Gamma-band activity reflects the metric structure of rhythmic tone sequences

Abstract

Relatively little is known about the dynamics of auditory cortical rhythm processing using non-invasive methods, partly because resolving responses to events in patterns is difficult using long-latency auditory neuroelectric responses. We studied the relationship between short-latency gamma-band (20–60 Hz) activity (GBA) and the structure of rhythmic tone sequences. We show that induced (non-phase-locked) GBA predicts tone onsets and persists when expected tones are omitted. Evoked (phase-locked) GBA occurs in response to tone onsets with ∼50 ms latency, and is strongly diminished during tone omissions. These properties of auditory GBA correspond with perception of meter in acoustic sequences and provide evidence for the dynamic allocation of attention to temporally structured auditory sequences.

Introduction

Human auditory communication signals can be characterized as sequences of approximately discrete events, such as musical notes and speech syllables. Rhythm refers to the temporal patterning of such event sequences, especially the timing of events and the emergence of accent patterns. In music and speech, event onsets are perceived as being periodic with inter-onset intervals (IOIs) in the 100–1000 ms range. This time scale corresponds to the range over which humans are most sensitive to changes in tempo for periodic and nearly periodic sequences [6] and the rates for which anticipation is observed in sensorimotor synchronization tasks [7]. Rhythms that display multiple levels of periodic temporal structuring are called metric rhythms, and give rise to perception of alternating strong and weak accents [17], [18]. Metric rhythms are easier to remember and reproduce than nonmetric rhythms [8], and result in characteristic patterns of functional brain activation [24]. Moreover, individual acoustic events are perceptually facilitated when presented in the context of metrically regular sequences [12], [14]. Thus, attention may be allocated more efficiently to events as metric rhythms unfold in time, because they are more predictable [14].

Despite the significance of event timing in the perception of music and speech, no direct manifestation of metric expectancy for individual events has been reported in electroencephalography (EEG) or magnetoencephalography (MEG) recordings of human cortical activity. A recent study used an oddball paradigm to show that metrically strong beats in a sequence elicit a larger late response (∼500 ms) possibly reflecting greater post-stimulus attention allocation [3]. However, it is likely that other aspects of meter perception occur earlier, in anticipation of the occurrence of sounded events [12], [14]. Moreover, given that alternating accents are characteristic of meter perception, we would expect to observe differential activity on strong versus weak beats within the metrical cycle.

Long-latency auditory responses have been studied extensively but are not typically studied at tempos characteristic of music and speech. This is in part because the responses diminish in amplitude at fast tempos and because responses from adjacent tone onsets begin to overlap at rhythmic rates with IOIs shorter than 1000 ms [5]. In contrast, short-latency evoked (phase-locked) gamma-band (20–60 Hz) activity (GBA) quickly follows stimulus events at a range of tempos [27], while induced (non-phase-locked) GBA can occur at various latencies and has been linked to integrative sensory, motor, and cognitive functions¹ [22], [28]. The short latency and transience of auditory GBA suggests that it might underlie onset processing in auditory patterns, and we further hypothesized that it might underlie onset expectancy. The fact that induced GBA does not require precise phase locking to tone onsets suggests that it might exhibit activity prior to tone onset or around the time of missing tones. Such anticipatory activity is predicted by theories of meter perception that posit internal attentional processes that synchronize with external rhythmic patterns forming the basis of metric percepts [14], as illustrated in Fig. 1. Therefore, we investigated the relationship between evoked and induced GBA and the structure of metric rhythms in subjects while they listened to simple metrically structured tone patterns that included missing tones at strongly and weakly accented temporal positions.