The Matilda Effect—Role Congruity Effects on Scholarly Communication

Abstract

Using role congruity theory as the basis for the study, an analysis of 1,020 articles published 1991-2005 in Communication Research and Journal of Communication, as well as the ISI citations these articles received and the citations these articles included, was conducted. In line with a hypothesized “Matilda effect” (underrecognition of female scientists), articles authored by female communication scientists received fewer citations than articles authored by males. Hypotheses on moderating impacts of research topic, author productivity, and citing author’s sex, as well as on change in the effect’s extent across time were derived from the theoretical framework. Networking conceptualizations led to an additional hypothesis. Five of six hypotheses were supported.

Keywords

citations science communication role congruity gender scientists

Science has long been conceptualized as a social system (e.g., Merton, 1957, 1973), and communication is seen as the common ground of all sciences (e.g., Lievrouw, 1989; Paisley, 1989). “Communication is the only general scientific behavior; other behaviors are mostly specific and technical” (Griffith, 1989, p. 600). Moreover, communication is the basis for the reward system in this social system, namely in the form of recognition (e.g., Cole & Cole, 1967; Merton, 1957). Given the crucial role of science for society in generating new knowledge, it is highly desirable that the communication-based reward system in science works in an unbiased fashion to foster maximum knowledge output and effective academic training. The current investigation examines whether male communication scientists, as opposed to female communication scientists, receive more recognition, as reflected in use of journal article references (citations)—referred to here as the “communication-based reward system.”

Many obstacles to the ideal of equity in the communication-based reward system of science have been identified in research (e.g., National Science Foundation, 2004). For this study, we will focus on sex differences among scientists regarding the recognition they receive. For instance, Etzkowitz, Kemelgor, and Uzzi (2000) noted that “The human price for the Ph.D. is higher for women than for men, and the rewards are often lower” (p. 95). The pattern has been labeled the Matilda effect (Rossiter, 1993)—a systematic underrecognition of female scientists. The term has been coined with reference to the well-known Matthew effect in the reward and communication systems of science (Merton, 1968)—overrecognition of those at the top of the scientific profession, even credit misallocation to scholars that are well known already.

Citations are often thought to be the reward currency of science (Merton, 1968). Citation rates and citations in what are considered to be top journals in various disciplines strongly influence decisions about hires, tenure/promotion, and faculty salaries (e.g., Cameron, 2005; Holden, Rosenberg, & Barker, 2005; Toutkoushian, 1994). Building on Rossiter’s (1993) normative claim “if science is to be meritocratic [. . .], similar or equal achievements should receive similar reputations or recognition” (p. 327), it is to be expected that publications from female and male authors in high-quality outlets would be recognized in similar ways. Thus, one would expect that male and female scientists’ research would be acknowledged and referenced, on average, about the same number of times. In other words, the share of publications by female authors should be proportionately reflected in citations to these publications—unless a Matilda effect applies and leads to undercitating of women’s work.

The present study examines the Matilda effect through a citation analysis of publications in communication research, as well as moderators of the effect. Our rationale for this approach is to base the comparison on the same achievements, namely, publication in leading communication research journals. Before doing so, we will first explain and describe the Matilda effect and then discuss role congruity theory (e.g., Eagly & Karau, 2002) as basis of five hypotheses. Given that citations represent a reciprocal behavior that goes beyond social cognitions, we will also draw on a network analysis framework (Ibarra, 1992, 1993) for our last hypothesis.

The Matilda Effect

Rossiter (1993) suggested the Matilda effect based on an historical review of female scientists’ contributions going unnoticed or undervalued. Quantitative evidence concerning the undervaluing of female scientist contributions based on referencing citations has emerged in a number of studies in fields such as sociology, anthropology, linguistics, and astrophysics (Baldi, 1998; Davenport & Snyder, 1995, Lutz, 1990; McElhinny, Hols, Holtzkener, Unger, & Hicks, 2003; Toutkoushian, 1994). However, findings have not been entirely consistent: For library and information science, Hâkanson (2005) found a Matilda effect in citation patterns whereas Sánchez Peñas and Willett (2006) did not. For psychology, Helmreich, Spence, Beane, Lucker, and Matthews (1980) detected that males received more citations than females, whereas Over (1990) and Haslam et al. (2008) did not. This inconsistency may result from different levels of acceptance of female scholars in different disciplines (Ferber, 1988), possibly resulting from different levels of representation of women in an academic field and some academic areas being perceived as more compatible with feminine roles than others (see also following section). The examined time periods can also result in different findings, as it has been argued that women will gain greater acceptance in academia over time (Ferber, 1988). The present investigation concerns whether a Matilda effect exists in the field of communication research based on citations to articles published during 1991-2005.

Role Congruity Theory

A suitable theoretical framework to explain the Matilda effect is role congruity theory (e.g., Eagly & Karau, 2002). The theory is based on social role theory (Eagly, 1987), which conceptualizes gender through the construct of gender roles. Social roles are socially shared expectations pertaining to individuals with a certain social position or from a particular social category. Thus gender roles are consensual beliefs about characteristics of women and men. “These beliefs are more than beliefs about the attributes of women and men: Many of these expectations are normative in the sense that they describe qualities or behavioral tendencies believed to be desirable for each sex” (Eagly, 1987, p. 13). A central proposition of social role theory is that most of the beliefs about the sexes concern communal and agentic characteristics (Eagly, 1987). Communal characteristics, more strongly ascribed to women, pertain to being concerned with others’ well-being (e.g., helpful, kind, sympathetic). In contrast, agentic attributes, ascribed more strongly to men, refer primarily to an assertive and self-assured disposition (e.g., competitive, self-confident, objective, ambitious, and able to lead) that match stereotypes of scientists (Nosek et al., 2009) well.

Grounded in social role theory, role congruity theory was developed by Eagly and colleagues (e.g., Eagly & Karau, 2002). Although it was initially applied to women in leadership roles, we will extend it here to women in research roles. Role congruity theory goes beyond social role theory to consider the congruity between gender roles and other roles such as the scientist role. It specifies factors and processes that affect congruity perceptions and consequences for perceptions and behaviors. This framework suggests that prejudice against female scientists arises from dissimilarity between a female gender role and the expectations commonly held toward individuals in a scientist role (e.g., Nosek et al., 2009). Such prejudice results when perceivers judge women in roles as scientists’ roles because of perceived inconsistency between the mostly communal qualities that perceivers commonly connect with women and the more agentic and masculine qualities commonly believed to apply to scientists. As people tend to have dissimilar beliefs about scientists and women and similar beliefs about scientists and men, seeing a person in a scientist role may induce gender-based expectations in competition with scientist role-based expectations.

When a stereotyped group member and an incongruent social role become connected in the perceiver’s perspective, this inconsistency affects the evaluation of the group member negatively with regard to performance in that role. Applied to the context of citations, role congruity theory suggests that an author who is reviewing others’ research is likely to perceive the publications by women less favorably. This social-psychological phenomenon of role incongruity may account for the Matilda effect of women being undercited. Based on these considerations, we propose our first hypothesis:

Hypothesis 1 (H1): Publications in top-quality communication research journals authored by males receive more citations than those authored by females only.

Based on role congruity theory, we will consider four moderating influences on the generally expected Matilda effect. The second and the third hypotheses pertain to the scientist role and the fourth and the fifth hypotheses pertain to the female gender role, as the extent of incongruity that leads to prejudice results from discrepancies between expectations toward individuals in scientist roles and expectations toward individuals in the female gender role. Thus both of these components should influence the extent of the Matilda effect.

Individuals may have stereotyped perceptions of the scientist’s role based on a specific research discipline or topic. For instance, mathematics and physics are particularly connected with scientist attributes commonly also associated with men (e.g., objective, skilled in math), whereas the humanities and social sciences may entail more emphasis on human relationships, which is more congruent with the communal qualities commonly attributed to women. As a result, a female scientist working in the male-typed area of physics may induce more incongruity between the female gender role and the scientist role than a female scientist who studies social work as a more female-typed discipline.

In addition, research topics may vary by the extent to which they are associated with stereotypically male or female attributes. Within communication research, the area of “children and media” may elicit more stereotypically female qualities, such as nurturing, caring, and the like, whereas the area of political communication could be more linked to stereotypically male characteristics such as competitiveness, assertiveness, and so forth. Thus, a female communication researcher who studies political communication may instigate more incongruity between female gender role and scientist role than a female scholar who studies “children and media.”

Along these lines, investigations for other academic disciplines suggested that women’s underrecognition in citations might result from research topic choices that differ by researchers’ biological sex and attract more or less citations (e.g., Davenport & Snyder, 1995; Hâkanson, 2005). However, the existing evidence is inconsistent: Sánchez Peñas and Willett (2006) as well as Haslam et al. (2008) did not find a topic influence on citations. Although it needs to be acknowledged that categorization of research topics per gendered stereotypes presents a challenge, the current study aims to test the proposition that research topic affects the Matilda effect within communication research through the following hypothesis:

Hypothesis 2 (H2): The gender gap in citation suggested in H1 is particularly strong for publications that focus on male-typed topics.

A possible reason for the Matilda effect has been suggested, stating that female authors are cited less frequently simply because women publish less than men and are more likely to drop out of science altogether (e.g., Cole & Zuckerman, 1984; Long, 1992). Research indicates that male communication researchers produce more publications on average (Bunz, 2005) and may be more likely to “become a name,” resulting in more citations due to greater visibility.

Role congruity theory, however, offers another perspective than that of a simple main effect of author productivity on citations. The theory suggests that a woman who meets the requirements of the scientist particularly well should reduce gender-stereotypical inferences about her. However, perceiving a female scientist as very similar to her male peers is likely to be disadvantageous for her. This disadvantage may arise from the injunctive norms linked to the female gender role. As women who are prolific, successful scientists appear to violate standards applied to their gender when they exhibit male-stereotypical, agentic characteristics and do not show female-stereotypical, communal characteristics, they are likely to be unfavorably evaluated for their gender role violation. Accordingly, the gender gap in citations should be particularly strong among prolific scholars.

Hypothesis 3 (H3): The gender gap in citations suggested in H1 is particularly strong for citations to scholars with high productivity.

As another potential moderator, perceivers’ gender ideology should affect the extent to which prejudice is applied to scientists. If the onlooker holds more traditional gender-role attitudes, this incongruity and the resulting prejudice would be stronger. Based on this assumption, it can be inferred that male authors in particular will undercite female scientists because men overall hold more traditional gender-role attitudes (e.g., Bolzendahl & Myers, 2004). It is plausible to assume that this sex difference in gender role ideology in the general population also exists among scientists—even though education leads to more egalitarian gender ideology, it does not reduce the sex difference pertaining to gender ideology (Bryant, 2003). Earlier citation analyses showed that men are less likely to cite female authors than women are (Davenport & Snyder, 1995; Ferber, 1988; Hâkanson, 2005; McElhinny et al., 2003). Hence, we propose a hypothesis on a sex difference in the extent to which citing authors contribute to an overall Matilda effect reflected in a gender gap in citations.

Hypothesis 4 (H4): Male authors of top-quality communication research journal publications are less likely than female authors to cite female peer scholars.

It has been suggested that greater gender equality in science may be achieved as time progresses. For example, Ferber (1988) suggested that as more women continue to enter a discipline, their group will receive more recognition. Along these lines, communication research should not have as much of a problem with gender inequality as some other disciplines, as the field has seen a considerable number of women earning their doctorate for about 30 years (e.g., 41% in 1980-1981, 53% in 1999-2000, see National Communication Association, n.d.). For earlier times, however, Burroughs, Christophel, Ady, and McGreal (1989) found only six women among the 99 most prolific authors in communication studies during 1915-1985. Hickson, Stacks, and Amsbary (1992) treated women as a separate group of scholars and ranked female prolific authors in communication specifically (see Blair, Brown, & Baxter, 1994, for commentary). Yet Cooper, Stewart, and Friedly (1989) noted an increase in publications from female authors for the 1967-1986 time frame and suggested that the blind review system benefits women, an argument very much in line with role congruity theory.

In contrast to attaching importance to the numeric share of female scientists, role congruity theory offers another perspective. Given that gender-role attitudes have generally become more egalitarian over time (e.g., Bolzendahl & Myers, 2004), it is plausible to assume that this change occurs among scientists as well and should over time result in smaller role incongruity perceptions regarding female scientists. This smaller incongruity should result in less prejudice and a reduced Matilda effect over time.

Hypothesis 5 (H5): The share of citations that women’s publications in top-quality communication research journals receive increases over time.

The Network Perspective

It is important to consider that citations manifest reciprocal behaviors that go beyond the processes conceptualized by role congruity theory. Indeed, a commonly cited obstacle for women’s success in science is the idea that women have smaller, weaker networks and do not fit seamlessly into the male-dominated networks in academia (Fox, 2010). Ibarra (1993) provided a conceptual framework for studying women’s networks in organizational contexts, which also guided the present perspective concerning sex differences in citation networks. Ibarra suggested that the genders differ both in the approaches they take to forming relationships in organizational settings as well as in constraints such as availability of homophilous ties with same-sex peers. Indeed, networking research has found such sex differences—men are more likely to form homophilous ties across multiple networks and to have stronger homophilous ties; they furthermore reap greater network returns from similar individual and positional resources, as well as from homophilous relationships (Ibarra, 1992; van Emmerik, 2006). Specifically in academia, male colleagues often network closely with other male colleagues (e.g., Fox, 2010). Within the male-dominated environment of the university, social proximity has been shown to be lower for women than for men (van Emmerik, 2006), in part a result of having comparatively few women in male-dominated academic working environments. We will examine whether these networking circumstances will be reflected in citation networks, as less availability of homophilous ties should result in lower network density and differences in networking approaches may result in less cohesive networks among women compared with men.

Hypothesis 6 (H6): The citation network among female authors of top-quality communication research journal publications is less (H6a) dense and (H6b) cohesive than the citation network among their male counterparts.

Method

Data Retrieval

On October 31, 2009, all entries for publications in Communication Research (CR) and Journal of Communication (JoC) were downloaded from the ISI Web of Knowledge database. We focused on these outlets because two citation analyses examining knowledge linkage structures within communication studies identified them as most influential based on being most frequently cited by communication journals (Barnett, Huh, Kim, & Park, 2011; Park & Leydesdorff, 2009). The ISI database features citations counts but does not include authors’ first names. The downloaded entries thus included the authors’ last names and initials, the publication title, the publication year, the source, the cited references (which included last name and initials of lead author, outlet and year for each cited publication), as well as number of times this publication had been cited per ISI records. Book reviews were excluded from the download.

Coding of Authors’ Biological Sex

From this database, we extracted the author names for publications in the 15-years range between 1991 and 2005. To keep the coding manageable, we focused on this period. We used a cut-off year that was 4 years before the download to allow also the most recent included publications some time to receive citations. Three coders coded author names for biological sex based on entries for these articles in ComAbstracts, Communication and Mass Media Complete, and PsychInfo—in contrast to the ISI database, these databases feature authors’ first names. Several databases were employed, as occasionally a database would lack a publication, not cover the entire timespan (CMMC), or record communication publications only selectively (PsychInfo). For instances where the first name of an author was not clearly associated with a biological sex (e.g., Kelly, Chris, foreign names), the coders searched for the authors’ personal websites to obtain the information. When no adequate information could be found, the author’s sex was recorded as missing value. To establish intercoder-reliability, 10% of the author names were coded independently by all three coders—Krippendorff’s alpha was .91. Despite the reliable coding, all entries were cross-checked by at least two coders.

Furthermore, in response to a reviewer comment, we established that communication researchers can detect an author’s biological sex based on the name only, which is a precondition for role congruity theory to explain the citation patterns as hypothesized. For this purpose, we drew a random sample of 109 names from the list of authors and asked two female and two male communication researchers (with PhD degrees) to code them for biological sex (female, male, not sure) simply based on their impressions or preexisting knowledge of the authors, without further researching individuals. Together with the coding as used in the analyses, there were thus five categorizations for each name, based on which we computed a Krippendorff’s alpha of .91. Hence, we can assume that communication researchers spontaneously detect an author’s biological sex.

Coding of Cited Lead Authors’ Biological Sex

In addition, citations in these publications in CR and JoC made to other CR and JoC publications were then coded for biological sex of the lead author of this cited publication. The information was based on the citations recorded by the ISI database, not the original list of references. The rationale is to examine how the “pool” of scholars who were able to place publications in top-quality communication research outlets cited other individuals from this pool. We refer to these references as peer scholar citations to differentiate them from other citations. (For the other, “nonpeer scholar” citations, citers’ sex was not coded because we estimated that for more than 16,000 citations received from typically multiauthored publications, approximately 40,000 coding decisions would be needed, based on researching a greater variety of literature databases. This was not feasible for us.) For this purpose, we used the list generated through the process described in the prior paragraph. Any cited authors that did not appear in this list, as they had published before the examined time period, were coded through the same procedure as described above.

Categorization of Article Topics

The second hypothesis pertained to whether article topics affected how often an article was cited. For this purpose, we wrote a script to categorize the articles into “female-typed topic” and “male-typed topic” based on strings contained in the article title. If the title contained “child,” “parent,” “body,” or “women,” it was categorized as “female-typed topic.” If the title contained “CMC,” “computer,” “news,” “politic,” “journalis,” “Internet,” or “persua,” it was categorized as “male-typed topic.” These terms were established to reflect gender-typing in research topics per empirical data with the following two approaches.

To examine what research topic keywords were more associated with articles authored by female or male scholars, we employed data collected from a class of 20 graduate students, 11 women and 9 men, who had already completed a graduate overview course on contemporary communication theories. Eighteen research keywords were presented with this instruction: “Please respond quickly and spontaneously to the following questions. There are no right or wrong answers, so please just check boxes to report your intuitions.” Only the question on “gender” was of interest; others served to veil the purpose of the data collection. “Please picture a communication scholar who works on the topic <research topic>. What characteristics describe this scholar? Country of origin (Asian country/European country/USA, North America/Other), age group (20-29/30-39/40-49/50-59/60-69/70+), Gender (male/female), Ethnicity (African American/Asian/Hispanic, Latino/White/Other).” The following list of topics that were presented is sorted by the extent to which they were associated with male researchers: computer-mediated communication (100% male attributed), news (95%), politics (84%), journalism (84%), Internet, (79%), persuasion, (79%), organizations (74%), entertainment (63%), minorities (63%), health (42%), intercultural communication (37%), teenagers (26%), interpersonal communication (26%), children (21%), parenting (11%), body (5%), women (0%). Any research keyword that at least 75% of the respondents associated with one of the sexes was considered to indicate a “gender-typed” research topic term.

As a second approach to establish gender-typing in research topics, we analyzed whether the categorization of articles into female-typed and male-typed topic areas was related to authors’ sex, as reported in the results section under preliminary analyses.

Data preparation for analyses with authors as unit

The data were aggregated based on author names. Only the authors who had published in the 1991-2005 period were included in the analyses (N = 1,334). The aggregated data set contained the author’s biological sex, number of publications in CR/JoC total, total number of citations pieces received from 1991 to 2005, percentage of publications with both-sexes authorship, percentages for publications that were categorized to have a “female-typed topic” or a “male-typed topic,” and the average publication year. Before the aggregation, the list of names was examined for similar entries (e.g., “Zook, E” and Zook, EG”) to assign the same name label to both entries when the first name was equivalent. Likewise, slightly different hyphenated names were examined and received the same labels when they referred to the same individual (for instance, “BALLROKEACH” and “BALL-ROKEACH”). Two coders worked on this task for all names (Krippendorff’s α = .89). Great care was taken to ensure accuracy, as each unit was coded by two coders and cross-checked. With this procedure, even subtleties such as two different authors with almost the exact same name (e.g., K. L. Johnson and K. K. Johnson) could be identified and coded accordingly.

Data preparation for analyses with publication year as unit

The data were rearranged to examine changes over time. For each year, the percentages of publications with (a) female lead authors, with (b) all female authors/teams versus (c) both-sexes author teams versus (d) all male authors/teams were computed. Additional measures referred to the citations that these publications received—the percentages (with all citations that CR/JoC publications of that year had received as base) of citations received by publications from (e) female lead authors, by (f) all female-authored publications, (g) both-sexes author teams, and (h) all male authors.

Data preparation for network analysis

We extracted the authors that had at least three publications in CR/JoC in the examined time span, to ensure that a minimal “give and take” of citations could possibly occur. This sample consisted of 141 authors, 41 women and 100 men. Thus a 141 × 141 matrix was compiled to indicate who cited whom how often within this sample of communication scholars. Based on this file, we could also establish that male and female authors did not differ in average number of publications that contained self-citations (p = .33), when considering total number of publications of an author.

Results

Analyses Overview

After preliminary analyses, H1 and H2 were tested based on the citations the examined articles received. We first used lead authors’ sex as independent variable and then considered all authors’ sex—the findings for these two approaches are largely the same, but both analyses are reported to show the robustness of the pattern. If men publish more articles in the examined outlets, it is to be expected that overall more citations to males’ articles exist—this effect of proportions is avoided by reporting average number of received citations per article to clarify a potential Matilda effect. Furthermore, H3 was tested based on citations received (see “Data Preparation for Analyses With Authors as Unit” in the method section). Moreover, we tested H1, H2, and H4 with the citations made by the examined articles in their own list of references (see “Coding of Cited Lead Authors’ Biological Sex” in the method section). Again, to clarify the robustness of the results, we used lead authors’ sex and then all authors’ sex as independent variable, even though the findings for these two approaches are largely the same. A follow-up section assesses impacts in line with H1 and H4 based on how many articles were actually published by female versus male authors and thus could get cited. Last, we tested H5 (see “Data Preparation for Analyses With Publication Year as Unit” in the method section) and H6 (see “Data Preparation for Network Analyses” in the method section).

Preliminary Analyses

Number of articles and citations received

During the 1991-2005 time span, 444 research articles and reviews appeared in CR and 576 in JoC, resulting in a total of 1,020 publications for the study’s “population.” The average number of citations a publication had received by October 2009 was 15.86 (SD = 30.82).

Authorship by sex

The 1,020 publications were authored by 1,334 different individuals. Of these authors, 41.7% were women and 56.2% were men (for 2.1%, the author’s sex could not be determined). However, as many authors published several pieces in these outlets, the number of authorship indications was higher than 1,334, namely 1,993. For the authorship indications, 38.7% referred to women and 59.9% referred to men (1.5% unidentified sex). The number of authors on a publication was 1.95 on average (SD = 1.26). For 22.5% of the publications, the sole author or all authors were female; 47.6% had a male sole author or only male authors. Hence, 29.1% publications were authored by teams that included at least one person from each sex. The lead authors were women for 36.2% of the publications.

Gender-typing of research topics

Regarding gender-typed topics, 48 pieces (4.7%) fell into the “female-typed” category, 236 (23.1%) fell into the “male-typed” category, and 17 pieces (1.7%) were categorized to fall into both categories based on featuring strings associated with stereotypes for both genders. The vast majority of 711 (70.5%) emerged as “gender-neutral” based on the categorizations of gender-typed research topics.

To establish gender-typing in research topics, we analyzed whether the categorization of articles into female-typed and male-typed topic areas was related to authors’ sex. As reported in Appendix A, for the articles with either gender-neutral or male-typed topics, about half in both groups were authored by male authors only; about a fifth of these articles were authored by female authors only in both these groups. However, articles on female-typed topics were twice as likely to be authored by female authors only compared with the other articles, and articles with only male authors accounted for only a fifth in this group. The observed frequencies in these groups differed significantly from expected frequencies, χ²(6, N = 1,020) = 26.28, p < .001. Moreover, the first author of an article was about twice as likely to be female if the topic was categorized as female-typed compared with the other articles (approximately two thirds vs. one third; see Appendix A). Again, the observed frequencies differed significantly from expected frequencies, χ²(3, N = 1,009) = 40.82, p < .001. These preliminary analyses attest to gender-typing in research topics.

Citations made by examined articles

The examined “population” of 1,020 CR/JoC articles made 2,958 references to other publications that appeared in the same outlets (which we call “peer scholar citations”). The articles referenced 2.90 peer scholar publications on average (SD = 3.82; compared with the M = 47.30 for the total number of references used, SD = 29.90). The average article cited 0.76 (SD = 1.55) female peer scholars and 2.14 (SD = 3.03) male peer scholars as lead author (The ISI database records only the lead author of cited publications). Almost half (46.5%) of the articles included at least one peer scholar reference, with 46.1% citing at least one female peer scholar and 59.6% citing at least one male peer scholar.

Testing H1 and H2: Impacts of Lead Authors’ Sex and Topic on Citations Received

Addressing H1 and H2, we conducted an ANOVA with lead authors’ biological sex as between-group factor, a two-step between-factor that simply differentiated between male-typed topic versus other topics (to avoid small cell sizes), publication year as covariate, and citations received as dependent variable. It yielded two main effects. The impact of lead author’s biological sex emerged as significant, F(1, 1005) = 6.47, p = .011, η² = .006. Publications with female lead authors were cited 12.77 times on average (SD = 20.57), whereas publications with male lead authors were cited 17.73 times on average (SD = 35.34). This finding supports the Matilda effect suggested in H1.

It also was relevant whether the research topic was “male-typed” or not, F(1, 1005) = 5.90, p = .015, η² = .006, with male-typed publications receiving significantly more citations (M = 21.04, SD = 38.63 vs. M = 14.44, SD = 28.08). It should be noted that the Matilda effect emerged despite inclusion of research topic impact, which implies it is not simply an “epiphenomenon” resulting from different research interests that female and male scholars are likely to pursue and that attract more or less research attention. However, the interaction between these two factors hypothesized in H2 did not approach significance (p = .158). Thus, even though gender-typing of the research topic did affect citations, this impact was independent of the Matilda effect. Hence, H2 was not supported based on citations received as dependent measure (later analyses based on citations made will, however, provide support for H2).¹

Finally, publication year as covariate had a significant impact, F(1, 1005) = 14.48, p < .001, η² = .014. Interestingly, the effect sizes reflect that the two main effects together had almost as much impact as the length of time a publication had been out.

Testing H1 and H2: Impacts of All Authors’ Sex and Topic on Citations Received

For further tests of H1 and H2, we conducted an ANOVA with authors’ biological sex as two-step between-group factor (only female author[s] vs. male author[s] and authors from both sexes), two-step between-factor that simply indicated whether a publication fell into the “male-typed” category or not (to avoid small cell sizes), publication year as covariate, and citations received as dependent variable. The factor of featuring at least one male author emerged as significant, F(1, 1008) = 5.63, p = .018, η² = .006. Publications with at least one male author received significantly more citations with M = 17.11 (SD = 33.38), compared with M = 11.93 citations (SD = 19.84) for publications from female authors. In addition, the research topic had an influence as well, F(1, 1008) = 5.91, p = .015, η² = .006, as noted for the prior analysis. The interaction between the two between-group factors did not approach significance (p = .414). Hence, gender-typing of the research topic influenced citations but independently from the Matilda effect, meaning H2 was not supported based on citations received (later analyses based on citations made will, however, provide support for H2). The impact of publication year as covariate was significant, F(1, 1008) = 16.06, p < .001, η² = .016, and may be of interest regarding the effect size as a comparison for the main effects’ sizes.

Testing H3: Impact of Author Productivity

Addressing H3, analyses were conducted using authors as units. An ANOVA with citations received as dependent variable employed the biological sex of the lead author and whether the author had published five or more articles published in CR/JoC since 1980 as a proxy for productivity as two-step between-group factors. Five ore more articles was used as cut-off for high productivity based on a scree plot. The average publication year of the author’s publications and the percentage of publications with both-sexes authorship served as covariates. This analysis confirmed the importance of authors’ biological sex for citations received, F(1, 1300) = 21.57, p < .001, η² = .016, as reported above. Not surprisingly, productivity also affected citations received, F(1, 1300) = 428.77, p < .001, η² = .248. These two main effects, however, were qualified by an interaction, F(1, 1300) = 18.16, p < .001, η² = .014. This significant interaction supports H3, as the Matilda effect was much smaller among authors with up to four publications (M = 20.12, SD = 29.84 for females vs. M = 23.19, SD = 33.78 for males) than among prolific authors with five or more publications (M = 126.03, SD = 104.82 for females vs. M = 184.84, SD = 190.89 for males). The covariates were also significant (for average publication year F[1, 1300] = 30.70, p < .001, η² = .025, and for percentage of publications with both-sexes authorship F[1, 1300] = 5.88, p < .015, η² = .005). A regression analysis with productivity as metric predictor variable reconfirmed the findings regarding H3.²

Testing H1, H2, and H4: Impacts of Lead Authors’ Sex and Topic on Citations Made

So far, we have analyzed citations that the CR and JoC articles received. In the following analysis, we examine citations of other CR and JoC publications that were made in CR and JoC articles during the 1991-2005 period as recorded by ISI. Again, the rationale is to examine how the “pool” of scholars who were able to place publications in top-quality communication research outlets cited other individuals from this pool.

To test H1, H2, and H4, an ANOVA was conducted with biological sex of the lead authors from the citing publications as a two-step between-group factor, another between-group factor that differentiated the publications into those on “male-typed” topics versus others, publication year and total number of references as covariates, and the number of references to female versus male peer scholars as repeated measures. It should be noted that combined impacts of cited peers’ sex (including interactions) accounted for 9.8% of variance (see details below).

For the general number of citations to peer scholars, which is not relevant for the hypotheses, the analysis yielded an effect of the research topic, F(1, 1003) = 8.83, p = .003, η² = .009, as publications with “male-typed” topics cited peer scholars significantly more frequently (M = 3.59, SD = 3.85 vs. M = 2.69, SD = 3.79). Both covariates affected the number of citations to peer scholars, with F(1, 1003) = 31.81, p < .001, η² = .031 for publication year and F(1, 1003) = 174.86, p < .001, η² = .148 for total number of references.

Pertaining to H1, the analysis showed that female peer scholars were significantly less often cited than male peer scholars, F(1, 1003) = 12.92, p < .001, η² = .013, with M = 0.77 (SD = 1.44) compared with M = 2.15 (SD = 3.03). With regard to H2, the research topic affected how many female versus male peers were cited, as indicated in an interaction, F(1, 1003) = 8.75, p = .003, η² = .009. Male authors of publications in CR and JoC were especially more likely to get cited (for not male-typed topics: M = 1.95, SD = 2.97; for male-typed topics M = 2.76, SD = 3.15), compared with female authors (for not male-typed topics: M = .74, SD = 1.46; for male typed-topics: M = .84, SD = 1.38), when the topic of the citing publication was “male-typed.” Hence, the interaction suggested in H2 was supported for citations made by CR/JoC publications. In other words, references of the examined CR/JoC publications reflected that male authors of CR/JoC benefitted especially from publishing on male-typed topics.

The interaction hypothesized in H4 emerged as significant, F(1, 1003) = 24.38, p < .001, η² = .024. The effect is illustrated in Figure 1; means are reported in Table 1. Further elaboration is provided below in the section “Assessing Lead Authors’ and All Authors’ Sex’ Impact on Citations Made.”

Figure 1.

Proportion of female and male lead author articles cited by female and male lead authors, by all-female authored, both-sexes team-authored, and all-male authored publications in comparison to percentage of articles published by female and male lead authors

Table 1.

Impacts of Lead Author’s Sex and All Authors’ Sex on Citations of Female versus Male Peer Scholars

	Citing lead author’s sex		All citing authors’ sex
	Female	Male	Female only	Female and male	Male only
	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)
Female lead authors of CR/JoC publications cited	1.02 (1.80)	0.62 (1.18)	0.92 (1.88)	1.07 (1.52)	0.51 (1.06)
Male lead authors of CR/JoC publications cited	1.72 (2.90)	2.38 (3.08)	1.39 (2.70)	2.44 (2.98)	2.31 (3.15)

Note: CR = Communication Research; JoC = Journal of Communication.

It is worthwhile noting that the covariates of publication year and total number of references not only had a direct effect on how many peers were cited (see above) but also had influenced peer citations through interactions with cited peers’ sex, with F(1, 1003) = 12.99, p < .001, η² = .013 for the interaction with publication year and F(1, 1003) = 40.96, p < .001, η² = .039 for total number of references. Citations to male peer scholars were disproportionately frequent in CR/JoC publications with more references and that came out more recently (see details in Appendix B).

Testing H1, H2, and H4: Impacts of All Authors’ Sex and Topic on Citations Made

For further testing of H1, H2, and H4, the following analysis considered all authors instead just the lead author’s biological sex, but employed again citations made. An ANOVA with combinations of biological sex of authors as between-group factor (female only, male only, or both female and male), another between-group factor that simply differentiated between male-typed topics versus others (to avoid small cell sizes), publication year and total number of references as covariates, and the number of references to female versus male peer scholars as repeated measures was conducted. Notably, combined impacts of cited peers’ sex (including interactions) accounted for 9.6% of variance (see details below).

Aside from hypothesized effects, impacts on the general number of citations to peer scholars emerged: The analysis yielded an effect of the between-group factor for authors’ sex, F(2, 1005) = 7.43, p = .001, η² = .015, as the both-sexes author teams cited peer scholar publications more often (M = 3.52, SD = 3.76 vs. M = 2.31, SD = 4.10, for all female-authored pieces and M = 2.82, SD = 3.68 for all male-authored pieces). Furthermore, the research topic, F(1, 1005) = 7.77, p = .005, η² = .008, and an interaction between research topic and authors’ sex was significant, F(2, 1005) = 3.56, p = .029, η² = .007. Both covariates affected the number of citations to peer scholars, with F(1, 1005) = 26.23, p < .001, η² = .026 for publication year and F(1, 1005) = 175.80, p < .001, η² = .149 for total number of references.

Regarding differences between citations to female versus male peers as suggested in H1, the examined publications were again shown to cite female peer scholars less often than male peer scholars, as a significant impact of the within-group factor showed F(1, 1005) = 11.21, p < .001, η² = .011. In support of H2, the interaction between the within-factor of cited author sex and research topic was again significant, F(1, 1005) = 6.98, p = .008, η² = .007. The interaction hypothesized in H4 emerged as significant, F(2, 1005) = 15.12, p < .001, η² = .029. The effect is illustrated in Figure 1; means are reported in Table 1 (further elaboration is provided in the section “Assessing Lead Authors’ and All Authors’ Sex’ Impact on Citations Made”). The interaction between the within-group factor and the two between-group factors fell short of significance (p = .074). For the within-group factor of citing female versus male peers, interactions with the publication year, F(1, 1005) = 11.28, p = 001, η² = .011, and with the total number of references, F(1, 1005) = 39.95, p < 001, η² = .038, emerged as before (see details in Appendix B).

Assessing Lead Authors’ and All Authors’ Sex’ Impact on Citations Made

In the two prior sections, a significant impact in line with H1 was reported. However, a significant difference for citations to female versus male peer scholar is not surprising, given that only 36% of the articles in the 1991-2005 period had a female lead author (see preliminary analyses). Yet the difference may be considered larger than what is to be expected because the average percentage of citations referencing a peer publication by a female lead author was only 26% (SD = 33). We computed the percentage of female peer scholars cited of all cited peer scholars for each publication and ran a one-sample t test with 36% as test value, which showed that female peer scholars were cited significantly less frequently, t(668) = −7.55, p < .001, than what should be expected based on their share of lead-authored publications.

As discussed in the two prior sections, H4 received support when examining impacts of lead authors’ sex as well as all authors’ sex. Male lead authors cited less female peers than female lead authors did. Publications authored by both-sexes author teams and all male authors cited significantly more male peer scholars than publications by female authors only. Moreover, publications authored by male authors only cited significantly less female peer scholars than publications by both-sexes teams and all female authors.

To assess these impacts, the “pool” of CR and JoC lead authors that citing peers could draw on needs to be considered. Figure 1 shows the proportions in the citation behaviors of communication scholars, differentiated per biological sex of the lead author and of all authors. The bar on top reflects the proportions of articles published by female versus male lead scholars in CR and JoC. The second bar shows proportions for how often female authors cited articles from female versus male lead authors of publications in CR and JoC. The fact that the proportions of the first and the second bar are essentially the same indicates that female authors cited articles from female versus male authors in proportion to what was available for them to cite. The third bar displays proportions for how often male authors cited articles from female versus male lead authors of publications in CR and JoC reflects that these male authors cited publications by male scholars disproportionately often.

For further testing of the impact suggested in H4 based on the impact of the lead authors’ sex, we computed the percentage of female peer scholars cited of all cited peer scholars for each publication and ran one-sample t tests to see for which groups of publications this percentage differed from the 36.2% of female lead authored pieces in CR/JoC. Publications from female lead authors had an average percentage of 37.5%, which was not significantly different from the test value (p = .572). In contrast, publications from male lead authors had an average percentage of 21.1%, which was significantly different from the test value, t(436) = −10.7, p < .001. These findings support H4, as only male citers showed a disproportionate preference for male authors.

The same approach was extended to find out if one of the author groups (only female, only male, female-male author teams) showed a disproportionate preference for citing female or male lead authors. For this purpose, we again used the percentage of female peer scholars cited of all cited peer scholars for each publication in one-sample t tests with 36.2% as test value, as this was the percentage of female lead authored publications in CR/JoC. Articles from all female sole authors or teams had an average percentage of 37.7%, which was not significantly different from the test value (p = .615). In contrast, publications from mixed author teams, M = 31.7%, t(223) = −2.02, p = .045, and male sole authors or all male author teams, M = 18.6%, t(311) = −10.84, p < .001, undercited female authors in comparison with their share in authorship. These findings again support H4.

Testing H5: Changes Across Time

Linear regression analyses were performed to detect changes across time per H5 (see method section for data preparation details). The share of publications from female lead authors increased over the 15 years examined (see Cooper et al., 1989, for a similar trend), as publication year had a significant positive impact on their percentage among all CR/JoC articles of a given year (β = .714, p = .003, b = .010). The share of females among the lead authors of top-quality communication research rose about 1 percentage point each year. The share of citations received by publications with lead female authors increased at about the same rate (β = .530, p = .042, b = .012) when using publication year as predictor in a linear regression.

The percentage of publications from all female authors or a sole female author did not change over time (p = .467). However, articles from both-sexes authorships increased significantly (β = .647, p = .009, b = .012) whereas articles from all-male author teams and sole-author males decreased significantly (β = −.779, p = .001, b = −.014). Changes in citations received reflect these developments somewhat, though the strengths of changes differ. The share of citations received by articles from solely female authors did not change significantly. Citations that articles from both-sexes authorships received showed some increase, but it fell short of significance (β = .454, p = .089, b = .011). Citations received by articles from male-only authors decreased significantly (β = −.519, p = .047, b = −.014) but not as much as the share of all-male authored publications did.

Finally, when focusing on the citation behaviors among the authors of top journal publications in communication based on the percentage of female peer scholars cited of all cited peer scholars for each publication, we did not find a significant correlation with publication year for the total or subgroups by authors’ sex. Overall it appears that communication scholars have not changed in their gender-typed citation patterns.

Testing H6: Networking Effects

We used UCINET network analysis software (Borgatti, Everett, & Freeman, 1999) in order to examine H6 and computed measures based on a matrix of 141 authors with at least three publications in CR and JoC in the examined time span (see method section for data preparation details). Regarding H6a, network density (proportion of ties in a network relative to the total number possible) for the subsets of female and male scholars was not substantially different, with .13 for females and .15 for males. However, the networks among only female scholars and only male scholars differed in cohesion measures, reflecting that female scholars’ work was less interrelated per publication references. The average distance between females was 3.08 versus 2.61 for males, distance-based cohesion (“compactness,” range 0 to 1; larger values indicate greater cohesiveness) among females was 0.316 compared with 0.420 among males, and distance-weighted fragmentation (“breadth”) was 0.684 among females versus 0.580 among males. Thus H6b was supported for network cohesion.

Discussion

This study used the field of communication to examine differences in citation patterns in scholarly communication, drawing on gender-based reward inequities resulting from the public communication system of science. The so-called Matilda effect occurs when female scientists are underrecognized for their work as a result of their gender. We studied this effect based on citations because they represent the “currency” of reward and recognition in academia. Role congruity theory led to five hypotheses on moderators of the Matilda effect; another hypothesis resulted from a networking perspective. Five of our six hypotheses were supported; only H5, which suggested that the Matilda effect weakens over time, was not supported.

H1 proposed that males’ publications in top communication research journals receive more citations than those authored by females. H1 was supported—both for citations the examined publications received and for citations to same-journals articles the examined publications made themselves. The effect for citations received was weaker than the effect evident in CR/JoC authors’ own references pertaining to authors who published in these two journals. For the former, the proportion was 1 citation to a female lead author compared with 1.39 citations to male lead authors; for the latter, the proportion was 1 citation to a female lead author compared with 2.79 citations to male lead authors. This difference for citations received versus made suggests that citations within the smaller circle of scholars who publish in the same top journals are more strongly affected by the Matilda effect.

H2 proposed that the Matilda effect is particularly strong for publications on topics that are stereotypically more associated with male than female scholars. It was not supported for citations the examined 1,020 publications received but for the citations made to peer authors in these publications. When examining citations to peer scholars, it became apparent that only male scholars benefit from working on a male-typed topic. Again, the difference for citations received versus made suggests that citations within the smaller circle of scholars who publish in the same top journals are more strongly affected by the Matilda effect. Possibly, role incongruity perceptions affect citations more strongly when examined in a more cohesive group of individuals who publish in the same journals and may attend the same conferences. In addition, the fact that the effect remained significant despite inclusion of research topic impact implies that it is not simply an “epiphenomenon” resulting from different topic interests that female and male scholars are likely to pursue and that attract more or less research attention.

H3, which was supported, proposed that the Matilda applies more strongly to more productive authors, as highly productive female scientists violate the female gender role not only by holding the scientist role but also by fulfilling that latter role particularly well. In H4, we hypothesized that male authors would be less likely than female authors to cite female peer scholars. Male scholars indeed showed a disproportionate preference for citing male scholars, whereas female scholars did not show a preference for citing male or female scholars. Furthermore, we hypothesized in H5 that the Matilda effect would weaken over time. Yet H5 was not supported, possibly because academia is slow in changing toward greater gender egalitarianism or because the examined time frame of 15 years was too short to show the suggested change. Finally, even though no sex difference emerged for citation network density (H6a), H6b on sex differences in networking was supported due to female’ scholars lower network cohesion reflected in citations.

There are limitations to our investigation. Caution is in place concerning the classification of communication research topics into gender-typed categories. We based this on only 20 graduate students’ judgments and were able to confirm this with proportions of female and male authors working in these areas (see Appendix A). No broader data from a survey among communication scholars about their perceptions of research topics as being more or less associated with a gender was available to back this up. However, the applied categorization is plausible because considerable gender-typed differences exist between the sexes in the realm of vocational interests (Su, Rounds, & Armstrong, 2009), even though psychological research shows that sex differences are otherwise actually small (Hyde, 2005), Furthermore, we examined only two journals in a specific discipline. Even though these two journals were identified as the two “key players” (Park & Leydesdorff, 2009, p. 166; also Barnett et al., 2011) within communication studies, other studies have come to slightly different conclusions (Feeley, 2008; Levine, 2010). More intensive examination should test whether these findings hold true regardless of journal focus. For instance, the JoC has been found to include more diverse areas (Barnett et al., 2011) including women’s studies, which may actually result in a weaker Matilda effect than for journals with a stricter social science approach such as Human Communication Research because certain topics in the JoC could result in less role congruity perceptions regarding female authors.³ In addition, other studies in humanities, the sciences and so forth would need to be conducted to see if the same patterns hold true across disciplines. Last, the present network analysis is merely a first step into exploring how networking among communication scholars affects scholarly recognition reflected in citations. Future analyses could extend this examination, for instance, by using two-modes network analysis (e.g., Barnett et al., 2011).

The importance of the Matilda effect deserves some comment and illustrative speculation. The effects detected based on citations a publication received from any other academic sources were relatively small (1%-3% of variance explained when including interactions). However, when using citations made in two flagship journals of the communication research discipline as criterion, an author’s sex accounted for almost 10% of the variance when including interactions. It appears as if an author’s sex is of particular relevance within the core scientific community, in which scholars may more often know authors and be more aware of their biological sex but not so much when considering citations from any discipline. If citations reflect a scholar’s general recognition and reception, it is an astonishing finding that a tenth of that recognition is affected by one’s biological sex. Such substantial difference in recognition is very likely to affect tenure, promotion, and even hiring decisions. Furthermore, this Matilda effect is likely to translate into salary differences between female and male faculty, as citations are a determinant of earnings (Diamond, 1986). To elucidate the importance of the matter in strong terms: A 10% salary difference resulting from the Matilda effect would probably be considered huge by most academics.

The Matilda effect in citations may be just the tip of an iceberg, as role incongruity perceptions in the academic communication context may occur on a much broader basis. Women have been gaining ground with regard to refereed publications, which may not be influenced by role congruity perceptions due to blind reviewing, but areas in which gender is apparent seem to reflect how Eagly and Karau’s (2002) theory accounts for gender gaps in academia. Female scholars are faced with related drawbacks at all career fronts: in peer reviewing of grant and stipend applications, academic socialization and mentoring, academic networking, teaching evaluations, the myth that family obligations undermine women’s research productivity, lower likelihood of obtaining tenure-track positions and tenure potentially related to family circumstances, along with impacts of age at entry into profession and maternity, and different impacts of length of academic experience (see, for instance, Basow, Phelan, & Capotosto, 2006; Clark & Corcoran, 1986; National Science Foundation, 2004; Sax, Hagedorn, Arredondo, & Dicrisi, 2002; Sinclair & Kunda, 2000; van Emmerik, 2006; Wennerås & Wold, 1997; for an overview, see Zuckerman, Cole, & Bruer, 1991). In brief, women in academia are clearly faced with a much more complicated and difficult situation than men.

This “accumulative disadvantage” (Clark & Corcoran, 1986) ends up being substantial. The overall results of these patterns in scholarly communication may be cumulative, leading to an inability for female scholars to forward themselves in their academic careers. Regarding the example of communication research, assistant professors in communication are indeed often female (almost two thirds), but women are drastically underrepresented in tenured ranks (less than a third; data from Bunz, 2005, p. 707), which shows barriers in their career progress. Ultimately, women’s underrecognition and underrepresentation in higher academic ranks may threaten the goals of excellence in communication research, as has been suggested for other science disciplines (Cozzens, 2004).

Footnotes

Appendix

Appendix B

Impacts of Publication Year and Total Number of References on Citations of Female Versus Male Peer Scholars

	Total number of references of citing publication (median split)		Publication years of citing publication
	0-43	44 or more	1991-1998	1999-2005
	M (SD)	M (SD)	M (SD)	M (SD)
Female lead authors of CR/JoC publications cited	0.48 (1.03)	1.06 (1.72)	0.51 (1.09)	1.06 (1.71)
Male lead authors of CR/JoC publications cited	1.29 (2.06)	3.00 (3.56)	1.48 (2.42)	2.89 (3.45)

Note: CR = Communication Research; JoC = Journal of Communication.

Acknowledgements

The authors wish to thank Holly Hagner, Yuan Gong, and Laura Kerbeykian for their help with coding, Dr. Dongyoung Sohn for supporting the network analyses, Michael Huge for his help with manuscript preparation and classifying a subset of author names, as well as Dr. Axel Westerwick and Dr. Stacie Renfro Powers for classifying a subset of author names. Moreover, the authors thank the graduate students who helped with the categorization of article topics.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

Bios

Silvia Knobloch-Westerwick (PhD, University of Music, Drama & Media Hanover, Germany) is an Associate Professor at the School of Communication, The Ohio State University, Columbus/ OH. Her research interests include media effects and selective exposure to media messages in the contexts of news, health information, entertainment, political communication, and new communication technologies.

Carroll J. Glynn (PhD, University of Wisconsin-Madison) is a Professor and Director of the School of Communication, Ohio State University, Columbus, OH. Her research interests focus on understanding public opinion from a social and normative perspective.

References

Aiken

L. S.

West

S. G.

(1991). Multiple Regression: Testing and interpreting interactions. Newbury Park, CA: Sage.

Baldi

(1998). Normative versus social constructivist processes in the allocation of citations: A network-analytic model. American Sociological Review, 63, 829-846.

Barnett

G. A.

Huh

Kim

Park

H. W.

(2011). Citations among communication journals and other disciplines: A network analysis. Scientometrics, 88(2), 449-469.

Basow

S. A.

Phelan

J. E.

Capotosto

(2006). Gender patterns in college students’ choices of their best and worst professors. Psychology of Women Quarterly, 30(1), 25-35.

Blair

Brown

J. R.

Baxter

L. A.

(1994). Disciplining the feminine. Quarterly Journal of Speech, 80, 383-409.

Bolzendahl

C. I.

Myers

D. J.

(2004). Feminist attitudes and support for gender equality: Opinion change in women and men, 1974-1998. Social Forces, 83, 759-789.

Borgatti

S. P.

Everett

M. G.

Freeman

L. C.

(1999). UCINET 6.275 version. Natick, MA: Analytic Technologies.

Bryant

A. N.

(2003). Changes in attitudes toward women’s roles: Predicting gender-role traditionalism among college students. Sex Roles, 48, 131-142.

Bunz

(2005). Publish or perish: A limited author analysis of ICA and NCA journals. Journalism of Communication, 55, 703-722.

10.

Burroughs

N. F.

Christophel

Ady

J. C.

McGreal

(1989). Top published authors in communication studies. ACA Bulletin, 67, 37-45.

11.

Cameron

B. D.

(2005). Trends in the usage of ISI bibliometric data: Uses, abuses, and implications. Portal-Libraries and the Academy, 5(1), 105-125.

12.

Clark

S. M.

Corcoran

(1986). Perspectives on the professional socialization of women faculty: A case of accumulative disadvantage? Journal of Higher Education, 57(1), 20-43.

13.

Cole

J. R.

(1967). Scientific output and recognition: A study in the operation of the reward system in science. American Sociological Review, 32, 377-390.

14.

Cole

J. R.

Zuckerman

(1984). The productivity puzzle. In Maehr

M. L.

Steinkamp

M. W.

(Eds.), Advances in motivation and achievement (pp. 217-258). Greenwich, CT: JAI Press.

15.

Cooper

P. J.

Stewart

L. P.

Friedly

S. A.

(1989). Twenty years of research by and about women in major communication journals: 1967-1986. ACA Bulletin, 67, 46-61.

16.

Cozzens

S. E.

(2004). Gender issues in US science and technology policy: Equality of what? In European Commission (Ed.), Gender and excellence in the making (pp. 169-174). Luxembourg: Office for Official Publications of the European Communities.

17.

Davenport

Snyder

(1995). Who cites women? Whom do women cite? An exploration and scholarly citation in sociology. Journal of Documentation, 51, 404-410.

18.

Diamond

(1986). What is a citation worth? Journal of Human Resources, 21, 200-215.

19.

Eagly

(1987). Sex differences in social behavior: A social-role interpretation. Hillsdale, NJ: Erlbaum.

20.

Eagly

A. H.

Karau

S. J.

(2002). Role congruity theory of prejudice toward female leaders. Psychological Review, 109, 573-598.

21.

Etzkowitz

Kemelgor

Uzzi

(2000). Athena unbound: The advancement of women in science and technology. Cambridge, NY: Cambridge University Press.

22.

Feeley

T. H.

(2008). A bibliometric analysis of communication journals from 2002 to 2005. Human Communication Research, 34, 505-520.

23.

Ferber

M. A.

(1988). Citations and networking. Gender & Society, 2(1), 82-89.

24.

Fox

M. F.

(2010) Women and men faculty in academic science and engineering: Social-organizational indicators and implications. American Behavioral Scientist, 53, 997-1012.

25.

Griffith

B. C.

(1989). Understanding science: Studies of communication and information. Communication Research, 16, 600-614.

26.

Hâkanson

(2005). The impact of gender on citation. College & Research Libraries, 66, 312-323.

27.

Haslam

Ban

Kaufmann

Loughnan

Peters

Whelan

Wilson

(2008). What makes an article influential? Predicting impact in social and personality psychology. Scientometrics, 76, 169-185.

28.

Helmreich

R. L.

Spence

J. T.

Beane

W. E.

Lucker

G. W.

Matthews

K. A.

(1980). Making it in academic psychology: Demographic and personality correlates of attainment. Journal of Personality and Social Psychology, 39, 896-908.

29.

Hickson

Stacks

D. W.

Amsbary

J. H.

(1992). Active prolific female scholars in communication: An analysis of research productivity, II. Communication Quarterly, 40, 350-356.

30.

Holden

Rosenberg

Barker

(2005). Bibliometrics: A potential decision making aid in hiring, reappointment, tenure and promotion decisions. Social Work in Health Care, 41(3-4), 67-92.

31.

Hyde

J. S.

(2005). The gender similarities hypothesis. American Psychologist, 60, 581-592.

32.

Ibarra

(1992). Homophily and differential returns: Sex differences in network structure and access in an advertising firm. Administrative Science Quarterly, 37, 422-447.

33.

Ibarra

(1993). Personal networks of women and minorities in management: A conceptual framework. Academy of Management Review, 18(1), 56-87.

34.

Levine

T. R.

(2010). Rankings and trends in citation patterns of communication journals. Communication Education, 59, 41-51.

35.

Lievrouw

L. A.

(1989). The invisible college reconsidered: Bibliometrics and the development of scientific communication theory. Communication Research, 16, 615-628.

36.

Long

J. S.

(1992). Measures of sex differences in scientific productivity. Social Forces, 71(1), 159-178.

37.

Lutz

(1990). The erasure of women’s writing in sociocultural anthropology. American Ethnologist, 17, 611-625.

38.

McElhinny

Hols

Holtzkener

Unger

Hicks

(2003). Gender, publication, and citation in sociolinguistics and linguistic anthropology. Language in Society, 32, 299-328.

39.

Merton

R. K.

(1957). Priorities in scientific discovery. American Sociological Review, 22, 635-659.

40.

Merton

R. K.

(1968). The Matthew effect in science. Science, 159(3810), 56-63.

41.

Merton

R. K.

(1973). The sociology of science: Theoretical and empirical investigations. Chicago, IL: University of Chicago Press.

42.

National Communication Association. (n.d.). Earned degrees in communication by level of degree and gender (1970-2000) (WWW document). Retrieved on August 1, 2011, from http://www.natcom.org/Default.aspx?id=635&libID=656

43.

National Science Foundation. (2004). Gender differences in the careers of academic scientists and engineers. Arlington, VA: National Science Foundation.

44.

Nosek

B. A.

Smyth

F. L.

Sriram

Lindner

N. M.

Devos

Ayala

. . . Greenwald

A. G.

(2009). National differences in gender-science stereotypes predict national sex differences in science and math achievement. Proceedings of the National Academy of Sciences, 106, 10593-10596.

45.

Over

(1990). The scholarly impact of articles published by men and women in psychological journals. Scientometrics, 18, 71-80.

46.

Paisley

(1989). Bibliometrics, scholarly communication, and communication research. Communication Research, 16, 701-717.

47.

Park

H. W.

Leydesdorff

(2009). Knowledge linkage structures in communication studies using citation analysis among communication journals. Scientometrics, 81(1), 157-175.

48.

Rossiter

M. W.

(1993). The Matilda effect in science. Social Studies of Science, 23, 325-341.

49.

Sánchez Peñas

Willett

(2006). Gender differences in publication and citation counts in librarianship and information science research. Journal of Information Science, 32, 480-485.

50.

Sax

L. J.

Hagedorn

L. S.

Arredondo

Dicrisi

F. A.

(2002). Faculty research productivity: Exploring the role of gender and family-related factors. Research in Higher Education, 43, 423-446.

51.

Sinclair

Kunda

(2000). Motivated stereotyping of women. Personality & Social Psychology Bulletin, 26, 1329-1342.

52.

Rounds

Armstrong

P. I.

(2009). Men and things, women and people: A meta-analysis of sex difference in interests. Psychological Bulletin, 135, 859-884.

53.

Toutkoushian

R. K.

(1994). Using citations to measure sex discrimination in faculty salaries. Review of Higher Education, 18(1), 61-82.

54.

van Emmerik

I. J. H.

(2006). Gender differences in the creation of different types of social capital: A multilevel study. Social Networks, 28, 24-37.

55.

Wennerås

Wold

(1997). Nepotism and sexism in peer-review. Nature, 387, 341-343.

56.

Zuckerman

Cole

J. R.

Bruer

J. T.

(Eds.). (1991). The outer circle: Women in the scientific community. New York, NY: Norton.