Mentoring practices that predict doctoral student outcomes in a biological sciences cohort

Abstract

Despite the importance of a diversity of backgrounds and perspectives in biological research, women, racial and ethnic minorities, and students from non-traditional academic backgrounds remain underrepresented in the composition of university faculty. Through a study on doctoral students at a research-intensive university, we pinpoint advising from faculty as a critical component of graduate student experiences and productivity. Graduate students from minority backgrounds reported lower levels of support from their advisors and research groups. However, working with an advisor from a similar demographic background substantially improved productivity and well-being of these students. Several other aspects of mentoring practices positively predicted student success and belonging, including frequent one-on-one meetings, empathetic and constructive feedback, and relationships with other peer or faculty mentors. Our study highlights the need to renovate graduate education with a focus on retention-not just recruitment-to best prepare students for success in scientific careers.

Fig 1. Composition of survey sample compared with enrollment data.

Responses of participants to demographic questions about their (a) departmental affiliation, (b) gender identity, and (c) racial identity. Within each panel, the composition of our sample (n = 129) is on the left, and enrollment data from the UC Berkeley Office of Planning and Analysis (n = 538) are available for comparison on the right. Color is used to indicate similar categories, but because the two surveys used different phrasing, not all categories have a direct analogue.

Fig 2. Differences in responses among demographic groups.

(a) Top five questions with the most disparate responses between male and female/non-binary students. (b) Top five questions with the most disparate responses between students who started their graduate degrees before or after the age of 30. Points and error bars represent the mean and standard error, respectively, of Likert responses converted to a 1–5 numerical scale. Statistical significance for each question was measured based on the distribution of differences between members of different groups, which was compared to zero (the null hypothesis, that no differences exist) using a one-sample t-test with degrees of freedom equal to the number of graduate students in the minority category minus one. Benjamini-Hochberg correction was applied to the resulting p-values to correct for multiple hypothesis testing. Asterisks indicate corrected p-values: (***) p≤0.001; (**) 0.001<p≤ 0.01; (*) 0.01<p≤0.05. There were no questions for which female/non-binary students reported significantly better outcomes than male students or for which students who started graduate school after age 30 reported better outcomes than students who started before age 30. (c-e) Responses to select questions among respondents who identified as multiple underrepresented identities (female or non-binary, started program after age 30, demographic mismatch with advisor, non-traditional academic background, other identity as specified in comments).

Fig 3. Frequency and scheduling of individual advising meetings.

(a) Graduate students with as-needed meetings saw their advisors less frequently than students with drop-in or regular meetings (cumulative link mixed mode, p<0.001). (b) Graduate students in the largest and smallest labs saw the least of their advisors (nonlinear regression, p = 0.007). (c-e) Graduate student outcomes according to meeting scheduling format and gender (cumulative link mixed model, p<0.05 for gender by meeting schedule interaction in all cases). Points and error bars represent the mean and standard error, respectively, of Likert responses converted to a 1–5 numerical scale.

Fig 4. Demographic representation in advisors.

(a) Proportion of students who identified as an underrepresented minority in any respect, and of those, proportion who worked with advisors of a similar demographic. (b) Effect of advisor representation was strongest at the early stages of the PhD. (c-e) Graduate student outcomes among students who worked with advisors of a similar demographic, students whose advisors were highly empathetic (answered “Agree” or “Strongly Agree” to the statement “My advisor is empathetic to my concerns and needs”), and students whose advisors were neither. Points and error bars represent the mean and standard error, respectively, of Likert responses converted to a 1–5 numerical scale. Asterisks indicate p-values: (***) p≤0.001; (**) 0.001<p≤0.01; (*) 0.01<p≤0.05.

Fig 5. Informal mentorship and community.

(a) Support from other students appears to compensate for lack of advisor support (cumulative link mixed model, main effect of advisor support p<0.001, main effect of student support p<0.001, interaction effect p = 0.008). (b) Research advice from other faculty appears to compensate for infrequent meetings with advisor (cumulative link mixed model, main effect of advisor meetings p<0.001, main effect of other faculty p = 0.048, interaction effect p = 0.078). (c) A collaborative departmental culture appears to compensate for less guidance from advisors (cumulative link mixed model, main effect of advisor guidance p<0.001, main effect of collaborative culture p = 0.019, interaction effect p = 0.048). Points and error bars represent the mean and standard error, respectively, of Likert responses converted to a 1–5 numerical scale. Asterisks indicate p-values: (***) p≤0.001; (**) 0.001<p≤0.01; (*) 0.01<p≤0.05.