| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
The Journal of Neurophysiology Vol. 80 No. 5 November 1998, pp. 2417-2432
Copyright ©1998 by the American Physiological Society
Department of Physiology and Waisman Center, University of Wisconsin, Madison, Wisconsin 53705
 |
ABSTRACT |
|---|
 Abstract
 Introduction
Methods
Results
Discussion
References
|
|---|
Brugge, John F., Richard A. Reale, and Joseph E. Hind. Spatial receptive fields of primary auditory cortical neurons in quiet and in the presence of continuous background noise. J. Neurophysiol. 80: 2417-2432, 1998. Spatial receptive fields of primary auditory (AI) neurons were studied by delivering, binaurally, synthesized virtual-space signals via earphones to cats under barbiturate anesthesia. Signals were broadband or narrowband transients presented in quiet anechoic space or in acoustic space filled with uncorrelated continuous broadband noise. In the absence of background noise, AI virtual space receptive fields (VSRFs) are typically large, representing a quadrant or more of acoustic space. Within the receptive field, onset latency and firing strength form functional gradients. We hypothesized earlier that functional gradients in the receptive field provide information about sound-source direction. Previous studies indicated that spatial gradients could remain relatively constant across changes in signal intensity. In the current experiments we tested the hypothesis that directional sensitivity to a transient signal, as reflected in the gradient structure of VSRFs of AI neurons, is also retained in the presence of a continuous background noise. When background noise was introduced three major affects on VSRFs were observed. 1) The size of the VSRF was reduced, accompanied by a reduction of firing strength and lengthening of response latency for signals at an acoustic axis and on-lines of constant azimuth and elevation passing through the acoustic axis. These effects were monotonically related to the intensity of the background noise over a noise intensity range of~30 dB. 2) The noise intensity-dependent changes in VSRFs were mirrored by the changes that occurred when the signal intensity was changed in signal-alone conditions. Thus adding background noise was equivalent to a shift in the threshold of a directional signal, and this shift was seen across the spatial receptive field. 3) The spatial gradients of response strength and latency remained evident over the range of background noise intensity that reduced spike count and lengthened onset latency. Those gradients along the azimuth that spanned the frontal midline tended to remain constant in slope and position in the face of increasing intensity of background noise. These findings are consistent with our hypothesis that, under background noise conditions, information that underlies directional acuity and accuracy is retained within the spatial receptive fields of an ensemble of AI neurons.
Under natural conditions a listener often must use acoustic cues to detect, identify, and locate particular sounds in environments filled with competing background noise. For the listener to accomplish these feats, the auditory system had to develop mechanisms not only to minimize the interference that one sound inflicts on another but to do so over a wide range of intensity of both signal and noise. There has been a long history of psychophysical studies of auditory masking, and indeed the auditory system does employ different processing modes depending on the temporal relationships and interaural configurations of signal and masker (for reviews see McFadden 1975 Detailed descriptions of the methods of animal preparation, synthesis of virtual space signals, extracellular single-neuron recording, and data collection are found in previous publications (Brugge et al. 1994 Acoustic stimulus generation
Acoustic stimuli were tone pips (50-ms duration, 5 ms rise-fall time), continuous broadband noise, and impulsive transients (6.4-ms duration); the latter mimicked in their spectrum and time waveform sounds arriving from a source in free space. Tone pips delivered monaurally or binaurally were used to estimate the characteristic frequency (CF; that frequency for which threshold is lowest) of a neuron. We did not examine systematically other tone-evoked response properties of these cells, as the main goal of the experiment was to study parameterically their directional responses to acoustic transients. Continuous broadband noise by itself did not evoke responses (other than at onset) from any AI neuron in this study, although in all cells studied it modified the responses to directional transient signals. The directional signal was a broadband (3-dB corner frequencies at 800 Hz and 40 kHz) or narrowband (3-dB bandwidth = 2 kHz) impulsive transient waveform originally generated by a sound source in free space and realized as the impulse response of a linear-phase finite-impulse-response (FIR) filter (Chen et al. 1994
Spatial receptive field measurements
VSRF properties were studied with a protocol that contained three stimulus paradigms. Each paradigm was executed both without and with a continuous noise background being present. For the "signal-in-noise" conditions, the continuous noise was turned on minutes before the start of data collection and remained on until the full compliment of directional signals was presented. The effects of changing intensity of the signal or the background noise were studied systematically. The noise attenuators were set to 110 dBA between paradigms. Several hours of recording from each neuron were required to complete the full protocol.
Results were obtained from seven cats drawn from a larger population of experiments in which we studied the VSRFs of several hundred neurons under various VAS conditions. Thirty of these neurons responded to directional transients and remained in contact with the electrode long enough to be studied parametrically and quantitatively for the effects of background noise on their spatial receptive field properties. The majority of VSRFs were categorized as falling predominantly in the contralateral (56%) or ipsilateral (19%) acoustic hemifield. The remainder were considered frontal (7%), omnidirectional (15%), or complex (3%). These proportions are close to those reported earlier (Brugge et al. 1994 Effects of noise on the size of the VSRF
The effects of continuous background noise on the spatial extent of a VSRF are introduced in Fig. 2 with data from two neurons. Each black square on a VSRF map denotes an effective stimulus. The left hemisphere of each connected pair of hemispheres represents frontal acoustic space as seen by the cat. The front right acoustic quadrant is always contralateral with respect to the left AI under study. The right hemisphere of the pair represents the rear acoustic hemifield, with the right rear quadrant adjoining the right front quadrant at +90°.
AA functions
Changes in the spatial extent of an AI neuron's VSRF consequent to the addition of continuous background noise reveal nothing about the effects such addition might have on the possible information bearing response properties within the VSRF. To study this question, we first examined the changes in a cell's discharge properties that took place at the acoustic axis as a function of noise intensity (AA functions). Thus, for AI cells with measurable frequency tuning and CF, VAS transient signals positioned on or near the acoustic axis typically have the lowest detection threshold. Illustrative data from one AI neuron are presented in Fig. 4A. Here the AA functions are plotted for both spike count (filled symbol) and mean first spike latency (open symbol). Plotted data were obtained when the intensity of the signal was held constant and that of the background noise was raised in steps of 5 dB. As is typical for AI neurons under the conditions of this experiment, the response to each of the effective VAS transient signals was a single action potential tightly time locked to stimulus onset. The neuron responded to almost every signal of the 40 presented with a mean response latency of 16.5 ms. The dynamic range, from maximal to minimal response, required only a 15- to 20-dB change in noise intensity. The fall in response strength was accompanied by a concomitant rise in response latency of~4 ms. We encountered no exceptions to these forms of AA functions, 13 of which are plotted in Fig. 4, B and C, to illustrate the generality of the findings. Here response strength was normalized to the maximum count.
AZ and EL functions
We examined the changes in a cell's discharge properties as a function of noise intensity that took place at multiple directions in the VSRF. A set of directions with constant elevation (AZ function) or constant azimuth (EL function) were chosen to pass through the acoustic axis and thereby cross regions in the VSRF where, for many AI neurons, response gradients tend to be steepest (see Brugge et al. 1996
Background noise results in a change in signal threshold
During the course of this study and from our earlier findings (Brugge et al. 1996
There are three major findings in the work presented. First, the presence of continuous wideband background noise reduced the size of the spatial receptive fields of AI neurons and simultaneously reduced firing strength and lengthened response latency throughout the VSRF. Detailed data were obtained for signals at the acoustic axis and on-lines of constant azimuth and elevation that passed through the acoustic axis. These effects were monotonically related to the intensity of the background noise over a noise intensity range of ~30 dB. At the highest noise intensities studied, the discharge of the cell could be suppressed entirely at all signal directions studied. Second, the changes in these spatial receptive field properties as a consequence of increasing noise intensity were mirrored by the changes that occurred when the signal intensity was reduced in signal-alone conditions. Thus for these response parameters the addition of background noise was equivalent to a shift in the threshold of a directional signal, and this shift was seen across the spatial receptive field. Third, the spatial gradients of response strength and latency remained evident over the range of background noise intensity that reduced spike count and lengthened onset latency. In particular, those gradients along the azimuth that spanned the frontal midline tended to remain relatively steep and fixed in azimuthal position in the face of increasing intensity of background noise.
INTRODUCTION
Abstract
Introduction
Methods
Results
Discussion
References
; Moore 1993). Most psychophysical and physiological studies of binaural masking have been designed with the idea that signal extraction by the central auditory system is used by the listener to achieve maximal detection performance rather than to derive information about the signal's spatial location. There have been few reported psychophysical studies of masking of sound-source location or direction under continuous masking conditions. Good et al. (1997) and Good and Gilkey (1996) explored how in the free field the presence of continuous background noise influences a listener's ability to make an absolute judgment on the true location of a sound. Their results indicate that, whereas accuracy of localization judgments is clearly affected by the presence of masking noise, relatively good localization performance can be observed even where signals are near detection threshold. This implies that the neural mechanisms underlying localization of a sound source are relatively tolerant of competing background sound.
, 1996, 1997a,b; Reale et al. 1996, 1998). For many AI neurons, there exists within the spatial receptive field a gradient of response latency or strength, or both, with latency shortest and discharge strength greatest at or near an acoustic axis. We postulated that such gradients in the receptive field could serve to encode sound-source direction in the output of an ensemble of AI neurons whose receptive fields were organized this way, and a quantitative model was proposed that upholds these predictions (Jenison 1998). If our hypothesis is correct that the direction of acoustic transients is encoded in the internal structure of AI spatial receptive fields, then one would predict that this structure would remain relatively intact with wide swings in the levels of both the signal and the background sound. Previous studies indicated that for a single AI cell spatial gradients could remain relatively constant across changes in signal intensity (Brugge et al. 1996). We might therefore expect that a similar structural constancy would obtain when a background noise is introduced into the acoustic environment. To test this we have begun to study AI spatial receptive fields in noisy environments, taking advantage of our ability to generate sounds in VAS. The experiments presented in this paper address the issue of whether and to what extent directional sensitivity to a transient signal, as reflected in the VSRF of AI neurons, is retained in the presence of a continuous background noise. This binaurally uncorrelated noise corresponds to a diffuse sound field that might represent a highly reverberant environment or an environment containing many independent sound sources.
METHODS
Abstract
Introduction
Methods
Results
Discussion
References
, 1996; Reale et al. 1996, 1998). Briefly, cats were anesthetized with sodium pentobarbital (40 mg/kg ip), and a venous catheter was inserted for supplemental administration of the drug (8 mg/kg iv) to maintain a surgical plane of anesthesia. Fluid replacement was maintained with a continuous infusion (3 ml·kg
1·h1) of 5% dextrose in 0.9% sodium chloride. Additional daily medications included atropine sulfate (0.1 mg/kg sc), dexamethasone sodium (0.2 mg/kg iv), and procaine penicillin (300K units im). Body temperature was maintained ~38° by a thermostatically controlled DC heating blanket. A tracheal cannula was inserted, and the pinnae and other soft tissue removed from the head. Hollow earpieces for sound delivery were inserted into the ear canals, sealed in place, and connected to specially designed earphones (Chan et al. 1993). The transfer characteristics of the left and right ear sound delivery systems were measured in vivo near the tympanic membrane. A chamber was cemented to the skull over the left auditory cortex, filled with warm silicone oil and hydraulically sealed with a glass plate on which a Davies-type microdrive was mounted.
). The latter had its center frequency set equal to the CF of the neuron and was employed for those neurons where broadband transients did not evoke consistent responses. Digitally synthesized signals (i.e., tone pips and impulsive transients) were compensated for the transmission characteristics of the earphones and sound delivery pathways (Brugge et al. 1994; Chen et al. 1994) before D/A conversion with a digital stimulus system (DSS). Separate DSS attenuators controlled the intensity of signals delivered to each ear, which in these experiments were set to equal values.
; Rice et al. 1992). The simulation of these directional sounds constitutes a VAS. Improvements in the synthesis and applications of VAS for neurophysiological studies is an ongoing process in our laboratory. Only a brief description and chronology of the relevant work is given here, as the details are found in the published articles (for review see Reale et al. 1996, 1998). First, we obtained an accurate approximation, by using FIR filters (Chen et al. 1994), of the free field to eardrum transformation for each sound-source direction with empiric free-field measurements obtained by Musicant and colleagues (1990). Second, we designed a closed field sound delivery/measurement system that did not modify the spectrum of signals delivered to the eardrums (Chan et al. 1993). Third, we designed and acoustically validated a frequency-domain general mathematical model of VAS that used as input the FIR-filter-approximated transformations and provided as output a veridical interpolation of VAS (Chen et al. 1995). Last, the VAS model was enhanced to be used as an interactive tool in binaural electrophysiological studies and was implemented in the time domain by a quasi-real-time algorithm (Reale et al. 1996; Wu et al. 1997). With this tool, any number of simulated signals can be synthesized for selected sound-source directions positioned in a spherical coordinate system [0-360° azimuth, 
36° to +90° elevation] and centered on the cat's head. For each individually modeled cat and a particular sound source (e.g., broadband transient), there is a unique relationship between sound-source direction and the peak-to-peak amplitudes of synthesized VAS signals. The intensity of any VAS signal is expressed simply as dB attenuation (dBA) relative to the maximum peak-to-peak amplitude for that cat. The "signal alone" condition in this study was equivalent to the signal in the presence of noise attenuated by 110 dB (110 dBA_NOISE).
3 dB) was produced by separate analog noise generators (MDF 9302, MDF Products, Danbury, CT), for the left and right ear sound delivery channels. For each channel the noise was led through a separate attenuator (HP model 350D) and audio transformer (Y-27220, Pico Electronics, Mt. Vernon, NY) before being mixed with a VAS transient signal produced by the DSS. The separate HP attenuators controlled the intensity of background noise delivered to each ear and, in these experiments, were set to equal values. The combined VAS signal and noise was fed to a power amplifier and then to the DSS attenuator before earphone transduction. Thus the spectrum of the VAS signals, but not that of the continuous noise, was corrected for the spectral changes introduced by the sound delivery system. Figure 1 presents the average and range of the magnitude spectrum for the stimulus-delivery system's transfer function for all the animals in this study. In general, the extreme values depart from the mean by 
5 dB. The ordinate of the curves show the maximal SPL attainable for tones at frequencies between 156 Hz and 35 kHz, and their shapes represent the expected spectrum of the broadband continuous noise measured near the tympanic membrane. The maximum peak-to-peak amplitude of the continuous noise was~5 dB below that of pure tones over the range of frequencies studied; the intensity of the noise was expressed as dBA relative to this maximum.
View larger version (24K):
[in a new window]
FIG. 1.
Average (solid line) and range (dashed lines) of the magnitude spectrum for the stimulus-delivery system transfer function for all animals in this study. The ordinate shows the maximal SPL attainable for tones at frequencies between 156 Hz and 35 kHz, and their shapes represent the expected spectrum of the broadband continuous noise measured near the tympanic membrane. The maximum peak-to-peak amplitude of the continuous noise was~5 dB below that of pure tones over the range of frequencies studied. Intensity of the noise is expressed as decibel attenuation (dBA_NOISE) relative to this maximum.
, 1996). As a rule, we initially obtained a VSRF at an intensity 20-30 dB above the threshold determined at the acoustic axis (AA), which is that direction, for a single frequency, where maximum pressure gain is recorded near the tympanum. For each neuron in our sample the CF frequency was used to define an AA that was then assigned to the cell. Unless the ears are perfectly symmetrical, there is a different set of acoustic axes for each ear (Musicant et al. 1990). The VAS constructed for our experiments employed symmetrical models for the left and right ears; thus we further defined the cell's AA to be on the side that yielded the lowest threshold. For the signal-in-noise condition, VSRFs were determined at one or more intensity of the continuous background noise. Second, we recorded the responses to 40 repetitions of a VAS transient signal that had a direction coincident with the acoustic axis. The strength and latency of the response were estimated from the total number of spikes and mean first spike latency, respectively. For the signal-in-noise condition, these response metrics were plotted as a function of intensity of the background noise and for the "signal-alone" condition, as a function of signal intensity. Third, we recorded the responses to 40 repetitions of VAS transient signals that had regularly spaced (9°) directions distributed along either the azimuthal or elevational dimension and that passed through the acoustic axis. Graphs of response strength or latency as a function of direction are termed azimuth (AZ) and elevation (EL) functions, respectively (see also Imig et al. 1990; Rajan et al. 1990). For the signal-in-noise condition, AZ and EL functions were determined at one or more intensity of the continuous background noise.
RESULTS
Abstract
Introduction
Methods
Results
Discussion
References
). The CFs of these neurons ranged from 7.3 to 28.6 kHz.
View larger version (57K):
[in a new window]
FIG. 2.
Virtual space receptive fields (VSRFs) from 2 AI neurons under conditions of signal alone (top, 110 dBA
NOISE) and signal in the presence of continuous background noise. Noise intensity shown above each VSRF.
View larger version (80K):
[in a new window]
FIG. 5.
Effects of noise on VSRFs and azimuth (AZ) and elevation (EL) functions for a single AI neuron. Dots on the top left globe indicate directions from which a signal was presented. Trajectories pass through the acoustic axis. VSRFs plotted as in Fig. 2. Spike count, normalized spike count, mean 1st spike latency, and change in mean 1st spike latency as a function of AZ (A-D) or EL (E-H). Noise intensity given in the inset. See text for details.
View larger version (58K):
[in a new window]
FIG. 7.
VSRFs obtained under conditions in which the transient signal was presented in the presence of background noise (left column) and when the VSRFs were obtained with the signal alone (right column). VSRF shown at top middle obtained with signal alone at 26 dBA. Intensity of the signal or noise shown above each VSRF. SA: spherical angle (degrees).
). In Figs. 2 and 7 the SA is shown to the left of its corresponding VSRF. SA is expressed in spherical degrees; one spherical degree is that portion of a sphere enclosed by a spherical triangle with two sides each having arcs of 90° and a third side having an arc of 1°. Figure 3 summarizes the results of this analysis. In Fig. 3A, SA is plotted as a function of noise intensity for four neurons, including one of those shown in Fig. 2. The size of the VSRF obtained with the signal alone varied among the four neurons as shown here by the range of SA between ~150 and 400°. When background noise was introduced and its intensity raised systematically, the SA fell abruptly and in a nearly linear fashion, reflecting the changes seen in the VSRF maps. We fit a straight line to the linear segment of the SA versus noise-intensity function for each of the 22 neurons in our sample for which we have such data. The correlation coefficients ranged from 0.854 to 0.995, although the slopes and intercepts of the functions varied among neurons, as can be seen in the three examples plotted in Fig. 3B. We derived from the slopes of these linear functions a measure that we call the "noise-suppression index." The noise-suppression index is defined as the change of noise intensity (in dB) required to reduce the size of the VSRF by 50%. Figure 3C illustrates the distribution of the noise-suppression index, including mean (14.7 dB) and SD (5.6 dB), for the entire sample. Thus, on average, a VSRF was reduced to one-half its size when a background noisewas raised by~15 dB above noise threshold and was completely obliterated when the noise was raised by 30 dB.
View larger version (21K):
[in a new window]
FIG. 3.
A: spherical area (SA) of the VSRF plotted as a function of intensity of the continuous background noise for 4 AI neurons. B: normalized SA plotted as a function of the intensity of continuous background noise for those noise intensities that reduced the size of the VSRF. Dashed lines are linear fits to the data. C: distribution of the noise suppression index. See text for details.
View larger version (21K):
[in a new window]
FIG. 4.
A: number of spikes evoked (filled symbol) and mean onset latency (open symbol) to signal alone at the acoustic axis (110 dBA
NOISE) and in the presence of continuous background noise at increasing intensity. B and C: same as (A) for 13 AI neurons, with spike count normalized to maximum.
View larger version (45K):
[in a new window]
FIG. 6.
Same as Fig. 5, A-D.
). Figure 5 presents results from one neuron that illustrate response features exhibited by all neurons in the series save those few exhibiting frontal VSRFs. In Fig. 5 VSRFs are illustrated along with AZ and EL functions obtained when transient signals were presented alone and when the same signals were delivered in the presence of continuous background noise. The trajectories of the AZ and EL functions are shown as dotted lines (left column, top) intersecting at the acoustic axis for this cell (CF = 15.8 kHz). The dots along the trajectories represent the signal directions from which data were obtained in both azimuth and elevation. For these data both absolute and relative values of spike count and latency are plotted in the AZ and EL functions. We do this because from our previous studies there was evidence that directional information might be preserved in the relative changes in these response parameters (Brugge et al. 1996). Thus the ordinates are total spike count (Fig. 5, A and E), spike count normalized to the maximal count (Fig. 5, B and F), mean latency (Fig. 5, C and G), and changes in mean latency relative to the minimum (Fig. 5, D and H). The dashed vertical line on each graph represents the frontal midline and the arrow the acoustic axis.
View larger version (30K):
[in a new window]
FIG. 8.
SA plotted, for 4 AI neurons, as a function of change in the intensity of either the background noise with the signal at a constant intensity (S + N closed symbol) or the intensity of the signal when presented alone in quiet (S open symbol). The abscissa was adjusted to align the curves.
18°. For directions behind the animal the count gradually declined. The EL functions (Fig. 5, E and F) show that the neuron fired maximally around the acoustic axis and maintained near-maximal firing at all other elevations studied. Under these quiet background conditions, onset latency tended to mirror the spike count (Fig. 5, C, G, D, and H, open symbols). Latency was shortest and spike count was highest for signals around azimuth +27° and elevation +18°, which is close to the direction of this acoustic axis. Latency increased and count decreased systematically for directions on either side of the acoustic axis, with steepest gradients occurring along the azimuth across the frontal midline.
View larger version (43K):
[in a new window]
FIG. 10.
AZ functions obtained under conditions of background noise (A-D) or signal alone (E-H). Inset indicates the intensity of noise and signal associated with these functions.
View larger version (38K):
[in a new window]
FIG. 11.
Elevational functions obtained under conditions of background noise (A-D) or signal alone (E-H). Inset indicates the intensity of noise and signal associated with these functions. Same neuron as illustrated in Fig. 10.
) it became evident that the presence of a steady background noise influenced VSRF properties in ways that were similar to those associated with changing signal level under listening conditions in which no background noise was present. To examine more closely this possible relationship we compared directly, in the same neuron, the VSRF properties obtained in the presence of increasingly intense background noise to those same properties recorded when no noise was present but the signal intensity was decreased.
). The series of VSRFs obtained with 5-dB steps under both signal-in-noise and signal-alone conditions were aligned by eye to obtain the best visual comparison. In Fig. 7 we see that each of the VSRFs (and their SAs) in the left column are similar to those with which they are paired on the right. We further analyzed and compared the VSRF data by plotting the SAs obtained under signal-alone conditions and those obtained under conditions of signal-in-noise. Figure 8 presents such data for four neurons, including the one whose VSRFs are illustrated in Fig. 7. For each cell, the signal-alone and signal-in-noise results are plotted on the same graph with the abscissae simply shifted appropriately to bring the curves into best alignment. For all intents and purposes the curves superimpose.
View larger version (31K):
[in a new window]
FIG. 9.
Normalized spike count (A and C) and mean 1st spike latency plotted as a function of change in intensity of either the background noise with the signal at a constant intensity (S + N closed symbol) or the intensity of the signal when presented alone in quiet (S open symbol) at the acoustic axis. The abscissa was adjusted to align the curves.
DISCUSSION
Abstract
Introduction
Methods
Results
Discussion
References
; Derbyshire and Davis 1935). These experiments showed that the amplitude of the N1 potential recorded from the round window of the anesthetized cat in response to a click or noise stimulus was reduced when that signal was mixed with a background of continuous broadband noise. Three decades later Kiang and colleagues (1965) went on to demonstrate that for a single auditory nerve fiber the presence of continuous broadband noise reduced the amplitude, but not the latency, of peaks in the poststimulus time histogram obtained with click stimuli. They suggested that the effect of increasing the level of noise "may be compared with the effect of decreasing click level." Rosenblith (1950) and Goldstein and Kiang (1958) studied the effects of noise on the click-evoked potential from auditory cortex and on the N1 potential at the round window in anesthetized cats, repeating and extending the earlier experiments of Davis and Derbyshire. They found that the response of both auditory nerve and cortex to clicks was affected "in proportion to the amount of noise masking," although constant noise alone seemed to evoke no measurable response from either site. Erulkar and colleagues (1956) clearly demonstrated, in recordings from single AI neurons in the anesthetized cat, that a tonal background increased markedly the latency of the click response and diminished the average number of spikes per click stimulus. In other words, an intense click when delivered against a certain tonal background caused a unit to discharge "as if the clicks delivered were weak." Similar observations were made by Phillips and Cynader (1985) and Phillips and Hall (1986), who studied the responses of AI neurons to contralateral CF tones in the presence of continuous wideband noise (see also Phillips 1985, 1990). As in our experiments, none of their cells responded in a sustained fashion to continuous noise. The main consequence of continuous noise background was to elevate tone threshold. This general decrease in sensitivity was reflected in a simple linear shift in the input-output function along the signal-intensity axis. The magnitude of the shift was equal to the change in noise level in decibels. Our results are in accord with these findings, namely, the magnitude of change in receptive field properties was a monotonic function of the magnitude of change in the background noise intensity. Phillips and colleagues also noted that noise-induced threshold shifts were the same for cells exhibiting monotonic or nonmonotonic rate-level functions. Imig et al. (1990) on the other hand reported that for AI neurons there was a high correlation between the azimuthal tuning and the growth of spike count with stimulus level; neurons narrowly tuned to azimuth tended to exhibit nonmonotonic spike count-versus-level functions. We did not study rate level functions over a range sufficient to address this issue directly, although it seems clear from our results that functional gradients in VSRF will remain intact over a relatively wide range of signal or noise level (see also Brugge et al. 1996).
reported that in decerebrate cats the noise masking of tone-burst responses of neurons in the ventral cochlear nucleus was qualitatively similar to that recorded in auditory nerve fibers. Hence, under our conditions of a continuous diffuse noise background, the output of the two cochleas and ventral cochlear nuclei might be expected to be equivalently masked by the noise, as the spectra and overall level of the noises were kept constant at the two ears. We might infer then that in the lower brain stem, where high-frequency binaural interactions take place, changes in background noise intensity would alter equivalently the absolute strength of left and right ear excitation and inhibition, thereby maintaining a balance among the magnitudes of these bilateral inputs. If this assertion is correct it suggests that, in the steady state, diffuse noise exerts its binaural effects at levels below the auditory cortex, which is also consistent with the observation that AI neurons exhibit no demonstrable response to the steady-state component of the background noise.
). Our results imply that, under natural free-field listening conditions in which a signal and the background noise are fluctuating in intensity, the spatial receptive fields of active AI neurons are continuously expanding and contracting. Yet, for a listener in such an environment, sound localization ability can be unexpectedly good even near detection threshold and therefore over some range of signal-to-noise ratio (see Good and Gilkey 1996; Good et al. 1997). Thus, if AI neurons are normally engaged in determining signal direction, then some information-bearing property or properties of their spatial receptive fields, other than size, must remain relatively tolerant of the intensity of the signal, of the background noise, or of both.
, 1996). Functional gradients of spike rate within a peripheral VSRF are established mainly by the spectral filtering properties of the pinna (Poon and Brugge 1993). This monaural spectral information may be retained to a greater or lesser extent within VSRFs of AI neurons (Brugge et al. 1996). Samson et al. (1993, 1994) reported that AZ tuning in some AI neurons survives plugging one ear, which suggests that these cells derive AZ sensitivity, in part at least, from monaural spectral cues. The structure of other VSRFs may be modified substantially by the spectral and temporal filtering properties of neuronal circuits that intervene between the cochlear nuclei and AI cortex. These complex relationships between pinna filtering and the structure of AI spatial receptive fields are poorly understood and remain under study. Binaural interactions are commonly observed in the VSRF of AI neurons (Brugge et al. 1994). In a substantial proportion of recorded AI neurons and at signal levels that exceed the acoustic attenuation created by the head, signals falling in one acoustic hemifield suppress neuronal firing. For these neurons the response is lateralized to the opposite hemifield, and the receptive field boundary at or near the frontal midline becomes sharply defined. Rajan et al. (1990) found that the most common response to variations in CF-tone intensity was maintenance of the shape and frontal midline border of the AZ function, although the proportions of constant AZ functions varied among cell classes. Imig et al. (1990), by using noise-burst signals, also observed that for a high proportion of cells the slope of the AZ function around the midline was relatively constant across level. For other classes of neurons as signal intensity is raised the receptive field may expand to engulf most or all of acoustic space, either because for all directions the signals to both ears are excitatory or because only one ear contributes to the response; yet these fields may retain directional information in their functional gradients. Thus, while binaural interactions no doubt play a role in directional sensitivity of AI neurons, the mechanisms by which spectral information extracted by each cochlea is transmitted to and through monaural and binaural channels to eventually reach AI cortex are poorly understood.
, 1997a,b) we emphasized that onset latency was a major direction-dependent response metric of AI neurons. Direction-dependent firing strength mirrors to a large extent that recorded in single auditory nerve fibers under the same stimulus conditions (Poon and Brugge 1993). Because onset latency and firing strength are often highly correlated, firing strength could serve as well as onset latency in encoding sound direction. On the other hand, direction-dependent onset latency is an emerging property of central auditory neurons because in discharges of auditory nerve fibers latency changes little with sound-source direction (Poon and Brugge 1993) or intensity (Kiang et al. 1965). The central auditory system devised strategies at the cellular and molecular levels that faithfully preserve time information (e.g., see Trussell 1997). We favor the proposition that discharge timing as well as rate and firing probability are an important metric in the coding of directional signals at the level of AI, especially the transient sound used in this study (see also Middlebrooks 1998; Middlebrooks et al. 1994). Although onset latency is but a scalar estimate of such timing information, Jenison (1998) has shown that the gradients of response latency measured in VSRFs for an ensemble of AI neurons are sufficient for an ideal observer to achieve the level of spatial acuity exhibited by human (Makous and Middlebrooks 1990) and cat (May and Huang 1996) listeners.
and Good et al. (1997) have the only reported human psychophysical data on noise masking of sound direction in the free field. In their experiments, signals and noise maskers were presented from the same or different loud speakers arrayed in a three-pole coordinate system within an anechoic chamber. Subjects were required to report the direction of the source in the presence or absence of background noise. It was found that localization judgments were clearly compromised by the presence of background noise and that the impact of the noise was complex, depending on the signal-to-noise ratio, on the target position, and on the position of the masker with respect to that of the subject. Of particular interest is the target angle and dynamic range. They showed (e.g., their Fig. 7, Good et al. 1997) that the major effects on average performance across subjects occurred when signal-to-noise ratio spanned a range of~20 dB. Our results depicted by the dynamic range of AA functions (e.g., Fig. 9) are in very good agreement with this observation. They also reported that, although the ability to determine the direction of a signal in noise decreased nearly monotonically with lowered signal-to-noise ratio in all three dimensions, at any signal-to-noise level there were fewer errors made for signals on the frontal horizontal plane (left-to-right) than for signals distributed horizontally front-to-back or vertically (up-down). Furthermore, relatively good localization performance was observed in the frontal (left-to-right) dimension compared with the other dimensions even at low signal-to-noise ratio. This observation is in accord with our finding of steep functional gradients being maintained across the frontal midline of AI spatial receptive fields.
). Dereverberation is lost when the sound streams arriving at the left and right ear become perfectly correlated. In this study signal direction was detected by AI neurons in the presence of steady, nondirectional, uncorrelated background noise. Under these conditions AI neurons were found to be relatively tolerant of wide changes in the signal-to-noise ratio, as would be required if these neurons are to serve directional sensitivity under certain natural conditions. The extent to which such directional information might be maintained under other competing sound conditions is presently under study.
| |
ACKNOWLEDGEMENTS |
|---|
This work was supported by National Institutes of Health Grants DC-00116 and HD-03352.
| |
FOOTNOTES |
|---|
Address for reprint requests: J. Brugge, 627 Waisman Center, University of Wisconsin, Madison, WI 53705.
Received 30 December 1997; accepted in final form 7 July 1998.
| |
REFERENCES |
|---|
|
Abstract
Introduction
Methods
Results
Discussion
References
|
|---|
This article has been cited by other articles:
|
|
 |
|
  L. Las, A.-H. Shapira, and I. Nelken Functional Gradients of Auditory Sensitivity along the Anterior Ectosylvian Sulcus of the Cat J. Neurosci., April 2, 2008; 28(14): 3657 - 3667. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
U. Werner-Reiss and J. M. Groh A Rate Code for Sound Azimuth in Monkey Auditory Cortex: Implications for Human Neuroimaging Studies J. Neurosci., April 2, 2008; 28(14): 3747 - 3758. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. F. Linden, R. C. Liu, M. Sahani, C. E. Schreiner, and M. M. Merzenich Spectrotemporal Structure of Receptive Fields in Areas AI and AAF of Mouse Auditory Cortex J Neurophysiol, October 1, 2003; 90(4): 2660 - 2675. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. A. Reale, R. L. Jenison, and J. F. Brugge Directional Sensitivity of Neurons in the Primary Auditory (AI) Cortex: Effects of Sound-Source Intensity Level J Neurophysiol, February 1, 2003; 89(2): 1024 - 1038. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
O. Bar-Yosef, Y. Rotman, and I. Nelken Responses of Neurons in Cat Primary Auditory Cortex to Bird Chirps: Effects of Temporal and Spectral Context J. Neurosci., October 1, 2002; 22(19): 8619 - 8632. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Furukawa and J. C. Middlebrooks Sensitivity of Auditory Cortical Neurons to Locations of Signals and Competing Noise Sources J Neurophysiol, July 1, 2001; 86(1): 226 - 240. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. A. Reale and J. F. Brugge Directional Sensitivity of Neurons in the Primary Auditory (AI) Cortex of the Cat to Successive Sounds Ordered in Time and Space J Neurophysiol, July 1, 2000; 84(1): 435 - 450. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
