Wednesday, October 5, 2022
HomeBiologyAffect of gender on the formation and consequence of formal mentoring relationships...

Affect of gender on the formation and consequence of formal mentoring relationships within the life sciences


Regardless of growing illustration in graduate coaching applications, a disproportionate variety of ladies depart educational analysis with out acquiring an unbiased place that permits them to coach the subsequent technology of educational researchers. To know elements underlying this pattern, we analyzed formal PhD and postdoctoral mentoring relationships within the life sciences through the years 2000 to 2020. Pupil and mentor gender are each related to variations in charges of pupil’s continuation to positions that enable formal educational mentorship. Though trainees of girls mentors are much less more likely to tackle positions as educational mentors than trainees of males mentors, this impact is lowered considerably after controlling for a number of measurements of mentor standing. Thus, the impact of mentor gender will be defined a minimum of partially by gender disparities in social and monetary sources obtainable to mentors. As a result of trainees and mentors are typically of the identical gender, this affiliation between mentor gender and educational continuation disproportionately impacts ladies trainees. On common, gender homophily in graduate coaching is unrelated to mentor standing. A notable exception to this pattern is the particular case of scientists having been granted an impressive distinction, evidenced by membership within the Nationwide Academy of Sciences, being a grantee of the Howard Hughes Medical Institute, or having been awarded the Nobel Prize. This group of mentors trains males graduate college students at increased charges than their most profitable colleagues. These outcomes recommend that, along with different elements that restrict profession decisions for ladies trainees, gender inequities in mentors’ entry to sources and status contribute to ladies’s attrition from unbiased analysis positions.


In academia, mentorship performs a key position as a determinant of success for each trainee and mentor [14]. Tutorial trainees spend a number of years coaching with simply 1 or 2 mentors, first to acquire a PhD after which typically as a postdoctoral fellow. In profitable mentoring relationships, trainees develop each their mental experience (via the training of recent abilities and ideas) and their skilled community (via the mentor’s sharing of educational connections and sponsorship). Conversely, mentors profit in the long term from their trainees’ success, because it allows an extra extension of their skilled networks and will increase peer recognition.

At current, there isn’t a consensus on whether or not the gender of the mentor influences the result of educational mentoring relationships. A number of research have examined correlations between mentor gender and trainee publication output, time to acquire a level or tenure-track place, or continued curiosity in pursuing a profession within the discipline. Nonetheless, the literature is combined on this level, with numerous research reporting constructive results of same-gender mentoring on the graduate or undergraduate stage [510], constructive results of mixed-gender mentoring [11], or no impact of gender on mentoring outcomes [12,13]. Proposed mechanisms linking mentor gender to trainee outcomes embrace elevated confidence and sense of belonging in traditionally male-dominated disciplines amongst ladies mentored by ladies [9] and variations within the stage of help or diploma of gender bias encountered by ladies in analysis teams predominantly composed of males or ladies [8].

Variations in coaching by males versus ladies mentors may replicate structural bias associated to the gender of the mentor [11]. Regardless of some current positive aspects in illustration, ladies stay underrepresented as school in lots of analysis fields [14] and lag behind males mentors in response to typical metrics of success. Gender-associated variations in mentor standing could in flip impression the power of trainees to carry out high-impact analysis and thus cut back the chance of constant on to formal educational mentorship roles. Such an impact could be according to earlier analysis inspecting cumulative benefit processes in educational careers (e.g., earlier outcomes suggesting that trainees of mentors with excessive trainee counts are inclined to grow to be mentors themselves [3,4]).

However the uncertainty across the impression of mentor and trainee gender on coaching outcomes, most research agree that there’s a tendency in mentoring relationships towards homophily, the formation of same-gender educational mentor–trainee pairs throughout each PhD [5,6,8,11,1519] and postdoctoral coaching [20]. Surprisingly little is understood in regards to the drivers that affect homophily. Specifically, it’s presently unknown if there are field-level variations. Most earlier work centered on a single scientific discipline (e.g., economics in [11,1517]), and the few research that encompassed a number of fields didn’t report the variation of homophily between them [13,20]. Maybe extra importantly, long-term developments in homophily inside or throughout scientific fields have by no means been investigated. If gender is a significant driver within the consequence of mentoring relationships, then the prevalence of homophily could be an essential issue shaping these outcomes.

The potential affiliation of homophily with different traits of researchers, and, specifically, their relative success, is usually uncharted. A noteworthy exception is a current survey in life science that reported a larger tendency for males school which can be recipients of a prestigious award to coach males college students and postdocs, in comparison with their males colleagues [20]. This examine has but to be replicated, and it’s unknown if its results generalize to different assessments of status.

To handle these questions, we examined a multidisciplinary database of PhD and postdoc-level coaching relationships [3], cross-referenced with information on publication, funding, and gender (as inferred from first names). We discover that gender homophily in graduate coaching is ubiquitous throughout fields, regardless of variations within the proportion of girls college students and college. Specializing in information from the life sciences, the place sampling of mentorship information is extra exhaustive, we discover that each pupil and mentor gender are related to totally different charges of trainee development to a tutorial mentorship position (i.e., a place that permits one function a proper supervisor to PhD college students and/or postdocs). Nonetheless, the results of mentor gender on this coaching consequence are considerably lowered after controlling for a number of measurements of the mentor’s standing. Gender homophily within the life sciences is usually unrelated to mentor standing. Nonetheless, a notable exception to this pattern is the particular case of scientists having been granted an impressive distinction, evidenced by membership within the Nationwide Academy of Sciences (NAS), being a grantee of the Howard Hughes Medical Institute (HHMI), or having been awarded the Nobel Prize. This group of mentors trains males graduate college students at increased charges than their most profitable colleagues. These outcomes recommend that institutional biases that have an effect on the careers of girls mentors not directly impression the careers of their trainees and that interventions to extend illustration of girls as trainees could also be focused at elite scientists.


Multidisciplinary educational mentorship dataset

We analyzed information from Tutorial Household Tree (AFT; obtainable at, a crowdsourced database of educational family tree [3,21]. The database integrates user-contributed and public information on educational coaching relationships and publications. A mentoring relationship was both explicitly indicated by database customers or inferred from authorship and supervision of a dissertation listed in ProQuest’s assortment of dissertations and theses.

We inferred mentor and trainee gender solely from first names. Gender inference was carried out utilizing Genderize, an algorithm that estimates the likelihood {that a} typical person of the identify identifies as a person or a lady based mostly on social media information recording how the identify is usually used [22]. Gender possibilities had been obtainable for 93.7% of people in AFT. Amongst this group, we excluded information from 4.3% of people whose first names didn’t have excessive likelihood of affiliation with one gender (see Strategies; S1 Fig and S2 Desk).

We examined coaching relationships with finish dates between 2000 and 2020, excluding information from coaching areas centered on enterprise and scientific medication (2.9% of coaching relationships excluded). The ensuing dataset included 109,784 mentors, 23,721 postdocs, and 365,446 college students from a variety of analysis areas in science, know-how, engineering, and arithmetic (STEM), humanities, and the social sciences (S2A Fig). For coaching at establishments in the US, the gender composition of graduate college students and postdocs throughout analysis areas was according to demographic information collected by the Nationwide Science Basis (NSF) (S2B and S2C Fig).

Gender homophily in graduate coaching

Homophily, the tendency for people to type relationships with these much like themselves, happens to various levels for a lot of facets of social life (race, class, gender, age, schooling, habits, attitudes and beliefs, and so forth.) [23]. To quantify gender homophily in mentoring relationships, we calculated the diploma to which same-gender mentoring relationships exceeded the proportion anticipated if trainees matched to mentors randomly. Distinguishing results of particular person preferences from constraints imposed by inhabitants construction is a perennial subject in research of homophily [2325]. When mentors of 1 gender are scarce relative to college students of that gender, full homophily is unimaginable: the larger the shortage, the decrease the utmost stage of homophily attainable (S3A Fig). We due to this fact normalized the worth of homophily in order that 0% signifies random trainee–mentor gender pairing and 100% signifies the utmost attainable worth, given the gender composition of the mentor and trainee swimming pools (S3B Fig and Eq 3).

Gender homophily happens amongst all common analysis areas we examined (Fig 1A and 1B). In all fields and all years, homophily was constructive, indicating a bent for mentors and college students of the identical gender to affiliate (median homophily throughout all fields and years = 20.5%). This pattern can be obvious on the stage of narrower fields (Figs 1C and S4 and S1 Desk, median homophily throughout 73 fields with any ladies mentors = 20.3%).


Fig 1. Gender homophily in graduate coaching.

(A) Gender ratio of all PhD college students from 2000 to 2020, break up throughout broad educational fields and mentor gender. Colours point out mentor’s gender. Subject abbreviations are reported within the legend. (B) Temporal developments in homophily, the tendency towards same-gender pairing between trainees and mentors. Every panel reveals information from one main discipline. Strains present a linear regression of homophily as a operate of commencement yr. P values point out significance of temporal pattern. (C) Homophily inside slim analysis areas (n = 73 areas). Strong line signifies median. The information and code wanted to generate this determine can be found on Zenodo (DOI: 10.5281/zenodo.4722020).

The diploma of homophily various significantly throughout analysis areas, with the strongest homophily in humanities and social sciences and the least in bodily sciences and engineering (S4A Fig). The diploma of homophily inside a analysis space was uncorrelated with its gender composition (S4B and S4C Fig). Nonetheless, evaluating narrower analysis areas with a minimum of 1,000 college students sampled confirmed a pattern towards correlation between homophily and the fraction of girls mentors or college students (S5 Fig, n = 29 analysis areas, Pearson’s correlation coefficient, homophily versus fraction ladies college students, r = 0.36, p = 0.06, homophily versus fraction ladies mentors, r = 0.37, p = 0.05), according to current work on gender homophily in coauthorship [25].

Gender homophily is reducing over time in some fields. In 6 of the 7 broad analysis areas, there was a major linear lower in homophily between 2000 and 2020 (Fig 1B, p<0.05, t take a look at on linear regression with time as unbiased variable and homophily as dependent variable). On the stage of slim analysis areas with greater than 1,000 college students, 10/29 confirmed a major lower and 19/29 confirmed no vital temporal developments (S6 Fig and S1 Desk).

The temporal developments noticed on the stage of analysis areas will be noticed on the stage of mentors grouped by educational seniority. We examined the subset of mentors with a minimum of 2 trainees and unbiased profession begin dates after 1970 (n = 37,962 mentors). We quantified the fraction of girls mentors, fraction of girls college students, and homophily as a operate of each mentor’s profession begin date (1970 to 2015) and pupil’s commencement date (2000 to 2015). The fraction of girls starting careers as mentors elevated over time from 1970 to 2015 (S7A Fig). After controlling for mentor’s coaching finish date, there was no relationship between the fraction of girls mentors and trainee’s coaching finish date. This end result means that the rise in ladies mentors was not pushed by mentor retirements between 2000 and 2015. The fraction of girls college students skilled by the mentors additionally elevated (S7B Fig). The lower in homophily throughout this era was associated to time (i.e., commencement yr), however not mentors’ educational age (S7C Fig).

Gender inequity in mentor standing and trainee continuation to educational mentorship roles

In line with earlier investigations into the attrition of girls throughout the educational profession monitor (generally referred to as the “leaky pipeline”) [14,2628], our outcomes present that the proportion of girls in social science and STEM fields is decrease at progressively later levels of the educational profession monitor, from graduate pupil to postdoc to educational mentor (S11A Fig). This end result signifies the inhabitants of educational mentors stays skewed towards males, even in analysis areas with pupil populations near gender parity. Nonetheless, it doesn’t in itself point out whether or not ladies graduate college students proceed on to educational mentorship positions on the identical price as males graduate college students. As well as, it doesn’t point out whether or not structural gender biases that have an effect on ladies as mentors not directly have an effect on retention of their college students in academia as formal mentors.

To handle these questions, we examined the proportion of graduate college students and postdocs within the life sciences that continued on to educational mentorship, accounting for elements that will impression continuation (see S1 Appendix). We hypothesized that if women and men mentors differ in standing (outlined right here as entry to funding, labor, and status markers comparable to citations) as a result of gender bias, these disparities would possibly result in variations in trainees’ continuation to educational mentorship roles. We due to this fact compiled a number of extensively used metrics to quantify mentor’s standing: h-index (a measurement of quotation price and publication manufacturing [29] based mostly on information from the Nationwide Library of Medication and Semantic Scholar), trainee depend (whole variety of PhD college students and postdocs mentored, a metric carefully associated to laboratory measurement [3]), the speed of funding granted by the US governmental businesses NSF and Nationwide Institutes of Well being (NIH), and the rank of the mentor’s educational establishment within the Quacquarelli Symonds World College Rankings, an annual evaluation that closely weights the establishment’s repute amongst teachers. Funding price, h-index, trainee depend, and institutional status had been all correlated with each other, suggesting that every one 4 metrics measured a standard trait of “mixture standing” (S8 Fig). To match mentors of the identical standing, we sorted mentors of all genders by every standing metric, then grouped them into as much as 10 bins of roughly equal measurement, such that mentors with the identical worth for a standing metric had been by no means positioned in several bins.

We restricted our evaluation of mentor standing to the life sciences. Sampling was extra full in these fields for two causes. First, information on publication and funding had been drawn from sources particular to this discipline (together with funding information from the NIH and publication information from the Nationwide Library of Medication). Second, as a result of the AFT started as an effort to crowdsource the educational family tree of neuroscience, its sampling of mentorship information is most dense for the life sciences. To keep away from false negatives in our identification of trainees that continued to educational mentorship roles, we additional restricted the evaluation to the subset of coaching relationships with cease dates earlier than 2010 and whose data had been manually edited by AFT customers (closing n = 11,112 mentors, 26,420 trainees, 26% of life science coaching relationships assembly different standards for evaluation). Though these standards lowered the scale of the dataset, they minimized the prospect of false negatives in our identification of development to educational mentorship. As a consequence of our strict definition of continuation as progress to educational mentorship, it’s probably that the continuation charges reported right here (see S1 Appendix) underestimate the precise proportion of trainees that stay in academia.

In comparison with ladies mentors, males mentors had increased imply charges of funding, trainee depend, and h-index, however not establishment rank (Fig 2A, p<0.00, Welch’s unequal variances t take a look at). In line with this discovering, males mentors had been overrepresented on the highest standing deciles for funding, trainee depend, and h-index, whereas ladies mentors had been overrepresented in decrease standing deciles (Fig 2B).


Fig 2. Relationship between mentor gender and mentor standing.

(A) Distribution of mentor-status metrics amongst women and men mentors in life sciences (n: whole variety of mentors within the continuation dataset with information for the standing metric, p: p-value of Welch’s unequal variances t take a look at for distinction in imply between women and men mentors). For improved visualization, 2 outliers are usually not included within the histograms (trainee depend = 75, h-index = 230). (B) Gender distribution of mentors after binning by standing (smaller numbers increased rank). Strong line signifies share of males mentors inside every bin. Dashed line signifies share of males mentors throughout all bins. The information and code wanted to generate this determine can be found on Zenodo (DOI: 10.5281/zenodo.4722020).

To check the speculation that structural gender bias amongst mentors not directly impacts the speed at which trainees proceed to educational mentorship positions, we match logistic regression fashions that predicted pupil and postdoc continuation based mostly on every mentor standing metric individually, trainee and mentor gender, mentor seniority, and coaching finish date (Fig 3, left). Together with coaching finish date as an unbiased variable accounted for long-term adjustments within the variety of trainees, standing variables and continuation charges [30]. Mentor seniority was included to manage for the chance that phenomena apparently associated to gender disparities in mentor standing may very well be defined as an alternative by gender variations in mentors’ educational age. We additionally match a mannequin that included the primary principal part (PC) of all 4 standing metrics as a single “mixture standing” variable (Fig 3, proper). To quantify the diploma to which variations in mentors’ standing account for variations in trainee continuation charges, we in contrast every mannequin to at least one by which mentor standing had been shuffled throughout trainees (Figs 3 and 4).


Fig 3. Affiliation between mentor standing, mentor gender, and trainee continuation in academia.

(A) Imply continuation price in academia for PhD college students and postdocs in life sciences. Every panel reveals information sorted in response to a distinct measure of mentor standing. Factors present imply continuation price of trainees of mentors with a given standing (smaller numbers point out increased standing), grouped by trainee and mentor gender. Strains present prediction of logistic regression mannequin, match to those variables in addition to coaching finish date and mentor seniority. Titles point out variety of mentors with respective standing information obtainable. (B) Marginal results of every unbiased variable on trainee continuation price, predicted by logistic regression fashions incorporating the standing variables in (A). Marginal results of gender present the impression of the mentor or trainee being a person relative to a lady. Marginal impact of mentor standing reveals impression of a 1-decile improve (worsening) in rank in comparison with others within the discipline. Marginal impact of mentor seniority reveals impression of a ten% improve within the variable. Error bars present 95% confidence intervals. Mild grey bars present marginal results for identical mannequin match to information with mentor standing shuffled throughout trainees. The information and code wanted to generate this determine can be found on Zenodo (DOI: 10.5281/zenodo.4722020).


Fig 4. Discount in mentor-gender results after controlling for mentor standing.

(A) Efficiency of logistic regression mannequin predicting trainee continuation based mostly on coaching finish date, mentor seniority, trainee and mentor gender, and mentor standing earlier than (strong line) and after shuffling (dashed) mentor standing throughout trainees. Fashions are match to the subset of life sciences information with info obtainable for the corresponding standing metric (see Fig 3). Efficiency of the standing shuffled fashions various as a result of the pool of mentors with obtainable information various between metrics. (B) Marginal impact of gender and temporal variables for fashions by which mentor standing is shuffled (dashed line) or shouldn’t be (strong strains), quantified as in Fig 3B. (C) P.c lower in marginal impact of mentor gender after incorporating every mentor standing variable into the logistic regression mannequin (p indicated for t take a look at between marginal impact for fashions with and with out every standing metric shuffled. The information and code wanted to generate this determine can be found on Zenodo (DOI: 10.5281/zenodo.4722020).

For all measures of mentor standing, increased rank was related to larger charges of trainee continuation to roles as educational mentors (observe constantly sloping strains in Fig 3). Being a person or the trainee of a person was additionally related to larger continuation charges. Nonetheless, this disparity was considerably lowered if one thought-about the overrepresentation of males in increased mentor standing ranks (Fig 2). Together with a measure of mentor standing within the mannequin considerably lowered the impact of mentor gender. For all measures of mentor standing, the magnitude of the mentor-gender impact was lowered relative to a mannequin by which mentor standing was randomized. This randomization had minimal impression on trainee-gender and temporal results (Figs 3B and 4). Thus, controlling for mentor standing reduces the obvious impact of mentor gender on trainee retention as educational mentors by as much as 49% (p<0.002, t take a look at, for all metrics besides trainee depend, Fig 4C). The utmost discount occurred within the mannequin that included h-index, and the combination standing metric didn’t have larger predictive energy than h-index.

Incorporating standing metrics additionally accounted for some results of mentor seniority (Fig 4B). In a stepwise comparability, randomizing information for each standing and seniority resulted in a larger discount of mentor-gender results than randomizing both alone, suggesting that gender variations in seniority don’t account for results of gender variations in standing (S9 Fig).

Individually analyzing information for graduate college students and postdocs (S10 Fig) confirmed constant results for mentor standing, mentor seniority, and trainee gender. Mentor-gender results didn’t attain significance amongst all subsets of the info, probably due to the lowered statistical energy obtainable in these smaller datasets.

Gender homophily and mentor standing

The variations in standing between mentors of various gender (Fig 2) means that gender homophily will pair trainees with mentors whose structural benefits or disadvantages reinforce their very own [31]. This impact could also be lowered or amplified if homophily differs throughout mentor standing ranges. We due to this fact analyzed how homophily and trainee gender differ in response to mentor standing. As well as, we individually analyzed information for extremely elite mentors, as indicated by a Nobel Prize, membership within the NAS, and/or funding by the HHMI. It’s nicely established that analysis communities are biased to worth the output of researchers who’ve established a excessive profile [32]. Furthermore, it has been reported that males mentors on this group practice fewer ladies than their friends, probably exaggerating the impression of homophily on this group [20].

Greater standing—as measured by funding price, h-index, trainee depend, establishment rank, or the combination standing metric—didn’t present a constant relationship with the proportion of males college students skilled by males or ladies mentors (Fig 5A). The diploma of homophily, which collapsed throughout each women and men mentors, didn’t present a relationship with mentor standing for any particular person standing metric or the combination standing metric (Fig 5B, p≥0.05, linear regression predicting homophily from mentor standing bin).


Fig 5. Relationship between mentor standing and gender homophily.

(A) Imply share of males college students amongst life sciences mentors divided by analysis space, mentor standing, and acquisition of elite awards. Horizontal dashed strains point out share of males college students for all mentors. Shade of strong strains signifies significance of a linear regression predicting the proportion of males college students for every mentor, based mostly on decile rank in standing and receipt of an award (see legend). Error bars present SEM. Numbers (n) point out whole mentors with legitimate standing measures for that group. (B) Imply homophily amongst life sciences mentors, grouped as in (A). Shade of strong line signifies results of linear regression predicting homophily based mostly on mentor standing. The information and code wanted to generate this determine can be found on Zenodo (DOI: 10.5281/zenodo.4722020).

In contrast, mentors that obtained a prestigious award tended to coach extra males college students than high-status mentors that weren’t prizewinners (Fig 5A). The proportion of males skilled by mentors that obtained prestigious awards was larger than mentors on the highest standing mixture standing rank (50% males in high decile versus 61% amongst awardees, n = 2,490 mentors, p = 6.1 × 10−22, t take a look at). This pattern was noticed amongst each women and men mentors (males mentors: 53% males in high decile versus 65% amongst awardees, n = 1,991 mentors, p = 4.3 × 10−21, t take a look at, ladies mentors: 36% versus 47%, n = 499, p = 1.7 × 10−6). The fraction of males skilled differed between award recipients and mentors ranked within the high decile in response to a number of standing metrics, suggesting that this discovering generalizes throughout a number of measurements of status (funding price: 49% males in high decile versus 61% amongst awardees, n = 2,075 mentors, p = 8.0 × 10−21, t take a look at, trainee depend: 44% versus 61%, n = 2,480, p = 3.0 × 10−53, h-index: 51% versus 61%, n = 2,480, p = 2.3 × 10−15, establishment rank: 51% versus 61%, n = 1,844, p = 1.3 × 10−11).

To extra rigorously take a look at the interplay between award receipt and the gender composition of pupil trainees, we turned to multivariate regression. A regression predicting fraction of males college students for every life science mentor based mostly on award receipt, mentor gender, and success rank confirmed an impact for award (β = 0.077, p = 2.1 × 10−7) and mentor gender (β = 0.16, p<1×10−29, n = 17,205 mentors) however solely a pattern for standing (β = −0.0017, p = 0.057).


Our outcomes point out that graduate-level mentoring relationships are fashioned at increased charges between trainees and mentors of the identical gender. This phenomenon of homophily happens constantly throughout wide-ranging analysis areas. Gender teams are additionally related to variations in mentoring outcomes, as measured by trainees’ continuation to positions that enable for formal supervision of doctoral college students and postdocs. Most prominently, ladies graduate college students and postdocs have decrease continuation charges than males, according to many earlier studies [14,2628]. We additionally discover that continuation charges within the life sciences depend upon mentor gender, as trainees of males mentors have increased continuation charges than trainees of girls mentors. Nonetheless, we discover a substantial portion of this latter affiliation is accounted for by the statement that males mentors in life sciences even have increased common rank in conventional measures of standing. These findings help a mannequin by which mentors’ entry to funding and labor, in addition to status markers comparable to citations, are distributed erratically by gender and in flip have an effect on trainees’ retention in academia as formal mentors. Compounded with gender homophily, these disparities in mentor standing could contribute to ongoing challenges across the retention of girls trainees in educational analysis.

The leaky pipeline phenomenon is widespread throughout fields

Our discovering that charges of continuation to educational mentorship are decrease for ladies graduate college students and postdocs replicates prior analysis that has recognized postgraduate profession transitions as factors of girls’s attrition from academia, though we observe that there’s a distinction between performing as formal mentor, holding a tenure-track place, and taking part in educational analysis in any capability [2628,33]. Though the fraction of girls graduate pupil trainees elevated between 2000 and 2010, the magnitude of “pipeline leak” didn’t change. Quite a few elements have been proposed to elucidate this phenomenon [3438]. Along with bias in evaluation and hiring, ladies could expertise larger obligations to household and childcare relative to males or lack of institutional help for balancing household and profession. Our information don’t instantly handle the relative position of those and different elements in inflicting the leaky pipeline, which stays a subject of debate. Nonetheless, we do present that disparities in retention related to trainee gender persist even after controlling for commencement yr (a proxy measure of the growing competitiveness of the educational job market [39]) in addition to (for the precise case of the life sciences) a number of measurements of mentors’ educational achievement. This end result signifies that pupil gender is related to differential outcomes even when evaluating college students which can be equally located when coaching ends. As well as, our information point out that attrition of girls throughout postgraduate profession transitions happens throughout a number of space of STEM and social science and finally impacts the scale of the pool of girls graduate mentors.

The position of educational mentor gender in trainee retention

We provide proof that gender-based disparities within the standing of mentors within the life sciences contribute to disparities within the price at which their trainees proceed to educational mentorship roles. As a result of our sources for information on quotation, funding, and mentor receipt of an award are most full for the life sciences, our outcomes on mentor standing are restricted to this discipline, not like our outcomes on homophily and the leaky pipeline phenomenon, which spanned a number of fields.

Our outcomes are according to earlier findings of disparities between women and men school on a number of typical measures of success. These embrace decrease variety of publications by ladies researchers [6,11,14], decrease charges at which their work is cited [40], and decrease ranges of funding [4145]. Gender variations additionally prolong to slower development via skilled positions [26,4649], in addition to extra delicate markers of profession achievement comparable to invites to current work at seminars [50].

The causes of gender-based disparities in typical measures of educational success are complicated. They might replicate long-standing bias within the educational group, which leads ladies’s competence or efficiency to be assessed on a distinct scale from males or in response to totally different qualities [34]. Proof of bias has been present in research of gender variations within the outcomes of fellowship functions [41], hiring assessments [36], and the content material of reference letters [51,52]. Results of bias could also be amplified by constraints and cultural expectations associated to household life, notably childcare, which impression ladies greater than males researchers, particularly in early and mid-career levels [14,27,53]. Latest longitudinal research point out that parenthood impacts ladies’s participation within the STEM workforce greater than males [38] and accounts for a big share of gender variations in analysis manufacturing [54]. Geographical constraints as a result of companions’ skilled employment may impression ladies in academia to a larger diploma than males [55].

Proposed mechanisms linking mentor and trainee gender to coaching outcomes embrace elevated confidence and sense of belonging in a traditionally male-dominated discipline amongst ladies mentored by ladies [9]. College students have additionally reported qualitatively totally different help relying on gender. Particularly, ladies graduate college students usually tend to report benefiting from psychosocial help from a mentor throughout their PhD coaching (comparable to offering emotional help and taking an curiosity within the pupil’s private life, cf. [56]), whereas males usually tend to report operational help (comparable to being concerned in chairing a convention, collaborating on papers, or being really useful to colleagues, cf. [5759]). It’s not clear from this descriptive proof whether or not such variations in help are pushed by mentor preferences, trainee wants, or results of gender stereotypes on trainee expectations of mentors [60]. Girls school usually assume a larger share of educational service duties than males [61], and variations in mentoring calls for could add to this disparity.

Future analysis might take a look at whether or not covariates of mentor gender, along with standing, have an effect on retention in academia as mentors. For instance, trainees of women and men mentors could also be judged in a different way, even when the mentors have comparable {qualifications}, reflecting a common gender bias in academia [34,36] (however see [37]). Women and men mentors’ social networks could differ in gender composition, as recommended by analyses of coauthorship [25,62], or in different options comparable to measurement. Lastly, gender variations in self-promotion kinds (e.g., self-citation and use of constructive language to explain analysis outcomes) could also be imitated by trainees [63,64]. Future analysis might additionally study whether or not subfields or subjects throughout the life sciences differ within the relationship between mentor gender, mentor standing, and trainee continuation.

A number of research which have documented constructive results of same-gender mentoring for ladies college students are appropriate with our particular end result. Specifically, our findings are appropriate with proof that contact with ladies position fashions has constructive results on ladies’s persistence in STEM careers at levels previous to graduate faculty [9,10]. Our outcomes are additionally appropriate with proof that ladies’s careers profit from a social community that features ladies throughout graduate schooling [65]. Our outcomes differ from a current examine that discovered increased continuation charges amongst ladies in chemistry that work with ladies PhD mentors, after controlling for college students’ analysis productiveness [8]. Nonetheless, our common discovering that obvious variations between women and men mentors are lowered after controlling for standing is according to a current preprint on gender and analysis productiveness [66]. We view such outcomes, like our personal, as cause to work for fairness in institutional help for ladies mentors in academia.

Given the numerous elements that may result in gender disparities, it’s unlikely that fairness will be achieved with a single coverage change. For instance, evaluation of NIH information means that outcomes of grant assessment are usually not biased towards ladies candidates [67]. Nonetheless, a disparity in funding ranges persists, which should replicate variations within the price at which ladies apply for funding or within the measurement of their awards. On this examine, the standing variable that accounted finest for the disparity in mentorship outcomes was h-index. Whereas h-index will be criticized as overly slim in its scope, this statement is according to the concept that a researcher with excessive publication impression is more likely to be well-known and revered inside their discipline. By this logic, funding by universities {and professional} societies in applications that allow ladies mentors to extend the impression of their work in ways in which will increase metrics comparable to h-index may profit trainees searching for positions as educational mentors.

Homophily in educational mentorship

A current survey of life science researchers suggests each college students’ and mentors’ preferences affect gender homophily in mentoring: Applicant swimming pools are skewed towards the gender of the mentor, however the gender composition of the ultimate analysis group matches the mentor’s gender extra carefully than the applicant pool [19]. Comparisons of homophily throughout differing ranges of group, such because the subfield, division, or analysis group might assist to increase this image. For instance, researcher gender composition varies throughout subfields [68]. If a subfield incorporates many college students and mentors of 1 gender, this might improve the diploma of homophily throughout the discipline of which it’s a half (see [25] for comparable observations on homophily in coauthorship). Gender variations in subfields could also be influenced by the diploma to which explicit analysis subjects or strategies match with internalized gender stereotypes, or whether or not the tradition of the subfield makes college students of a specific gender really feel that they belong [69,70]. Comparisons of gender homophily throughout subfields might due to this fact reveal the extent to which drivers of gender homophily lie outdoors the method of making use of for analysis supervision. A examine of gender homophily in coauthorship of life sciences articles discovered solely a weak relationship between the diploma of homophily and the journal’s self-discipline [25], suggesting that homophily is pushed by selection of coauthors reasonably than gender variations in analysis subject selection. Our discovering that there’s solely a weak correlation between homophily and analysis space suggests comparable causes for gender homophily in mentorship. One other attainable affect on homophily is the analysis group itself, which is a supply of casual mentorship, acculturation, and help, notably in STEM. A way of affinity for the analysis group might additionally draw college students to work with explicit mentors. It will due to this fact be fascinating to know whether or not there is kind of gender homophily in fields the place mentoring is extra one-on-one. Future analysis might additionally study how gender homophily interacts with different demographic variables, notably race and ethnicity.

We present that homophily is unrelated to markers of mentor standing, excluding receipt of a prestigious award. Our outcomes thus construct on the discovering that elite males mentors within the life sciences make use of fewer ladies than their colleagues [20], suggesting that it doesn’t generalize to different indicators of status (as it could if we had discovered a constant lower within the share of males college students throughout success ranges). As an alternative, mentors that obtain prestigious awards are a particular case, however an essential one, given their position as feeder labs for unbiased researchers [20].

Limitations of the present examine

Provided that we depend on observational information, our means to establish causes is proscribed. We’ve got tried to make use of appropriately certified language to explain our findings, and to debate their relationship to managed experiments on gender. Nonetheless, we additionally observe that there’s a wealthy custom of utilizing observational information and statistical fashions to check how gender impacts educational careers below real-world circumstances. A crucial side of this method is to establish underlying elements that designate variations between noticed teams [71]. The evaluation of mentor standing illustrates this method, the place an obvious impact of mentor gender will be defined by variations within the sources and status related to males versus ladies mentor teams. As fashions are refined with extra detailed and quantifiable variables, they could be used to drive experiments that take a look at causal relationships.

Static, binary gender classes are a simplification of the complicated social and organic actuality of intercourse and gender [72,73]. As a result of probabilistic nature of the strategies used on this examine, we had been unable to establish transgender, intersex, and/or nonbinary people within the information. There may be proof that transgender standing influences experiences in academia. An account by a distinguished transgender scientist signifies that gender transition affected his therapy by colleagues [74]. Survey information point out that transgender graduate college students expertise stress in day-to-day interactions with friends and college as a result of their gender id [75,76]. We hope that future analysis on gender and mentoring will combine findings from research that leverage the big pattern sizes obtainable via automated evaluation of first names with evaluation of survey information that includes extra complicated understandings of gender. Ideally, these research would think about the range of experiences throughout the transgender inhabitants (e.g., transmasculine and transfeminine, age and profession stage of gender transition, nonbinary and binary).

An essential limitation of this examine is its concentrate on placement in positions that allow formal supervision of doctoral college students and postdocs as a measurement of coaching consequence. Formal educational mentorship does play an outsized position in shaping analysis communities and thus deserves particular consideration. On the identical time, this mentorship position is just one of many alternative and beneficial outcomes that educational coaching makes attainable. People that go right into a teaching-focused job at a non-PhD granting establishment could have an effect on undergraduates’ persistence in STEM [9,10] or different fields. Trainees may instantly advance scientific data via work in business or conducting and supervising analysis at non-PhD granting educational establishments. We concentrate on doctoral and postdoctoral supervisors due to the position they play in defining analysis issues and coaching the subsequent technology of school, in addition to the chance that gender disparities on this particular trajectory could also be an indicator of extra widespread sexism in academia.

Supplies and strategies

Knowledge for the present examine had been drawn from the AFT (, a web based database of mentoring relationships [3,21]. The database data the id of the mentor and trainee, the kind of coaching (graduate or postdoctoral), and the beginning and finish yr of the coaching.

The AFT derives info on coaching relationships from 2 sources: crowdsourced (i.e., user-provided) information and ProQuest’s assortment of dissertations and theses. In a earlier examine, a portion of the crowdsourced information was validated by comparability with information on formal mentoring relationships indicated on school internet pages [21]. To restrict heterogeneity in the kind of mentoring relationships included [77], ProQuest information had been restricted to data of dissertations that resulted in doctoral levels inside a current time interval (2000 to 2015) and listed the dissertation advisor. When the identify and institutional affiliation of an advisor listed in ProQuest matched the identify and discipline of examine of a mentor included in crowdsourced information, the 2 had been thought-about to confer with the identical particular person. If no identify match was discovered at any establishment, a brand new node was added for the advisor. If a reputation match was discovered at one other establishment, the coaching relationship was not added till it may very well be reviewed manually as being a match to an present mentor or a brand new mentor.

As of November 2020, the AFT contained information on 724,657 researchers and 695,045 coaching relationships. Knowledge for 397,008 coaching relationships (57%) had been populated based mostly on the ProQuest dissertation database. Trainees and mentors for present AFT information had been crammed in based mostly on identify and institutional affiliation matches to the ProQuest information.

We analyzed information from 79 labelled analysis areas (S2 Fig). As a result of these labels for areas are added to the AFT by public contributors, the scale and specificity of their respective analysis communities varies. We due to this fact grouped information into 8 broad fields based mostly on the classes used within the NSF’s Survey of Earned Doctorates (S2 Fig).

There’s a correlation between the fraction of PhD-level coaching relationships in every broad discipline within the AFT dataset and concurrent information from the NSF’s census of PhD-level coaching relationships (r = 0.97, p = 0.0003, AFT information restricted to coaching relationships at US establishments). Thus, the sampling within the general AFT dataset matches that reported by the NSF. Inspecting crowdsourced information alone reveals an overrepresentation of the life sciences and bodily sciences relative to their proportion in NSF information. This pattern could also be as a result of historical past of AFT, which started as an effort to trace the educational family tree of neuroscience via crowdsourcing [21].

Geographic places had been obtainable for 91.8% of coaching relationships and indicated that the overwhelming majority (90.2%) of information was drawn from US establishments. The gender composition of US graduate college students throughout analysis areas was extremely correlated with equal information from the NSF Survey of Earned Doctorates, an annual demographic examine of US graduate applications (Pearson’s correlation coefficient, r = 0.97, p = 0.0003, S2B Fig). Complete demographic information on postdocs by gender, discipline of examine, and yr of coaching finish date weren’t obtainable. Nonetheless, the gender composition of US postdocs in STEM and social sciences was correlated with information from the 2015 NSF Survey of Graduate College students and Postdoctorates in Science in Engineering, a cross-sectional survey (r = 0.98, p = 0.003, S2C Fig).

Knowledge evaluation was applied in Python and R [7883]. The subset of AFT information analyzed on this undertaking, together with code for producing figures, is accessible on Zenodo (DOI: 10.5281/zenodo.4722020).

Gender inference

Researcher gender was inferred from first names utilizing, a web based portal that depends on social media information to estimate the likelihood {that a} first identify is related to a person figuring out as a person or a lady [22]. Statistical evaluation of authors’ first names have beforehand been to used to check gender variations in educational publication and quotation [40,68,84,85]. In earlier evaluations, it has been proven to have excessive ranges of accuracy when utilized to editorial boards of educational journals [86] and writer lists [22,85]. As a result of some names present variations of their typical gender (e.g. “Robin” is usually a person’s identify in the UK, however not within the US), utilization information for gender inference had been drawn from the nation by which a person’s educational establishment is positioned. When no location info was obtainable, we used gender estimates based mostly on utilization information pooled throughout all nations.

We excluded information for researchers whose names weren’t clearly related to one gender. To keep away from attainable bias in measurement error [84], we adjusted the thresholds barely for assigning gender to names to stability the likelihood of error for women and men. Our process for adjusting the edge was as follows:

(1) For every likelihood p (0 = girl, 1 = man), discover the variety of names, n(p). (2) For every likelihood, calculate the variety of names anticipated to be falsely assigned to every gender (e.g., for males anticipated error at likelihood p = n(p)−p*n)(p)). (3) Utilizing the outcomes from (2), calculate the cumulative likelihood of error if the edge is ready at every attainable worth of p. (4) Select a threshold to make use of for one gender. (5) Utilizing the outcomes from (3), discover the corresponding cumulative likelihood of error for one gender at this threshold. Select the edge for the remaining gender that provides an equal cumulative likelihood of error.

Following this process with a hard and fast threshold of p≥0.75 for together with putative males leads to a threshold of p≤0.24 for together with putative ladies (see S1 Fig). A comparability with stricter thresholds on key outcomes confirmed no constant distinction between fastened and balanced thresholds (S2 Desk).

We manually validated the gender estimated via on a randomly chosen subset of n = 2,184 researchers with a profile image on the internet portal of the AFT. Particularly, every profile photograph was offered to 2 totally different scorers (out of three whole scorers) who had been instructed to report the obvious gender of the particular person, within the absence of every other clue beside the image. The scorers might report one of many 3 choices: “male,” “feminine,” and “ambiguous/laborious to inform from this image”. Gender possibilities based mostly on first names had been obtainable for 2001 researchers within the pattern. We excluded footage reported as ambiguous by one or each scorers (n = 24 reported ambiguous by each scorers, n = 189 reported as ambiguous by one scorer). There have been no cases by which one scorer perceived the person within the {photograph} as a person and the opposite perceived the person as a lady. Scorers usually marked photographs as ambiguous as a result of technical errors in loading the photograph reasonably than uncertainty in regards to the particular person’s gender presentation. We discovered a excessive price of settlement between classification based mostly on first names by way of and scorers’ studies based mostly on photographs (n = 1,788 researchers, space below ROC curve = 0.99). An in depth breakdown of error charges is supplied in S3 Desk. The magnitude of error charges is much like these reported in a earlier examine that used algorithmic identification of gender based mostly on names [84].


We measured homophily because the diploma to which same-gender mentoring relationships exceeded the proportion anticipated if trainees matched to mentors randomly. Homophily was first calculated individually for women and men:

General homophily was then computed as their sum, weighted by the entire variety of coaching relationships with mentors in every group:

Optimistic values point out that college students and mentors of the identical gender are inclined to work collectively, whereas destructive values point out that college students and mentors of various gender are inclined to work collectively. A worth of 0 signifies that college students of any gender have an equal likelihood of coaching with mentors of any gender. Values of homophily had been normalized in order that 100% signifies the utmost attainable worth, given the gender composition of the mentor and trainee swimming pools.

To show results of the gender composition of a analysis discipline on homophily, we carried out a simplified simulation (S3 Fig). We assumed that every one people within the discipline had a set propensity to type same-gender mentor–trainee pairs, described by a parameter (a) between 0 and 1. This parameter decided the preliminary variety of man–man and girl–girl mentor–trainee dyads within the discipline:

The remaining mentors and trainees within the pool had been matched randomly. Homophily was measured from the ensuing inhabitants of mentor–trainee pairs. S3 Fig reveals outcomes from simulations together with 1,000 mentors and 1,000 trainees.

To extra clearly present change over time, the plot of temporal developments of the gender ratio of scholars and mentors in S6 Fig was smoothed with a 5-year transferring common. For the regression evaluation of temporal developments, homophily was computed from the uncooked information (Fig 1B and S1 Desk).

For the supplemental determine of temporal developments in slim analysis areas (S6 Fig), in addition to comparisons of homophily throughout analysis areas, solely fields with a minimum of 1,000 college students (29/73 fields with any ladies mentors) had been included. Knowledge from analysis areas with lower than 1,000 college students had been included in evaluation reported elsewhere within the paper, together with mixture statistics on homophily (see Outcomes, S1 Desk).

Mentor standing

We computed 4 metrics of a mentor’s educational standing:


The utmost quantity h such that the mentor has h publications with a minimum of h citations and all different publications have ≤h citations [29]. Quotation information had been drawn from the Semantic Scholar database for papers linked to researchers based mostly on string matches to their identify and the names of related trainees and mentors [3].

Establishment rank.

The rank of the mentor’s establishment within the 2015–2016 Quacquarelli Symonds World College Rankings (QS rankings). Establishment names in AFT had been matched to college names within the QS rankings utilizing fuzzy string matching [87]. Matches with lower than 95% similarity between characters had been excluded. The place the supply information supplied an interval reasonably than an actual rank, the midpoint of the interval was used because the rank for all establishments inside it.

To acquire an mixture measure of standing, we calculated the primary PC of trainee depend, funding, h-index, and establishment rank throughout mentors, which we confer with as “mixture standing.” Solely mentors with obtainable information for h-index had been included on this evaluation. When different standing metrics weren’t obtainable, they had been imputed with the imply worth for that educational discipline. Knowledge for every metric had been normalized by subtracting the imply and scaling to unit variance earlier than performing principal part evaluation (PCA).

To look at the connection between mentor standing and retention in academia (Figs 4, 5, S9 and S10), we match logistic regression fashions predicting the likelihood that trainees will themselves proceed to mentorship (p) based mostly on their coaching finish date (yr), mentor seniority on the time of coaching finish date (seniority), mentor and trainee gender, and mentor standing:
the place standingrank signifies the mentor’s rank (approximate decile) on one of many standing measures mentioned above relative to others within the discipline. To quantify the fraction of variance defined by including mentor standing (
Fig 4), we match the fashions above to information by which mentor standing had been shuffled throughout trainees. The change within the marginal impact (dy/dx) is:

In all logistic regression fashions, gender was coded as “1” for males and “0ˮ for ladies and the minimal coaching yr within the dataset was subtracted from all years. Mentor seniority was transformed to a 1- to 10-point scale to permit for comparability of marginal results throughout fashions.

To look at the connection between mentor standing and homophily on the particular person stage, we match linear fashions predicting the fraction of males college students skilled by every mentor based mostly on the mentor’s standing and receipt of a prestigious award (see beneath):
the place standingrank signifies the rank (approximate decile) that the mentor’s standing falls into compared with different mentors within the discipline, and award is a categorical variable indicating whether or not the mentor obtained a Nobel, NAS membership, or HHMI funding.

We in contrast this to a mannequin predicting homophily as outlined in previous analyses from mentor standing:

As a result of the measurement of homophily is outlined on the stage of the group, we couldn’t embrace award receipt as a variable within the regression mannequin.

To establish Nobel laureates, members of the NAS, and HHMI grantees, information had been drawn from official web sites (;;, then linked to researcher first names utilizing fuzzy string matching. Hyperlinks with lower than 95% similarity between characters had been excluded. AFT profiles manually recognized as Nobel laureates by contributors to the database had been additionally included.

Supporting info

S3 Fig. Homophily versus gender composition.

(A) Simulation of how measurements of gender homophily are affected by the gender composition of mentor and trainee pool inside a discipline (see Strategies). Every heatmap reveals simulations for a distinct propensity for mentors and trainees to type same-gender pairs, starting from 0 (none) to 1 (most). Scatter plot reveals precise mentor and trainee gender composition for information used within the homophily evaluation. Every level represents common composition for one slim analysis discipline. When mentors or trainees of 1 gender are scarce, homophily doesn’t replicate the underlying propensity to type same-gender pairs. (B) Simulation, after correcting for results of gender composition of pool, with precise common compositions overlaid, as in (A). The information and code wanted to generate this determine can be found on Zenodo (DOI: 10.5281/zenodo.4722020).


S2 Desk. Various thresholds for gender classification.

Every row reveals median gender homophily throughout all fields and years, marginal impact of trainee and mentor gender on trainee continuation to mentorship (see Fig 3), and alter within the impact of mentor gender after controlling for mentor mixture standing for a single threshold. Threshold refers back to the minimal/most likelihood of a primary identify figuring out a person for classification as a person/girl. Excluded signifies the p.c of particular person excluded for having ambiguous names in response to that threshold. For all thresholds, results of mentor gender weren’t vital in fashions that accounted for standing (see Fig 4). **: p<0.0001, *: p<0.05. Knowledge utilizing different thresholds can be found from the authors on request. The code wanted to generate this desk is accessible on Zenodo (DOI: 10.5281/zenodo.4722020).



  1. 1.
    Kram KE. Mentoring at work: Developmental relationships in organizational life. Lanham, MD: College Press of America; 1988.
  2. 2.
    Ragins BR, Kram KE, Ragins B, Kram Okay. The roots and that means of mentoring. The handbook of mentoring at work: Principle, analysis, and apply. Thousand Oaks, CA: Sage; 2007. p. 3–15.
  3. 3.
    Liénard JF, Achakulvisut T, Acuna DE, David SV. Mental synthesis in mentorship determines success in educational careers. Nat Commun. 2018;9(1):4840. pmid:30482900
  4. 4.
    Malmgren RD, Ottino JM, Nunes Amaral LA. The position of mentorship in protégé efficiency. Nature. 2010;465(7298):622–626. pmid:20520715
  5. 5.
    Principal JB. Gender homophily, Ph.D. completion, and time to diploma within the humanities and humanistic social sciences. Rev Excessive Ed. 2014;37(3):349–375.
  6. 6.
    Pezzoni M, Mairesse J, Stephan P, Lane J. Gender and the publication output of graduate college students: A case examine. PLoS ONE. 2016;11(1):e0145146. pmid:26760776
  7. 7.
    Lee SE. Profession Development Outcomes in Tutorial Science, Know-how, Engineering and Arithmetic (STEM): Gender, Mentoring Sources, and Homophily [PhD]. Tempe, AZ: Arizona State College; 2017. Obtainable from: with.web/2286/R.I.46343.
  8. 8.
    Gaule P, Piacentini M. An advisor like me? Advisor gender and post-graduate careers in science. Res Coverage. 2018;47(4):805–813.
  9. 9.
    Dennehy TC, Dasgupta N. Feminine peer mentors early in school improve ladies’s constructive educational experiences and retention in engineering. Proc Natl Acad Sci U S A. 2017;114(23):5964–5969. pmid:28533360
  10. 10.
    Hernandez PR, Adams AS, Barnes RT, Bloodhart B, Burt M, Clinton SM, et al. Inspiration, inoculation, and introductions are all crucial to profitable mentorship for undergraduate ladies pursuing geoscience careers. Commun Earth Environ. 2020;1(1):7.
  11. 11.
    Hilmer C, Hilmer M. Girls serving to ladies, males serving to ladies? Similar-gender mentoring, preliminary job placements, and early profession publishing success for economics PhDs. Am Econ Rev. 2007;97(2):422–426.
  12. 12.
    Blake-Beard S, Bayne ML, Crosby FJ, Muller CB. Matching by race and gender in mentoring relationships: Preserving our eyes on the prize. Aust J Soc Points. 2011;67(3):622–643.
  13. 13.
    Tenenbaum HR, Crosby FJ, Gliner MD. Mentoring relationships in graduate faculty. J Vocat Behav. 2001;59(3):326–341.
  14. 14.
    Ceci SJ, Ginther DK, Kahn S, Williams WM. Girls in educational science: A altering panorama. Psychol Sci Public Curiosity. 2014;15(3):75–141. pmid:26172066
  15. 15.
    Neumark D, Gardecki R. Girls serving to ladies? Position mannequin and mentoring results on feminine Ph.D. college students in economics. J Hum Resour. 1998;33(1):220–246.
  16. 16.
    Kalbfleisch PJ. Similarity and attraction in enterprise and educational enviornments: Similar and cross-sex mentoring relationships. Rev Bus. 2000;21(1/2):58.
  17. 17.
    Hale G, Regev T. Gender ratios at high PhD applications in economics. Econ Educ Rev. 2014;41:55–70.
  18. 18.
    Kaplan DH, Mapes JE. The place are the ladies? Accounting for discrepancies in feminine doctorates in US geography. Prof Geogr. 2016;68(3):427–435.
  19. 19.
    Begin D, McCauley S. Gender underlies the formation of STEM analysis teams. Ecol Evol. 2020;10(9):3834–3843. pmid:32489614
  20. 20.
    Sheltzer JM, Smith JC. Elite male school within the life sciences make use of fewer ladies. Proc Natl Acad Sci U S A. 2014;111(28):10107–10112. pmid:24982167
  21. 21.
    David SV, Hayden BY. Neurotree: A collaborative, graphical database of the educational family tree of neuroscience. PLoS One. 2012;7(10):e46608. pmid:23071595
  22. 22.
    Wais Okay. Gender prediction strategies based mostly on first names with genderizeR. The R Journal. 2016;8(1):17.
  23. 23.
    McPherson M, Smith-Lovin L, Cook dinner JM. Birds of a feather: Homophily in social networks. Annu Rev Sociol. 2001;27(1):415–444.
  24. 24.
    Bojanowski M, Corten R. Measuring segregation in social networks. Soc Networks. 2014;39:14–32.
  25. 25.
    Holman L, Morandin C. Researchers collaborate with same-gendered colleagues extra typically than anticipated throughout the life sciences. PLoS ONE. 2019;14(4):e0216128. pmid:31026265
  26. 26.
    Martinez ED, Botos J, Dohoney KM, Geiman TM, Kolla SS, Olivera A, et al. Falling off the educational bandwagon. EMBO Rep. 2007;8(11):977–981.
  27. 27.
    Ginther DK, Kahn S. Girls’s careers in educational social science: Progress, pitfalls, and plateaus. The Economics of Economists. Cambridge, UK: Cambridge College Press; 2014. p. 285–315.
  28. 28.
    Lerchenmueller MJ, Sorenson O. The gender hole in early profession transitions within the life sciences. Res Coverage. 2018;47(6):1007–1017.
  29. 29.
    Hirsch JE. An index to quantify a person’s scientific analysis output. Proc Natl Acad Sci U S A. 2005;102(46):16569–16572. pmid:16275915
  30. 30.
    Blume-Kohout ME, Clack JW. Are Graduate College students Rational? Proof from the Marketplace for Biomedical Scientists. PLoS ONE. 2013;8(12):e82759. pmid:24376573
  31. 31.
    Avin C, Keller B, Lotker Z, Mathieu C, Peleg D, Pignolet YA. Homophily and the glass ceiling impact in social networks. Proceedings of the 2015 Convention on Improvements in Theoretical Laptop Science. ACM; 2015. p. 41–50.
  32. 32.
    Merton RK. The Matthew impact in science. Science. 1968;159(3810):56–63.
  33. 33.
    Shaw AK, Stanton DE. Leaks within the pipeline: Separating demographic inertia from ongoing gender variations in academia. Philos Trans R Soc Lond B Biol Sci. 2012;279(1743):3736–3741. pmid:22719028
  34. 34.
    Valian V. Past gender schemas: bettering the development of girls in academia. Hypatia. 2005;20(3):198–212.
  35. 35.
    Ceci SJ, Williams WM. Understanding present causes of girls’s underrepresentation in science. Proc Natl Acad Sci U S A. 2011;108(8):3157–3162. pmid:21300892
  36. 36.
    Moss-Racusin CA, Dovidio JF, Brescoll VL, Graham MJ, Handelsman J. Science school’s delicate gender biases favor male college students. Proc Natl Acad Sci U S A. 2012;109(41):16474–16479. pmid:22988126
  37. 37.
    Williams WM, Ceci SJ. Nationwide hiring experiments reveal 2:1 school choice for ladies on STEM tenure monitor. Proc Natl Acad Sci U S A. 2015;112(17):5360–5365. pmid:25870272
  38. 38.
    Cech EA, Blair-Loy M. The altering profession trajectories of recent dad and mom in STEM. Proc Natl Acad Sci U S A. 2019;116(10):4182–4187. pmid:30782835
  39. 39.
    McDowell GS, Gunsalus KTW, MacKellar DC, Mazzilli SA, Pai VP, Goodwin PR, et al. Shaping the Way forward for Analysis: a perspective from junior scientists. F1000Res. 2015;3:291. pmid:25653845
  40. 40.
    Dworkin JD, Linn KA, Teich EG, Zurn P, Shinohara RT, Bassett DS. The extent and drivers of gender imbalance in neuroscience reference lists. Nat Neurosci. 2020;23(8):918–926. pmid:32561883
  41. 41.
    Wennerås C, Wold A. Nepotism and sexism in peer-review. Nature. 1997;387(6631):341–343. pmid:9163412
  42. 42.
    Jagsi R, Motomura AR, Griffith KA, Rangarajan S, Ubel PA. Intercourse variations in attainment of unbiased funding by profession improvement awardees. Ann Intern Med. 2009;151(11):804–811. pmid:19949146
  43. 43.
    Pohlhaus JR, Jiang H, Wagner RM, Schaffer WT, Pinn VW. Intercourse variations in utility, success, and funding charges for NIH extramural applications. Acad Med. 2011;86(6):759. pmid:21512358
  44. 44.
    Jagsi R, DeCastro R, Griffith KA, Rangarajan S, Churchill C, Stewart A, et al. Similarities and variations within the profession trajectories of female and male profession improvement award recipients. Acad Med. 2011;86(11):1415–1421. pmid:21952061
  45. 45.
    Kaatz A, Lee YG, Potvien A, Magua W, Filut A, Bhattacharya A, et al. Evaluation of Nationwide Institutes of Well being R01 utility critiques, impression, and standards scores: does the intercourse of the principal investigator make a distinction? Acad Med. 2016;91(8):1080–1088. pmid:27276003
  46. 46.
    Lengthy JS, Allison PD, McGinnis R. Rank development in educational careers: Intercourse variations and the results of productiveness. Am Sociol Rev. 1993;58(5):703–722.
  47. 47.
    Wolfinger NH, Mason MA, Goulden M. Issues within the pipeline: Gender, marriage, and fertility within the ivory tower. J Excessive Educ. 2008;79(4):388–405.
  48. 48.
    Rotbart HA, McMillen D, Taussig H, Daniels SR. Assessing gender fairness in a big educational division of pediatrics. Acad Med. 2012;87(1):98–104. pmid:22104061
  49. 49.
    Holliday EB, Jagsi R, Wilson LD, Choi M, Thomas CR Jr, Fuller CD. Gender variations in publication productiveness, educational place, profession length and funding amongst US educational radiation oncology school. Acad Med. 2014;89(5):767.
  50. 50.
    Nittrouer CL, Hebl MR, Ashburn-Nardo L, Trump-Steele RCE, Lane DM, Valian V. Gender disparities in colloquium audio system at high universities. Proc Natl Acad Sci U S A. 2018;115(1):104–108. pmid:29255050
  51. 51.
    Dutt Okay, Pfaff DL, Bernstein AF, Dillard JS, Block CJ. Gender variations in suggestion letters for postdoctoral fellowships in geoscience. Nat Geosci. 2016;9:805–808.
  52. 52.
    Madera JM, Hebl MR, Martin RC. Gender and letters of advice for academia: Agentic and communal variations. J Appl Psychol. 2009;94(6):1591–1599. pmid:19916666
  53. 53.
    Stack S. Gender, kids and analysis productiveness. Res Excessive Educ. 2004;45(8):891–920.
  54. 54.
    Morgan AC, Manner SF, Hoefer MJD, Larremore DB, Galesic M, Clauset A. The unequal impression of parenthood in academia. Sci Adv. 2021;7(9):eabd1996. pmid:33627417
  55. 55.
    Rivera LA. When two our bodies are (not) an issue: Gender and relationship standing discrimination in educational hiring. Am Sociol Rev. 2017;82(6):1111–1138.
  56. 56.
    Noy S, Ray R. Graduate college students’ perceptions of their advisors: Is there a scientific drawback in mentorship? J Excessive Educ. 2012;83(6):876–914.
  57. 57.
    Fried LP, Francomano CA, MacDonald SM, Wagner EM, Stokes EJ, Carbone KM, et al. Profession improvement for ladies in educational medication: A number of interventions in a division of drugs. JAMA. 1996;276(11):898–905. pmid:8782639
  58. 58.
    Seagram BC, Gould J, Pyke SW. An investigation of gender and different variables on time to completion of doctoral levels. Res Excessive Educ. 1998;39(3):319–335.
  59. 59.
    Curtin N, Malley J, Stewart AJ. Mentoring the subsequent technology of school: Supporting educational profession aspirations amongst doctoral college students. Res Excessive Educ. 2016;57(6):714–738.
  60. 60.
    El-Alayli A, Hansen-Brown AA, Ceynar M. Dancing backwards in excessive heels: feminine professors expertise extra work calls for and particular favor requests, notably from academically entitled college students. Intercourse Roles. 2018;79(3–4):136–150.
  61. 61.
    Guarino CM, Borden VMH. School service hundreds and gender: are ladies taking good care of the educational household? Res Excessive Educ. 2017;58(6):672–694.
  62. 62.
    AlShebli BK, Rahwan T, Woon WL. The preeminence of ethnic range in scientific collaboration. Nat Commun. 2018;9(1):5163. pmid:30514841
  63. 63.
    King MM, Bergstrom CT, Correll SJ, Jacquet J, West JD. Males set their very own cites excessive: Gender and self-citation throughout fields and over time. Soc Forces. 2017;3:1–22.
  64. 64.
    Lerchenmueller MJ, Sorenson O, Jena AB. Gender variations in how scientists current the significance of their analysis: observational examine. BMJ. 2019;367:l6573. pmid:31843745
  65. 65.
    Yang Y, Chawla NV, Uzzi B. A community’s gender composition and communication sample predict ladies’s management success. Proc Natl Acad Sci U S A. 2019;116(6):2033–2038. pmid:30670641
  66. 66.
    Yu H, Willis KA, Litovitz A, Harriman RM, Davis MT, Meyer P, et al. The impact of mentee and mentor gender on scientific productiveness of candidates for NIH coaching fellowships. bioRxiv. 2021:p. 2021.02.02.429450.
  67. 67.
    Ginther DK, Kahn S, Schaffer WT. Gender, race/ethnicity, and Nationwide Institutes of Well being R01 analysis awards: is there proof of a double bind for ladies of coloration? Acad Med. 2016;91(8):1098. pmid:27306969
  68. 68.
    West JD, Jacquet J, King MM, Correll SJ, Bergstrom CT. The position of gender in scholarly authorship. PLoS ONE. 2013;8(7):1–6. pmid:23894278
  69. 69.
    Cheryan S, Ziegler SA, Montoya AK, Jiang L. Why are some STEM fields extra gender balanced than others? Psychol Bull. 2017;143(1):1–35. pmid:27732018
  70. 70.
    Cheryan S, Plaut VC, Davies PG, Steele CM. Ambient belonging: How stereotypical cues impression gender participation in pc science. J Pers Soc Psychol. 2009;97(6):1045–1060. pmid:19968418
  71. 71.
    Acuna DE. Some concerns for finding out gender, mentorship, and scientific impression: commentary on AlShebli, Makovi, and Rahwan (2020). OSF Preprints. 2020;10.31219.
  72. 72.
    Serano J. Homogenizing versus holistic views of gender and sexuality. Excluded: making feminist and queer actions extra inclusive. Berkeley, CA: Seal Press; 2013. p. 138–168.
  73. 73.
    Roughgarden J. Evolution’s rainbow: range, gender, and sexuality in nature and folks. Berkeley, CA: College of California Press; 2004.
  74. 74.
    Barres BA. Does gender matter? Nature. 2006;442(7099):133–136. pmid:16848004
  75. 75.
    Goldberg AE, Kuvalanka Okay, Dickey L. Transgender graduate college students’ experiences in increased schooling: A mixed-methods exploratory examine. J Divers Excessive Educ. 2019;12(1):38–51.
  76. 76.
    Atherton TJ, Barthelemy RS, Deconinck W, Falk ML, Garmon S, Lengthy E, et al. LGBT local weather in physics: constructing an inclusive group. Faculty Park, MD: American Bodily Society; 2016.
  77. 77.
    Sugimoto CR. Tutorial family tree. Past Bibliometrics: Harnessing Multidimensional Indicators of Scholarly Affect. Cambridge, MA: MIT Press; 2014. p. 365–382.
  78. 78.
    Hunter JD. Matplotlib: a 2D graphics surroundings. Comput Sci Eng. 2007;9(3):90–95.
  79. 79.
    McKinney W. Knowledge buildings for statistical computing in Python. Python in Science Convention. Austin, Texas; 2010. p. 56–61. Obtainable from:
  80. 80.
    SciPy 1 0 Contributors, Virtanen P, Gommers R, Oliphant TE, Haberland M, Reddy T, et al. SciPy 1.0: elementary algorithms for scientific computing in Python. Nat Strategies. 2020;17(3):261–272. pmid:32015543
  81. 81.
    Seabold S, Perktold J. Statsmodels: Econometric and statistical modeling with Python. In: Python in Science Convention. Austin, Texas; 2010. p. 92–96. Obtainable from:
  82. 82.
    Waskom M, Gelbart M, Botvinnik O, Ostblom J, Hobson P, Lukauskas S, et al. mwaskom/seaborn: v0.11.1 (December 2020). 2020. Obtainable from:
  83. 83.
    Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-Study: Machine studying in Python. J Mach Study Res. 2011;12(85):2825–2830.
  84. 84.
    Larivière V, Ni C, Gingras Y, Cronin B, Sugimoto CR. Bibliometrics: World gender disparities in science. Nature. 2013;504(7479):211–213. pmid:24350369
  85. 85.
    Holman L, Stuart-Fox D, Hauser CE. The gender hole in science: How lengthy till ladies are equally represented? PLoS Biol. 2018;16(4):e2004956. pmid:29672508
  86. 86.
    Topaz CM, Sen S. Gender illustration on journal editorial boards within the mathematical sciences. PLoS ONE. 2016;11(8):e0161357. pmid:27536970
  87. 87.
    Bachmann M. 2020. Obtainable from:


Please enter your comment!
Please enter your name here

Most Popular

Recent Comments