The effects of auditory attention measured from human electrocorticograms

Abstract

Objective

A central question in auditory electrophysiology has been whether selective attention can modulate exogenous components of the scalp-recorded N1 (the ‘N1 effect’). Intracranial electrocorticograms were used in the current work to investigate this issue in greater anatomical detail.

Methods

Data were recorded from subdural electrodes placed across temporal cortex in 6 patient-volunteers undergoing diagnostic procedures for medically intractable epilepsy. Patients performed a dichotic listening task in which they alternately attended to a series of tones presented to both ears (mean ISI 800 ms) by responding to rare frequency deviants.

Results

Effects of attention were measured on the largest negative and positive waveform deflections observed between 70 and 220 ms post-stimulus for stimuli presented contralateral to grid location. Peak deflections were most often recorded from the upper bank of the posterior superior temporal gyrus at approximately 89 and 173 ms on average (labeled N90stg and P170stg, respectively). Selective attention had little effect on peak latencies but significantly increased the N90stg for 3 subjects, increased the P170stg for two subjects, and decreased the P170stg for two other subjects.

Conclusions

Selective auditory attention can modulate neural response in auditory cortex.

Significance

The effects of attention on the scalp-recorded N1 component may arise in part from the enhancement of exogenous responses in temporal cortex.

Introduction

Many attempts have been made over the last several decades to provide a physiological foundation to the classic psychological debate regarding when and how selective attention affects sensory processing. This debate has traditionally centered around two opposing models of attentional intervention, the early and late selection theories. Both views are based on the assumption that cognitive resources are limited, and hence only some sensory input may be selected for continued processing en route to conscious awareness or a behavioral response. The role of selective attention is thought to determine when and how signals are selected for further processing.

The early-selection theory (Broadbent, 1958) posits that the attentional ‘bottleneck’ occurs at an early stage before sensory processing is complete. Such a view suggests an anatomical locus early in the sensory stream, perhaps at the thalamus or modality-specific cortex. Late-selection models argue that some information from the unattended channels is processed along with attended information (Treisman, 1960), or may even undergo full semantic analysis (Deutsch and Deutsch, 1963), before attention intervenes to select information for further cognitive processing. This latter view also implies a possibly distinct, endogenous source of the effects of selective attention on sensory processing (Kastner and Ungerleider, 2000, Kemner et al., 2004, Posner and Petersen, 1990).

Neurophysiological work in attention has investigated these implications by measuring when, where and how attention modulates neural response in the perceptual stream. Single-unit and neuroimaging studies have both found that selective attention can enhance visual neural activity, most commonly in extra-striate cortex (Maunsell and Cook, 2002, Treue and Maunsell, 1999), but also as early as V1 (Kastner and Ungerleider, 2000; Martínez et al., 1999). Neurophysiological research in auditory selective attention has been dominated by studies using non-invasive measures in humans. Electro- and magnetoencephalography (EEG, MEG) have been especially useful in this regard because their inherent temporal resolution is well suited to measuring the engagement of cognition (e.g. attention) on sensory events.

As in the psychological literature, two competing models of selective attention have emerged from these studies. In one view, Hillyard and colleagues have proposed an early-selection enhancement of the evoked auditory response (or event related potential, ERP) component occurring approximately 70–120 ms post-stimulus (the ‘N1 effect’, Hillyard et al., 1973, Näätänen and Picton, 1987, Woods, 1990). Though the N1 deflection in the ERP waveform may have several sub-components (Giard et al., 2000, Näätänen and Picton, 1987, Woods, 1990), its primary neural generator is believed to reside in auditory cortical fields along the supratemporal plane (Godey et al., 2001, Halgren et al., 1995, Liégeois-Chauvel et al., 1994, Näätänen, 1992, Picton et al., 1999, Woldorff et al., 1993, Yvert et al., 2001). This location, along with its early latency, tonotopic organization, and sensitivity to the state of the listener, has lead many to view the N1 as an exogenous (i.e. automatic) sensory response (see Näätänen, 1992 for a review).

An alternative position has been proposed by Näätänen and colleagues in which the effects of selective attention in auditory cortex arises primarily from an endogenous generator distinct from that producing the N1 (Näätänen, 1992). In this view, selective attention creates a generally broad negative component that adds to the N1 amplitude. The neural source of this negativity can purportedly be distinguished from that causing the N1 deflection mainly by the increased latency in the peak negative difference wave (Nd), formed by subtracting the ERP of the unattended signal from that of the attended signal. This negative enhancement has been labeled the ‘processing negativity’ (PN) and is viewed as the physiological index of an endogenous matching process occurring when subjects listen for targets in auditory input streams. Under conditions of high attentional load (e.g. signals with short inter-stimulus intervals (ISIs)), the peak PN latency may decrease, causing it to temporally overlap with the N1 in the ERP waveform and thus create an apparent N1 effect. The PN can also have a broader temporal effect such that later components like the P2, a positive deflection occurring on average 170–200 ms post-stimulus, may also exhibit a significant negative shift in attend relative to ignore conditions (Näätänen, 1992, Teder et al., 1993).

There is other evidence, however, to suggest that exogenous sensory components may indeed be directly modulated by selective attention. Woldorff and colleagues reported the attentional modulation of the positive mid-latency ERP deflection occurring around 20–50 ms post-stimulus (known alternately as P20–50 or P1), but only under conditions of high attentional load (e.g. ISIs of 200 ms) (Woldorff and Hillyard, 1991, Woldorff et al., 1993). These early attentional effects occur very near in time to the earliest evoked potentials recorded from human auditory cortex (Celesia, 1976, Howard et al., 2000), suggesting a potentially very early involvement of attention in sensory processing. In addition, the scalp distribution of some sub-components of the N1 effect can match that of the underlying N1 component itself, supporting the view of a common generator underlying the two phenomena (Giard et al., 2000, Woldorff et al., 1993, Woods et al., 1994). Nonetheless, a PN may still co-occur with attentional modulation of exogenous components due to its possibly distinct generator (Alho et al., 1994, Giard et al., 2000, Teder et al., 1993, Woods et al., 1991).

Patients undergoing evaluation for the surgical treatment of medically intractable epilepsy provide a unique opportunity to measure the effects of selective attention directly from human cortex. As part of their treatment, these patients at times require electrodes to be implanted either subdurally on the cortical surface or directly into the parenchyma of the brain in order to identify the locus of seizure activity and/or map cortical areas involved in speech and language. Electrophysiological recordings made from these implants are known as electrocorticograms (ECoG), and their increased signal-to-noise ratio and anatomical precision can help illuminate the sub-processes contributing to the scalp EEG.

Most ECoG studies have investigated aspects of auditory processing other than the potential influence of active attention on exogenous auditory response. This work includes general mapping of acoustically responsive cortex and its cortical connections (Brugge et al., 2003, Celesia, 1976, Halgren et al., 1995, Howard et al., 2000, Kropotov et al., 2000, Liégeois-Chauvel et al., 1994, Rosburg et al., 2004); mapping of responsiveness and discrimination of speech stimuli (Boatman et al., 1994, Crone et al., 2001, Liégeois-Chauvel et al., 1999, Matsumoto et al., 2004, Steinschneider et al., 1999); attention-independent ERP components such as the mismatch negativity (MMN) (Kropotov et al., 1995, Kropotov et al., 2000, Rosburg et al., 2005); and correlates of endogenous cognitive processing such as the late waveform component, P3/P300 (Halgren et al., 1995).

The goal of the current work is to investigate more explicitly the potential effects of selective auditory attention on human peri-Sylvian cortex as produced in the classic dichotic auditory oddball paradigm (Hillyard et al., 1973). It is hoped that results from this study will augment earlier depth electrode work which only indirectly considered the physiological effects of auditory attention (e.g. Halgren et al., 1995), and contribute to extant knowledge about how selective attention may modulate neural responses in auditory cortex.

It should be noted that the relationship between intracranial ERP components and their putative scalp counterparts is necessarily complex and can be difficult to ascertain (Nunez, personal communication, August 7, 2005). In the present work, no scalp EEG recordings to the experimental stimuli were available for any of the patients. However, even with such comparison recordings it may not always be theoretically possible to establish the desired isomorphisms between intra- and extra-cranial components. Such limits are primarily due to (1) electrical signal volume conduction across the scalp, and (2) the imposition by the high conductance skull of a strong low-pass filter on emanating electrical signals, in both the temporal and spatial domains (Srinivasan, 1999). One possible effect of these physical factors may be to allow a focal intracranial source with strong low temporal frequency content to impact scalp-recorded ERPs (e.g. on lower bandwidth components such as N1, P2, etc.), but simultaneously attenuate its higher spatial frequencies and thus obscure its cortical location.

Given the generally greater signal-to-noise ratio and spatial sensitivity of ECoG recordings, intracranial waveforms most likely represent only sub-components of scalp-recorded ERP components. It is also unclear, in light of the analysis complexities described earlier, whether such sub-components can be unequivocally isolated from scalp ERP component complexes even though they may contribute to the scalp ERP waveform. Hence, to distinguish intracranial sources from their possible scalp counterparts, we will identify the intracranial deflections under current investigation using a nomenclature of polarity (negative/positive), mean latency (in meter second), and anatomical location.