Many people in the US use more than one language at home, yet English remains the dominant (L1) language in US society, which can complicate medical encounters. In this study we ask in what ways effective communication can be ensured in health care settings when speakers differ in language proficiency. One strategy people use is second language (L2) speech accommodation, which is characterized by slowed speech, less complex words, and clearer enunciation. We employ a mixed-reality platform called MURSION to document how a group of Physician Assistant students use speech accommodation during a healthcare encounter. MURSION is a computer-based virtual environment where participants interact with an Avatar controlled by a human interactor in a standardized environment. We record 5-minute interactions between the student and a high or low English proficiency Avatar. Our analyses evaluate lexical choices in L1-L2 interactions with SCOPE (South Carolina Psycholinguistic Metabase) and acoustic properties with PRAAT. Results show that clinical students use slower speech and high frequency words when speaking to a low proficiency virtual patient, indicating a sensitivity for the communicative needs of L2 English users. Speech accommodation results will contribute to communication training modules for clinicians to interact efficiently with linguistically diverse populations.