Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Gender, Family Characteristics, and Publication Productivity among Scientists

Author(s): Mary Frank Fox


Source: Social Studies of Science , Feb., 2005, Vol. 35, No. 1 (Feb., 2005), pp. 131-150
Published by: Sage Publications, Ltd.

Stable URL: http://www.jstor.com/stable/25046632

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Sage Publications, Ltd. is collaborating with JSTOR to digitize, preserve and extend access to
Social Studies of Science

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
SISIS
ABSTRACT This paper concentrates upon the relationship between marriage,
parental status, and publication productivity for women in academic science, with
comparisons to men. Findings indicate that gender, family characteristics, and
productivity are complex considerations that go beyond being married or not
married, and the presence or absence of children. For women particularly, the
relationship between marriage and productivity varies by type of marriage: first
compared with subsequent marriage, and occupation of spouse (in scientific
compared with non-scientific occupation). Further, type of family composition is
important: women with preschool children have higher productivity than women
without children or with school-age children. Women with preschool children are
found to be a socially selective group in their characteristics, particularly in their
allocations of time.

Keywords children, family, gender, household, productivity, scientists

Gender, Family Characteristics, and


Publication Productivity among Scientists
Mary Frank Fox

In the study of gender and science, publication productivity is important


for two (related) reasons. First, publication productivity is a central social
process of science. It is through publications that research findings are
communicated and verified, and that scientific priority is established
(Price, 1963; Merton, 1973; Mullins, 1973).1 Second and accordingly,
until we understand factors that are associated with productivity, and
variation in productivity by gender, we can neither assess nor correct
inequities in rewards, including rank, promotion, and salary. This is
because publication productivity operates as both cause and effect of status
in science. Publication productivity reflects women's depressed rank and
status, and partially accounts for it. 'Partially' is a key term: comparable
levels of publication do not produce the same rewards for women and men.
This is particularly conspicuous in advancement in academic rank for
women compared with men (see Sonnert & Holton, 1995).
Numbers of studies indicate gender disparity in publication produc
tivity.2 In recent years, the disparity has narrowed somewhat in life
sciences, but has persisted in scientific fields outside life sciences (Black
burn & Lawrence, 1996). Further, data indicate that while women and
men publish at different rates, the publication of both is positively skewed -

Social Studies of Science 35/l(February 2005) 131-150


? SSS and SAGE Publications (London, Thousand Oaks CA, New Delhi)
ISSN 0306-3127 DOI: 10.1177/0306312705046630
www. sagepublications. com

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
132 Social Studies of Science 35/1

so that most published work is produced by a minority of scientists among


both women and men (Cole & Zuckerman, 1984).
Documenting gender variation in productivity is one thing, and ac
counting for it another. The variation in publication may be clear, the
explanations are not. This paper concentrates specifically upon the rela
tionship between marriage, children, and publication productivity for
women, with comparison with men, in academic science.
Much has been made about the impact of marriage and children up
on women's scientific productivity and performance. The mythology of
science (Bruer, 1984) has it that good scientists are either men with wives,
or women without husbands and children. Evidence, however, tends to
contradict this conventional wisdom. Studies across scientific disciplines
indicate that married women publish as much as or more than unmarried
women (Helmreich et al., 1980; Astin & Davis, 1985; Cole & Zuckerman,
1987; Kyvik, 1990). Further, the presence of children has no effect on
women's productivity (Cole & Zuckerman, 1987), a slightly negative, non
significant effect (Reskin, 1978; Long, 1990), or a positive effect (Astin &
Davis, 1985; Fox & Faver, 1985). These patterns remain puzzling and
somewhat counter-intuitive.
This paper goes beyond the limits of current findings that emphasize
the effects of being married or not married, or the presence or non
presence of children. It takes a new, extended approach to the study of
gender, productivity, and family characteristics among scientists, examin
ing: (1) types of household and marriage patterns; (2) types of family
composition in relationship to productivity of women and men. It then
addresses factors that may account for the patterns through: (1) advan
tages of marriage by first or subsequent marriage, and type of occupation
of spouse; (2) the association between productivity and parenthood of
young children, as this relates to seven sets of conditions.
In the study of gender, status, and science, analysis of the complexities
of the relationships between household/marriage patterns, family composi
tion, and publication productivity is important. This is because emphasis
upon motherhood alone can operate as an 'intrinsic' (or deterministic)
explanation of the status of women in science, with implications that
gender and status are governed by factors that are very difficult to alter,
socially and organizationally (Luukkonen-Gronow & Stolte-Heiskanen,
1983).

Method
Data
The data are from a national mail survey, addressing aspects of resea
and graduate-level teaching as well as background characteristics.
survey, conducted in 1993-94, was of 1215 full-time, tenured, or ten
track faculty in doctoral-granting departments in computer science, c
istry, electrical engineering, microbiology, and physics. These fields r
sent a range of the scientific classifications of the National Rese

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
Fox: Gender, Family Characteristics, and Publication Productivity 133

Council (NRC) and National Science Foundation (NSF), including spe


cifically, physical, life, computing, and engineering sciences (the only two
NRC/NSF scientific categories3 not represented in the survey are earth
and atmospheric sciences, and mathematics).
The study is distinguished by sampling of faculty from known popula
tions. Details on sampling design are given in the Appendix. The sampling
design of faculty from departments that are low, high, and improved in
proportions of doctoral degrees awarded to women reflects the interest in
issues of departmental context for students and faculty, including diversity
of participants in science - and in turn, this represents a key concern
of national science policy (Pearson & Fechter, 1994; National Academy of
Sciences, 1995).
Of the 1215 faculty sent questionnaires, 26 were ineligible because of
the faculty member's departure from the department or being deceased.
The response rate was 65% (removing ineligible from the base). Those
faculty in computer science (58%) and electrical engineering (63%) had
lower, and those in chemistry (69%) and physics (65%) higher, rates than
average; and women's response rate (71%) was higher than men's (64%).

Variables
Independent Variables

Household composition is assessed through respondents' reports of being


married, never married, cohabiting with partner, divorced or separated, or
widowed.
Marriage pattern is assessed through two sets of responses for those
married. One set is an indication of present marriage as a 'first' or
'subsequent' marriage. The second is an indication of occupation of
spouse. Spousal occupations reported by respondents are then classified
into categories of: (1) academic professor in same scientific field as
respondent; (2) academic professor in different scientific field than re
spondent; (3) academic professor outside of scientific fields; (4) science
occupation outside of academia (non-academic engineer, mathematical
specialist, life and physical scientist, scientific researcher); (5) traditional
professional (attorney, judge, physician, dentist, or architect); (6) other,
nontraditional professional (accountant, nurse/dietitian/physical therapist,
health technologist, social worker, teacher in other than college, other
technician, writer/artist, computer specialist, librarian); (7) not scientific or
professional occupation, but employed; (8) homemaker, retired, or not
employed.
Type of family composition is assessed through respondents' reports of
the presence of children and their ages. The responses are then categorized
as: no children (childless); preschool children only; some elementary and/
or secondary school children; and only college and/or adult children.
These categories, in turn, reflect blocks of time and scheduling for chil
dren, with consequences for schedules of parents. Preschool children
constitute a separate category because they have special needs in blocks of

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
134 Social Studies of Science 35/1

time for care, since the USA does not have universal 'day care' for young
children. Elementary and/or secondary school children constitute a com
bined category because schedules of children in both of these groups are
governed by universal schooling in USA, with consequences, in turn, for
parents' schedules of'time away from home'. Further, it is not feasible or
desirable to cut the categories more 'thinly', because the resulting numbers
of women, in particular, would become very small.4

Dependent variable

Publication productivity is the number of papers published or accepted for


publication in refereed journals within the 3 years before the survey. This
measure takes into account (1) types of publications; (2) time lags; (3)
period of time; (4) self-reported data.
First, the survey instrument asks respondents to list separately the
number of items that were published or accepted for publication within
categories (papers in refereed journals; papers in other publications; con
ference proceedings; critiques, responses, and comments; abstracts; book
reviews; chapters, papers in newspapers/newsletters; and other). This
itemization helps eliminate mis-categorizing (for example, categorizing
critiques, responses, and comments or abstracts as papers) and thus helps
provide more valid counts of papers in referred journals, which represent
the core of publication productivity in scientific fields. Second, the inclu
sion of both papers published and accepted for publication within the
period helps take into account the time lags between research, submission,
and publication.5 Third, the specification of data for the 3-year period
controls for the effects of seniority levels of scientists, and thus the available
publishing period. Fourth, with respect to the validity of self-reported
publications, self-reports correlate highly (0.94-0.85) with those listed in
indices of abstracts (Clark & Centra, 1985; Creswell, 1985).
In the (multivariate) regression analyses of productivity, publication
productivity takes the form 1 + log (number of published papers + number
of accepted papers + 0.5). For regression analyses, the logarithmic form
normalizes the skewed distribution of productivity. The addition of 0.5
(equivalent of half a paper) avoids the quantity of the log of zero (minus
infinity) and thus allows one to deal with those with no papers published or
accepted for publication in the period. The addition of a constant of 1
avoids negative scores and their inconvenience in interpretation (see also
Pelz & Andrews, 1976).

Findings and Discussion


Gender and Publication Productivity

To begin, let us look at general features of women's and men's publication


productivity in this study. These scientists are all full-time, tenured, or

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
Fox: Gender, Family Characteristics, and Publication Productivity 135

tenure-track academics in doctoral-granting, university departments. In


the 3-year period before the survey, women published or had accepted for
publication 8.9 papers; this compares with 11.4 for men (Figure 1). This is
a significant (p = .008)6 gender difference. However, the gender difference
in publication is not as large as for samples that include women and men
scientists across a broader range of institutional types, both research and
non-research oriented colleges and universities (see Blackburn & Lawr
ence, 1996), or as large as for women and men over a longer span of time,
especially over the entire career (Cole & Singer, 1991).
In addition, two other features of the distributions of productivity are
notable (Figure 1). First, the overall difference between women's and
men's productivity owes especially to gender disparity at both extremes of
productivity: women are almost twice as likely as men to publish zero or
one paper7 in this period (women 18.8%, men 10.5%); men are twice as
likely as women to be at the other extreme, publishing 20 or more papers
in the period (men 15.8%, women 8.4%). Gender differences are not as
marked in the middle categories, for those publishing 2-9 papers (men
44.2%, women 51.4%), or even 10-19 papers (men 29.4%, women
21.0%).
Second, the productivity of both men and women is strongly positively
skewed: 14.5% of women account for one-half (50.8%) of papers pub
lished by women; 19.8% of men account for one-half (49.3%) of papers
published by men. The pattern of highly variable and strongly skewed
publication was documented more than 75 years ago in Lotka's (1926)
analysis of papers published in physics journals - and is characteristic of
both the men and women scientists in this study.

FIGURE 1
Number of refereed journal papers (accepted + published) by gender
12

N Mean SD
631 11.4 11.1
138 8.9 12.2

Men
Women
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 9

Papers (A/)

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
136 Social Studies of Science 35/1

FIGURE 2
Household and marital status and productivity by gender
100 , 20
18
80 16
14
m 60
CD 12
10
F%
40 8
6

20 r 4
2

Never married Married Widowed


Cohabiting Divorced/separated
Men Women x Men Women

Gender, Marital!Household Status, and Productivity

Family characteristics are assessed through marital/household status, as


well as parental status. Women and men differ significantly in their marital/
household statuses (p = .0000). We find (see bars at bottom of Figure 2)
that most men are married (86%); this compares with 62% of the women.
Women are twice as likely as men to be never married (16 vs 8%); almost
three times as likely to be divorced or separated (11.1 vs 4.2%); and five
times more likely to be cohabiting but not married (11.1 vs 1.8%).
How does marital/household status relate to productivity (Figure 2)?
The publication productivity of married women (and cohabiting) exceeds
that of never married women and of divorced and separated women.
Among men, the never married have the lowest productivity. Men of all
other marital/household statuses, except the never married, have higher
productivity than women of comparable status. The gender difference in
productivity is most significant among married scientists (p = .096) and
among divorced scientists (p = .05).
In the data collected, categories were provided for current marriage as
a first marriage or subsequent marriage, that is, re-married. Previous
studies have put the married into one category. In data separated by first
and subsequent marriage, women in subsequent marriages have produc
tivity that is nearly twice that of women in a first marriage (15.43 vs 8.08
papers; p = .033) (Figure 3). For men, productivity varies little for those
in subsequent compared with first marriage (10.55 vs 11.87 papers).

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
Fox: Gender, Family Characteristics, and Publication Productivity 137

FIGURE 3
Household and marital status and productivity by gender, noting marriage
100 , 20
18
80
16 1
14 I jQ
12 g.
10 TR
40 -j m
20

Never married First marriage Divor


Cohabiting Subsequent marr
Men y?=? Women Men Women

This then leads another question: to whom are these scientists married,
and how does that relate to productivity? Specifically, what is the occupation
of spouse for men and women who are married, and for those in first or
subsequent marriages? How does occupation of spouse relate to
productivity?
For married scientists across first and subsequent marriages (that is,
combining the two categories), the most common pattern among women is
marriage to another scientist (see bars at bottom of Figure 4). Most of the
women (59%) are married to another scientist (this combines occupations
of academic professor in same field as respondent, academic position in
another science field, and scientific occupation outside academia). This
contrasts with 17% of men who are married to another scientist (combin
ing the same three science occupations).
For women, the highest productivity is among those married to a
scientist outside academia or to someone in one of the non-traditional
professions (such as accountant, social worker, librarian, physical thera
pist, writer/artist). Among men, productivity varies less than women's by
the occupation of their spouse (see the graph lines in Figure 4).
What about the relationship between spousal occupation and pro
ductivity for those in first marriages compared with subsequent marriages?
Does the higher productivity of women in subsequent marriages relate to
the occupations of spouses? It is the case that women in subsequent,
compared with first, marriages are somewhat more likely to be married to
another scientist; 63% of the women in subsequent marriages, compared
with 56% of those in first marriages, have a spouse employed in science

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
138 Social Studies of Science 35/1

FIGURE 4
Spouse's occupation and productivity by gender
100- - - 20
18
80 i
16 |
14 I CO

60 12 I a.

10 ?
40
6 r
CL
20 CO
a.

_____

Academic same Academic Traditional


i^A Other/
science other professional student
Academic different Nonacademic Other Home/
science science professional retired

Men Women Men ? Women

(Figure 5).8 Further, for the group of women in subseq


productivity is highest among those married to another
academic in their field or is a nonacademic scientist. I
subsequent compared with first marriages have higher pr
and for women in subsequent marriages, productivity is
among those married to an academic scientist in the
scientist outside of academia.
Men in subsequent marriages are also more likely than
marriages to have a spouse employed in science (32%
14%) (Figure 6). For men in subsequent marriages, produc
for those married to an academic in the same field.
Overall, then, subsequent compared with first marria
tists are marriages that are more 'synchronized' in matc
tions of spouses. In subsequent marriages, especially, sci
chosen spousal circumstances that support their producti
ance. It may also be the case that for those in subsequen
'matched occupations' (two scientists), science is esp
'outside interests' are limited.

Gender, Family Composition, and Productivity

Children form more complex circumstances than mar


looks beyond having or not having children to the curre

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
Fox: Gender, Family Characteristics, and Publication Productivity

FIGURE 5
Spouse's occupation and productivity by marriage, for women
100 i 30
27
80 24
21

2o 60 18
15
40 12
9
20 4 6
3
0 311_ _XZL 0
Academic same Academic Traditional Other/
science other professional student
Academic different Nonacademic OtherHome/
science science professional retired

First Subsequent First Subsequent

FIGURE 6
Spouse's occupation and productivity by marriage, for men
100 30
27
80 4 24
21

2o 60 18
15
40 12
9
20 6

II
3
M n 0
Academic same Academic Traditional Other/
science other professional student
Academic different Nonacademic OtherHome/
science science professional retired

Subsequent First Subsequent


[~] First

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
140 Social Studies of Science 35/1

tion: no children (childless); preschool children only; some elementary


and/or secondary school children; and only college and/or adult children.
Women and men differ significantly in these family compositions
(p = .0000). Most noteworthy is that most women (52%) have no chil
dren; this contrasts with 21% of the men (Figure 7). Men are 2.5 times
more likely than women to have college or adult children (men 33.3%,
women 13.2%), and 1.5 times more likely to have elementary/secondary
children (men 31.4%, women 22%). However, the same proportions of
men and women have preschool children only (men 13.4%, women
13.2%).
Considering the relationship between family composition and pro
ductivity, especially notable is the high productivity of women with pre
school children only, in comparison with women without children or
women with school children. The productivity of men varies less by family
composition than does women's.
Given the demands of young children upon time, energy, and atten
tion, how do we account for this anomaly of women with preschool
children having higher productivity than women without children or
women with school-aged children? Who are these women, and what are
their circumstances? I address seven sets of considerations.
The first consideration is the presence of children as a possible artifact
in relationship to productivity, reflecting the effect of age and stage of life.
The question is: are women bearing and caring for young children at a
stage of already established productivity? For this group of women scien
tists, the relationship between family composition and productivity does

FIGURE 7
Family composition and productivity by gender

Childless Elementary/secondary
Pre-school only Only college/adult
Men Women Men Women

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
Fox: Gender, Family Characteristics, and Publication Productivity 141

not owe to life stage.9 In analysis of covariance (with covariates of age,


age2, time since PhD, and time since PhD2), family composition does
remain a significant determinant of productivity for women.
A second consideration is number of children. The question is: do
women with preschool children only have fewer children than other
groups? Women with preschool children do have fewer children (1.4) than
do women with elementary and secondary school children (2.0) or only
adult/college children (2.2) (Figure 8).
A third consideration is the occupation of spouse. The question is: does
the productivity of women with preschool children reflect the occupation
of spouse - particularly, marriage to a scientist? Marriage to another
scientist can provide women access to mainstream networks of informa
tion, funding, and resources for research (Astin & Davis, 1985; Astin &
Milem, 1997; Creamer, 1999). Figure 8 shows that women with preschool
children are the group most likely to be married to another scientist.
A fourth consideration is type of marriage, specifically a subsequent
(compared with first) marriage, which, as discussed earlier, is associated
with higher publication productivity among the women in this study. The
question is: compared with women with other family compositions, are
women with preschool children especially likely to be in subsequent,
compared with first, marriages? To the contrary, women with preschool
children are less likely to be in subsequent marriages; 12.5% of women
with preschool children, compared with 21 % of women in other family
compositions, are in subsequent marriages (Figure 8).

FIGURE 8
Factors by family composition and gender
100

80

60 Q)
"O

O
40

20

Childless Elementary/ Childless Elementary/


secondary secondary
Pre-school College/adult Pre-school College/adult

WIn subsequent marriageSpouse in science field


No. of children
fill Heavy research interest

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
142 Social Studies of Science 35/1

A fifth consideration is social selectivity: women who marry, and those


who currently have preschool children, may have especially strong stamina
for and commitment to research - with implications for their productivity
(Fox & Faver, 1985). These women with preschool children are holding
full-time academic appointments in science. They have not phased out of,
or scaled down in, scientific careers at this particular parental-stage, and
they may then be a 'super-select' group in stamina and commitment.
Accordingly, we find that compared with women without children and
women with older children, the women in this study with preschool
children are the group with the highest proportion (50%) reporting that
their work interest is 'heavily in research' (compared with the other
categories of work interest: 'heavily in teaching' 'both, but leaning toward
teaching' or 'both, but leaning toward research') (Figure 8).
A sixth consideration comes from interviews with scientists (Cole &
Zuckerman, 1987), which suggest that women with children make espe
cially 'disciplined allocations' of time - allocating time and attention to
work and children, but little else. One way to assess this is with data
collected in the study on reported hours per week in work activities
(Figure 9).
These activities are: research and writing; advising graduate students;
preparation and administration of grants; editorial and review boards;
preparation for instruction; instruction; advising undergraduates; and serv
ing on departmental/university committees. The first four activities are
more research-related; the second four, less research-related. (In Figure 9,

FIGURE 9
Work activity by family composition and gender
60
Men Women
50

| 40
I 30
?20
10

Childless Elementary/ Childless Elementary/


secondary secondary
Pre-school College/adult Pre-school College/adult
Research and writing Advising graduates "| Grant preparation
J and administration
Reviewing and editorial Course preparation Course instruction

Advising undergraduates ! Committees

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
Fox: Gender, Family Characteristics, and Publication Productivity 143

horizontal lines separate the more compared with less research related
activities. These data do not include time in non-work [for example,
domestic] activities.)
Interestingly and with implications for 'disciplined allocation of time',
compared with women in other family compositions, women with pre
school children spend more time advising graduate students, and spend
somewhat less time advising undergraduates and serving on departmental
and university committees. In Figure 9, we see that it is not differences in
time in research and writing that are notable among women in different
family compositions; rather, what is more notable is time in activities
related to research, particularly more time advising graduate students and
less time advising undergraduates.
A final, seventh consideration is that women with preschool children
may be located in departments that have climates more attuned to the
participation of women. One limited way to assess this consideration, with
data that are available, is to see whether women with preschool children are
especially likely to be located in departments that are high or improved in
doctoral degrees awarded to women (the variable of the sampling design).
Figure 10 shows that women with preschool children are, in fact, the group
who are least likely to be in departments that are high or improved, without
support for this consideration.
To assess further the association for women between productivity,
family composition of preschool children, and family composition of
school-age children, the relationships are expressed in regression models in
two stages (Table 1). The first stage represents the relationship between

FIGURE 10
Departmental context by family composition and gender

100 | Men ; Women

Childless Elementary/ Childless Elemen


secondary secondary
Pre-school College/adult Pre-school

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
144 Social Studies of Science 35/1

TABLE 1
Regression of publication productivity on family composition variables (stage I) and
family composition variables plus factors (stage II) for women

Stage I Stage II
Variable Coefficent SE Significance Coe
Family composition:
Preschool children .657 .294 .027** .381 .319 .234
School-age children -.345 .241 .155 -.440 .287 .128
No. of children .063 .105 .549
Married to scientist .214 .210 .310
Interest: heavily in .086 .218 .696
research
No. of hours per week:
Advising graduate .040 .022 .069*
students
Advising -.092 .041 .027**
undergraduates
Committees -.012 .034 .715
Constant 2.664 2.532
R2 .063 .140
N 131 131
SE of estimate 1.107 1.086

* significant at < .05; **significant at < .01.

productivity and family composition of p


with other family compositions = no chil
adult/college-aged children only) and sch
other family compositions = no childre
adult/college-aged children only). This is
stage represents the relationship between
of preschool children, family composit
factors found here to be related to famil
number of children, marriage to another
riage to person outside of science or n
research (compared with interest 'heavily
toward teaching', or 'both, but leaning to
hours per week in activities. This is the
simultaneously.
In stage/model 1, we see among wom
nificant effect (p = .027) for productivity
children and the negative effect (p = .155
children. In stage/model 2, with factors c
that the family composition variables are no
productivity.
In model 2, the variables that emerge
number of hours per week advising gradu
number of hours advising undergraduates

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
Fox: Gender, Family Characteristics, and Publication Productivity 145

that, among the factors considered here as they account for the relation
ship between productivity and family composition, hours in activities
related to research (advising graduate students) compared with hours in
non-research (advising undergraduates) are most notable. The pattern of
investments of time in research-related activities and in teaching-related
activities, as related to productivity, has been found in analyses among
social scientists as well (Fox, 1992). The finding here suggests that time
investments in research-related and teaching-related activities operate in
opposite directions in relationship to productivity. These time investments
account, in larger part than other factors here, for the relationship between
family composition and productivity among women scientists in doctoral
granting departments - insofar as they emerge, in model 2, as the sig
nificant predictors of productivity and the family composition of preschool
children is no longer significant.
At the same time, factors beyond the range considered here should be
the subject of further analyses of family composition and productivity, as
total variation (R2) explained in model 2 is 15% (.140).

Summary and Conclusions


In summary, gender, productivity, and family characteristics are complex
considerations that go beyond being married or not married, and the
presence or absence of children. For women, particularly, the relationship
between marriage and productivity varies by type of marriage: that is,
subsequent compared with first marriage, and occupation of spouse.
Women in subsequent marriages have higher productivity than women in
first marriages. This relates to their greater likelihood to be married to
another scientist; and when married to a scientist, the effects for produc
tivity are positive.
In family composition, the predominant pattern for women scientists
is that of'no children', found among 52% of women (compared with 21%
of men). Among types of family compositions, however, the productivity of
women with preschool children is higher than that of women without
children or those with school-aged children. In pursuing factors that may
be associated with this anomalous pattern, women scientists who have
preschool children show signs of being a socially selective group in mar
riage and family patterns, research interests, and allocations of time. In a
multivariate model of productivity for women, allocations of more time in
research-related activity and less in non-research-related activity are the
most significant factors, among those considered.
The data do not indicate particular policies and practices in the work
environments of the women scientists. At issue, for example, are the
implications for productivity of flexible-time policies on campus or pro
grams of parental leave and child-care. It may - or may not - be the case
that women with preschool children are apt to be in settings with such
policies or programs.

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
146 Social Studies of Science 35/1

In addition, it is important to emphasize that these data do not indicate


that marriage and young children have no effect upon women in science.
Marriage and young children have a multitude of effects in personal
sacrifices as well as rewards, and extraordinary arrangements of accom
modation (Grant et al., 2000). What these data show is that marriage and
young children are not associated with depressed publication productivity
among women who do hold academic positions in science.
In the interpretation of the data on marriage, parenthood, and pro
ductivity, it is important to point out this: these data are based upon
women who have survived a rigorous and demanding process of scrutiny,
selection, and evaluation in science. Family demands may take their toll
along the way, through graduate school and early career, so that a propor
tion of women are eliminated from scientific careers and do not even fall
into such cross-sectional data of professional, employed scientists (see
Long, 1987).
Thus in continuing steps, we need to understand more about the way
that productivity and productivity differences unfold over time and in
relationship to family and household characteristics - with implications for
sustained participation and performance in science.

Notes
The research reported here was supported by grants from the National Science Foundation
(SED-9153994 ad SBE-0123532). For his reading of and comments on an earlier version
of this paper, I thank J. Scott Long.

1. Unpublished work and informal exchange are also consequential for the development
and communication of knowledge. But informal communication is haphazard, and
unpublished work cannot be widely evaluated.
2. See reviews in Creamer (1998), Long & Fox (1995), and Zuckerman et al. (1991).
When one considers citations, however, which are not at focus in this paper, women's
papers may receive as many citations as men's on a per paper basis (Long, 1992).
3. 'Scientific categories' as considered here include those fields outside of social sciences.
4. The numbers of women in these family composition groups are: 70 with no children;
18 with preschool children only; 30 with some elementary and/or secondary school
children; 18 with college or adult children only.
5. Inclusion of accepted and published papers also reduces a 'floor' effect of publications
in analyses.
6. Probability levels of significance appear within the text of the paper. The tests of
significance are these: for two groups, r-test; for more than two groups, analysis of
variance; and for categorical variables, %2-test. Because of small sizes of groups of
women, particularly when divided by variables, differences among women may be
notable but not statistically significant. They are indicated and treated as such.
7. 8.7% of women and 7.1% of men published zero papers in this time span; 10.1% of
women and 3.3% of men published one.
8. Note that for women the percentage 'married to another scientist' is relatively high
(56%) among those in a first marriage. It is yet higher (63%) among women in
subsequent marriages. However, this is against a comparative point that is already high
(56%), such that a 'ceiling' effect may be operating here.
9. Relatedly, although the publication productivity of women with preschool children is
higher than that of women with no children or elementary and/or secondary aged

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
Fox: Gender, Family Characteristics, and Publication Productivity 147

school children, this does not appear to be a function of women with preschool
children being an untenured group, prompted to publish by this status. The clear
(61%) majority of women with preschool children only are tenured; this compares with
41% of women without children and 73% of those women with elementary and/or
secondary aged school children.
10. As discussed, family composition remained a significant determinant of productivity in
the presence of age, age2, time since PhD, and time since PhD2; age and time since
PhD are not then included in the model. As discussed also, women with preschool
children, compared with women with other family compositions, are not more likely to
be in a subsequent (compared with first) marriage or to be in a department that is high
or improved (compared with low) in doctoral degrees awarded to women. Thus, neither
of these variables is included in the model.

References
Astin, Helen & Diane Davis (1985) 'Research Productivity Across the life- and Career
Cycles: Facilitators and Predictors for Women', in M.F. Fox (ed.), Scholarly Writing
and Publishing: Issues, Problems, and Solutions (Boulder, CO: Westview): 147-60.
Astin, Helen & Jeffrey Milem (1997) 'The Status of Academic Couples in U.S.
Institutions5, in M.A. Ferber & J.W Loeb (eds), Academic Couples: Problems and
Promises (Urbana, IL: University of Illinois Press): 128-55.
Blackburn, Robert & Janet Lawrence (1996) Faculty at Work (Baltimore, MD: Johns
Hopkins University Press).
Bruer, John (1984) 'Women in Science: Toward Equitable Participation', Science, Technology,
and Human Values 9: 3-7.
Clark, Mary Jo & John A. Centra (1985) 'Influences on the Career Accomplishments of
Ph.D.s', Research in Higher Education 23: 256-69.
Cole, Jonathan R. & Burton Singer (1991) 'A Theory of Limited Differences: Explaining
the Productivity Puzzle in Science', in H. Zuckerman, J. Cole & J. Bruer (eds), The
Outer Circle: Women in the Scientific Community (New York: WW Norton): 277-10.
Cole, Jonathan R. & Harriet Zuckerman (1984) 'The Productivity Puzzle: Persistence and
Change in Patterns of Publication Among Men and Women Scientists', in M.W
Steimkamp & M. Maehr (eds), Advances in Motivation and Achievement, vol. 2
(Greenwich, CT: JAI Press).
Cole, Jonathan R. & Harriet Zuckerman (1987) 'Marriage, Motherhood, and Research
Performance in Science', Scientific American 255: 119-25.
Creamer, Elizabeth (1998) Assessing Faculty Publication Productivity: Issues of Equity, ASHE
ERIC Higher Education Report, vol. 26, no. 2 (Washington, DC: The George
Washington University).
Creamer, Elizabeth (1999) 'Knowledge Production, Publication Productivity, and Intimate
Academic Partnerships', Journal of Higher Education 70 (May/June): 261-77.
Creswell, John W (1985) Faculty Research Performance: Lessons from the Sciences and Social
Sciences, ASHE-ERIC Higher Education Report no. 4 (Washington, DC: Association for
the Study of Higher Education).
Fox, Mary Frank (1992) 'Research, Teaching, and Publication Productivity: Mutuality
versus Competition in Academia', Sociology of Education 65 (October): 293-305.
Fox, Mary Frank & Catherine Faver (1985) 'Women, Men, and Publication Productivity',
Sociological Quarterly 26: 537-49.
Grant, Linda, Ivy Kennelly & Kathryn Ward (2000) 'Revisiting the Gender, Marriage, and
Parenthood Puzzle in Scientific Careers', Women's Studies Quarterly 28 (spring/
summer): 62-83.
Helmreich, Robert, Janet Spence, WE. Beane, G.W Lucker & K.A. Matthews (1980)
'Making It in Academic Psychology: Demographic and Personality Correlates of
Attainment', Journal of Personality and Social Psychology 39 (November): 896-908.

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
148 Social Studies of Science 35/1

Kyvik, Svein (1990) 'Motherhood and Scientific Productivity', Social Studies of Science 20:
149-60.
Long, J. Scott (1987) 'Discussion: Problems and Prospects for Research on Sex
Differences', in L.S. Dix (ed.), Women: Their Underrepresentation and Career Differentials
in Science and Engineering (Washington, DC: National Research Council): 157-69.
Long, J. Scott (1990) 'The Origins of Sex Differences in Science', Social Forces 68:
1297-315.
Long, J. Scott (1992) 'Measures of Sex Differences in Science', Social Forces 71: 159-78.
Long, J. Scott & Mary Frank Fox (1995) 'Scientific Careers: Universalism and
Particularism', Annual Review of Sociology 21: 45-71.
Lotka, A.J. (1926) 'The Frequency Distribution of Scientific Productivity', Journal of the
Washington Academy of Sciences 26: 317-23.
Luukkonen-Gronow, Terttu & Veronica Stolte-Heiskanen (1983) 'Myths and Realities of
Role Incompatibility of Women Scientists', Acta Sociol?gica 26: 267-80.
Merton, Robert K. (1973) 'Priorities in Scientific Discovery', in N.W. Storer (ed.), The
Sociology of Science (Chicago, IL: The University of Chicago Press).
Mullins, Nicholas C. (1973) Science: Some Sociological Perspectives (Indianapolis, IN: Bobbs
Merrill).
National Academy of Sciences (1995) Reshaping the Graduate Education of Scientists and
Engineers (Washington, DC: National Academy Press).
Pearson, Willie & Alan Fechter (1994) Who Will Do Science? Educating the Next Generation
(Baltimore, MD: Johns Hopkins University Press).
Pelz, Donald & Frank M. Andrews (1976) Scientists in Organizations (Ann Arbor, MI:
Institute for Social Research, The University of Michigan).
Price, Derek (1963) Little Science, Big Science (NewYork: Columbia University Press).
Reskin, Barbara (1978) 'Scientific Productivity, Sex, and Location in the Institution of
Science', American Journal of Sociology 83: 1235-43.
Sonnert, Gerhard & Gerald Holton (1995) Gender Differences in Science Careers (New
Brunswick, NJ: Rutgers University Press).
Zuckerman, Harriet, Jonathan Cole & John Bruer (1991) The Outer Circle: Women in the
Scientific Community (NewYork: WW. Norton).

Mary Frank Fox is NSF Advance Professor in the School of Public Policy and
Co-director of the Center for the Study of Women, Science, and Technology
at Georgia Institute of Technology. Her research focuses upon gender and
scientific and academic organizations and occupations. Her current research
includes a Study of Programs for Women in Science and Engineering,
supported by the National Science Foundation. Her publications appear in
more than 40 different scholarly and scientific journals, collections, and
books. She has been a member of the National Research Council/National
Academy of Sciences (NRC/NAS) study panel on Trends in the Early Careers
of Life Scientists, and consultant to the NRC/NAS study panel on Gender
Differences in Science and Engineering. She was awarded the WEPAN
(Women in Engineering Programs) Betty Vetter Research Award, 2002, for
notable achievement in research on women in engineering, and the SWS
Feminist Lecturer Award, 2000, for a prominent feminist scholar who has
made a commitment to social change.

Address: School of Public Policy, Georgia Institute of Technology, Atlanta,


Georgia, GA 30332-0345, USA; fax +1 404 371 8811; email:
mary.fox@pubpolicy.gatech.edu

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
Fox: Gender, Family Characteristics, and Publication Productivity 149

Appendix
Sampling Design
For each of the fields (except microbiology, discussed subsequently), the
doctoral-granting departments sampled were those, identified on the basis
of data from the National Research Council (NRC), Survey of Doctoral
Recipients, as being: (1) consistently low; (2) consistently high; or (3) most
improved in proportions of doctoral degrees awarded to women over a
17-year period.
For chemistry, computer science, electrical engineering, and physics,
the rate of degrees awarded to women was computed for the first 5 years
and the last 5 years of the period for which data were available. The first 5
years were 1974-78 except for computer science (1978-82) and electrical
engineering (1977-81); the last 5 years were 1986-90. Within each field,
the departments were ranked based upon the total number of PhDs
produced during the period. The largest PhD departments - those produc
ing 70% of all doctorates in the field - were selected as the 'population of
interest', except for computer science where a 50% cutoff was used.
Within each of the four fields, departments were then ranked, based
on the difference between the ending rate of women PhDs (that is, the rate
over the last 5 years) and the beginning rate (that is, the rate over the first
5 years). For chemistry, the 'most improved' departments were those with
an increase of 15% or more in the rate of women PhDs. For computer
science and engineering, the most improved were 8% or more, and for
physics, 9% or more. For each of the four fields, the 'consistently low' and
'consistently high' departments were those for which the difference in the
rate of women PhDs was within a change of not more than ?5%. Within
each field, the consistent departments were then ranked, based on the
average of the beginning and ending rates. Depending upon the field,
cutoffs were selected by 'low' consistent rate and 'high' consistent rate. By
field, the cutoff points are: chemistry, high = > 15%, low = < 13%;
computer science, high = > 10%, low = <8%; electrical engineering,
high = >5%, low = <1%; physics, high = >8%, low = <4%.
The study is distinguished by sampling of faculty from known popula
tions. To accomplish this, I obtained rosters of faculty from the respective
departments determined through the NRC data. Because of the low
proportions of women compared with men faculty in these four science
and engineering fields, and the aim for sufficient numbers to allow gender
comparisons, sampling fractions were applied separately for male and
female faculty. For the departments in these four fields, all women and
40% of men were sampled. Likewise, because the study included focus
upon differences in organizational and outcome variables among de
partmental categories (low, high, improved), it is desirable to put those
categories in the design and not leave sample outcome to randomness of
departmental categories.
Microbiology cannot be sampled with the same design (of depart
ments that have been low, high, or improved in proportions of degrees

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms
150 Social Studies of Science 35/1

awarded to women). That is because for microbiology, the field in which


students identify degree (in NRC Survey of Doctoral Recipients) corre
sponds more loosely with department. For example, a degree listed as field
of biology or microbiology may be from variable departments, such as
molecular genetics, neurobiology, or other units. Thus, for microbiology,
rosters of faculty and students were obtained from 69 responding depart
ments of the 103 US doctoral-granting departments of microbiology in the
listing of 'Colleges and Universities Granting Degrees in the Microbio
logical Sciences', American Society of Microbiology. The sample of faculty
was drawn from the 19 departments granting 50% of all microbiology
degrees. Sampling fractions were applied separately for women and men
faculty, with 40% of male faculty and 50% of female faculty sampled.
Departments were then classified as 'high' if more than 50% of the
doctoral students in the department were women (based upon the rosters
of doctoral students collected from the departments).

This content downloaded from


196.1.114.18 on Mon, 03 Oct 2022 13:42:33 UTC
All use subject to https://about.jstor.org/terms

You might also like