Metrical expectations from preceding prosody influence spoken word recognition

Metrical expectations from preceding prosody influence spoken word recognition Meredith Brown (mbrown@bcs.rochester.edu) Department of Brain & Cognitive Sciences, University of Rochester Meliora Hall, Box 270268, Rochester, NY 14627-0268 Anne Pier Salverda (asalverda@bcs.rochester.edu) Department of Brain & Cognitive Sciences, University of Rochester Meliora Hall, Box 270268, Rochester, NY 14627-0268 Laura C. Dilley (ldilley@msu.edu) Department of Communicative Sciences & Disorders, Michigan State University 116 Oyer, East Lansing, MI 48824 Michael K. Tanenhaus (mtan@bcs.rochester.edu) Department of Brain & Cognitive Sciences, University of Rochester Meliora Hall, Box 270268, Rochester, NY 14627-0268 Abstract Two visual world experiments tested the hypothesis that ex- pectations based on preceding prosody influence the percep- tion of suprasegmental cues to lexical stress. Experiment 1 showed that phonemically overlapping words with different initial stress patterns compete for recognition. Experiment 2 further demonstrated that fundamental frequency and sylla- ble timing patterns across material preceding the target word can influence the relative activation of competing alternatives with different initial stress patterns. The activation of alter- natives with initial stress was higher when preceding stressed syllables had suprasegmental acoustic characteristics similar to the initial syllable of the target word. These findings sug- gest that expectations about the acoustic realization of an ut- terance include information about metrical organization and lexical stress, and that these expectations constrain the initial interpretation of suprasegmental stress cues. These results are interpreted as support for expectation-based forward models in which acoustic information in the speech stream is interpreted based on expectations created by prosody. Keywords: Prosody; spoken word recognition; lexical stress; visual world paradigm; expectations; lexical competition Introduction A growing body of work indicates that expectations about the acoustic realization of the phonemes and prosody of a spoken sentence influence how listeners initially interpret incoming acoustic-phonetic cues during spoken language processing. For example, manipulations of pitch and duration early in an utterance influence the interpretation of cues to prosodic and morphophonemic constituency several syllables downstream (Dilley & McAuley, 2008; Dilley et al., 2010; Brown et al., 2011; Dilley & Pitt, 2010). However, little is known about the types of representations that contribute to these perceptual ex- pectations. The present study investigates whether perceived prosodic and metrical patterning across preceding portions of an utterance can influence listeners’ expectations about the metrical organization of upcoming material, modulating their interpretation of proximal cues to lexical stress and therefore influencing the activation of potential lexical candidates. Lexical stress is a key contributor to sentence-level promi- nence patterns and rhythmicity. Listeners are sensitive to segmental and suprasegmental cues to stress during spoken word recognition (Cutler, Dahan & van Donselaar, 1997). Although vowel quality is the most potent stress cue to influ- ence lexical processing in English, other suprasegmental cues such as duration also distinguish stressed from unstressed syllables, and judgments about these suprasegmental stress cues are modulated by surrounding prosody in off-line tasks (Niebuhr, 2009). Cues to stress may influence not only the recognition of particular words, but also the process of seg- menting the speech stream more generally. For example, lis- teners are more likely to misperceive phrases like “she’s a must to avoid” as “she’s a muscular boy” than they are to mishear “in closing” as “enclosing”, suggesting that listeners preferentially posit word boundaries prior to prominent sylla- bles (Cutler & Butterfield, 1992). This metrical segmentation strategy is substantiated by the distribution of stressed sylla- bles within the English lexicon: approximately 90% of con- tent words in conversational English have initial stress (Cutler & Carter, 1987). Perceived metrical patterning is a potentially powerful source of expectations in speech perception. Speech prosody often exhibits characteristics that listeners perceive as pattern- ing (Couper-Kuhlen, 1993; Pierrehumbert, 2000). For ex- ample, listeners tend to hear stressed syllables in English as perceptually isochronous, i.e., as occurring at regular inter- vals (e.g., Lehiste, 1977). In addition, previous work using non-linguistic auditory stimuli (e.g. sequences of alternating tones) has demonstrated that pitch, temporal, and/or ampli- tude patterning in distal (i.e. non-local) auditory context can influence the processing of proximal material (e.g. the per- ceived relative prominence of high vs. low tones; Woodrow, 1911; Thomassen, 1982). The tendency for speakers to use recurring sequences of pitch accents within an intonational phrase (Couper-Kuhlen, 1993; Pierrehumbert, 2000) may likewise contribute to the perceived metrical structure across syllables in an utterance. We conducted two visual world experiments to test the hy-