INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Cats and other mammals are capable of localizing the source of high-frequency, broadband sounds with considerable accuracy. Binaural level disparities are most important for left-right (azimuthal) localization. Diffraction of sound with the pinna leads to direction-dependent variation in high-frequency spectral shape (spectral shape cues) at the eardrum (cat: Musicant et al. 1990; Rice et al. 1992, Xu and Middlebrooks 2000; human: Batteau 1967, Shaw 1974). Some unilaterally deaf humans can localize broadband high-frequency sounds (Butler 1975; Hausler et al. 1983; Slattery and Middlebrooks 1994) presumably on the basis of monaural spectral shape. Monaural spectral shape cues are important for up-down (elevational) localization (cats: May 2000; Sutherland et al. 1998a,b; humans: Middlebrooks and Green 1991; Wightman and Kistler 1997). They also contribute to front-back localization and to a lesser extent to left-right localization (Butler 1986; Butler et al. 1990, Fisher and Freeman 1968; Hebrank and Wright 1974; May 2000; Oldfield and Parker 1986; Slattery and Middlebrooks 1994; Wightman and Kistler 1997).

High-frequency neurons in the inferior colliculus (IC) exhibit three patterns of binaural interactions that act on binaural disparities to shape azimuth tuning (Delgutte et al. 1999; Geisler et al. 1969; Irvine 1986; Irvine and Gago 1990; Rose et al. 1966). Binaural inhibition suppresses responsiveness to sound directions on one side of the head thus producing hemi-field selectivity. Binaural facilitation enhances responsiveness to sound directions near the midline to produce central-field selectivity. Mixed inhibitory and facilitatory interactions suppress responsiveness to sound directions on one side of the head, enhance responsiveness on the other, and create receptive fields that vary from central-field to hemi-field. Similar patterns of binaural interactions and azimuth tuning have been described in the cat's superior colliculus (Hirsch et al. 1985; Middlebrooks 1987; Wise and Irvine 1985), medial geniculate body (MGB) (Ivarsson et al. 1988; Samson et al. 2000), and primary auditory cortex (AI) (Samson et al. 1994).

We previously classified forebrain (MGB and AI) cells on the basis of whether they derived directionality predominantly from binaural cues (binaural directional or BD cells), monaural cues, or both monaural and binaural cues (monaural directional or MD cells) (Clarey et al. 1995; Samson et al. 1993, 1994, 2000). Under monaural conditions, MD cells were responsive to noise bursts at certain azimuths, and relatively unresponsive at others, even at high sound pressure levels (SPLs). The direction-dependent variation in responsiveness suggests that these cells are sensitive to monaural spectral shape cues. Particular spectral shapes can apparently suppress or inhibit a cell's response to noise, a phenomenon referred to as spectral inhibition. Furthermore, the dependence of such cells' directionality on monaural spectral shape cues is consistent with the observation that they are monaurally responsive at each azimuth (nondirectional) to tone bursts (Clarey et al. 1995; Imig et al. 1997; Samson et al. 1993, 2000). Tone burst spectra are dominated by a single frequency and thus show relatively little direction-dependent spectral-shape variation. Spectral-shape sensitivity has also been shown to contribute to the directionality of neurons in the dorsal cochlear nucleus (DCN). Monaural directionality of DCN neurons is stronger than that of ventral cochlear nucleus (VCN) neurons or auditory nerve fibers but is much weaker than that seen in AI and the MGB (Imig et al. 2000; Poon and Brugge 1993; Rice et al. 1995). Although sensitivity to monaural spectral shape cues has been shown to contribute to the directionality of neurons at lower and higher levels of the auditory pathway than the IC, there are no published accounts of this phenomenon in the IC. Thus our first goal was to determine the extent to which IC neurons exhibit monaural directionality.

Side-band inhibition appears to contribute to spectral inhibition in the MGB as MD cells showed stronger two-tone inhibition at directions where they were unresponsive to noise stimulation than at directions where they were responsive (Imig et al. 1997). According to the side-band inhibition model, a MD neuron's responsiveness reflects the relative amounts of acoustic energy in its excitatory and inhibitory frequency domains (Imig et al. 1997). Discharge rate increases at directions where there is more energy in excitatory than inhibitory domains and decreases at directions where there is more energy in inhibitory than excitatory domains. A second goal of this study was to test this hypothesis in the IC by determining if MD cells show stronger two-tone inhibition at directions where they were unresponsive to noise stimulation than at directions where they were responsive.

The side-band inhibition model predicts that MD cells should have stronger and/or more extensive inhibitory side-bands than BD cells, thus giving rise to their stronger monaural directionality. In accord with the prediction, excitatory frequency domains of MD cells are about half as wide as those of BD cells in the MGB (Imig et al. 1997). Additionally, MD cells in both the MGB and AI show more strongly nonmonotonic level response functions to noise stimulation as compared with BD cells. Finally in both the MGB and AI, MD cells show lower responsiveness to noise than tones, whereas BD cells show no difference in responsiveness (Clarey et al. 1995; Imig et al. 1997; Samson et al. 2000). These differences suggest the presence of stronger inhibitory processes in MD cells than in BD cells. The third goal of this study was to determine whether BD and MD cells in the IC exhibited similar differences in these response properties.

Discharges of DCN neurons are inhibited when a spectral notch is centered at best frequency (BF), a phenomenon known as notch inhibition (Nelken and Young 1994; Spirou and Young 1991). Spectral notches are natural features of high-frequency spectra at the eardrum. The number, depth, width, and center frequencies of notches vary with sound direction, and they have been implicated as playing an important role in directional hearing (Huang and May 1996a,b; May and Huang 1996; Musicant et al. 1990; Rice et al. 1992). Notch inhibition contributes to the directionality of DCN neurons (Imig et al. 2000). Delgutte et al. (1999) looked for evidence of notch inhibition in the responses of IC single units, but their results were inconclusive. They used binaural stimuli that were presented at relatively low SPLs, and these may not be best suited for revealing spectral inhibition. Thus the fourth goal of this study was to search for notch inhibition in the IC using stimuli that we have found to reveal monaural directionality previously, i.e., noise bursts presented over a broad range of SPLs under monaural conditions.

Previous studies show that spectral inhibition is weaker in the cochlear nucleus (CN) (Imig et al. 2000) than in the MGB and AI (Samson et al. 2000). The fifth goal of this study is to compare the strength of spectral inhibition in the IC with that at lower and higher levels of the auditory system. This comparison should indicate the extent to which spectral inhibition is generated between the CN and the IC, between the IC and the MGB, and between the MGB and AI. Such a comparison is possible because recordings in the CN, MGB, and AI studies were carried out under virtually identical conditions to those used in the IC in this present study. Some of these findings have been presented previously in abstract form (Poirier et al. 1996).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Twelve healthy young adult cats with clean external ears, translucent eardrums and low-threshold single-unit responses were used in the IC recording experiments. Experiments were carried out using protocols approved by the Institutional Animal Care and Use Committee of the University of Kansas Medical Center. We also performed an analysis of strength of spectral inhibition on some previously published data in the MGB and AI (Samson et al. 2000) to compare response properties of cells at different levels of the auditory system. Details concerning single-unit recording, computer control of data collection, data analysis, and sound generation have been described previously (Barone et al. 1996; Imig et al. 2000; Samson et al. 1993, 1994, 2000).

Chronic recording procedures were used to increase the amount of data collected from each cat. The number of recording sessions per cat varied between 2 and 11 with an average of 4.7 per cat. Cats were prepared for chronic recording during an initial aseptic surgical procedure using general anesthesia. A recording chamber was positioned over a craniotomy either at a 10° angle from rostrodorsal to caudoventral or a 60° angle from caudodorsal to rostroventral to allow recording electrodes to pass beneath the tentorium and reach the left IC. During recording sessions, pentobarbital sodium anesthesia was used to eliminate pinna reflexes and spontaneous movements. Electrolytic marking lesions were placed during terminal recording sessions to aid in electrode track reconstruction.

Single-unit recordings were carried out in an electrically shielded, anechoic, sound-isolation chamber. The anesthetized cat rested in a sling. Its head was supported with the horizontal Horsley-Clarke plane tilting forward and down at an angle of ~18° from horizontal, and the ears were pulled to an upright position This approximates the head and pinna position of an alert cat looking forward.

Auditory waveform synthesis, acoustic calibration, stimulus timing and sequencing, and data collection were controlled by a PDP 11/73 computer. Stimulus waveforms were generated at an output sample rate of 100 kHz using a 16-bit D/A converter (Boys Town National Research Hospital), low-pass filtered at 40 kHz (Kemo VBF/8, 180 dB/octave) to prevent aliasing, attenuated with computer controllable attenuators, and amplified. Stimuli were 50 ms in duration including 5-ms linear rise-fall times, typically repeated 10 times and presented at 2-5 Hz. Spike times were stored in a computer data file with a resolution of 0.01 ms. A peristimulus time window of 0-60 ms was used for the analysis of spike counts although in most cases this was unnecessary as spontaneous activity was low or nonexistent and there were no responses outside of this window.

A horizontal array of loudspeakers (Radio Shack 40-1310B) with similar frequency response characteristics allowed the presentation of sounds from different azimuths. The array was composed of 13 loudspeakers spaced at 15° intervals along a 180° arc of an imaginary circle (0.79 m radius) that was centered on the cat's head. Azimuth could be varied by presenting sound from different loudspeakers or by rotating the array about the head. Each loudspeaker was calibrated by placing a microphone (B&K type 4133 1/2 in) at the center of the circle, aiming it at the loudspeaker, and performing a fast Fourier transform on the impulse response. Noise SPL was measured using the RMS voltage from the microphone. Tables of maximum SPLs attainable at different frequencies and for noise stimuli for each loudspeaker were stored in computer files for use during experiments. Loudspeaker output increased from 4 kHz to a peak at 8 kHz at 20 dB/octave, decreased by 5 dB/octave 35 kHz, and then decreased at 60 dB/octave. A random number generator produced a frozen noise waveform with a flat spectrum (0-50 kHz) and random amplitude distribution. The actual spectrum of the noise delivered from the loudspeaker was shaped by the sound system (mainly the loudspeaker).

Single-unit waveforms were identified using noise burst search stimuli that varied in azimuth and SPL. Once a unit was isolated, tone bursts were presented from a loudspeaker located at an azimuth that produced the strongest response to noise stimulation, and the unit's BF was determined using audiovisual criteria. BF is the frequency that produced the maximum response at a SPL near lowest threshold. Next, responses of high-BF (>4 kHz) units to noise and in some cases BF tone bursts were recorded to characterize azimuth tuning. Azimuth-level data sets consisted of a single-unit's responses to sounds that varied in azimuth (typically frontal field -90 to +90° in 30° steps) and SPL (typically from 0 to 80 dB in 10- or 20-dB steps). For all units, azimuth was varied by presenting sound from different loudspeakers, and these data are referred to as "multi-loudspeaker data sets." Additionally, some units were presented sounds from different azimuths using a single loudspeaker that was moved by rotating the loudspeaker array. These data are referred to as "single-loudspeaker data sets." Low-BF units were not studied as low-frequency direction-dependent variation in spectral shape is relatively small (Musicant et al. 1990; Rice et al. 1992) and would not be expected to strongly influence unit directionality. Frequency-level data sets consisted of a single-unit's responses to tone bursts of varying frequency (8 steps per octave) and SPL (10- or 20-dB steps, from below threshold to 80 dB SPL). These were recorded for the purpose of assessing breadth of frequency tuning and were graphically displayed as excitatory frequency response areas. Tone bursts were presented from a loudspeaker located at the azimuth where the cell was most responsive to noise stimulation.

Stimuli were presented under binaural and "quasi-monaural" (henceforth, "monaural") conditions, the latter produced using an earplug to attenuate free-field sound reaching one ear. Ears were plugged by injecting ear mold compound (Audalin, All American Mold Lab) into the concha and ear canal and firmly pressing the material in place to ensure a tight seal. The ear mold compound was a viscous fluid when injected, but it cured to a soft plastic consistency that was easy to remove, leaving no visible residues in the ear canal. New plugs were made each time an ear was occluded. Earplug attenuation was estimated using both acoustical and physiological measures. For acoustical measurements (Samson et al. 1993), the external auditory canal was surgically opened near its junction with the bulla, and a probe tube microphone was sealed in the opening with its tip near the eardrum. The transfer function from the loudspeaker to the eardrum was measured with and without the ear plugged (the earplug did not reach the tip of the probe tube). The resulting frequency spectra were subtracted. The difference spectrum (attenuation) varied between 32-70 dB in the range of 4 - 32 kHz (the useable frequency range for the measurements). At most frequencies, attenuation ranged between 40 and 60 dB, with less attenuation occurring over a narrow range of frequency that corresponded with a notch in the spectrum of the unplugged ear. Earplug attenuation was also measured physiologically. We compared a single unit's noise burst thresholds with and without bilateral earplugs as shown in Fig. 1. Sounds were presented at the azimuth that produced the lowest threshold response without earplugs (e.g., 30°, Fig. 3D). Threshold was ~0 dB SPL without ear plugs (no plug). Bilateral ear plugging caused the threshold to increase to ~70 dB SPL. The earplugs were then removed and threshold returned to 0 dB. This was typical of all cells so tested showing the reversibility of the method. The 31 units for which thresholds were compared under no plug and bilateral plug conditions showed an average increase in threshold of 51 dB after bilateral ear plugging. This is an underestimate of the actual threshold shift because 8/31 units did not respond at the highest level tested, and in these cases, we took earplug attenuation to be equal to the difference between threshold with no plugs and the highest stimulus level tested with earplugs. Although thresholds increased by 40 dB in 29/31 units, two units showed threshold increases of only 10-20 dB, presumably due to poorly fitted plugs. Nevertheless, these data show that ear plugging typically produced attenuation of 40 dB, consistent with the acoustical measurements.

View larger version (23K): [in this window] [in a new window]   Fig. 1. Effect of ear plugging on single-unit level-response functions. Level-response functions were obtained by presenting noise bursts at 30° azimuth. Three level-response functions were obtained without ear plugs (no plug), and 1 with bilateral ear plugs. The threshold difference between no plug and bilateral plug functions is ~70 dB. Responses of this unit (9403-24) are also shown in the 2nd row of Fig. 3.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Spike counts at each azimuth-level combination in a data set were displayed as an azimuth-level response area (ALRA, e.g., Fig. 3A). Spike counts at each azimuth were averaged over SPL to obtain an azimuth function (e.g., Fig. 3C). Azimuth function modulation, the percentage difference between minimum and maximum function values, was used as an index of azimuth sensitivity (directionality). Cells with function modulation 75% were classified as azimuth sensitive (directional). Such units typically responded well with low thresholds at certain azimuths and relatively poorly and with higher thresholds at other azimuths. Many were completely unresponsive at certain azimuths even at high SPLs, (e.g., Fig. 3A). We refer to azimuths where responsiveness is relatively low (25% of maximum) as azimuth function troughs, and azimuths where responsiveness is relatively high (75% of maximum) as azimuth function peaks.

Multi-loudspeaker ALRA data sets were obtained for 425 single units using noise stimulation under binaural conditions (B-stim, i.e., neither ear plugged). Sixty-nine percent of the sample of 425 units was directional. This report focuses on the responses of 129 of the directional units whose monaural responses were also recorded. This sample excludes the responses of eight cells with unreliable responses or minimum thresholds 50 dB SPL. The latter were excluded as identification of spectral inhibition depends on recording responses over a broader range of SPLs than was available for such cells. Monaural and binaural responses were compared to assess binaural interactions, the presence of monaural excitatory inputs, and the relative contributions of binaural and monaural mechanisms to azimuth tuning. Repeated monaural and/or binaural multi-loudspeaker ALRA data sets were available for 85/129 units and were used to assess response reliability. Response property measures (e.g., azimuth functions, level functions, etc.) were based on averages of the repetitions. Single-loudspeaker data sets were additionally obtained for 44 of the 85 units but were not used in the numerical analyses because the single-loudspeaker data were not always collected using the same SPL steps as were the multi-loudspeaker data. Later we show that the single- and multi-loudspeaker data were indistinguishable, so limiting analyses to multi-loudspeaker data has no significant effect on the results. Most units (105/129) in this sample lacked spontaneous activity. The remaining 24 units exhibited very low spontaneous activity, of ~1 spike/s or less, and this represented ~1% or less of the maximum rate of discharge during the stimulus.

Histological reconstruction of electrode tracks allowed identification of recording sites for most (120/129) units in the sample. These sites were found in either the central nucleus (n = 94/120) or the dorsal cortex (n = 26/120) of the IC (Berman 1968; Morest and Oliver 1984). There were no obvious differences in response properties between these two groups, so they were combined. BFs could be assigned for 127/129 units, and these ranged from 4 to 37 kHz [14.7 ± 6.5 (SD) kHz]. The two units for which BFs were indeterminate were responsive to a broad range of high-frequency tone bursts.

Binaural directional (BD) and predominantly binaural (PB) cells

The BD group was composed of 68 cells that were nondirectional (azimuth function modulation <75%) to noise bursts presented under monaural conditions, showing that their directionality was dependent on binaural cues. Comparison of monaural and binaural responses revealed the presence of three patterns of binaural interactions: inhibition, facilitation, and mixed inhibition and facilitation. Qualitative classification of binaural interactions was usually clear-cut as binaural inhibition created azimuth function troughs, binaural facilitation created azimuth function peaks, and mixed inhibition and facilitation created troughs and peaks, respectively. Nevertheless classification of some cells was somewhat arbitrary, so we used two objective criteria developed by Delgutte et al. (1999) for classifying binaural interaction type and strength (Fig. 2A). Binaural interaction strength (BIS) characterizes the difference between monaural and binaural responses, and ranges from 0 (no difference) to 1 (large difference). Binaural interaction type (BIT) indicates the relative balance of binaural inhibition and facilitation with positive values indicating greater facilitation and negative values indicating greater inhibition. A scatter plot showing the joint distribution of BIS versus BIT values for the sample of BD and PB units is shown in B. Delgutte et al. (1999) used BIT criteria of -0.4 and +0.7 to divide their sample into three binaural interaction classes of facilitation, inhibition, and mixed facilitation and inhibition. We have used these same criteria as they match closely our qualitative classification.

View larger version (30K): [in this window] [in a new window]   Fig. 2. Quantitative classification of binaural interactions for noise burst stimuli presented in the free field. A: method of Delgutte et al. (1999) for computing binaural interaction strength (BIS) and binaural interaction type (BIT). Azimuth functions are averages of the multi-loudspeaker functions shown in Fig. 3F. Superposition of monaural and binaural azimuth functions defines 3 areas: an area of facilitation (AF), an area of inhibition or suppression (AS), and an area common to both curves (A0). B: a scatter plot shows BIS vs. BIT values for the binaural directional (BD) and predominantly binaural (PB) samples. C: scatter plot showing BIS vs. BIT values for the monaural directional (MD) sample.

BD-EI cells (n = 45) exhibited binaural inhibition (smaller binaural than monaural responses, 1.0  BIT < 0.4, Fig. 2B). Based on their responses to ear plugging, we inferred that BD-EI cells in our sample received excitatory (E) input from the contralateral ear and inhibitory (I) input from the ipsilateral ear. Under binaural conditions, they were most responsive to contralateral sound directions, i.e., were contralateral preferring. A typical example is shown in Fig. 3, top. This unit's ALRAs obtained under binaural (B-stim, i.e., neither ear plugged) and contralateral monaural conditions (C-stim, i.e., ipsilateral ear plugged) are shown in A and B, respectively. The unit was unresponsive to stimulation under ipsilateral monaural conditions (C, I-stim, i.e., contralateral ear plugged), although ipsilateral stimulation had an inhibitory effect on responses in the ipsilateral field (compare C- and B-stim). Of the 10 units tested under ipsilateral monaural conditions (i.e., stimulation of the inhibitory ear), 9 were unresponsive or responded only at high SPLs, the latter responses presumably resulting from sound leakage through the earplug. One remaining unit responded with weak low-threshold responses.

View larger version (41K): [in this window] [in a new window]   Fig. 3. Responses of BD and PB cells to noise burst stimulation under monaural and binaural conditions. Top: a BD cell exhibiting binaural inhibition (BD-EI cell, 9506-19). A: azimuth-level response area (ALRA) for binaural noise stimulation (B-stim). An ALRA displays the normalized spike count response (% of maximum) as a joint function of azimuth and sound pressure level (SPL). Azimuth representation: 0°, median plane in front of the head; +90°, contralateral (right) pole; 90°, ipsilateral (left) pole. , azimuths and SPLs of stimuli presented during data collection. ,100% response (Rmax). B: ALRA for noise stimulation under monaural contralateral conditions (C-stim). C: average azimuth functions for B-, C-, and I-stim (stimulation under monaural ipsilateral conditions). These were obtained by averaging over SPL, and identical SPLs were used for each function in a monaural/binaural comparison. Numbers preceding the stimulus identifier (e.g., 1 B-stim, 2 C-stim, etc.) indicate the sequential order of data collection. Middle: a BD cell exhibiting mixed binaural facilitation and inhibition (BD-FI cell, 9403-24). D: ALRA for B-stim. E: ALRA for C-stim. F: average azimuth functions for B-, C-, and I-stim. Bottom, G: average azimuth functions for a BD cell that exhibited binaural facilitation (BD-F cell, 940--89). H and I: a PB cell that exhibited binaural facilitation and was unresponsive under monaural conditions (9603-44). H: ALRA for B-stim. I: average azimuth functions for B-, C-, and I-stim. Rmax values: A, 15.2 spikes * stimulus1; B, 16.3 spikes * stimulus1; D, 8.9 spikes * stimulus1; E, 2.6 spikes * stimulus1; H, 2.5 spikes * stimulus1. BF, best frequency; BIS and BIT, see Fig. 2. Data were collected using the multi-loudspeaker paradigm to vary azimuth except for those azimuth functions identified with an asterisk (e.g., 5 B-stim*, C). These were collected using a single loudspeaker and were not necessarily obtained using the same levels as the other multi-loudspeaker azimuth functions.

BD-FI cells (n = 21) exhibited facilitation (larger binaural than monaural responses) at some azimuth-level combinations and inhibition at others (-0.4  BIT  +0.7, Fig. 2B). A typical example is shown in Fig. 3, middle. Comparison of monaural and binaural azimuth functions (F) shows facilitation at contralateral directions and inhibition at ipsilateral directions. This cell was responsive to C-stim but not to I-stim, but some other BD-FI cells were responsive to both C-stim and I-stim. Azimuth function peaks of BD-FI cells varied in location. Most had contralateral or midline preferences and a few had ipsilateral preferences.

BD-F (n = 2) and PB (n = 18) cells exhibited binaural facilitation (+0.7 < BIT  +1.0, Fig. 2B). BD-F cells responded under monaural conditions to stimulation of one or both ears (Fig. 3G), whereas PB cells were unresponsive or poorly responsive to stimulation of either ear under monaural conditions (Fig. 3, H and I). Most PB cells in the monaurally poorly responsive group responded only at high levels under monaural conditions. We interpret the responses as resulting from sound leakage though the earplug. A few PB cells responded very weakly at low levels under monaural conditions. In these cases, the distinction between PB and BD-F cells was rather arbitrary. Most PB and BD-F cells had response peaks located near the midline. The azimuth tuning of PB cells clearly depended on binaural disparities but lack of response under monaural conditions precluded testing their monaural directionality. Consequently we consider them as a separate group that is not included in either the BD or monaural directional (MD) groups.

Single-loudspeaker data sets were obtained for 20 BD cells and 12 PB cells under monaural or binaural conditions or both. Azimuth functions obtained using a single loudspeaker are identified with an asterisk in the figures (e.g., Fig. 3C, see 5 B-stim*). The multi- and single-loudspeaker data showed no consistent differences, indicating that any differences in the frequency response characteristics between loudspeakers were so small that they did not significantly affect the response properties that we measured.

Patterns of monaural and binaural azimuth tuning in MD cells

MD cells (n = 43) were distinguished from BD and PB cells by their directionality to noise bursts presented under monaural conditions (i.e., azimuth function modulation 75%). Recall that our definition of directionality depends on a cell's response over a broad range of SPL. Thus just because both MD and BD cells responded under monaural conditions to restricted ranges of azimuth at SPLs near threshold, this does not make both types of cell directional. Directionality (or lack thereof) is revealed in responses to higher noise levels at which BD cells were responsive at all azimuths whereas MD cells were not. MD-E0 cells (n = 17) had rather similar monaural and binaural azimuth function profiles showing that sensitivity to monaural cues made a major contribution to their azimuth tuning. Most MD-E0 cells (15/17) received excitatory input from the contralateral ear. Figure 4 shows examples of three MD-E0 cells with azimuth functions characterized by relatively narrow peaks and broad troughs. The cell in Fig. 4, top, responded well throughout the contralateral rear quadrant, and it was relatively unresponsive at any direction in the frontal field except at 90°. Cells in Fig. 4, middle and bottom, had response peaks located at 0 and 60°, respectively. Figure 5 shows two MD-E0 cells whose azimuth functions were characterized by narrow troughs. In each case, the trough occurred at the same location regardless of whether the cell was stimulated under monaural or binaural conditions.

View larger version (38K): [in this window] [in a new window]   Fig. 4. Neurons with azimuth tuning dependent predominantly on a monaural mechanism (MD-E0 cells) that exhibited broad troughs and narrow peaks in azimuth functions. Similarity between monaural and binaural noise azimuth functions shows that binaural interactions make little contribution to azimuth tuning. Top, middle, and bottom: responses of units 9502-30, 9508-06, and 9502-14, respectively. I: 6 B-stim (17 kHz) and 7 C-stim (17 kHz) show responses to 17-kHz tonal stimulation. Left: ALRAs for B-stim. Middle: ALRAs for C-stim. Right: average azimuth functions under monaural and binaural conditions. Rmax values: A, 1.0 spikes * stimulus1; B, 1.0 spikes * stimulus1; D, 2.5 spikes * stimulus1; E, 2.3 spikes * stimulus1; G, 0.8 spikes * stimulus1; H, 0.7 spikes * stimulus1.

View larger version (34K): [in this window] [in a new window]   Fig. 5. Neurons with azimuth tuning dependent predominantly on a monaural mechanism (MD-E0 cells) that exhibited narrow troughs in azimuth functions. Similarity between monaural and binaural azimuth functions shows that binaural interactions make little contribution to azimuth tuning. Top: responses of unit 9601-09. C: 3 C-stim (8.5 kHz), responses to tonal stimulation. Bottom: responses of unit 9605-01. F: 3 C-stim (11 kHz), responses to tonal stimulation. Left: ALRAs for B-stim. Middle: ALRAs for C-stim. Right: average azimuth functions for B- and C-stim. Rmax values: A, 4.1 spikes * stimulus1; B, 4.0 spikes * stimulus1; D, 2.4 spikes * stimulus1; E, 3.8 spikes * stimulus1.

Some MD-E0 cells may be strictly monaural (low BIS values in Fig. 2C), but in other cases, repeated data sets revealed consistent differences between monaural and binaural responses. The cell in Fig. 4, bottom, shows weak binaural inhibition at 60 and 90° to noise stimulation. Weak binaural inhibition also appears to be present in its responses to 17-kHz tonal stimulation (Fig. 4I). Nevertheless, binaural inhibition by itself was insufficient to produce directional responses to tonal stimulation, and the azimuth tuning to noise stimuli is so similar under monaural and binaural conditions that it must be attributed predominantly to a monaural mechanism. The binaural responses of the cell in Fig. 4F were consistently larger than the monaural response, revealing binaural facilitation. The BIS is above the average for the MD-E0 sample, but again, the azimuth tuning is so similar under monaural and binaural conditions that it must be attributed predominantly to a monaural mechanism. The cell in Fig. 5F shows weak mixed interactions but these do not produce the peak and trough in the azimuth function that is typical of stronger mixed interactions. Although this binaural interaction doesn't have an obvious effect on azimuth tuning, it does contribute to level tuning as responses to C-stim were consistently more strongly nonmonotonic than to B-stim (Fig. 5, D and E). Thus the MD-E0 designation does not imply that cells were necessarily monaural only that azimuth tuning was predominantly determined by a monaural mechanism.

The azimuth tuning of MD-BD cells (n = 26) showed strong contributions from both monaural and binaural mechanisms. Depending on pattern of binaural interactions, such cells were classified as MD-EI (n = 14), MD-FI (n = 7), or MD-F (n = 5) according to the same BIT criteria used to classify BD cells (MD-BD, Fig. 2C). Binaural interactions had the same effect on responses of MD-BD cells as they had on BD cells, i.e., binaural inhibition created azimuth function troughs, mixed inhibition and facilitation created troughs and peaks, and binaural facilitation created or enhanced peaks in the azimuth function. MD-EI cells were unresponsive at the same location in both monaural and binaural azimuth functions (e.g., Figure 6, top at +30° and middle top at +90°). Most (12/14) MD-EI cells received excitatory input from the contralateral ear. An example of a MD-FI cell that had a response peak located near the midline is shown in Fig. 6, middle bottom. It was responsive to C-stim but not I-stim. C- and B-stim noise-burst azimuth-functions showed troughs at +90°, suggesting that monaural spectral inhibition also contributed to binaural azimuth tuning (I). A MD-F cell (Fig. 6, bottom) showed minimal and maximal responsiveness at the same locations in monaural and binaural azimuth functions, but it was considerably more responsive at 0 and -30° to B-stim than to C-stim (responses to I-stim were not obtained). A monaural response trough at +60° contributes to binaural azimuth tuning in this cell.

View larger version (51K): [in this window] [in a new window]   Fig. 6. Neurons with azimuth tuning dependent on both monaural and binaural mechanisms (MD-BD cells). Top: responses of a MD cell that exhibited binaural inhibition (MD-EI cell, 9603-42). Middle top: responses of a MD-EI cell (9407-41). Middle bottom: responses of cell that exhibited mixed binaural facilitation and inhibition (MD-FI, 9601-05). I: 5 C-stim (17 kHz), responses to tonal stimulation. Bottom: responses of a MD cell that exhibited binaural facilitation (MD-F, 9502-08). Left: ALRAs for B-stim. Middle: ALRAs for C-stim. Right: average azimuth functions under monaural and binaural conditions. Rmax values: A, 1.0 spikes * stimulus1; B, 1.4 spikes * stimulus1; D, 4.7 spikes * stimulus1; E, 6.2 spikes * stimulus1; G, 8.5 spikes * stimulus1; H, 4.3 spikes * stimulus1, J, 1.0 spikes * stimulus1; K, 1.1 spikes * stimulus1.

Some MD-BD cells exhibited monaural response troughs that were not seen in their binaural azimuth functions (MD-EI, n = 4; MD-FI, n = 4; MD-F, n = 3). Although none of these responses are illustrated, an example of a monaural trough that does not obviously appear in the binaural response is seen in Fig. 6, middle top. This cell shows a trough to C-stim at -90°, but it isn't obvious that this monaural trough contributes to binaural azimuth tuning because it lies within a broader trough that is created by binaural inhibition. Two MD-FI cells and one MD-F cell showed troughs at a particular azimuth under monaural but not binaural conditions. These cells exhibited monaural directionality that did not obviously contribute to their binaural directionality.

The use of a multi-loudspeaker array to present sounds from different azimuths presents a potential ambiguity in interpretation of these results. Loudspeakers had similar but slightly different frequency-response characteristics, therefore use of different loudspeakers to vary azimuth also produced some loudspeaker-dependent variations in spectral composition of noise stimuli. Thus one might question whether the response modulation that occurred was a function of single-unit sensitivity to sound direction or to differences among loudspeaker frequency response characteristics. As a control, ALRA data sets for noise stimulation under monaural and/or binaural conditions were obtained for 12 MD cells using a single loudspeaker and moving its location to vary azimuth. This control removes the contribution of differences in loudspeaker frequency response characteristics to sound spectra at different directions, leaving only the contribution of direction-dependent differences in acoustic diffraction. These single-loudspeaker data are shown in azimuth functions marked by asterisks in Figs. 4I (5 B-stim*), 5, C (5 C-stim* and 8 B-stim*) and F (7 B-stim*), and 8A (3 B-stim*). There were no systematic differences between multi- and single-loudspeaker data sets showing that azimuth function response modulation was due to direction-dependent differences in monaural spectral shape produced by acoustical diffraction rather than differences in the frequency-response characteristics of loudspeakers.

Monaural directionality is greater to noise than to tone bursts

If monaural directionality depends on monaural spectral shape cues, then MD cells should exhibit greater monaural directionality to noise bursts that show relatively large direction-dependent variation in spectral shape than to BF-tone bursts whose narrow spectra allow for relatively little variation. Monaural ALRA data sets were obtained for both BF-tone and noise burst stimulation for a sample of 16 MD cells. Each MD cell showed considerably greater azimuth function modulation for noise than tones (e.g., Figs. 4I, 5, C and F, and 6I). Figure 7 shows a comparison of tone and noise modulations for the entire MD sample (). The average modulation for noise (90.9%) was significantly greater than that for BF tones (38.8%, paired 2-tailed t-test, P < 0.0001).

View larger version (19K): [in this window] [in a new window]   Fig. 7. Relationship of azimuth function modulation for noise and tone stimulation. The scatter plot compares azimuth-function modulation for BF-tone bursts vs. noise bursts for various classes of cells. MD cells were tested under monaural conditions. BD and PB cells were tested using binaural stimulation.

Binaural directionality to BF tones and noise was compared for eight BD-EI and BD-FI cells (Fig. 7). The average azimuth function modulation for noise (92.6%) was slightly greater than that for BF-tones (90.6%), and the difference was not statistically significant. This suggests that BD cell directionality does not strongly depend on spectral shape cues. PB cells (n = 8) showed greater variation in their sensitivity to bandwidth. The average azimuth function modulation for noise (94.9%) was greater than that for BF-tones (81.5%, paired 2-tailed t-test, P = 0.11). For most PB cells, modulation was similar for noise and tones, but for a few, tone modulation was much less than noise modulation. The lack of PB cell response under monaural conditions precluded any inference regarding the contribution of monaural spectral-shape cues to their azimuth tuning. However, these results indicate that the directionality of some PB cells was more dependent on stimulus bandwidth than was the directionality of others.

Effect of sound direction on strength of two-tone inhibition

According to the side-band inhibition model, a MD cell derives monaural directionality from a noise stimulus by comparing the amount of energy in its excitatory and inhibitory frequency domains. Spectral shape changes with sound direction, thus the relative amount of energy in excitatory and inhibitory frequency domains also changes. At sound directions corresponding with azimuth function peaks, there is presumably a relatively greater amount of energy in excitatory domains than inhibitory domains causing an increase in responsiveness. At sound directions corresponding with azimuth function troughs, there is presumably relatively more energy in inhibitory domains than excitatory domains causing a decrease in responsiveness. We have tested this hypothesis using a two-tone stimulation paradigm, in which stimuli are composed of two frequency components that were presented simultaneously. A constant frequency component, F1, located at BF, caused the cell to discharge when presented alone. F1 was presented in combination with a variable frequency component (F2) that by itself often had no effect on discharge, but when presented simultaneously with F1 it could cause a decrease (2-tone inhibition) or increase (2-tone facilitation) in responsiveness relative to the response to F1 alone. According to the side-band inhibition hypothesis, two-tone inhibition in MD cells should be stronger at azimuth function troughs than at peaks.

Figure 8B shows a MD cell's ALRA that was obtained under monaural conditions and encompasses the full 360° of azimuth. Responses were similar under monaural and binaural conditions including a response trough at 60° (A). This cell's frequency response area is shown in C. Tone bursts were presented in eighth octave steps between 2 and 40 kHz, and a single excitatory domain was revealed with a BF of 17 kHz. A `constant-SPL F1' two-tone paradigm, as described in the figure legend, revealed two low-threshold, inhibitory domains (gray shading C), one overlapping the excitatory domain and extending to higher frequencies, and the other centered near 6 kHz.

View larger version (61K): [in this window] [in a new window]   Fig. 8. Magnitude of 2-tone inhibition is greater at an azimuth trough than peak for a MD-E0 cell (unit 9403-93). A: azimuth functions for noise stimulation obtained under binaural (B-stim) and monaural conditions (C-stim). B: ALRA for noise stimulation under monaural conditions. Rmax is 1.9 spikes/stimulus. C: frequency response area. The excitatory domain is shown in the color contour plot. Rmax is 2.0 spikes/stimulus. Inhibitory domains, shaded in gray, were identified using a "constant-SPL F1" 2-tone paradigm. The 2-tone stimulus consisted of a fixed-frequency (BF) and constant-SPL (10-15 dB above BF threshold) F1 component, and variable frequency and variable SPL F2 component. F1 and F2 were simultaneously presented. In this case, F1 was presented at 17 kHz and 0 dB SPL. F2 varied in frequency between 4 and 26 kHz and varied in SPL between 0 and 50 dB. Inhibitory domains in C indicate where the 2-tone response was <52.5% of the response to F1 alone. D and E: responses to equal-amplitude 2-tone stimuli delivered from the azimuth function trough (60°) and peak (30°) directions. Responses to F1 alone are shown by individual symbols at 3 and 38 kHz. See text for additional information. F: averaged, normalized, equal-amplitude two-tone response functions obtained from stimuli delivered from trough (60°) and peak (30°) directions. Two-tone data were obtained under monaural conditions.

To compare two-tone responses at different sound directions, we used an "equal-amplitude" two-tone paradigm in which the amplitudes of F1 and F2 were identical in the electrical two-tone signal that was generated by the computer. The equal-amplitude frequency components in the two-tone signal paralleled the equal-amplitude composition of frequency components in the flat spectrum of the electrical signal of the noise stimulus. The relative amplitudes of frequency components in both noise and two-tone stimuli reaching the eardrum were in general unequal due to the transfer function of the sound system (mainly the loudspeaker) and diffraction of sound by the head and pinna. Nevertheless, the relative amplitudes of the frequency components in the two-tone stimulus were equal to the relative amplitudes of the corresponding frequency components in the noise stimulus because the sound system and acoustical diffraction had the same effect on both. Thus direction-dependent differences in two-tone response functions should reflect direction-dependent differences in cross-frequency excitatory-inhibitory interactions produced by the noise stimulus.

Figure 8D shows the cell's responses to six levels of the equal-amplitude two-tone stimuli that ranged from 80 dB attenuation (dBA) to 30 dBA. F2 varied between 5 and 31 kHz, and individual symbols at 3 and 38 kHz indicate responses to the F1 tone alone at the appropriate level (dBA) that were obtained before and after each two-tone series. A two-tone response that is smaller than the F1-alone response indicates two-tone inhibition. A prominent area of two-tone inhibition was seen near 20 kHz, corresponding with the high-frequency inhibitory domain seen in C. Additionally there was a region of weak inhibition that corresponds with the 6-kHz inhibitory domain in C. A two-tone response that is larger than the F1-alone response indicates two-tone facilitation, and facilitation was seen in two-tone responses between 7 and 15 kHz, and between 26 and 31 kHz.

Equal-amplitude two-tone stimuli were presented from directions corresponding to a trough (60°, D) and a peak (30°, E) in the azimuth function. Two-tone inhibition and facilitation are seen at each direction but inhibition is stronger at the trough than at the peak. The two-tone response decreases to near 0 spikes per stimulus at the trough (D) and to a minimum of about one spike per stimulus at the peak (E).

The individual equal-amplitude data (D and E) were averaged across level, normalized as a percentage of response to F1 alone, and the normalized functions are shown in F. Responses <100% represent two-tone inhibition and responses >100% represent two-tone facilitation. The normalized functions exhibit the major patterns of inhibition and facilitation seen in the functions for individual levels. Prominent two-tone inhibition is seen at both directions between 18 and 24 kHz but is stronger in response to stimuli delivered at the trough than at the peak. Additionally both functions show weak two-tone inhibition in the range of 6-7 kHz, although inhibition is a bit stronger to stimuli delivered from the trough than from the peak. Two-tone facilitation is seen in both functions, and the patterns are not identical. If two-tone facilitation contributes to monaural directionality, facilitation would be expected to be stronger at peak than at trough directions. The greatest difference in two-tone facilitation appears in the range of 12-15 kHz, but facilitation is stronger at the trough than at the peak direction. This is opposite of the expectation if two-tone facilitation was to contribute to increased responsiveness at the peak direction.

Figure 9 shows frequency response areas (left) and normalized, equal-amplitude, two-tone functions (right) for four different MD cells. Azimuth functions for the cells in Fig. 9, middle top, middle bottom, and bottom, are illustrated in other figures. The lower two frequency response areas do not show inhibitory domains because those data were not collected. The strongest inhibitory response to the entire set of two-tone stimuli occurred at the trough in each of these cells as was the case for all MD cells that were tested. The two cells in Fig. 9, top and middle top, showed inhibition in frequency domains above and below F1 that was stronger at the trough than the peak. The cells in the third and fourth rows showed inhibition at F2 frequencies above F1, and the inhibition was stronger at the trough than at the peak. Consistent direction-dependent differences in inhibition were not apparent at F2 frequencies below F1 in these cells.

View larger version (64K): [in this window] [in a new window]   Fig. 9. Examples of peak and trough 2-tone response functions for 4 MD cells. Each cell's response is shown in a row with its frequency response area (left) and its average, normalized, equal-amplitude, two-tone response functions (right). Red 2-tone functions were obtained at an azimuth function trough, black functions were obtained at a peak. Individual symbols plotted at 3 and 38 kHz (right) represent responses to F1 alone. No F1 is plotted at 38 kHz in B. All data were obtained under monaural conditions. A and B: responses of unit 9407-7. Two-tone response functions were obtained using the same loudspeaker placed at 2 different azimuths. The noise azimuth function is not shown, but normalized responses were 100% at 90° and 13.1% at 0° azimuth. C-H: azimuth functions for units 9502-8, 9601-9, and 9605-1 are shown in Figs. 6L and 5, C and F, respectively. Rmaxs for A, C, E, and G were 6.8, 4.0, 3.0, and 2.3 spikes/stimulus, respectively. See Fig. 8 for additional details.

Eight MD cells were tested using the equal-amplitude two-tone paradigm. Normalized two-tone data (e.g., Fig. 9, right) were plotted for each cell as a scatter plot in Fig. 10 to further evaluate the relationship between responses to two-tone stimuli from trough and peak directions. Each point in the scatter plot represents a cell's averaged, normalized response at peak and trough directions to two-tone stimuli with the same F2 frequency. Different points represent normalized responses to two-tone stimuli that differ in F2 frequency. Shading of different quadrants indicates whether two-tone inhibition occurs at both peak and trough (dark), either peak or trough but not both (light) or neither peak or trough. (unshaded). Conversely, shading indicates whether two-tone facilitation occurs at peak and trough (unshaded), either peak or trough but not both (light), or neither peak or trough (dark). We use the term facilitation to include all instances in which the response to the two-tone stimulus was larger than the response to F1 alone. In some cases, the cell did not respond to F2 alone (e.g., Figure 8), and in other cases, both F1 and F2 produced excitatory responses when presented alone (e.g., Fig. 9, bottom). In Fig. 10A, all points were located in the lower left quadrant of the graph, showing that each two-tone pair produced inhibition at both directions. In all other cases, responses showed a mixture of two-tone inhibition and facilitation, depending on sound direction, and the particular frequency composition of the two-tone stimulus.

View larger version (51K): [in this window] [in a new window]   Fig. 10. Comparison of responses to 2-tone stimuli at trough and peak directions for 8 MD units. A-H: average, normalized, equal-amplitude, 2-tone response data are represented for each unit in a scatter plot. For example, data from Fig. 9D are shown in A for unit 9502-8. A point in a scatter plot represents a unit's responses at trough and peak directions to a 2-tone stimulus as a percentage of the unit's responses to F1 alone. Different points in each scatter plot represent responses to 2-tone stimuli with different F2. I: sum of data represented in the individual panels. See text for additional details.

In some cells, most points were located above the diagonal line of unity slope showing that normalized two-tone responses for most F2 frequencies were smaller at trough that at peak directions. A paired t-test was performed on data from each cell, and in the case of three cells (A, B, and F), responses at the trough were found to be significantly smaller than at the peak (P values are shown in figure panels). I shows the sum of data for the eight cells. Responses at the trough (91.9 ± 39.3%, mean ± SD) were on average smaller than responses at the peak (96.4 ± 29.7%), and the difference was statistically significant (P < 0.03). To more directly answer the question whether two-tone inhibition was stronger at trough than at peak directions, a t-test was performed on all normalized responses that showed two-tone inhibition to stimuli delivered from the peak, trough, or both directions. These data are located in the darkly and lightly shaded quadrants of I and exclude responses that showed facilitation at both directions (unshaded quadrant). The difference in response magnitude was statistically significant (n = 93, trough, 72.5 ± 30.4%; peak, 83.6 ± 23.8%, P < 0.0000006) showing that two-tone inhibition was stronger at trough than at peak directions in the tested sample of MD cells. Two-tone facilitation could in theory also contribute to monaural directionality if two-tone facilitation were stronger at the peak than at the trough. To address this question, a t-test was performed on all responses that showed two-tone facilitation to stimuli delivered from either the peak, trough, or both directions (unshaded and lightly shaded quadrants of panel I). There was no significant difference in response (n = 73, trough, 117.3 ± 27.6%; peak, 115.0 ± 20.6%, P < 0.43) showing that two-tone facilitation does not contribute significantly to monaural directionality of this sample.

MD cells exhibit stronger inhibitory response attributes than BD cells

According to the side-band inhibition model, MD cells exhibit stronger monaural directionality than BD cells because they have stronger inhibitory side-bands. Stronger inhibitory side-bands might be expected to cause MD cells to exhibit narrower frequency tuning and more strongly nonmonotonic rate level functions than BD cells and cause MD cells to respond with lower discharge rates to noise than tone-burst stimulation. Of the 93 cells for which we had frequency tuning data, 80 had a single excitatory frequency domain and 13 had two excitatory domains. In these latter cells, the two excitatory domains had thresholds within 20 dB of each other and were separated by a frequency range throughout which the cell was not excited by tones <60 dB above threshold. There was no obvious relationship between unit classification and the presence of two excitatory domains (13%, 6/46 BD cells; 8%, 1/12 PB cells; 17%, 6/35 MD cells). Tonal responses for 87 cells were sufficiently complete to allow measurement of excitatory bandwidth at 30 dB re threshold. In the case of a few cells with strongly nonmonotonic level tuning (e.g., Fig. 9A), width was measured at a lower level because the cell was unresponsive at 30 dB. In the case of cells with two excitatory domains, the two bandwidths were averaged to obtain a single value. Responses to tone bursts were obtained under binaural conditions for PB cells and under monaural conditions for most BD and MD cells (see legend of Fig. 11). ANOVA showed a significant bandwidth difference among groups (P < 0.0001). The BD group average bandwidth (1.27 ± 0.84 octaves, n = 46) was significantly greater than the MD (0.63 ± 0.34 octaves, n = 30, P < 0.001) and PB averages (0.63 ± 0.31 octaves, n = 11, P < 0.05). There was no significant difference between the MD and PB groups (post hoc t-test with Bonferroni correction).

View larger version (13K): [in this window] [in a new window]   Fig. 11. Comparison of maximal responsiveness (Rmax) of PB, MD and BD cells to noise and tone bursts. Tone bursts were presented under binaural conditions for PB cells (n = 16) and under monaural conditions for the majority of MD (n = 28/35), BD-EI (n = 31/38), and BD-FI (n = 11/19) cells. Noise and tone Rmaxs were obtained for each cell under identical monaural or binaural conditions. MD cells were less responsive to noise than tones, whereas BD and PB cells showed similar responsiveness to noise and tones. See the text for additional information.

Level-response-function nonmonotonicity for binaural noise bursts was compared for different groups of azimuth-sensitive units (Table 1). Nonmonotonic (NM) strength was defined as the greatest percentage reduction from maximum response with increasing SPL and could vary between 0% (maximum response at highest SPL) and 100% (no response at an SPL higher than that at which the maximum occurred). ANOVA showed a significant difference in nonmonotonic strength among groups for noise (P < 0.001) and BF tone stimulation (P < 0.002). For both noise and BF tone stimulation, average NM strength of MD cells was significantly greater than that of BD cells but not significantly different from that of PB cells. The difference between the BD and PB averages closely approached significance (Table 1).

The maximum responsiveness (Rmax) of individual cells to noise and tone bursts was obtained from azimuth-level data sets and frequency-level data sets, respectively. The noise and tone Rmaxs were obtained under identical binaural or monaural conditions for each cell. Most BD and MD cells were tested under monaural conditions, and PB cells were tested under binaural conditions (see legend of Fig. 11). Figure 11 is a scatter plot that compares each cell's Rmax for noise and tone bursts. As a group, MD cells were significantly less responsive to noise (2.6 ± 2.0 spikes/stimulus) than tones (4.5 ± 4.4 spikes/stimulus, P < 0.005, paired t-test). In contrast, the combined BD-EI and BD-FI cells showed no significant differences in responsiveness to noise (3.6 ± 3.4 spikes/stimulus) and tones (3.4 ± 3.0 spikes/stimulus P < 0.35) nor did the PB group (noise, 2.7 ± 1.9, vs. tone, 2.4 ± 1.5 spikes/stimulus, P < 0.52). These comparisons of frequency tuning, nonmonotonicity of level-response functions, and responsiveness to noise and tones are consistent with the idea that MD cells have stronger inhibitory sidebands than BD cells.

Relationship between spectral notches and azimuth function troughs

Monaural azimuth functions of some MD cells exhibited a response trough at a single azimuth (e.g., Figs. 5 and 6). Response troughs may occur in azimuth functions of DCN neurons when BF coincides with spectral notch center frequency (Imig et al. 2000). We examined our data to determine whether a similar phenomenon occurs in the IC. Spectral notches with center frequencies in the range of 9-13 kHz vary systematically as a function of sound-source direction (Rice et al. 1992). In the frontal horizontal plane, notches with lower center frequencies are located further from the stimulated ear than those with higher center frequencies (spectral notch data, Fig. 12). If spectral notches produce azimuth function troughs, then troughs should occur at the same locations as spectral notches with center frequency equal to unit BF. Three MD cells are illustrated with BFs in this range, and they show the expected relationship. The cell with the lowest BF (8.5 kHz, Fig. 5, top) has a trough located at 30°, furthest from the stimulated ear. The cell with the highest BF (13 kHz, Fig. 6, top) has a trough located at +30°, closest to the stimulated ear. And a cell with an intermediate BF (11 kHz, Fig. 5, bottom) has a trough in an intermediate location. Figure 12 shows the relationship between BF and response trough location for all MD cells that exhibited single-azimuth troughs. The contralateral ear was excitatory for all but one of these cells, and for this one, the location of the trough was reversed in sign so that data are plotted as if the contralateral ear were excitatory in each case. Eight units had BFs between 8.5 and 13 kHz. The slope of the linear regression line fit to these data closely parallels the slope of the line connecting the spectral notch data points, although it is displaced 10-15° toward the left ear.

View larger version (15K): [in this window] [in a new window]   Fig. 12. There is a linear relationship between MD-cell azimuth-function response-trough location in the horizontal plane and BF for BFs in the range of 8.5-13 kHz. The linear regression line fitted to these data closely parallels the line connecting spectral-notch azimuth vs. center-frequency data in the same frequency range.

Comparison of strength of spectral inhibition in the IC, MGB, and AI

We compared the strength of spectral inhibition in the IC with that in the MGB and AI. A measure of strength of spectral inhibition for MD cells was obtained from their monaural-noise azimuth functions. We counted the number of azimuth function values that were 25% of maximum out of the seven azimuths at which each cell was tested (±90, ±60, ±30, 0°). We reasoned that spectral inhibition caused responsiveness to decrease to 25% in the monaural responses of MD cells, so that the greater number of azimuths at which responsiveness was 25%, the stronger the spectral inhibition. Data for the MGB and AI were obtained from previous work from this laboratory (Samson et al. 2000). The number of azimuths at which responsiveness was 25% showed a small increase at successively higher levels (IC, n = 43, 2.3 ± 1.5; MGB, n = 30, 2.7 ± 1.6; AI, n = 28, 3.0 ± 1.7), but the differences were not statistically significant.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Comparison of single-unit responses to noise bursts delivered under monaural and binaural conditions allowed identification of monaural and binaural contributions to directionality in the IC. Directional cells were by definition those that exhibited 75% azimuth-dependent modulation in responsiveness, where responsiveness is taken as an average response to a wide range of SPLs. Thus directional cells responded well at certain azimuths and relatively poorly, if at all, at others, even at high SPLs. BD cells were directional under binaural but not monaural conditions. On the other hand, MD cells were directional under both binaural and monaural conditions. This shows that MD cells were more sensitive to directional cues present in monaural noise bursts than were BD cells.

MD cells in the IC were directional to noise but not tone bursts under monaural conditions. Similar findings have been reported for MD cells in AI and the MGB (MGB: Imig et al. 1997; AI: Samson et al. 1993; MGB and AI: Clarey et al. 1995). From these results, we infer that MD cells derive directionality from direction-dependent spectral shape variation that is present to a much greater extent in noise than in tone bursts. We are unaware of any other potential directional cue present in noise bursts and so feel that this assumption is reasonable.

All high-frequency neurons in the auditory system exhibit directional responses to low-level stimulation under monaural conditions, i.e., they are responsive at some directions and not others (e.g., Moore et al. 1984) (monaural ALRAs in this report). If one views the magnitude of a neuron's spike discharge as reflecting the amount of energy falling in its excitatory frequency domain (energy detector model), then low-level monaural directionality is expected due to sound amplification by the pinna and sound shadowing by the head. Given high-level stimuli, an energy detector would be expected to respond well to all directions. Auditory nerve fibers (Poon and Brugge 1993; Rice et al. 1995) and many BD cells in the IC, MGB, and AI do respond well at all directions to monaural stimulation (Samson et al. 1994, 2000). In contrast, MD cells are relatively unresponsive to noise bursts at certain azimuths even at high SPLs, a characteristic that is inconsistent with the energy detector model. A deep spectral notch, for example, may cause a reduction in sound level of 20 dB or so at its center frequency. A notch at BF might cause a low-level noise stimulus to drop below threshold. If a cell is responding only to the amount of energy near BF, then an increase in stimulus level by 20 dB should result in a response. But this does not occur as can be seen in the ALRAs of the MD cells. These cells failed to respond or responded only weakly at certain azimuths regardless of SPL.

According to the side-band inhibition hypothesis, the responsiveness of a MD cell depends on the relative balance of acoustic energy in the cell's excitatory and inhibitory frequency domains. Spectral shapes that increase energy in excitatory relative to inhibitory frequency domains produce an increase in responsiveness, shapes that increase energy in inhibitory relative to excitatory domains produce a decrease in responsiveness. Assume that a notch (or some other spectral feature) reduces energy in an excitatory domain relative to adjacent inhibitory domains. Raising SPL of a noise burst does not cause the cell to respond because energy increases not only in the excitatory domain but also in inhibitory domains. Because inhibition predominates, the cell does not discharge. Although there is no direct evidence that "spectral inhibition" is a result of synaptic inhibition, neither is "side-band inhibition" believed to result entirely from synaptic inhibition. Nevertheless, because the term "side-band inhibition" is commonly accepted, then "spectral inhibition" should be likewise accepted as both appear to reflect common mechanisms.

The side-band inhibition hypothesis gains support from finding that two-tone inhibition is stronger at directions where an MD cell is relatively unresponsive to a noise stimulus than at directions where it is responsive. Similar findings have been previously reported in the MGB (Imig et al. 1997). MD cells commonly showed two-tone facilitation in addition to two-tone inhibition, and two-tone facilitation has previously been reported in the IC (cat: Ehret and Merzenich 1988; mouse: Egorova et al. 2001, bat: Mittmann and Wenstrup 1995; Wenstrup et al. 1999; Yan and Suga 1996). In theory, two-tone facilitation could also contribute to monaural directionality, but our data suggest that it is not a significant factor.

Noise bursts containing artificially produced spectral notches inhibit the discharge of DCN neurons when center frequency coincides with BF (Nelken and Young 1994; Spirou and Young 1991). Noise bursts containing naturally occurring spectral notches also inhibit the responses of DCN neurons. Troughs in noise-burst azimuth functions occurred at locations where spectral-notches' center frequencies were expected to coincide with unit BF (Imig et al. 2000). The present results strongly suggest that notch inhibition can also account for response troughs in some IC MD cells' azimuth functions based on the same reasoning. When a notch is centered at a cell's BF, it reduces energy in an excitatory frequency domain with respect to inhibitory domains, thus causing a decrease in responsiveness. Changing sound-source direction causes the notch to shift in frequency away from the cell's excitatory domain, energy increases in excitatory relative to inhibitory domains, and the cell's responsiveness increases. The result is a trough in the cell's azimuth function where the notch was centered at BF. Spectral notches with center frequencies in the range of 9-13 kHz show a linear relationship to azimuth in the horizontal plane (Rice et al. 1992) that parallels azimuth function trough location for IC neurons with BFs in the same range. This is the expected result if a trough is created when a spectral notch is centered at a cell's BF.

There is about a 10-15° disparity between the acoustic and IC neural data. It is not surprising that a similar disparity was found in the DCN because both the IC and DCN studies used the same protocol for head and pinna orientations. At least two factors could account for the disparity between the neural and acoustic data. The disparity may be due to differences in pinnae and head orientation used to obtain the acoustic and neural data as spectral notch location is known to depend on pinna orientation (Young et al. 1996). Additionally, disparities of this magnitude are found in acoustical measurements among individual animals (Xu and Middlebrooks 2000).

The side-band inhibition model of spectral inhibition can in theory account for the peak-shaped azimuth functions exhibited by some MD cells in the IC and in the MGB and AI (Imig et al. 1997; Samson et al. 1993, 2000). One can imagine that a cell with a strong inhibitory and relatively weak excitatory frequency domain would be inhibited by noise stimulation except at locations where energy in the inhibitory domain was reduced relative to the excitatory domain. This could occur when a spectral notch is centered on the inhibitory domain. In this case, the cell is unresponsive at most azimuths because of spectral inhibition and responds at one or a few azimuths because of spectral disinhibition. While we have focused on notches, there is no reason to believe that notches are the only spectral shape to which MD cells are sensitive.

Delgutte et al. (1999) described directional responses of IC neurons using virtual space stimuli. Ten neurons showed troughs in their midline elevation functions, and in 4/10, there was a spectral notch corresponding to unit BF at that location. This finding appears somewhat inconsistent with our results that indicated spectral notches corresponded with response troughs for all MD cells with BFs in the range of 8.5-13 kHz. It is quite possible that the apparent inconsistencies between the two sets of data reflect methodological differences. Delgutte et al. (1999) used a single rather low level of stimulation. We used a broad range of SPLs, and thus response troughs in our data are not level-dependent and indicate locations where spectral inhibition is strongest. Additionally, Delgutte et al. (1999) used binaural stimulation so it is possible that binaural interactions may have contributed to troughs in their data. We presented stimuli under monaural conditions and thus only monaural cues could contribute to troughs.

According to the side-band inhibition model of spectral inhibition, MD cells should have stronger and/or broader inhibitory frequency domains than BD cells. Response properties of MD and BD cells in the IC showed several significant differences that are consistent with this prediction. MD cells exhibited significantly narrower excitatory frequency domains and greater nonmonotonic strength to noise and BF tone bursts than BD cells. Additionally, MD cells had lower discharge rates to noise than tones, whereas BD cells had similar discharge rates to tones and noise. Together these differences suggest that MD cells receive stronger inhibitory input than BD cells. MD and BD groups in the MGB and AI show similar differences in frequency tuning, nonmonotonic strength to noise stimuli, and discharge rates to noise and tones (Clarey et al. 1995; Imig et al. 1997; Samson et al. 2000). These data suggest a common mechanism of monaural directionality at these different sites in the auditory system.

Monaural spectral inhibition makes a relatively modest contribution to the azimuth tuning of BD-EI, BD-FI, and BD-F cells in the IC. The contribution to PB cells cannot be determined using the method of monaural/binaural response comparison because PB cells do not respond under monaural conditions. Stimulus bandwidth appears to be important for the directionality of at least some PB cells as their directionality to tonal stimulation is much less than that to noise. These findings regarding BD, MD, and PB cells are consistent with previous observations in the MGB and AI (Clarey et al. 1995). Our data are insufficient to determine whether spectral-dependent directionality reflects monaural spectral inhibition in monaural inputs to PB cells, a binaural mechanism of spectral sensitivity, or both.

The azimuth preferences and patterns of binaural interactions that we observed using free-field stimulation and ear plugging are consistent with those expected from previous studies using independent earphone stimulation of each ear. Most cells had azimuth preferences that varied from near the midline to the contralateral pole, although a few had ipsilateral preferences in agreement with previous studies. Three patterns of binaural interactions could be recognized that contributed strongly to azimuth tuning. These include binaural inhibition, mixed binaural facilitation and inhibition, and binaural facilitation. Previous studies reported similar patterns of binaural interactions in the IC of cats using noise, tones, clicks, and virtual space noise bursts delivered through earphones sealed to the ears (e.g., Delgutte et al. 1999; Geisler et al. 1969; Irvine 1986; Irvine and Gago 1990; Rose et al. 1966).

Previous work shows that directional tuning of some IC neurons in the cat is dependent on stimulus bandwidth. Aitkin and Martin (1987) observed that 28% of their sample of high-frequency neurons was sensitive to the azimuth of broadband noise but not tone bursts. Neurons in the IC of the bat are more sensitive to the azimuth of two-tone stimuli consisting of a BF tone and a second tone in an inhibitory frequency domain than to BF tones presented alone (Zhou and Jen 2000). These results show that directionality depends on stimulus bandwidth but don't indicate whether the mechanism was binaural or monaural.

Azimuth tuning of IC neurons to noise burst stimulation may reflect predominantly binaural interaction, predominantly monaural spectral inhibition, or a combination of both. It is possible that the variety of response patterns described here are the result of two independent processes, binaural mechanisms that determine different types and strengths of binaural interactions, and monaural mechanisms that determine the strength of spectral inhibition. While this seems to be consistent with our IC data, we would not have suggested this based on our data from the MGB and auditory cortex as MD cells showing facilitation or mixed binaural interactions were much less commonly identified in those structures (Samson et al. 1993, 2000). Nevertheless, there were a few. It is possible that these types of responses really are less frequently encountered in the forebrain or that we simply failed to recognize them as often.

Binaural interactions that contribute to IC neural directionality are believed to originate in the superior olivary system, the dorsal nucleus of the lateral lemniscus and IC (literature reviewed by Delgutte et al. 1999; Tollin and Yin 2002a). The sources of MD sensitivity are much less certain. MD sensitivity may be transmitted from lower levels of the auditory system to the IC, generated locally within the IC, or both.

DCN neurons are more sensitive to spectral shape and show greater monaural directional sensitivity than ventral cochlear nucleus (VCN) neurons (Imig et al. 2000; Young et al. 1992) or auditory nerve fibers (Poon and Brugge 1993; Rice et al. 1995). DCN neurons provide excitatory input to the IC (Semple and Aitkin 1980), and Davis (2002) has demonstrated that some IC neurons receive input predominantly from the DCN. Thus it is possible that at least some MD cells in the IC receive input directly from the DCN.

Although VCN neurons appear less sensitive to spectral shape than DCN neurons, it is unknown whether or not neurons in most of the brain stem nuclei that receive input from the VCN are sensitive to spectral shape. The exception is the lateral superior olive (LSO) that receives excitatory input from the ipsilateral VCN and inhibitory input from the contralateral VCN via relay in the medial nucleus of the trapezoid body (Schwartz 1992). Tollin and Yin (2002b) used virtual space stimuli to experimentally investigate the contribution of binaural disparities and spectral shape in producing azimuth tuning in LSO neurons. They found that binaural inhibition acting on interaural level differences was most important and that spectral shape variation made little contribution. They suggested that LSO neurons are likely major contributors to ILD-sensitive IC neurons that exhibit binaural inhibition. In our study, BD-EI cells might be expected to be recipients of LSO input as their directionality depends predominantly on binaural disparities and not on spectral shape.

Although neurons in DCN pathways may be more sensitive to spectral shape than neurons in VCN pathways, both appear to be important for accuracy of monaural, spectral-shape-dependent sound localization in cats. Lesion of the dorsal acoustic stria, that carries DCN fibers to the IC, causes a decrease in accuracy of reflex orientation to noise bursts that vary in elevation, suggesting that the DCN plays an important role in this spectral-shape dependent localization behavior. On the other hand, the same lesion does not cause an increase in minimum audible angle in elevation (May 2000; Sutherland et al. 1998a,b), suggesting that the VCN pathways convey sufficient monaural spectral shape information for this localization task. It appears that both the DCN and VCN pathways could potentially convey information that ultimately contributes to the MD responses in the IC.

IC neurons exhibit greater monaural directionality than DCN or VCN neurons showing that much of the MD sensitivity in IC neurons appears to be generated above the level of the CN and quite possibly within the IC. DCN neurons are more sensitive to monaural spectral shape cues than VCN neurons or auditory nerve fibers (Poon and Brugge 1993; Rice et al. 1995), but DCN spectral sensitivity cannot account for that seen in the IC. Only ~2% (1/61) of the presumed output neurons recorded in the DCN exhibited noise-burst azimuth-function modulation 75% (Imig et al. 2000), whereas an estimated 23% of the high BF cells in the IC exhibit noise-burst azimuth-function modulation 75%. This estimate is based on the fact that 69% of the IC neurons that we recorded were directional and that 33% (43/129) of the directional cells that were tested showed monaural directionality (0.33 * 0.69 = 0.23). Furthermore, peak-shaped monaural azimuth functions are found in the IC but not the DCN (Imig et al. 2000). These are presumably a result of spectral disinhibition that depends on very strong inhibitory side-bands. Such strong inhibition is not present in the CN and could be produced locally within the IC. Iontophoretic injection of GABAergic and glycinergic agonists and antagonists show that inhibitory mechanisms within the IC contribute to side-band inhibition (Davis 2002; Faingold et al. 1991; Fuzessary and Hall 1996; Pollak and Park 1993; Vater et al. 1992; Yang et al. 1992). As side-band inhibition is believed to be an important basis for monaural directionality, this suggests that much MD sensitivity may be generated with the IC.

Our data do not show compelling evidence for a significant increase in strength of spectral inhibition in the MGB and AI as compared with the IC. The estimated proportion of high-frequency cells that are monaurally directional is slightly lower in the forebrain as compared with the IC. MD cells comprise ~23% of the high-frequency sample in the IC and ~18% of the samples in the MGB and AI (Barone et al. 1996; Samson et al. 2000). The estimated strength of spectral inhibition is sequentially greater from the IC to the MGB to AI, but the difference is not statistically significant. It appears that there are significant increases in spectral inhibition between the level of auditory nerve fibers and the DCN and between the DCN and the IC but relatively little additional change between the IC and the forebrain.

On the basis of this comparison, it is not unreasonable to assume that MD responses in the MGB and AI at least in part reflect response properties that are transmitted from lower levels. Nevertheless, inhibitory side-bands are also locally generated in each structure (MGB: Suga et al. 1997; auditory cortex: Chen and Jen 2000; Wang et al. 2000), and this could contribute to MD responses in the forebrain. Some cells in the MGB and IC have frequency tuning characterized by two or more excitatory domains. In the MGB, a significantly higher proportion of MD cells than BD cells had multiple excitatory domains (Imig et al. 1997), but this was not true in the IC sample. Overall, the MGB sample contained a somewhat higher proportion of cells with two or more excitatory domains (15/65, 23%) than the IC sample (13/93, 14%). It is possible that generation of MD responses with multiple excitatory frequency domains in the MGB could account for these differences.

Delgutte et al. (1999) questioned the validity of the ear plugging method due in part to the effect of sound leakage through the earplug and in part due to the reproducibility of attenuation produced by multiple earplug insertions. There is no doubt that greater interaural cross-talk occurs with the ear-plugging method than with earphones sealed to each ear. Nevertheless attenuation produced by earplugs is adequate to answer the questions we asked. Ear plugging by injection of ear mold compound typically produces attention of 40-60 dB in the frequency range to which our sample of cells was most sensitive (>4 kHz). This amount of attenuation was sufficient to unambiguously identify a neuron's binaural interactions and monaural directionality. There was a minor problem of interpreting whether the high-threshold monaural responses obtained during ear plugging in some cells were really monaural responses or responses to sound leaking though the ear plug. We opted for the latter interpretation although this has no major bearing on the major findings in this report. Additionally, similarity of neural responses to repeated sets of binaural stimuli (e.g., azimuth functions in Figs. 1, 3, 4-6, and 8) separated by insertion and removal of one or more earplugs shows that ear plugging had an effect only when the earplug was in place. Plugging had no lasting effect on a unit's response after the plug was removed. Similarity of neural responses to repeated stimulus sets under monaural conditions separated by removal and reinsertion of the earplug shows that similar attenuation was typically produced each time the ear was plugged. Our data are completely consistent with previous studies regarding the types of binaural interactions and the relationship of binaural interaction types to receptive field locations, including those reported by Delgutte et al. (1999). Sound leakage through the earplug does not have a significant effect on interpretation of binaural interactions. We cannot think of a way that sound leakage could cause the patterns of spectral inhibition that were apparent in our data. Thus while the issues raised by Delgutte et al. (1999) are certainly potential problems, our results suggest that they do not seriously hinder data interpretation.

Some unilaterally deaf humans are able to localize sounds with considerable accuracy in all three dimensions, showing that monaural spectral shape cues potentially could provide information useful for right-left localization (Slattery and Middlebrooks 1994). Nevertheless it is generally believed that normal listeners, under binaural conditions, do not rely on spectral shape cues for localization in this dimension (Middlebrooks 1992; Middlebrooks and Green 1991; Wightman and Kistler 1992, 1997). This raises the question of the possible function of MD cells in the IC that derive azimuth sensitivity from monaural spectral shape cues. Front-back confusion is believed to result because binaural disparities are identical or nearly so at directions that are symmetrical with respect to the interaural axis. MD cells presumably could provide useful information to resolve front-back confusion because monaural spectral shape cues are not symmetrical about the interaural axis. Cell's deriving azimuth tuning predominantly from binaural disparities might be expected to show symmetrical azimuth tuning about the interaural axis, and an example of a BD-EI cell in the MGB with symmetrical tuning has previously been published (Fig. 1 in Samson et al. 2000). Spectral shape cues do not exhibit interaural axis symmetry (Musicant et al. 1990; Rice et al. 1992), so MD cells would be expected to show nonsymmetrical azimuth tuning about the interaural axis, as is the case (e.g., Figs. 4, top, and 8, A and B). Examples of other MD cells with nonsymmetrical front-back azimuth tuning have also been illustrated in AI (Samson et al. 1993) and the MGB (Samson et al. 2000). Thus azimuth sensitivity dependent on monaural spectral shape variation may play a role in front-back localization. Monaural spectral shape sensitivity may also play a role in minimum audible angle (MAA) tasks that require detection of whether the location of two sound sources is the same or different not the actual location of the source. Unilaterally deafened cats can detect differences in the elevation of a sound source with similar accuracy to normal cats, suggesting that cats normally utilize monaural shape sensitivity in this task (Sutherland et al. 1998a). We are unaware of studies that have tested unilaterally deafened cats for MAA sensitivity in azimuth. Binaural disparities are generally believed to contribute strongly to MAA sensitivity in azimuth at least near the median plane (Middlebrooks and Green 1991; Mills 1958). The degree to which monaural spectral shape cues may contribute to MAA sensitivity in azimuth, especially near the lateral poles, under normal listening conditions is unknown. Experiments using virtual space stimuli in which monaural spectral-shape and binaural disparity cues are independently manipulated offer an opportunity to shed light on this question.