10648 • The Journal of Neuroscience, November 16, 2005 • 25(46):10648 –10657

Behavioral/Systems/Cognitive

Interaural Phase and Level Difference Sensitivity in

Low-Frequency Neurons in the Lateral Superior Olive
Daniel J. Tollin and Tom C. T. Yin
Department of Physiology, University of Wisconsin–Madison, Madison, Wisconsin 53706

The lateral superior olive (LSO) is believed to encode differences in sound level at the two ears, a cue for azimuthal sound location. Most
high-frequency-sensitive LSO neurons are binaural, receiving inputs from both ears. An inhibitory input from the contralateral ear, via
the medial nucleus of the trapezoid body (MNTB), and excitatory input from the ipsilateral ear enable level differences to be encoded.
However, the classical descriptions of low-frequency-sensitive neurons report primarily monaural cells with no contralateral inhibition.
Anatomical and physiological evidence, however, shows that low-frequency LSO neurons receive low-frequency inhibitory input from
ipsilateral MNTB, which in turn receives excitatory input from the contralateral cochlear nucleus and low-frequency excitatory input
from the ipsilateral cochlear nucleus. Therefore, these neurons would be expected to be binaural with contralateral inhibition. Here, we
re-examined binaural interaction in low-frequency (less than ⬃3 kHz) LSO neurons and phase locking in the MNTB. Phase locking to
low-frequency tones in MNTB and ipsilaterally driven LSO neurons with frequency sensitivities ⬍1.2 kHz was enhanced relative to the
auditory nerve. Moreover, most low-frequency LSO neurons exhibited contralateral inhibition: ipsilaterally driven responses were sup-
pressed by raising the level of the contralateral stimulus; most neurons were sensitive to interaural time delays in pure tone and noise
stimuli such that inhibition was nearly maximal when the stimuli were presented to the ears in-phase. The data demonstrate that
low-frequency LSO neurons of cat are not monaural and can exhibit contralateral inhibition like their high-frequency counterparts.
Key words: lateral superior olive; medial nucleus of the trapezoid body; interaural time delay; interaural level difference; sound localiza-
tion; phase locking

Introduction (IE) binaural interaction]. The dual brainstem pathways provide

For ⬃100 years, the Duplex theory has posited that low- and anatomical and physiological correlates of the Duplex theory.
high-frequency sounds are localized using two different acousti- LSO neurons encode ILDs for high-frequency sounds in
cal cues, interaural time delays (ITDs) and interaural level differ- which the cues are physically available (Boudreau and Tsuchi-
ences (ILDs), respectively (Rayleigh, 1907). Psychophysical data tani, 1968; Tollin, 2003), but there are discrepancies regarding
have generally supported the theory for pure tones (Stevens and low-frequency neurons. The classical descriptions of LSO report
Newman, 1936; Mills, 1958). Anatomical and physiological stud- that low-frequency neurons are not binaural but rather are mon-
ies have revealed two parallel brainstem pathways that appear to aural in that they respond to stimuli presented to only one ear
encode ITDs and ILDs separately (Yin, 2002). ITDs are extracted (Boudreau and Tsuchitani, 1968; Guinan et al., 1972b; Tsuchi-
by medial superior olive neurons in which ITD sensitivity results tani, 1977). Batra et al. (1997) were also unable to find low-
from the coincident arrival of excitatory inputs from the two ears frequency IE neurons in the rabbit brainstem, although it appears
[excitatory– excitatory (EE) binaural interaction], causing maxi- they only used monaural responses to assess the presence of in-
mal responses when the ITD in the stimulus offsets differences in hibitory input via reductions in spontaneous activity. However,
neural conduction times from the ears. ILDs are extracted by anatomical and physiological evidence predicts that low-
lateral superior olive (LSO) neurons via a subtraction-like pro- frequency LSO neurons should exhibit IE interaction and thus
cess resulting from inhibitory inputs from the contralateral and respond to ILDs. First, these neurons receive excitatory input
excitatory inputs from the ipsilateral ear [inhibitory– excitatory from neurons in the ipsilateral cochlear nucleus and frequency-
matched input from glycinergic, inhibitory cells of the ipsilateral
medial nucleus of the trapezoid body (MNTB), which in turn
Received April 23, 2005; revised Sept. 28, 2005; accepted Oct. 1, 2005. receives excitatory input from neurons in the contralateral co-
This work was supported by National Institute on Deafness and Other Communication Disorders Grants DC-00116
chlear nucleus (Glendenning et al., 1985; Smith et al., 1993,
and DC-02840 to T.C.T.Y. and DC-005122 and DC-006865 to D.J.T. Special thanks to Dr. Philip Joris for data. We
acknowledge the assistance of J. Sekulski and R. Kochhar for help with computer programming and Dr. Nathan 1998). Low-frequency LSO neurons would also be expected to be
Schoppa for comments on this manuscript. sensitive to ITDs in the ongoing “fine structure” of low-
Correspondence should be addressed to Dr. Daniel J. Tollin, Department of Physiology and Biophysics, University frequency sounds, because the afferents preserve temporal infor-
of Colorado Health Sciences Center, Mail Stop 8307, P.O. Box 6511, Aurora, CO 80045. E-mail:
[email protected].
mation in the fine structure of the sounds from both ears: low-
DOI:10.1523/JNEUROSCI.1609-05.2005 frequency neurons of the ipsilateral cochlear nucleus (Joris et al.,
Copyright © 2005 Society for Neuroscience 0270-6474/05/2510648-10$15.00/0 1994) and the MNTB exhibit phase-locked action potentials that
Tollin and Yin • Interaural Phase Coding in the LSO J. Neurosci., November 16, 2005 • 25(46):10648 –10657 • 10649

occur at particular phases of low-frequency tones (Smith et al., dated the small probe-tube microphone by which the sound delivery
1998; Kopp-Scheinpflug et al., 2003). We hypothesize that the system to each ear was calibrated for tones between 50 Hz and 40 kHz in
ITD that offsets the differences in neural conduction times from 50 Hz steps. The calibration data were used to compute digital filters that
the two ears should result in the coincident arrival of excitation equalized the responses of the acoustical system and typically yielded flat
frequency responses within ⫾2 dB for frequencies ⬍25 kHz.
and inhibition from the ipsilateral and contralateral ears, respec-
Tone bursts of varying frequency were used as search stimuli with the
tively. Consequently, LSO neurons will be maximally inhibited at
sound pressure level of the tone to the ipsilateral ear being 5–10 dB higher
this delay. Low-frequency LSO neurons sensitive to ILDs have than the tone to the contralateral ear so that the IE cells of the LSO would
been reported (Caird and Klinke, 1983; Finlayson and Caspary, not be missed. Once a single unit was isolated and the excitatory ear
1991; Spitzer and Semple, 1995; Joris and Yin, 1995), but few determined, its characteristic frequency and threshold level were mea-
studies in the cat (Caird and Klinke, 1983; Joris and Yin, 1995) sured using an automated tracking procedure based on that by Kiang et
have reported ITD-sensitive cells of the type just described nor al. (1970). Discharge rate was measured as a function of sound pressure
have they been studied in detail. Here, we report results from level by presenting 200 repetitions of a 50 ms duration tone (3.9 ms
low-characteristic-frequency (less than ⬃3 kHz) LSO and MNTB rise/fall times) every 100 ms at the characteristic frequency at different
neurons. sound pressure levels in 5–10 dB steps. From these rate–level functions,
the resulting poststimulus time and period histograms were examined.
Phase locking (or synchrony), which measures the tendency of a neuron
Materials and Methods to fire action potentials that are synchronized to a particular phase of a
General sinusoidal stimulus, was measured by computing the synchronization
All procedures used were approved by the University of Wisconsin Ani- coefficient (Goldberg and Brown, 1969). To determine the presence and
mal Care and Use Committee and also complied with the National Insti- nature of any binaural interaction, a characteristic frequency tone or
tutes of Health guidelines for animal use. Adult female cats were anes- broadband noise (300 ms duration presented every 500 ms with a rise/fall
thetized initially with ketamine hydrochloride (20 mg/kg) along with time of 4 ms) was presented to the ipsilateral ear at 10 –20 dB above the
acepromazine (0.1 mg/kg). Atropine sulfate (0.05 mg/kg) was also ad- threshold level while the level of a corresponding characteristic frequency
ministered to reduce mucous secretions. A venous cannula was im- tone or noise presented to the contralateral ear was varied. This proce-
planted in the femoral vein through which supplemental doses of sodium dure reveals whether ipsilaterally evoked neural responses can be inhib-
pentobarbital (3–5 mg/kg) were administered as needed to maintain ited by contralateral stimulation, a hallmark of LSO cells.
areflexia. The cat’s temperature was monitored continuously with a rec- Binaural beat. Sensitivity to ITDs of low-frequency tones was assessed
tal thermometer and maintained with a heating pad at 37°C. A tracheal in two ways. The first method used the binaural-beat stimulus (Kuwada
cannula was inserted to facilitate respiration. Both pinnas were removed, et al., 1979). To create binaural beats, tones of 5 s in duration were
and tight-fitting hollow earpieces were fitted snugly into the external presented to both ears ⬃20 –30 dB above ipsilateral-ear threshold but
auditory meati. Polyethylene tubing (PE-90; 30 cm; inner diameter, 0.9 with a frequency difference of 1 Hz, usually with the tone to the ipsilateral
mm; Clay Adams, New York, NY) was glued into a small hole made in ear having the higher frequency. The binaural-beat stimulus results in a
each bulla to maintain normal middle-ear pressure. dynamically changing difference in the phases of the tones presented to
A ventral transpharyngeal approach was used, and the LSO was ac- the two ears, or interaural phase difference (IPD), over a 1 s period (see
cessed by drilling small holes into the basioccipital bone. Small slits were Fig. 5C, inset, for example binaural beat). ITDs can be computed from
then made in the dura through which parylene-coated tungsten micro- IPDs by simply dividing the IPD by the stimulus frequency. For example,
electrodes (1–2 M⍀; Microprobe, Clarksburg, MD) were advanced ven- an IPD of 0.5 cycles with a 500 Hz stimulus is equivalent to an ITD of 1
tromedial to dorsolateral at an angle of 26 –30° into the brainstem by a ms, but the same IPD is equivalent to 0.5 ms for a 1000 Hz stimulus. As
hydraulic microdrive affixed to a micromanipulator (Trent Wells, discussed below, we use IPDs in the analysis of the data because it allows
Coulterville, CA) that could be advanced remotely from outside the the binaural sensitivity of a neuron to be classified in a common frame-
double-walled sound-attenuating recording chamber (Industrial Acous- work, regardless of the absolute sensitivity to ITDs. If the neuron is
tics Company, Bronx, NY). Electrical activity was amplified and filtered sensitive to IPD, then its response will be modulated systematically by the
between 300 and 3000 Hz (model CR4; Princeton Applied Research, Oak variation in IPD present in the binaural beats, which can be converted to
Ridge, TN). Unit responses were discriminated with an amplitude–time an ITD.
window discriminator (model DIS-1; Bak Electronics, Mount Airy, Responses of LSO neurons to the 1 Hz binaural-beat stimuli were
MD), and spike times were stored at a precision of 1 ␮s via custom-built obtained at several different frequencies near the characteristic frequency
equipment. of each neuron. At each frequency, the stimulus was repeated three times
Histology with a 1 s interstimulus interval. To characterize the tendency of these
In most experiments, electrolytic DC lesions were made to differentiate neurons to phase lock or fire action potentials at particular IPDs, period
electrode tracks, mark locations of interest, and assist in estimating tissue histograms were constructed by binning (64 bins) the responses on the
shrinkage after histological processing. At the conclusion of each exper- period (1 s) of the binaural-beat frequency (1 Hz). Note that stimuli
iment, the brain was fixed in 10% formalin by immersion or perfusion presented in the configuration we used, with the ipsilateral ear having the
through the heart. The fixed tissue was cut into 50 ␮m frozen sections higher frequency, were equivalent to a “negative” binaural beat by our
and stained with cresyl violet so that electrode tracts and lesions could be convention (Kuwada et al., 1979; Yin and Kuwada, 1983). To be consis-
seen. Figure 1 shows a section of brainstem from an experiment in which tent with previous studies, the interaural phase was “corrected” by flip-
two low-frequency LSO neurons were recorded from two parallel elec- ping the period histograms around the point of IPD of 0.5 cycles when
trode penetrations. this stimulus configuration was used. From these corrected histograms,
the mean IPD and the interaural synchronization coefficient, a measure
Stimuli of the precision of phase locking, were determined for each stimulating
General. All stimuli were generated digitally at 16 bit resolution and frequency. Significance of interaural phase locking was evaluated with
converted to analog at a rate of 100 kHz via the custom-built digital the Rayleigh test (Mardia, 1972) at the p ⬍ 0.001 level. The characteristic
stimulus system designed by Rhode (1976). Overall stimulus level was delay (CD) (Yin and Kuwada, 1983) and characteristic phase (CP) were
controlled using programmable attenuators. The conditioned output of determined by fitting a line (least-squares procedure) to a plot of the
the digital-to-analog converter was sent to an acoustic assembly (one for mean IPD of the response to different frequencies, with each IPD being
each ear) comprising an electrodynamic speaker (40-1377; Realistic, Fort weighted by the product of the synchronization coefficient and the total
Worth, TX), a calibrated probe-tube microphone (0.5 inch; Brüel & number of spikes (Batra et al., 1997). CD is given by the slope of the
Kjær, Norcross, GA), and a hollow earpiece that was fit and sealed snugly phase–frequency plot and represents the relative difference in neural
into the cut end of the auditory meatus. The hollow earpiece accommo- conduction delays between the inputs to the neuron from the two ears.
10650 • J. Neurosci., November 16, 2005 • 25(46):10648 –10657 Tollin and Yin • Interaural Phase Coding in the LSO

Figure 1. Light micrograph of the left side of the superior olivary complex showing two
electrode penetrations through the LSO. An electrolytic DC lesion marks the locations of one of
the low-characteristic-frequency LSO neurons (arrow; 108-5, characteristic frequency, 1.35 kHz;
and 108-9, characteristic frequency, 1.75 kHz). The major nuclei and landmarks of the superior
olivary complex are denoted. MSO, Medial superior olive; TB, trapezoid body; Pyr, pyramidal
tract.

The CP is determined from the y-intercept value of the phase–frequency

plot at 0 Hz and indicates whether coincident arrival of inputs from the
two ears results in maximal or minimal responses. In our convention,
positive CDs (CD ⬎ 0) indicate a delay to the ipsilateral input, simulating
a stimulus in the contralateral sound field. A CP of 0.0 or 1.0 cycles Figure 2. Phase-locked responses of a low-characteristic-frequency MNTB neuron to pure-
indicates that the CD results in a maximal response from the neuron, tone stimuli. A, Frequency–threshold curve (characteristic frequency, 317 Hz). B, Both the
which is consistent with EE neurons. A CP of 0.5 cycles indicates that the discharge rate (left ordinate; filled circles) and response synchrony (Sync; right ordinate; open
CD results in a minimal response from the neuron, consistent with IE circles) increase with increasing level of a 50 ms duration characteristic frequency tone. Spon-
neurons. taneous rate is indicated by the dashed line (Spon; left ordinate). SPL, Sound pressure level. C, D,
ITD sensitivity was also examined by presenting identical tokens of Dot rasters to 200 presentations of a 50 ms duration characteristic frequency tone (C) and
broadband noise with equal amplitude to the two ears but systematically associated period histograms (D) at three different stimulus levels. The responses are phase
varying the ITD. Here, ITDs were created by delaying the entire noise locked to the stimulus frequency, as evidenced by the tendency of spike times to line up verti-
waveform. The resulting plots of mean discharge rate as a function of the cally in the rasters (C) and for responses to occur at a particular phase angle resulting in peaked
ITD are called noise– delay functions. On occasion, we also assessed ITD period histograms (D). The synchronization coefficients (r) of the responses in D are indicated in
sensitivity to pure tones at or near the characteristic frequency by pre- the inset. The time waveform of the characteristic frequency tone is shown below C.
senting equal amplitude tones to the two ears but varying the ITD in the
whole waveform, rather than using binaural beats.
Neuron classification. The LSO and MNTB neurons reported in this Results
paper were classified based on two main factors: their physiological re- Results are based on single-unit recordings of 16 LSO and 17
sponses to tones and their anatomical location in the brainstem derived MNTB neurons, all with characteristic frequencies ⬍3.2 kHz re-
from histology. Physiologically, LSO neurons were identified by excita- corded from 15 cats. Within this sample, 11 of 16 LSO and 8 of 17
tory responses to ipsilateral and inhibitory responses to contralateral MNTB neurons had characteristic frequencies ⬍1.5 kHz. All
stimulation. Anatomically, these low-frequency LSO neurons were lo- low-frequency LSO neurons were recorded during electrode pen-
cated in the lateral limb of the LSO. The low-frequency LSO neurons
etrations that traversed the LSO (see Materials and Methods).
were invariably encountered deeper in the electrode penetration after
having encountered high- to mid-frequency-sensitive IE neurons that The locations of 12 of 16 LSO neurons were verified through
were recorded in the medial and middle limbs, respectively, during the histology and electrode track reconstruction. Figure 1 shows an
same penetrations. Their depth relative to the mid- to high-frequency example histological section with two well stained parallel pene-
neurons ranged from 441 to 1286 ␮m (mean, 731 ␮m). Given the angle trations from which recordings of two low-frequency LSO neu-
of electrode penetration used in these experiments (26 –30°), the maxi- rons were obtained. The arrow indicates an electrolytic DC lesion
mum distances separating the medial and the middle limbs of the LSO made at the location of one of the neurons (characteristic fre-
from the lateral limb measured in a sample of our histological sections quency, 1.35 kHz), the responses of which are shown in Figure 7.
(see Fig. 1) ranged from ⬃1300 to 750 ␮m, respectively. The depths of all All MNTB neurons responded only to contralateral stimulation,
the low-frequency LSO neurons relative to high-frequency LSO neurons exhibited prepotentials in their action potential waveforms, and
recorded in the same penetration fell below these maximums. MNTB
were encountered in the dorsolateral portion of the MNTB nu-
neurons were identified by the shape of their extracellularly recorded
action potentials, which exhibit a prepotential that precedes the action cleus. Because of limited recording time, not all stimulus manip-
potential by ⬃500 ␮s (Guinan et al., 1972a; Guinan and Li, 1990; Smith ulations were presented to each neuron.
et al., 1998) and by responses only to monaural contralateral stimulation. The reason for the small number of neurons in our sample
They were also invariably located early in the electrode track shortly after deserves some explanation. Historically, LSO, and to a lesser de-
encountering trapezoid body fibers (see Fig. 1) gree MNTB, neurons, which are located deep in the brainstem in
Tollin and Yin • Interaural Phase Coding in the LSO J. Neurosci., November 16, 2005 • 25(46):10648 –10657 • 10651

Figure 4. Low-frequency MNTB and LSO neurons exhibit phase locking to characteristic
frequency tones that is enhanced over that seen in ANFs. Maximum (Max) synchronization
coefficients are plotted as a function of the characteristic frequencies for MNTB (filled circles)
and ipsilaterally stimulated LSO (Ipsi LSO; filled triangles) neurons from this study and from the
population of ANF (small open circles) from the study by Johnson (1980). For characteristic
frequencies ⬍1.2 kHz, LSO and MNTB neurons showed higher synchrony than ANFs with similar
characteristic frequencies but lower synchrony for characteristic frequencies ⬎1.2 kHz. Note
that data points from three MNTB neurons overlap near 1 kHz, and two overlap near 2 kHz at a
synchronization coefficient of 0.6.

Figure 3. Phase-locked responses of a low-characteristic-frequency LSO neuron to pure-

cleus that provide the inputs to the MNTB and the ipsilateral
tone stimuli. A–D, The same as in Figure 2, but, for a low-frequency LSO neuron stimulated with inputs to the LSO (Joris et al., 1994). Figure 2 shows the phase-
characteristic frequency (440 Hz; A), tones were presented to the excitatory ipsilateral ear only. locking ability of an MNTB neuron stimulated monaurally at the
The spontaneous rate for this neuron was 0 spikes/s. Note that because the response happened contralateral ear. Figure 2 A shows the frequency tuning curve of
to fall near a stimulus phase of 0 cycles, the phase axis has been plotted from ⫺0.5 to 0.5 cycles this neuron, which had a characteristic frequency of 317 Hz. The
for ease of presentation. SPL, Sound pressure level; Sync, synchrony. frequency tuning curve shapes and bandwidths of these neurons
were similar to that seen in the auditory nerve and in cochlear
the cat, have proven to be difficult to access and record from with nucleus neurons of similar characteristic frequencies. Figure 2 B
microelectrodes. As a result of the over-representation of high- shows that when presented with a 50 ms duration tone at the
frequency neurons in the LSO, the lateral limb where the low- characteristic frequency, the discharge rate (filled circles) and the
frequency neurons are located is relatively small. The difficulty in synchrony (open circles), which is a measure of the precision of
recording from the low-characteristic frequency LSO neurons in phase locking, increased monotonically with increasing stimulus
the lateral limb is evident in the classic large-scale survey studies level. To illustrate how the synchronous firing of the neuron
of the LSO. Of 432 neurons from all of the superior olivary com- increases with stimulus level, Figure 2, C and D, shows the dot
plex (SOC) nuclei, Guinan et al. (1972b) recorded from only 22 rasters and cycle histograms for 200 presentations of the charac-
LSO neurons, despite its relatively large size compared with other teristic frequency tone at the three levels indicated. At high stim-
SOC nuclei (Fig. 1), and only a handful of these had low- ulus levels (e.g., ⬎35 dB), this neuron exhibits enhanced syn-
characteristic frequencies; and Tsuchitani (1977) recorded from chrony [relative to auditory nerve fibers (ANFs) with similar
244 LSO neurons, but only 16 (7%) of those had characteristic characteristic frequencies; see below] and entrainment, firing an
frequencies ⬍2 kHz. As for low-frequency (⬍2 kHz) MNTB neu- action potential on virtually every single cycle of the stimulus.
rons, Smith et al. (1998) and Kopp-Scheinpflug et al. (2003) re- This is indicated by the discharge rate of the neuron, which as-
corded from only two and three cells, respectively. Clearly, there ymptotes to almost exactly 317 Hz at these higher stimulus levels.
is a need for additional studies of these neurons. Figure 3 shows in the same format the responses of a LSO neuron
with a low characteristic frequency of 440 Hz (Fig. 3A). Figure
Low-frequency LSO and MNTB neurons exhibit highly 3B–D shows that when stimulated with a characteristic frequency
synchronized phase locking tone presented to the ipsilateral, excitatory ear, this LSO neuron
In order for central neurons such as those in the medial superior also exhibits a high degree of phase locking, with synchronization
olive and the LSO to encode ITDs, the afferent inputs from both coefficients of 0.9 or greater. This LSO neuron exhibited a weakly
ears to these neurons must accurately encode the temporal char- nonmonotonic rate–level function.
acteristics of the stimuli. Here, we examine whether low- Figure 4 summarizes the synchronization coefficients of 17
characteristic-frequency LSO and MNTB neurons exhibit phase- MNTB and 10 LSO neurons as a function of the characteristic
locked responses to pure-tone stimuli. Previous studies from our frequency of the neurons. For reference, also plotted are the syn-
laboratory have already demonstrated the excellent, and often chronization coefficients from the population of auditory nerve
enhanced (relative to their auditory nerve inputs), phase-locking fibers from the work by Johnson (1980). LSO (t(145) ⫽ 5.7; p ⬍
abilities of globular and spherical bushy cells of the cochlear nu- 0.0001) and MNTB (t(146) ⫽ 5.4; p ⬍ 0.0001) neurons with char-
10652 • J. Neurosci., November 16, 2005 • 25(46):10648 –10657 Tollin and Yin • Interaural Phase Coding in the LSO

Figure 5. Example of a low-frequency (characteristic frequency, 440 Hz) LSO neuron that is sensitive to ILD and IPD. Same neuron as in Figure 3. A, Mean discharge rate as a function of ILD of
characteristic frequency tones. ILD (bottom abscissa) was varied by fixing the level at the ipsilateral ear at 30 dB and changing the level of a simultaneously presented characteristic frequency tone
to the contralateral ear (top abscissa). Consistent with IE binaural interaction, increases in the level of the contralateral tone inhibited ipsilaterally driven responses resulting in a sigmoidally shaped
rate–ILD function. Error bars indicate ⫾1 SEM. B, Poststimulus time histograms (PSTHs; left column) to binaural beat stimuli created by presenting a tone at the stimulating frequency (top right,
left column) to the ipsilateral ear and a tone at a frequency 1 Hz less to the contralateral ear. An example binaural beat (C, inset) is shown. IPD sensitivity is evidenced by the response modulation
at 1 Hz. Period histograms (right column) were constructed from the PSTHs by binning the responses on the period of the beat frequency (1 s). As described in Materials and Methods, the period
histograms were corrected by flipping them around the point of IPD of 0.5 cycles. Using this representation, responses were low when the tones to the two ears were in-phase (0.0 cycles) and
increased when the tones were out-of-phase (0.5 cycles). Vector averaging of the period histograms for each stimulating frequency yields a mean interaural phase (␾; top left of panels in right
column) and synchronization coefficient (r). Asterisks indicate significant ( p ⬍ 0.001) phase locking to the 1 Hz beat frequency. C, Mean interaural phase (␾b) from the binaural-beat stimuli (B)
as a function of frequency. The CD and CP are estimated from the slope and y-intercept of the best-fitting line to the phase–frequency data (solid line) and are indicated in the top left of the panel.
D, The IPD sensitivity of the neuron as expressed in the binaural-beat period histograms (B) can be converted to an equivalent ITD sensitivity by dividing the IPD by the stimulus frequency and
reorganizing the abscissa, as indicated by the drop lines. Stimulus frequency was 400 Hz. The location of the CD is indicated with an arrow. The solid line in bottom panel shows a smoothed version
of the data. IPSI or Ipsi, Ipsilateral; CONTRA or Contra, contralateral.

acteristic frequencies ⬍1.2 kHz exhibited significantly enhanced ing the level of a characteristic frequency tone presented to the
synchronization to pure-tone stimuli. The synchronization is ipsilateral excitatory ear constant at 30 dB (⬃20 dB with respect
greater than that observed in auditory nerve fibers over the same to threshold), while increasing the level of a characteristic fre-
ranges of characteristic frequencies. This enhanced synchrony is quency tone simultaneously presented in-phase to the contralat-
similar to the synchrony observed in low-characteristic frequency eral inhibitory ear (Fig. 5A, top abscissa). Consistent with IE
spherical and globular bushy cells (Joris et al., 1994), suggesting binaural interaction, increases in sound level at the contralateral
that the enhanced phase locking at those peripheral levels is pre- ear progressively inhibited the ipsilaterally driven responses re-
served at the MNTB when stimulated contralaterally and at the sulting in a classical sigmoidally shaped rate–ILD function that is
LSO when stimulated ipsilaterally. For neurons with characteris- characteristic of LSO neurons. Figure 5B shows that the neuron
tic frequencies greater than ⬃1.2 kHz, phase locking was clearly was also sensitive to the dynamically changing IPDs in binaural
poorer than that exhibited by ANFs. This finding also mirrors the beat stimuli (Fig. 5C, inset) in that the responses were clearly
poorer phase locking observed in bushy cells at higher modulated by the 1 Hz binaural beat frequency. As indicated by
frequencies. the period histograms on the right (the asterisks indicate signifi-
cant phase locking at the p ⬍ 0.001 level to the 1 Hz beat fre-
Low-frequency LSO neurons can be sensitive to both ILDs quency), the response was minimal when the stimuli were nearly
and IPDs in-phase (at the IPD of 0 cycles) at the two ears and nearly max-
The excellent phase locking exhibited by low-characteristic- imal when the stimuli were nearly out-of-phase (at 0.5 cycles).
frequency LSO and MNTB neurons suggests that individual low- Recall that the period histograms were created by flipping the
frequency LSO neurons that exhibit IE binaural interaction histograms created from the responses in Figure 5B (left column)
might indeed be sensitive to IPDs in addition to ILDs. In Figure 5, around 0.5 cycles (see Materials and Methods). Observing these
we show that the same LSO neuron as shown in Figure 3 exhibits corrected period histograms for this neuron, it is clear that inhi-
sensitivity to both ILDs and IPDs, and it illustrates how these bition tended to occur over the first half of the cycle where the
neurons were studied. This neuron showed highly phase-locked binaural beat stimulus was such that the stimulus delivered to the
responses to characteristic frequency tones (440 Hz) presented to contralateral ear (the input to LSO of which is inhibitory) was
the ipsilateral ear (Fig. 3). The function relating discharge rate to leading that to the ipsilateral ear (i.e., ipsilateral delay). Minimal
ILD, the rate–ILD function, for this neuron was studied by hold- inhibition occurred over the last half of the cycle, in which the
Tollin and Yin • Interaural Phase Coding in the LSO J. Neurosci., November 16, 2005 • 25(46):10648 –10657 • 10653

must be reorganized for this procedure.

Figure 5D (bottom) shows the ITD sensi-
tivity of this neuron. The location of the
CD (187 ␮s) is shown. This LSO neuron
was jointly sensitive to both ILD and IPD
(and, equivalently, ITD), consistent with
the expectations given IE binaural interac-
tion and phase-locked inputs.
Figure 6 shows the results of another
low-frequency LSO neuron with a charac-
teristic frequency of 566 Hz (Fig. 6 A). This
neuron was also sensitive to ILDs (Fig. 6 B)
and IPDs of binaural beats (Fig. 6C,D).
Consistent with IE binaural interaction,
the CP for this neuron was near 0.5 (0.47)
cycles with a CD of 235 ␮s. With this neu-
ron, we were able to test the hypothesis
that the CD is that delay which just offsets
the relative neural delay in the excitatory
and inhibitory inputs to the LSO from the
two ears. To do this, we measured the
noise– delay function constructed by plot-
ting the mean discharge rate of the neuron
as a function of the ITD between identical
broadband noise bursts of equal ampli-
tude presented to the two ears (Fig. 6E).
Based on the CP and CD of this neuron, we
hypothesize that the response as a function
of ITD should exhibit a minimum, or
trough, at an ITD that just offsets the CD,
that is, ⬃235 ␮s (indicated by an arrow). Ac-
cordingly, there is a rate minimum at 200 ␮s,
the nearest ITD tested.
Figure 7 shows yet another example
(the histological location of this neuron is
Figure 6. Example of a low-characteristic-frequency LSO neuron that is sensitive to ILDs and IPDs. A, Frequency–threshold shown by the arrow in Fig. 1). This neuron
curve (characteristic frequency, 566 Hz). B–D, The same as in Figure 5. ␾, Mean interaural phase; ␾b, mean interaural phase from had a characteristic frequency of 1345 Hz
binaural-beat stimuli; *p ⬍ 0.001. E, Noise– delay function created by plotting the mean discharge rate as a function of the ITD (Fig. 7A) and was determined audiovisu-
of a broadband noise presented to the two ears with equal amplitude. Negative delays indicate that the contralateral stimulus lags
ally to be ILD sensitive, as expected of LSO
the ipsilateral stimulus. Consistent with the CD (235 ␮s) and the CP (0.47 cycles) obtained from the binaural-beat stimulus for this
neuron (C, D), the response minimum in the noise– delay function occurs at an ipsilateral delay of 200 ␮s, which is near the delay
cells. The neuron was also clearly IPD sen-
of 235 ␮s predicted from the CD. At the CD, the ipsilateral excitation and the contralateral inhibition arrive coincidentally at the sitive yielding a CD of ⫺299 ␮s and a CP of
LSO neuron, resulting in maximal inhibition. Error bars in B and E represent ⫾1 SEM. SPL, Sound pressure level; IPSI or Ipsi, 0.66 cycles (Fig. 7 B, C). A CD ⬍0 indicates
ipsilateral; CONTRA or Contra, contralateral. that the excitatory input from the ipsilat-
eral ear lags the inhibitory input from the
contralateral ear. Figure 7D shows the
stimulus configuration was reversed (Contra delay). In addition noise– delay function for this neuron, which has a clear response
to the ILD sensitivity, this type of response to IPDs is wholly minimum at a delay of 0 ␮s, along with the predicted location of the
consistent with IE binaural interaction as well. response minimum at ⫺299 ␮s.
From the binaural-beat data at several different frequencies, Figure 8 shows a summary of the IPD and ILD sensitivity of
the CP and CD were computed (Fig. 5C). The slope of the best- low-characteristic frequency LSO neurons. To the extent to
fitting line to the phase–frequency data indicates the CD, and the which the inputs to these neurons were phase locked (Fig. 4), we
intercept indicates the CP. Indicative of IE binaural interaction, expected these neurons to be IPD sensitive. All neurons tested (10
the CP was near 0.5 (0.61 cycles). The positive CD (⫹187 ␮s) of 10) with binaural beats were modulated with changes in IPD
indicates that the inhibitory input from the contralateral ear lags for at least one carrier frequency. Two additional neurons were
the excitatory input from the ipsilateral ear by 187 ␮s. The IPDs not tested with binaural beats, but their responses were system-
that occur over one cycle of the binaural-beat stimulus (Fig. 5C, atically modulated by ITDs in characteristic frequency tones. Fig-
inset) can be converted to equivalent ITDs by dividing the IPD ure 8 A shows the phase–frequency plots for the eight neurons
axis of the period histograms by the stimulus frequency (see Ma- that were tested most extensively with binaural beats. Note that
terials and Methods). The period histograms can then be replot- for two neurons, 110-47 and 023-14, the phase–frequency func-
ted as a function of this equivalent ITD. Figure 5D (top) shows an tions are based on only two points. Each point, however, pro-
example period histogram created from the binaural-beat stim- duced significant phase locking to the binaural beats. All neurons
ulus for a frequency of 400 Hz, which is replotted in terms of yielded CPs that were nearer to 0.5 cycles (mean CP, 0.52 cycles;
equivalent ITD (bottom). The drop lines show how the abscissa SD, 0.12 cycles; n ⫽ 8) than they were to 0.0 or 1.0 cycles, and CDs
10654 • J. Neurosci., November 16, 2005 • 25(46):10648 –10657 Tollin and Yin • Interaural Phase Coding in the LSO

that ranged from ⫺299 to 1065 ␮s (mean

CD, 186 ␮s; SD, 451 ␮s; n ⫽ 8). Eleven of
13 neurons exhibited IE binaural interac-
tion as evidenced by their ILD sensitivity;
three were documented audiovisually, and
the eight neurons tested with ILD func-
tions showed the characteristic sigmoi-
dally shaped rate–ILD functions illus-
trated in Figure 8 B. Of the 11 units that
were tested for both ILD sensitivity and
IPD sensitivity, nine were sensitive to both
in a manner consistent with IE binaural
interaction, and two responded to IPDs
consistent with IE interactions but were
not modulated by ILDs over the range
tested. Of the seven units that were tested
with both binaural beats and tested for
sensitivity to ITDs in noise, five were mod-
ulated by both noise delays and IPDs con-
sistent with IE interaction, and two were
modulated by IPDs consistent with IE in-
teraction but not modulated by noise de-
lays. Although the noise– delay functions
based on discharge rate for these latter two
units did not change systematically with
ITD, close examination of the dot rasters
of the responses (data not shown) indi-
cated that there was some form of binaural
interaction, as evidenced by the systematic
appearance and disappearance of re-
sponses as a function of ITD. Such re-
sponses would not be expected for a mon-
aural unit. Finally, one neuron was
monaural, with no ILD or IPD sensitivity,
and could only be stimulated ipsilaterally. Figure 7. Another example of a low-characteristic-frequency LSO neuron that is sensitive to IPDs and ITDs in broadband noise
stimuli. A, Frequency–threshold curve (characteristic frequency, 1.35 kHz). B–D, Same as in Figure 6, C and D. Here, the noise
Discussion delay function has a response minimum for an ITD of 0 ␮s, compared with the delay predicted from the CD of ⫺299 ␮s computed
Previous evidence for low-frequency from B and C. ␾, Mean interaural phase; ␾b, mean interaural phase from binaural-beat stimuli; *p ⬍ 0.001. Error bars represent
IPD and ILD sensitivity in LSO neurons ⫾1 SEM. SPL, Sound pressure level; Contra, contralateral; Ipsi, ipsilateral.
Reports of low-frequency LSO neurons
with IPD sensitivity are sparse (Caird and
Klinke, 1983; Finlayson and Caspary, 1991; Joris and Yin, 1995;
Batra et al., 1997). In the most comprehensive study, Finlayson
and Caspary (1991) recorded from 96 LSO neurons with charac-
teristic frequencies ⬍1.2 kHz in chinchilla. These neurons were
sensitive to ILDs, and their responses could be modulated by
IPDs. However, 31 of 61 neurons tested with both in-phase and
out-of-phase stimuli showed maximum inhibition to stimuli
180° out-of-phase, rather than in-phase as would be predicted
from IE binaural interaction. Thirty-two additional cells in their
study that were only tested with in-phase stimuli had strong in-
hibition to in-phase tones consistent with our data.
Batra et al. (1997) were surprised not to find any low-
frequency neurons with IE interaction in the brainstem of the Figure 8. Phase–frequency (A) and ILD (B) functions for the population of low-
characteristic-frequency LSO neurons. In support of the hypothesis that low-characteristic-
awake rabbit. It is not clear whether they tested for ILD sensitivity
frequency LSO neurons can be sensitive to IPDs and ILDs in manner consistent with IE binaural
by stimulating the ipsilateral ear and looking for inhibitory input interaction, the phase–frequency functions of these neurons yield CPs near 0.5 cycles (A), and
from stimulation of the contralateral ear, as we have done. Low- all neurons tested were sensitive to ILD (B). ␾b, Mean interaural phase from binaural-beat
frequency LSO neurons showing ILD sensitivity have been re- stimuli. Error bars represent ⫾1 SEM. Ipsi, Ipsilateral; Contra, contralateral.
ported occasionally (Caird and Klinke, 1983; Finlayson and
Caspary, 1991; Joris and Yin, 1995; Spitzer and Semple, 1995). In
general, these studies examined either IPD or ILD sensitivity but LSO where neurons were reported to be predominantly monau-
rarely both. All of these reports, in addition to our data, differ ral (Boudreau and Tsuchitani, 1968; Guinan et al., 1972a,b; Tsu-
from the classical descriptions of the low-frequency region of the chitani, 1977). Only 1 of 16 neurons in our sample was monaural.
Tollin and Yin • Interaural Phase Coding in the LSO J. Neurosci., November 16, 2005 • 25(46):10648 –10657 • 10655

The average CD of our LSO neurons was ⫹186 ␮s and is in sitivity in low-frequency inferior colliculus neurons is consistent
agreement with delay measurements made in high-frequency with the under-representation and over-representation of low-
LSO neurons. The high-frequency data were obtained using frequency neurons in LSO and medial superior olive, respectively
amplitude-modulated stimuli, because high-frequency neurons (Guinan et al., 1972b). Some inferior colliculus neurons exhibit
phase lock to the low-frequency envelope of amplitude-modulated nonlinear phase–frequency functions when studied with binau-
stimuli with high-frequency carriers but not to pure tones. For 15 ral beats (Stanford et al., 1992; McAlpine et al., 1998). The shapes
high-frequency LSO neurons, the delay was calculated from the re- of these functions could be explained by converging inputs from
sponse to monaural amplitude modulated stimuli from the slope of medial superior olive-like neurons and LSO-like neurons (Shack-
the phase–frequency plot. The difference in the slopes for ipsilateral leton et al., 2000) that exhibited IPD sensitivity like that demon-
and contralateral stimulation averaged ⫹182 ␮s with the contralat- strated here.
eral response delayed (Joris and Yin, 1998). For binaural stimulation
of high-frequency LSO neurons with amplitude-modulated tones, Functional considerations
Joris (1996) found that the mean CD was 200 ␮s. These three studies According to the duplex theory, localization of high-frequency
show that although the contralateral inhibitory input to LSO has a sounds is based on ILDs. What is the function, then, of low-
longer overall path length through the brainstem and an additional frequency LSO neurons that are sensitive to ILDs and IPDs? First,
synapse in the MNTB, the inhibition lags the ipsilateral excitation by acoustic measurements of ILD cues in cats reveal moderately
only ⬃190 ␮s in both the high- and low-frequency limbs of the LSO. sized ILDs of ⫾5–10 dB, even for frequencies ⬍2.0 kHz (Wiener
This remarkable finding is likely caused by the fast synapse through et al., 1966; Tollin, 2004). The LSO neurons in our sample could
the calyx of Held from the contralateral globular bushy cells to the encode these ILDs (Fig. 8 B). It is unknown whether cats use
MNTB (Joris and Yin, 1998) and the large caliber axons of the glob- low-frequency ILDs for localization. The IE mechanisms of low-
ular bushy cells (Spangler et al., 1985). In fact, three of eight of our frequency LSO neurons could allow for other binaural computa-
neurons and 7 of 21 of the high-frequency LSO cells in the study by tions. For example, the classical model of binaural hearing based
Joris (1996) had CDs ⬍0, indicating that the contralateral inhibition on ITDs is the Jeffress coincidence detection model (Jeffress,
led the ipsilateral excitation. 1948). The anatomical correlate of the Jeffress model is usually
The reports by Boudreau and Tsuchitani (1968) and Guinan based on EE binaural interaction occurring at the medial superior
et al. (1972a,b) of predominantly monaural responses in low- olive (Yin and Chan, 1990). However, inhibitory mechanisms
frequency LSO neurons were followed by anatomical studies that may shape the ITD selectivity of medial superior olive neurons
tended to support those findings. The projections from MNTB to (Brand et al., 2002; Grothe, 2003). Anatomical evidence suggests
the lateral limb of the LSO were less dense (Glendenning et al., that inhibitory input to medial superior olive derives in part from
1985, 1991; Spangler et al., 1985; Sanes et al., 1987; Saint Marie et the MNTB (Kuwabara and Zook, 1991, 1992; Banks and Smith,
al., 1989), and the gradient of glycine receptors was less dense in 1992; Smith, 1995; Smith et al., 1998). The enhanced phase lock-
the lateral than the more medial limbs (Sanes and Wooten, 1987). ing by MNTB neurons to low-frequency stimuli may play an
On the other hand, there was also substantial physiological important role in the encoding of ITD by medial superior olive
(discussed previously) and anatomical evidence predicting the neurons.
existence of IPD- and ILD-sensitive LSO neurons in the lateral Our data suggest that it might also be possible that excitatory
limb. First, these neurons receive excitatory input from low- inputs to LSO from the contralateral cochlear nucleus (Glenden-
frequency neurons in the ipsilateral cochlear nucleus and ning et al., 1985) shaped the IPD selectivity observed here. For
frequency-matched inhibitory input from the ipsilateral MNTB, example, although pure IE-type IPD sensitivity would produce
which receives excitatory input from low-frequency neurons in CPs of exactly 0.5 cycles and linear phase–frequency plots, our
the contralateral cochlear nucleus (Glendenning et al., 1985; data exhibited a range of CPs around 0.5 (Fig. 8 A), and some
Smith et al., 1993, 1998). Second, low-frequency cochlear nucleus neurons (Fig. 8 A, 232-15) exhibited nonlinear phase–frequency
neurons (Lavine, 1971; Bourk, 1976; Palmer et al., 1986; Black- plots. Such alterations in CP and phase–frequency linearity are
burn and Sachs, 1989; Joris et al., 1994) exhibit highly phase- consistent with the model by Batra et al. (1997), which is based on
locked responses to low-frequency tones. We demonstrated here the anatomical observations that some LSO neurons can also
that this enhanced phase locking is preserved at the LSO when receive excitatory inputs from the contralateral ear. Finally, low-
stimulated ipsilaterally (Figs. 3, 4). Moreover, we demonstrated frequency neurons in the lateral limb of LSO project predomi-
enhanced phase locking to low-frequency tones in MNTB neu- nantly to the ipsilateral inferior colliculus and often stain posi-
rons (Figs. 2, 4), confirming previous reports from a few cells [cat tively for the inhibitory neurotransmitter glycine (Saint Marie et
(Smith et al., 1998) and gerbil (Kopp-Scheinpflug et al., 2003)]. al., 1989; Glendenning et al., 1992). These inhibitory inputs con-
Finally, the existence of low-frequency IPD-sensitive LSO neu- verge in inferior colliculus with the excitatory projections from
rons has been predicted from studies of low-frequency IPD- the ipsilateral medial superior olive neurons (Loftus et al., 2004).
sensitive neurons in the inferior colliculus. Inferior colliculus It is conceivable that EE and IE ITD processing by the medial
neurons with CPs near 0.5 have been reported in the cat, rabbit, superior olive and LSO, respectively, could interact at the inferior
and guinea pig. However, the proportion of neurons exhibiting colliculus to produce neurons with novel ITD sensitivities with a
CPs in the range consistent with IE binaural interaction (⬃0.25– larger range of CPs and CDs than the medial superior olive and
0.75 cycles) was relatively small with only ⬃20 and ⬃25% in cat LSO produce independently (previous section).
[for characteristic frequencies ⬍1 kHz (Yin and Kuwada, 1983)] In contrast to the Jeffress model, many models of low-
and rabbit (Kuwada et al., 1987) inferior colliculus, respectively, frequency binaural psychophysical phenomena require IE binau-
the rest being distributed around 0.0 cycles, consistent with EE ral mechanisms rather than EE. These models assume neurons
interaction. Only ⬃10% of 520 low-frequency inferior colliculus that respond to ITDs and/or ILDs like the LSO neurons demon-
neurons in the guinea pig (McAlpine et al., 2001) had noise– strated here. For example, models of the binaural-masking-level
delay functions (Fig. 6 E) that exhibited a trough, as would be difference assume a subtractive process, including the equaliza-
expected from IE interaction. The predominance of EE IPD sen- tion– cancellation model (Durlach, 1963; Green, 1966; Breebaart
10656 • J. Neurosci., November 16, 2005 • 25(46):10648 –10657 Tollin and Yin • Interaural Phase Coding in the LSO

et al., 2001). Models of binaural speech segregation (Culling and Grothe B (2003) New roles for synaptic inhibition in sound localization. Nat
Summerfield, 1995) and the illusory binaural pitches (Huggin’s, Rev Neurosci 4:540 –550.
Guinan Jr JJ, Li RYS (1990) Signal processing in brainstem auditory neurons
binaural edge, and Fourcin) require subtractive mechanisms
which receive giant endings (calyces of Held) in the medial nucleus of the
(Fourcin, 1970; Culling et al., 1998a,b; Akeroyd and Summer- trapezoid body. Hear Res 49:321–334.
field, 2000). Models incorporating subtraction from the two ears Guinan Jr JJ, Guinan SS, Norris BE (1972a) Single auditory units in the
have been able to account for phenomena not easily explained by superior olivary complex. I: responses to sounds and classifications based
Jeffress-like EE interaction (de Cheveigne, 1993; Hartmann and on physiological properties. Int J Neurosci 4:101–120.
Zhang, 2003). These models require fast inhibitory interaction of Guinan Jr JJ, Norris BE, Guinan SS (1972b) Single auditory units in the
temporally structured or phase-locked inputs. This is precisely superior olivary complex. II: location of unit categories and tonotopic
organization. Int J Neurosci 4:147–166.
the type of interaction that we found here in low-frequency LSO
Hartmann WM, Zhang PX (2003) Binaural models and the strength of di-
and MNTB. chotic pitches. J Acoust Soc Am 114:3317–3326.
Jeffress LA (1948) A place theory of sound localization. J Comp Physiol
References Psychol 41:35–39.
Akeroyd MA, Summerfield AQ (2000) A fully temporal account of the per- Johnson DH (1980) The relationship between spike rate and synchrony in
ception of dichotic pitches. Br J Audiol 33:106 –107. responses of auditory-nerve fibers to single tones. J Acoust Soc Am
Banks MI, Smith PH (1992) Intracellular recordings from neurobiotin- 68:1115–1122.
labeled cells in brain slices of the rat medial nucleus of the trapezoid body. Joris PX (1996) Envelope coding in the lateral superior olive. II. Character-
J Neurosci 12:2819 –2837. istic delays and comparison with responses in the medial superior olive.
Batra R, Kuwada S, Fitzpatrick DC (1997) Sensitivity to interaural temporal J Neurophysiol 76:2137–2156.
disparities of low- and high-frequency neurons in the superior olivary Joris PX, Yin TCT (1995) Envelope coding in the lateral superior olive. I.
complex. I. Heterogeneity of responses. J Neurophysiol 78:1222–1236. Sensitivity to interaural time differences. J Neurophysiol 73:1043–1062.
Blackburn CC, Sachs MB (1989) Classification of unit types in the antero- Joris PX, Yin TCT (1998) Envelope coding in the lateral superior olive. III.
ventral cochlear nucleus: PST histograms and regularity analysis. J Neu- Comparison with afferent pathways. J Neurophysiol 79:253–269.
rophysiol 62:1303–1329. Joris PX, Carney LH, Smith PH, Yin TCT (1994) Enhancement of neural
Boudreau JC, Tsuchitani C (1968) Binaural interaction in the cat superior synchronization in the anteroventral cochlear nucleus. I. Responses to
olive S segment. J Neurophysiol 31:442– 454. tones at the characteristic frequency. J Neurophysiol 71:1022–1036.
Bourk TR (1976) Electrical responses of neural units in the anteroventral
Kiang NYS, Moxon EC, Levine RA (1970) Auditory-nerve activity in cats
cochlear nucleus of the cat. PhD thesis, Massachusetts Institute of
with normal and abnormal cochleas. In: Sensorineural hearing loss (Wol-
Technology.
stenholme GEW, Knight J, eds), pp 241–273. London: Churchill.
Brand A, Behrend O, Marquardt T, McAlpine D, Grothe B (2002) Precise
Kopp-Scheinpflug C, Lippe WR, Dorrscheidt GJ, Rubsamen R (2003) The
inhibition is essential for microsecond interaural time difference coding.
medial nucleus of the trapezoid body in the gerbil is more than a relay:
Nature 417:543–547.
comparison of pre- and postsynaptic activity. J Assoc Res Otolaryngol
Breebaart J, van de Par S, Kohlrausch A (2001) Binaural processing model
4:1–23.
based on contralateral inhibition. I. Model structure. J Acoust Soc Am
Kuwabara N, Zook JM (1991) Classification of the principal cells of the
110:1074 –1088.
medial nucleus of the trapezoid body. J Comp Neurol 314:707–720.
Caird D, Klinke R (1983) Processing of binaural stimuli by cat superior
Kuwabara N, Zook JM (1992) Projections to the medial superior olive from
olivary complex neurons. Exp Brain Res 52:385–399.
the medial and lateral nuclei of the trapezoid body in rodents and bats.
Culling JF, Summerfield Q (1995) Perceptual separation of concurrent
J Comp Neurol 324:522–538.
speech sounds: absence of across-frequency grouping by common inter-
aural delay. J Acoust Soc Am 98:785–797. Kuwada S, Yin TC, Wickesberg RE (1979) Response of cat inferior colliculus
Culling JF, Summerfield AQ, Marshall DH (1998a) Dichotic pitches as illu- neurons to binaural beat stimuli: possible mechanisms for sound local-
sions of binaural unmasking. I. Huggins’ pitch and the “binaural edge ization. Science 206:586 –588.
pitch.” J Acoust Soc Am 103:3509 –3526. Kuwada S, Stanford TR, Batra B (1987) Interaural phase-sensitive units in
Culling JF, Marshall DH, Summerfield AQ (1998b) Dichotic pitches as illu- the inferior colliculus of the unanesthetized rabbit: effects of changing
sions of binaural unmasking. II. The Fourcin pitch and the dichotic rep- frequency. J Neurophysiol 57:1338 –1360.
etition pitch. J Acoust Soc Am 103:3527–3539. Lavine RA (1971) Phase-locking in response of single neurons in cochlear
de Cheveigne A (1993) Separation of concurrent harmonic sounds: funda- nuclear complex of the cat to low-frequency tonal stimuli. J Neurophysiol
mental frequency estimation and a time-domain cancellation model of 34:467– 483.
auditory processing. J Acoust Soc Am 93:3271–3290. Loftus WC, Bishop DC, Saint Marie RL, Oliver DL (2004) Organization of
Durlach NI (1963) Equalization and cancellation theory of binaural binaural excitatory and inhibitory inputs to the inferior colliculus from
masking-level differences. J Acoust Soc Am 35:1206 –1218. the superior olive. J Comp Neurol 472:330 –344.
Finlayson PG, Caspary DM (1991) Low-frequency neurons in the lateral Mardia KV (1972) Statistics of directional data. London: Academic.
superior olive exhibit phase-sensitive binaural inhibition. J Neurophysiol McAlpine D, Jiang D, Shackleton TM, Palmer AR (1998) Convergent input
65:598 – 605. from brainstem coincidence detectors onto delay-sensitive neurons in the
Fourcin AJ (1970) Central pitch and auditory localization. In: Frequency inferior colliculus. J Neurosci 18:6026 – 6039.
analysis and periodicity detection in hearing (Plomp R, Smoorenburg GF, McAlpine D, Jiang D, Palmer AR (2001) A neural code for low-frequency
eds), pp 319 –328. Leiden, Germany: Sijthoff. sound localization in mammals. Nat Neurosci 4:396 – 401.
Glendenning KK, Hutson KA, Nudo RJ, Masterton RB (1985) Acoustic chi- Mills W (1958) On the minimum audible angle. J Acoust Soc Am 30:237–
asm II: anatomical basis of binaurality in lateral superior olive of cat. 246.
J Comp Neurol 232:261–285. Palmer AR, Winter IM, Darwin CJ (1986) The representation of steady-
Glendenning KK, Masterton RB, Baker BN, Wenthold RJ (1991) Acoustic state vowel sounds in the temporal discharge patterns of the guinea pig
chiasm. III: nature, distribution, and sources of afferents to the lateral cochlear nerve and primarylike cochlear nucleus neurons. J Acoust Soc
superior olive in the cat. J Comp Neurol 310:377– 400. Am 79:100 –113.
Glendenning KK, Baker BN, Hutson KA, Masterton RB (1992) Acoustic Rayleigh L (1907) On our perception of sound direction. Philos Mag (6 ser.)
chiasm V: inhibition and excitation in the ipsilateral and contralateral 13:214 –232.
projections of LSO. J Comp Neurol 319:100 –122. Rhode WS (1976) A digital system for auditory neurophysiological re-
Goldberg JM, Brown PB (1969) Response of binaural neurons of dog supe- search. In: Current computer technology in neurobiology (Brown P, ed),
rior olivary complex to dichotic tonal stimuli: some physiological mech- pp 543–567. Washington, DC: Hemisphere.
anisms of sound localization. J Neurophysiol 32:613– 636. Saint Marie RL, Ostapoff EM, Morest DK, Wenthold RJ (1989) Glycine-
Green DM (1966) Signal detection analysis of EC model. J Acoust Soc Am immunoreactive projection of the cat lateral superior olive: possible role
40:833– 838. in midbrain ear dominance. J Comp Neurol 279:382–396.
Tollin and Yin • Interaural Phase Coding in the LSO J. Neurosci., November 16, 2005 • 25(46):10648 –10657 • 10657

Sanes DH, Wooten GF (1987) Development of glycine receptor distribution sensitivity of neurons in the inferior colliculus and thalamus of the un-
in the lateral superior olive of the gerbil. J Neurosci 7:3803–3811. anesthetized rabbit. J Neurosci 12:3200 –3216.
Shackleton TM, McAlpine D, Palmer AR (2000) Modelling convergent in- Stevens SS, Newman EB (1936) The localization of actual sources of sound.
put onto interaural-delay-sensitive inferior colliculus neurones. Hear Res Am J Psychol 48:297–306.
149:199 –215. Tollin DJ (2003) The lateral superior olive: a functional role in sound source
Smith PH (1995) Structural and functional differences distinguish principal localization. Neuroscientist 9:127–143.
from nonprincipal cells in the guinea pig MSO slice. J Neurophysiol Tollin DJ (2004) The development of the acoustical cues to sound location
73:1653–1667. in cats. Assoc Res Otol Abstr 27:161.
Smith PH, Joris PX, Yin TCT (1993) Projections of physiologically character- Tsuchitani C (1977) Functional organization of lateral cell groups of cat
ized spherical bushy cell axons from the cochlear nucleus of the cat: evidence
superior olivary complex. J Neurophysiol 40:296 –318.
for delay lines to the medial superior olive. J Comp Neurol 331:245–260.
Wiener FM, Pfeiffer RR, Backus ASN (1966) On the sound pressure trans-
Smith PH, Joris PX, Yin TCT (1998) Anatomy and physiology of principal
formation by the head and auditory meatus of the cat. Acta Otolaryngol
cells of the medial nucleus of the trapezoid body (MNTB) of the cat.
J Neurophysiol 79:3127–3142. 61:255–269.
Spangler KM, Warr WB, Henkel CK (1985) The projections of principal Yin TCT (2002) Neural mechanisms of encoding localization cues in the audi-
cells of the medial nucleus of the trapezoid body in the cat. J Comp Neurol tory brainstem. In: Integrative functions in the mammalian auditory path-
238:249 –262. way, (Oertel D, Fay RR, Popper AN, eds), pp 99 –159. New York: Springer.
Spitzer MW, Semple MN (1995) Neurons sensitive to interaural phase dis- Yin TCT, Chan JC (1990) Interaural time sensitivity in medial superior olive
parity in gerbil superior olive: diverse monaural and temporal response of cat. J Neurophysiol 64:465– 488.
properties. J Neurophysiol 73:1668 –1690. Yin TCT, Kuwada S (1983) Binaural interaction in inferior colliculus of the
Stanford TR, Kuwada S, Batra R (1992) A comparison of the interaural time cat. III. Effects of changing frequency. J Neurophysiol 50:1020 –1042.