Purpose <p>Ethnic/racial bias was found in MSPEs of general surgery residency applicants. However, which portion(s) of the MSPEs is/are associated with the bias remains unknown. We aim to answer the question: what portion(s) of the MSPEs is/are associated with ethnic/racial bias as measured by differential use of communal and agentic terms?</p> Methods <p>Retrospective study evaluating the source of ethnic/race bias, as measured by differential use of agentic and communal terms, in all MSPEs of residency applicants to a single institution from two consecutive match cycles. A separate bias score was calculated for the different portions of the MSPE, and multivariable regression was used to assess the association between each score (of the different portions) and ethnicity/race (URiM versus non-URiM). US medical students applying for a categorical surgery residency position at a single academic institution for two consecutive Match cycles were included.</p> Results <p>1314 MSPEs from 146 medical schools were included. Baseline characteristics were comparable between application cycles. Genders were similarly distributed (women, 51.6%). Most applicants identified as white or Asian (52.7% and 25.3%, respectively), and 17% self-reported as URiM. Applicants identifying as URiM had lower median overall and “structure”/summary bias scores (9.09 and 13.33, respectively) when compared to white applicants (13.04 and 20, respectively), indicating differential word use. Additionally, MSPEs of URiM applicants were more likely to be communally-biased in the “structure”/summary portions of their MSPE when compared to white applicants.</p> Conclusions <p>The source of biased writing in the MSPEs of general surgery residency applicants was found within the “structural”/summary portion of the document.</p>