Ship pro- grsm aimed at Usher upgrading the study potent of current Census Bureau data products and making them more directly usable by the research com- munity. Programs of on-she study at federal agencies of which the American Statistical Association research Allows program is s mode~rovide an op- portunity far researchers to analyze cats that may be too expensive to prepare far public release. These programs also increase the opportunities far dsts- collecting agencies to evaluate the research importance of requests to include particular critical hems on repeatedly administered survey instruments. The Edemas stst~tical system, when ~ ~ able to serve both agency missions and research needs, is s very efhcient resource. Moreover, researchers over the years have conthhuted substantially to the technical improvement, enrichment,
224 / The Behavioral and Social Sciences and utility of mission-oriented federal statistics. Yet research access is increas- ingly jeopardized by concerns about anonymity, confidentiality, and privacy rights despite the fact that researchers using survey-type data are hardly ever interested in discovering the individual identities of respondents and not at all interested in revealing such identities or using them for nonresearch purposes. Nor have researchers in fact ever been accused of improper disclosure or use. The most intense concerns about the privacy or confidentiality of records actually refer not to research use but to the routine use by state and federal agencies of each other's automated records for administrative, investigative, and intelligence purposes. These practices were recently reviewed in a report by the Office of Technology Assessment, Federal Government Information Tech- nology: Electronic Record Systems and Individual Privacy (19863. This report noted that "the widespread use of computerized data bases, electronic record searches and matches, and computer networking is leading rapidly to the creation of a de facto national data base containing personal information on most Americans" (p. 35. Virtually all the activities reviewed involve access to and study of files that completely enumerate a given population and provide exact identifiers. Research uses, in contrast, involve taking a small proportion of a set of files, generally, a structured random sample, either ignoring or actually deleting from the selected records any individual identifying data before the analysis begins. Other concerns involve more sophisticated technology. As government agen- cies and a variety of private companies (such as insurance clearinghouses, direct-mail advertisers, and credit bureaus) amass more extensive and detailed data files, the possibility increases that someone with a large-scale record system in which identities are known could use computer-matching software to iden- tify individual respondents even in a set of anonymous microdata, that is, in files containing individual records from which all identifiers (such as names, addresses, and social security numbers) have been removed. The accumulation of large and detailed microdata files both in government and in business may make the "signatures" of unusual individual records easier to discern through matching techniques. Federal data-base administrators are becoming very concerned about their own liability in this connection. Agency staff who have been working with researchers for many years perceive increasing pressure to reduce the amount and nature of data they release, despite the fact that the need for these data in research is expanding and despite existing practices that guard against un- wanted disclosure. These practices include releasing anonymous data in which geographical location is masked or the minimal identifiable geographic unit is no fewer than 100,000 residents; combining detailed items into composites; truncating information about such items as age or family income into broad categories; and withholding certain details about the sample design. When the problems of identification are especially severe- such as with data concerning
The Research Support System / 225 very large business firms or very wealthy individuals, because there are rela- tively few of them scientists have developed and used the technique of mi- croaggregation for the limited release of confidential data. In microaggregation a series of"average" firms are created from the data, and research files are produced on these average units rather than on actual individual firms. Be- havioral and social scientists have been in the forefront of the development of such techniques. Moreover, research users can be placed under legal obligation (and associated sanctions) not to make or permit nonresearch use of protected data to which they may gain authorized access. The liability for any breach then lies with research users, greatly reducing the liability exposure of the agency that is the source of the data. Despite such safeguards, there have recently been instances of denial of research access to federal statistics that were previously made available in sam- ple form, for example, the Continuous Work History Sample from the Internal Revenue Service. Researchers are concerned that such denials do not become a trend. An overly stringent interpretation of acceptable disclosure risk could drastically decrease present levels of access to some very important data, at a time when austere statistical budgets already threaten established standards for data preparation, publication, and dissemination. Significant research advances in several areas are likely to depend critically on the availability of large-scale sample microdata bases that can be created by merging administrative records with sample survey results. Such merging of records on a sample basis for research use has encouraging precedents. For example, an earlier effort linked a sample of tax files from the Internal Revenue Service, data from the Current Population Survey, and selected Social Security files for the year 1973 and rendered them suitable, including suitably anony- mous, for research. A good candidate for producing new research advances is the proposed linkage of Social Security and tax data to information from the nearly 30,000 households sampled in the longitudinal Survey of Income and Program Participation now being conducted by the Census Bureau; such link- age would provide an unparalleled resource for studying the microdynamics of household economic behavior. Similarly, the availability of geographically disaggregated microdata on individuals, firms, and households would im- measurably benefit research on migration, regional growth, and urban ecology. This discussion has focused on access to quantitative survey-type data, but there are other critically important types of governmental data that, for reasons of confidentiality, security concerns, and lack of funding, have not been de- veloped as research resources. Development of some or better research access to such data- whose costs of collection have been and will continue to be covered by their mission uses is among the most cost-effective research ex- penditures that could be made in any area of behavioral and social sciences.
226 / The Behavioral and Social Sconces Co~orale and [oca1 Goverumen1 ArchKes ~ recommend $2 miNion ~ berm "~ studies ,~ determine 'he research voice, me'hodol~"I problems, ~~d colas So "~ Clinic i~roveJ "cress to moor private record centers and unused Concorde and HI gove~men, "~hives jar research purposes. In contrast to most ordeal agencies, many organizations that collect and store potentially useful data on social and behavioral processes have no history of making them svailsEle ~r research. Principal examples here are most of the bias of major corporations and the archives of local governments: in most cases, the data are paper ales that are stored in warehouses. There are three stages in gaining access to such data. The heat ~ to keep co~oradons and local governments Tom destroying or throwing away their records without regard to their potential Clue far research. The second is to gain permission Tom their custodians to study them. The third ~ to pace such records or samples of them, as appropriste,into computer-readable form and twin technical suE or develop appropriate documentation to assist potential users. In some cases these stages may involve significant costs, and the relation between potential research benefits sod costs of preservation, s~ilshiLty, and conversion must be carefully considered. In many instances the benefits may outweigh the costs, but detailed plans and criteria far archival practices have not been developed. To increase access to research-relevant dsu far basic studies as well as to evaluate Ederal programs, nongovernmental archives with large record systems should be considered far receipt of research hands. One deskstle easy to in- cresse the usefulness of data archives~ublic and privat~ould be to pro- vide them with trained star or statistical analysis. Capacides far producing properly protected data tales, or periling stst~tical analyses when it is im- possible or too costly to protect the ales adequately far release to researchers, Could improve the range of potential resources at relatively low COSL ~uld- variste analysis open requires little more than covarisnce matrices far a sample and far some important subpopuladons, a level of a~regadon that surely protects the privacy of individual records. The data held by Blue Cross/Blue Shield and other carders of medical insurance, major automobile and lid in- surance companies, sad the like could be made more accessible to research uses ~ each major record center had stsE Ash research-oriented stst~tical and programming expertise. With such capachies, data archives could then respond to requests far analysis of their holdings either by releasing appropriately pro- tected cats or by performing some (or aL) of the statistical analyses requested by the researcher, who could be direcdy charged far some Cor sH) of the costs involved.
The Research Support System / 227 FUNDING RESOURCES In light of the unpredictability of research, there is inevitably a need for patience and adaptability on the part of those who produce scientific work and those who underwrite its costs. The ideal is to maintain a funding system that is fiscally responsible, flexible, and responsive to new developments. It should encourage competition among new ideas, sustain productive investments, and close off research avenues that are no longer productive. It is impressive that present funding arrangements manifest most of these ideal components most of the time. The changes needed and recommended in this chapter, while important, should be regarded as improvements at the margin of a basically sound system of support. Probably the most vexing problem facing researchers and the agencies that fund them is how to decide between funding individual scientists to conduct highly specific, time-limited projects and providing support for more extended, less specific research designs and structures. The latter include arrangements for shared access to expensive technologically advanced equipment, groups of investigators working together on difficult multidisciplinary problems, and longitudinal studies. Such facilities and organizations are beyond the scope of individual investigation and require making choices that can have far-reaching consequences and be difficult to reverse. The development of a complex data base or the creation of a research center may well constitute valuable enterprises that can contribute to the scientific work of many investigators; but when a large-scale project may command, say, an annual sum equal to one-third of the federal budget lines presently devoted to basic research in one of the core disciplines, a serious debate is joined. Modes of Support We recommend that the major mechanismfor supportingfundamental research in the behavioral and social sciences continue to be individual grants awarded under a scheme of competition among intellectually similar projects, with evaluation con- ducted by scientifically qualified and organizationally disinterested individuals. We recommend an aggregate increase of $70 million annually in the level of support for such research (as detailed in Previous chanters of this report). , ~ ~ The mainstay of research support in the behavioral and social sciences is the modestly sized (roughly, $30,000-$70,000 per year), short-term (up to 3 years), competitively awarded grant, administered through a standing organization (often a university), for the part-time support of research by an individual investigator with one or more assistants who are often graduate students work- ing part-time on research. Such grants provide resources to carry out the study
228 / The Behavioral and Social Sconces specified in the invesd~stor~ proposst with the requirement that technical and Rnancisl reporo be submitted periodicsHy to the Ending agency. The research resulo are expected to be Driven up and published in the scientific literature. Most scientists believe that such grants are s very good easy to nourish the highest quality and productivity of research. The investigstor-inidsted grant process is s system of repeated direct competition far new Rinds among discrete proposals. In each proposal the prospective invesdgator discusses his or her past research sccompl~hments, detach the methods he or she proposes to use sod the theories to be investigated, and estimates the ~gniEcance and promise of the proposed work. Evaluation of these propossL is generally carried out (oRen on an anonymous or conhdendsl bash to ensure candor) by other re- sesrche~ who possess the scientific expertise needed to understand and sp- prsise the work. Evaluators are, as s rule, precluded Tom judging the merits of soy proposal if they are closely involved with the psr~cular applicant or the investigator? institution. In most agencies, including the National Science Foundation, the Nshonsl Institutes of Health, sod the Alcohot Drug Abuse, and Dental Health Admin~t~tion, the technical evaluadon ~ conducted by an extemal scientific review panel supplemented by ad hoc consultants In other agencies, including the U.S. Department of De~nse, the scientific evsE ustion is intemal On-house), and the system ~ usually subject to s general periodic spprs~sl of the program by outside evaluators. There are also s variety of mixed systems, far exempt, in the U.S. Department of Education. It ~ essential to preserve what the scientific community widely agrees is especially valuable in the current system of allocating research support Grant Size and Duration reromme"d [~1~ndi~r individu~I research grands in Beh"~ior"1 and soci"I sciences be revised. E~ch pr~"I and "~ sho"Id include Unit fo purchase rim research m"~eri"I~ conga' sulfa or "squire and maintain rat animal, coII"~e and communicable Ails grit Ji~ersed col- Ic~ues, acquire or Kingpin android Nat research dig", and Signor' research saw incIuJi~ [he princ~1 ins. The proboscis [h"' receive [he most Sporadic ~"Iu"fions in co-~hive relics should be~l~nded even {~e [~"I number Agents "~d must Be constr~incd in Aver [o do so. In "~diNon,~ndi~ agencies shouId increase [~e present "verge Jur~fions ~~"rds in [~e BeN"vior"1 and soci"I sciences in order fo secure confinui~ ~rese~~h and rouge [~c amount Mme ~en' on prairie and revic~i~ propose. Much research in the behavioral and social sciences is both labor intensive and equipment intensive. For example, Usher cadences in the knowledge of how fundamental mental capachies develop Dig depend on using video devices
The Research Support System / 229 that can record where infants look or place their hands and computers pro- grammed to track and analyze facial movement. Research of this sort also requires staff to arrange with parents for visits to laboratories and to conduct interviews to obtain other relevant social, behavioral, and physiological data. Similarly, without considerable investment in both personnel and equipment, it is virtually impossible to run a modern laboratory studying the neurophys- iology of memory, the conditioning of complex primate behavior, or the var- iations in simulated commodity trading under varying prices and market ar- rangements. A field research facility to study fertility determinants across a larger area's population requires a substantial staff to develop a valid and re- liable sample, track the respondents, and periodically conduct interviews about births, deaths, marriages, illnesses, family finances, and intimate beliefs and practices over a lifetime. Such projects also need substantial space to house staff, records, and computational equipment. We stress these facts about personnel, facilities, and the resulting expenses of research, because, all too often, funding sources or officials who want the results of research or even universities still think that state-of-the-art work in the behavioral and social sciences requires only a principal investigator, one or two graduate students, and a slightly larger office allocation. The size of the typical research grant in the behavioral and social sciences often precludes substantial use of modern equipment or hiring trained staff, including the technicians, predoctoral assistants, and support staff, needed to carry out high- quality research. Because of year-by-year administrative decisions made by staff and review panels in key federal agencies such as the National Science Foun- dation, the trend of decreasing budgets experienced over the past decade has not led to appreciably fewer proposals being funded, but to reductions in the size of each grant in many research programs (see Appendix A). Another response to budgetary stringency by granting agencies has been to decrease the duration of grants. For example, although a maximum duration of 5 years per competitively reviewed proposal is possible at the National Science Foundation, the average standard (fixed-term) grant is typically 1 to 2 years. Continuing grants, which can continue for up to S years, comprise only one-third of all National Science Foundation awards in the behavioral and social sciences. In the National Institutes of Health, which has not been con- strained in the same way, investigators and review panels have nevertheless been oriented to short-term proposals and awards even when an investigator's proposal is on a more extended timetable. This pattern of short-term funding significantly increases the number of applications that investigators must pre- pare and reviewers must evaluate. As a result, significantly less time is available for research. If the costs of doing state-of-the-art work have come to exceed the sine and duration of the typical current grant, funding programs must resolve a difficult dilemma: whether to continue with too-small awards across the board, which
230 / The Behavioral and Social Sciences may inhibit or cut the scope of new projects, or to constrain the number of grants funded in order to foster highest quality work. The best solution would be more and larger grants, but at this time growth in size should take priority over growth in number. The Disciplines and Interdisciplinary Research We recommend that procedures for evaluating and funding interdisciplinary re- search proposals be reexamined. Most of the reviewers of those proposals should have scientific interests and competence in interdisciplinary areas. Stalking needs to be adequate to permit appropriate handling of interdisciplinary projects, which often require greater time and attention. Wefurther recommend that the National Institutes of Health and the Alcohol, Drug Abuse, andMental Health Administration reconsider the criteria they use to evaluate behavioral and social sciences research. Supportfor the study of behavioral and social phenomena should not be restricted to factors specific to particular diseases, but should extend to factors appearing important in the etiology and treatment of a number of diseases. The National Science Foundation should reconsider its programs for supporting studies of intel- ligent systems natural ones (humans and animals) and artificial ones (com- puters) which were separated after a promising start at developing an interdis- . Hi. clplmary program. Most research in the behavioral and social sciences (as in most science) is carried out within the confines of familiar academic disciplines. The largest behavioral and social sciences discipline, in terms of personnel trained, is psychology, followed by economics, sociology, political science, and then the substantially smaller disciplines of anthropology, linguistics, and geography. In addition, many scholars whose degrees or primary affiliations are in the fields of education, history, law, management, operations research, philosophy, police science, psychiatry, public health, and statistics identify with the be- havioral and social sciences on the basis of their research interests, methods, ot theories. Researchers oriented to the behavioral and social sciences are suf- Fciently numerous and mutually cognizant in some of these nearby disciplines to have created specialized research societies. The discipline of statistics, which is practiced, taught, and contributed to by scientists from a variety of back- grounds, is often considered and treated as fundamentally a part of the behav- ioral and social sciences. * The behavioral and social sciences disciplines are durable structures for professional certification, advancement, communication, and academic in- struction. Research emphases, theories, and methods come and go, but uni- *Descriptions of the main substantiative concerns of the respective disciplines can be found in Chapter 2 of an earlier report of this committee, Behavioral and Social Sciences Research: A National Resource. Washington, D.C.: National Academy Press (1982).
The Research Support System / 231 versity departments and disciplinary associations continue. Advisory and hon- orary groups such as the National Academy of Sciences and the American Academy of Arts and Sciences categorize their memberships along standard disciplinary lines. However, many frontier areas of research sprawl across dis- ciplinary boundaries and have hence to be considered interdisciplinary. For example, many researchers who are studying the development of human ex- pertise are as much computer scientists as they are psychologists. Research on addictive behaviors combines biological and behavioral principles in devel- oping new theories and research designs. The comparative analysis of languages and religions and the study of contemporary international structures and pro- cesses draw on the contributions of anthropologists, economists, political sci- entists, sociologists, and historians. Another feature that contributes to interdisciplinary work is the transfer of methods and theoretical approaches. For example, formal models originally developed in logic, mathematics, and computer science now pervade many other disciplines. Formal mathematical or philosophical training has come to be common among those who do research on the nature and development of spatial, mathematical, and logical thought. Developments in measurement the- ory have found applications in marketing, the study of risk and utility, and theories of sensory processes. A number of general mathematical topics, in- cluding stochastic models, systems of linear and nonlinear equations, and other geometric and algebraic structures, have found applications in econometrics, models of conflict resolution, and the sociology of occupational attainment. Over time, interdisciplinary research can be best seen as an interaction that develops across and works back into the disciplinary structures. Because re- search frontiers often involve problem definition, exploratory knowledge, de- velopment of new methods, or strong policy interest, creative scientists from various disciplines can often enter them relatively easily. Progress in such research usually accelerates when the problems are sufficiently defined to en- able the application of the most advanced technical tools from one or more disciplines tools that are constantly being improved and refined. As progress in research crystallizes into new bodies of accepted knowledge, continued work in the area may become incorporated into the core curriculum (and other durable structure) of one or more disciplines- in some cases directly displacing other content and transforming the disciplines or may become the basis of a newly recognized specialty. Over time, such a specialty may even approach or achieve the status of a discipline. Examples are the emergence of linguistics, mainly from anthropology and comparative languages, and of criminology, from sociology, law, and other fields. These kinds of developments- which count significantly among the research frontiers highlighted in this report have stimulated seed grants from private foundations, interest from time to time on the part of federal agencies, and occasional support from colleges and universities. But they go against the grain
23~- / The Behavioral and Social Sciences of standard procedure. For example, consider a collaborative study of language leading that brings together linguists and philosophers located in a humanities division; neurologists in a school of medicine; psychologists who, depending on the institution, may be in either social, biological, or physical sciences divisions; and computer scientists and applied mathematicians in schools of engineering. Developing a collaborative proposal means dealing with different assumptions about normal teaching loads, different mechanisms of support for graduate students, and even different academic timetables. If such an interdisciplinary group applies to a diversified funding agency such as the National Science Foundation, the proposal is likely to be of partial interest to different divisions within a directorate and even to different direc- torates. Unless the agency staff has sufficient reserves of time and energy to adopt such a proposal and give it special handling, shepherding it diligently through all the possible sources of support, the review system is likely to treat it more as an unworkable processing problem than an unusual scientific op- portunity. When agency staff are reduced or their grant loads are increased in the interest of greater processing efficiency, the capacity to respond creatively and effectively to such opportunities diminishes and, ultimately, disappears. Unless funding agenciesand universitiesperiodically consider whether they have the capacity to facilitate good research that does not fit the established patterns, they are not likely to have such work in progress, and new interdis- ciplinary research will be inhibited. The most challenging aspect of interdisciplinary research is to assess whether new developments are genuinely powerful or are passing fads. The general issues in evaluating interdisciplinary projects are, first, to determine in which budgetary pool a proposal competes and, second, to decide the intellectual criteria on which it is to be judged. New efforts are needed to consider those issues. If review is primarily by disciplinary panels, they must include research- ers who are experienced in interdisciplinary projects. Alternatively, agencies could constitute a specialized panel or panels to evaluate all interdisciplinary proposals as a group, with details of program-based funding to be worked out separately and subsequently. The most radical step would be to generate a separate resource pool for innovative interdisciplinary projects, which would not compete on a case-by-case basis with disciplinary proposals. Two recent examples of lack of receptivity to interdisciplinary developments are of particular concern. First, in recent years two major research support agencies in the U.S. Department of Health and Human Services, the National Institutes of Health and the Alcohol, Drug Abuse, and Mental Health Admin- istration, have increasingly stressed that behavioral and social research should directly address specific disease syndromes. This emphasis, reinforced by the sizable representation of clinically oriented physicians on the behavioral re- search review panels of these agencies, excessively narrows their investments in many useful areas of basic research. A focus on interdisciplinary areas of
The Research Support System / 233 etiology, prevention, treatment, and services research would alleviate this un- fortunate tendency and strengthen the research portfolios of these agencies. The second specific concern is the former Division of Information Science and Technology at the National Science Foundation, which in 1986 was trans- ferred from the Directorate for Biological, Behavioral, and Social Sciences, where it had drawn information science research into close contact with research on cognition, language, and other areas of the behavioral and social sciences. It was transferred to the new Directorate for Computer and Information Science and Engineering and reorganized as a Division of Information, Robotic, and Intelligent Systems. As a result of the move, the unit has ceased to fulfill its important role in supporting basic behavioral and social sciences research proj- ects with strong information science components, to the detriment of both areas and especially of their interdisciplinary intersection. It would be sub- stantially more productive in our view for the National Science Foundation to redirect this program, to bring studies of artificial information processing sys- tems back into contact with studies of natural ones. Interdisciplinary Research Centers We recommend that private and jederalfunding agencies encourage submission of proposals for relatively long-term multidisciplinary research activities focused in research centers, housed either in single institutions or in consortia. Funding should be provided for new centers as well as for centers of proven effectiveness. Review of center proposals should be at least partly independent of single- investigator proposals. A program to provide development or pilot funds for center proposals is necessary in nearly all behavioral and social sciences fields. We recommend that research centers receive approximately $2S million in new funding. Interdisciplinary research centers are intended to solve two closely related problems. One is the need to bring together a critical mass of scientists for research that by its nature requires interdisciplinary input. The other is to provide an appropriate physical, logistical, and administrative setting for ad- vancing interdisciplinary research that may not be limited to a particular focus or set of problems. There are several examples of national and international centers and institutes that have endured for decades and have contributed to significant advances in the behavioral and social sciences. Many of the same basic questions that were discussed in connection with evaluating new large-scale data collections apply to new research centers: How worthwhile scientifically are the anticipated results? What fraction of the new funding of the relevant fields should go into a center or centers? Can the same results be achieved as efficiently by encouraging new directions of research through individual grants or already established centers? Additional consid- erations are the frequency and nature of evaluation of the work of a center; the
234 / The Behavioral and Social Sciences ways in which less successful centers might be phased out; and the institutional locus (for example, in a university or an independent institute). Most initiatives for centers discussed in Chapters 1-S are oriented either toward the efficient use of a multiuser laboratory or toward a large-scale data collection project. In such cases, the sheer size of the laboratory or survey project generates the need for a centralized unit to facilitate the work. In some other cases, a center is seen principally as a mechanism to ensure interdisci- plinary communication. For example: · Learning and memory. Centers in this area could bring together inves- tigators in psychobiology, neurobiology, cognitive science, mathematical neural modeling, and the experimental analysis of behavior to focus on particular memory circuits, networks, or aspects of learning and memory. A particular advantage of such arrangements is to facilitate the sharing of equipment, such as imaging devices. One or more centers could be devoted to studies using animal subjects, which would introduce economies of scale in financing improvements in the conditions under which the animals are maintained and studied. One or more research centers could focus on particular aspects of clinical learning and memory disorders, for example, . aging anc amnesia. · Affect and motivation. Centers in this area could bring together research- ers among whom there is at present too little sustained communication, including psychiatrists and neurologists who are concerned with research on and treatment of affective and motivational disorders and psychologists and sociologists who are engaged in research on the interpersonal and social factors that shape and condition emotional development and expression. · Demographic behavior. There is substantial warrant for a new demo- graphic research facility in the developing world, particularly in Africa, to engage in epidemiological, ethnographic, and demographic work similar to that now carried out by the International Centre for Diarrheal Disease Research in Bangladesh. Data are needed on biological characteristics, pat- terns of individual decision making regarding marriage, childbearing, and childrearing, and institutional and cultural features in other regions, such as Africa. Sources used would include demographic records and surveys, field observations, and historical records. Such a center would enable re- searchers to track discrete cohorts of individuals through several decades of life and would be a stable foundation for research on childbearing, mortality, morbidity, population, and the institutional and cultural factors that affect these phenomena among hundreds of millions of people about whom very little is known. In general, the funding of such research centers but not a number of other large-scale activities such as longitudinal studiescan be arranged on a two-
The Research Support System / 23S tier basis. One tier is a core grant designed to provide the autonomous infra- structure and flexibility needed by the center. A core grant covers basic oper- ating and administrative costs, some shared laboratory or computing facilities, common staff members, and a certain amount of discretionary research funding mainly for new investigators before they move to the second tier. That second tier consists of proposals submitted by individual investigators and reviewed in competition with all other investigator-initiated grants. This second tier, of course, forms the backbone of the research endeavor of a center. The advantage of such a partition of funding is that research progress will be regularly and appropriately monitored by the relevant specialists. THE PROBLEM OF VOICE The place and role of the behavioral and social sciences in the administrative arrangements of relevant federal agencies should be critically reappraised. It is necessary to ensure continuous high-level representation of the scientific needs and opportunities in these fields. The coordination of these sciences and their general advisory roles should be strengthened by establishing a mechanism to coordinate interagency policy on behavioral and social sciences research. One feature of the present situation that works against the best interests and the best utilization of the behavioral and social sciences is the way they are situated within the administrative structures of federal research agencies. If one looks to research-oriented academic institutions, it is customary for faculty members to be affiliated with divisions of humanities, physical sciences, life sciences, and social sciences or schools of business, law, medicine, and other professions. (History is located sometimes with the social sciences and some- times with the humanities, and psychology is located sometimes with the life sciences and sometimes with the social sciences.) This division permits behav- ioral and social scientists to negotiate effectively the immediate administrative environments in which they work and, in most universities, to gain reasonable access to (and occasionally to find colleagues in) high administrative positions. If one looks at the behavioral and social sciences in federal research agencies, there is no parallel to academia. For example, the National Science Foundation has its Directorate for Biological, Behavioral, and Social Sciences. The title of the directorate, except for its nonalphabetic sequence, suggests perhaps a cer- tain degree of parity, but that is not so: the budget for the behavioral and social sciences was about $50 million in fiscal 1987, compared with $200 million for biological research. Moreover, since its inception in 1974, this directorate has been headed by biologists. There are similar or even more disproportionate relationships between biomedical and behavioral and social sciences research in every institute in the Alcohol, Drug Abuse, and Mental Health Administration and National Institutes of Health.
236 / The Behavioral and Social Sciences This kind of subsumption under related fields has led in the past and could lead at some future date to misperceptions at the highest policy levels about the scientific opportunities and needs of the behavioral and social sciences. This problem of misperception is aggravated by the paucity of high-level co- ordination between agencies of the federal government that fund behavioral and social sciences research, including the Departments of Health and Human Services, Defense, Labor, Education, Justice, Commerce, State, and Housing and Urban Development, and independent agencies, especially the National Science Foundation and the Smithsonian Institution. To deploy scarce re- sources wisely and effectively, there should be efforts to develop complemen- tary (not, of course, monolithic) research and funding policy for the behavioral and social sciences across the federal government.
