Women: Their Underrepresentation and Career Differentials in Science and Engineering: Proceedings of a Conference (1987)

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.

Social Sciences 1971 1973 1975 YEAR OF DOCTORATE 1977 SOURCE: Survey of Earned Doctorates, National Research Council. Figure 3 Percent of doctoral degrees in science and engineering awarded to women, l970-l977. marked differences in the representation of women among specialties in psychology (Russo and O'Connell, l980:3l) but no significant differ- ences in specialty choice among men and women in chemistry [National Research Council, Committee on Education and Employment of Women in Science and Engineering, l979 (hereafter, CEEWISE); Syverson, l980:24]. in any event, the marked differences that do exist between men and women in the fields they choose must be taken into account in examining salary differentials and differentials in rates of unemployment, since these differ greatly field by field. Recent data show that rates of unemployment are higher and salaries lower in the fields in which women are more numerous, but this has not always been so. Thus, these dif- ferences seem not to arise exclusively from the gender composition of these fields. Men and women scientists and engineers also work in different sectors of the economy. Relatively more men than women (69 percent versus 53 percent) work in industry and relatively more women than men (23 percent versus l0 percent) work in educational institutions (NSB, l985:Table 3-7). A similar pattern of underrepresentation in industry and overrepresentation in education holds for women doctorates, with some variation by field, but gender differences in sector of employment are not as great among doctorates as among all scientific and technical workers (NSB, l985:Table 3-8). l27

While rates of unemployment for scientists and engineers are rela- tively low compared to those for all American workers, they have been consistently higher for women than for men (e.g., 3.4 percent versus l.3 percent in l984) (NSF, l986:5). For doctorates, rates of unemploy- ment are even lower: 2.5 percent for women and l.2 percent for men (NSF, l986:7). While such percentages are comparatively small, they must loom large for those unable to find work. Women are also more apt than men (8 percent versus 2 percent) to be underemployed—that is, involuntarily working part-time or outside science and engineering, a difference that results in part from the small numbers of women in fields, like engineering, which have comparatively good employment prospects for both sexes (NSF, l986:6-7). Rates of underemployment for Ph.D.s are lower for men and women—l.2 and 2.5 percent, respectively (NSF, l986:7). All this means that women's work histories are briefer, on aver- age, than men's and more often marked by part-time work. Since these patterns hold in every field, field differences in employment rates cannot fully account for them. Nor can they be attributed solely to women's domestic and parental responsibilities. Labor force participa- tion rates (that is, the proportions working or seeking work) for men and women scientists and engineers are now high and almost identical (96 percent for men and 94 percent for women). Similar patterns hold for doctorates (NSF, l986:4; l3lff.). Yet larger proportions of women than men doctorates interrupt their careers for a year or more (l7 per- cent versus 5 percent), and, as Centra (l974:32) shows, spells of unemployment last longer for such women than for men. On average, how- ever, women Ph.D.s remain out of work for a brief time (Astin, l969: 58). Yet career interruptions do not explain much of the difference in salary paid to men and women scientists (Lewis, l986). Contrary to widely-held belief, women's work patterns are weakly related to their family obligations; women scientists with young children under six years of age are more apt to be working or seeking work than those with older children (NSF, l986:5). However women's family obligations are generally assumed to affect their work histories, and these assumptions are socially consequential. From this brief report, it should be evident, then, that the num- ber of women in science and engineering has risen rapidly in the last l5 years. However, women scientists and engineers work in different fields and in different sectors of employment than men, are younger, and have less professional experience and education. These differ- ences, obscured by simple bivariate distributions, cannot be ignored when comparing the career attainments of men and women scientists and engineers. The sources of gender difference in the choice of fields and sector of employment are not well understood and require more in- tensive study than they have received. Research on Career Attainments of Men and Women Scientists and Engineers Studies of career attainments in science and engineering not sur- prisingly reflect researchers' distinctive disciplinary interests, l28

styles, and methods of inquiry.^ These studies are sharply focused: o On scientists, not engineers—with the notable exception of the research program of LeBold and his group (LeBold, et al., l983). This limitation is significant, since engineers actually out- number scientists and comprise 56 percent of all those working in science and engineering posts. o On academics rather than industrial or government scientists, this also being a significant limitation. Just l2 percent of all scientists and engineers worked in educational institu- tions in l983, although a larger share of doctorates (53 per- cent) did so (NSB, l985:Tables 3-7 and 3-8). o On holders of the doctorate rather than those with lower level degrees, the former being just ll percent of the population of scientists and engineers (ibid.) o On the current period o On individual career histories, not the effects of institu- tionalized processes of evaluation and allocation of resources and rewards on the careers of men and women scientists or their subjective experiences and attitudes. Thus, the findings of research reviewed here are partial, limited to particular subgroups and to a particular time (see Zuckerman and Cole, l975, for an earlier review). They are also complicated. To make gender comparisons clear, I take up phases of the career one by one rather than emphasizing their connections in a model. Such models have been proposed (Cole, l979; Helmreich, et al., l980; Long, this volume; Reskin, l978), but longitudinal data for men and women are largely limited to the first l5 years of their careers, making models for the entire career more schematic and speculative than empirically grounded. Initial Qualifications The qualifications of men and women beginning careers are now similar in three important respects. First, among doctorates, the only group for which data are available, the intellectual calibre of men and women insofar as it is measured by standardized tests and academic per- formance is much the same. Women do as well as or better than men on such tests (Cole, l979:62; CEEWISE, l979:23-25). However, measured ability seems unrelated to research performance in science (Bayer and Folger, l966). Second, the proportions of men and women getting degrees from top-ranking research university departments do not differ overall. This is so when departmental rankings are measured by receipt of research funds or by ratings of the quality of doctoral programs (CEEWISE, l979:37; l983:2.7). Such differences, however, appear in 5These researchers are drawn mainly from sociology, but also from psychology and economics; with a cadre of policy researchers being associated with the Committee on the Education and Employment of Women in Science and Engineering of the National Research Council, research- ers in federal agencies have also made important contributions. l29

some fields: significantly fewer women than men get degrees in top- ranked departments in mathematics and physics, but significantly fewer men than women get degrees from these departments in microbiology and psychology. Given the long term consequences of scientists' doctoral origins for their careers (Long, l978; Long, et al., l979), this over- all similarity is worth underscoring. Third, men and women are much the same age when they get doctoral degrees. In the doctoral cohort of l98l, the median ages of men and women were 30.3 years and 3l years, respectively, with variation, of course, among fields (CEEWISE, l983: 2.3). In the mid l960s, however, new women Ph.D.s were markedly older: the comparable figures were 30.9 years of age for men and 32.5 for women (Harmon, l978:54). Women also took longer to get their degrees, and a larger share were 40 years old or older when completing their degrees (Harmon, l978:54-55). By l98l, however, women began careers substantially later than men only in the medical sciences. In all other fields, their ages were approximately the same—that is, within a year of one another (CEEWISE, l983:2.8). Thus, women doctorates are quite similar to men in where they got degrees and when and in those attributes that are measured by standard- ized tests of ability. Men and women are also equally apt (30 percent and 28 percent, respectively, in l98l) to take postdoctoral appoint- ments in the sciences and engineering (Coggeshall, l98l:l48), and about the same proportions were accepted for postdoctoral fellowships by top-rated institutions. Such fellowships help to transform recent graduates into independent scientists and provide the opportunity to establish a program of research before teaching begins. The gender similarity here suggests that men and women start their careers as equals in these respects as well. However, finer-grained data indicate that the gender similarities in postdoctoral appointments are not as great as they seem at first. The reasons that men and women give for taking postdoctoral fellowships differ. Married women particularly emphasize the geographic location of the fellowship more than do single women and men, married or single {Coggeshall, l98l:l5l). Moreover, Reskin's (l976) study of chemists suggests that women may receive postdoctoral fellowships as often as men, but those they hold have less prestige. This, Reskin conjectures, is indirect evidence that women more often than men accept postdoctoral fellowships because real jobs are unavailable. Further analyses of these same data show that holding a prestigeful fellowship helped men more than women get tenure-track positions after completing their fel- lowships (Reskin and Hargens, l979:ll8). Other studies show that men who had held postdoctoral appointments earned more than women in every field seven years after the fellowship, after taking sector of employ- ment into account (Coggeshall, l98l:l56). Such differences in earnings doubtless reflect gender differences in academic and organizational rank and possibly the gender differences in publication that begin to appear at this stage of the career. Job Histories What happens to men and women when they search for their first l30

jobs? In engineering, men and women fresh out of college encounter quite similar job opportunities, at least as gauged by salary and by proportions given supervisory and technical responsibility in the year following graduation (LeBold, et al., l983:22-23). Circumstances are more complex for new college graduates seeking jobs in the sciences. Women now do about as well as men (again using relative salaries as an indicator) in the aggregate, in all fields but the biological sciences (Vetter, l98l:l3l8); but there is evidence that grade levels and sal- aries differ greatly for men and women scientists starting work in industry and government (CEEWISE, l980:4-5). Among new Ph.D.s, gender differences in job opportunities are less marked. This was not always so. The once common pattern of women doctorates having to choose be- tween teaching in a women's college or serving as a research associate in a university laboratory no longer holds. Although women continue to work in academia more often than men, an increasing share of new women doctorates find jobs outside the academy; and similar proportions of men and women doctorates now plan to work in industry (CEEWISE, l983:2.l3). Confining our attention to academics and to doctorates, current data show, contrary to expectation, that men and women find jobs in much the same kinds of educational institutions. About the same pro- portions are hired by top-rated universities, whether one uses reputa- tional measures or rankings based on receipt of funds for research. In l977, for example, about l3 percent of women doctorates employed in academic institutions were affiliated with the top 25 in terms of funding as against l4 percent of men (calculated from CEEWISE, l979: Table 4.3).^ These data are consistent with those reported by Cole (l979:70) and by Ahern and Scott (l98l:47) for matched samples of men and women doctorates and are significant given the important effects organizational context has on scientists' research performance— although these effects have been demonstrated only for men (Long and McGinnis, l98l). However, being on the faculty at MIT or Berkeley is quite a dif- ferent matter from being employed at these same institutions as a research associate or in other off-ladder positions. Too few studies combine data on institutional affiliation with data on organizational rank. Those that do show that women are slightly overrepresented among assistant professors, given their numbers among Ph.D.s generally, with some variation by class of institution. In l977, l9 percent of assistant professors in the sciences at the top 25 universities were women, l5 percent in the second tier, and l8 percent in other institu- tions (calculated from CEEWISE, l979:Table 4.3), this at a time when women earned l2-l3 percent of all doctorates in the sciences and engi- neering. However, a larger proportion of women than men also held instructorships, lectureships, and other off-ladder posts—the outcome, in part but not entirely, of fewer women proportionately holding the 6Moreover, the proportions of women doctorates holding faculty posi- tions at all ranks increased more sharply in the top 50 institutions in the l970s than in all other institutions (CEEWISE, l979:76). l3l

doctorate (Bayer and Astin, l975:797), since the same pattern holds among men and women Ph.D.s (CEEWISE, l983:4.ll). The same data look altogether different when examined from the perspective of the distributions of men and women Ph.D.s among academic ranks. These cross-sectional data show, of course, that women are heavily concentrated in the lower ranks as compared to men. For exam- ple, 37 percent of women scientists and engineers held assistant pro- fessorships in l977 as against 22 percent of men (calculated from CEEWISE, l979:60, Table 4.3) while the situation was reversed at the top of the academic ladder, where 39 percent of men held full profes- sorships versus l5 percent of women (CEEWISE, l979:6l).7 Such con- centrations are often taken to mean that women are not promoted but are kept in lower ranks. In fact, women do not get promoted to high rank at the same rate as men, but such cross-sectional data cannot show this, since they do not take into account the differing age distribu- tions of the pools from which men and women professors are drawn. So far then, comparisons confined to new Ph.D.s and to newly hired assistant professors show that women become assistant professors at about the rate that would be expected, given their representation among new degree holders.& For this limited group and for the cur- rent period, it would appear that gender parity has been achieved, especially at the top-rated institutions. At the same time, it is clear that earlier cohorts of men and women scientists encountered quite different job prospects when they began their careers. Gender parity in the hiring of new Ph.D.s in academia or industry has not, the evidence suggests, been around for long. Later Jobs What has happened to earlier age cohorts—to men and women who began their careers a decade or more ago? Cross-sectional data on the ranks that they have attained in educational institutions, industry, and government are less fine-grained than one would like, but they are relentlessly consistent: such women, on average, started out in lower ranks than men, and the disparity in their ranks continues. 9 Avail- able longitudinal data on men and women of the same professional ages do not contradict the cross-sectional evidence. For example, Ahern and Scott (l98l:l8, Table 2.5) report that among men and women scientists who received Ph.D.s in the l940s and l950s and were matched for field, 'In industry, the distributions are similar: 29 percent of men and l6 percent of women work mainly in managerial jobs (CEEWISE, l983:5.3). ^Not all assistant professors are newly minted Ph.D.s, but the great majority are. ^Comparable data on scientists and engineers in industry are sketchy but largely consistent with those for academics. The large differences in rank and managerial responsibility for the aggregate of men and women scientists and engineers working in industry suggest that such differences are concentrated among older workers (CEEWISE, l983:5.3). See Perrucci (l970) for an earlier analysis. l32

reputation of degree-granting department, and race, 86 percent of men had become full professors by l979 as against 64 percent of women (a ratio of l.3:l). In the cohort that got Ph.D.s in the l960s, a smaller proportion of both men (52 percent) and women (30 percent) had become full professors by l979, but men proportionately outnumbered women by a ratio of l.7:l (calculated from l98l:25, Table 3.5). And finally, among those who had gotten degrees between l970 and l974, just 6 per- cent of men and 3 percent of women had already become full professors. However, 4l percent of the men and 29 percent of the women in this young group, a ratio of l.4:l, had already been promoted to associate professorships (calculated from l98l:33, Table 4.4). In short, the evidence suggests that men are ranked consistently higher than their age-peers among women. Gender differences in rank turn up in all classes of academic in- stitutions but are most accentuated in the top-ranking ones (just the opposite of what is observed among assistant professors). On average, the higher the prestige of the institution, the lower the proportion of women in full professorships (CEEWISE, l979:6l; CEEWISE, l983:4.7) and the smaller the proportion of women assistant professors who are promoted to associate professorships (CEEWISE, l983:4.l3). These dif- ferences between institutions need to be treated with some caution, however, since women employed by the higher-ranked institutions may be younger, on average, than those employed in other institutions. Among those who earned degrees in the l940s and l950s, women are disproportionately bunched in what are called "off-ladder appoint- ments"—that is, instructorships and non-faculty posts such as research associateships and other miscellaneous research jobs (Ahern and Scott, l98l:l8). Estimates vary on the proportion of women in these posts, ranging from l3-20 percent or about 2-3 times the proportion of men (CEEWISE, l983:4-l5). Such positions are insecure, especially in times when research money is scarce. in many universities, holders of off- ladder appointments have difficulty applying for research funds as principal investigators; they are dependent on others and cannot set their own research programs. Perhaps the only unmitigated virtue of holding such appointments is that they require less frequent attendance at committee meetings than do regular faculty posts. Finally, trend data on the distribution of women among academic ranks indicate that important changes have occurred. Since the l970s, there have been increases in the proportions of women scientists in every academic rank in all classes of educational institutions, with the rates of increase being largest in the top-ranked institutions— this being partly the result of so few women being on the faculties of these institutions before the l970s (CEEWISE, l983:4.6-4.l3). At the same time, it is worth emphasizing that the top-ranked institutions grew very slowly, if at all, in the l970s and early l980s. Increasing numbers of women faculty did not simply result from increasing overall faculty size. These trend data, like those examined earlier, show growing similarities in the career attainments, particularly in aca- demic rank, of successive cohorts of men and women scientists. The gap between them is narrowing but it has not been eliminated, especially at the highest ranks. l33

Tenure As with rank, so with tenure (the two being strongly intertwined): academic women scientists and engineers are less apt than men to be tenured and also less apt to hold tenure-track jobs (CEEWISE, 1983: 4.l5). Among those in tenure-track posts in l983, two-thirds of men as compared with 40 percent of women were already tenured (NSF, l986:4). Such marked cross-sectional differences shrink, of course, when professional age is taken into account, but they do not disappear. Among men and women scientists studied by Ahern and Scott (l98l) and matched, it will be remembered, for field, doctoral department, race, and age, women in each age cohort were less apt than the matched men to be tenured. Among those who took degrees in the l940s and l950s, 98 percent of men and 88 percent of women were tenured by l979, more of course than among those who took degrees in the l960s. In this age group, 89 percent of men and 78 percent of women in the sciences had received tenure. Gender differences in rates of tenure are significant in some sciences but not in others. Among younger men and women, the tenure picture is more complex. Among those who received degrees be- tween l970 and l974 and were associate professors, women, were just as apt as men to be tenured—in all fields, not just in the sciences. But women scientists, it turns out, were less likely than men to have be- come associate professors (29 percent versus 4l percent) and thus less likely, in the aggregate, to be tenured. Other evidence from large samples of scientists and engineers working in colleges and universi- ties are consistent with these data; the differences in the tenure status of men and women professors appear to be narrowing (CEEWISE, l983:4.l4), suggesting movement toward gender parity in this important aspect of academic careers. Timing of Promotion and Tenure For academics, it is not just holding high rank and tenure that matters, but how long it takes to achieve them. Women are promoted more slowly than men; and among those promoted, they are slower to re- ceive tenure. Again, turning to the carefully selected matched samples of men and women studied by Ahern and Scott (l98l:27), 24 percent of the women who took degrees in the l960s and who had tenure by l979 waited nine years or more to get Ph.D.s as compared to l4 percent of the men (these data are for all fields). Correlatively, younger women doctorates also wait longer for promotion than men. A smaller share of women (44 percent) than men (62 percent) who earned Ph.D.s between l970 and l974 and were assistant professors in l977 had been promoted to associate professorships three years later. However, these data do not tell the whole story. Among those who had already been made asso- ciate professors, equal proportions of men and women were promoted to full professorships in this same period (Ahern and Scott, l98l:35). The dominant pattern in all cohorts, however, is that a larger share of men are promoted to high rank than women in the same group and they are also promoted more quickly (CEEWISE, l983:4.l3). Taken together, the greater incidence and more rapid promotion of men enlarge gender l34

differences in attainments among age peers over time. Other data sug- gest that gender differences in "time to tenure" are more pronounced in top-rated universities than in others—that is, the more prestigious the institution, the longer women wait to be promoted (Ahern and Scott, l98l:40). It is worth reiterating, however, that without data perti- nent to role performance of men and women, it is difficult to say much about the causes of gender differences in rank and tenure. In the case of rank and tenure, the general pattern of growing disparities with age holds up to a point and then narrows toward the end of the career. Large majorities of men and women who remain in academia long enough, do eventually get tenure and most do eventually get promoted. Yet, disparities in rank and tenure persist among older men and women scientists. Among those who have had their doctorates for 30 years or more, as we have already seen, a considerably larger share of men than women hold full professorships, although the vast majority of both are tenured (Ahern and Scott, l98l:l8). Salary Differences In light of gender differences in rank, it comes as no surprise that women scientists and engineers earn less than men. On average, in l984, their median salaries were 71 percent as large as those of men (NSF, l986:7).^ The disparity varies among fields, is smaller in engineering than in the sciences, and is also somewhat smaller among Ph.D.s than holders of master's and bachelor's degrees. in l983, for example, women doctorates earned 78 percent as much as men. But since salary is closely tied to age, experience, and rank and since women scientists and engineers, on average, are younger, have fewer years of experience, and hold lower ranks, some differences in salary are to be expected. However, among doctorates in the sciences, "about one-half of the differential in female-male salaries remains unexplained after standardizing for field, race, sector of employment and years of pro- fessional experience" (NSF, l986:8).H In academia, this pattern holds for men and women associate and full professors. Women earn less than men in the same ranks who got degrees in the same years. The most recent data available (l98l) show that salary differences between men and women are larger for full than difference is absolutely large, but women in the sciences and engineering in fact generally do somewhat better than women college graduates, who earn 67 percent as much as men with the same educational attainments (NSF, l986:7). H-Again, the data available for industrial scientists are far less detailed than those for academics. These are quickly summarized: women scientists who work in industry have lower median salaries than men, the salary differential increases with years of experience, but (unlike academics) such differences turn up among newly hired men and women as well as among older ones. it is not known whether such dif- ferences between men and women are explained by their different field distributions (CEEWISE, l983:5.5). l35

for associate professors but absent for men and women assistant pro- fessors (CEEWISE, l983:4.2l). Among full professors, however, women's salaries are not lower than men's across the board; they are as large in some fields—computer sciences, earth sciences, and engineering—but far smaller in others. The largest difference—a median of $6,200 per year (l5 percent less than men's salaries)--appears in the medical sciences, and this is followed closely by a difference of $5,800 in economics (l4 percent less than men's) (CEEWISE, l983:4.22). Among associate professors, similar patterns are evident. Women's salaries are comparable to men's in some fields but not in others, with women doing least well in the medical sciences and economics (ibid.). Such differences cannot be attributed to a larger share of women than men working part-time, since these data are limited to full-time employees alone. These data are consistent with those reported by Bayer and Astin (l975), whose de- tailed analyses reduce salary differences between men and women at the full professor level by holding a variety of variables constant, but do not eliminate them altogether. They report that a residual R of -.l2 remains between gender and salary for full professors after the effects of all statistically significant predictors they studied were considered (see also Fox, l985, on gender differences in salary at- tributable to rank and function). Although it is useful to consider rank in comparing the salaries of men and women scientists, doing so makes for systematic underesti- mates of gender differences in salary. Discrimination against women in promotion and appointments has the second-order effect of discrimi- nating against them in salary, an effect that is erased when compari- sons are confined to men and women in the same ranks. Comparisons of the salaries of men and women in any given year also systematically understate long-term gender differences in salary, since these differ- ences accumulate year by year. Even modest median differences in an- nual salary quickly add up to large differences when considered over the course of scientists' careers. l2 As with rank, so then with salary. The three patterns noted earlier of persisting gender difference, growing gender parity—par- ticularly among the young—and increasing difference over the course of men's and women's careers hold fairly well. But in the absence of longitudinal data on salary, it is not clear whether the salary gap between men and women narrows toward the end of the career as women catch up to men in rank. However, some hint that this does not occur can be found in the data reported by Ahern and Scott (l98l:77). They examined salary residuals for matched triads of men and women faculty who were in the same fields and who had received doctoral degrees at I2AS more than one older woman scientist has observed, having earned small salaries and being promoted late has the further consequence of producing small pensions on retirement. Since retirement pay is pegged to the amount of past salary and the number of years that contributions to retirement plans were made, women's annuity incomes tend to be smaller than men's. l36

the same time. They showed that, on average, women in these triads earned less than the men more often than the reverse was so. Deficits in women's earnings, as compared to those of matched men, increased among older Ph.D.s and were greatest for those in the sample who had the doctoral degree longest (l6-2l years). These cross-sectional data on scientists of the same professional ages suggest that disparities in salary may increase with age, but they cannot, of course, substitute for the longitudinal data needed to determine whether the gender dis- parity in salary actually grows larger with age. Convergences in rank, tenure, and (to some extent) salary among men and women over the past decade and a half coincide with and may be the outcome of efforts to comply with Affirmative Action legislation. Enacted first in l972, this legislation made institutions of higher education responsible for correcting gender discrimination in employ- ment, rank, and salary. Such convergences may be consequential for women's later careers. There is good reason to suppose that high rank provides greater resources and opportunities for research, including less teaching, more access to graduate students and postdoctoral fel- lows, and more space. Thus, recent improvements in rank and the op- portunity structure for research for women, possibly associated with Affirmative Action, may enhance women's future research performance while improving their current rewards. Role Performance So far, I have focused on gender differences in rank, tenure, and salary and noted repeatedly that such differences need to be examined in light of possible gender differences in the quality of role per- formance—that is, how effectively men and women do their jobs. DO men and women who do their jobs equally well receive the same rank, tenure, and salary? This question is central because some claim that women are not rewarded at the same level as men because they do not perform as well as men. But while this question is central analytically, it can- not be answered in quite the way it is put. There is no satisfactory aggregate measure of role performance of scientists and engineers con- sidered in their multiple roles as teachers, administrators, research- ers, managers, and citizens of the scientific community, even though we do attempt to measure some of these separately. Thus, while we know that this question needs answering, we also know we cannot answer it now. Research Performance Some data are available on one significant aspect of scientists' and engineers' role performance: the extent to which they contribute to the advancement of knowledge through publication. The number of papers that scientists publish and the number of times that they are cited—that is, the number of times that their papers are used in the research literature—are, at best, crude measures of research perfor- mance. However, many studies show that these are correlated with other, more direct measures of extent of contribution such as peer l37

judgments and honorific awards (Cole and Cole, l973; Garfield, l979, l982; Gaston, l978; Zuckerman, l977). These studies suggest, in the aggregate if not in any individual cases, that the extent of publica- tion and citation do register differences in evaluation of contribu- tions to scientific knowledge. The use of publication and citation counts is by no means uncon- troversial (Edge, l979). And it is important to note that some who find them acceptable for men scientists and engineers object to their use in the special case of assessing the relative research performance of women. These critics claim that few women hold high rank in the top research universities and, thus, few women have as much access as men to the resources needed to do research that leads to high rates of publication and citation (CEEWISE, l979:xiv). There is merit in this criticism insofar as it emphasizes gender-related inequalities in opportunities to do research and to publish. At the same time, this criticism is not entirely valid. First, the number of women in high-ranking positions is not too small for comparative analysis and, of course, the number of women in lower-level positions is substantial enough to allow for comparisons with men who are similarly situated. Second, research performance is almost uni- versally taken into account in science in decisions on allocating re- sources and rewards. Like it or not, research performance is conse- quential for the careers of men and women scientists, and it is important to know how men and women compare in this respect. Third, it is not only feasible to compare men's and womens1 performance (using these admittedly crude indicators), but it is necessary to do so if the sources and consequences of gender differences in research per- formance are to be identified. And last, should we find that men and women whose research performance is much the same are differently re- warded, this would provide justification for concluding that gender, rather than functionally relevant criteria, really does affect the allocation of resources and rewards—better justification than would be available if such assessments were not made because they were con- sidered unacceptable and even taboo. With that lengthy preamble, what is known first about the compara- tive access of men and women scientists to resources for research and then about their research performance? Do men and women have equal access to research support? The evidence here is exceedingly sparse. Limited data on funding at the National Institutes of Health during the early l970s (NIH, l98l:23-25) show that a small proportion of women apply for research funding, smaller than their numbers in the pool of life scientists. But among those who do apply, women's applications are as apt to be as successful as men's. l3 There are small gender differences in the size of awards that NIH made, but these appear to result from women more often than men applying to programs with small budgets. Access to research funding is crucial to research perfor- !3Tnese findings are consistent with those reported for NIH between l966 and l972 (Douglass and James, l973) and for NSF's program in political science (Sigelman and Scioli, l986). l38

> § UJ 1942 1946 1950 1954 1970 1974 1978 1982 SOURCE: From J. R. Cole and H. Zuckerman, "Marriage, Motherhood and Research Performance in Science," Scientific American 256(2):ll9-l25. Copyrighted l987 by Scientific American, Inc. All rights reserved. Figure 4 Mean cumulative productivity of men and women scientists who earned Ph.D.s in l942 and l970. nance, and not nearly enough is known about patterns of application for funds by men and women, what success they have, how they fare in the budget-cutting process, and how much money they are ultimately awarded. When it comes to rates of publication, more than 50 studies of scientists in a variety of scientific disciplines, types of institu- tions, and different countries show that women publish fewer papers than men of the same ages—on average, 50-60 percent as many (see Cole and Zuckerman, l984, for a review of the studies since l975). The weight of the data is persuasive. Moreover, we find that gender dif- ferences in publication are smaller earlier in the career than later: data on matched pairs of men and women scientists who received Ph.D.s in the same departments in the same years in the same fields, illus- trated in Figure 4, show that gender differentials in publication start early in the career and grow as scientists get older, and this has been so for some time (Cole and Zuckerman, l984). Such differentials are, of course, reduced considerably when rank and type of institution are held constant, but they are not eliminated. Detailed analysis of men's and women's publication patterns indi- cate that the aggregate gender differences are mainly the outcome of differences in the proportions of men and women who publish at a very high rate. As Figure 5 shows, a smaller share of women than men turn up among those who publish large numbers of papers. However, these data, when juxtaposed with those for earlier age cohorts, suggest that this pattern may be changing. Just 8 percent of l39

70- J2 60 - n i 5°- hi o 40- H & 30- A) MEN 20 40 60 80 NUMBER OF PAPERS 100 70- 60- 50 - 40 - 30- 20- 10 - 0 B) WOMEN 20 40 60 80 NUMBER OF PAPERS 100 SOURCE: J. R. Cole and H. Zuckerraan, "The Productivity Puzzle: Per- sistence and Change in Patterns of Publication on Men and Women Scien- tists" in Advances in Motivation and Achievement, P. Maehr and M. W. Steinkamp, eds., Greenwich, Conn.: JAI Press, l984, pp. 2l7-256. Figure 5 Distribution of total publications of men and women scien- tists who earned Ph.D.s from the same departments and in the same fields, l968-l979. women scientists who got their degrees in l957-l958 published as many as 20 papers in the first dozen years of their careers as against 26 percent of women who got their degrees in l970-l97l (Cole and Zuckerman, l984:229). Should these increases continue among still younger cohorts of women scientists, aggregate gender differences in publication will begin to narrow. Figure 5 also shows considerable intra-gender variation in publi- cation. A great many men publish few papers, and a small fraction publish many. The same is true for women. Moreover, the degree of inequality in rates of publication within each gender is much the same. Gender is, therefore, a poor predictor of published research perfor- mance. Knowing whether scientist-authors are men or women will tell little about their rate of publication. The reasons for gender differences in publication are complex and not well understood. Two comparatively simple explanations—(l) that women co-author papers less often and, therefore, publish less than men, and (2) that women have a harder time than men getting papers accepted for publication—do not square with the evidence in hand. On the first, women are as apt as men to publish co-authored papers (Cole and Zuckerman, l984). On the second, the evidence usually cited for this claim is a report showing that "blinded refereeing" (that is, re- moval of author's names from papers) increased the acceptance rate of l40

women's papers submitted for presentation at the meetings of the Modern Language Association (Lefkowitz, l979:56). It is not clear whether data of this sort from the humanities are quite apropos. Still, a systematic assessment of rates of rejection of papers submitted by men and women is in order under refereeing systems in which authorship is both anonymous and identified. Close attention will have to be paid to match submitted papers not just for substantive significance, but also for other attributes (such as research method and specialty) that might affect the probability of acceptance for publication. Increasing disparities in rates of publication by men and women as they grow older may indicate that women's access to resources rela- tive to men's diminishes with time and possibly that their commitment to research and publication wanes as they receive fewer rewards and incentives to continue. These are reasonable conjectures that require further inquiry before they can be properly assessed.l4 Impact or Influence of Research The extent to which scientists' research is cited in the litera- ture is often used as an indicator of the impact or influence of that research. Just as analysts differ on the usefulness of publication counts in assessing role performance, they also differ on the useful- ness of citation counts (see Cole and Cole, l973; Edge, l979; Garfield, l979; Zuckerman, l987). Recent work by Ferber (l986) suggests that citation counts are particularly biased against women. Her studies show a tendency for men to cite men and women to cite women, at least in economics. Given the small number of women in the pool of citing authors, women's papers receive lower rates of citation. This inter- esting idea also needs to be followed up in other fields and in spe- cialties with gender-neutral and gender-related subject matters. How- ever, the high frequency of mixed gender author sets in many sciences should reduce the effects of this bias should it occur outside econom- ics. Other data suggest that women's lower rates of citation have quite a different explanation. Analysis of citations to the work of matched men and women in six sciences indicates that gender differences in citation are a function of their differences in rates of publication [Cole and Zuckerman, l984; Helmreich and Spence (l982) report contrary findings for psychology]. Papers by women authors are cited, on aver- age, just as often as papers by men authors; therefore, aggregate dif- ferences between men and women in numbers of publications account for their differences in citation.l5 and I (l984) have also suggested that women may respond some- what differently than men to positive and negative reinforcement of their work (as gauged by citation in the literature). This hypothesis also deserves further examination. l5As Long noted (see his paper in this volume), this record of equal citation per paper is surprising given the fact that women, in general, have poorer appointments than men (and thus poorer research facilities) and also less active research programs. l4l

Gender, Research Performance, and Rank Are differences between men and women in research performance suf- ficient to explain women's lower ranks and slower rates of promotion? The answer is that they are not. Evidence from carefully analyzed studies by Bayer and Astin (l975) for academics generally, by Reskin (l976; l978) for chemists, and by Cole (l979) for scientists in five fields show, other things being equal, that men and women with equiva- lent records of research performance do not hold the same ranks. Men are apt to hold higher ranks than women and this, Cole (l979:57-58) re- ports, is especially so in the prestigeful research universities. In rank and affiliation, then, comparable men and women do not get equal rewards. And, as I noted earlier, high rank and appointments in uni- versities that facilitate research are, in one sense, rewards, but in another sense, they are resources for further work. To the extent that this is so, these gender differences in rank and affiliation may help account for women's lower aggregate rates of publication later in their careers. Honor and Repute How do women fare relative to men in the allocation of recognition for contributions? Again, the evidence on hand is fragmentary. Cole (l979:Chap. 4) has examined the "reputations" of men and women scien- tists—that is, how visible men and women scientists are to their peers, how their work is assessed, and how often they are named as ma- jor contributors to their fields. Women, he finds, are, on average, less visible than men; their work is perceived to be of lower quality; and they are rarely mentioned as being major contributors [see Davis and Astin (in press) for further evidence on reputation—using differ- ent measures and having somewhat different results]. Cole also finds that while department, rank, awards, age, and doctoral origins contrib- ute to scientists' reputations, research performance (here the extent of publication and citation) is far and away its strongest correlate. Once research performance is taken into account, being a women does not detract from or add to a scientist's reputation (Cole, l979:ll9). How- ever, he also finds that women scientists of the very first rank, Nobel laureates and others of Nobel class, are less well known than men hav- ing the same social statuses (Cole, l979:l20). In the absence of fur- ther evidence, it is difficult to say why this is so. Moreover, Cole (l979:l42) reports that the process by which men and women achieve re- nown appears to differ, again a finding that is not readily explained with the data in hand but plainly important in understanding gender differences in career attainments. Scientists are formally honored by an array of prizes and awards. These may seem trivial, but they are not—for recipients, the groups that give them, or the scientific community at large. They reassure honorees that the work they have done matters, they provide incentives for future work, and they call attention to excellence and indirectly help in the intense competition for resources (Zuckerman, l977). There are no great differences in the sheer number of awards con- ferred on men and women scientists; most have received few or none l42

TABLE l: Women Who are Members of National Academies of Science and Nobelists in the Sciences (number and percent) and Proportions of Women among Holders of Scientific Doctorates in the Selected Countries All Women among Members Women Women Doctorates (N) (N) (%) <%)<* Academic des Sciences (l982) l30 3 2.3 l9.0 (l970) Deutsche Akademie der l,000a 2l 2.l 4.8 Naturf orscher Leopoldina (l973) (l982) National Academy of Sciences l,477b (l986) Royal Society of London (l982) 909 Nobel Laureates (l90l-l986) 39l 50 29 8C 3.4 3.2 2.0 9.8 (l960-l969) 9.3 (l968) aEstimated figure reported by the Akademie. bExcludes foreign associates. Total membership may be slightly smaller owing to delayed reporting of deaths of members. cMadame Curie, laureate in chemistry for l9ll and in physics for l903, is counted twice. Seven different women have been named Nobel laureates. ^Data on proportions of women among holders of doctoral degrees (or in the case of England, holders of "higher diplomas") are given for the years indicated in parenthesis. Those for France, Germany, and England are from the Office of Economic Cooperation and Development, Educa- tional Trends in the l970s: A Quantitative Analysis, New York: United Nations, l984, pp. 26-3l, and for the United States from Lindsey Harmon, A Century of Doctorates: Data Analysis of Growth and Change, Washington, D.C.: National Academy of Sciences, l978, pp. ll7-ll9. eThere are no reliable estimates of the proportion of women in the pool from which Nobel laureates are drawn. (Cole, l979:59). But among those who have, it is difficult to estimate whether women are underrepresented and, if so, by how much. Little is known about the demographic composition of the pools from which award winners of various sorts are drawn. As Table l shows, women now com- prise 2-3 percent of the members of the major academies and 2 percent of Nobel laureates, the top-most levels of the reward system. We know that academicians and laureates are typically selected from among older, full professors or senior researchers in major research univer- sities and that the number of women in those posts is small. Thus, the l43

extent of underrepresentation of women may be less marked than the last column in Table l implies, since the populations of those holding doc- toral degrees in the sciences, country by country, include a larger share of young people and particularly of women than the population of potential laureates and academicians. It is not possible to say, then, whether women are now underrepresented in the major national academies and, if so, by how much. In the United States, we do know that the proportion of women elected recently to the Academy parallels their representation on senior faculties (CEEWISE, l979:l02), suggesting that women are now being elected in proportion to their numbers in the pool. However, this conclusion holds only if it is assumed that women are on senior faculties in the appropriate proportions. Some would question whether this is so. In recent years, however, the number of women elected to the U.S. National Academy of Sciences has risen sharply, by five times since l972, while the membership as a whole has grown l.6 times. Moreover, research done some years ago on the ages of members at the time they were elected to the Academy (l865-l969) showed that women were elected nine years later than men in the same fields, yet there was no reason to believe that the research that led to their election was done any later (Zuckerman and Cole, l975:98). Comparable data for the period since then shows that the average age of women at the time of election has been dropping, though it has not quite reached the averages for men. On the basis of this limited evidence, it appears that two of the general patterns noted at the outset hold here: cross-sectional dif- ferences persist between men and women in reputation with the evidence on awards being scanty but providing some indication of growing gender parity. The absence of longitudinal data on awards that scientists receive over the course of their careers makes it impossible to say whether gender disparities grow as scientists age; but the overall tendency, observed among men, of awards being conferred on those who already have them (Zuckerman, l977) would seem to suggest this might be so. Taken together, what does the evidence tell us about career dif- ferences of men and women scientists and engineers? As I indicated at the outset, it provides a fragmentary quantitative picture focused mainly on scientists, academics, doctorates, and researchers in the current period. As a consequence, not much is known systematically about the past or about careers of the majority of scientists who work in industry and government or of engineers. What is known suggests that the general patterns noted earlier seem to hold: women scientists and engineers in industry and government occupy lower ranks, advance more slowly, are paid less, and have less managerial responsibility than men (CEEWISE, l980), but nothing is known about their relative role performance. Among academics, a consistent picture emerges: women scientists' career attainments, for the most part, do not match those of men. In some measure, these differences are attributable to their being younger, having different work histories, and being in different fields than men. We have also seen that they get degrees from equally distinguished institutions, have had postdoctoral fellow- l44

ships equally often, and now are proportionately represented among assistant professors. Older cohorts of women scientists, however, differ from men in rank, salary, research performance, and reputation, with these differences being almost always in the direction of compara- tive disadvantage for women. However, there are also signs of growing parity in career attainments, particularly among younger scientists. It is not yet clear whether this trend will ultimately dampen the third pattern, observed earlier, of increasing disparity in men's and women's attainments as they grow older. Some Explanations Proposed for Gender Differences in Career Attainments There is little disagreement that women scientists' career attain- ments, on average, do not equal those of men. There is, however, much disagreement about why this is so. In general, these fall into four classes that emphasize the following: • Gender differences in scientific ability, • Gender differences arising from social selection, based on (a) gender discrimination or (b) gender differences in role per- formance and the allocation of resources and rewards in accord with these differences, • Gender differences arising from self-selection, including (a) marriage and motherhood and their consequences and (b) gender differences in career commitment, and • Outcomes of accumulation of advantage and disadvantage. How well does the available evidence square with each of these ex- planations? Not well. The evidence on all is ambiguous, not because the theories are unclear but because the data are complex, often vex- ingly incoherent, and frequently partial. Gender Differences in Scientific Ability There is no support, as I noted earlier, for the claim that the different career attainments of men and women scientists result from gender differences in ability or competence. To the extent that these can be measured by intelligence tests or academic performance, women's abilities equal or surpass those of men. However, there is evidence that girls do less well than boys in mathematics and also that girls turn up in disproportionately small numbers among youngsters with high scores on tests of mathematical ability.i6 Benbow and Stanley (l980; l983) conclude that "superior male mathematical ability, an expression of both endogenous and exogenous variables," accounts for this finding. jane B. Kahle and Marsha L. Matyas, "Equitable Science and Mathematics Education: A Discrepancy Model," included earlier in this volume. l45

However, there is also enough evidence for marked gender differences in socialization and exposure to mathematicsl? to raise serious ques- tions about this and to warrant further examination of their effects on variability in achievement scores of boys and girls. For our pur- poses, these findings are not quite entirely pertinent. Men and women who do science are a highly selected sample of all adult men and women. Data on youngsters, even highly selected ones, are not helpful in un- derstanding field and specialty choice, much less differences in adults'career attainments. Yet women do, more often than men, select fields and specialties of science that are comparatively less demanding mathematically, but there is no systematic evidence available indi- cating why this is so. Processes of Social Selection Explanations of gender differences in career attainments that em- phasize gender discrimination, on the one hand, and women's poorer re- search performance, on the other, both rest on notions of social selection. Social selection processes involve decision making about individuals (here, about their careers) over which they have limited control. They contrast with processes of self-selection in which decisions are controlled by individuals and are not, except perhaps indirectly, attributable to socially structured arrangements for selec- tion. In practice, social selection and self-selection are inter- twined; but in principle, they are and should be analytically separate. When scientists apply for research support, they engage in self-selec- tion; some apply, some do not. Those who do apply subject themselves to social selection—in this instance, to peer review. It is plainly useful to know whether men and women differ in patterns of self- and social selection and, if so, why. It is not useful to assume that gender differences arise exclusively from one or the other selective process. Gender discrimination occurs when unequal treatment of men and women is based on the functionally irrelevant criterion of gender rather than on functionally relevant criteria such as role performance. Discrimination can affect men's and women's career attainments when their opportunities for role performance are unequal, when the same quality of role performance is judged according to different standards, and when the same quality of role performance evokes different rewards. Gender discrimination, as with social discrimination generally, treats some social status "as relevant when intrinsically it is functionally irrelevant" (Merton, l972:20). Proponents of the view that gender discrimination best explains the unequal career attainments of men and women point to instances of women having poorer facilities and re- sources for research, to their being judged by harsher standards, and to their being promoted and paid less than comparable men. jane B. Kahle and Marsha L. Matyas, "Equitable Science and Mathematics Education: A Discrepancy Model," included earlier in this volume. l46

It is no easy matter, however, to assess the extent of gender dis- crimination and how it affects scientists' careers. Discrimination is often subtle and, therefore, difficult to identify, much less measure. It can be entangled with other forms of particularism (age, for exam- ple); and because it appears throughout the career, a full accounting of its effects is hard to make. As a consequence, researchers have come to rely on indirect rather than direct measures of discrimination. They have assumed that differences in the career attainments of men and women that remain after taking all functionally relevant criteria into account (that is, after holding all relevant variables constant) must be the outcome of gender discrimination. Measuring discrimination by the use of residual differences has its problems, not the least that it requires that appropriate evidence be available on all functionally relevant criteria that could account for gender differences in career attainment (on "residualism" in its various guises in the law and in social science research, see Cole, l979:36ff). When this mode of analysis is used, the evidence shows that gender discrimination affects promotions, tenure, and salary allocation among academic men and women with similar records of research performance. It also shows that discrimination is receding, especially for younger women. However, available data are limited only to position and to salary and do not register gender discrimination in informal social interaction or its subjective effects on women (see, for example, Briscoe, l984; Keller, l977). There is also evidence suggesting that gender discrimination may operate in different ways for different groups of women. Women who make important contributions to science may fare less well relative to comparable men than do the journeywomen of science relative to their performance peers. Cole's (l979:l20) studies show that women who have done important scientific work are less apt to be considered major contributors to their fields than are comparable men. Conversely, it has also been suggested that it is the journeywomen of science who fare poorly compared to men in the absence of any clear-cut evidence of re- search performance (Zuckerman and Cole, l975). Thus, the incidence and dynamics of gender discrimination in science have neither been satis- factorily described nor fully explained. Discrimination need not be reflected only in differences in career outcomes but can also be found in differential processes of status attainment. Reskin (l978) and Reskin and Hargens (l979) suggest that the connections between role performance and rewards for women are less consistent than those that apply in men's careers; women are more often rewarded for poor performance and less often for good performance. And, as noted, Cole (l979) observes that the processes of reputation building differ for men and women. If so, then women do not have as clear incentives for role performance as men and may, as a result, perform less well. Social selection, as I have indicated, also includes differential evaluation of men's and women's role performance on functionally rele- vant criteria and differential treatment on this basis. Some believe that women are judged fairly and that gender differences in career at- tainments result from their performing less well than men. As I have l47

repeatedly noted, we do not know how men and women do in their various roles as teachers, administrators, managers, and citizens of the sci- entific community. We do know that, on average, women publish less and are cited less than men. To the extent that these actually gauge re- search performance, then women's poorer performance is related to their lower career attainments, especially in those institutions that put a premium on publication. However, as we have also seen, women's re- search performance does not explain all such differences; indeed, some remain after research performance and many other variables are taken into account. Moreover, the conclusion that women perform less well than men does not take into account unequal opportunities to do re- search. The sources of differential role performance may well reside in structured inequalities of opportunity. How much, we do not know. Gender Differences as Outcomes of Self-Selection Career attainments are, of course, also shaped by decisions indi- viduals make for themselves, by self-selection as well as by social selection. Women's decisions to marry and to have children, to take on their distinctive domestic and parental roles, are said to interfere with their scientific work and to lead to career decisions that benefit their families but damage their careers. The evidence here is mixed: • The work histories of women scientists and engineers differ from those of men. They are more often employed part-time or not seeking work and are out of work longer than men. Women attribute these work patterns to their family obligations more often than men. However, actual family obligations (having young children) are a poor predictor of unemployment among women scientists and engineers.l8 • Married and single women academics are less mobile geographi- cally than men (Harwell, et al., l979), and married women say that mobility decisions are affected by their family obligations (Coggeshall, 198l; Rosenfeld, l98l). Since promotion and pay increases are often tied to changing jobs, women's limited geo- graphic mobility may, in part, account for gender differences in career attainments. • Marriage and motherhood are widely believed to account for women scientists' lower rates of publication (e.g., Lester, l974:42). However, on balance, the evidence suggests that this is not so. Married women Ph.D.s in the sciences publish as much as single women and having successive children is not associated with re- capital economists emphasize that women's lower educational attainment, intermittent work histories, and part-time employment ac- count in large measure for gender differences in occupational attain- ments in the work force at large. Indeed, they do account, in part, for the gross cross-sectional differences observed here. They cannot be the whole explanation, however, since gender differences in career attainments hold within groups with the same human capital investments. l48

duced rates of publication (Astin and Bayer, l979; Centra, l974; Cole, l979; Cole and Zuckerman, l986; Ferber and Huber, l979; Hargens, et al., l978; Helmreich, et al., l980; Simon, et al., l966; and Wanner, et al., l98l). • Moreover, marriage and parenthood are not associated with lower rank and salary among women, in some fields and classes of in- stitutions, the correlations are positive and in others nega- tive. Overall, however, they are not large (Ahern and Scott, l98l:Chap. 6). This is so in spite of the widespread belief that married women have poorer career opportunities than single women. In short, women's domestic obligations are not the simple explanation of gender differences in career attainments since, in many respects, married women and women with children fare as well or better than single and childless women. However, the diverse career consequences of marriage and parenthood for men and women have not been identified in anything like the needed detail. The judgment on this explanation is still out. Gender Differences in Career Commitment There is little or no systematic data on the career commitments of men and women scientists and engineers, especially those holding doctoral degrees. That is, there is no evidence bearing on whether women care less, more, or the same as men about rank, salary, responsi- bility, and recognition. There is indirect evidence that the prefer- ences of academic men and women differ: for teaching as against re- search [or at least they did two decades ago (Bayer, l970)] and for living in urban as against less populous areas (Harwell, et al., l979). But the connections between such career-related preferences and career commitments are not established, nor are the ways these preferences are shaped by opportunities perceived and real. This hypothesis lacks any support, pro or con. Accumulation of Advantage and Disadvantage The notion of the accumulation of advantage and disadvantage has been repeatedly used in studies of stratification in science (Allison and Stewart, l974; S. Cole, l970; Cole and Cole, l973; Hargens, et^ al., l980; Long, l978; Mittermeir and Knorr, l979; Price, l976; and Zuckerman, l970, l977, l979). It is plainly pertinent also to the disparity observed in the attainments of men and women scientists. Accumulation of advantage refers to social processes by which certain groups receive greater resources and rewards, such that recipients are enriched at an accelerating rate and, conversely, non-recipients become relatively impoverished [see Merton, (l942) l973; Zuckerman, l977:59- 60]. When greater resources consistently go to those who can use them effectively—that is, when they are allocated on functionally relevant criteria, the gap in performance that separates recipients from non- recipients grows increasingly large (Zuckerman, l977:248ff). Advantage l49

also accumulates for certain groups when resources are allocated to them on functionally irrelevant criteria, such as gender. The result- ing disparity in achievement between the "haves" and "have nots" is not as great as when functionally relevant criteria are applied, but in both instances the "haves" perform better than the "have-nots" and are rewarded more copiously. The idea of accumulation of advantage helps to account for the observed cross-sectional differences between men and women scientists in research performance, rewards, and recognition, for observed intra-gender variation (since not all women are equally dis- advantaged nor are all men equally advantaged), and (not least) for the growing divergence in performance and attainments of men and women sci- entists as they grow older. It is also consistent with or, more pre- cisely, does not exclude, the third pattern we observed of growing con- vergence in attainments of men and women, especially among the young. To the extent that processes of accumulation of advantage and disadvantage are supplemented by self-selection, by women making deci- sions that benefit their families but have the effect of damaging their careers, disparities between their attainments and those of men will be amplified and accentuated. Although the ideas of accumulation of advantage and disadvantage have been further elaborated and examined empirically not just in science but in other occupations (e.g., Broughton and Mills, l980; Clark and Corcoran, l986), detailed evi- dence is needed on the constituent processes. Although differential access to resources plays a central role in the theory, not much is known directly about how differential access to resources affects per- formance nor is much known about how rewards are transformed into re- sources. Before we can conclude that accumulation of advantage and disadvantage and related processes of self-selection really do explain why the career attainments of men and women differ and why these dif- ferences increase with time, these gender differences in the allocation of resources and rewards need to be examined. A Limited Research Agenda: Domains of Specified Ignorance Finally, a limited number of questions should be earmarked as do- mains of specified ignorance (Merton, l987), what we now know that we need to know and why: l. On the research performance of men and women scientists: We need to know why women publish less than men and the extent to which this results from discrimination, differential access to the means of scientific production, and women's preferences or choices. More specifically, we need to know the relative access men and women have to such important research resources as funds, space, appropriate co-workers, and instrumentation. How do women's organizational ranks and institutional affiliations, in combination, affect their resources and their research per- formance over the course of their careers? l50

We also need to learn whether there are significant differ- ences in the research strategies and practices used by men and women scientists. Are there greater differences between men and women in these respects than among them? If so, do they affect how much research is done and its impact? 2. On disparities in career attainments of men and women: Why do these disparities grow as men and women get older? Does the accumulation of advantage explain this pattern fully, or is there evidence also for other explanations such as women's grow- ing discouragement and reduced aspirations? [See Zuckerman and Cole (l975) on the "Triple Penalty" against women, which links discrimination to lowered aspirations.] Longitudinal studies on multiple age cohorts are needed to answer these questions.l9 Why do women fare better in certain sciences and less well in others? Is it the case, as Rossiter (l978) suggests, that women do better in new and growing fields? Are the "cultures" of some fields more consistent with feminine values than others, as Keller (l985) and Traweek (l984) imply? 3. On gender discrimination—its incidence, forms, and consequences: To what extent is discrimination conditional or practiced against all women, regardless of their status characteristics? How does discrimination in its less blatant forms affect women's informal relations with their colleagues and their networks of associa- tions? Rose (l985) observes that the networks of young men and women scientists differ not only in composition but also in how useful they are believed to be. More research attention needs to be paid to the consequences of informal associations for scien- tists' careers. And, finally, how does the experience of gender discrimination affect women's motivation and career commitment? Do men have equivalent experiences not associated with gender that have similar effects? 4. On changing labor markets: In what measure are the career attainments of men and women sci- entists and engineers determined by changing labor market condi- tions? Is the move toward gender parity likely to be permanent, or will it wane if jobs become scarce? 5. On the career consequences of marriage and parenthood: To what extent is women's limited geographic mobility (perceived and real) related to their poorer attainments? How do the re- *lt would be useful to have similar studies done on women and men in other professions, since there is reason to believe that gender disparities in the career attainments of lawyers and managers also grow as they age [C. Epstein, l987 (private communication); Gallese, l985; White, l967]. l5l

quirements of dual-career couples affect their attainments? This is of no small moment given the fact that a majority of married women scientists are married to men scientists. 6. On commitment to careers: So little is known about the career commitment of men and women scientists (and other professionals) that this is a thoroughly uncharted domain. It would indeed be useful to know whether men and women differ or are the same with respect to career aspira- tions, concern with promotion, income, and fame. Believing that they do, some attribute gender differences in career attainments to these attitudinal differences. Others contend that such dif- ferences are negligible and that structural barriers faced by women account for differences in career attainments. In either event, this domain of specified ignorance about attitudes requires further examination and needs to be linked to evidence on the behavior of men and women scientists. So much for the future research agenda. How much have women sci- entists' careers changed since the Sabin period? They have changed considerably, but it has taken a long time. Sabin1s career, then a succession of "firsts," is now far less atypical. It would have been inconceivable in l925 that women scientists would be hired as faculty members by universities at about the rate they were getting Ph.D.s, and it would have been thought highly unlikely that women would be promoted into senior posts in all classes of universities at the rate they have. At the same time, their career attainments continue, on average, to be more modest than those of men in all sectors—academia, industry, and government—and the gap in attainments grows as men and women age. Moreover, while distinguished women scientists and engineers have be- come insiders and members of the scientific establishment, some still consider themselves to be outsiders and on the margin. It is not clear at this juncture whether parity will be achieved in the careers of men and women scientists and engineers and, if so, when. It is clear that equal attainments, on average, will require equal access to opportunity for high-level role performance, it will also require the application of functionally relevant criteria first in assessing role performance and then in the allocation of resources and rewards. Bibliography Ahern, N. C., and E. L. Scott. l98l. Career Outcomes in a Matched Sample of Men and Women Ph.D.s: An Analytical Report. Washington, D.C.: National Academy Press. Allison, P. D., and J. A. Stewart. l974. Productivity differences among scientists: Evidence for accumulative advantage. American Sociological Review 39:596-606. Astin, H. S. l969. The Women Doctorate in America. New York: Russell Sage Foundation. Astin, H. S., and A. E. Bayer. l979. Pervasive sex differences in the academic reward system: Scholarship, marriage, and what else? In l52

Academic Rewards in Higher Education, D. R. Lewis and W. E. Becker, eds. Cambridge, Mass.: Bal linger Publishing Co., pp. 2ll-229. Bayer, A. E. l970. College and University Faculty: A Statistical De- scription. ACE Research Reports 5:5. Washington, D.C.: American Council on Education. Bayer, A. E., and H. S. Astin. l975. Sex differentials in the academic reward system. Science l88:796-802. Bayer, A. E., and J. Folger. l966. Some correlates of a citation mea- sure of productivity in science. Sociology of Education 39:38l- 390. Benbow, C. P., and J. C. Stanley. l980. Sex differences in mathemati- cal ability: Fact or artifact? Science 2l0:l262-l264. Benbow, C. P., and J. C. Stanley. l983. Sex differences in mathemati- cal ability: More facts. Science 222:l029-l03l. Breiger, G. H. l980. Florence Rena Sabin. In Notable American Women: The Modern Period, B. Sicherman, C. H. Green, I. Kantrov, and H. Walker, eds. Cambridge, Mass.: Belknap Press of Harvard Univer- sity, pp. 6l4-6l7. Briscoe, A. M. l984. Scientific sexism: The world of chemistry. In Women in Scientific and Engineering Professions, V. B. Haas and C. C. Perrucci, eds. Ann Arbor: University of Michigan Press, pp. l47-l59. Broughton, W., and E. W. Mills. l980. Resource inequality and accumu- lative advantage: Stratification in the ministry. Social Forces 58:l289-l30l. Centra, J. A., with N. M. Kuykendall. l974. Women, Men and the Doc- torate. Princeton, N.J.: Educational Testing Service. Clark, S. M., and M. Corcoran. l986. Perspectives on the professional socialization of women faculty: A case of accumulative disadvan- tage? Journal of Higher Education 57:20-43. Coggeshall, P. E. l98l. Postdoctoral Appointments and Disappointments. Washington, D.C.: National Academy Press. Cole, J. R. l979. Fair Science: Women in the Scientific Community. New York: Free Press. Cole, J. R., and S. Cole. l973. Social Stratification in Science. Chicago: University of Chicago Press. Cole, J. R., and H. Zuckerman. l984. The productivity puzzle: Persis- tence and change in patterns of publication of men and women sci- entists. In Advances in Motivation and Achievement, P. Maehr and M. W. Steinkamp, eds. Greenwich, Conn.: JAI Press, pp. 2l7-256. Cole, J. R., and H. Zuckerman. l987. Marriage and motherhood and re- search performance in science. Scientific American 256(2):ll9-l25. Cole, S. l970. Professional standing and the reception of scientific discoveries. American Sociological Review 76:286-306. CEEWISE (Committee on the Education and Employment of Women in Science and Engineering), National Research Council. l979. Climbing the Academic Ladder: Doctoral Women Scientists in Academe. Washing- ton, D.C.: National Academy of Sciences. CEEWISE. l980. Women Scientists in Industry and Government: How Much Progress in the l970s? Washington, D.C.: National Academy of Sciences. l53

CEEWISE. l983. Climbing the Ladder: An Update of the Status of Doc- toral Women Scientists and Engineers. Washington, D.C.: National Academy Press. Davis, D., and H. S. Astin. In press. Reputational standing in aca- deme. Journal of Higher Education. Douglass, C. D., and J. C. James. l973. Support of new principal in- vestigators by N.I.H.: l966-l972. Science l8l(July 20):24l-244. Edge, D. l979. Quantitative measures of commmunication in science: A critical review. History of Science l7:l02-l34. Ferber, M. l986. Citations: Are they an objective measure of work of women and men? Signs ll:38l-389. Ferber, M., and J. Huber. l979. Husbands, wives, and careers. Journal of Marriage and the Family 4l:3l5-325. Fox, M. F. l985. Location, sex-typing, and salary among academics. Work and Occupations l2:l86-205. Gallese, L. R. l985. Women Like Us. New York: William Morrow. Garfield, E. l979. Is citation analysis a legitimate evaluation tool? Scientometrics l:359-375. Garfield, E. l982. The l,000 most-cited contemporary authors, Part 2A: Details on authors in the physical and chemical sciences and some comments about Nobels and Academy memberships: Current Contents (March l):5-l3. Gaston, J. l970. Originality and Competition in Science: A Study of the British High Energy Physics Community. Chicago: University of Chicago Press. Hargens, L., J. McCann, and B. Reskin. l978. Productivity and repro- ductivity: Marital fertility and professional achievement among research scientists. Social Forces 52:l29-l46. Hargens, L., N. Mullins, and P. Hecht. l980. Research areas and strat- ification processes in science. Social Studies of Science l0:55-74. Harmon, L. R. l978. A Century of Doctorates: Data Analyses of Growth and Change. Washington, D.C.: National Academy of Sciences. Helmreich, R. L. and J. T. Spence. l982. Gender differences in pro- ductivity and impact. American Psychologist 37:ll42. Helmreich, R. L., J. T. Spence, W. E. Beane, G. W. Lucker, and K. A. Matthews. l980. Making it in academic psychology: Demographic and personality correlates of attainment. Journal of Personality and Social Psychology 39:896-908. Keller, E. F. l977. The anomaly of a woman in physics. In Working It Out, S. Ruddick and P. Daniels, eds. New York: Pantheon Books, pp. 77-9l. Keller, E. F. l985. Reflections on Gender in Science. New Haven, Conn.: Yale University Press. LeBold, W. K., K. W. Linden, C. M. Jagacinski, and K. D. Shell. l983. National Engineering Career Development Study: Engineers Profiles of the Eighties. West Lafayette, Ind.: Purdue University. Lefkowitz, M. R. l979. Education for women in a man's world. The Chronicle of Higher Education (August 6):56. Lester, R. A. l974. Antibias Regulation of Universities: Faculty Problems and Their Solutions. New York: McGraw-Hill. l54

Lewis, G. L. l986. Career Interruptions and Gender Differences in Salaries of Scientists and Engineers. Working paper prepared for the Office of Scientific and Engineering Personnel. Washington, D.C.: National Research Council. Long, J. S. l978. Productivity and academic position in the scien- tific career. American Sociological Review 43:889-908. Long, J. S., P. D. Allison, and R. McGinnis. l979. Entrance into the academic career. American Sociological Review 44:8l6-830. Long, J. S., and R. McGinnis. l98l. Organizational context and scien- tific productivity. American Sociological Review 46:422-442. Harwell, G., R. Rosenfeld, and S. Spilerman. l979. Geographic con- straints on women's careers in academia. Science 205:l225-l23l. Merton, R. K. l957. Priorities in scientific discovery: A chapter in the sociology of science. American Sociological Review 22:635-659. Merton, R. K. l968. The Matthew Effect in science. Science l59: l56- l63. Merton, R. K. l972. The perspectives of insiders and outsiders. Amer- ican Journal of Sociology 77:9-47. Merton, R. K. [l942] l973. The normative structure of science. In The Sociology of Science. Chicago: University of Chicago Press, pp. 267-278. Merton, R. K. l987. Three fragments from a sociologist's notebooks: Establishing the phenomenon, specified ignorance, and strategic research materials (SRMs). Annual Review of Sociology l3:(in press). Mittermeir, R. and K. D. Knorr. l979. Scientific productivity and ac- cumulative advantage: A thesis reassessed in the light of inter- national data. R&D Management 9:235-239. NIH (National Institutes of Health), Special Programs Office, Office of Extramural Research and Training. l98l. Women in Biomedical Research. Publication No. 8l-249. Washington, D.C.: The Insti- tutes. NSB (National Science Board). l977. Science Indicators l976. NSB 85-l. Washington, D.C.: U.S. Government Printing Office. NSB. l985. Science Indicators. Washington, D.C.: U.S. Government Printing Office. NSF (National Science Foundation). l986. Women and Minorities in Sci- ience and Engineering. Washington, D.C.: U.S. Government Printing Office. OECD (Office of Economic Cooperation and Development). l984. Educa- tional Trends in the l970s: A Quantitative Analysis. New York: United Nations. Perrucci, C. C. l970. Minority status and the pursuit of professional careers: Women in science and engineering. Social Forces 49:245- 259. Price, D. J. de S. l976. A general theory of bibliometric and other cumulative advantage processes. Journal of the American Society for Information Science 27:292-306. Reskin, B. F. l976. Sex differences in status attainment in science: The case of postdoctoral fellowships. American Sociological Re- view 4l:597-6l2. l55

Reskin, B. F. l978. Scientific productivity, sex and location in the institution of science. American Journal of Sociology 83:l235- l243. Reskin, B. F., and L. L. Hargens. l979. Scientific advancement of male and female chemists. In Discrimination in Organizations, R. Alvarez, H. Lutterman, and associates, eds. San Francisco: Jossey- Bass, pp. l00-l22. Rose, S. M. l985. Professional networks of junior faculty in psychol- ogy. Psychology of Women Quarterly 9:533-547. Rosenfeld, R. l98l. Academic men and women's career mobility. Social Science Research l0:337-363. Rossiter, M. l978. Sexual segregation in the sciences: Some data and a model. Signs 4:l46-l5l. Rossiter, M. l982. Women Scientists in America: Struggles and Strate- gies to l940. Baltimore, Md.: Johns Hopkins Press. Russo, N. F. and A. N. O'Connell. l980. Models from our past: Psychol- ogy's foremothers. Psychology of Women Quarterly 5:ll-54. Sigelman, L. , and F. P. Scioli, Jr. l986. Retreading familiar terrain: Bias, peer review and the NSF political science program. Unpub- lished manuscript. Simon, R. J., S. M. Clark, and L. L. Tifft. l966. Of nepotism, mar- riage, and the pursuit of an academic career. Sociology of Edu- cation 39:344-358. Syverson, P. D. l980. Summary Report l979: Doctorate Recipients from United States Universities. Washington, D.C.: National Academy of Sciences. Traweek, S. l984. High energy physics: A male preserve. Technology Review 87(Nov/Dec):42-43. Vetter, B. l98l. Women scientists and engineers: Trends in participa- tion. Science 2l4:l3l3-l32l. Vetter, B., and E. M. Babco, eds. l986. Manpower Comments 23:7. Wash- ington, D.C.: Commission on Professionals in Science and Technol- ogy. Wanner, R. A., L. S. Lewis, and D. I. Gregorio. l98l. Research pro- ductivity in academia: A comparative study of the sciences, social sciences, and humanities. Sociology of Education 54:238-253. White, J. l967. Women in the law. Michigan Law Review 65:l05l. Zuckerman, H. l970. Stratification in American science. Sociological inquiry 40:235-257. Zuckerman, H. l977. Scientific Elite: Nobel Laureates in the United States. New York: Free Press. Zuckerman, H. l987. Citation analysis and the complex problem of in- tellectual influence. Scientometrics ll(in press). Zuckerman, H., and J. R. Cole. l975. Women in American science. Minerva l3:82-l02. Zuckerman, H., and J. R. Cole. l987. Marriage, motherhood and research performance in science. Scientific American 256(2):ll9-l25. l56

PROBLEMS AND PROSPECTS FOR RESEARCH ON SEX DIFFERENCES IN THE SCIENTIFIC CAREER* /. Scott Long Introduction The literature on the graduate education and careers of women in science is diverse, rich, and complex. Unfortunately, the literature is also fragmented and contradictory, with fundamental questions re- maining unanswered. Many of the limitations of this literature can be attributed to a set of common, methodological flaws. if research in this area is to contribute fully to understanding the scientific career, future research must be guided by an understanding of the com- plexity of the scientific career and the implications of this complex- ity for the design of research. To provide a framework for a methodological critique of the lit- erature on women in science, two classes of facts, related to partici- pation and to performance, are suggested and must be explained. Explanations for these facts are then reviewed and critiqued, with a focus on the methodological limitations of past research. The Facts To Be Explained The preceding reviews by Hornig and Zuckerman suggest two sets of findings that are fundamental, clearly established, and demanding of explanation. The first set involves the demographics of science; the second set involves gender differences in scientific performance. Participation in Science There are two components of participation in science that require explanation. First, being a female decreases one's chances of obtain- ing a degree in science. Second, if a degree is obtained, being a fe- male decreases the chances of the degree-holder progressing through the "normal" stages of a career, where the normal career may be thought of *The author would like to thank Lowell L. Hargens, Lilli Hornig, Robert McGinnis, and Rachel A. Rosenfeld for their comments on an earlier version of this paper. l57

as a typical male career. In short, women are less likely to be sci- entists, either through failure to obtain degrees or to keep mainline scientific positions. Arlie Russell Hochschild (l975) quotes figures from the University of California at Berkeley that highlight the demographics of the aca- demic sector: Why, at a public university like the University of Cali- fornia at Berkeley in l972, do women compose 4l percent of the entering freshman, 37 percent of the graduating seniors, 3l percent of the applicants for admission to graduate school, 28 percent of the graduate admissions, 24 percent of the doctoral students, 2l percent of advanced doctoral students, l2 percent of Ph.D.s, 38 percent of instructors, 9 percent of assistant professors, 6 percent of associate professors, and 3 percent of full professors? These figures give an exaggerated impression of attrition since the increasingly smaller percentages of women found in the higher ranks reflects the smaller percentages of women obtaining Ph.D.s in the past as well as the higher attrition of women. Nonetheless, they vividly portray the fact that being a women decreases one's chances of a suc- cessful career in science. Recent figures suggest that the percentage of degrees in science and engineering that are given to females has increased dramatically. Yet there is little evidence that the differences in attrition between men and women after the receipt of the degree have declined. if this is true, the gains made in granting degrees to females may have sub- stantially less impact on science than they might. Performance and Recognition The basic facts regarding the performance of women in science are also clear. First, the performance of women in the coursework and in the standardized examinations leading up to the Ph.D. are, as noted by Hornig, superior to the performance of men. Second, women are less productive than men as indicated by numbers of papers published and citation received. The average female Ph.D. publishes fewer articles and receives fewer citations per year than the average male Ph.D. This finding of lesser productivity must be understood in the context of a third, intriguing, although less well-established dimension of pro- ductivity. Namely, the average female scientist receives fewer cita- tions for her work over the course of her career and often finds her- self in marginal scientific employment, but individual papers by women receive as many citations as individual papers by men. Research I have recently completed, based on nearly 40,000 articles published by 550 male and 600 female biochemists, indicates that while the average male published approximately 60 percent more articles than the average fe- male, the average publication by a female received approximately 20 percent more citations than the average paper by a male. l58

Each of these facts is important for understanding the position of women in science. It is a mistake to focus solely on the lesser pro- ductivity of women. The Explanations Research on gender differences in science must come to an under- standing of the processes that generate differential participation and performance. Yet in going beyond the statement of these basic facts, research is, to borrow a phrase from Zuckerman's paper, "vexingly incoherent and frequently partial." The explanations that have been provided are less convincing than the facts that require explanation. Hornig's and Zuckerman's reviews have suggested five potential sources of gender differences in participation and performance can be isolated: (l) differences in ability; (2) processes of evaluation governed by both universalistic and particularistic criteria; (3) dif- ferences in motivation and dedication; (4) role expectations of women with respect to the family; and (5) processes of cumulative advantage and disadvantage. While these explanations are distinguished for pur- poses of presentation, they should not be viewed in isolation. They operate together, reinforcing one another, and consequently must be understood as a whole. Further, none of these explanations is fully satisfactory. In most cases the evidence is mixed. Ability Hornig and Zuckerman point out that there is no evidence that males have more ability than females. Not only are differences between sexes absent in measured ability, but what research there is suggests that variation in such ability does not explain variation in produc- tivity within sexes (Cole and Cole, l973:67, 68). It is important to note that both reviews were careful to add the qualification that ability is being operationalized as performance on standardized exams and coursework. This suggests that differences in ability may still be a useful explanation of performance if the definition is broadened to go beyond that commodity that allows a person to fill in the correct circles on an examination. This broader definition involves the concept of tacit knowledge. While graduate education involves the mastery of the technical knowledge that is often measured by standardized exams, there is also an emphasis on the transmission of tacit knowledge. Such knowledge results in a different sort of ability that is essential for scien- tific success. Michael A. Overington (l977:l45) explains tacit knowl- edge in this way: Tacit knowledge—requires an engagement in research on the model of some skillful practitioner in whose person there is incarnated both the general culture of science and par- ticular traditions within the culture. One can no more l59

discover the culture of scientific research from its written results than one can construct a Stradivarius from measure- ments of an original. Tacit knowledge, by its very nature, is difficult to measure. None- theless, that dimension of ability that reflects mastery of the tacit knowledge of a discipline should not be dismissed as an explanation of differences in scientific performance. Given that tacit knowledge is often transmitted through informal contacts, female graduates may be at a disadvantage, in part due to discrimination based on sex. Discrimination and Evaluation Differences in performance and participation may result from the evaluation of females by members of the scientific community. To the extent that universalistic characteristics are used in the evaluation, the system is fair and operates as a meritocracy. To the extent that particularistic characteristics are used, discrimination occurs. Discrimination exists in both blatant and subtle forms. Hornig in this volume describes discrimination in funding during graduate school. Vivian Gornick (l983) documents many cases of a more virulent discrimination by exclusion, where males simply deny females the right to be full-fledged scientists. In addition to blatant forms of discrimination, it is likely that subtle forms discourage women from active careers and keep them from being as productive as their male counterparts. Subtle forms of discrimination undoubtedly occur in interpersonal relationships. It may occur through disparaging remarks about a woman's ability to have a family and be a scientist. it can occur when females are excluded from "extra-scientific" activities that may result in the transmission of essential, tacit knowledge. The importance of more subtle forms of discrimination, which may appear to some to be minor inconveniences, is through processes of cumulative advantage, which are discussed below. While discriminatory evaluations may be the focus of gender dif- ferences in science, it is important to keep in mind that differences in advancement and participation can also result from universalistic evaluations. For example, slower advancement in rank may result from the lesser seniority and fewer publications of women. However, such explanations require the answer to the prior question of why women differ in either seniority or publications. Motivation and Dedication Differences in motivation are undoubtedly important in explaining the fewer females than males applying to graduate school. Evelyn Fox Keller (l977:l30), in her feminist critique of science, has suggested that "perhaps the single most powerful inhibitor [of women entering science] was the widespread belief in the intrinsic masculinity of scientific thought." l60

Differences in motivation and dedication may also explain differ- ences in performance and career advancement. While Zuckerman reports no strong evidence for this, this possibility requires further consid- eration within the context of domestic obligations. Further, if a woman drops out of science due to lack of motivation or dedication, it is important to know whether discrimination or demands of family re- quired levels of dedication exceeding those of the most dedicated male. Domestic Obligations Differences in motivation and dedication are likely to arise from conflicts between the demands of family and the demands of science. While there is evidence that women are more strongly affected by family and domestic obligations, the consequences of these obligations are unclear in the literature and, I believe, underestimated. Consider the two seemingly contradictory effects of females' greater family obligations that were reported by Zuckerman. First, family obligations cost women in terms of underemployment and unem- ployment. Second, marriage and motherhood are not associated with lower levels of publication among female scientists. Married women and women with children publish as much or more, on average, as single and childless women. Further, parenthood is positively correlated with rank and salary for women as well as men. These conclusions seem contradictory and contrary to much of what we know about the institution of science and the institution of the family. First, science demands a great deal of time. Arlie Hochschild (l975) has made this point with respect to the cost of raising a family: The academic career is based on peculiar assumptions about the relationship between work and competing with others, competing and getting credit, getting credit and building a reputation, building a reputation and doing it while young, doing it while young and hoarding time, hoarding time and minimizing family life, minimizing family life and leaving it to your wife. Even if the meritocracy worked perfectly, even if women did not cool themselves out, I suspect there would remain in a system that defines careers this way only a handful of women at the top. Given that the number of hours a week may be an important factor af- fecting productivity (Cole and Cole, l973:7l),l and to the extent that greater domestic demands are placed on women, it is hard to imagine that this will not adversely affect the scientific careers of women with children. ^However, Hargens, (l975) found support for this relationship in chemistry, but not in mathematics and political science. l6l

Second, family obligations for women are reported to increase underemployment and unemployment, which in turn should adversely affect later participation and performance. In academia, colleagues who take a term in administration or are absent for illness lament the diffi- culty of reentering research at their prior level of productivity. Women face this difficulty in reentering the career after raising or giving birth to a child. There is little evidence that the university or other employers of women scientists accommodate the interruptions often required by giving birth and raising a child. In a sense, by not accommodating the biological fact that women give birth, and the consequences of that interruption on the career, there is what might be called cross-institutional discrimination—the demands of the institution of family place women as a group at a dis- advantage in the institution of science.2 How, then, is the reported result of no effect of family on pro- ductivity to be explained? This will be discussed in the section on methodological concerns. Cumulative Advantage A line of research generated by Robert Merton's (l968) essay on the Matthew effect has clearly established the importance of cumulative advantage in science. Scientists are experts at pressing their advantages, in reinvesting success to generate greater success. In understanding sex differences in performance, it is important also to consider the converse, the accumulation of disadvantage. Small differences that result from any of the above explanations can be expanded with time. A short interruption due to the birth of a child may initially be a small disadvantage. To the extent that it accumulates, perhaps as the result of opinions that motherhood and dedication to science are incompatible (even if they are not), the disadvantage may grow, and the career may decline. Subtle forms of discrimination may encourage self-evaluations that define oneself as unlikely to succeed, an evaluation that may be a self-fulfilling prophecy. The importance of accumulative advantage/disadvantage lies in its ability to amplify other differences, thereby giving other explanations greater effect. This fact is essential for understanding the impact of several of the methodological flaws that are now to be considered. considering the effects of domestic obligations on the produc- tivity and performance of female scientists, it must be kept in mind that female scientists are much less likely to be married than male scientists, and among those who are married, females are less likely to have children. Consequently, the lower levels of productivity and participation by women who are not married must be explained by other factors. l62