International Clinical Phonetics and Linguistics Association 2016

Listener bias is a well-documented phenomenon in linguistics and sociolinguistics research. Listeners have demonstrated bias in assigning personality traits such as intelligence, selfconfidence and likeability to speakers from different language backgrounds (Delamere, 1996). A study by Derwing and Munro (1997), demonstrated that intelligibility scores assigned to nonnative speech were better when listeners were able to accurately identify the NL background of the speaker. While studies have demonstrated the effects of listener bias on measures of intelligibility and accentedness, fewer studies have investigated listener bias and goodness ratings. Accentedness is defined as a listener’s perception of the ELL’s speech as different from the listener’s own language community (Derwing & Munro, 2005). Goodness, is defined as a listener’s judgment of an utterance’s accuracy compared to the listener’s concept of an ideal prototype (Franklin & Stoel-Gammon, 2014). Few studies have examined the relationship between accentedness and goodness as measures of pronunciation proficiency. Therefore, the purpose of the study is to compare goodness and accentedness ratings of speech tokens among naïve and aware listeners and to investigate the role of listener bias with respect to these ratings. The research questions are as follows: 1. Do listeners rate goodness of /hVt/ tokens differently than they rate accentedness? 2. Do listeners who are aware of the speaker’s language background rate accentedness and goodness differently than listeners who are naïve to the language background? Method: Twenty eight monolingual speakers of U.S. English served as listeners. Twelve speakers from U.S. English, Spanish, Korean and Japanese native language (NL) backgrounds contributed /hVt/ tokens. Listeners were presented with five blocks of /hVt/ tokens each representing a different target vowel: /hit/, /hɪt/, /het/, /hæt/, /hɑt/. In addition to correct productions, each block included incorrect productions of each target vowel. The “incorrect” productions were constructed by including other /hVt/ productions to create foils. For example, the /hit/ block comprised correct productions of /hit/ as well as the following 3 foils: /hɛt/, /het/, /hɪt/. This design ensured that listeners heard the same incorrect productions regardless of the