Listeners unconsciously index race, gender, and sexual orientation from little phonetic input. Psycholinguistic theory attributes this to a representation of language where episodic traces encode fine, speaker-specific acoustic detail. We examined this unconscious linguistic bias in a task with native English speaker participants. Participants were presented with two pictures, both of which were either Hispanic or Caucasian males. Simultaneously, they listened to a semantically-neutral English sentence spoken by either a bilingual Spanish-English male or a monolingual English male and were told to choose which pictured man said the sentence. When the auditory stimulus was bilingual speech, speakers were quicker to choose between the two Hispanic faces than between the Caucasian faces. However, when participants heard a monolingual voice, there was no difference in reaction time. In the second phase of the experiment, participants were instead presented with two different pictures, one Hispanic and one Caucasian. Listeners were more likely to associate bilingual English with the Hispanic male than the Caucasian male and vice versa (monolingual English with Caucasian male). These results suggest that listeners quickly index male Spanish-English speech, but monolingual English-Caucasian associations may not be as robust.