| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1Department of Physiology and 2The Ear Institute at University College London, London, United Kingdom; and 3Departments of Neuroscience and Surgery, University of Connecticut Health Center, Farmington, Connecticut
Submitted 4 August 2004; accepted in final form 27 January 2005
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
), human sound localization is subserved by 2 mechanisms: At low frequencies (<1,500 Hz), sources of sounds are localized using interaural time differences (ITDs), whereas at high frequencies interaural intensity differences (IIDs) mediate localization. The dichotomy suggested by the duplex theory is, to a first approximation, reflected in the anatomy and physiology underlying sensitivity to ITDs and IIDs in mammals (Joris et al. 1998; Tollin 2003). Electrophysiological recordings confirm that phase-locked action potentials converge from each cochlear nucleus onto coincidence detection neurons in the medial superior olive (MSO), generating sensitivity to ITDs (Batra et al. 1997a; Goldberg and Brown 1969; Spitzer and Semple 1995; Yin and Chan 1990). Sensitivity to IIDs occurs in lateral superior olive (LSO) neurons, which receive excitatory ipsilateral and inhibitory contralateral inputs (Boudreau and Tsuchitani 1968).
These brain stem pathways are not wholly independent; e.g., both involve glycinergic inhibition from the medial nucleus of the trapezoid body (Brand et al. 2002; Smith et al. 1998). Physiological recordings suggest that LSO neurons do not exclusively encode IIDs and MSO neurons do not exclusively encode ITDs. Rather, some overlap occurs (Batra et al. 1997a; Caird and Klinke 1983; Joris and Yin 1995). The strict dichotomy suggested by the duplex theory is also contradicted by psychophysical experiments that demonstrate sensitivity to ITDs conveyed by the envelopes of high-frequency complex sounds (David et al. 1959; Henning 1974; Klumpp and Eady 1956; McFadden and Pasanen 1976; Yost et al. 1971).
Human sensitivity to changes in ITD conveyed by high-frequency stimuli has typically been found to be poorer than that measured with low-frequency stimuli (Bernstein and Trahiotis 1982; Blauert 1997; Jones and Williams 1981; Yost et al. 1971). In addition, the lateral extent of intracranial images produced by ITDs conveyed by high-frequency, complex stimuli is typically smaller than that conveyed by low-frequency stimuli (Bernstein and Trahiotis 1985). The relatively poor sensitivity to ITDs at high frequencies may be explained by a lack of specialization for the processing of envelope ITDs within central auditory centers. In the LSO, ITD sensitivity in response to high-frequency stimuli has been considered a by-product of neural circuitry specialized for IID sensitivity (Joris and Yin 1995, 1998; Tollin 2003), whereas the MSO is viewed as being specialized for detection of ITDs in low-frequency stimuli.
Colburn and Equissaud (1976) hypothesized that the differences in ITD sensitivity at low and high frequencies observed psychophysically could be accounted for by differential effects of peripheral processing on low- and high-frequency stimuli. Firing of action potentials in auditory nerve fibers (ANFs) can be modeled by band-pass filtering, half-wave rectification, and low-pass filtering of the sound waveform (Geisler 1998). This produces a "phase-locked" response to low-frequency tones (Fig. 1A). When a high-frequency sound is modulated in amplitude, ANF responses exhibit phase locking to the envelope (Johnson 1980; Joris and Yin 1992; Palmer 1982; Palmer and Russell 1986). Sinusoidally amplitude modulated (SAM) tones produce firing patterns in ANFs that follow the sinusoidal envelope (Fig. 1B). Although this timing information supports sensitivity to ITDs, the firing pattern lacks the distinct "OFF periods" (where the firing probability is zero), produced by half-wave rectification, that characterize responses to low-frequency tones (compare Fig. 1, A and B).
|
demonstrated, using "transposed" tones, that ITD sensitivity at high frequencies could be comparable to ITD sensitivity at low frequencies. Transposed stimuli were first designed by van de Par and Kohlrausch (1997) to provide high-frequency ANFs with envelope-based information similar to the waveform-based information available from low-frequency tones. The presumption is that, despite the differences in the sound pressure waveforms, the firing probability in ANFs is similar for high-frequency transposed tones and low-frequency tones (Fig. 1C).
The current study investigated neural responses to ITDs conveyed by high-frequency SAM and transposed tones; extracellular recordings were obtained from single neurons in the inferior colliculus (IC) of the guinea pig. Neural responses reflected greater sensitivity to ITDs for transposed tones than for SAM tones. Consistent with the psychophysical findings of Bernstein and Trahiotis (2002), neural discrimination of ITDs within transposed tones was comparable to neural discrimination of ITDs within low-frequency tones. Also consistent with these, and other, psychophysical (Bernstein and Trahiotis 1994; McFadden and Pasanen 1976; Nuetzel and Hafter 1981) and physiological (Joris and Yin 1998) findings, ITD sensitivity in neurons was limited to low-modulation frequencies (
250 Hz) for both SAM and transposed tones. Our results suggest that, for those low rates of modulation, the central mechanisms that mediate sensitivity to envelope-based ITDs are essentially equivalent to those that mediate sensitivity to fine-structure–based ITDs.
Parts of this work were previously published in abstract form.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
All experiments were carried out in accordance with the Animal (Scientific Procedures) Act of 1986 of Great Britain and Northern Ireland. Single-neuron recordings were made from the central nucleus of the right IC of 27 adult guinea pigs under urethane anesthesia [Sigma-Aldrich, Poole, UK; 25% solution of 0.9% NaCl; 1 g/kg, administered intraperitoneally (ip)]. Additional analgesia was provided using fentanyl–fluanisone (Hypnorm; Janssen-Cilag, High Wycome, UK; 0.1 ml, administered intramuscularly), supplementary doses of which were administered as required. Atropine sulfate [Animalcare, York, UK; 0.6 mg/ml; 0.1 ml, administered subcutaneously (sc)] was dispensed to reduce bronchial secretions, and lidocaine hydrochloride (Martindale Pharmaceuticals, Romford, UK; 2%, sc) was administered locally before any surgical incision. A tracheal cannula was inserted and core temperature was maintained at 37°C with a heating blanket and rectal probe (Harvard Apparatus, Kent, UK).
Animals were placed in a sound-attenuating chamber (IAC, Winchester, UK) and held in a stereotaxic frame with hollow ear speculae (modified from model 1730, David Kopf Instruments, Tujunga, CA). Before positioning the animal, the tragus was cut to obtain clear access to the tympanic membrane. A craniotomy was performed to expose the cortex overlying the IC, the covering dura was removed, and agar (about 2%) was applied to prevent drying and deterioration of the cortex. The bullae were vented to equalize air pressure in the middle ears, by insertion of cannulae and sealing with petroleum jelly. A parylene-coated tungsten microelectrode [1–5 M
; World Precision Instruments, Sarasota, FL; or made in house (Bullock et al. 1988)] was positioned stereotaxically 2 mm above the IC (Medvedeva 1977) and advanced ventrally from outside the recording chamber using a piezo-stepped microdrive (Burleigh Instruments, Westbury, NY). At the end of each experiment, animals were administered a lethal dose of sodium pentobarbitone (Pentoject; Animalcare; 60 mg/ml; 1–2 ml, ip or Sagatal, 60 mg/ml).
Stimulus production and presentation
Sounds were produced using Tucker Davis Technologies (TDT, Alachua, FL) digital signal processing hardware. Tonal stimuli used for the isolation and characterization of single units were generated using custom software (T. Shackleton and A. Palmer, MRC Institute of Hearing Research, Nottingham, UK; 100-kHz sampling rate) and TDT system II hardware. TDT Brainware, Real Time Processor Visual Design Studio (RPvds), and system III hardware were used to generate the SAM and transposed stimuli (50-kHz sampling rate).
Stimuli were generated and scaled such that their peak voltages were at the maximum available voltage (±10 V) of the D/A converters (DACs). The outputs were attenuated to achieve the desired level for the experiments using PA4 (system II) or PA5 (system III) modules (TDT). (In initial experiments using system III, the attenuation was achieved by scaling within the signal-generation software.) Fixed amplification [Rotel (Worthing, UK) RB971 power amplifier or a Beyerdynamic (Burgess Hill, UK) A150 Blueprint stereo-amplifier] was followed by 60 dB of "final" attenuation. Using a high-level signal followed by attenuation maximized the signal-to-noise ratio through the stimulus-generation pathway. Specifically, this technique yielded a greater signal-to-noise ratio than if amplification alone were used to achieve the desired intensity. Sounds were delivered by Beyerdynamic DT48A loudspeaker drivers, modified and fitted with a probe-tube attachment to allow insertion and sealing into the hollow ear speculae. Knowles Acoustics (Burgess Hill, UK) FG3452 microphones [attached to steel tubes placed in the ear speculae and calibrated against a Brüel & Kjær (Stevenage, UK) Type 4136 -in. microphone] were used to measure the stimulus within a few millimeters of the tympanic membrane to ensure that the sounds delivered to each ear were well matched.
Spike collection
Electrical signals from the electrode were conducted by a headstage to a preamplifier (TDT Medusa RA16PA) where they were amplified and digitized at a 25-kHz sampling rate. The signal was then conducted by a fiber-optic cable to the RA16 base station, which produced fixed amplification and filtering (x1,000 gain, 300-Hz high-pass filter, 10-kHz low-pass filter, and 50-Hz notch filter).
Using system II hardware, a spike voltage–level discriminator (TDT ET1) was used to detect spikes from the background noise. Spikes were monitored on a Tektronix TD210 oscilloscope for continuity of spike characteristics (i.e., shape and amplitude) to ensure recordings were from a single neuron.
Using system III hardware, spike data passed from the RA16 base station to TDT Brainware and spikes that crossed a user-defined trigger level were counted. Spikes were then additionally sorted, in Brainware, according to spike characteristics, to ensure data were collected from a single neuron.
Sound stimuli
Isolation and characterization stimuli. Binaural "search stimuli," consisting of 50-ms presentations of diotic single tones, repeated at a rate of 5/s, were used to isolate neurons and estimate their threshold (the lowest-intensity sound required to evoke firing, determined audiovisually, by the experimenter) and characteristic frequency (CF; the frequency at which the lowest threshold was obtained). The CF and threshold were confirmed by recording a frequency-versus-level response area spectrally flanking the CF estimate (2 octaves above and 4 octaves below CF) and between 10 and 90 dB of attenuation from the maximum system output of about 110 dB SPL. If the unit remained isolated for a sufficient length of time, neurons were further characterized by presentation of 50- or 200-ms diotic, ipsilateral, and contralateral tones with frequency equal to CF. Tones were presented at 20 dB above threshold and 100 or 150 repetitions were presented.
Sam and transposed tones. Two parameters of the stimuli were varied: ITD (of primary interest) and frequency of modulation (fm). SAM and transposed stimuli were constructed with a carrier frequency (fc; in sine phase) equal to neuronal CF and were modulated by multiplication with either a lower-frequency, DC-shifted, sinusoidal waveform (in sine phase) to produce a SAM tone (Fig. 1B) or a half-rectified sinusoid (90° phase advanced with respect to the SAM modulating tone) to produce a transposed tone (Fig. 1C). Both stimuli were modulated by 100%. For each neuron, the peak voltages of SAM and transposed tones presented were equal and between 10 to 30 dB above the peak voltage of the sinusoid at CF that defined threshold. For 75/82 neurons, peak voltages were 20 dB above that of the tone at threshold. This choice resulted in the SAM and transposed tones being presented at 15.7 and 14.0 dB, respectively, above the rms level of the tone. In all cases, the rms levels of the transposed tones were within 2 dB of their SAM counterparts.
ITD generation
Stimuli were 500 ms in duration and were gated ON and OFF with 2-ms cosine-squared ramps. An ITD was created by delaying the entire waveform (both fine-structure and envelope) in one channel and advancing it in the other by an equal amount. The resulting stimuli thus contained onset, ongoing, and offset ITDs. A positive ITD was created by delaying the stimulus in the ipsilateral (left) ear and advancing the stimulus in the contralateral ear. A negative ITD was created by delaying the stimulus in the contralateral ear and advancing the stimulus in the ipsilateral ear. The values of ITD that were chosen depended on the fm of the stimuli such that 17 steps of an interaural phase difference (IPD) were presented over ±0.5 cycle of the fm. The IPDs applied remained constant irrespective of the fm, ensuring that the neural tuning was measured over a complete cycle of each fm. Data are also presented from experiments in which the stimuli were either 800 ms in duration or, alternatively, were 500 ms in duration with no onset or offset ITD. In the latter case, stimuli were gated ON and OFF at the same time in both ears.
Modulation frequency
Neural responses from one neuron, to any one fm and over all 17 IPDs, will be referred to as a single "recording." SAM and transposed tones were both presented at each fm. Initially, for each neuron, 6 or 7 fm values were presented pseudo-randomly at each IPD. Modulation frequencies varied between 10 and 640 Hz (or CF/6 if this was lower) in logarithmic steps or between 10 and 650 Hz (or CF/6) in linear steps. The rationale for limiting the fm to CF/6 is explained below. If the neuron remained well isolated, further data sets were collected at more narrowly spaced fm. For 11 neurons, in later experiments, a smaller range of fm was presented (between 10 and 200–250 Hz) because it was over this range of fm that sensitivity to ITDs was most likely. At least 3 (3–6, median = 3) repetitions of each stimulus condition were presented. The exact fm values presented were determined by the experimenter according to the response of the neuron under investigation.
Spectral components and the fm limitation
SAM tones are characterized by 3 frequency components: the carrier (fc) and 2 "sidebands" [(fc + fm) and (fc – fm)]. Transposed tones have additional sidebands spaced at multiples of 2 x fm (Fig. 2). In psychophysical experiments van de Par and Kohlrausch (1997) demonstrated that the 5 central frequency components of a high-frequency transposed stimulus were sufficient to yield improved binaural performance at high frequencies. For transmission of the temporal structure of the modulation to a single ANF, the spectral components of SAM tones and the central 5 spectral components of transposed tones must fall within the ANF’s effective "filter" or spectral receptive field. As the fm increases, sidebands are attenuated by the ANF’s filter in proportion to their spectral "distance" from the CF. This alters the temporal structure of the modulation transmitted to ANFs. Accordingly, the fm in the present study was limited to a maximum of CF/6 to include fm values at, and slightly above, the range over which we expected that the temporal structure would be preserved. This was judged from the shape of the frequency-versus-level response areas obtained from each neuron, and from reports of the equivalent rectangular bandwidth (ERB) of guinea pig ANFs (Evans et al. 1992). The maximum fm was also limited to 640 or 650 Hz because this included, and slightly exceeded, the range of fm at which sensitivity to ITDs has been observed at high frequencies (e.g., Bernstein and Trahiotis 1994; Joris and Yin 1998).
|
limited the spectra of transposed stimuli to prevent the use of energy outside the psychophysically determined "auditory filter" (see Moore 1997) surrounding the center frequency of the signal under investigation. They low-pass filtered their half-wave rectified tones before multiplication with high-frequency carriers. It was not necessary to impose such a spectral limitation in the present study because we investigated the response of single IC neurons. The spectral components influencing single neurons’ responses are necessarily limited by their spectral receptive fields. Spectral components distant from fc (outside the central 5 components) are unlikely to contribute any appreciable energy to a neuron with a CF equal to fc and at the intensities at which transposed tones were presented (10–30 dB above pure tone threshold). High-resolution functions
"High-resolution" IPD functions were collected from 14 neurons. A value of fm was selected to which a given neuron exhibited a substantial modulation in its firing rate as a function of IPD (i.e., was found to be IPD sensitive). SAM and transposed tones with 101 (or 102) IPDs between +0.5 and –0.5 cycle, including an onset and offset ITD, were presented at this fm. The duration of the stimulus was 500 ms and 
6 repetitions (6–13, median = 10) were presented, pseudo-randomly, of each stimulus condition.
Data analysis
Spike times were exported into Matlab 6.5 (The MathWorks, Natick, MA) for off-line analysis.
Sensitivity to itds. Spike rates were determined by counting spikes over 2 time windows. First, spike rates were calculated by counting the number of spikes occurring during a 600-ms window from the onset of the stimulus at one ear to the offset of the stimulus at the other ear. This took account of the fact that for the lowest modulation rate used (10 Hz) with an ITD equivalent to 0.5 cycle of interaural phase difference, the 500-ms stimulus was gated ON at one ear 50 ms before the stimulus was gated ON at the other ear, and gated OFF 50 ms before the stimulus. Thus spikes counted over this time period included the onset and offset response to the stimuli, which could include responses to monaural stimulation at large values of ITD. Second, spikes were also counted over the middle 350 ms of the stimulus presentation, and over a time period that ensured a whole number of modulation periods within the 350 ms. This middle time period was used to exclude neural responses to the onset and offset of the stimuli and to eliminate periods of monaural stimulation at large values of ITD.
In a small number of early experiments the stimulus duration was either 800 or 500 ms but IPDs were generated with no onset ITD. Recordings from these experiments (from 13 neurons) were analyzed over the middle 350 ms of the stimulus and are included in the current study.
Tuning to ITDs was calculated from the mean spike rates at each fm and over ±0.5 cycle of IPD. The following calculations were performed for each fm (i.e., for each recording) in response to SAM and transposed tones. The mean phase vector was calculated using the method of Goldberg and Brown (1969). At every fm, there were 2 measurements for an IPD of 0.5 cycle (at +0.5 and –0.5 cycle) and the mean of these was used in the calculation. The mean phase vector was termed the "best" phase (BP); this was considered the IPD to which the recording was tuned. The normalized length of the vector (the vector strength, which varied between 0 and 1) was used to calculate the Rayleigh coefficient (Kuwada et al. 1987). The Rayleigh coefficient was the test static for a chi-squared distribution with 2 degrees of freedom (Mardia and Jupp 1999). An individual recording was classified as "ITD sensitive" if the Rayleigh coefficient was >13.815, P < 0.01 (Mardia and Jupp 1999; Yin and Kuwada 1983a), indicating that firing rates were not evenly distributed around a cycle of IPD. Individual neurons were classified as ITD sensitive if one or more recordings were classified as ITD sensitive.
High-resolution functions.
High-resolution IPD functions were obtained to calculate neural discrimination thresholds. ROC (receiver operating characteristic) analysis was used to estimate the smallest ITD that could be discriminated from 0 ITD, using the firing rates of a single neuron. IPDs were converted into ITDs with reference to the modulation frequency. ROC analysis has been used previously (Shackleton et al. 2003; Skottun et al. 2001) to compare neural ITD sensitivity to human psychophysical performance (see Shackleton et al. 2003 for a detailed description of the method.) For this analysis, spike rates were calculated over the middle 350-ms analysis period. To perform ROC analysis, a neurometric function was calculated for each high-resolution IPD function. The distribution of firing rates obtained at 0 ITD was compared with the distribution of firing rates at every IPD presented. The spike rates recorded from all repetitions (6–10) of each ITD form a sample distribution of the spike rates at that ITD. If 101 ITDs were presented, 101 pairs of spike rates are compared, including the comparison of the spike rates at 0 ITD with itself. The neurometric functions describe the probability of randomly selecting a spike rate from each pair of distributions and finding the spike rate at 0 ITD to be lowest.
Neurometric functions were smoothed by averaging over 3 consecutive ITDs. ITDs were considered discriminable from zero ITD if the probability (from the smoothed function) was 0.75 or 
0.25 (giving 2 discrimination thresholds). We defined the just noticeable difference (JND) in ITD as the smallest ITD that is discriminable from zero ITD. No interpolation was carried out to estimate the point where the neurometric functions crossed 0.75 or 0.25; thus JNDs were, if anything, slightly overestimated. For 5 high-resolution functions, 102 (instead of 101) values of IPD were presented. In these cases no response to zero ITD was collected, and discrimination was then calculated from the closest negative ITD to zero.
Period histograms.
Entrainment of spike times to the modulation period for SAM and transposed stimuli was examined by binning spike times over one cycle of the modulation rate. Spike times over the middle 350 ms of the stimulus presentation, and over a time period that ensured a whole number of modulation periods within the 350 ms, were analyzed. The cycles of the modulation period at which spikes occurred were adjusted such that 0 cycle was always with reference to the onset of sound in the contralateral ear. To quantify the degree of phase locking to the stimulus period, at each IPD and fm, the vector strength was calculated. Phase locking was considered significant if the Rayleigh coefficient was >13.815 (P < 0.01) (Mardia and Jupp 1999).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Responses to SAM and transposed tones were obtained from 82 IC neurons with CFs >2 kHz (2.0–13 kHz, median = 4.5 kHz). For 55/82 neurons, peristimulus time histograms (PSTHs) were obtained to monaural and diotic CF tones. All neurons exhibited greater firing rates in response to tones at the contralateral ear than at the ipsilateral ear. For 70% of the sample (39/55), ipsilateral stimulation resulted in inhibition, exhibited as either a reduction in the spontaneous firing rate in response to ipsilateral stimulation or as a reduction in the rate of firing in response to a contralateral sound when stimulation was binaural.
Sixty-nine neurons were investigated using SAM and transposed tones of 500-ms duration with ITDs generated by delaying the entire waveform. Over both analysis windows (350 and 600 ms), more recordings (the response of a single neuron at one fm) showed sensitivity to ITDs within transposed tones than in SAM tones. Correspondingly, more neurons were sensitive to ITDs within transposed than in SAM tones. Responses from a further 13 neurons were recorded using stimuli that were 800 ms in duration or had no onset ITDs. The responses of these 13 neurons were analyzed over the middle 350 ms of the stimulus duration and appeared qualitatively similar to responses recorded from neurons using stimuli of 500 ms (with onset ITDs). They are therefore included in all subsequent analyses examining responses to the middle 350 ms of the stimulus. The first column in Table 1 shows the number of neurons from which recordings were made, as well as the total number of recordings, using analysis windows of 600 and 350 ms. The second column in Table 1 shows the number of neurons and recordings in which ITD sensitivity was observed for transposed, but not SAM, stimuli. The third column shows the number of neurons and recordings in which ITD sensitivity was observed for both transposed and SAM stimuli, and the fourth column shows the number of neurons and recordings in which ITD sensitivity was observed for SAM stimuli only. The final column shows the number of neurons and recordings in which ITD sensitivity was not observed for either transposed or SAM stimuli.
|
). In addition, the modulation in firing rates occurred over a cycle of interaural phase with respect to the modulation frequency and not the carrier frequency. For the 69 neurons from which recordings were made using stimuli of 500-ms duration, more recordings were classified as ITD sensitive when spikes were analyzed over the full (600-ms) analysis window than the middle 350-ms analysis window. Recordings were considered ITD sensitive if the Rayleigh coefficient was >13.815 (see METHODS). However, the vector strengths of the recordings were, on average, lower over the 600-ms analysis window than over the 350-ms analysis window (Wilcoxon rank-sum test; P < 0.001 in response to both SAM and transposed tones). Because the Rayleigh coefficient increases as the vector strength and/or the spike count increases, the likely reason for the increased number of neurons classified as ITD sensitive when the full 600-ms window was analyzed is a result of an increase in the number of spikes counted. This appears to compensate for the reduced vector strength over the 600-ms analysis window than that over the 350-ms analysis window. All subsequent analyses pertain to responses from all 82 neurons, analyzed over the middle 350-ms window of the stimulus.
Responses of a typical IC neuron sensitive to ITDs in the envelope of SAM and transposed tones are shown in Fig. 3. The neuronal CF was 3.4 kHz, which is confirmed by the frequency-versus-level response area (Fig. 3A). PSTHs for monaural and diotic CF tones indicate the neuron to be excited by contralateral stimulation but unresponsive to ipsilateral stimulation (Fig. 3B). The response to diotic stimulation was characterized by a peak spike rate lower than that to contralateral stimulation alone, but longer in duration.
|
Responses to SAM (left) and transposed (right) tones across all fm and IPDs are displayed as 3-dimensional (3D) mesh plots in Fig. 3I. The neuron had a band-pass rate modulation transfer function (rMTF), with maximum firing rates at 110 Hz, a preference for IPD close to zero, and enhanced responses to transposed tones compared with SAM tones; peak discharge rates at the preferred fm were almost double those evoked by SAM tones.
Responses to IPDs imposed on SAM and transposed stimuli for a second IC neuron, with a CF of 4.3 kHz, are shown in Fig. 4. PSTHs in Fig. 4B indicate that contralateral stimulation evoked higher discharge rates than binaural stimulation; ipsilateral stimulation caused discharge rates to fall below the spontaneous rate of the neuron. Raster plots (Fig. 4, C–H) show spike times for 6 recordings in response to both SAM and transposed tones from 10 to 650 Hz fm. Spike times were clearly phase locked to the envelope at 10-Hz modulation. The neuron was sensitive to IPDs in both stimuli at 10 Hz; the pattern of the phase locking to the modulation period changed with IPD and the spike rate fell around 0 IPD. At 138 Hz fm (Fig. 4D), the neuron was IPD sensitive to transposed, but not to SAM, tones. The neuron was insensitive to IPDs in either stimulus at the higher modulation rates (Fig. 4, E–H).
|
Examples of recordings from 2 ITD-insensitive neurons are shown in Figs. 5 and 6. The neuron referred to in Fig. 5 (CF = 3.1 kHz) exhibited a strong onset response to diotic tones, followed by a pause and then a more sustained response (Fig. 5B). The PSTH to contralateral stimulation was similar but with a less well defined pause, whereas ipsilateral stimulation resulted in a small onset response followed by inhibition of spontaneous firing (Fig. 5B). Raster diagrams (Fig. 5, C–H) and the 3D mesh plot (Fig. 5I) indicate that the neuron’s discharge rate increased with increasing fm; that is, the discharge rate was high-pass with respect to fm. At higher values of fm, the discharge rate was greater for transposed than for SAM tones, whereas at lower values of fm, the opposite was the case. Thus the discharge rate exhibited greater modulation as a function of fm in response to transposed tones. Visual inspection suggests some form of binaural interaction occurred at fm values of 10 and 20 Hz; phase locking appeared weaker near zero IPD (Fig. 5, C and D). Further, at 320 Hz, the discharge rate was lower for values of IPD near zero, although the criteria for ITD sensitivity were not met (Fig. 5I).
|
|
Mean firing rates of a further 9 neurons, chosen to illustrate the variety of responses obtained, are displayed as 3D mesh plots in Fig. 7. Neurons classified as sensitive to envelope-based ITDs and with firing rates that peak close to zero IPD are arranged in the top row and neurons with firing rates that peak close to IPDs of 0.5 cycle in the middle row. The bottom row shows neurons insensitive to envelope-based ITDs. For each ITD-sensitive neuron, the maximum Rayleigh coefficient (which was used as a measure of sensitivity to ITDs) was determined across all recordings in response to SAM and transposed tones and both the top and middle rows (Fig. 7, A–C and D–F) are arranged from left to right in order of decreasing maximum Rayleigh coefficient.
|
Discharge rates as a function of IPD were also more modulated in response to transposed tones for the responses in Fig. 7, D and E. Neurons were insensitive to ITDs in the envelope of SAM tones, whereas in Fig. 7, D and E, sensitivity occurred for more than one fm in response to transposed tones. Atypically for our sample of neurons, the firing rate of the neuron in Fig. 7F was more modulated as a function of IPD in SAM tones compared to that in transposed tones at 250 Hz fm (although the reverse was true at 330 Hz). There was a response minimum at 250 Hz fm, which was lower for transposed tones and may explain the reduced modulation in firing rate as a function of IPD to transposed tones. As in Fig. 7, B and C, the IPD tuning in Fig. 7F was similar for SAM and transposed tones (i.e., peaks and troughs in firing rate were similarly positioned) and, for each neuron, there was a modulation rate at which sensitivity was greatest (210 Hz in 7D, 10 Hz in 7E, and 250 Hz in 7F).
The vector strengths of all ITD-sensitive recordings in response to transposed tones were, on average, higher than the vector strengths of the ITD-sensitive recordings in response to SAM tones (Wilcoxon rank-sum test, P = 0.0011), as were the Rayleigh coefficients (calculated using the Rayleigh coefficients from all recordings; Wilcoxon rank-sum test, P < 0.001). Both observations are consistent with transposed tones evoking discharge rates that are better modulated with envelope ITDs than SAM tones.
Neurons for which responses are shown in the bottom row of Fig. 7 were not sensitive to envelope ITDs in SAM or transposed tones. However, their discharge rates were modulated as a function of fm. For each of the 3 neurons, changes in discharge rate as a function of fm were more pronounced in response to transposed tones than to SAM tones. Although not of central interest to the current study, neural sensitivity to the fm was examined, in the form of rate modulation transfer functions (rMTFs). Responses of 71 neurons were examined where SAM and transposed tones were presented with zero ITD and at fm between 10 and 300 Hz. Responses were considered modulated as a function of fm if discharge rates varied by 70% of the maximum discharge rate over the range of fm values tested. In response to SAM tones, 12 neurons did not meet this criterion whereas, in response to transposed tones, only one neuron did not meet the criterion (P < 0.01, test for independence of paired proportions). Further characterization of the rMTFs is not shown because a comprehensive range of fm values was not examined for all neurons. We conclude, however, that neural responses to transposed tones were more modulated as a function of fm.
Sensitivity to ITD as a function of fm
The number of ITD-sensitive recordings as a proportion of the total number of recordings is displayed in Fig. 8, calculated as a function of fm, in 50-Hz bins. Sensitivity to ITDs could be described as band-pass as a function of fm, with the greatest proportion of ITD-sensitive recordings between 60 and 310 Hz fm. The sharpest reduction in the proportion of ITD-sensitive recordings occurred as the fm increased >310 Hz and no recordings were considered sensitive to ITDs at fm >550 Hz. The total number of recordings in each fm bin is shown in the inset of Fig. 8.
|
; McAlpine et al. 2001; Yin and Kuwada 1983b; Yin et al. 1984, 1986).
|
|
). Phase plots (BP plotted as a function of fm) are shown in Fig. 10B for the same neurons for which responses are shown in Fig. 10A. Theoretically, for neurons described as "peak" or "trough" neurons, BP changes systematically with pure-tone frequency or modulation frequency (Batra et al. 1997b; Joris 1996; Rose et al. 1966). The slope of a linear regression describing the relation between BP and frequency is referred to as the characteristic delay (CD) and the intercept with the ordinate axis is referred to as the characteristic phase (CP) (Rose et al. 1966; Yin and Kuwada 1983b). For "peak" neurons, the CD is equal to the ITD at which the neuron responds maximally. A CP of 0 would be predicted from perfect coincidence of excitatory (EE) inputs. A CP of 0.5 would be predicted from coincidence of excitatory and inhibitory inputs (EI). Although it is possible to fit a linear regression to some of the phase plots in Fig. 10B (e.g., Fig. 10B, d and h in response to SAM tones, and in Fig. 10Ba in response to transposed tones have R2 >0.6), for other neurons the linear regression is not appropriate (e.g., Fig. 10Bd). Previous investigations of sensitivity to ITDs of pure tones and high-frequency SAM tones in the IC reported a similar range of neural responses; phase plots are often not well fitted by linear regression and, where phase plots are well fitted by linear regression, CPs may not be close to 0 or 0.5 (Batra et al. 1997; Kuwada et al. 1987; McAlpine et al. 1996; Yin and Kuwada 1983b; Yin et al. 1986). Detection thresholds for ITDs in the envelope of transposed tones: comparison with SAM tones and low-frequency pure tones
Of primary interest for this investigation was the neural sensitivity to ITDs in SAM as compared with transposed tones. To quantify the sensitivity to ITDs in response to transposed tones and SAM tones, high-resolution IPD functions were obtained and neural discrimination thresholds were calculated using ROC analysis (see METHODS).
High-resolution IPD functions were obtained in response to SAM and transposed tones from 14 neurons. For 12 neurons, functions were obtained only at a single fm. For 2 neurons, functions were obtained at 2 fm values (for one of these 2 neurons, a second high-resolution function was obtained in response to transposed, but not SAM, tones). This gave a total of 16 functions obtained in response to transposed tones and 15 functions obtained in response to SAM tones. Six examples of such functions are shown in Fig. 11, A–F. On the left, the mean discharge rates and SDs (error bars) are shown in response to SAM (blue) and transposed (red) tones. The abscissae differ across plots because they reflect the presentation of different values of fm that, given the constant range of IPDs presented, resulted in different ranges of ITDs being presented. Neurons in Fig. 11, A, C, and E responded best to IPDs around zero cycle, whereas those in Fig. 11, B, D, and F responded best to IPDs around 0.5 cycle. Consistent with the responses shown in Figs. 3, 4, and 7, discharge rates were more highly modulated to transposed tones than to SAM tones.
|
Figure 11A depicts data obtained with fm of 100 Hz. At that rate of modulation, the neural JNDs were 644 µs for SAM and 258 µs for transposed tones. Figure 11C shows data obtained from the same neuron in Fig. 11A, but for fm of 140 Hz. For that rate of modulation, JNDs were 672 and 177 µs for SAM and transposed tones, respectively. Consistent with the greater modulation in firing rates in response to transposed tones as a function of ITD, JNDs derived from these functions were correspondingly lower for transposed than for SAM tones. This is also the case in Fig. 11, B, D, and E. For the responses shown in Fig. 11E, the neural JNDs were large, being 10 and 3.3 ms in response to SAM and transposed tones, respectively. These large values were, at least in part, a result of the large variability observed in the measures of firing rate compared with the mean firing rate change over a cycle of IPD. For the responses depicted in Fig. 11F, the JND in response to transposed tones was >2 ms and, in response to SAM tones, there were no ITDs that were determined to be discriminable from an ITD of zero.
JNDs 4 ms for SAM and transposed tones are plotted as a function of fm in Fig. 11G. All 16 JNDs calculated in response to transposed tones were 4 ms and are marked by red diamonds in Fig. 11G. For 2 neurons, JNDs were calculated from high-resolution functions at 2 fm values. These are indicated by a square or a circular black outline around the red diamonds. For one of these neurons (the JND marked by a circular outline) a high-resolution IPD function was obtained in response to SAM tones at only one fm. Of the 15 JNDs obtained in response to SAM tones, only 10 were 4 ms, and these are plotted in Fig. 11G as blue circles. The remaining 5 JNDs were either >4 ms or were immeasurable. The horizontal dotted lines in Fig. 10G indicate estimates of the maximum ITD experienced by the guinea pig. The lower value is from a theoretical consideration of the maximum difference in time for sound traveling around a spherical head (McAlpine et al. 2001) and the higher value is from head-related transfer functions measured in the guinea pig (Sterbing et al. 2003) (see DISCUSSION).
Also shown in Fig. 11 are neural JNDs for ITD measured with low-frequency tones <400 Hz (Shackleton and Palmer 2004; Shackleton et al. 2003) for low-CF neurons in the guinea pig IC (squares). Median JNDs in response to SAM (blue), transposed (red), and low-frequency tones (black), are marked by asterisks. Overall, JNDs in ITD are lower for transposed tones than for SAM tones (Wilcoxon rank-sum test for equal medians, z = 2.6, P < 0.05). In the case of just one neuron (also shown in Fig. 7F), and only at fm of 200 Hz, was the JND lower in response to SAM tones than to transposed tones (0.5 vs. 1.4 ms). When fm was increased to 280 Hz, the JND for that same neuron was 320 µs in response to transposed tones, lower than either JND at 200 Hz.
The lowest JNDs in response to transposed tones were comparable to JNDs obtained in response to low-frequency tones, at corresponding frequencies, whereas the lowest JNDs in response to SAM tones were substantially higher. For fm >280 Hz, no neural JNDs could be calculated. As previously discussed (see Fig. 8), responses indicating sensitivity to changes in ITD were rarely recorded at these rates and when such responses were observed they were only weakly modulated with variations in ITD. In contrast, neural JNDs in response to low-frequency tones decrease as the tone frequency increases up to nearly 500 Hz (Shackleton and Palmer 2004). JNDs in response to low-frequency tones between 280 and 400 Hz contribute to the low median JND for low-frequency tones.
Phase locking to the envelope modulation
Transposed tones are designed to evoke a similar temporal pattern of action potentials from high-frequency ANFs to that evoked in low-frequency fibers by low-frequency tones. As previously discussed, high-frequency ANFs would not be expected to show this pattern in response to SAM tones. It is thus predicted that ANFs should show "tighter" phase locking to the period of a transposed than a SAM waveform (compare Fig. 1, B and C, right). We examined phase locking to the modulation rate in our sample of IC neurons.
Period histograms of the number of spikes occurring throughout the period of SAM and transposed waveforms were calculated for the 4 example neurons in Figs. 3–6 and are displayed in Fig. 12, A–D, respectively. Paired responses to SAM (top) and transposed (bottom) tones are shown in response to each modulation rate, with the response to 10-Hz modulation on the left. The response is clearly phase locked to 10 Hz in Fig. 12A. In response to transposed tones, spikes occur over a narrower range of the stimulus cycle than they do in response to SAM tones. In the second pair of histograms, measured at 110-Hz modulation, phase locking occurred in response to both stimuli but was tighter in response to transposed tones. As the modulation rate increased, phase locking became weaker and was no longer apparent at 310 Hz. A similar pattern is observed in Fig. 12, B, C, and D; the phase locking is limited to the lower-modulation rates in response to both stimuli and in response to transposed tones is "tighter" than to SAM tones. This likely reflects the shorter "ON period" per cycle of the transposed tones.
|
To examine the effect of stimulus type and fm on the vector strength, the vector strength values (for all recordings that showed significant phase locking) were grouped into logarithmically spaced fm bins. The means for each fm bin are shown in Fig. 13. Overall, phase locking was greater in response to transposed than to SAM tones (t-test, P < 0.001) and was low pass as a function of fm for both stimuli. The vector strengths in response to both stimuli decreased from similar fm values (about 100 Hz) and the characteristics of the reduction were qualitatively similar. Phase locking is essential for coding temporal information that mediates sensitivity to ITDs in binaural neurons. The reduction in phase locking with increasing fm might underlie the limit on sensitivity to envelope-based ITDs observed at around 300 Hz (see Figs. 8 and 10) (see DISCUSSION).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
). More neurons were sensitive to ITDs within the envelope of transposed than of SAM tones. Firing rates were more modulated as a function of ITD, and JNDs for ITD were consistently lower, in response to transposed tones than to SAM tones. We conclude that neurons are more sensitive to ITDs within transposed tones than within SAM tones. Comparison with human psychophysics using transposed tones
The current study of neural responses to ITDs was motivated by the finding that, for human listeners, transposed tones confer an improvement in sensitivity to ITDs at high carrier frequencies compared with SAM tones (Bernstein and Trahiotis 2002). The neural JNDs we observed follow a pattern similar to that obtained from human listeners; threshold ITDs obtained with transposed tones were smaller than those obtained with SAM tones. It is noteworthy that Bernstein and Trahiotis (2002) found threshold ITDs to be immeasurable with the high-frequency stimuli for rates of modulation >256 Hz fm. A similar limitation was observed in the neural data presented in the current study.
Bernstein and Trahiotis (2002) used a 300-ms presentation and a 2-cue, 2-alternative forced choice adaptive task to determine ITD thresholds. Clearly, one should be cautious when comparing human psychophysical data with neural recordings from guinea pigs, given the species differences and questions concerning the contribution of the response of single neurons to behavior. [These are well discussed in Skottun et al. (2001).] Additionally, guinea pigs were subject to urethane anesthesia. Despite this, a similar dependency of ITD-discrimination thresholds on fm and the type of stimulus was observed between neural responses and human listeners (Bernstein and Trahiotis 2002).
Comparison with previous in vivo electrophysiology
Previous authors have reported mammalian neural sensitivity to ITDs in the envelopes of high-frequency stimuli including data using SAM tones recorded from the MSO (Batra et al. 1997a, b), LSO (Joris 1996; Joris and Yin 1995), and IC (Batra et al. 1989, 1993), noise recorded from the IC (Joris 2003), clicks recorded from the IC (Caird and Klinke 1987), and high-frequency tones with trapezoid envelopes recorded from the IC (Yin et al. 1984). The neural responses to SAM and transposed tones, in the current study, were qualitatively similar to these previous studies; BPs were predominantly leading at the contralateral ear and phase plots were not necessarily well described by linear regression. Clustering of BPs close to 0 and 0.5 cycle was reminiscent of EE and EI interactions described previously in response to high- (Batra et al. 1997a) and low-frequency sounds (Yin and Kuwada 1983b).
We compared neural JNDs obtained in the
