| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1Departments of Anatomy and Neurobiology and 3Cognitive Science, University of California, Irvine, California; 2Department of Biology, School of Life Science, East China Normal University, Shanghai, People's Republic of China
Submitted 23 November 2004; accepted in final form 25 May 2005
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
50 ms before each of the stimuli comprising the level response area. An effective preceding stimulus alters the shape and severely reduces the size and response magnitude of the level response area. This ability of the preceding stimulus depends on its proximity in the level domain to the level response area, not on its absolute level or on the size of the response it evokes. Preceding stimuli evoke a nonlinear inhibition across the level response area that results in an increased selectivity of a cortical neuron for its preferred binaural stimuli. The selectivity of AI neurons during the processing of a stream of acoustic stimuli is likely to be restricted to a portion of their level response areas apparent in the tone-alone condition. Thus rather than being static, level response areas are fluid; they can vary greatly in extent, shape and response magnitude. The dynamic modulation of the level response area and level selectivity of AI neurons might be related to several tasks confronting the central auditory system. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Brosch and Schreiner 1997; Calford and Semple 1995; Costalupes 1985; Costalupes et al. 1984; Harris and Dallos 1979; Kaltenbach et al. 1993; Phillips 1990; Phillips and Cynader 1985; Phillips and Hall 1992, 1986; Rhode et al. 1978; Shore 1995; Smith 1977, 1979). Thus responses to many of the originally effective stimuli are reduced or entirely suppressed.
Maps of monaural suppressive stimuli have been determined in the frequency-level domain (Brosch and Schreiner 1997; Calford and Semple 1995; Fuzessery and Hall 1996; Harris and Dallos 1979; Sutter and Loftus 2003). These maps typically are determined by identifying the stimuli that reduce the response of a neuron to a test stimulus within its frequency-threshold tuning curve. They vary considerably across neurons, and are commonly used as estimates of inhibitory side bands. As there is nothing particularly unique about the test stimuli, these studies imply that other stimuli within the frequency-threshold tuning curve are also affected, perhaps to a greater or a lesser extent, by the inhibitory stimuli. In turn, by possibly reducing a neuron's response to many stimuli within its frequency-threshold tuning curve, this suggestion implies that a preceding or an on-going stimulus modifies the tuning of a neuron to acoustic stimuli. To assess the validity of these inferences, the present study examined how a preceding stimulus modifies a neuron's response area and its selectivity for preferred stimulus parameters.
Although almost all previous studies regarding the influence of a preceding or ongoing stimulus used monaural stimulus parameters, the present study focused on binaural stimuli because the response of neurons in the primary auditory cortex (AI) are almost always (>90%) influenced by stimulation of both ears (Kelly and Judge 1994; Semple and Kitzes 1993a; Zhang et al. 2004). Responses of cortical neurons to binaural stimuli (Reale and Brugge 2000), as well as neurons throughout most of the auditory neuroaxis (Faure et al. 2003; Finlayson 1999; Finlayson and Adam 1997; Fitzpatrick et al. 1999; Litovsky and Yin 1998b; Yin 1994), can be altered by a preceding sound in a frequency-, level-, and time-dependent manner. The focus of the present study, therefore, is the fluidity of the binaural level response area of cortical neurons.
AI neurons respond most strongly to a small range of level combinations at the two ears and less strongly to a much larger range of binaural level combinations. The magnitude of responses to stimuli that vary in level at the two ears constitute a neuron's level response area. Such level response areas have been demonstrated in both free field (Clarey et al. 1994; Imig et al. 1990) and in dichotic studies (Irvine et al. 1996; Nakamoto et al. 2004; Semple and Kitzes 1993a,b; Zhang et al. 2004). Previously (Nakamoto et al. 2004; Zhang et al. 2004), we defined the set of binaural level combinations that evoke 80% of the maximal response as a measure of a neuron's preferred binaural combinations (PBC).
The present study assessed the effect of the binaural levels of a preceding, conditioning stimulus on the tuning of AI neurons to binaural stimuli in the level domain, i.e., stimuli that vary inlevel at each ear. We demonstrate that a conditioning stimulus can reduce the size of the level response area and the magnitude of responses to stimuli remaining within the contracted area. Through a nonlinear reduction of responses across the level response area, conditioning stimuli increase the selectivity of AI neurons for their preferred binaural-level stimuli. Thus rather than being fixed and immutable, the configuration of the level response area of AI neurons is mutable and fluid.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Sixteen healthy adult cats, free from external and middle ear infection, were used in this study. The institutional animal care and use committee (IACUC) of the University of California Irvine approved the procedures of the experiment. The cats were anesthetized with pentobarbital sodium (40 mg/kg ip). Anesthesia and body fluid level were maintained throughout the surgery and the physiological recording by intravenous drip of 5% dextrose and 0.1% pentobarbital sodium in lactated Ringer solution. The level of anesthesia was periodically checked by ensuring the lack of a withdrawal reflex to the pinch of a front paw. When necessary, additional pentobarbital sodium was given through a syringe attached to a three-way valve in the drip system. Close attention was paid to the interval between the supplementary dose of anesthesia and to the threshold of auditory responses evoked at the neuron's characteristic frequency (CF, i.e., the frequency that evokes a response at the lowest level). A steady increase in threshold was considered to be an indication of the progressive decline of the preparation and resulted in termination of the experiment. Atropine sulfate (0.048 mg) and dexamethasone sodium phosphate (0.044 mg) were injected (subcutaneously) at the beginning of surgery and at 24 h intervals to reduce bronchial secretions and cerebral edema, respectively. Body temperature was monitored by a rectal probe and maintained at 38°C by a feedback–controlled heating blanket. A tracheotomy was performed to allow unobstructed breathing and reduce respiration noise. The head was immobilized by a metal rod that was attached to the exposed dorsal surface of the skull with screws and dental cement and then secured to a stereotaxic-like recording frame. The pinnae were resected and earpieces inserted and acoustically sealed within the transected external auditory canals. A small hole was made through the skull, and a small incision was made in the dura to expose part of the primary auditory cortex. Warm saline or mineral oil was applied to the cortex during the experiment to prevent drying.
Acoustic stimuli
Pure tone stimuli were 50 ms in duration with a 6-ms rise-decay time, presented at 800-ms inter-trial intervals at the neuron's CF, and repeated 30 times at each tested stimulus level. Tone pips were generated digitally by a MALab system (Kaiser Instruments, Irvine, CA) controlled by a Macintosh computer. Stimuli were transduced by STAX earphones housed in metal canisters that were coupled directly to the earpieces. Sound pressure level (SPL) near the tympanic membrane was measured (re. 20 µPa) at each ear from 100 Hz to 30 kHz under computer control with a calibrated probe tube and a 0.5-in condenser microphone (Brüel and Kjær, Cleveland, OH). The calibrations were stored in the computer for use in controlling attenuators to obtain desired SPLs.
Recording system
Responses of single neurons were recorded with parylene-insulated tungsten microelectrodes (Microprobe, Potomoc, MD) with a 5-µm tip. Electrodes were oriented orthogonal to the pial surface. The electrode signal was amplified (1,000 times), filtered (0.3–3.0 kHz), and then sent to a digital oscilloscope for display. The MALab system was used to generate stimuli and to analyze data, both on- and off-line. Recording was conducted in a double-wall, sound-insulated chamber. Electrode advancement was controlled from outside of the sound-insulated chamber.
Data collection
Single neurons were recorded in the third or fourth layer (estimated from the depth of penetration indicated by the microdrive controller) of the primary auditory cortex of the cat. Once a single neuron was isolated, its CF and threshold at CF were determined. Only high-frequency (CF 5 kHz) neurons were selected. After determining the response of the neuron to monaural CF tones delivered at increasing levels (the monaural rate-level function), a large binaural stimulus matrix was presented at CF. Binaural level within the matrix varied in interaural level difference (ILD: contralateral SPL minus ipsilateral SPL in dB) from –20 to +20 in 10-dB steps and in average binaural level (ABL: the sum of the stimulus levels at the 2 ears divided by 2) (Goldberg and Brown 1969) from 0 to 70 in steps of 10. Positive ILDs favor the contralateral ear and negative ILDs favor the ipsilateral ear. Responses to 30 repetitions of each stimulus within the binaural matrix were recorded and plotted as response contours in 20% divisions. The binaural level combinations in the stimulus matrix that excited the neuron comprise the neuron's level response area (LRA). Binaural stimulus combination that evoked the top 20% of the responses were defined as the neuron's PBCs. After determining the LRA and the PBC of the neuron, a forward masking paradigm was presented. The frequency, duration and rise-decay time were the same for the conditioning tone (tone 1) and the probe tone (tone 2). The interval between the two tones was 50 ms. This time interval was selected because it severely reduces or eliminates cochlear effects (Harris and Dallos 1979) and, thereby, emphasizes the central auditory system contribution to sequential interactions. The influence of the interval between the two tones will be presented in a subsequent communication (manuscript in preparation).
Statistical analyses of response functions employed SPSS software or routines coded in MatLab.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
6 kHz. Due to the length of time required to collect a set of data from each neuron, control LRAs and monaural rate level functions were obtained periodically during and at the end of the recording period to assess response stability. Neurons were considered stable if the configuration of their control LRAs was relatively constant and their rate-level functions varied by <20%. Only stable neurons were included in the data analysis. We first present data demonstrating how the binaural levels of a preceding, conditioning stimulus influence the response of AI neurons to a stimulus in their PBC. We then present data demonstrating how a preceding binaural stimulus affects the extent, shape and magnitude of the LRA of AI neurons. An analysis is then presented that demonstrates the lack of an association between the magnitude of the responses to the preceding and probe stimuli at the level of individual trials. Next we present data demonstrating that a preceding binaural stimulus reduces the LRA of AI neurons nonlinearly. Finally, we present population data indicating that the binaural level selectivity of the great majority of AI neurons depends on the parameters of a preceding stimulus.
Sequential inhibition of responses to a preferred binaural stimulus level combination
To determine how the binaural levels of a preceding stimulus affect the response of AI neurons to their preferred binaural-level stimuli, the stimuli in the binaural matrix served as tone 1, and a binaural stimulus within the PBC of the neuron served as tone 2. Thus tone 2 was fixed, and the binaural parameters of tone 1 were stepped through the stimulus matrix. Both tones 1 and 2 were delivered at the neuron's CF. The primary finding is that, regardless of their absolute levels, stimuli within the PBC of a neuron exert stronger sequential inhibition than do non-PBC stimuli.
Data from three representative neurons are shown in Fig. 1. A–C illustrate the control level response areas of the three neurons to the stimulus matrix (tone 1). D–F illustrate responses to the indicated stimulus (white circle in A–C) within each neuron's PBC as tone 1 was stepped through the matrix. Three facts are evident in these panels. First, tone 1 PBC stimuli (Fig. 1, A–C, red areas) are associated with the strongest suppression of responses to tone 2 (Fig. 1, D–F, gray or white areas). Non-PBC tone 1 stimuli are associated with weaker or no suppression of responses to tone 2 (Fig. 1, A vs. D, B vs. E, and C vs. F). Second, responses to PBC stimuli are still robust despite presentation of preceding non-PBC stimuli (Fig. 1, D–F). Third, a high-level tone 1 does not necessarily exert more sequential inhibition of the response to a PBC stimulus than does a low-level tone 1. This is particularly evident in the data from the monotonic neuron shown in the middle column (Fig. 1, B and E) and in the data from the nonmonotonic neuron shown in the right column (Fig. 1, C and F). Such nonmonotonic suppression is reminiscent of the monaural, nonmonotonic masking functions described previously in AI (Calford and Semple 1995). These data suggest that the strength of sequential inhibition depends more on the binaural level preference of a neuron than on the absolute binaural level of the preceding stimulus.
|
These positive correlations raise the question whether sequential inhibition is linked to the parameters of the preceding stimulus or to the response it evokes. The most accurate way to distinguish between these two alternatives is at the level of the individual trial, i.e., single presentation of tones 1 and 2. The statistical question is whether the size of the response to tone 2 is contingent on or associated with the size of the response to tone 1. A Phi analysis, sometimes called Pearson's coefficient of mean-square contingency, was performed on the set of 30 trials of each pair of tone 1 and tone 2 stimuli in which both kinds of stimuli evoked more than three spikes within the set of 30 trials. The x and y coordinates of the contingency table were the number of spikes evoked by tone 1 and 2 stimuli, respectively. The variables of the Phi analysis were scaler, e.g., 0, 1, 2, or 3 evoked spikes. The bottom row in Fig. 1 shows, for the respective neuron in each column, the probabilities of the obtained Phi values. In each case, the probabilities were distributed broadly. Using a criterion of P < 0.05 for statistical significance, two of 32 of the tone 1 –tone 2 stimuli pairings illustrated in D and E were statistically significant (G and H) and only 2 of 29 pairings illustrated in F were statistically significant (I). The probabilities of the Phi values (within trial contingency analysis) obtained from the sample of 20 neurons that had significant correlations between the magnitude of responses to tone 1 and the suppression of responses to tone 2 (across-trials analysis) are pooled together in Fig. 4A. The distribution of probabilities is flat, with only 23 of 408 sets having a probability < 0.05. The nonsignificant results of the contingency analysis suggest that, at the level of the individual trial, responses to tone 1 and to tone 2 do not directly affect each other. Thus at the level of the individual trial, sequential inhibition of AI neurons is not dependent on their response to the preceding stimulus.
|
; Fitzpatrick et al. 1995; Litovsky and Yin 1998b) and AI (Calford and Semple 1995; Reale and Brugge 2000). In the present study, the across-trials analysis indicates that responses to tones 1 and 2 are negatively correlated. The within-trial contingency analysis (Phi) suggest that responses to tones 1 and 2 are not causally related to each other. These two observations suggest that the excitatory drive evoked by the preceding stimulus determines both a response and the strength of the sequential inhibition. Thus sequential inhibition is related to the preceding stimulus, not to the strength of the response it evokes. Sequential interactions shape the level response area of AI neurons
Thus far we have demonstrated that the strength of sequential inhibition varies as a function of the parameters of the preceding binaural stimulus relative to the PBC. Now we present evidence that the extent, shape, and magnitude of the entire LRA of a neuron depend on the level parameters of a preceding binaural stimulus. In this analysis, a single stimulus within the binaural matrix served as tone 1; the stimuli comprising the binaural matrix served as tone 2. Thus tone 1 was fixed, and tone 2 was stepped through the binaural parameters of the stimulus matrix.
Data from three representative neurons are shown in Figs. 2 and 3. The control LRA, shown in Fig. 2A, is located in the lower half of the stimulus matrix. This neuron responded nonmonotonically to both contralateral and ipsilateral level. The LRA of the neuron was shaped (Fig. 2, B–G) by tone 1 presented at 0–40 ABL at 0 ILD (Fig. 2, 
). The change is evident in both the shape of the LRA and the number of evoked spikes. However, the binaural level preference of the neuron was maintained despite the various modifications of its LRA. Overall, the sequential inhibition exerted by tone 1 presented at 10 dB ABL, within the PBC, was stronger than the sequential inhibition exerted by tone 1 presented at higher or lower ABLs. The sequential inhibition diminished systematically as tone 1 departed from the PBC. This nonmonotonic effect demonstrates most clearly that sequential inhibition is PBC dependent rather than level dependent.
|
|
The LRAs of the two neurons separated and were modified in different ways when the same tone 1 preceded each of the stimuli in the binaural matrix (Fig. 3, B–D). Tone 1 delivered at 50 dB contralateral and 70 dB ipsilateral compressed the LRA of neuron a more than it did the LRA of neuron b (Fig. 3B), whereas the converse was true for tone 1 delivered at 40 dB contralateral and 60 dB ipsilateral (Fig. 3C). Tone 1 delivered binaurally at 30 dB had the strongest inhibitory effect on neuron b and the weakest inhibitory effect on neuron a (Fig. 3D). Consistent with the nonmonotonic suppression presented in Figs. 1 and 2, a lower level binaural stimulus had stronger sequential inhibitory effects on the LRA of the nonmonotonic neuron than did the two higher binaural level stimuli (Fig. 3, D vs. B and C). The extent and magnitude, in terms of number of evoked spikes, of the LRAs varied as a function of the proximity of the preceding stimulus to the PBC of each neuron rather than its absolute level (Fig. 3, B–D). Low levels of tone 1 can result in the suppression of responses to appreciably stronger tone 2 stimuli (Fig. 3, D–F). Finally, despite changes in their shape and magnitude, the LRAs remain organized even if the response area outside the PBC is greatly or completely suppressed by the preceding stimulus, i.e., maximal responses to tone 2 are evoked by stimuli within the control PBC and decrease as the level combinations depart from the control PBC.
Determinant factor of sequential inhibition
The data shown in Figs. 2 and 3 again raise the question whether the observed suppression is likely the result of an inhibitory process or from adaptation or fatigue of the neuron due to its response to tone 1. This question was raised earlier with regard to the significant across-trials correlation between the magnitude of the response to tone 1 and the suppression of the response to tone 2 in the stimulus paradigm used to generate the data shown in Fig. 1. The within-trials Phi analysis presented in Fig. 4B strongly suggests that, in the paradigm used to generate the data shown in Figs. 2 and 3, the suppression of responses to tone 2 was not causally associated with the magnitude of the response to tone 1. The distribution of the probabilities of Phi values is rather flat, with P < 0.05 for 28 of 597 sets of 30 tone 1–tone 2 trials obtained from 27 neurons subjected to this paradigm.
The significant correlation between the suppression of responses to tone 2 and the response to tone 1 that is based on the total numbers of spikes evoked in each set of 30 trials strongly suggests that these two parameters are determined by the excitatory drive exerted by tone 1. The lack of a significant contingent relation between these two variables based on the individual trial demonstrates that the response to tone 2 does not result from a factor such as fatigue or adaptation of the neuron caused by its response to tone 1. It is possible, of course, that the observed behavior resulted from the fatigue or adaptation of a neuron in the afferent pathway of the neuron being studied.
Nonlinear sequential suppression of responses across the LRA
The data presented in Figs. 2 and 3 also raise the question whether sequential inhibition reduces responses uniformly across the LRAs of AI neurons. To answer this question, 25 neurons were analyzed using the following paradigm: tone 1 was a stimulus within the binaural matrix; tone 2 consisted of a sequence of stimuli in the ILD or ABL dimensions that crossed through the PBC. The ILD of tone 2 was varied within the range of –30 to +30 dB in a 10-dB step and the ABL of tone 2 was varied from 0 to 70 in steps of 10. This analysis of the monotonic neuron the LRA of which is shown in Fig. 3 (neuron a) is presented in Fig. 5. Whether examined across ILD (Fig. 5, A–C) or across ABL (Fig. 5, D–F), the preceding stimuli reduced responses to non-PBC stimuli to a much great extent than responses to PBC stimuli. Responses to PBC stimuli were still robust or even at control values, whereas responses to non-PBC stimuli were greatly diminished or entirely absent.
|
|
Data obtained from 23 neurons were used in this analysis. The slope values and whether the slope differed significantly from 0 were determined individually for each function. The results of this analysis are illustrated in Fig. 7, A and B, for the ILD and ABL dimensions, respectively. The slope of the difference functions differed significantly from 0 in the great majority of cases, 83% (34 of 41) in ILD and 93% (40 of 43) in ABL. Among the 23 neurons, 20 had significant slopes in the ABL and ILD dimensions, 2 had significant slopes in either the ABL or ILD dimensions, and 1 neuron had significant slopes in neither dimensions. The significant positive slope in our group data demonstrates that the LRA is reduced in the masked condition in a systematic but nonuniform manner, with responses to the PBC stimuli being reduced least. This nonlinear suppression of the LRA contributes to the increased selectivity of AI neurons in the masked condition.
|
The nonuniform reduction of responses across an LRA results in an increased selectivity for preferred binaural stimulus level. In the following section, we show the selectivity changes across a population of neurons. For each neuron, selectivity for binaural stimulus level was quantified by determining its preference for ILD and ABL. The level response area data were averaged at each ILD across ABL and at each ABL across ILD, creating single averaged ILD and ABL functions from the responses to each stimulus matrix. Then each of these averaged response functions was normalized to its maximal value. Data obtained from a monotonic unit are shown in the form of such normalized ILD (Fig. 8 A) and ABL response functions (Fig. 8B). It is apparent in these data that the highest 0.2 of the normalized functions occupy a smaller portion of the ILD and ABL dimensions in the sequential-tone condition than in the tone-alone condition, indicating an increased selectivity for both ILD and ABL. The increasing difference between the normalized and control functions at more negative ILDs (Fig. 8A) and at lower ABLs (Fig. 8B) reflect the nonlinear reduction of the LRA discussed earlier and demonstrate that the increase in stimulus selectivity is also nonlinear.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
The picture emerging from the present data, as well as from other studies (Imig et al. 1990; Irvine et al. 1996; Nakamoto et al. 2004; Semple and Kitzes 1993a,b; Zhang et al. 2004), is that neurons in the auditory cortex commonly respond to a broad range of stimuli that vary in ABL and ILD. These level response areas vary in shape, extent, and the number of spikes evoked by the stimuli comprising it. The centers of the PBCs cover the entire binaural level range, with those of monotonic neurons occurring most commonly in the middle to high level range and those of nonmonotonic neurons occurring in the low to middle level range (Zhang et al. 2004). We have suggested that this differential selectivity of cortical neurons enables AI to deal with the entire range of binaural stimulus level (Nakamoto et al. 2004; Zhang et al. 2004).
The occurrence of an effective conditioning stimulus causes the LRAs of AI neurons with CFs that equal the stimulus to contract in a nonlinear manner. This contraction increases the selectivity of cortical neurons for particular binaural level combinations. The degree of contraction depends on the proximity of the preceding stimulus to the PBC, not on its absolute level. A salient feature of this contraction is that the occurrence of a conditioning stimulus can reduce or entirely suppress the ability of a cortical neuron to respond to a large set of other stimuli within its control level response area. Indeed, the occurrence of a conditioning stimulus that is either within or near a neuron's PBC can prevent the neuron from responding to the great majority of stimuli within its control LRA, many of which differ significantly in ILD and/or ABL from the conditioning stimulus.
The occurrence of an effective conditioning stimulus does not lead to a uniform contraction of a level response area or a uniform reduction of responses across the remaining level response area. Compared with responses to non-PBC stimuli, responses to PBC stimuli have a reduced sensitivity to the sequential inhibition exerted by a conditioning stimulus. Although we observed this nonlinear sequential inhibition in AI, the origin and mechanism of this nonlinearity is presently unknown. It could arise at subcortical levels or be an emergent property of cortical processing. Further study is necessary to determine the level or levels of the auditory pathway where this nonlinear property is generated.
Which binaural combinations the neuron remains responsive to depend on the LRA of the neuron in quiet and the location in the level domain of the conditioning stimulus relative to the LRA. However, it invariably includes a portion of the original PBC. Moreover, as demonstrated by the data in Fig. 6, an LRA that is broadly tuned to either ILD or ABL can be transformed into a highly selective LRA by the occurrence of a preceding stimulus.
Dynamic modulation of the LRA of AI neurons by preceding stimuli
The presence of a previous sound reduces the size of receptive fields in virtual acoustic space of AI neurons (Reale and Brugge 2000). The spatial gradients, based on firing probability and/or response latency, are maintained in the contracted spatial receptive field. The authors suggested that the maintenance of such gradients might provide a mechanism by which directional acuity remains intact in an acoustical environment containing competing acoustical transients. Although our methods differed greatly from those used by Reale and Brugge, our data are consistent with this suggestion in that the preferred level acuity is maintained despite prior stimulation. The contracted LRA is highly organized, with the stimuli remaining within the control PBC evoking the largest responses. Indeed, our data suggest that the selectivity for preferred binaural stimulus levels is enhanced for those neurons affected by the prior stimulation.
Although we have not as yet explored the impact of stimulus frequency, it is anticipated that the magnitude of the sequential interaction is greatest when the frequency of the conditioning and probe tones are the same. Previous research indicated that the strength of sequential inhibition depends on the inter-stimulus interval (Brosch and Schreiner 1997; Calford and Semple 1995; Reale and Brugge 2000). This strongly implies that the dimensions of the level response areas of cortical neurons vary dynamically. Thus rather than being a static, "hard-wired" property of a cortical neuron, the level response areas are quite fluid, varying greatly in shape, extent. and magnitude in a noisy environment. The extent of this variation depends on the binaural level preference of the neuron in quiet, the binaural levels of the preceding stimulus, very likely the difference in frequency between conditioning and probe stimuli and the time interval between stimuli (manuscript in preparation).
Aspects of our observations may have commonalties with recent descriptions of the receptive fields of cortical neurons in other sensory systems. For example, responses of neurons in the visual and somatosensory systems are context dependent (Brumberg et al. 1996; Das and Gilbert 1999; Gilbert 1996; Gilbert and Wiesel 1990; Greek et al. 2003; Pettet and Gilbert 1992). In addition, compelling data indicate that the receptive fields of visual cortex cells change over time after presentation of a stimulus, suggesting that such receptive fields should be considered to have two spatial dimensions plus a temporal domain (Felsen et al. 2002; Levitt and Lund 1997; Worgotter and Eysel 2000). This suggests that receptive fields in the visual cortex are also dynamic rather than static structures.
Circuitry determining the sequential inhibition
On a trial-by-trial basis, there is no association between the size of the response to tone 1 and the size of the response to tone 2. This lack of association is inconsistent with fatigue or some sort of adaptation of the neuron under study as the cause of the reduced response to tone 2. It also is inconsistent with a feedback inhibitory circuit reducing the response to tone 2. Yet over the entire set of trials, the total number of spikes evoked by tone 1 is negatively correlated with the total number of spikes evoked by tone 2. These two observations, taken together, strongly suggest that the suppression of the response to tone 2 is linked to the excitatory drive exerted by tone 1. As suggested by Calford and Semple (1995), the linkage is probably through a feed-forward inhibitory circuit. Two facts strongly suggest that the activity driving the feed-forward inhibition must first traverse the same circuitry that establishes the location and configuration of the level response area: first, the overall suppression of responses to tone 2 is correlated with the overall size of responses to tone 1, and second, the ability of a preceding stimulus to exert an inhibitory influence depends on its proximity to the PBC, not on its absolute level. Thus the feed-forward inhibitory circuit appears to be driven by the output of the same circuitry that determines the location and shape of the level response area. Whether the feed-forward inhibition is generated in the cortex or at a sub-cortical locus remains to be determined.
Broad and narrow tuning to binaural level by AI neurons
The configuration of a cortical cell's level response area determines which stimuli it is able to respond to. The fluidity of level response areas is therefore likely to be fundamental to the processing of acoustic stimuli by cortical cells. Possible benefits of fluid level response areas over fixed, hard-wired, static level response areas are discussed in the following text in terms of stimulus detection and discrimination, localization of sound sources, and the processing of speech. Rather than being a complete analysis, the following discussion is intended to illustrate the varied tasks facing the auditory system that are likely to be impacted by the fluidity of level response areas.
Viewing the LRAs of neurons in AI as tuning functions in the level domain, the relevant question is whether the LRAs contracted by the occurrence of appropriate preceding stimuli are more effective or less effective in detecting a stimulus and in discriminating among subsequent stimuli. From the perspective of information theory, whether a broadly tuned neuron is more effective or less effective than a narrowly tuned neuron depends on several factors, e.g., how many dimensions are to be discriminated, whether noise levels remain proportional to the bandwidth of the filters (Eurich and Wilke 2000; Pouget et al. 1999; Wilke and Eurich 2002; Zhang and Sejnowski 1999). It is generally agreed that narrow tuning is more effective in coding a single dimension; however, severely narrow tuning that prevents adequate overlap of receptive fields results in degraded coding (Eurich and Wilke 2000). A recent analysis of Fisher information concluded that optimal coding strategies may not be separable on the basis of a simple dichotomy between narrow and broad tuning (Wilke and Eurich 2002). These authors estimated that a population of neurons with variable tuning widths would be superior to a population of neurons with identical tuning curves. Thus whether broad or narrow tuning is more effective in discriminating stimulus dimensions remains a topic of intense study.
Responses of AI cells to virtual stimuli simulating a large array of locations varying in azimuth and elevation have been analyzed in terms of Fisher information by Jenison and his colleagues (Jenison 2000; Jenison and Reale 2003). These analyses have demonstrated that correlated activity as well as incorporating responses to multiple acoustic parameters should improve the localization of a sound source.
The binaural stimuli used in this study varied in ILD and ABL. To the extent that these stimuli are representative of acoustic space, the present data suggest that in a quiet environment AI neurons are able to detect the occurrence of a sound across a rather broad expanse of the acoustic environment. The occurrence of the stimulus results in the contraction of the LRAs that are near or include the stimulus. This contraction excludes stimuli present in the original filter and focuses the selectivity of AI neurons on a smaller expanse of acoustic space. Which part of acoustic space remains capable of evoking a response from a neuron depends on the position of its original LRA in the binaural level domain and the ILD and ABL of the preceding stimulus. Such behavior is most evident in the data illustrated in Fig. 3. Given the caveats from information theory mentioned earlier, it seems reasonable to expect that the enhanced selectivity of the contracted LRAs would provide greater acuity in discriminating among stimuli arising in the reduced portion of acoustic space still capable of exciting the neuron.
The sequential stimuli comprising conversational speech are more likely to be coded by contracted LRAs than by the broad LRAs that occur in a quiet environment. This suggestion arises from the fact that speech by a given speaker has a limited variation in level and frequency. Most typically the speech will occur during a conversation at a stable ILD. Given the relatively continuous nature of speech, the shape, size, and response amplitude of the LRAs of the stimulated neurons would likely fluctuate continuously but usually remain smaller than they would be in a quiet environment. Indeed, any LRA that is repeatedly stimulated at short intervals would remain small, potentially consisting of only its PBC or a portion of its PBC. As discussed in the preceding text regarding localization, it is very possible that the enhanced selectivity that characterizes contracted level response areas improves the ability of cortical cells to discriminate among the often subtle differences between speech sounds.
Schema of level coding by AI neurons
Previous physiological studies of sequential and simultaneous interactions were often based, explicitly or implicitly, on the perceptual phenomena of masking, i.e., the increased threshold of detecting the occurrence of a test stimulus as a consequence of a preceding or ongoing masker. Those studies emphasized the reduction of responses to a test stimulus as a consequence of a masking stimulus (e.g., Calford and Semple 1995; Faure et al. 2003; Phillips 1985, 1990; Phillips and Cynader 1985; Phillips and Hall 1986; Sutter and Loftus 2003). Another set of physiological studies focused on the precedence effect, i.e., a phenomenon that is based on the interactions among binaural stimuli occurring within a 10-ms time frame (Litovsky and Delgutte 2002; Litovsky and Yin 1998a,b; Mickey and Middlebrooks 2001; Yin 1994). The paradigms used in the present study were designed to assess the interactions between stimuli occurring in sequences of sounds, such as speech. The time interval between successive stimuli is many tens of milliseconds longer than the interval pertinent to the precedence effect.
The data in the present study suggest a schema of level coding by AI neurons (Fig. 9). In a quiet environment, cortical neurons are broadly tuned to binaural stimulus level (Fig. 9, A and E), enabling a large population of neurons to detect the occurrence of a stimulus. During a limited time window after a stimulus, the sequential inhibition it evokes causes the contraction of LRAs in the ABL and ILD range of the stimulus. The LRAs with PBCs that are closest to or include the binaural levels of the stimulus will undergo the greatest contraction (Fig. 9, B–D). Thus the particular binaural level properties of the stimulus determine which subpopulation of AI neurons is affected and the extent of the LRA modulation of the affected neurons. This contraction focuses the selectivity of the affected neurons on a subset of level combinations in their original LRA. The contraction occurs nonlinearly, such that responses to PBC or a subset of the PBC stimuli are least susceptible to suppression (Fig. 9, B–D and F). That is, the nonlinear contraction of the LRA leaves the most affected neurons able to respond mainly or only to stimuli in their original PBCs. Our previously published paper (Nakamoto et al. 2004) described a topographic organization of binaurally monotonic and nonmonotonic neurons but only a very rough organization according to ILD. Consequently, Fig. 9 does not imply a spatial organization in auditory cortex of the neurons being affected by the conditioning stimulus.
|
Moore et al. (1999) arrived at a very similar conclusion with regard to the processing of stimuli by the barrels in somatosensory cortex. Citing studies in both the ventrobasal complex in the thalamus and the barrel fields in somatosensory cortex, they suggested that whisking of vibrissae results in a trade-off between sensitivity and specificity. Contact with an object when the vibrissae are not whisking leads to a broad divergence of excitation, whereas contact during active whisking differentially inhibits smaller inputs from adjacent vibrissae and, thereby, increases the relative amount of activity evoked by the primary vibrissa of each stimulated barrel. Thus barrel cortex and AI appear to have in common a functional distinction between receptive fields in nonstimulated and stimulated conditions. In the absence of stimulation, their receptive fields are broad and therefore well suited to detect a stimulus. With stimulation, their receptive fields contract, making them differentially selective for the stimuli that are most effective in driving the cortical neurons.
|
GRANTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: L. M. Kitzes, Dept. of Anatomy and Neurobiology, University of California Irvine, Irvine, CA 92697-1275 (E-mail: lmkitzes{at}uci.edu)
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Brosch M and Schreiner CE. Time course of forward masking tuning curves in cat primary auditory cortex. J Neurophysiol 77: 923–943, 1997.
Brumberg JC, Pinto DJ, and Simons DJ. Spatial gradients and inhibitory summation in the rat whisker barrel system. J Neurophysiol 76: 130–140, 1996.
Calford MB and Semple MN. Monaural inhibition in cat auditory cortex. J Neurophysiol 73: 1876–1891, 1995.
Clarey JC, Barone P, and Imig TJ. Functional organization of sound direction and sound pressure level in primary auditory cortex of the cat. J Neurophysiol 72: 2383–2405, 1994.
Costalupes JA. Representation of tones in noise in the responses of auditory nerve fibers in cats. I. Comparison with detection thresholds. J Neurosci 5: 3261–3269, 1985.[Abstract]
Costalupes JA, Young ED, and Gibson DJ. Effects of continuous noise backgrounds on rate response of auditory nerve fibers in cat. J Neurophysiol 51: 1326–1344, 1984.
Das A and Gilbert CD. Topography of contextual modulations mediated by short-range interactions in primary visual cortex. Nature 399: 655–661, 1999.[CrossRef][Medline]
Eurich CW and Wilke SD. Multidimensional encoding strategy of spiking neurons. Neural Comput 12: 1519–1529, 2000.
Faure PA, Fremouw T, Casseday JH, and Covey E. Temporal masking reveals properties of sound-evoked inhibition in duration-tuned neurons of the inferior colliculus. J Neurosci 23: 3052–3065, 2003.
Felsen G, Shen YS, Yao H, Spor G, Li C, and Dan Y. Dynamic modification of cortical orientation tuning mediated by recurrent connections. Neuron 36: 945–954, 2002.[CrossRef][ISI][Medline]
Finlayson PG. Post-stimulatory suppression, facilitation and tuning for delays shape responses of inferior colliculus neurons to sequential pure tones. Hear Res 131: 177–194, 1999.[CrossRef][ISI][Medline]
Finlayson PG and Adam TJ. Excitatory and inhibitory response adaptation in the superior olive complex affects binaural acoustic processing. Hear Res 103: 1–18, 1997.[CrossRef][ISI][Medline]
Fitzpatrick DC, Kuwada S, Batra R, and Trahiotis C. Neural responses to simple simulated echoes in the auditory brain stem of the unanesthetized rabbit. J Neurophysiol 74: 2469–2486, 1995.
Fitzpatrick DC, Kuwada S, Kim DO, Parham K, and Batra R. Responses of neurons to click-pairs as simulated echoes: auditory nerve to auditory cortex. J Acoust Soc Am 106: 3460–3472, 1999.[CrossRef][ISI][Medline]
Furukawa S and Middlebrooks JC. Sensitivity of auditory cortical neurons to locations of signals and competing noise sources. J Neurophysiol 86: 226–240, 2001.
Fuzessery ZM and Hall JC. Role of GABA in shaping frequency tuning and creating FM sweep selectivity in the inferior colliculus. J Neurophysiol 76: 1059–1073, 1996.
Gilbert CD. Plasticity in visual perception and physiology. Curr Opin Neurobiol 6: 269–274, 1996.[CrossRef][ISI][Medline]
Gilbert CD and Wiesel TN. The influence of contextual stimuli on the orientation selectivity of cells in primary visual cortex of the cat. Vision Res 30: 1689–1701, 1990.[CrossRef][ISI][Medline]
Goldberg JM and Brown PB. Response of binaural neurons of dog superior olivary complex to dichotic tonal stimuli: some physiological mechanisms of sound localization. J Neurophysiol 32: 613–636, 1969.
Greek KA, Chowdhury SA, and Rasmusson DD. Interactions between inputs from adjacent digits in somatosensory thalamus and cortex of the raccoon. Exp Brain Res 151: 364–371, 2003.[CrossRef][ISI][Medline]
Harris DM and Dallos P. Forward masking of auditory nerve fiber responses. J Neurophysiol 42: 1083–1107, 1979.
Imig TJ, Irons WA, and Samson FR. Single-unit selectivity to azimuthal direction and sound pressure level of noise bursts in cat high-frequency primary auditory cortex. J Neurophysiol 63: 1448–1466, 1990.
Irvine DR, Rajan R, and Aitkin LM. Sensitivity to interaural intensity differences of neurons in primary auditory cortex of the cat. I. Types of sensitivity and effects of variations in sound pressure level. J Neurophysiol 75: 75–96, 1996.
Jenison RL. Correlated cortical populations can enhance sound localization performance. J Acoust Soc Am 107: 414–421, 2000.[CrossRef][ISI][Medline]
Jenison RL and Reale RA. Likelihood approaches to sensory coding in auditory cortex. Network 14: 83–102, 2003.[ISI][Medline]
Kaltenbach JA, Meleca RJ, Falzarano PR, Myers SF, and Simpson TH. Forward masking properties of neurons in the dorsal cochlear nucleus: possible role in the process of echo suppression. Hear Res 67: 35–44, 1993.[CrossRef][ISI][Medline]
Kelly JB and Judge PW. Binaural organization of primary auditory cortex in the ferret (Mustela putorius). J Neurophysiol 71: 904–913, 1994.
Levitt JB and Lund JS. Contrast dependence of contextual effects in primate visual cortex. Nature 387: 73–76, 1997.[CrossRef][Medline]
Litovsky RY and Delgutte B. Neural correlates of the precedence effect in the inferior colliculus: effect of localization cues. J Neurophysiol 87: 976–994, 2002.
Litovsky RY and Yin TC. Physiological studies of the precedence effect in the inferior colliculus of the cat. I. Correlates of psychophysics. J Neurophysiol 80: 1285–1301, 1998a.
Litovsky RY and Yin TC. Physiological studies of the precedence effect in the inferior colliculus of the cat. II. Neural mechanisms. J Neurophysiol 80: 1302–1316, 1998b.
Mickey BJ and Middlebrooks JC. Responses of auditory cortical neurons to pairs of sounds: correlates of fusion and localization. J Neurophysiol 86: 1333–1350, 2001.
Moore CI, Nelson SB, and Sur M. Dynamics of neuronal processing in rat somatosensory cortex. Trends Neurosci 22: 513–520, 1999.[CrossRef][ISI][Medline]
, the selectivity change for that neuron generated by various tone 1 stimulus levels.