(How) does majoring in the social sciences and humanities relate to specialty choice?: a cross-sectional study Authors: Laura E. Hirshfield, Tirza Ochrach-Konradi, Alan Schwartz Published: 2026 ABSTRACT BackgroundScholars have previously demonstrated a link between backgrounds in the social sciences and humanities and humanistic qualities, yet the literature exploring the relationship between undergraduate college major and specialty choice is equivocal, in part because previous studies have rarely used nationally representative datasets. Further, there are many potential mediating factors that may influence this relationship.MethodsThis cross-sectional study aimed to investigate whether undergraduate college major relates to future specialty and explored potential mediating factors in this relationship. In particular, the authors investigated the relationship between social science and humanities college major and choice of either primary care or non-procedural specialties. A regression modeling approach was used with a U.S.-based national dataset that included 49,214 individuals who participated in the 2018–2020 Association of American Medical Colleges (AAMC) Graduation Questionnaires (GQ). Study participants represent about 82% of the 59,887 graduates from U.S. allopathic medical schools between 2018 and 2020.ResultsOverall, medical students who majored in the humanities and social sciences were more likely than their science, technology, engineering, and mathematics (STEM) major peers to choose primary care specialties and to choose non-procedural specialties. These relationships were consistent even when considering other social identity and performance factors. Additionally, medical students who took “patient-focused” electives were more likely than those who did not to choose primary care and non-procedural specialties.ConclusionsA promising approach to increase the number of future primary care physicians in the US may be to recruit more social science and humanities majors.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12909-026-09464-5. Introduction When choosing who to select as students for their future classes, medical school administrators and admissions committees within the United States must grapple with two key questions: what characteristics make successful medical students and what characteristics make successful doctors? For a variety of reasons, admissions committees frequently concentrate on the former question, focusing on the demonstration of natural science aptitude and background because of an assumption that students with these backgrounds and skills will be most successful in medical school environments [1, 2]. In contrast, students with backgrounds in the humanities and social sciences are often seen as “riskier bets”, despite little evidence that they struggle academically [3]. However, there may be other, non-academic reasons for increasing their representation in medical schools and in the profession; specifically, in relation to their potential to become doctors whose career intentions align with national healthcare needs. Indeed, scholars have demonstrated a link between experience or backgrounds in the social sciences and humanities and humanistic qualities [4–6]. Students with humanities backgrounds have been observed to be better communicators, have stronger interpersonal skills, and greater empathy than their peers [4, 5]. Scholars have also argued that exposure to humanities coursework in the premedical period of training may lead students to be more capable of understanding and accepting the validity of more than one form of knowledge production, which in turn may contribute to students’ ability to grasp the intellectual component of empathy [7]. This ability to integrate and make sense of diverse ways of knowing is essential for success in medicine and this skill may, at least in part, be facilitated by prior training in the social sciences and humanities [8]. In fact, in terms of empathy development, medical students have highlighted the important role of premedical coursework that asks them to think about their future patients in terms of the diverse, broader contexts that they may be situated in, including subjects like sociology, psychology, philosophy, and ethics [6]. These skills may be particularly essential at a time when numerous concerns have been raised about the increasing shortfall of primary care physicians both in the United States [9–12] and globally [13]. Scholars also caution that current medical training may underemphasize and underdevelop our trainees’ humanism [14–16]. In light of the growing demand for primary care physicians in the workforce, scholars have sought to identify influences that guide students to primary care specialties. However, many findings in this area remain equivocal because studies on specialty choice often focus primarily on singular specialties and have rarely used nationally representative datasets [3]. Some scholars have found no relationship between undergraduate major and residency choice [4, 17, 18], while others found that medical students with humanities or social science backgrounds are more likely to choose specialties that emphasize patient-physician interaction [19, 20]. Some researchers have also reported that higher empathy is associated with a preference for “patient-oriented” specialties, while others find no relationship [21]. Studies further show that interest in patient service contributes to medical students’ specialty selection, and higher empathy levels may be associated with preference for patient focused specialties [21–23]. Extracurricular activities such as interest groups, clinical experiences, and working with physician mentors have also been shown to increase students’ interests in direct patient care [24–26]. Finally, there have been mixed findings on the influence of student debt; however, studies suggest that higher debt may correlate with selection of higher paying specialties [27]. In this study, we used a cross-sectional, regression modeling approach with a national dataset to investigate whether undergraduate college major relates to future choice of specialty, as well as explore potential mediating factors in this relationship. Based on previous findings that highlight the relationship between social science and humanities backgrounds and empathy [5], as well as the role of higher empathy and specialty choice [21–23], our team developed the following hypotheses: Hypothesis 1: Medical students with premedical majors in the humanities and social sciences will be more likely than their peers in the natural sciences to:Hypothesis 1a: choose primary care specialties, which emphasize and value patient-physician interaction, andHypothesis 1b: choose non-procedural specialties, which are more likely to involve direct patient-physician interactionHypothesis 2: Medical students with premedical majors in the humanities and social sciences will be more likely than their peers in the natural sciences to participate in medical school activities or electives that emphasize patient experience and contact, which, in turn, will explain some of the relationship between college major and specialty choice. Hypothesis 1: Medical students with premedical majors in the humanities and social sciences will be more likely than their peers in the natural sciences to:Hypothesis 1a: choose primary care specialties, which emphasize and value patient-physician interaction, andHypothesis 1b: choose non-procedural specialties, which are more likely to involve direct patient-physician interaction Hypothesis 1: Medical students with premedical majors in the humanities and social sciences will be more likely than their peers in the natural sciences to:Hypothesis 1a: choose primary care specialties, which emphasize and value patient-physician interaction, andHypothesis 1b: choose non-procedural specialties, which are more likely to involve direct patient-physician interaction Hypothesis 1a: choose primary care specialties, which emphasize and value patient-physician interaction, and Hypothesis 1a: choose primary care specialties, which emphasize and value patient-physician interaction, and Hypothesis 1b: choose non-procedural specialties, which are more likely to involve direct patient-physician interaction Hypothesis 1b: choose non-procedural specialties, which are more likely to involve direct patient-physician interaction Hypothesis 2: Medical students with premedical majors in the humanities and social sciences will be more likely than their peers in the natural sciences to participate in medical school activities or electives that emphasize patient experience and contact, which, in turn, will explain some of the relationship between college major and specialty choice. Hypothesis 2: Medical students with premedical majors in the humanities and social sciences will be more likely than their peers in the natural sciences to participate in medical school activities or electives that emphasize patient experience and contact, which, in turn, will explain some of the relationship between college major and specialty choice. Methods Data source We obtained data through a licensing agreement with the Association of American Medical Colleges (AAMC) after submitting a data request through the AAMC.org website; the AAMC also helped us to obtain data from the National Board of Medical Examiners (NMBE) through a separate data request process, then coordinated the linkage of data sets. Our final data set included 49,214 individuals who participated in the 2018–2020 Graduation Questionnaires (GQ), graduated in the 2017–2018, 2018–2019, or 2019–2020 academic years, and indicated a specialty choice. These individuals represent about 82% of the 59,887 graduates from U.S. allopathic medical schools in those academic years. For each participant, variables were provided from their medical school application (i.e., information regarding their demographics and pre-medical school assessments) and from their Graduation Questionnaire (GQ) (i.e., information about their experiences and accomplishments in medical school, as well as about their decision-making processes regarding their future specialties). In addition to data on the specialty chosen by each student after medical school, we obtained questionnaire data about students’ interest in primary care at the time of matriculation to medical school from the AAMC Matriculating Student Questionnaire (MSQ) and at the end of year two of medical school using AAMC’s Year Two Questionnaire (Y2Q). Our data set did not include the ultimate specialties that our participants matched into; however, specialty preference and ultimate match rates are highly correlated, especially for primary care specialties. We chose to use data from individuals that had graduated from medical school prior to 2020 to limit potential effects from two substantial shifts in the medical education landscape: (1) the COVID-19 pandemic, and (2) the change in the scoring system for the USMLE Step 1 examination. In the US, the USMLE Step 1 and 2CK exams are taken during the last two years of medical school. Though these licensing exams were not developed as a source of information for program directors to rank candidates for residency interviews, historically, scores from both exams have been used to filter applicants for a limited number of positions. Training time and pathways also vary considerably across the various specialties; specifically, residency can be anywhere from 3 years to 7 years in length, and some future physicians may choose to add to their training by completing an additional fellowship, or post-residency training. The data set includes most recent Medical College Admission Test (MCAT) score, first-time United States Medical Licensing Examination (USMLE) Step 2CK scores, and first-time USMLE Step 1 pass/fail data. Although participants in this study chose their specialty as one of the last groups of medical students with a scored Step 1 exam, we were not given access to these scores due to National Board of Medical Examiners (NBME) policies. This affected our abilities to control adequately for the role of Step 1 score in contributing to specialty selection. However, there is good evidence that MCAT and Step 2CK scores (both of which were included in our models, along with Step 1 pass/fail data) are correlated with Step 1 score [28, 29]. These scores were included to control for the important role of these high-stakes exams in students’ specialty selection [30]. Data included sex (male, female, and no response, which we coded as non-male) and self-identified underrepresented minority (URiM) status (Black, Hispanic, and Native participants vs. all others), but not specific race or ethnicity. Specifically, URiM status included any U.S. citizen or permanent resident who self-identified as one or more of the following race/ethnicity categories: American Indian or Alaska Native; Black or African American; Hispanic or Latino; or Native Hawaiian or Pacific Islander. The data sets generated and analyzed during the current study are not publicly available due to restrictions by the AAMC and NBME; however details about the variables included in the data set are available from the corresponding author on reasonable request. Categorization of college majors, specialties, and medical school electives Using a team-based consensus process, we categorized college majors in the data into four categories: STEM (science, technology, engineering, and mathematics), social sciences, humanities, and miscellaneous (Supplemental Appendix 1 shows the categorization of each major in the data set.) We also categorized each practice specialty as either primary care or not, and as either procedural or not. (Supplemental Appendix 2 shows the categorization of each specialty in the data set.) We excluded respondents who did not plan to practice or listed specialty as other or undecided. Notably, there is no standard definition of “primary care” specialties. For example, while some scholars include obstetrics/gynecology and psychiatry in their definitions, others see these physicians as specialists. In this article, we have taken a fairly expansive definition of these specialties, following the example of several previous studies on the topic and using a broad view of the construct related to their foundational, preventive, and integrated practice [31, 32]. Finally, we categorized each student-reported medical school elective as either patient-focused or research-focused. We considered electives in medical or patient education, health communications, home care, foreign language, interpreting services, health disparities, cultural awareness, nursing, or free clinic work to be patient-focused. We considered independent studies, research experiences, authorship, and research presentations to be research-focused. We excluded “other” electives. Supplemental Appendix 3 shows the categorization of electives. We computed, for each medical student, the proportion of their total electives that were patient-focused. Analyses To test hypotheses 1a and 1b, we first performed simple chi-squared tests on the college major category x primary care and college major category x nonprocedural specialty contingency tables using R Statistical Software (v4.4; R Core Team 2021). We then fitted a set of three nested logistic regression models to the primary care and nonprocedural specialty outcomes. In the first model for each outcome, we included only college major category as predictors. In the second model, we added gender (nonmale vs. male), URiM status, undergraduate grade point average (GPA) in science courses, undergraduate GPA in non-science courses, and MCAT score as additional predictors. In the third model, we further added USMLE Step 1 outcome (pass vs. fail), USMLE Step 2 CK score, and whether the student reported on their Graduation Questionnaire that debt, interest in fellowship, salary, and/or training time required were influential in their specialty choice (any influence vs. no influence for each factor). To test hypothesis 2, we conducted a causal mediation analysis using the proportion of patient-focused electives as the mediator of the relationship between college major and choice of primary care specialty. We fitted a linear regression predicting proportion of patient-focused electives from college major category, gender (nonmale vs. male), URiM status (Black, Hispanic, and Native participants vs. all others), undergraduate GPA in science courses, and undergraduate GPA in non-science courses. We fitted a logistic regression model to choice of primary care specialty using the proportion of patient-focused electives and college major category, gender (nonmale vs. male), URiM status (Black, Hispanic, and Native participants vs. all others), undergraduate GPA in science courses, and undergraduate GPA in non-science courses as predictors. Using the models, we computed the average direct effect and average causal mediation effect of major category on choice of primary care specialty [33]. This study involved the secondary analysis of existing data that was gathered for another purpose; thus, it did not qualify as human subjects research according to United States research ethics guidelines according to 45 CFR 46 (“the Common Rule”). Results Preliminary analyses Sample characteristics are presented in Table 1. The descriptive statistics in Fig. 1 illustrate the effect of college major on specialty choice. There was a significant association between college major and choice of both primary care specialties (χ 2 (3) = 120.4, p < 0.001) and non-procedural specialties (χ2(3) = 68.5, p < 0.001). Table 1Participant characteristics (including demographics, premedical, and medical school factors)GenderOverall (N = 49,214)Male24,417 (49.8%)Non-male24,602 (50.2%)Missing195URiM Not URiM37,115 (82.8%) URiM7,719 (17.2%) Missing4,380College major category STEM32,500 (68.1%) Social Science4,991 (10.5%) Humanities2,756 (5.8%) Other7,446 (15.6%) Missing1,521College GPA (science courses) Mean (SD)3.65 (0.30) Missing1,469College GPA (other courses) Mean (SD)3.78 (0.22) Missing1,492MCAT score (most recent attempt) Mean (SD)512.02 (6.53) Missing1,605USMLE Step 1 result (first attempt) Fail1,231 (2.5%) Pass47,933 (97.5%) Missing50USMLE Step 2CK score (first attempt) Mean (SD)245.35 (15.44) Missing41Specialty influence – Debt No25,375 (54.9%) Yes*20,829 (45.1%) Missing3,010Specialty influence – Fellowship No9,046 (19.6%) Yes*37,129 (80.4%) Missing3,039Specialty influence – Salary No10,144 (22.0%) Yes*36,070 (78.0%) Missing3,000Specialty influence – Training length No11,508 (24.9%) Yes*34,713 (75.1%) Missing2,993Proportion of people-oriented electives in med school Mean (SD)0.66 (0.18) Missing2,043* “yes” denotes any influence; please see manuscript for details Table 1Participant characteristics (including demographics, premedical, and medical school factors)GenderOverall (N = 49,214)Male24,417 (49.8%)Non-male24,602 (50.2%)Missing195URiM Not URiM37,115 (82.8%) URiM7,719 (17.2%) Missing4,380College major category STEM32,500 (68.1%) Social Science4,991 (10.5%) Humanities2,756 (5.8%) Other7,446 (15.6%) Missing1,521College GPA (science courses) Mean (SD)3.65 (0.30) Missing1,469College GPA (other courses) Mean (SD)3.78 (0.22) Missing1,492MCAT score (most recent attempt) Mean (SD)512.02 (6.53) Missing1,605USMLE Step 1 result (first attempt) Fail1,231 (2.5%) Pass47,933 (97.5%) Missing50USMLE Step 2CK score (first attempt) Mean (SD)245.35 (15.44) Missing41Specialty influence – Debt No25,375 (54.9%) Yes*20,829 (45.1%) Missing3,010Specialty influence – Fellowship No9,046 (19.6%) Yes*37,129 (80.4%) Missing3,039Specialty influence – Salary No10,144 (22.0%) Yes*36,070 (78.0%) Missing3,000Specialty influence – Training length No11,508 (24.9%) Yes*34,713 (75.1%) Missing2,993Proportion of people-oriented electives in med school Mean (SD)0.66 (0.18) Missing2,043* “yes” denotes any influence; please see manuscript for details Participant characteristics (including demographics, premedical, and medical school factors) * “yes” denotes any influence; please see manuscript for details Fig. 1Percentage of medical students graduating in years 2018–2020 by college major category who entered non-procedural (left panel) or primary care (right panel) specialties. Error bars represent 95% confidence intervals Percentage of medical students graduating in years 2018–2020 by college major category who entered non-procedural (left panel) or primary care (right panel) specialties. Error bars represent 95% confidence intervals Hypothesis 1 Table 2 presents the results of the generalized linear model regression analysis testing the relationship between college major and primary care specialty choice. Overall, as shown in Model 1 (n = 41,223), Social Science majors were 1.32 times (CI 1.24–1.41, p < 0.001) more likely and Humanities majors were 1.39 times (CI 1.28–1.52, p < 0.001) more likely than STEM graduates to go into primary care. Model 2 introduced several demographic (gender and URiM status) and pre-matriculation (MCAT, science GPA, and other GPA) variables; the impact of college major remained significant. Model 3 introduced a number of medical school variables, including Step 1 pass/fail, Step 2 CK score, and factors related to specialty choice. The relationship between college major and specialty choice remained significant after adjusting for these factors. Specifically, although URiM students were more likely to choose primary care specialties in univariate analyses (OR = 1.16, 95% CI 1.10–1.22, p < 0.001, n = 38,780), URiM was no longer a significant predictor when adjusting for college major. On the other hand, the impact of gender remained even when adjusting for college major. Table 2Logistic regression analysis predicting primary care specialty choicePredictorsModel 1College Major OnlyModel 2College Major + Premedical FactorsModel 3College Major + Premedical + Medical School FactorsOdds RatiosCI p* Odds RatiosCI p Odds RatiosCI p (Intercept)1.271.24–1.306318.9978.4–40890.0218.130.4–1565.2College major STEM (reference)111 Social science1.321.24–1.41 < 0.001 1.261.18–1.36 < 0.001 1.261.17–1.36 < 0.001 Humanities1.391.28–1.52 < 0.001 1.371.25–1.51 < 0.001 1.381.25–1.52 < 0.001 Other0.990.94–1.040.680.960.90–1.020.170.970.91–1.030.27Non-male gender2.202.11–2.30 < 0.001 2.112.01–2.20 < 0.001 Under-represented in medicine1.010.95–1.070.780.980.91–1.040.47College GPA (science courses)0.850.78–0.94 0.001 1.010.91–1.110.90College GPA (non-science courses)1.751.54–1.98 < 0.001 1.881.65–2.14 < 0.001 MCAT score0.980.98–0.98 < 0.001 1.000.99–1.000.087Pass on USMLE Step 1 first attempt0.330.26–0.40 < 0.001 Score on USMLE Step 2CK first attempt0.980.98–0.98 < 0.001 Specialty influence – Debt0.820.78–0.86 < 0.001 Specialty influence – Fellowship1.651.55–1.75 < 0.001 Specialty influence – Salary0.320.30–0.34 < 0.001 Specialty influence – Training lengths1.931.82–2.04 < 0.001 N41,22336,86336,526R20.0030.0530.121*p ≤ 0.05 was considered to be statistically significant Table 2Logistic regression analysis predicting primary care specialty choicePredictorsModel 1College Major OnlyModel 2College Major + Premedical FactorsModel 3College Major + Premedical + Medical School FactorsOdds RatiosCI p* Odds RatiosCI p Odds RatiosCI p (Intercept)1.271.24–1.306318.9978.4–40890.0218.130.4–1565.2College major STEM (reference)111 Social science1.321.24–1.41 < 0.001 1.261.18–1.36 < 0.001 1.261.17–1.36 < 0.001 Humanities1.391.28–1.52 < 0.001 1.371.25–1.51 < 0.001 1.381.25–1.52 < 0.001 Other0.990.94–1.040.680.960.90–1.020.170.970.91–1.030.27Non-male gender2.202.11–2.30 < 0.001 2.112.01–2.20 < 0.001 Under-represented in medicine1.010.95–1.070.780.980.91–1.040.47College GPA (science courses)0.850.78–0.94 0.001 1.010.91–1.110.90College GPA (non-science courses)1.751.54–1.98 < 0.001 1.881.65–2.14 < 0.001 MCAT score0.980.98–0.98 < 0.001 1.000.99–1.000.087Pass on USMLE Step 1 first attempt0.330.26–0.40 < 0.001 Score on USMLE Step 2CK first attempt0.980.98–0.98 < 0.001 Specialty influence – Debt0.820.78–0.86 < 0.001 Specialty influence – Fellowship1.651.55–1.75 < 0.001 Specialty influence – Salary0.320.30–0.34 < 0.001 Specialty influence – Training lengths1.931.82–2.04 < 0.001 N41,22336,86336,526R20.0030.0530.121*p ≤ 0.05 was considered to be statistically significant Logistic regression analysis predicting primary care specialty choice *p ≤ 0.05 was considered to be statistically significant Supplemental Appendix 4 presents the results of the generalized linear model regression analysis examining the relationship between college major and non-procedural specialty choice. Compared with STEM graduates, Social Science majors were 1.29 times (CI 1.20–1.38, p < 0.001) more likely and Humanities majors were 1.24 times (CI 1.13–1.35, p < 0.001) more likely to go into a non-procedural specialty (overall n = 41,223). As with primary care specialty, the relationship between college major and non-procedural specialty choice remained significant after introducing both demographic and pre-matriculation variables (Model 2) and medical school variables (Model 3), despite slightly different relationships with other predictors. In addition to data on the specialty actually chosen by each student after medical school, we obtained questionnaire data about students’ interest in primary care at the time of matriculation to medical school and at the end of year 2 of medical school, and conducted a repeated-measures regression (once in the full data set and once limited to students with responses at all three time points) to determine whether changes in primary care interest over time were associated with college major. Although there was a main effect of time and the previously described main effect of college major, there were no significant interactions between college major and time of measurement on interest in primary care (Supplemental Appendix 5). Hypothesis 2 The proportion of “patient-focused electives,” or electives that emphasize patient experience and contact, taken by medical students in our sample is shown in Table 1. Students that took a higher proportion of patient-focused electives were also more likely to go into both primary care (OR = 7.55, 95% CI 6.63–8.60, p < 0.001) and into non-procedural specialties (OR = 12.1, 95% CI 10.6–13.8, p < 0.001) than their peers (Supplemental Appendix 6). However, the models in Table 3 demonstrate that college major seems to have only a small relationship with students’ choice of patient-focused electives. In a causal mediation analysis, the impact of college major on primary care (and on non-procedural specialty) was not significantly mediated by the proportion of patient-focused electives. Table 3Linear regression predicting proportion of patient-focused electives taken in medical school to explore mediationPredictorsModel 1College Major OnlyModel 2College Major + Premedical FactorsEstimatesCI p* EstimatesCI p (Intercept)0.670.66–0.670.700.67–0.73College Major STEM (reference group)00 Social Science-0.00-0.01–0.000.41-0.01-0.01 – -0.00 0.004 Humanities0.00-0.01–0.010.61-0.00-0.01–0.000.52Other-0.01-0.01 – -0.00 0.005 -0.01-0.01 – -0.01 < 0.001 Non-male gender0.030.03–0.03 < 0.001 Underrepresented in medicine0.020.01–0.02 < 0.001 College GPA (science courses)-0.03-0.04 – -0.03 < 0.001 College GPA (non-science courses)0.020.01–0.03 < 0.001 N45,77441,870R2< 0.0010.013*p ≤ 0.05 was considered to be statistically significant Table 3Linear regression predicting proportion of patient-focused electives taken in medical school to explore mediationPredictorsModel 1College Major OnlyModel 2College Major + Premedical FactorsEstimatesCI p* EstimatesCI p (Intercept)0.670.66–0.670.700.67–0.73College Major STEM (reference group)00 Social Science-0.00-0.01–0.000.41-0.01-0.01 – -0.00 0.004 Humanities0.00-0.01–0.010.61-0.00-0.01–0.000.52Other-0.01-0.01 – -0.00 0.005 -0.01-0.01 – -0.01 < 0.001 Non-male gender0.030.03–0.03 < 0.001 Underrepresented in medicine0.020.01–0.02 < 0.001 College GPA (science courses)-0.03-0.04 – -0.03 < 0.001 College GPA (non-science courses)0.020.01–0.03 < 0.001 N45,77441,870R2< 0.0010.013*p ≤ 0.05 was considered to be statistically significant Linear regression predicting proportion of patient-focused electives taken in medical school to explore mediation *p ≤ 0.05 was considered to be statistically significant Discussion There is a growing shortfall of primary care specialists in the physician population in the United States [1, 10–12], not to mention worldwide [13], which is forecast to increase in future. At the same time, scholars consistently find that patient-centeredness and empathy decrease during medical training [21]. Both of these issues – decreasing numbers of primary care physicians and physicians with low rates of empathy – threaten the capacity of physicians to best serve the public, particularly as, at least in the US, our patient population becomes progressively more aged and infirm [9, 10]. Findings from our study suggest that matriculation of more medical students who completed pre-medical majors in the humanities and social sciences could be a partial solution. Indeed, our findings demonstrate that medical students who majored in the humanities and social sciences were more likely than their STEM major peers to choose primary care specialties and to choose non-procedural specialties. These relationships were consistent even when considering other social identity and performance factors, such as gender, race, GPA, and performance on USMLE examinations. However, there were differences between the two groups when examining College GPA for science vs. non-science courses. This could be, in part, because those students who were better at the natural sciences were more inclined to choose non-primary care specialties or could be because those students who were better at science courses were also more successful at getting into non-primary care specialties. Additionally, we found that medical students who took “patient-focused” electives were more likely than those who did not to choose primary care and non-procedural specialties. However, though our analyses demonstrate that both college major and taking patient-focused electives affect the likelihood of going into primary care and non-procedural specialties, there is no interaction between these two effects. In other words, we did not find support for our hypothesis that the relationship between going into primary care and college majors could be partially explained by engagement in patient-focused electives. Unfortunately, without longitudinal or qualitative data, we are unable to fully explain this finding. That said, there may be a more complex relationship between elective course-taking patterns and specialty choice than we had originally imagined. While previous proclivities or interests may influence decisions regarding elective choice, students may also use electives as signals to future specialty program directors. Future research may be needed to explore these patterns. This study has several key limitations. First, as a cross-sectional, rather than longitudinal study, our findings do not allow for strong causal claims. For example, while our data do demonstrate a clear relationship between undergraduate college major and future specialty-type, there is no way to identify whether there is an additional external factor that may influence both of these variables, as well. This is especially likely in the case of our Model 1, since the model explained a comparatively low amount of the variance in the relationship between our variables of interest. These external relationships are particularly hard to examine because in our US-based sample, college major is not chosen at random and students with aspirations to become physicians often face substantial pressure to choose their undergraduate majors according to common expectations and norms that push students towards the natural sciences (i.e., STEM majors). Likewise, those with particular worldviews may be both more likely to choose the humanities or social sciences in their undergraduate (college) training and more likely to choose primary care and non-procedural specialties. Another limitation is the US context within which this study took place. There are important differences between medical education pathways in the US and other countries, in which medical education is provided as part of undergraduate (university) training. As such, our findings may not be as transferable to contexts in which premedical pathways are more linear than in the US. That said, recent studies in the UK and Australia have explored the role of previous degrees and different pathways into medical training [34, 35]. Further future research on the more direct effects of humanities and social science content within such contexts may be interesting to investigate in relation to specialty choice. In addition, the data we received did not disaggregate race or ethnicity beyond URiM status; it would be interesting to note whether there are any variations between racial/ethnic groups in these patterns. Another limitation involves the age of the data; while we made a conscious effort to control for the effects of key changes in the medical education landscape due to COVID-19 and changes to Step 1’s scoring rubric by using data from prior to 2020, this may have affected the generalizability of these findings for the current context. Finally, given past arguments about the beneficial effects of health humanities coursework [36–39], a closer examination of the relationship between these courses and specialty selection is warranted. In the current US premedical training model, the pathway to, and at times, through medical school may be more challenging for those who pursue social science or humanities college majors [40]. These barriers may be overcome, however, with support from college premedical advisors building clearer roadmaps to provide students with guidance on the best approach to completing prerequisites for both their majors and for medical school, as well as the numerous additional extracurricular and application-related work that must be undertaken to make students viable candidates for medical school [40]. We hope that the findings from this study, combined with those from previous scholarship demonstrating the benefits of humanities and social science college training for both communication and empathy, may be used by admissions committees and medical school administrators to help continue to shift their thinking about what types of students will make successful, “good” students [41] and physicians and may provide support for the continued push for the social sciences and humanities in pre-medical and medical pathways [39, 42, 43]. Supplementary Information Supplementary Material 1. Supplementary Material 1.