Auditory stream segregation processes operate similarly in school-aged children and adults

Abstract

Our previous research with adults suggests that pre-attentive (bottom-up) brain processes govern auditory stream segregation [Sussman et al., 1998. Brain Res. 789, 130–138; Sussman et al., 1999. Psychophysiology 36, 22–34; Winkler et al., submitted for publication]. We investigated whether the pre-attentive mechanisms underlying auditory stream segregation operate similarly in school-aged (7–10 years of age) children and adults. We used an electrophysiological index of auditory change detection that does not require the experimental participant to focus on the sounds to be evoked. In Experiment 1, children were presented with mixtures of high and low frequency tones in different conditions and were instructed to watch a silent video and ignore the sounds. In Experiment 2, children were asked to listen to the same sets of sounds as presented in Experiment 1 and tell whether they heard one or two auditory streams. The pre-attentive processing of the mixture of sounds as one or two auditory streams (Experiment 1), matched with the perception of the sounds as one or two distinct streams (Experiment 2). Our results demonstrate that the mechanisms for auditory stream segregation operate similarly in school-aged children and adults when frequency proximity is the cue for segregation.

Introduction

From infancy, the human auditory system is confronted with overlapping sources of acoustic information. An important function of the auditory system is to sort the sounds to their original sources, creating a veridical representation of the auditory environment. Behavioral research with adults shows that the perceptual organization of sequential sounds to separated streams relies on a variety of acoustic cues (Bregman, 1990). When the mixture of sound is composed of generally similar acoustic characteristics, the sequence is usually heard as coming from a single source of sound; whereas when the characteristics of sound penetrates a multidimensional acoustic space, the mixture is usually heard as originating from more than one source. This process involves the grouping into a common auditory stream those acoustic properties that show coherence across time, within an ecologically valid set of temporal, spatial, spectral and timbral characteristics of the sound. Segregation of auditory input to sources allows us to hear a single voice in a crowd, detect an approaching train, or distinguish between a male voice and a female voice at a cocktail party.

Bregman (1990) hypothesized that auditory stream segregation is based on ‘primitive’ processes, attributable to innate mechanisms. Recent evidence (Sussman et al., 1998, Sussman et al., 1999, Ritter et al., 2000, Winkler et al., submitted for publication) supports this view showing that bottom-up sensory processes handle the segregation of sounds to distinct sources at a pre-attentive level of acoustic processing. This indicates that the auditory streams are sorted prior to stimulus selection. If automatic, low level systems handle the sorting of the incoming mixture of sounds this could facilitate the ability to select sources for further processing. That is, if the information about the separated sounds is maintained by low level systems, attentional resources would be free for higher cognitive processing of complex acoustic within-stream patterns such as is needed for understanding speech.

It is not known to what degree perceptual experience contributes to the acuity of this type of auditory scene analysis. A few previous studies have investigated auditory stream segregation in infants. They used various acoustic cues to test whether infants could perceive separated sound sources (timbre and spatial location, McAdams and Bertoncini (1997); and frequency proximity of pure tones, Demany (1982)). These studies found that segregation of auditory information could occur in infants but that much larger separations between the acoustic dimensions of the sounds tested were needed than for adults. Moreover, infants were not able to segregate sounds when they were presented at a rapid pace, despite the fact that speed promotes stream segregation in adults (Bregman, 1990, van Noorden, 1977). The results are consistent with literature that shows that infants have higher auditory thresholds than adults for the detection of differences in pitch (Nozza and Wilson, 1984, Ruben, 1992, Sinnott et al., 1983, Trehub et al., 1980, Werner et al., 1992), spatial location (Clifton et al., 1981, Muir and Field, 1979) or temporal resolution (Morrongiello and Trehub, 1987, Werner et al., 1992, Whightman et al., 1989). Furthermore, this suggests that some changes occur during the development of the auditory system that may contribute to the acuity and automaticity of the segregation process observed in adults.

Although the results of such studies do suggest that the basic mechanisms of stream segregation are set from birth, there is no research on the development of the underlying mechanisms between infancy and adulthood. Here we investigate whether the pre-attentive mechanisms underlying stream segregation operate on a similar basis in school-aged (7–10 years of age) children and adults. We manipulated frequency proximity as the basis of the segregation process and used stimulus parameters (brief tone duration (50 ms) and rapid interstimulus intervals (ISI)) that pre-attentively induced streaming in adults (Sussman et al., 1999). Discrimination of the frequency of brief duration tones (20 ms) reaches adult-like characteristics after the age of 7 years (Thompson et al., 1999). Therefore, if the segregation process is based primarily upon the child’s ability to discriminate between the frequency of tones, these parameters are suitable for determining whether the stimulus parameters that promote automatic segregation in adults operate similarly in the children tested. In order to determine whether stimulus-driven properties of a tone sequence promote stream segregation automatically in children, we used an electrophysiological index of auditory processing that does not require the experimental participant to actively process the sounds or to indicate his/her perception of them. The mismatch negativity (MMN), an event-related brain potential generated in or near primary auditory cortex, is the outcome of a change detection process that occurs when the incoming sound deviates from a detected regularity of the preceding auditory input. The location of the MMN generators in the auditory cortex can account for the scalp topography of the waveform, which is maximally negative over the fronto-central regions of the scalp and inverts in polarity at electrode sites located below the Sylvian fissure. MMN can be elicited most simply in the auditory oddball paradigm by infrequent sounds that differ from the frequently repeating sound in some stimulus feature (e.g. frequency, intensity or duration). It has been determined that the change detection process reflected by the MMN component is based upon the cortical sound representation of the regularity that is formed on the basis of the previous sound stimulation independently of the direction of focused attention (for detailed discussions, see Näätänen, 1990, Näätänen, 1992, Ritter et al., 1995, Schröger, 1997, Näätänen and Winkler, 1999). MMN, therefore, is an especially useful tool for measuring auditory sensory processes in infants and children because a behavioral task is not required to elicit the brain’s discriminative response. The MMN, unlike some other event-related potential (ERP) components, shows adult-like characteristics very early in development. MMN has been recorded in pre-term infants, newborns and school-aged children (4–12 years of age; Alho et al., 1990, Cheour et al., 1998, Cheour et al., 2000, Csépe, 1995, Cheour-Luhtanen et al., 1996, Gomes et al., 1999, Shafer et al., 2000). However, consistent with the findings that infants have higher auditory thresholds than school-aged children, larger pitch separations between the standard and deviant stimuli were needed to elicit reliable MMNs in subjects under 4 years of age (Morr et al., submitted for publication).

To summarize, MMN is an automatic brain response elicited by detected changes with respect to the sensory memory trace of the ongoing regularities of sound stimulation that can be used to test pre-attentive processing in school-aged children.

The main question addressed in the current study is whether the process of auditory stream segregation is pre-attentive in childhood. To test this, we presented a simple MMN auditory oddball paradigm in three different conditions. MMN will be elicited by the oddball (or ‘deviant’) tones with respect to the sensory memory trace of the regularity (i.e. the frequently repeating (or ‘standard’) tone). By presenting tones that vary randomly around the standard tone the regularity can be obscured, which could prevent MMN from being elicited by the deviant tones. The paradigm was manipulated so that the regularity of the repeating standard tone would be obscured unless the tones segregated to different frequency streams. The segregation of tones to high and low frequency streams was cued by the frequency separation between the standard and randomly varying tones, whereas intensity changes were used to elicit MMN. In this way, we separated the cues that would be used in the process of deviance detection (intensity change) from those used in the segregation process (frequency separation). If the segregation process occurs prior to the MMN-generating process, then the intensity deviants should elicit an MMN when the tones are segregated, similar to the MMN elicited by those tones when they are presented alone in a simple oddball sequence. In contrast, if the segregation process does not occur pre-attentively (e.g. if attention is needed to segregate the sounds), then MMN should only be elicited in the simple oddball sequence. In a second experiment we assessed the perceptual organization of the sounds into one or two auditory streams by measuring a behavioral response that reflected whether or not the children could actively segregate the test sounds used in the ERP experiment.