Bias against parents in science hits women harder

Gender bias is predominant in science, especially in the STEM fields, where women's presence decreases sharply during the academic career. This phenomenon is known as the scissors effect or leaky pipeline and many studies showed that women mostly leave academia after graduate school at the post-doc level. Factors leading women to abandon academia are multiple, including a gendered workplace, lower funding, harassment, implicit bias, and, likely the most important, motherhood. Bias—defined as prejudicial judgments that violate impartiality—can have cumulative effects on women's scientific careers. Motherhood often triggers negative assumptions of competence and commitment, producing a motherhood penalty, while fathers may face no penalties or even advantages. In academia, flexibility stigma is reported by faculty with young children and is associated with intentions to leave. Despite this, studies on bias against parents in academia, particularly self-perceived bias, are scarce. Academia’s distinct culture and high workloads may create unique challenges for caregivers. The objective of this study was to evaluate self-reported negative bias that scientists with children suffer in their work environment and to identify factors influencing this perception. Using an online survey of Brazilian faculty with children, we assessed bias perceptions across variables such as gender, race, career stage/status, field, and number of children. We hypothesized that female faculty with children perceive greater negative bias than male peers and that race, career stage, and research area aggravate bias perception, with Black, early-career, and STEM faculty being most affected. By providing quantitative data, we aim to better understand the prevalence of bias against parents in academia.

Prior research documents persistent gender bias in academia, including implicit and explicit biases affecting hiring, mentoring, salaries, recommendation letters, and evaluations, with cumulative effects on promotion, tenure, publication, funding success, and respect. Experimental studies show equivalent applications receive more favorable evaluations when associated with men; race and gender can combine to disadvantage Black and Latin American women. A robust literature identifies a motherhood penalty in labor markets—mothers are judged less competent/committed and offered lower salaries—while fathers may receive a premium. In academia, flexibility stigma is associated with intentions to exit, and negative bias toward those raising families has been reported. Some studies suggest STEM’s masculine culture intensifies pressures on mothers, though gender bias can persist even in gender-balanced or female-dominated fields. Evidence on intersections with race indicates complex dynamics, but motherhood penalties may occur across races. Collectively, past work motivates investigating self-perceived parenthood-related bias within academia and its moderators.

Ethics approval was obtained from the Federal University of Rio Grande do Sul Ethics Committee (IRB 51551321.0.0000.5347). Survey instrument: An online questionnaire (Portuguese) assessed perceptions of bias against parents in academia. It included 22 demographic/academic/parenthood items (gender, age, race, education, position, years of experience, graduate supervision, number/ages of children, disability status of children) and seven 5-point Likert items (1 = completely disagree to 5 = completely agree) probing experiences surrounding parental leave reception, treatment after return, workload pressure, perceived commitment/competence, performance reviews, access to opportunities, and need to continually prove competence. Several items (1, 2, 4, 5, 6) were reverse-coded; summed total scores ranged 5–35, representing self-perceived negative parental bias. Sampling and data collection: The survey link was disseminated via social media, emails to Brazilian universities/research centers, and snowball sampling. It was open Oct 8–Nov 27, 2021, took ~5 minutes to complete, and was answered by 995 scientists; 105 were excluded for not having had at least one child after being hired as faculty, yielding n=890 faculty parents. Analysis: Descriptive statistics summarized demographics and parenthood profiles. Group comparisons on individual Likert items by gender used Mann–Whitney tests (ordinal data). Internal consistency of the 7-item bias scale was good (standardized Cronbach’s alpha = 0.81). Bivariate tests (Mann–Whitney or Kruskal–Wallis) examined associations of total score with candidate predictors: gender (men/women), race (Black/White; after excluding small categories: final n=861), graduate supervisor (yes/no), Productivity Scholarship (PS) holder (yes/no), number of children (one vs more than one), research field (8 areas), and hiring time (<15 vs ≥15 years). Variables with p ≤ 0.20 were entered into linear regression models predicting total score; model assumptions (residual normality, linearity, homoscedasticity, multicollinearity) were checked and met. Three regression models were fitted: (1) main effects only (gender, graduate supervisor, hiring time), (2) main effects plus gender × graduate supervisor, and (3) main effects plus gender × hiring time. Analyses were conducted in RStudio (R 2022) with Tidyverse, psych, car, and moments; significance threshold p ≤ 0.05. Analysis scripts and the full survey are provided in supplementary material.

Sample: n=890 faculty parents; 69.3% women. Group comparisons (individual items by gender; all p-values below refer to Mann–Whitney tests): • All seven items showed significant gender differences (p ≤ 0.035). • Mothers reported more negative experiences around parental leave and return. • Perceived commitment/competence: 63% of fathers vs 35% of mothers completely agreed that having children did not change colleagues’/superiors’ perceptions. • Fair performance evaluations: 74.4% of fathers vs 52.4% of mothers completely agreed. • Access to opportunities: 71.8% of fathers vs 42.8% of mothers completely agreed they had as much access as peers. • Constantly proving competence: 50.5% of fathers completely disagreed vs 28.8% of mothers. Bivariate analyses (total score, higher = more perceived negative bias): • Gender: women mean 15.87 (SD 6.6) vs men mean 11.96 (SD 4.83); W=49540; p<0.001. • Race (Black vs White): no significant difference; W=51033; p=0.091; means 15.41 (SD 6.4) vs 14.56 (SD 6.37). • Hiring time: <15 years mean 15.26 (SD 6.39) vs ≥15 years mean 13.52 (SD 6.21); W=65900; p<0.001. • Graduate supervisor: supervisors mean 14.23 (SD 6.2) vs non-supervisors mean 16.45 (SD 6.73); W=75985; p<0.001. • PS holder: holders mean 12.87 (SD 6.13) vs non-holders mean 15.17 (SD 6.37); W=72490; p<0.001 (bivariate). • Research field: no significant differences (Chi-sq=7.05, df=7, p=0.423); means similar across Agricultural (15.19), Biological (14.06), Health (14.78), Exact/Earth (14.73), Humanities (15.18), Social (15.41), Engineering (13.29), Linguistics/Language/Arts (14.85). • Number of children: one child mean 14.92 (SD 6.23) vs more than one mean 14.56 (SD 6.5); W=95531; p=0.234 (ns). Regression (dependent variable: total score): • Overall model significant: adjusted R^2=0.09; F(5,855)=19.08; p<0.001. • Significant predictors: gender (β=3.53, p<0.001; women higher than men), graduate supervisor (β=−1.26, p=0.017; supervisors lower), hiring time (β=0.983, p=0.036; <15 years higher than ≥15). • Non-significant predictors: race (β=0.303, p=0.576), PS holder (β=−0.908, p=0.107). Interactions: • Model including gender × hiring time improved fit (F=6.121, p<0.05). Pairwise comparisons (Bonferroni-corrected) showed women with <15 years score 4.46 points higher than men with <15 years (p<0.001). Women with ≥15 years still score higher than men with <15 years (mean difference 2.50, p=0.001). Hiring time was not significant among fathers, indicating a specific effect on mothers. Overall: Mothers perceive greater negative bias than fathers across items and total score; effects are not explained by field, number of children, or race in this sample; early-career status and lower career consolidation (non-supervisor, non-PS holder in bivariate) are associated with higher perceived bias; strongest predictor is gender.

Findings indicate a strong, gendered pattern of perceived negative bias against faculty parents in Brazil, disproportionately affecting mothers. Mothers reported more negative experiences around leave, treatment upon return, workload pressure, access to opportunities, fair evaluations, and the need to continually prove competence, while fathers reported more favorable assessments. Considering the composite bias score, gender had the largest effect, consistent with the concept of micro-aggressions undermining women’s performance and persistence in academia. Although the study measures perception rather than objective bias, extensive external evidence of maternal bias suggests mothers’ perceptions may accurately reflect workplace realities. Asymmetric parental leave policies (longer maternity vs short paternity leave) may exacerbate biases against mothers; contexts with more equitable parental leave might reduce the motherhood penalty. No differences by research field suggest the phenomenon is widespread across disciplines, including those with higher female representation. Career status moderated perceptions: graduate supervisors reported lower perceived bias, possibly due to greater career consolidation and team support during/after leave. Early-career parents, especially mothers with less than 15 years since hiring, reported higher perceived bias; experience may attenuate but does not eliminate bias, as mothers with ≥15 years still perceived more bias than early-career fathers. Race did not significantly predict perceptions in models, though limited sample sizes for some groups may mask differences; literature suggests motherhood penalties occur across races, warranting deeper intersectional inquiry. The results underscore the need for institutional and cultural interventions to counteract parenthood-related gender bias, including awareness-building, evidence-based bias mitigation, supportive policies, and workload adjustments, particularly for early-career parents.

This study quantitatively documents that parenthood-related bias in academia is perceived more strongly by mothers than fathers. Gender, early-career status (<15 years since hiring), and non-supervisory status are associated with higher perceived bias, whereas race, field, and number of children are not significant predictors in this sample. Interactions reveal that early-career mothers are particularly affected. To promote equity, institutions should pursue structural and cultural changes: normalize and support parenthood for all genders, design equitable parental leave, reduce overwork cultures, provide workload and timeline accommodations (especially early-career), enable flexible arrangements without stigma, build awareness and accountability around implicit bias, and establish transparent data collection with monitoring of intervention impacts. Future research should: (1) include academics without children as a comparison group; (2) use longitudinal and experimental designs to distinguish perception from objective bias; (3) examine intersectional effects with larger, more diverse samples (e.g., race/ethnicity); and (4) evaluate the efficacy of specific institutional policies and trainings in reducing parenthood-related bias.

• The study measures self-perception of bias, not objective bias or causal effects. • No comparison group of academics without children, limiting attribution specifically to parenthood versus gender alone. • Potential sampling biases due to online, voluntary, snowball recruitment. • Limited sample sizes for some racial categories (e.g., Indigenous, Asian) reduced analytic power; race was collapsed to Black vs White and some respondents excluded, possibly obscuring intersectional effects. • The cross-sectional design precludes causal inference and may be influenced by contemporaneous factors (e.g., COVID-19 pandemic) affecting caregiving burdens and perceptions. • Productivity Scholarship holder effects were significant in bivariate but not in multivariable models, potentially due to sample size or residual confounding. • All data are from Brazil; generalizability to other countries may be limited.