Transfer effects in the vocal imitation of speech and song.

Abstract

In this study, we investigated how practice imitations of speech impacted imitations of songs and vice versa. Participants were first asked to practice imitating sung or spoken sequences, and then to imitate a new sequence, which could differ with respect to domain (speaking or singing), global pitch contour (question vs. statement pattern), and/or words. Pitch accuracy during transfer was affected by changes to domain and contour, but not text. Somewhat surprisingly, best transfer was found either when both domain and contour remained the same or both changed. Transfer performance suffered when only one feature changed and the other remained consistent. Analyses of individual differences showed that poor-pitch imitators had a harder time adopting the pitch structure of new sequences, regardless of whether the sequence was speech or song. Results were not consistent with claims for either independence or complete integration of music and language, but instead argue for differences in domain possibly based on the salience of pitch structures in the signal.Keywords: vocal imitation, imitation learning, poor-pitch singers, generalization(ProQuest: ... denotes formula omitted.)Debate concerning the systems underlying music and language has centered around the processing of pitch. The influential modular model of Peretz and Coltheart (2003) focuses on processing of pitch as a critical difference between the domains. Specifically, this model predicts that pitch information within a musical context (e.g., in song) is processed by a tonal encoding module independent of spoken pitch. Other accounts, here termed integrationist, propose that pitch is processed similarly in both speech and musical contexts (e.g., Koelsch, 2011; Koelsch & Seibel, 2005). For instance, Koelsch and Siebel (2005; see also Koelsch, 2011) propose that the extraction of meaning from pitch and timbre occurs in similar modules for speech and music. Similarly, several auditory scene analysis models (e.g., Bregman, 1990; Patterson, Allerhand, & Giguere, 1995; Yost, 2007) propose that the perception of speech and music is formed by the same set of computations.The validity of independence versus integrationist accounts for pitch perception in music and language continues to be a source of debate. However, our perspective has to do with the ability to produce pitch patterns by imitation. Two points motivate this focus. First, research in both music and language cognition is dominated by perception as opposed to production, leaving open to question whether similar effects found in perception research also hold for production. Second, research has suggested that some individuals exhibit a production-specific musical pitch deficit, here termed poor-pitch singing (Pfordresher & Brown, 2007; Roberts & Davies, 1975; Welch, 1979), that may exist in the absence of any perceptual deficit such as congenital amusia (Peretz et al., 2002). Both points led us to explore whether the imitative production of speech incorporates a pitch processing system that is distinct from the imitative production of song.Independent Versus Integrationist Views of Music and LanguageModels of music and language processing can be said to fall along a continuum bracketed by the extreme views of full independence and complete integration. A fully independent view, following Fodor's description of modularity (Fodor, 1983, 2000), conceptualizes speech processing as comprising a distinct (domain specific) set of processing modules that are not shared with music processing, and furthermore are not influenced by information processing that occurs within the music system (information encapsulation). A similar set of constraints should exist for music processing. One view that approximates this approach, though focusing on domain specificity rather than encapsulation, is the aforementioned model of Peretz and Coltheart (2003). With respect to pitch processing-whether in perception or production-a fully independent view would predict that constraints or benefits associated with pitch processing in one domain (e. …