Neuroscience and Biobehavioral Reviews 36 (2012) 889–900
Contents lists available at SciVerse ScienceDirect
Neuroscience and Biobehavioral Reviews
journal homepage: www.elsevier.com/locate/neubiorev
Review
Stability of central binaural sound localization mechanisms in mammals, and the
Heffner hypothesis
Dennis P. Phillipsa,b
, Chelsea K. Quinlana
, Rachel N. Dinglea,∗
a
Department of Psychology, Dalhousie University, Halifax, Nova Scotia, Canada B3H 4J1
b
Department of Surgery, Dalhousie University, Halifax, Nova Scotia, Canada B3H 4J1
a r t i c l e i n f o
Article history:
Received 31 January 2011
Received in revised form
16 September 2011
Accepted 6 November 2011
Keywords:
Sound localization
Evolution of the auditory system
Binaural hearing
Mammals
Lower envelope principle
a b s t r a c t
Heffner (2004) provided an overview of data on the evolutionary pressures on sound localization acuity in
mammals. Her most important ﬁnding was that sound localization acuity was most strongly correlated
with width of ﬁeld of best vision. This correlation leaves unexplained the mechanism through which
evolutionary pressures affect localization acuity in different mammals. A review of the neurophysiology
of binaural sound localization cue coding, and the behavioural performance it supports, led us to two
hypotheses. First, there is little or no evidence that the neural mechanisms for coding binaural sound
location cues, or the dynamic range of the code, vary across mammals. Rather, the neural coding mechanism
is remarkably constant both across species, and within species across frequency. Second, there is
no need to postulate that evolutionary pressures are exerted on the cue coding mechanism itself. We
hypothesize instead that the evolutionary pressure may be on the organism’s ability to exploit a ‘lower
envelope principle’ (after Barlow, 1972).
© 2011 Elsevier Ltd. All rights reserved.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 889
2. Different mammals have highly similar mechanisms for the coding of binaural sound localization cues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 890
2.1. Interaural time differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 891
2.2. Interaural level differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 894
2.3. Departures from the general model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 896
3. A ‘lower envelope principle’ can explain the relation between Heffner’s evolutionary account and the central neurophysiological data . . . . . . 896
4. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 898
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 899
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 899
1. Introduction
Mammals vary widely in the acuity of their behavioural sound
localization performance, as measured by the smallest detectable
separation of sound sources symmetrically disposed around the
midline. Heffner’s (2004) thoughtful review on this topic provided
a collation of data on the midline acuity of mammals varying
widely in head size, hearing range, and ecological status. Her review
offered a developing hypothesis of the evolutionary pressures that
shape sound localization acuity across mammals. The key factors
∗ Corresponding author at: Department of Psychology, Dalhousie University, 1355
Oxford Street, PO Box 15000, Halifax, NS, Canada B3H 4R2. Tel.: +1 902 494 2383;
fax: +1 902 494 6585.
E-mail address: rdingle@dal.ca (R.N. Dingle).
examined were head size, limit of high-frequency hearing, and
width of the ﬁeld of best vision as inferred from retinal ganglion
cell density measurements (Heffner, 2004).
Historically, head size has been thought to matter for sound
localization acuity for two reasons. First, it sets an upper limit on
the range of interaural time disparities (ITDs) available to the animal.
Because sound travels at the same velocity to the two ears, the
interaural distance is important in determining the size of differences
in sound travel time to the two ears. ITDs come in two forms:
an arrival time disparity or onset time disparity, and for periodic
sounds, an interaural phase difference resulting from the imposition
of the interaural arrival time delay on all subsequent periods
of the stimulus waveform. In what follows, we use the term ITD to
mean interaural phase difference, unless otherwise speciﬁed. ITDs
are arguably most pertinent at low stimulus frequencies, for which
the period of the stimulus is long relative to the sound travel time
0149-7634/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.neubiorev.2011.11.003
890 D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900
between the ears; interaural phase difference becomes ambiguous
when it exceeds a half-period of the stimulus (Hartmann, 1999).
Second, head size determines the depth of the acoustic shadow cast
by the head (and pinnae), and therefore the magnitude of the interaural
level difference (ILD) available; in practice, shadows are cast
for sounds whose wavelengths are shorter than the head diameter
(or pinna size). Animals with large heads thus potentially beneﬁt
from the availability of a larger range of both ITDs and ILDs to use in
determining source azimuth. The limit of high-frequency hearing
matters because animals with small heads are potentially able to
capitalize on pinna directionality and head shadow effects if their
hearing range extends to frequencies high enough to engage those
acoustical factors.
Width of ﬁeld of best vision, quantiﬁed as the horizontal
width of the retinal region with the greatest ganglion cell density,
also matters. Heffner and Heffner (1992; see also Heffner, 2004)
hypothesized that one purpose of sound localization is to provide
information for visually orienting to the sound source; it follows
that the acuity of sound localization required to bring the source
into the region of most acute vision is inversely proportional to the
width of the ﬁeld of best vision (Heffner, 2004).
Perhaps the most surprising feature of Heffner’s work was an
analysis of the partial correlations between these three factors and
sound localization acuity (Heffner and Heffner, 1992). For what
follows, most important was the fact that when the correlation
between head size and acuity was partialed out, the correlation
between width of ﬁeld of best vision and sound localization acuity
remained strong, while the reverse was not true. Indeed, with the
visual factor partialed out, the correlation between head size and
sound localization acuity was rendered statistically insigniﬁcant
(Heffner and Heffner, 1992). The ﬁnding that sound localization
acuity was most strongly correlated with the width of ﬁeld of best
vision stands even after doubling the number of species studied
(Heffner et al., 2007).
Heffner was careful to emphasize that her account was of the
evolutionary pressures that determined sound localization acuity,
and not the mechanisms through which those pressures were
exerted. The purpose of this article is to address this lacuna. The
most obvious hypothesis linking the two ends of the correlation
(width of ﬁeld of best vision, and sound localization acuity) is that
possession of a visual fovea results in, or is associated with, a reﬁnement
in the precision of localization cue coding, i.e., a tailoring of
the accuracy of ILD and ITD coding to that required by the width of
ﬁeld of best vision. Surprisingly, our review of the neurophysiology
of binaural sound location cue coding ﬁnds no evidence supporting
that hypothesis. Indeed, we ﬁnd a remarkable stability of binaural
cue coding both across mammalian species, and across frequency
within species. This evidence is expressed in both neurophysiological
studies in animals, and psychophysical studies in man; it leads to
a conclusion that mammals as a biological class may have inherited
the same mechanisms for ITD coding and the same mechanisms for
ILD coding, and they show similar ranges of encoded cue sizes. As
such, we propose that evolutionary pressure is exerted not on the
cue coding itself, but on the ability of a species to exploit the precision
of coding by the ‘best’ neurons — an extension of Barlow’s
(1972) ‘lower envelope principle.’ An ability to base perceptual
judgements selectively on the activity of neurons whose responses
are the steepest functions of near-zero ITDs and ILDs may be the link
between Heffner and Heffner’s (1992) data on visual acuity and the
stability of ITD and ILD coding mechanisms across mammals seen
by others.
The review that follows is necessarily selective. Our focus is on
the possible neurophysiological intermediaries between the Heffners’
ﬁndings on the width of ﬁeld of best vision on the one hand, and
behavioural midline sound localization acuity on the other. Some
recent reviews on sound localization mechanisms in general and
their evolution in mammals and other classes can be found elsewhere
(Grothe, 2000; McAlpine and Grothe, 2003; Grothe et al.,
2010; King and Middlebrooks, 2011).
2. Different mammals have highly similar mechanisms for
the coding of binaural sound localization cues
Under normal (i.e., binaural hearing) circumstances, the two
main cues used to localize sounds are the ILD and ITD generated by
the spatial relationship between the sound source and the two ears.
ITDs increase from zero for sources at the midline, almost linearly
to azimuths near 90◦ in the lateral hemiﬁelds (Middlebrooks and
Green, 1990). Maximal ITDs depend on azimuth, head size, and to
some extent frequency, and can reach 750 ␮s in man (Middlebrooks
and Green, 1990) and 350 ␮s in cats (Roth et al., 1980). ILDs also
depend on source azimuth, head size, but more heavily than ITDs on
frequency. In cats, there is a signiﬁcant tendency for ILD magnitude
to saturate for sources deep in the lateral hemiﬁelds (Irvine, 1987).
In man, the patterns of ILD magnitude across azimuth and elevation
can show some marked individual differences (Middlebrooks
et al., 1989), but the general pattern is one of greater ILDs for sources
located more laterally, with some tendency for ILDs to be near maxima
over relatively broad ranges of lateral azimuths. The tendency
of ILDs to saturate in the lateral hemiﬁelds renders ILD amplitude
somewhat imprecise in specifying source azimuth when the source
is in the lateral hemiﬁelds, and, indeed, free-ﬁeld sound localization
is less accurate for lateral sources (Makous and Middlebrooks,
1990). Localization of sources in the lateral hemiﬁelds is likely
based on the dichotic cue information being supplemented by
other data, namely the distribution of energy across frequencies
(the ‘head-related transfer function’), which is potentially available
both monaurally and binaurally. In natural settings, this cue
information may be supplemented further by patterns of sound
reﬂection off environmental obstacles (Hartmann, 1999). The use
of monaural spectral cues is probably mediated by comparison of
the received source spectrum with some kind of internal template
or representation of the source spectrum itself or of the listener’s
own head-related transfer function. The temporal dynamics of this
processing are quite different to those of the ‘online’ analysis of
ITDs and ILDs (Hofman and Van Opstal, 1998; Ishigami and Phillips,
2008).
In man, behavioural minimal audible angles around the midline
supported by these mechanisms can be as small as 1◦ of auditory
azimuth (Mills, 1958). Human free-ﬁeld acuity is comparable to
that in elephants (about 1◦ of azimuth: Heffner and Heffner, 1982),
but most other species studied to date have midline acuities of 5◦ or
worse, and many species have acuities in the range from 10◦ to 40◦
(see Heffner et al., 2007). A subtle, but perhaps important, feature
of the Heffner et al. (2007) data is that the task required animals to
discriminate which of two sound sources symmetrically disposed
across the midline was activated. Technically, this is a lateralization
(as opposed to a localization) task because the animal is required
simply to differentiate left from right. As such, and on the assumption
that the stimulus information used in the discrimination is the
binaural ITD and/or ILD cue, the task reduces to the detection of a
change in sign of the ITD/ILD from favouring one ear to favouring
the other.
Under dichotic conditions, human subjects can detect ITDs as
small as 10–15 ␮s (Klump and Eady, 1956; Mills, 1958), and they
can detect ILDs as small as 1 dB (Mills, 1958, 1960; Grantham, 1984).
Unfortunately, there are few comparative studies of the sensitivity
of animals to ITDs and ILDs. The data available suggest that despite
differences in head size, there are similarities in threshold-level
ITDs and ILDs detectable across species. Dichotic studies show that
cats can discriminate ITDs as small as 15 ␮s and ILDs as small as 1 dB
D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900 891
(Cranford, 1979). Pig-tailed monkeys have slightly higher thresholds
(45 ␮s and 3.5 dB: Houben and Gourevitch, 1979). Japanese
macaque monkeys have ITD and ILD thresholds of about 26 ␮s and
1 dB, respectively (Boester, 1994). Cats and macaque monkeys have
similar widths of ﬁeld of best vision (about 5◦) while the human
width of ﬁeld of best vision is much smaller (less than 1◦; see
Heffner et al., 2007). Although preliminary, these data suggest that
there is little correlation between width of ﬁeld of best vision and
ITD/ILD acuity. Because these species (e.g., humans and cats) differ
in their free-ﬁeld midline sound localization acuity but not in their
ITD/ILD acuity, the data also suggest that the correlation between
ITD/ILD acuity and free-ﬁeld localization acuity may be weak. The
only hint that any mammal has greater acuity than those seen
in cats and humans comes from a free-ﬁeld localization study of
Jamaican fruit bats (Heffner et al., 2000) in which it was shown
that the bat’s ability to use ITD cues to distinguish left from right
speakers at 30◦ eccentricity extended to 6.3 kHz. This is remarkable
only in that 6.3 kHz is the highest frequency at which mammals
have ever been shown to use ITDs. A tone of 6.3 kHz has a period
of 160 ␮s and the Jamaican fruit bat’s interaural distance supports
ITDs of over 75 ␮s, so while the upper limit of ITD used in this
species may be high, it does not constitute evidence of behavioural
ITD acuity better than 15 ␮s. Interestingly, these species (humans,
cats, Jamaican fruit bats) also differ widely in their widths of ﬁeld
of best vision (about one, seven and 35◦, respectively: Heffner et al.,
2007). Thus, the available behavioural data suggest that variations
in head size and widths of ﬁeld of best vision are not accompanied
by variations in sensitivity to the binaural sound localization cues.
2.1. Interaural time differences
In the case of both ITDs and ILDs, neural sensitivity to the
stimulus cue information ultimately depends on a tracking of the
amplitude of cochlear output on the two sides. The cochlea’s transducers
are the inner hair cells (IHCs), and the cochlea’s output is
constituted by spiral ganglion cells, each of which contacts a single
IHC (Spoendlin, 1967). IHC depolarization is driven only by
upward motions of the basilar membrane. The probability of an
action potential being generated in a cochlear ganglion cell depends
on the magnitude of the IHC depolarization response. For low tone
frequencies, the stimulus period is so long that IHC depolarization
can track the amplitude of basilar membrane motion driven
by individual cycles of the stimulus; period histograms of auditory
nerve responses to low-frequency tones or tone combinations
show that the temporal distribution of spike discharges within the
stimulus period follows the amplitude of the positive-going portions
of the stimulus waveform quite precisely (‘phase-locking’:
Rose et al., 1967; Brugge et al., 1969a). That is, the temporal distribution
of spikes has an envelope that is akin to a half-wave
rectiﬁed version of the stimulus waveform (see below). For high
tone frequencies, the IHC depolarization also depends on the amplitude
of basilar membrane motion, but the IHC membrane response
is dominated by a direct current (‘pedestal’) depolarization component
(Russell and Sellick, 1978). This has the consequence that
IHC neurotransmitter release is more nearly continuous, the timing
of cochlear nerve cell action potentials is random with respect to
phase of basilar membrane motion (see Ruggero, 1992), and auditory
nerve responses, i.e., average ﬁring rate, follow the envelope
of the stimulus.
Neural comparisons of the cochlear outputs occur ﬁrst in the
superior olivary complex (SOC: Irvine, 1992), although it is embellished
at higher levels (Spitzer and Semple, 1998; Phillips, 2000;
Park et al., 2004). Indeed, de novo binaural interactions occur as
high as the auditory cortex (Kitzes et al., 1980). The SOC contains
a number of nuclei, the most important of which for the present
discussion are the medial (MSO) and lateral (LSO) superior olivary
nuclei. The inputs to both the MSO and LSO preserve the amplitudecoding
properties seen at cochlear output. The two nuclei, however,
have distinctly different tonotopic organizations in that the frequency
representation of the former is highly biased towards low
frequencies, while that of the latter is biased towards high frequencies
(Guinan et al., 1972). This offers MSO cells a specialized
opportunity to compare the instantaneous amplitudes of the two
cochlear outputs on a cycle-by-cycle basis for low-frequency stimulus
elements, and it affords LSO cells the opportunity to compare
the envelope amplitudes of high-frequency components of binaural
stimuli.
The MSO and LSO have unique cytoarchitectures and patterns
of afferent and efferent connectivity (Schwartz, 1992; Smith et al.,
1993). Irvine (1992) raises the question of whether the MSO and
LSO circuitries constitute separate, parallel pathways for the processing
of ITDs and ILDs respectively. He noted that species with
well-developed MSOs and poorly developed or absent LSOs may
nevertheless display good ITD discrimination at low frequencies
and ILD discrimination at high frequencies (see also below), suggesting
that ITD and ILD processing can be executed by different
neural populations within a single nucleus. It follows from this
that the anatomical separation of MSO/LSO circuitries seen in some
animals (e.g., cats) may not be paralleled in other species. An
interesting corollary of this point is that the functional distinction
between the MSO and LSO, which is usually cast in terms of ITD and
ILD coding, may in fact functionally be a distinction in frequency
representation, with which the coding of behaviourally-relevant
ITDs and ILDs of simple tones is correlated. On the other hand, as
we shall see below, in some species the poor development of one
or other nucleus does indeed appear to have behavioural correlates
for the processing of ITDs or ILDs.
The responses of MSO cells, and the neurons in higher centers
deriving their input from the MSO, are a sensitive function of the
relative phases of the signals at the two ears (Yin and Kuwada, 1983;
McAlpine et al., 2001; Brugge et al., 1970; Brugge and Merzenich,
1973; Spitzer and Semple, 1995, 1998; Hancock and Delgutte,
2004). The ﬁring rate of the binaural cells reﬂects the extent to
which phase-locked inputs from the two ears arrive in temporal
coincidence. For simple tones, this sensitivity manifests as a cyclical
relation of spike count to interaural delay, with the period of
the response cycle being equal to that of the tonal signal. This is
what one would expect from a coincidence-detection mechanism
operating on phase-locked inputs from the two sides.
At the level of the MSO, this coincidence detection is mediated
by interlaced, phase-locked excitatory/inhibitory inputs from the
two sides (see Grothe et al., 2010 for review; but see also Spitzer
and Semple, 1995). At higher levels, it can be mediated by a host of
different binaural interactions. For example, one can ﬁnd cells with
a phase-locked excitatory input from one ear, and an inhibitory one
from the other ear (Orman and Phillips, 1984); one can ﬁnd cells
receiving almost sub-(spike)-threshold excitatory inputs from the
two sides which sum to generate spike responses in the binaural
comparator only when a certain phase relation occurs in the stimuli
at the ears, and thus in the inputs to the binaural comparator
(Brugge and Merzenich, 1973); one can ﬁnd cells that receive interlaced
excitatory and inhibitory half-periods from one or both sides
(Brugge et al., 1970; Orman and Phillips, 1984). These all have in
common that stimulus amplitude-dependent responses from the
two sides are phase-locked because the long stimulus period permits
it, and this enables a moment-by-moment comparison of the
stimuli at the two ears within each period of the stimulus, executed
as a temporal coincidence detection. The diversity of binaural
interactions mediating the interaural correlation likely reﬂects the
convergence of inputs onto higher neurons, including those that
support de novo binaural interactions (e.g., van Adel et al., 1999;
Kitzes et al., 1980).
892 D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900
Fig. 1. Schematic diagram illustrating some general features of ITD coding. A shows an idealized spike count versus ITD function for one neuron tested at a single frequency.
B shows idealized normalized ITD functions for a single neuron tested separately at four frequencies. In this instance, the response functions coincide at a single ITD, which
is termed the ‘characteristic delay’ (CD) for that neuron. C shows an idealized average of the functions obtained when many frequencies are tested, each weighted by the
absolute ﬁring rates characterizing the function; it represents the composite delay function for that neuron. The ﬁlled circle depicts the CD for this neuron. D shows a
schematic representation of the distribution of best delays typically encountered across a population of neurons. The distribution of characteristic ‘peak’ delays (of the kind
schematically illustrated in B) would look very similar. More detailed description in text.
Central auditory neurons, of course, do not respond only to a
single frequency, but to a range of frequencies described by the
tuning curve (the envelope of best sensitivities across the neuron’s
limited effective frequency range). For each frequency within the
tuning curve, low-frequency cells can be tested for their sensitivity
to ITDs. An important concept to emerge from such studies is
that of the ‘composite delay function’ which is the average of the
ITD functions seen at all the frequencies tested, with the function
at each frequency weighted by the ﬁring rate characterizing it (Yin
and Kuwada, 1983). Fig. 1 schematically illustrates how composite
delay functions are obtained. Fig. 1A shows an idealized ITD function
for a single neuron tested at its preferred frequency. Fig. 1B
shows idealized and normalized ITD functions for the same neuron,
this time tested at four frequencies. Note that the functions coincide
(in this instance) at the peak of the functions, although in practice,
that coincidence can occur at any point on the functions. The ITD
associated with that coincidence is termed the ‘characteristic delay’
(CD) for that neuron. A composite delay function is computed by
taking the average of the ITD functions, with each function being
weighted by the absolute ﬁring rates that characterize it. Seven
frequencies were used to generate the idealized composite delay
function shown in Fig. 1C. Note that it preserves the CD. If ITD
functions are obtained for a sufﬁcient range of frequencies, then
the composite delay function is exceedingly close in form to that
obtained with a noise stimulus whose bandwidth spans the effective
frequency range of the neuron (Yin et al., 1986). Sometimes, the
peak of the composite function is termed the ‘best delay’ (McAlpine
et al., 2001; Hancock and Delgutte, 2004) to distinguish it from the
characteristic delay (ITD at which delay functions for tones of different
frequencies coincide). Composite delay functions typically,
though not always, have peaks favouring the contralateral ear (i.e.,
contralateral stimulus phase precedes the ipsilateral phase). Fig. 1D
shows an idealized distribution of best delays; it is based on the data
of Hancock and Delgutte (2004) and McAlpine et al. (2001). The bias
in best delays towards ones favouring the contralateral ear has the
result the steeply declining portion of the functions is often centered
over zero ITD (compare: McAlpine et al., 2001; Brand et al.,
2002; Hancock and Delgutte, 2004). The peaks and troughs of such
functions are often outside the behaviourally relevant range of ITDs,
suggesting that it is the steep portion of the function straddling
zero ITD, rather than the peak (or trough), which is the informative
part of the function (see Phillips and Brugge, 1985; Spitzer and
Semple, 1995; McAlpine et al., 2001; Brand et al., 2002; Hancock
and Delgutte, 2004).
Evidence has recently been presented in guinea pigs (McAlpine
et al., 2001), gerbils (Brand et al., 2002) and cats (Hancock and
Delgutte, 2004) that best delays are inversely correlated with
cell’s preferred tone frequencies. Cells with low preferred frequencies
have large best ITDs, and cells with higher preferred
frequencies have small best ITDs. The relationship is such that the
best interaural phase (as opposed to best ITD in units of time) is
nearly constant (at about 45◦) across cells with different preferred
frequencies. The ﬁnding has the corollary that the steep portion
of the composite ITD function is centered close to zero ITD. The
details of the neurophysiological mechanisms supporting the
relationship between best frequency and best ITD are still being
worked out, but it clearly involves precisely-timed inhibition at
the level of the inputs to the MSO (Brand et al., 2002; Pecka et al.,
2008; see also Grothe et al., 2010). These ﬁndings may require
further conﬁrmation because the relationships between best
frequency, best ITD and best interaural phase were not reported in
earlier detailed studies of ITD coding (e.g., Yin and Kuwada, 1983;
Spitzer and Semple, 1995). These new observations are, however,
interesting and important for three reasons. One is that the three
D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900 893
species in which the relationships have been reported vary widely
in head size and width of ﬁeld of best vision. The second is that
because the preferred interaural phase difference is constant across
the (low-frequency) tonotopic array, the mammalian ITD coding
mechanisms may have evolved to encode interaural phase rather
than interaural time. The third is that because best ITDs are dependent
on neural best frequency, any model of sound localization
mechanisms cannot be based on an ‘identity code,’ i.e., the lateral
position of a sound source cannot be encoded by which particular
neurons are discharging maximally.
These and other data led to the hypothesis that the lateral locus
of a sound is encoded by the relative activation of two ‘channels’ of
neurons: those activated maximally by sources with ITDs favouring
the contralateral and ipsilateral sides respectively, with steep borders
straddling the midline, i.e., zero ITD (see: Boehnke and Phillips,
1999; McAlpine et al., 2001; Phillips and Hall, 2005; Phillips, 2008;
Dingle et al., 2010). Thus, as the stimulus ITD shifts from favouring
one side to favouring the other, there is a maximal differentiation of
response strengths in the two populations; for large ITDs favouring
one side, there are less differentiated responses with comparable
changes in ITD. This account parallels human behavioural
sensitivity, in which azimuthal allocations of a source are steep
functions of ITD for near-zero values of ITD, with a ‘saturation’ of
azimuthal assignments for large ITDs (Yost, 1981). The two-channel
model has received recent support from human evoked response
studies (Magezi and Krumbholz, 2010) and from human magnetoencephalography
studies (Salminen et al., 2009, 2010).
The two-channel model has not been without challenge.
Harper and McAlpine (2004) modeled optimal coding strategies
for ITDs. They concluded that for small-headed animals and/or
low-frequency sounds, the two-channel model was optimal. For
large-headed animals and/or higher frequencies, a distributed
coding system based on a population of neurons with a homogeneous
distribution of ITD tunings was optimal. For human beings,
the two-channel model was deemed optimal for stimulus frequencies
below about 400 Hz, and a distributed population was
deemed optimal for higher frequencies. Nevertheless, empirical
psychophysical evidence in man supports the two-channel model
at frequencies below and above 400 Hz (Phillips and Hall, 2005;
Vigneault-MacLean et al., 2007). Magezi and Krumbholz’s (2010)
human electrophysiological study of responses to ‘inward’ and ‘outward’
changes in ITD employed noise stimuli, and also supported
the two-channel model. The use of noise stimuli makes it difﬁcult to
infer exactly which frequency-speciﬁc channels of processing contributed
to the responses, but it is perhaps reasonable to assume
that it was all of those that supported coding of the ITD stimuli.
Salminen et al.’s (2009, 2010) magnetoencephalographic studies in
man also produced data favouring a two-channel mechanism. Once
again, recall that human beings and the smaller mammals (guinea
pigs, gerbils, cats) whose ITD coding mechanisms were used as a
neurophysiological basis for the two-channel model have vastly
different widths of ﬁeld of best vision.
A second challenge to the two-channel model is the assertion
that it requires integration of information across the two cerebral
hemispheres (Joris and Yin, 2007). The challenge is based on
three facts. One is that each cerebral hemisphere is dominated by
cells encoding spatial information for the contralateral auditory
hemiﬁeld (e.g., Phillips and Brugge, 1985). The second is that animals
with unilateral lesions rostral to the LSO/MSO display sound
localization deﬁcits only for sources in the auditory hemiﬁeld contralateral
to the ablation (Jenkins and Masterton, 1982; Jenkins and
Merzenich, 1984; Heffner, 1997). The third is that experimental
lesion of the forebrain commissural pathways in animals (Moore
et al., 1974) or congenital agenesis of the corpus callosum (or early
callosotomy) in man (Lessard et al., 2002) has only minor consequences
for sound localization acuity. What this assertion misses,
however, is that each side of the auditory forebrain contains populations
of neurons that encode ipsilateral auditory space as well as
the more numerous cells that encode contralateral auditory space
(e.g., Stecker et al., 2005). The relative activity comparison required
by the two-channel model can thus be mediated within a single
cerebral hemisphere contralateral to the sound source (Dingle et al.,
2010).
A commonality in the neurophysiological literature is the mismatch
between the size of ITDs generated by the animal’s head size,
and the range of ITDs encoded by their central auditory systems. The
latter is typically much wider in range than is the former, especially
for small-headed animals (see Yin and Kuwada, 1983 [cats, with a
relatively small fovea]; Phillips and Brugge, 1985; Kelly and Phillips,
1991 [rats, with a relatively broad width of best vision, and studied
with click stimuli]; McAlpine et al., 2001 [guinea pigs]; Spitzer
and Semple, 1995; Brand et al., 2002 [gerbils, with an exceptionally
broad width of ﬁeld of best vision]). Indeed, although somewhat
anecdotal, the data suggest that the range of ITDs encoded by central
mechanisms is relatively constant across mammals. Thus, the
distribution of preferred (and characteristic) delays in cats (Yin and
Kuwada, 1983; Hancock and Delgutte, 2004), guinea pigs (McAlpine
et al., 2001) and gerbils (Spitzer and Semple, 1995) are extraordinarily
similar, despite those species having very different head
sizes and widths of ﬁeld of best vision. In all three species, the distribution
of best delays is highly peaked, centered near 250–300 ␮s
favouring the contralateral ear; most cells have best delays between
−100 and +400 ␮s, although the skirts of the distributions extend
from about −500 ␮s to +1000 ␮s (contralateral stimulus time re:
ipsilateral).
The mismatch between the available range of ITDs and those
encoded by central neurons in small-headed animals has been
apparent for many years, and led to the question of whether maximal
responses to ITDs outside the behaviourally relevant range
served a purpose other than direct coding of the cue for source
azimuth (McFadden, 1973). It could be argued, indeed, that once it
has been illustrated that the range of encoded ITDs across mammals
is relatively constant, the mismatch is to be expected: in a
two-channel model, the existence of neurons with best ITDs outside
the behaviourally relevant range is a simple consequence of
animal head size and audible frequency range. The alternative possibility,
apparently refuted by the empirical evidence, is that the
range of encoded ITDs could have covaried with head size. It is
the relatively constant range of encoded ITDs across species with
greatly different head sizes that ﬁrst prompted a hypothesis that
mammals as a class have a common set of mechanisms for ITD
coding, and one which serves larger headed animals better than
small headed ones (Phillips and Brugge, 1985; Kelly and Phillips,
1991). That is, large-headed animals have the advantage of being
able to exploit a greater portion of the response-vs-ITD dynamic
range simply because they have a larger range of ITDs available to
them, and this supports greater sound localization acuity. Indeed
one sees a modest correlation between head size and sound localization
acuity in earlier comparative studies (Heffner and Heffner,
1982). It was only later that midline sound localization acuity was
found to be independent of head size when width of ﬁeld of best
vision is partialed out (Heffner and Heffner, 1992).
The further question concerns the stability or generality of ITD
coding mechanisms across the frequency domain within species.
Historically, ITD coding at behavioural and neurophysiological
levels has been studied for low-frequency sounds, presumably
because low-frequency sounds are the most obvious instances
in which ITDs are behaviourally relevant. In practice, however,
even sounds with purely high-frequency spectra can contain lowfrequency
periodicities, and so it is an open question as to whether
such sounds are subject to ITD coding using the same mechanisms
as low-frequency tones.
894 D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900
Fig. 2. Schematic diagram illustrating stimuli used in studies of ITD coding, and the average temporal distribution of spikes evoked by them in the cochlear nerve. A shows
the case for a simple low-frequency tone. The upper panel shows the time waveform of the stimulus, and the lower panel shows the distribution of spikes, whose form
follows from the coupling of spike discharge selectively to upward motions of the basilar membrane. B shows the case for sinusoidally amplitude-modulated tones. The
uppermost panel shows the time waveform of the modulator, the next panel shows the waveform of the carrier tone, the third panel shows the waveform of the result of
modulating the latter by the former, and the bottom panel shows the expected distribution of spikes driven by the stimulus. Note that the neural response contains no ‘half
periods’ of inactivity of the kind seen in response to simple tones. C shows the case for transposed tones. In this instance, the low-frequency modulator (uppermost panel) is
half-wave rectiﬁed. Multiplication of the modulator by the carrier tone (next panel) creates a stimulus (next lower panel) with periods of silence between the tone bursts. In
turn, the temporal distribution of evoked spikes also contains ‘half periods’ of inactivity, and thus resembles that seen with a simple low-frequency tone, even though there
is no low-frequency energy in the transposed tone stimulus.
Perhaps the simplest high-frequency sounds with lowfrequency
periodicities are sinusoidally amplitude-modulated
tones. Henning (1974) was early to note that in man, such sounds
could be lateralized on the basis of an ITD of the modulating waveform
quite well. It has since become clear that central auditory
neurons encode the ITDs of the low-frequency envelopes of sinusoidally
modulated tones using binaural interaction mechanisms
very similar to those used for simple low-frequency tones (Yin et al.,
1984; Joris and Yin, 1995).
Low-frequency tones evoke auditory nerve responses whose
temporal distribution resembles a half-wave rectiﬁed version of
the stimulus waveform (see Fig. 2A). Sinusoidally amplitudemodulated
tones do not, because multiplication of a high-frequency
carrier by a low-frequency modulator results in a stimulus more
likely to afford a full-wave rectiﬁed distribution of auditory nerve
action potentials (at half the modulator frequency). This point is
illustrated schematically in Fig. 2B. When a high-frequency carrier
is multiplied by a low-frequency modulator, the resultant waveform
consists of sinusoidally-modulated bursts of the carrier, with
a zero-duration silent period between them. The expected average
temporal distribution of cochlear nerve spikes driven by such
a stimulus also has effectively a zero-duration period of silence
between responses.
Bernstein and Trahiotis (2002) used what are termed ‘transposed
stimuli’ in which a low-frequency ﬁltered, half-wave
rectiﬁed modulator tone was multiplied by a high-frequency
carrier. A transposed stimulus affords the auditory nerve an opportunity
to phase-lock to the modulated waveform in a fashion much
more akin to that seen in response to low-frequency tones. We
illustrate this point in Fig. 2C. When a half-wave rectiﬁed, lowfrequency
modulator is multiplied by a high-frequency carrier, the
resultant waveform is a series of high-frequency tone bursts with
equal-duration silent periods between them. The temporal distribution
of phase-locked spikes driven by this stimulus would be
expected to resemble much more closely that seen with the simple
low-frequency tone. Behavioural sensitivity to ITDs of the transposed
stimuli was typically better than that seen in response to
sinusoidally amplitude-modulated tones, and at least equal to that
seen in response to low-frequency tones of the modulator frequency
(Bernstein and Trahiotis, 2002). Still more recently, it has
been revealed that neural sensitivity to ITDs of transposed stimuli
follows the same pattern, and led to the conclusion that ‘the neural
mechanisms that mediate sensitivity to ITDs at high and low
frequencies are functionally equivalent’ (Grifﬁn et al., 2005). These
data suggest that the neural mechanisms of ITD coding are not only
stable across mammalian species, but also across frequency within
mammalian species.
2.2. Interaural level differences
The neural coding of ILDs is based on a comparison of leveldependent
responses from the two ears. It has been studied almost
exclusively in central neurons tuned to high tone frequencies, presumably
because it is high tone frequencies for which the head and
pinnae cast signiﬁcant acoustic shadows and thus provide signiﬁcant
(i.e., usable, behaviourally-relevant) ILDs. As mentioned above,
the lateral superior olive (LSO) has a tonotopic organization biased
towards high frequencies and is the ﬁrst major neural site of convergence
of stimulus level information from the two ears (Boudreau
and Tsuchitani, 1968). The most common pattern of binaural input
mediating this comparison is excitation from one ear and inhibition
from the other. Because both of those inputs are level-dependent
(and probably saturating), the output of the binaural comparator is
most commonly a sigmoidal function of ILD, with the steep portion
of the function centered near zero dB ILD. Rostral to the crossed
outputs of the LSOs, it is usually, though not always, the case that
response rates are at ceiling for ILDs signiﬁcantly favouring the contralateral
ear, and at minima for ILDs favouring the ipsilateral ear
(Phillips and Brugge, 1985; Park et al., 2004). Neurons vary in the
slopes of their spike rate-vs-ILD functions, but their ILD dynamic
D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900 895
Fig. 3. Normalized spike count vs ILD functions for primary auditory cortical neurons
in the cat. Each curve is a sigmoidal function of ILD, although they have
somewhat variable slopes and dispositions along the ILD axis. Nevertheless, all are
centered over small ILDs which themselves are associated with spatial locations
close to the midline. Data are all from the study by Phillips (1980).
ranges are usually less than about 25–30 dB wide. This sigmoidal
pattern of responsivity is seen throughout the auditory pathway
(Brugge et al., 1969b, 1970; Brugge and Merzenich, 1973; Phillips
and Irvine, 1981; Orman and Phillips, 1984).
The more rostral central auditory centers contain cells that
encode ILDs using a range of other binaural interactions. These
include nearly subthreshold excitatory inputs from the two sides
which sum to evoke spikes only for ILDs near zero dB (Kitzes et al.,
1980; Phillips and Irvine, 1981; Orman and Phillips, 1984), and cells
with a short-latency excitatory and longer-latency inhibitory input
from one side and an inhibitory input from the other (Phillips and
Irvine, 1981). Note that these are the same patterns of binaural
input mediating sensitivity to ITDs (above). Interestingly, at the
most rostral levels of the auditory system, neural responses are
quite often dominated by an onset transient (Phillips et al., 2002),
and one might construe response strength to binaural stimuli as
dependent on the degree of temporal synchrony with which inputs
from the two sides arrive, i.e., coincidence detection (e.g., Kitzes
et al., 1980).
Fig. 3 shows normalized ILD functions for a population of neurons
studied in the primary auditory cortex of cats (based on data
from Phillips, 1980). The functions are at ceiling for ILDs significantly
favouring the contralateral ear, and at minima for ILDs
favouring the ipsilateral ear. The functions have somewhat variable
slopes, and are disposed across somewhat variable ILD ranges.
Individually, neurons usually have ILD dynamic ranges of 10–20 dB
(though sometimes much wider). Collectively, the population code
spans about 10–15 dB on each side of zero ILD (i.e., a total width
of 25–30 dB). Like ITD functions, the steep parts of the ILD functions
are most often concentrated around relatively small ILDs, i.e.,
those that would be generated by sources located near the midline
(see Park et al., 2004). Recall that ILDs are a most precise indicator
of source azimuth for sources close to the midline. Comparable
data have been presented for the rat (Kelly et al., 1998) which has a
much wider ﬁeld of best vision (Heffner et al., 2007). It was the conjunction
of ﬁndings on the azimuth-ITD/ILD relationship and the
ITD/ILD-spike rate relationship that prompted Phillips and Brugge
(1985) to suggest that auditory forebrain neurons are most sensitive
to cues for source azimuth over cue ranges that themselves
most precisely specify source azimuth.
Once again, these data, and data from free-ﬁeld studies of sound
azimuth coding (see especially, Stecker et al., 2005) have led to the
hypothesis of a two (Boehnke and Phillips, 1999; Phillips and Hall,
2005; Phillips, 2008) or three (Dingle et al., 2010) channel model
of sound azimuth coding. The argument is that perceived sound
azimuth is speciﬁed by the relative outputs of two (left, right) or
three (left, right, midline) neural channels, much as in the case of
ITD coding. The outputs of the neural channels are most differentiated
for ILDs near zero (and therefore for source azimuths near the
midline), and this contributes to behavioural acuity being greatest
for near-midline source locations and poorer for sources in the
lateral hemiﬁelds.
The hypothesis that common sets of neural coding mechanisms
characterizes sound localization behaviour across the
mammalian frequency range arguably requires that ILD sensitivity
be manifested for low-frequency stimuli, even in the event
that low-frequency stimuli do not generate signiﬁcant ILDs. This
is indeed the case. First, the human scalp-recorded frequency
following response (an index of the phase-locked resposes to
low frequency stimuli) is sensitive to ILD over a range of ILDs
that matches the dynamic ranges of central neurons (Krishnan
and McDaniel, 1998). Second, human scalp-recorded steady state
responses to periodic reversals of the interaural phase of a noise
stimulus are sensitive to ILDs over roughly the same 25 dB range
(Massoud et al., 2011). This particular response to ILDs can be mediated
only by low-frequency neurons because only low-frequency
neurons will be sensitive to phase reversals of a noise carrier. Third,
not only do humans have ILD detection thresholds of less than 1 dB
at frequencies well under 1 kHz (Mills, 1960), but they show a mapping
of ILD to intracranial auditory azimuth at low frequencies as
they do for high frequencies (Yost, 1981). Fourth, neurophysiological
studies in cats show that low-frequency central neurons are
directly sensitive to ILD, and the ITD sensitivity of low-frequency
neurons is dramatically affected by the imposition of an ILD, in the
form of ILD-driven modulations of the depth of the ITD function, or
a time (latency)-intensity trading (Yin and Chan, 1990; Tollin and
Yin, 2005).
The foregoing data render it unequivocal that the neural ILD
sensitivity seen in response to high-frequency stimuli extends to
low-frequency sounds. We had previously seen that the neural sensitivity
to ITDs, typically studied for low-frequency sounds, extends
to the amplitude envelopes of high frequency carriers, i.e., highfrequency
cells can possess the same ITD coding mechanisms seen
in responses to single tones by low-frequency cells. Moreover, at
their root, both ITD and ILD coding depend on a comparison of
basilar membrane motion amplitude on the two sides; for low frequency
stimulus elements, the periods are sufﬁciently long that
cochlear output can track the phase of basilar membrane motion,
and this in turn permits central neurons to compare the phase of
the stimulus at the two ears in a process that can be described
as a temporal coincidence detection. Next, the ITD and ILD coding
mechanisms are phenomenologically similar across species of
different head sizes (and widths of ﬁeld of best vision), and the
data available suggest that there are not major differences in the
dynamic ranges of the ITD or ILD cues the species encode. Finally,
behavioural data in man (Klump and Eady, 1956; Grantham, 1984)
and in cats (Cranford, 1979) which have a much smaller head size,
indicate that the smallest detectable ITDs and ILDs are close to
10–15 ␮s and 1 dB, respectively. Taken together, these data suggest
that the mammalian biological class has common sets of mechanisms
for the encoding of ITDs and ILDs.
The foregoing might be interpreted as a challenge to the duplex
theory of sound localization. The duplex theory is based on freeﬁeld
studies of human minimal audible angles revealed in studies
using pure tones (see Moore, 2003 for review). It argues that we
use ITDs at low frequencies and ILDs at high frequencies. Free ﬁeld
studies (e.g., Mills, 1958) show that humans have small errors in
midline localization at low frequencies and at high ones, but not
at intermediate ones (around 2 kHz). The reasons for this are (1),
the intermediate frequencies are above those for which neural
coding of ITDs of simple tones is possible, and (2), the intermediate
896 D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900
frequencies are ones for which the head does not cast a signiﬁcant
acoustic shadow, so no ILDs are available to use to perform the
task. Nothing in our argument disputes this. The fact that we can
contrive high-frequency stimuli (e.g., transposed tones) to reveal
high-frequency sensitivity to ITDs does not dispute that the nervous
system is insensitive to ITDs of simple high-frequency tones.
The fact that the neural machinery for processing of low-frequency
stimuli is sensitive to ILDs does not dispute that ILDs do not exist for
low frequency sounds. Our point is only that the neural machinery
serving the processing of all frequencies is capable of encoding
ITDs and ILDs if the stimulus design offers it opportunity to do so.
2.3. Departures from the general model
There is behavioural evidence, however, of exceptions to this
general rule. In particular, hedgehogs (Masterton et al., 1975) and
some bats (e.g., Koay et al., 1998) have extraordinarily poor abilities
to use interaural phase difference cues to execute free-ﬁeld
sound localization. In addition, horses (Heffner and Heffner, 1986)
and pigs (Heffner and Heffner, 1989) are very poor in their ability
to use ILDs to localize free-ﬁeld sources. In each of these cases, animals
were tested for their ability to identify which of two sound
sources, located symmetrically about the frontal midline, was activated.
Tonal stimuli were used, and the frequencies were selected
to explore the ability of the animals to use ITD or ILD cues, by covering
the frequency range over which ITD and ILD cues were predicted
to be useful (or not) based on the animal’s head size.
For all of these species, sound localization performance could be
trained to excellence over the part of the frequency range expected
to support one or other of ITD or ILD coding, but was extremely
poor (performance at chance level) over the other. This suggests a
wholesale loss of the use of one or other cue. Because of the withinsubject
design, it is difﬁcult to make case that the animals were
simply more motivated to respond to one cue than to the other.
In the case of at least some species (hedgehogs, horses), the poor
behavioural performance over the affected frequency range was
paralleled by a poorly developed (or absent) MSO or LSO. This was
not so clear in the case of others (e.g., pigs), although it is possible
that anatomical defects, if present, were simply too subtle to be
detected by the methods used.
For the purposes of the present article, the important point is
that these species’ deviations from the general pattern are not so
much quantitative adjustments of a single mechanism or model,
but, apparently, the wholesale loss of one part of it. This is important
because it prompts the view that binaural interaction mechanisms
for ITD and ILD coding are already as optimal as the mammalian
nervous system can provide, so that the only deviations from that
optimum are impoverishments; these appear to take the form of
losses of MSO or LSO (or frequency-speciﬁc) function. At the other
end of the spectrum, extraordinary advances in sound localization
behaviour appear to have been achieved by the addition of a wholly
new mechanism, e.g., echolocation (Suga, 1982). To date, there is
no compelling evidence that the central coding of passive sound
localization cues in mammals in some way adjusts to species head
size or width of ﬁeld of best vision.
3. A ‘lower envelope principle’ can explain the relation
between Heffner’s evolutionary account and the central
neurophysiological data
The obvious question that emerges from the foregoing concerns
the nature of any variations in central neural coding mechanisms
that might explain the relationship between sound localization
acuity on the one hand, and width of ﬁeld of best vision on the
other. The foregoing review offered no evidence of specializations
of sound localization circuitry or range of cue sizes encoded in
animals with vastly different widths of ﬁeld of best vision and/or
signiﬁcantly different head sizes. Appeals to the possibility that
Heffner (2004) failed to identify all of the relevant evolutionary
pressures on midline sound localization acuity do not help because
such appeals still leave unanswered the identity of the mediating
auditory neural factor(s) that link an evolutionary pressure with
behavioural acuity, Our working hypothesis is that one does not
need to postulate covariation of behavioural sound localization
acuity and ITD/ILD coding mechanisms. Rather, there is a continuous,
if limited, distribution of the range of ITDs and ILDs encoded,
and in the precision with which central neural responses specify
the stimulus disparity size, and thus source azimuth. Evolutionary
differences in acuity may reﬂect adjustments not of that encoded
range or coding precision, but in the extent to which the animal is
selectively able to base behavioural performance on the subset of
neurons with the most informative code of small cue values.
This hypothesis is fundamentally similar to Barlow’s (1972)
‘lower envelope principle.’ In its simplest formulation, this states
that the detectability of an auditory event follows the envelope of
the sensitivities of the neurons with the greatest sensitivities to
that event. Consider the behavioural audiogram. Cochlear neurons
serving any given position along the cochlea have varying absolute
sensitivities (Ruggero, 1992; see his Fig. 2.4). The behavioural
audiogram, which depicts the maximum sensitivity of the animal
across the frequency domain, has a shape that reﬂects the thresholds
(‘follows the envelope’) of the most sensitive neurons across
the frequency range (e.g., Ruggero, 1992; see his Fig. 2.2). It is a
small step to extend this principle from detectability of an event
per se to discriminability of an event change. In this case, the lower
envelope principle now makes the general argument that the lowest
psychophysical threshold in discrimination reﬂects the activity
of the subset of neurons whose own ‘acuity’ is also the best (see
also Eggermont, 1998). At suprathreshold cue values, the population
response may well be a vector sum of the outputs of the contributing
neurons (Eggermont, 1998; Eggermont and Mossop, 1998). That
vector sum population response is the neural instantiation of the
two or three neural ‘channels’ whose outputs are compared in
lateralization judgments in the perceptual architecture proposed
by others (Phillips and Hall, 2005; Stecker et al., 2005; VigneaultMacLean
et al., 2007; Phillips, 2008; Dingle et al., 2010). Heffner’s
measurements of sound localization acuity are by design efforts
to measure the smallest discriminable (i.e., threshold) change in
source location around the midline. This might be construed as
equivalent to measuring the smallest ITD/ILD for which a change
in sign (from left-favouring to right-favouring) is detectable. The
neurons in the two channels that are best capable of indicating
the change in location are those with the steepest spike rate functions
for midline azimuths (or the cues for them). Our hypothesis is
that in the absence of evidence that the neural coding of sound
localization cues varies across species, evolutionary forces may
operate instead on the selectivity with which the animal can access
the ‘best’ neurons. Heffner’s argument (Heffner, 2004; Heffner and
Heffner, 1992; Heffner et al., 2007) is that the force shaping sound
localization acuity is the width of ﬁeld of best vision. Thus, animals
with broad visual streaks require only relatively poor sound
localization acuity to bring that region of the retinas in line with
the auditory target; such animals are under only weak pressure to
selectively access the ‘best’ auditory neurons. In contrast, animals
with small foveas require highly accurate information about sound
source location to align the foveas with the target; these species are
under greater pressure to access the outputs of the ‘best’ auditory
neurons. In a sound localization threshold study, this means that
the behavioural performance must be based on the subset of neurons
with the lowest neural thresholds for discriminating source
position. Eggermont (1999) has previously appealed to a lower
D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900 897
Fig. 4. Schematic illustration showing how Barlow’s (1972) lower envelope principle might be expressed in the ILD stimulus domain. A shows idealized spike rate-vs-ILD
functions for neurons with different ILD dynamic ranges. B shows the same data, this time expressed as the instantaneous slope of the ILD function, measured over a very
small ILD range. Sensitivity to ILD change (or detection of the presence of an ILD at all) reﬂects the inverse of the functions in B, shown in C as the dashed line. A listener
whose perceptual judgement is based selectively on the responses of the most sensitive neurons will have the lowest threshold for ILD detection. Any other weighting of the
outputs of the neurons will result in a poorer (higher) threshold for ILD detection.
envelope principle to link human auditory temporal gap detection
thresholds with the responses of cortical neurons studied with gap
detection stimuli.
We illustrate our argument schematically in Fig. 4. Fig. 4A shows
idealized curves intended to represent the spike rate-vs-ILD (or, in
principle, ITD) functions of neurons with different dynamic ranges,
and for the purpose of simplicity, all are centered on 0 dB ILD.
These are the responses of neurons contributing to one of the
two channels in the two-channel model. Some neurons have narrow,
steep functions, while others have broader encoded ranges
and thus shallower slopes in their response functions. The ability
of a neuron to respond differentially to slightly different ILDs
is expressed as the instantaneous slope of the function for the
stimulus range over which the small ILD change occurs. For the
functions shown in Fig. 4A, these slopes are shown in Fig. 4B. As
one would expect, the neurons with the greatest change in ﬁring
rate for a given change in ILD are those whose spike count-vsILD
functions are the steepest. The best sensitivity of the organism
to a small change in ILD is the envelope drawn around the functions
of Fig. 4B when they are inverted (dotted line in Fig. 4C).
In the two-channel account, this is of beneﬁt because as stimulus
azimuth around the midline alters, the rate of change of difference
in channel spike output will be greatest for precisely those neurons.
The listener whose perceptual judgments are based selectively on
the activity of the neurons with the greatest ILD discriminating
power will have the lowest threshold for ILD change detection, and
by extension, the smallest behavioural minimum audible angle.
Any other weighting of the outputs of the ILD-sensitive neurons
results in a poorer ILD change-detection threshold, and by extension,
a larger behavioural minimum audible angle. Parenthetically,
the danger in this argument lies in the assumption that it is only
sensitivity to binaural cues that mediates free-ﬁeld sound lateralization
acuity at the midline. In principle, it is possible that
monaural cues also play a role. It is, however, precisely at midline
azimuths that the binaural cues themselves are the most sensitive
function of source eccentricity, and it is difﬁcult to imagine
that monaural cues could be informative enough to support empirically
observed sound lateralization thresholds. In this regard, one
recent study showed that even the best long-term monaural listeners
do not have azimuthal localization judgments for near-midline
sources as good as do normal-hearing control listeners (Slattery
and Middlebrooks, 1994).
An objection to our use of a lower envelope principle in the
present context might be derived from the work of Hancock and
Delgutte (2004). They studied ITD coding in the cat’s auditory midbrain,
and then used modeling techniques to try to explain human
ITD discrimination based on the slopes of neurophysiological ITD
functions. In particular, they were concerned with the fact that ITD
acuity is greater for small ITDs than for large ones, especially for
noise stimuli. Their modeling data, far from supporting a lower
envelope principle, in fact suggested that pooling of neural outputs
across frequency provided a better account of human ITD discrimination,
and for both tonal and noise stimuli. This was due, at least in
part, to the alignment of ITD functions over zero ITD (see Hancock
and Delgutte, 2004, for details). In practice, however, their ﬁnding
does not contradict the present hypothesis. Variance in ITD
898 D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900
discrimination at different base ITDs within a species is simply not
the same thing as differences in lateralization acuity (i.e., detection
of a change in sign of ITD/ILD from favouring one ear to
favouring the other, which is what the Heffner and Heffner task
likely measures) across species that are equipped with the same
ITD/ILD coding mechanisms. In a sense, these are almost orthogonal
dimensions. It is quite possible that a pooling principle explains
acuity differences for ITD acuity across different base ITDs on one
or the other side, while differences in the application of a lower
envelope principle explain the acuity differences for left-right discriminations.
In this regard, acuity of left-right discriminations
tend to survive cortical lesions in cats (Cranford, 1979) and ferrets
(Kavanagh and Kelly, 1987), while localization discriminations
within the affected auditory hemiﬁeld(s) do not, in cats (Jenkins and
Masterton, 1982; Jenkins and Merzenich, 1984), primates (Heffner,
1997) or ferrets (Kavanagh and Kelly, 1987). These data suggest that
left-right laterality discriminations are mediated by fundamentally
different neural mechanisms than those mediating localization discriminations,
and thus support the contention that our hypothesis
is not at all incompatible with that of Hancock and Delgutte (2004).
A second objection to our argument might be that a lower
envelope principle in fact contradicts the two- (or three-) channel
localization model. This is because the model is based on the
premise that it is in some way the aggregate or vector response of
neuron response rates in the channels that contributes to relative
ﬁring rate comparison, while our application of a lower envelope
principle counters this by suggesting that the discrimination is
based selectively on the responses of the ‘best’ neurons. In practice,
this is precisely our point: that ‘expert’ threshold performance
develops from the ability to differentially weight the activity of
the ‘best’ neurons in the channels contributing to the perceptual
decision-making process. That is, the neurons within a channel
that are most critical in signaling a change in location will be
those with the steepest functions and it is the ability of a particular
species to differentially weight the responses of this group
of neurons in the decision-making process that determines their
localization/lateralization acuity.
It is not clear what mechanisms the brain implements to exploit
a lower envelope principle. It is, however, an axiom of perception
and psychophysics that subjects improve their thresholds for
detection or discrimination with practice. At least in part, that
improvement likely reﬂects that the superior (‘trained’) perceptual
performance is being based on better selective access to,
i.e., more heavy weighting of, the responses of the neurons with
the best thresholds rather than on any vector weighting of the
responses of a larger population of neurons. This is not to dispute
that perceptual development and perceptual learning cannot
exploit neural strategies that do indeed ‘ﬁne tune’ neural circuits
serving discriminations, especially complex ones (e.g., face
recognition, object recognition, speech discrimination). Indeed,
the construction of perceptual prototypes or templates is perhaps
fundamental to complex perceptual discriminations, and that construction
must have neural correlates. However, practice effects in
acuity along basic psychophysical dimensions (that is, dimensions
that are neurophysiologically close to parameters mapped by the
sensory transduction process), may not require appeal to this kind
of perceptual learning process; they may require only improved
efﬁciency at basing the perceptual judgement on the responses of
the ‘best’ neurons.
There are examples in the literature of apparently superior
sound localization performance among exceptionally practiced listeners,
including orchestra conductors (Münte et al., 2001) and
persons with early onset blindness (Lessard et al., 1998; Ohuchi
et al., 2006). It is important brieﬂy to consider these, to question
whether the apparently superior performance reﬂects a genuine
shift in the sensitivity of ITD/ILD coding mechanisms (some kind of
auditory ‘hyperacuity’) or some other practice effect. In this regard,
superior accuracy in sound localization by blind listeners is not
without dispute (e.g., Zwiers et al., 2001), although their processing
efﬁciency might be enhanced (e.g., shorter reaction times: Liotti
et al., 1998). The bulk of evidence to date suggests that any performance
superiority in blind listeners may be restricted to monaural
or other listening conditions that rely less on the coding of ITD/ILD
cues than on spectral or other information; this is because sound
localization under binaural conditions permitting use of dichotic
cues (e.g., at the frontal midline) is already at ceiling in both sighted
and blind listeners (Lessard et al., 1998; Zwiers et al., 2001). The
performance advantage in the blind listener is restricted to sources
in the lateral hemiﬁelds. That is, the blind listener becomes expert
in using information residing in details of the head-related transfer
function and/or in sound reﬂections from the local environment
that are used for localizing sources away from the midline.
The same conclusion is drawn from studies of minimum audible
angles in visually-deprived ferrets (King and Parsons, 1999). These
animals show no difference from control animals for minimum
audible angles around the midline, but better than control performance
(i.e., smaller minimum audible angles) for sources centered
on 45◦ in the lateral hemiﬁelds (King and Parsons, 1999). Compatible
data have been described for cats deprived of vision from
birth (Rauschecker and Kniepert, 1994). In this experiment, control
and visually-deprived cats had to walk towards whichever of eight
hidden free-ﬁeld speakers, disposed at 45◦ eccentricity spacings,
was activated in each trial. Performance was more accurate for the
visually deprived animals for all source locations, but mostly so for
sources at lateral azimuths or behind the animal.
Arguably the simplest interpretation of the evidence from
visually-impaired listeners is that extreme practice has little or no
effect on the use of ITD or ILD cues, but increases the listener’s perceptual
weighting of the more subtle cues of monaural spectrum
or sound reﬂections. Parenthetically, even if we acknowledge that
blind listeners who exercise extreme practice are more often at
the superior end of the performance spectrum, it is far from clear
that the same performance levels would not be reached by regular
listeners given equivalent practice. For the present hypothesis, it
is signiﬁcant that studies of blind individuals have not yet offered
evidence of a modulation in the neural coding of basic ITD and ILD
cues.
4. Summary
Across mammalian species, there is an inverse relation between
width of ﬁeld of best vision and sound localization acuity around the
midline, and it is independent of the animal’s head size (Heffner and
Heffner, 1992; Heffner, 2004; Heffner et al., 2007). This raises the
question of what it is about the central processing of sound localization
cue coding that varies among these species. Under optimal
stimulus conditions, human beings have minimal audible angles as
low as about 1◦ of azimuth (Mills, 1958), can detect ITDs as small
as 10–15 ␮s (Klump and Eady, 1956; Mills, 1958), and can detect
ILDs as small as 1 dB (Mills, 1958; Grantham, 1984). Cats can discriminate
ITDs as small as 15 ␮s and ILDs as small as 1 dB (Cranford,
1979), despite the fact that cats have a much smaller head size than
human beings. Likewise, Japanese macaque monkeys have comparable
thresholds (26 ␮s and 1 dB: Boester, 1994). These data suggest
that the ﬁdelity of binaural spatial cue coding in the mammalian
central auditory system is as ﬁne as it can get, and no advantage is
accrued by increases in interaural distance beyond that required to
be optimally exploited by the coding mechanism.
The neurophysiological data reviewed above suggest a stability
of ITD/ILD coding mechanisms across mammalian species, and
across frequency within species. The general form of the mechanisms
is a sensitivity to the amplitude of basilar membrane motion
D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900 899
on the two sides, with the extension that for low-frequency sound
components, signal period is so long that cochlear output can
track basilar membrane motion within the cycle, enabling a central
comparison of those tracking responses in a form described
as an ongoing temporal coincidence detection. This set of mechanisms
is general across the frequency range, with the exception of
some species that appear unable to use one or other of the cues.
The behavioural data also prompt the view that the neurophysiological
performance supported by this machinery is the best that
mammalian mechanisms can provide.
These ﬁndings collectively are the basis of our view that
mammals as a biological class have acquired common sets of mechanisms
for the coding of the dichotic sound localization cues (ITD
and ILD). This stability raises a question about the central neural
source of the species differences in free-ﬁeld sound localization
acuity that is coupled so strongly to width of ﬁeld of best vision
(Heffner and Heffner, 1992). Our hypothesis is that there is no direct
adjustment or variation in strictly central auditory processes correlated
with the visual factor. All that is required to relate the two
ﬁndings (width of ﬁeld of best vision, sound localization acuity) is
evolutionary pressure on the need to exploit Barlow’s (1972) lower
envelope principle, namely pressure on the need to base perceptual
judgments selectively on the outputs of the central neurons with
the best discriminative performance. There is little such pressure
in the case of species with broad visual streaks, but a great deal of
it in the case of species with small foveas.
Acknowledgements
Preparation of this manuscript, and execution of some of the
work cited herein, was supported in part by grants from NSERC of
Canada and the Killam Trust to DPP. RND is supported by an NSERC
of Canada PhD scholarship, and CKQ is supported by an NSERC of
Canada Julie Payette scholarship. Special thanks are due to Susan E.
Hall for helpful discussions at all stages of this work, and for preparation
of the illustrations. Thanks are also due to Katrin Krumbholz
and to two anonymous reviewers for their helpful comments on a
previous version of this manuscript.
References
Barlow, H.B., 1972. Single units and sensation: a neuron doctrine for perceptual
psychology? Perception 1, 371–394.
Bernstein, L.R., Trahiotis, C., 2002. Enhancing sensitivity to interaural delays at high
frequencies by using ‘transposed stimuli’. J. Acoust. Soc. Am. 112, 1026–1036.
Boehnke, S.E., Phillips, D.P., 1999. Azimuthal tuning of human perceptual channels
for sound location. J. Acoust. Soc. Am. 106, 1948–1955.
Boester, L.L., 1994. Binaural Time and Intensity Discrimination Following Unilateral
Auditory Cortex Ablation in Japanese Macaques (Macaca fuscata). Unpublished
MA dissertation, Department of Psychology, University of Toledo, Toledo, OH.
Boudreau, J.C., Tsuchitani, C., 1968. Binaural interaction in the cat superior olive S
segment. J. Neurophysiol. 31, 442–454.
Brand, A., Behrend, O., Marquardt, T., McAlpine, D., Grothe, B., 2002. Precise inhibition
is essential for microsecond interaural time difference coding. Nature 417,
543–547.
Brugge, J.F., Anderson, D.J., Aitkin, L.M., 1970. Responses of neurons in the dorsal
nucleus of the lateral lemniscus of cat to binaural tonal stimulation. J. Neurophysiol.
33, 441–458.
Brugge, J.F., Anderson, D.J., Hind, J.E., Rose, J.E., 1969a. Time structure of discharges
in single auditory nerve ﬁbers of the squirrel monkey in response to complex
periodic sounds. J. Neurophysiol. 32, 386–401.
Brugge, J.F., Dubrovsky, N.A., Aitkin, L.M., Anderson, D.J., 1969b. Sensitivity of single
neurons in auditory cortex of cat to binaural tonal stimulation; effects of varying
interaural time and intensity. J. Neurophysiol. 32, 1005–1024.
Brugge, J.F., Merzenich, M.M., 1973. Responses of neurons in auditory cortex of the
macaque monkey to monaural and binaural stimulation. J. Neurophysiol. 36,
1138–1158.
Cranford, J.L., 1979. Auditory cortex lesions and interaural intensity and phase-angle
discrimination in cats. J. Neurophysiol. 42, 1518–1526.
Dingle, R.N., Hall, S.E., Phillips, D.P., 2010. A midline azimuthal channel in human
spatial hearing. Hear Res. 268, 67–74.
Eggermont, J.J., 1998. Is there a neural code? Neurosci. Biobehav. Revs. 22, 355–370.
Eggermont, J.J., 1999. Neural correlates of gap detection in three auditory cortical
ﬁelds in the cat. J. Neurophysiol. 81, 2570–2581.
Eggermont, J.J., Mossop, J.E., 1998. Azimuth coding in primary auditory cortex of the
cat. I. Spike synchrony versus spike count representations. J. Neurophysiol. 80,
2133–2150.
Grantham, D.W., 1984. Interaural intensity discrimination: insensitivity at 1000 Hz.
J. Acoust. Soc. Am. 75, 1191–1194.
Grifﬁn, S.J., Bernstein, L.R., Ingham, N.J., McAlpine, D., 2005. Neural sensitivity to
interaural envelope delays in the inferior colliculus of the guinea pig. J. Neurophysiol.
93, 3463–3478.
Grothe, B., 2000. The evolution of temporal processing in the medial superior olive,
an auditory brainstem structure. Prog. Neurobiol. 61, 581–610.
Grothe, B., Pecka, M., McAlpine, D., 2010. Mechanisms of sound localization in mammals.
Physiol. Rev. 90, 983–1012.
Guinan, J.J., Norris, B.E., Guinan, S.S., 1972. Single units in the superior olivary complex.
II. Location of unit categories and tonotopic organization. Int. J. Neurosci.
4, 147–166.
Hancock, K.E., Delgutte, B., 2004. A physiologically based model of interaural time
difference discrimination. J. Neurosci. 24, 7110–7117.
Harper, N.S., McAlpine, D., 2004. Optimal neural population coding of an auditory
spatial cue. Nature 430, 682–686.
Hartmann, W.M., 1999. How we localize sound. Phys. Today 52 (11), 24–29.
Heffner, H.E., 1997. The role of macaque auditory cortex in sound localization. Acta
Otolaryngol. Suppl. 532, 22–27.
Heffner, R.S., 2004. Primate hearing from a mammalian perspective. Anat. Rec. 281A,
1111–1122.
Heffner, R.S., Heffner, H.E., 1982. Hearing in the elephant (Elephas maximus), absolute
sensitivity, frequency discrimination, and sound localization. J. Comp. Physiol.
Psychol. 96, 926–944.
Heffner, R.S., Heffner, H.E., 1986. Localization of tones by horses, use of binaural cues
and the role of the superior olivary complex. Behav. Neurosci. 100, 93–103.
Heffner, R.S., Heffner, H.E., 1989. Sound localization, use of binaural cues and superior
olivary complex in pigs Brain. Behav. Evoln. 33, 248–258.
Heffner, R.S., Heffner, H.E., 1992. Visual factors in sound localization in mammals. J.
Comp. Neurol. 317, 219–232.
Heffner, R.S., Koay, G., Heffner, H.E., 2000. Sound localization in a new world frugivorous
bat, Artibeus jamaicensis: acuity, use of binaural cues, and relationship to
vision. J. Acoust. Soc. Am. 109, 412–421.
Heffner, R.S., Koay, G., Heffner, H.E., 2007. Sound localiztion acuity and its relation to
vision in large and small fruit-eating bats. I. Echolocating species, Phyllostomus
hastatus and Carollia perspicallata. Hear Res. 234, 1–9.
Henning, G.B., 1974. Detectability of interaural delay in high-frequency complex
waveforms. J. Acoust. Soc. Am. 55, 84–90.
Hofman, P.M., Van Opstal, A.J., 1998. Spectro-temporal factors in two-dimensional
human sound localization. J. Acoust. Soc Am. 103, 2634–2648.
Houben, D., Gourevitch, G., 1979. Auditory lateralization in monkeys: an examination
of two cues serving directional hearing. J. Acoust. Soc. Am. 66,
1057–1063.
Irvine, D.R.F., 1987. Interaural intensity differences in the cat: changes in sound
pressure level at the two ears associated with azimuthal displacements in the
frontal horizontal plane. Hear Res. 26, 267–286.
Irvine, D.R.F., 1992. Physiology of the auditory brainstem. In: Popper, A.N., Fay, R.R.
(Eds.), The Mammalian Auditory Pathway: Neurophysiology. Springer-Verlag,
New York, pp. 153–231.
Ishigami, Y., Phillips, D.P., 2008. Effect of stimulus hemiﬁeld on free-ﬁeld auditory
saltation. Hear Res. 241, 97–102.
Jenkins, W.M., Masterton, R.B., 1982. Sound localization: effects of unilateral lesions
in central auditory system. J. Neurophysiol. 47, 987–1016.
Jenkins, W.M., Merzenich, M.M., 1984. Role of cat primary auditory cortex for sound
localization behavior. J. Neurophysiol. 52, 819–847.
Joris, P.X., Yin, T.C.T., 1995. Envelope coding in the lateral superior olive. I. Sensitivity
to interaural time differences. J Neurophysiol. 73, 1043–1062.
Joris, P.X., Yin, T.C.T., 2007. A matter of time: internal delays in binaural processing.
Trends Neurosci. 30, 70–78.
Kavanagh, G.L., Kelly, J.B., 1987. Contribution of auditory cortex to sound localization
by the ferret (Mustela putorius). J. Neurophysiol. 57, 1746–1766.
Kelly, J.B., Buckthought, A.D., Kidd, S.A., 1998. Monaural and binaural response properties
of single neurons in the rat’s dorsal nucleus of the lateral lemniscus. Hear
Res. 122, 25–40.
Kelly, J.B., Phillips, D.P., 1991. Coding of interaural time differences of transients in
auditory cortex of Rattus norvegicus: implications for evolution of mammalian
sound localization. Hear Res. 55, 39–44.
King, A.J., Middlebrooks, J.C., 2011. Cortical representation of auditory space. In:
Winer, J.A., Schreiner, C.E. (Eds.), The Auditory Cortex. Springer, New York, pp.
329–342.
King, A.J., Parsons, C.H., 1999. Improved auditory spatial acuity in visually-deprived
ferrets. Eur. J. Neurosci. 11, 3945–3956.
Kitzes, L.M., Wrege, K.S., Cassady, J.M., 1980. Patterns of responses of cortical cells
to binaural stimulation. J. Comp. Neurol. 192, 455–472.
Koay, G., Kearns, D., Heffner, H.E., Heffner, R.S., 1998. Passive sound-localization
ability of the big brown bat (Eptesicus fuscus). Hear Res. 119, 37–48.
Klump, R.G., Eady, H.R., 1956. Some measurements of interaural time difference
thresholds. J. Acoust. Soc. Am. 28, 859–860.
Krishnan, A., McDaniel, S.S., 1998. Binaural interaction in the human frequencyfollowing
response: effects of interaural intensity difference. Audiol. Neuro-Otol.
3, 291–299.
900 D.P. Phillips et al. / Neuroscience and Biobehavioral Reviews 36 (2012) 889–900
Lessard, M., Lepore, F., Villemagne, J., Lassonde, M., 2002. Sound localization in
callosal agenesis and early callosotomy subjects: brain reorganization and/or
compensatory strategies. Brain 125, 1039–1053.
Lessard, M., Paré, M., Lepore, F., Lassonde, M., 1998. Early-blind human subjects
localize sound sources better than sighted subjects. Nature 395, 278–280.
Liotti, M., Ryder, K., Woldorff, M.G., 1998. Auditory attention in the congenitally
blind: where, when and what gets reorganized? Neuroreport 9, 1007–1012.
Magezi, D.A., Krumbholz, K., 2010. Evidence for opponent-channel coding of interaural
time differences in human auditory cortex. J. Neurophysiol. 104, 1997–2007.
Makous, J.C., Middlebrooks, J.C., 1990. Two-dimensional sound localization by
human listeners. J. Acoust. Soc. Am. 87, 2188–2200.
Massoud, S., Aiken, S.J., Newman, A.J., Phillips, D.P., Bance, M., 2011. Sensitivity of
the human binaural cortical steady state response to interaural level differences.
Ear Hear. 32, 114–120.
Masterton, R.B., Thompson, G.C., Bechtold, J.K., RoBards, M.J., 1975. Neuroanatomical
basis of binaural phase-difference analysis for sound localization: a comparative
study. J. Comp. Physiol. Psychol. 89, 379–386.
McAlpine, D., Grothe, B., 2003. Sound localization and delay lines – do mammals ﬁt
the model? Trends Neurosci. 26, 347–350.
McAlpine, D., Jiang, D., Palmer, A.R., 2001. A neural code for low-frequency sound
localization in mammals. Nat. Neurosci. 4, 396–401.
McFadden, D., 1973. Precedence effects and auditory cells with long characteristic
delays. J. Acoust. Soc. Am. 54, 528–530.
Middlebrooks, J.C., Green, D.R., 1990. Directional dependence of interaural envelope
delays. J. Acoust. Soc. Am. 87, 2149–2162.
Middlebrooks, J.C., Makous, J.C., Green, D.M., 1989. Directional sensitivity of
sound-pressure levels in the human ear canal. J. Acoust. Soc. Am. 86,
89–108.
Mills, A.W., 1958. On the minimum audible angle. J. Acoust. Soc. Am. 30, 237–246.
Mills, A.W., 1960. Lateralization of high-frequency tones. J. Acoust. Soc. Am. 32,
132–134.
Moore, B.C.J., 2003. An Introduction to the Psychology of Hearing, 5th edn. Academic
Press, London.
Moore, C.N., Casseday, J.H., Neff, W.D., 1974. Sound localization: the role of the
commissural pathways of the auditory system of the cat. Brain Res. 82,
13–26.
Münte, T.F., Kohlmetz, C., Nager, W., Altenmüller, E., 2001. Superior auditory spatial
tuning in conductors. Nature 409, 580.
Ohuchi, M., Iwaya, Y., Suzuki, Y., Munekata, T., 2006. A comparative study of sound
localization acuity of congenital blind and sighted people. Acoust. Sci. Technol.
27, 290–293.
Orman, S.S., Phillips, D.P., 1984. Binaural interactions of single neurons in posterior
ﬁeld of cat auditory cortex. J. Neurophysiol. 51, 1028–1039.
Park, T.J., Klug, A., Holinstat, M., Grothe, B., 2004. Interaural level difference processing
in the lateral superior olive and the inferior colliculus. J. Neurophysiol. 92,
289–301.
Pecka, M., Brand, A., Behrend, O., Grothe, B., 2008. Interaural time difference processing
in the mammalian medial superior olive: the role of glycinergic inhibition.
J. Neurosci. 28, 6914–6925.
Phillips, D.P., 1980. Responses to Tonal Stimuli of Single Neurones in PhysiologicallyDeﬁned
Area AI of Cat Cerebral Cortex. Unpublished PhD Dissertation, Monash
University, Melbourne, Australia.
Phillips, D.P., 2000. Introduction to the central auditory nervous system. In: Jahn,
A.F., Santos-Sacchi, J.R. (Eds.), Physiology of the Ear. , 2nd edn. Singular, San
Diego, pp. 613–638.
Phillips, D.P., 2008. A perceptual architecture for sound lateralization in man. Hear
Res. 238, 124–132.
Phillips, D.P., Brugge, J., 1985. Progress in neurophysiology of sound localization.
Ann. Rev. Psychol. 36, 245–274.
Phillips, D.P., Hall, S.E., 2005. Psychophysical evidence for adaptation of central
auditory processors for interaural differences in time and level. Hear Res. 202,
188–199.
Phillips, D.P., Hall, S.E., Boehnke, S.E., 2002. Central auditory onset responses, and
temporal asymmetries in auditory perception. Hear Res. 167, 192–205.
Phillips, D.P., Irvine, D.R.F., 1981. Responses of single neurons in physiologically
deﬁned area AI of cat cerebral cortex: sensitivity to interaural intensity differences.
Hear Res. 4, 299–307.
Rauschecker, J.P., Kniepert, U., 1994. Auditory localization behaviour in visually
deprived cats. Eur. J. Neurosci. 6, 149–160.
Rose, J.E., Brugge, J.F., Anderson, D.J., Hind, J.E., 1967. Phase-locked response to
low-frequency tones in single auditory nerver ﬁbers of the squirrel monkey.
J. Neurophysiol. 30, 769–793.
Roth, G.L., Kochhar, R.K., Hind, J.E., 1980. Interaural time differences: implications
regarding the neurophysiology of sound localization. J. Acoust. Soc. Am. 68,
1643–1651.
Ruggero, M.A., 1992. Physiology and coding of sound in the auditory nerve. In: Popper,
A.N., Fay, R.R. (Eds.), The Mammalian Auditory Pathway: Neurophysiology.
Springer-Verlag, New York, pp. 34–93.
Russell, I.J., Sellick, P.M., 1978. Intracellular studies of hair cells in the mammalian
cochlea. J. Physiol. 338, 179–206.
Salminen, N.H., May, P.J.C., Alku, P., Tiitinen, H., 2009. A population rate code of
auditory space in the human auditory cortex. PLoS ONE 4, e7600.
Salminen, N.H., Tiitinen, H., Yrttiaho, S., May, P.J.C., 2010. The neural code for interaural
time difference in human auditory cortex. J. Acoust. Soc. Am. Express Lett.
127, EL60–EL65.
Schwartz, I.R., 1992. The superior olivary complex and lateral lemniscal nuclei. In:
Webster, D.B., Popper, A.N., Fay, R.R. (Eds.), The Mammalian Auditory Pathway:
Neuroanatomy. Springer-Verlag, pp. 117–167.
Slattery III, W.H., Middlebrooks, J.C., 1994. Monaural sound localization: acute verus
chronic unilateral impairment. Hear Res. 75, 38–46.
Smith, P.H., Joris, P.X., Yin, T.C.T., 1993. Projections of physiologically characterized
spherical bushy cell axons from the cochlear nucleus of the cat: evidence for
delay lines to the medial superior olive. J. Comp. Neurol. 331, 245–260.
Spitzer, M.W., Semple, M.N., 1995. Neurons sensitive to interaural phase disparity
in gerbil superior olive: diverse monaural and temporal response properties. J.
Neurophysiol. 73, 1668–1690.
Spitzer, M.W., Semple, M.N., 1998. Transformation of binaural response properties
in the ascending auditory pathway: inﬂuence of time-varying interaural phase
disparity. J. Neurophysiol. 80, 3062–3076.
Spoendlin, H.H., 1967. Innervation patterns in the Organ of Corti of the cat. Acta
Otolaryngol. 67, 239–254.
Stecker, G.C., Harrington, I.A., Middlebrooks, J.C., 2005. Location coding by opponent
neural populations in the auditory cortex. PLoS Biol. 3, e78.
Suga, N., 1982. Functional organization of the auditory cortex. In: Woolsey, C.N. (Ed.),
Cortical Sensory Organization, vol. 3. Multiple Auditory Areas. Humana, Clifton,
NJ, pp. 157–218.
Tollin, D.J., Yin, T.C.T., 2005. Interaural phase and level difference sensitivity in lowfrequency
neurons in the lateral superior olive. J. Neurosci. 25, 10648–10657.
van Adel, B.A., Kidd, S.A., Kelly, J.B., 1999. Contribution of the commissure of Probst
to binaural evoked responses in the rat’s inferior colliculus: interaural time
differences. Hear Res. 130, 115–130.
Vigneault-MacLean, B.K., Hall, S.E., Phillips, D.P., 2007. The effects of lateralized
adaptors on lateral position judgements of tones within and across frequency
channels. Hear Res. 224, 93–100.
Yin, T.C.T., Chan, J.C.K., 1990. Interaural time sensitivity in medial superior olive of
cat. J. Neurophysiol. 64, 465–488.
Yin, T.C.T., Chan, J.C.K., Irvine, D.R.F., 1986. Effects of interaural time delays of noise
stimuli on low-frequency cells in the cat’s inferior colliculus. J. Neurophysiol.
55, 280–300.
Yin, T.C.T., Kuwada, S., 1983. Binaural interaction in low-frequency neurons in inferior
colliculus of the cat. III. Effects of changing frequency. J. Neurophysiol. 50,
1020–1042.
Yin, T.C.T., Kuwada, S., Sujaku, Y., 1984. Interaural time sensitivity of high-frequency
neurons in the inferior colliculus. J. Acoust. Soc. Am. 76, 1401–1410.
Yost, W.A., 1981. Lateral position of sinusoids presented with interaural intensive
and temporal differences. J. Acoust. Soc. Am. 70, 397–409.
Zwiers, M.P., Van Opstal, A.J., Cruysberg, R.M.G., 2001. Two-dimensional sound localization
behavior of early-blind humans. Exp. Brain Res. 140, 206–222.