Vocal pitch shift in congenital amusia (pitch deafness)

Abstract

We tested whether congenital amusics, who exhibit pitch perception deficits, nevertheless adjust the pitch of their voice in response to a sudden pitch shift applied to vocal feedback. Nine amusics and matched controls imitated their own previously-recorded speech or singing, while the online feedback they received was shifted mid-utterance by 25 or 200 cents. While a few amusics failed to show pitch-shift effects, a majority showed a pitch-shift response and nearly half showed a normal response to both large and small shifts, with similar magnitudes and response times as controls. The size and presence of the shift response to small shifts were significantly predicted by participants’ vocal pitch matching accuracy, rather than their ability to perceive small pitch changes. The observed dissociation between the ability to consciously perceive small pitch changes and to produce and monitor vocal pitch provides evidence for a dual-route model of pitch processing in the brain.

Introduction

The voice is one of the most important tools for communication in the auditory domain, and pitch, which plays an integral role in both the linguistic and musical domains, represents a very important parameter of both vocal perception and production. In the current study, we aim to find evidence of a functional division between the abilities to perceive and produce pitch. One natural place to look for evidence of such a dissociation is in congenital amusia. This is a neurogenetic disorder (Peretz, Cummings, & Dube, 2007) estimated to affect about 4% of the population, which is characterized by impaired music perception and memory ability in the absence of other cognitive deficits, hearing loss, brain damage, or lack of exposure (Peretz, 2008, Stewart, 2011). Congenital amusia is formally diagnosed by the Montreal Battery of Evaluation of Amusia (MBEA; Peretz, Champod, & Hyde, 2003), and it is believed that the core deficit of congenital amusics is a disability to perceive fine-grained pitch differences. For example, amusics are impaired in their ability to discriminate between repeated tones which are less than a semitone (one twelfth of an octave) apart (Hyde and Peretz, 2004, Peretz et al., 2002) as well as in distinguishing between rising and falling pitches (Foxton et al., 2004, Liu et al., 2010).

Amusics typically report pitch production difficulties to complement their perceptual problems, and are generally judged by listeners as having impaired singing abilities (Ayotte, Peretz, & Hyde, 2002). Congenital amusics make more pitch errors than non-amusics when singing familiar melodies (Dalla Bella et al., 2009, Tremblay-Champoux et al., 2010) and when imitating single tones (Hutchins, Zarate, Zatorre, & Peretz, 2010). However, some amusic participants show relatively preserved singing abilities despite their severe perceptual problems. Dalla Bella et al. (2009) identified a subset of amusics who showed relatively well-preserved singing abilities, and Hutchins, Gosselin, and Peretz (2010) showed that even erroneous singing in amusics tended to show a consistent and linear relationship between target tones and sung responses, although their singing proficiency was not linked to their perceptual ability. Loui and colleagues (Loui, Guenther, Mathys, & Schlaug, 2008) provided evidence of a dissociation in amusics between the ability to perceive and produce the direction of a pitch change; congenital amusics were able to sing back a musical interval in the same direction (but not of the same size) as a model, but could not reliably categorize the same interval as going up or down in pitch. Williamson and colleagues (Williamson, Liu, Peryer, Greierson, & Stewart, 2012), however, failed to replicate the results of Loui et al. (2008), and suggested that differences in amusics between perception and production abilities may instead arise from task demands. In speech, Liu and colleagues (Liu et al., 2010) have also shown some evidence that amusics are better at imitating statements and questions than they are at discriminating the same sentences, although they are nevertheless impaired at both tasks. A recent study by Hutchins and Peretz (2012b) showed that when the meaning of the sentence is controlled for, amusics were unimpaired at imitating pitch changes in sentences, but were considerably worse than controls at discriminating those same pitch changes.

An interesting paradigm to study pitch production abilities further in congenital amusia is the pitch shift reflex. When the pitch of the feedback that speakers or singers receive is altered while they are vocalizing, there is generally an automatic compensatory pitch change in the opposite direction of the pitch shift (Burnett et al., 1998, Burnett and Larson, 2002, Elman, 1981). This response begins just over 100 ms after the onset of the pitch shift, and is considered to be automatic and reflexive (Burnett et al., 1998), although most studies did not test explicitly for awareness of the shift. There is some evidence that the pitch shift reflex can be divided into an early, automatic component and a later, more controlled component (Burnett et al., 1998, Hain et al., 2000), although often there is not a clear separation between the two parts in vocal responses. The response size seems to vary with the size of the shift (Liu & Larson, 2007), although it is possible that this may only be the case for sustained, rather than ephemeral (e.g. 100 ms) pitch shifts (Burnett et al., 1998), and only compensates for part, rather than all, of the shift.

Computationally, the pitch shift reflex is thought to be the result of the process whereby the efference copy of the motor command to the voice is compared to the actual auditory feedback received. If there is a mismatch between the actual and predicted outcomes, the vocal-motor command is modified to correct for this presumed error. This model makes use of the parity between motor and auditory representations of pitch, and predicts that auditory information about pitch is translated into a vocal-motor code representation (see Hutchins et al., 2010, Liberman and Mattingly, 1985). This process implies that reflexive pitch shifts could be made even in the absence of their conscious perception. Hafke (2008) showed that trained singers did in fact make adjustments to their pitch in a pitch shift paradigm even in cases where they did not explicitly recognize that a small shift (9 or 19 cents; 100 cents = 1 semitone) had occurred during a trial, providing evidence for a dual-route theory of pitch processing.

The current experiment constitutes a further test of dual-route theories of pitch processing. Here, we test congenital amusics and matched controls in a pitch shift paradigm, using 200 cent and 25 cent shifts. Both of these shifts evoke reliable pitch shift responses among normal participants (Burnett et al., 1998, Hawco et al., 2009, Liu and Larson, 2007, Zarate et al., 2010). However, 25 cent shifts are difficult for normal listeners to perceive (Hafke, 2008), and are substantially harder to detect for amusics than for controls (Hyde & Peretz, 2004). If we find evidence that amusics are making compensatory responses to these shifts, this provides evidence that their auditory system is processing the feedback from their own voice in terms of the vocal-motor code, even though conscious perception of the pitch is seriously impaired by their deficit, representing a dissociation between pitch perception and production.

Because congenital amusics often have difficulty in vocally matching a target pitch (Dalla Bella et al., 2009, Hutchins et al., 2010), we employ a self-matching method, in which participants first record instances of their own voice, and match the pitch of the recordings of their own voice. This method has been shown to significantly improve vocal pitch matching among nonmusicians (Hutchins & Peretz, 2012a), and eliminates the possibility of insufficient vocal range. We test participants using both spoken and sung material, to investigate whether any music-specific deficits can be found in amusia. Here, amusics and controls produce three one-syllable words and three different notes as targets. They imitate their own voice, holding the note or syllable for about 3 s. In the pitch shift trials, their imitations are shifted either up or down by 25 or 200 cents (or left unshifted as a control). Participants were told there would be pitch shifts during their responses, and were instructed to ignore them as best they could and continue vocalizing until the end of the trial. We hypothesize that controls will show a normal pitch-shift response, with different response magnitudes for large and small shifts. We also predict that amusics should show a normal pitch-shift response, showing significant differences from the non-shift condition in both the 25 cent and 200 cent shift conditions. This would show that amusics can in fact perceive production-relevant aspects of pitch, indicating that they have a preserved production pathway.