The purpose of this research was to investigate the nature of listeners' attitudes toward foreign-accented speech and the manner in which those attitudes are formed. This study measured 165 participants' implicit and explicit attitudes toward US- and foreign-accented audio stimuli. Implicit attitudes were measured with an audio Implicit Association Test. The use of audio stimuli as repeated tokens for their phonological attributes represents an innovation in IAT methodology. Explicit attitudes were elicited through self-report. The explicit task was contextualized as a fictional medical malpractice trial; participants heard the recorded audio testimony of two actors (one US-accented and one Korean-accented) portraying opposing expert witnesses. Four test conditions counterbalanced across participants were created from the recordings. Participants rated the experts on fourteen dependent variables ('traits'): believability, credibility, judgment, knowledge, competence, trustworthiness, likeability, friendliness, expertise, intelligence, warmth, persuasiveness, presentation style, and clarity of presentation. Participants were also asked for their attitudes toward the speakers relative to each other (i.e., Which doctor would you side with in this dispute?). The question of speaker preference was posed as a binary choice, an 11- point slider scale measure, and two confirmation questions asking participants to state how fair they thought an outcome for each party would be.

This study's hypothesis that participants' implicit and explicit attitudes toward the same speech would diverge was confirmed. The IAT results indicated an implicit bias [D=.33, p<.05] in favor of the US-accented speaker, while the self-report results indicated an explicit bias [ F(2,121)=3.969, p=.021, η²=.062] in favor of the foreign-accented speaker in the slider scale and confirmation questions [ F(2,121)=3.708, p=.027, η²=.058, and F(2,121)=3.563, p=.031, η²=.056]. While the binary choice question showed a trend toward favoring the foreign-accented speaker, the result was not significant. No discernable pattern was found to exist in attitudes toward the speaker by trait. This study's findings argue for the recognition of both implicit and explicit attitude constructs and the integration of implicit attitudes measurement methodologies into future language attitudes research. Additional theoretical implications of these findings for future language attitudes research are also discussed, including implications for selecting an appropriate cognitive processing model.