A matter of time: internal delays in binaural processing

As an animal navigates its surroundings, the sounds reaching its two ears change in waveform similarity (interaural correlation) and in time of arrival (interaural time difference, ITD). Humans are exquisitely sensitive to these binaural cues, and it is generally agreed that this sensitivity involves coincidence detectors and internal delays that compensate for external acoustic delays (ITDs). Recent data show an unexpected relationship between the tuning of a neuron to frequency and to ITD, leading to several proposals for sources of internal delay and the neural coding of interaural temporal cues. We review the alternatives, and argue that an understanding of binaural mechanisms requires consideration of sensitivity not only to ITDs, but also to interaural correlation.

Introduction

Spatial hearing offers a unique window on temporal processing in the nervous system. In contrast to the receptor organs for vision and touch, the cochlea does not have an explicit representation of the spatial position of sound sources, because this organ performs frequency analysis rather than spatial analysis. The spatial position of a sound source is computed in the CNS from implicit information sent downstream by the cochlea. It has long been known that the computation of azimuth (horizontal position of a source) is predominantly based on temporal differences between the two ears, but the underlying mechanisms are currently a matter of much controversy.

Sound sources off the midsagittal plane travel different distances to the two ears and thereby generate interaural time delays (ITDs), both in the arrival time of the stimulus wavefront (‘onset ITD’) and throughout the stimulus (‘ongoing ITD’) (Figure 1d). In humans, ongoing ITDs of low frequencies are the main source of information used to determine horizontal localization of sound 1, 2, 3. Even the largest ITDs, which occur for sound sources that face one ear, are tiny. Their extreme values (henceforth referred to as the ‘ecological range’) are ±700 μs in humans and ±400 μs in cats, but ITDs can be discriminated at values of 10–20 μs [4]. Considering that the duration of an action potential is ∼50 times longer, this acuity is an intriguing biological feat.

Neural sensitivity to ITDs was discovered in the 1960s 5, 6 in the midbrain inferior colliculus and brainstem medial superior olive (MSO), which have binaural neurons whose average firing rate depends on ITD (Figure 1a). Each neuron is tuned to a ‘best delay’ (BD), at which its response is maximal. Neurons differ in their BD, and are maximally excited by sound sources at correspondingly different positions in space. A general finding across studies is a clear bias for tuning to the contralateral hemifield: BDs are mostly at ‘positive’ ITDs, defined as ITDs at which the ear that is contralateral to the neuron is the first to receive the sound. For example, a sound source directly in front of a cat maximally excites neurons on both sides the brain that have a BD of 0 μs, whereas a source placed to the extreme right will excite neurons on the left (i.e. contralateral) side of the brain that have BDs near 400 μs. For each intermediary horizontal position between extreme right and the midline, there are neurons on the left side that are maximally excited.

These physiological observations, in combination with psychophysical work and an influential qualitative model [7], led to a general framework that seemed congruent with general neurobiological principles and that is commonly referred to as ‘the Jeffress model’. This model holds that populations of binaural neurons are tuned both to frequency and to ITD, and that there is a neural ‘display’ in which these neurons are arranged topographically in terms of the frequency by which they are maximally excited (best frequency, BF) and BD. Sound sources cause activity patterns on this BD–BF plane according to their spatial location and frequency characteristics.

In various incarnations, this general model has dominated the field [8] and is the basis of most models of binaural hearing, even though not all of its components are equally well established. However, new data have spawned alternative ideas, for which we here review the evidence. The existence of a BD–BF plane has been questioned, and there are several competing proposals for the physiological mechanisms that underlie the existence of BDs. Because these controversies mostly concern mammals, we do not cover the extensive work on binaural hearing in barn owls [9].