Atypical processing of auditory temporal complexity in autistics

Abstract

Autistics exhibit a contrasting combination of auditory behavior, with enhanced pitch processing abilities often coexisting with reduced orienting towards complex speech sounds. Based on an analogous dissociation observed in vision, we expected that autistics’ auditory behavior with respect to complex sound processing may result from atypical activity in non-primary auditory cortex. We employed fMRI to explore the neural basis of complex non-social sound processing in 15 autistic and 13 non-autistics, using a factorial design in which auditory stimuli varied in spectral and temporal complexity. Spectral complexity was modulated by varying the harmonic content, whereas temporal complexity was modulated by varying frequency modulation depth. The detection task was performed similarly by autistics and non-autistics. In both groups, increasing spectral or temporal complexity was associated with activity increases in primary (Heschl's gyrus) and non-primary (anterolateral and posterior superior temporal gyrus) auditory cortex Activity was right-lateralized for spectral and left-lateralized for temporal complexity. Increasing temporal complexity was associated with greater activity in anterolateral superior temporal gyrus in non-autistics and greater effects in Heschl's gyrus in autistics. While we observed similar hierarchical functional organization for auditory processing in both groups, autistics exhibited diminished activity in non-primary auditory cortex and increased activity in primary auditory cortex in response to the presentation of temporally, but not of spectrally complex sounds. Greater temporal complexity effects in regions sensitive to acoustic features and reduced temporal complexity effects in regions sensitive to more abstract sound features could represent a greater focus towards perceptual aspects of speech sounds in autism.

Introduction

Behavioral evidence indicates that the cognitive architecture of visual and auditory perceptual processing may be differently organized in autism (Behrmann et al., 2006, Dakin and Frith, 2005, Mottron et al., 2006, Samson et al., 2006). The diagnostic criteria for autism (Lord et al., 1997) include signs related to both hypo- and hyper-reactivity to sounds (Grandin and Scariano, 1986, Metz, 1967, Novick et al., 1980). While autistics¹ may display apparent disinterest in speech sounds, and aversive reactions to vacuum cleaner and crowd noises (Goldfarb, 1961), they may also have heightened musical interests, and enhanced auditory abilities such as superior pitch memory and pure tone discrimination (Bonnel et al., 2003, Heaton, 2003).

Auditory processing atypicalities in autism have been interpreted in two frameworks. Weak central coherence theory (Frith & Happe, 1994) hypothesizes that autistics have difficulty integrating local auditory features into larger ensembles at the global level (Kellerman et al., 2005, Nieto Del Rincon, 2008). However, reports of intact global auditory processing in autism challenge this hypothesis (Heaton, 2005, Mottron et al., 1999, Mottron et al., 2000). The enhanced perceptual functioning model (EPF) (Mottron et al., 2006) describes both the superiorities of processing local features and the intact global contour processing within hierarchical auditory patterns (e.g. melody). Moreover, this model emphasizes the link between the bias towards local elements in hierarchical auditory patterns and superior pitch detection for pure tones (Heaton, 2003, Mottron et al., 2000). However, the EPF predictions regarding the processing of psychophysically complex sounds are essentially derived from evidence reported from studies investigating early visual processing that have demonstrated superior abilities in autistics for simple, luminance-defined, information extracted by mechanisms operating within primary (V1) visual cortex (Bertone et al., 2005, Plaisted et al., 1998), and lower performance for tasks involving more complex visual processing requiring involvement of both primary (V1) and non-primary (V2, V3) regions of visual cortex (Bertone et al., 2003, Bertone et al., 2005, Blake et al., 2003, Milne et al., 2002, Pellicano et al., 2005, Vandenbroucke et al., 2009). Atypical integration between primary and non-primary regions of the visual cortex could underlie this dissociation (Bertone et al., 2005). This heuristic was recently extended to audition, resulting in predictions of differential processing by autistics for simple, compared to complex, auditory stimuli (Samson et al., 2006). In support of this idea, enhanced pitch processing has been documented in numerous behavioral (Bonnel et al., 2010, Bonnel et al., 2003, Heaton, 2003, Heaton, 2005, Jones et al., 2009, O’Riordan and Passetti, 2006) and electrophysiological studies of autistics (Ferri et al., 2003, Gomot et al., 2002, Lepisto et al., 2008, Lepisto et al., 2005). In some cases, the behavioral advantage reaches outstanding levels, extending beyond four and five standard deviations above the mean of the control group (Heaton, Davis, & Happe, 2008). Moreover, superior processing of individual sound components might underlie enhanced chord disembodying (Heaton, 2003, Miller, 1989, Mottron et al., 1999) or the unimpaired discrimination of non-social complex sounds in autistics if they were to achieve successful processing though the decomposition of complex sounds (Bonnel et al., 2010).

The relevance of studying auditory perception in autism is not limited to the peaks of ability, as most studies report diminished abilities in processing social auditory information in this population. This is the case for speech recognition in noise (Alcantara et al., 2004, Groen et al., 2009) or prosody perception (Kujala et al., 2005, Peppe et al., 2007). However, typical voice processing abilities in autistics have been reported (Boucher, Lewis, & Collis, 2000). Event-related potential studies have shown reduced cortical responses to complex speech-like sounds, including vowels (Ceponiene et al., 2003, Lepisto et al., 2005, Lepisto et al., 2006, Whitehouse and Bishop, 2008) and consonant–vowel syllables (Jansson-Verkasalo et al., 2003, Russo et al., 2009). Finally, reduced activation of the “voice area” in the superior temporal sulcus (STS) has been reported in autistic adults (Gervais et al., 2004), and a reduced leftward asymmetry has been observed for speech processing (Boddaert et al., 2003, Boddaert et al., 2004, Lepisto et al., 2005, Lepisto et al., 2006, Minagawa-Kawai et al., 2009). It is therefore plausible that atypical processing of psychophysical properties of complex sounds plays a role in the apparent disinterest for speech, evident in most autistics, at least in their early years.

As in the visual system (Grill-Spector & Malach, 2004), auditory cortical analysis is organized hierarchically, with simple feature extraction at the primary level providing input to non-primary fields that subsequently extract more complex features. This functional organization scheme receives empirical support from both animal and human studies. In non-human primates and cats, the primary or ‘core’ auditory region, located in Heschl's gyrus (HG), is more tonotopically organized by frequency, has sharper frequency tuning and shows lower thresholds to pure tones as compared to non-primary auditory fields (Merzenich and Brugge, 1973, Morel et al., 1993, Rauschecker et al., 1995, Schreiner and Cynader, 1984). The non-primary neurons within the superior temporal gyrus (STG), labeled as ‘belt’ and ‘parabelt’ auditory regions, show broader individual frequency tuning, collectively respond to a broader range of frequencies and are selectively responsive to more complex stimuli such as band-passed noise (Rauschecker et al., 1995, Recanzone, 2000). These physiological findings combined with the known anatomical connections among primary and associative auditory regions (Hackett, Stepniewska, & Kaas, 1998) lend support to hierarchical organizational accounts of information flow in non-human auditory cortex (Kaas and Hackett, 1998, Kaas and Hackett, 2000, Rauschecker, 1998). Similar organizational plans are evident in human auditory cortex. An fMRI study reported that pure tone presentation resulted in activity increases in primary auditory cortex (HG), whereas complex band-passed noise elicited activity increases extending to the surrounding non-primary auditory fields in the anterolateral aspect of HG and STG (Wessinger et al., 2001), consistent with the location of the belt region in macaques (Rauschecker, 1998). Similarly, imaging studies have consistently revealed that spectrally complex sounds, with multiple harmonic components, and temporally complex sounds, with varying frequency or amplitude in time, elicit activity increases extending to non-primary auditory areas in anterior, lateral and posterior STG, corresponding to the belt and parabelt regions (Binder et al., 2000, Giraud et al., 2000, Hall et al., 2002, Hart et al., 2003, Schonwiesner et al., 2005a, Schonwiesner and Zatorre, 2009, Thivard et al., 2000, Zatorre and Belin, 2001). Moreover, while primary areas (HG) are sensitive to acoustic variations in speech sounds, non-primary areas within anterior and posterior superior temporal regions are more responsive to abstract sound features like intelligibility than to acoustic signal variations (Okada et al., 2010), supporting the role of these non-primary fields in the processing of more complex auditory information (Rauschecker & Scott, 2009). In addition to a within-hemisphere hierarchical architecture, auditory processing models also incorporate lateralization features (Zatorre, Belin, & Penhune, 2002) mainly, a leftward asymmetry of the auditory cortical response to temporal sound variation (Belin et al., 1998, Jamison et al., 2006, Schonwiesner et al., 2005a, Zaehle et al., 2004, Zatorre and Belin, 2001) and rightward to spectral sound variation (Jamison et al., 2006, Patterson et al., 2002, Schonwiesner et al., 2005b, Zatorre and Belin, 2001).

We used fMRI to examine differential auditory cortical responses to stimuli of varying spectral and temporal complexity in autism. Before being able to account for differential behavioral performances in naturalistic situations, and specifically for atypical autistic performance in music, language and voice processing, a preliminary study exploring the effects of variation in fundamental sound properties is required. For this purpose, we exposed our participants to simple and controlled stimuli, that is pure tones, in addition to spectrally and temporally complex sounds. Although spectral and temporal acoustic features are important characteristics of more complex sounds like speech, it is important to note that the stimuli used here do not represent actual features of speech and remain very simple in comparison to components like vowels, consonants or the structure of formants.

We define sounds as spectrally complex when they include more than one frequency component or harmonic. Similar stimuli have been used to investigate cortical auditory spectral complexity processing (Hall et al., 2002, Hart et al., 2003). We define temporally complex sounds as having frequency variation over time. Similar frequency-modulated sounds have been used in imaging studies (Hall et al., 2002, Hart et al., 2003). Our experiment is novel in this regard, as we use a parametric design with three levels of temporal complexity with increasing frequency modulation depth while maintaining constant modulation rate.

On the basis of previous work in typical individuals, we predicted that sounds of higher spectral and temporal complexity would induce increased activity in primary (HG) and non-primary auditory cortex, mainly with spatial extension to anterolateral STG (Hall et al., 2002, Hart et al., 2003). Based on hypothesized reduced integration among auditory cortical regions (Bertone et al., 2005, Samson et al., 2006), autistics should exhibit reduced activity in response to complex auditory material in non-primary auditory areas, with higher sensitivity to complex sound features. Between-group effects would possibly be more important in response to temporal complexity, which is specifically important for speech recognition, particularly low modulation rates as the one used here (Houtgast and Steeneken, 1985, Tallal et al., 1993). In terms of response lateralization, we expected a rightward asymmetry associated with spectral complexity and leftward for temporal complexity.