| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Physiology, Hebrew University
Hadassah Medical School, Jerusalem 91120, Israel; and 2 Department of Biomedical Engineering and Center for Hearing Sciences, Johns Hopkins University, Baltimore, Maryland 21205
 |
ABSTRACT |
|---|
 Abstract
 Introduction
Methods
Results
Discussion
References
|
|---|
Nelken, Israel and Eric D. Young. Linear and nonlinear spectral integration in type IV neurons of the dorsal cochlear nucleus. I. Regions of linear interaction. J. Neurophysiol. 78: 790-799, 1997. The principal neurons of the dorsal cochlear nucleus have complex response properties, many of which are classified as type IV. These units integrate energy in the acoustic signal in a nonlinear fashion; for example, at high sound levels the response to a noise of narrow bandwidth and to a band-reject filtered noise with a spectral notch of the same bandwidth may both be inhibitory. However, the sum of these two stimuli, which is broadband noise (BBN), generally gives an excitatory response. In other situations, linear interactions among stimulus components are observed. In this paper, three regimes of approximate linearity were identified. First, best-frequency (BF) tones and equal-energy narrow noisebands centered at BF evoke almost the same response, which is consistent with a stage of linear filtering followed by a nonlinearity that generates the rate responses of the neuron. Second, for sounds close to threshold (10-15 dB re threshold), energy over the full bandwidth of the unit is integrated linearly. Within this regime, responses to the narrow noiseband and the spectral notch mentioned above do sum to equal the response to BBN. Finally, two noisebands centered at different frequencies, such that their sum is a notch in a broad band of noise, sum linearly at low sound levels; the degree of linearity improves as the separation between the noisebands increases. The results are interpreted in terms of a model of type IV response generation containing two inhibitory interneurons: type II units, which are active for narrowband stimuli, including tones, and the wideband inhibitor, which is active for broadband stimuli. In most cases, the onset of nonlinearity occurs for stimuli that significantly activate the type II inhibitory interneuron.
Most auditory neurons respond to sound in a nonlinear way, meaning that the response to the sum of two stimuli is not always equal to the sum of the responses to the individual stimuli. In the auditory nerve, the phenomena of two-tone suppression (Sachs and Kiang 1968 Animal preparation and unit recording
The methods have been described in detail elsewhere (Nelken and Young 1994 Acoustic stimuli and experimental protocol
Acoustic stimuli consisted of BF tones and BBN, used to determine unit type, and a variety of noiseband and notch-noise stimuli, used to test linearity of interaction of stimulus components. All stimuli were 200 ms long with 10-ms rise-fall times and a 1-s repetition period.
Superposition model
The simplest means to predict the responses of a unit to an arbitrary stimulus spectrum is to assume that the unit weights energy in different frequency regions linearly; the predicted response is then the sum of the responses to an appropriately chosen set of simpler stimuli. Figure 1 shows two stimulus situations that were used to test for linearity of DCN unit responses. In each case the spectrum of stimulus S3 equals the sum of the spectra of S1 and S2. Under linear interaction, the response to S3 should equal the sum of the responses to S1 and S2. In practice, spontaneous activity must be treated separately, so the actual linear hypothesis is
This paper is based on the responses of 48 type IV units to the stimulus paradigm described above. Data from 28 of these units were used in a previous report (Nelken and Young 1994 Narrow noisebands produce similar responses to equal-power tones
A simple model of type IV units would consist of a linear tuning stage, followed by a stage of nonlinear rate processing. The range within which such a model holds can be investigated by comparing responses to tones with responses to noisebands of various bandwidths; a prediction of this model is that responses to tones should be similar to responses to sufficiently narrow noisebands. Figure 2A shows the rate-level functions of a type IV unit to BF tones and to noisebands of various bandwidths. The tone level is given in dB SPL; the noiseband rate-level functions are shifted so that their energy is equal to the tone energy at each point on the abscissa. The shape of the BF tone rate-level function is typical. It is excitatory at low levels and inhibitory at high levels; this inhibition is produced by type II units (Voigt and Young 1980
Responses to near-threshold stimuli add linearly
Figure 5 shows a determination of a type IV unit's region of linearity with the use of the stimulus regime of Fig. 1A (Nelken and Young 1994
Responses to separate noisebands can be used to define a region of linearity
Figure 7 shows examples of the range of goodness of fit that can be achieved in the linearity analyses with the use of the stimuli of Fig. 1B. The responses to noisebands (S1 and S2 in Fig. 1B) have been used to predict the responses to a stimulus composed of two flanking bands with a notch between them (S3 in Fig. 1B); the width of the noisebands was chosen to be the same as the width of a flanking band in the notch stimulus. The width of the band into which the notch was introduced was usually slightly narrower than2 * BF, so the width of the flanking bands was BF-NW/2, where NW is the notch width. Usually, the notch widths started at 250 Hz and were tested in octave steps. Figure 7A shows the responses of a type IV unit to a noiseband of fixed spectrum level (
Recently, Nelken and Young (1994)
Linear regime of type IV units
We describe three regimes in which type IV units behave linearly. First, type IV units give essentially identical responses to narrow noisebands and to tones with the same energy. This behavior is consistent with a model in which a linear filter integrates energy over a certain bandwidth and then the energy at the output of that filter drives a nonlinear system that produces the rate-level functions of the neuron. Two small differences were noted: at low levels the maxima of the rate-level functions for narrow noisebands were somewhat lower than for tones (Fig. 3B), and at high levels narrow noisebands inhibited the units somewhat less than equal-energy tones (Fig. 2, A, C, and E). Both effects can be interpreted in terms of the WBI model. We assume that the WBI responds weakly to the narrow noisebands, but that the noiseband response is stronger than the response to equal-energy tones; this behavior is observed for onset units of the type that are likely to be the WBI (Palmer et al. 1996 Implications for DCN models
The results presented here can all be at least qualitatively understood in terms of the model in which type IV units receive inhibition from both type II units and WBIs (Fig. 9). Recently it has been argued that the WBI connection to type IV units is not necessary for explaining type IV responses (Blum et al. 1995
INTRODUCTION
Abstract
Introduction
Methods
Results
Discussion
References
) and combination-tone generation (Goldstein and Kiang 1968) provide examples of nonlinear processing. These and other auditory nerve nonlinearities are well described quantitatively, and models exist that are reasonably predictive of nonlinear effects in the auditory nerve representation of sounds. In the central auditory system, however, there are additional nonlinearities and more complicated models are required to reach a quantitative understanding of spectral and temporal integration processes.
; Osen et al. 1990). DCN units have appropriately complex response characteristics (Nelken and Young 1994; Young et al. 1988). Principal cells of the DCN in decerebrate cats most frequently have the so-called type IV response characteristic (Young 1980; Young and Brownell 1976): they are spontaneously active, and they are excited by low-level best-frequency (BF) tones but inhibited by high-level BF tones. Tone response maps of type IV units (rate of response as a function of tone frequency and sound level) typically show large inhibitory subfields both on BF and off BF. Prediction of responses to broadband noise (BBN) from tone response maps, assuming linear integration of spectral information, usually suggests that the response to noise should be inhibitory (Spirou and Young 1991). Narrowband noise is indeed inhibitory at high levels (as will be shown later), but most type IV units are excited by BBN. Even more surprisingly, it was found that type IV units are also inhibited by notch noise, meaning BBN containing a spectral notch, when the notch is centered on BF (Spirou and Young 1991). This finding illustrates dramatically the qualitative nonlinearity of type IV units: the response to a narrow band of noise centered on BF is inhibitory; the response to a notch stimulus with the same width may be also inhibitory; but the response to the sum of the stimuli, which is a BBN, is excitatory.
METHODS
Abstract
Introduction
Methods
Results
Discussion
References
; Spirou and Young 1991; Young and Brownell 1976). Briefly, 21 adult cats (1.5-4 kg) with clean external ears were used. The animals were premedicated with atropine (0.1 mg im) and tranquilized with xylazine (0.1 mg im). Anesthesia was induced by ketamine (75-100 mg im). After tracheotomy, the animals were placed in a head holder. Body temperature was maintained at 37-39°C with a heating pad under feedback control. Decerebration was performed by aspirating through the superior colliculus under visual control. No anesthetic was given after the decerebration. In four cases, artificial respiration was required when the animals showed signs of respiratory distress. The animals were paralyzed with gallamine triethiodide (10 mg/h iv) and artificially respired; the end-tidal CO2 was continuously monitored and maintained between 4.5% and 5.5% by varying respiratory volume and rate.
). No aspiration of neural tissue is needed with this approach.
). To achieve stability we infiltrated the exposure with agar.
); generally the calibration fluctuated within ±10 dB over the relevant frequency range.
. Of 177 units characterized in this way, 85 were type IV, 27 were type II, and 63 were type III or I/III.
View larger version (20K):
[in a new window]
FIG. 1.
Examples of stimulus combinations used for linear predictions. Plots show schematic frequency spectra of noise stimuli. A: combining noiseband (S1) with notch noise (S2) to predict responses to broadband noise (BBN, S3). B: combining responses to 2 noisebands (S1 and S2) to predict responses to notch between 2 noisebands (S3).
for details). In later experiments, noise stimuli were generated digitally: the spectrum of the stimulus was generated in the frequency domain by filling the real and imaginary parts of the Fourier transform of the signal with independent Gaussian pseudorandom deviates (passbands) or zeros (stopbands). The resulting complex spectrum was then inverse transformed back into the time domain. The transform length was 32,768 points, giving a frequency resolution of 5 Hz at a sampling frequency of 163.8 kHz. A 16-bit D/A converter was used to transform the digital samples into an analog waveform; stimulus level was set by attenuating the analog waveform after D/A conversion. Because the Nyquist rate (81.9 kHz) was above the audible range of cats, no antialiasing filter was used. Frequency bands generated in this way had much steeper slopes than the noisebands generated with the use of the analog modulation technique. Nevertheless, the responses to corresponding stimuli were found to be similar. In all cases, the noisebands and the notches were centered arithmetically around their nominal center frequency.
where R1, R2, and R3 are the response rates to S1-3 and Rspontaneous is spontaneous discharge rate. That is, we assume that the driven rate to S3, defined as (R3
(1)
Rspontaneous), equals the sum of the driven rates to S1 and S2.
RESULTS
Abstract
Introduction
Methods
Results
Discussion
References
) on related issues; additional data from 25 units are described in a related paper concerning nonlinear prediction of responses to noiseband stimuli from tone and BBN data (Nelken et al. 1997).
, 1990), whose thresholds for BF tones are ~10 dB higher than those of type IV units (Young and Brownell 1976). The narrower noisebands have rate-level functions that are similar to each other and to the tone rate level. At wider noise bandwidths, the rate functions differ from the tone function, both in shape and in threshold. Figure 2A shows an example with a relatively large discrepancy between the tone rate-level function and the narrow noiseband rate-level functions. Figure 2C shows an example where the tone and narrow noiseband data almost superimpose. Figure 2E shows an intermediate case. In most units, the BF tone rate-level function is very similar to the noiseband rate-level functions for narrow noisebands.
View larger version (26K):
[in a new window]
FIG. 2.
Analysis of noiseband rate-level functions. A: rate-level functions of type IV unit [best frequency (BF) = 16.9 kHz] for BF tones and noisebands at different bandwidths; noise bandwidth given in key. Stimulus levels explained in text. B: d value as function of noise bandwidth for unit in A. There is large increase in d at bandwidths >3.2 kHz; therefore this is designated critical bandwidth. C and D: same as A and B for another unit (BF = 12.7 kHz). Critical bandwidth: 1.6 kHz. E and F: same as A and B for another unit (BF = 9.8 kHz). Critical bandwidth: 6.4 kHz.
where r1(l) and r2(l) are the responses to the two stimuli at sound level l, var is an estimate of the variance in the responses, computed as the variance of the spontaneous rate measurements, and L is the number of sound levels at which both units were tested. Under the null hypothesis of equal responses, d * L is
![<IT>d</IT>= <FR><NU>1</NU><DE>2<IT>L⋅</IT>var</DE></FR><LIM><OP>∑</OP><LL><SUB><IT>l</IT>=1</SUB></LL><UL><SUP><IT>L</IT></SUP></UL></LIM>[<IT>r</IT><SUB>1</SUB>(<IT>l</IT>) − <IT>r</IT><SUB>2</SUB>(<IT>l</IT>)]<SUP>2</SUP>](/content/vol78/issue2/fulltext/790/img002.gif)
(2)
2 distributed with L degrees of freedom. The expected value for d is 1; values much larger than 1 are evidence for discrepancies between the rate-level functions that cannot be accounted for by added noise of variance var. In the text below, d is used as a measure of the amount of difference between two rate functions, and not as a statistical test for difference; in fact, most pairs of curves are significantly different in a statistical sense. For the units in Fig. 2, A and C, d is 2.38 and 1.03, respectively.
View larger version (33K):
[in a new window]
FIG. 3.
A: histogram of d distance (Eq. 2) between BF rate-level functions and 200-Hz noisebands. Distribution under assumption of equality is expected to peak at d = 1. This is the case, but there is a tail toward higher values of d that is larger than expected. B: scatter plot of maximum driven firing rate in response to BF tones and maximum driven firing rates in response to 200-Hz noisebands. Diagonal line: equal rates. Note that most points fall below line, so that maximum responses to 200-Hz noisebands are somewhat smaller than to BF tones. C: scatter plot of critical bandwidths against BF. Note weak correlation between the two. D: plots of d vs. bandwidth for population of units. These are all units whose BFs were between 8 and 16 kHz, and which were tested at >5 bandwidths. Note large scatter of critical bandwidths.
; Young and Brownell 1976). The critical bandwidth was determined by finding the bandwidth at which d started to increase rapidly for all the units that were tested with the use of this paradigm. Figure 3C shows a scatter plot of the critical bandwidths against the BFs of the units. There is only a weak dependence of the critical bandwidth on BF (r = 0.44, df = 41, P < 0.01). This weak dependence is illustrated also by the examples in Figs. 2 and 3D. In Fig. 2 the three examples are ordered by decreasing BF, going from 16.9 to 9.8 kHz. The critical bandwidth does not follow the BF, however: it is 3.2 kHz for the unit with the highest BF, 1.6 kHz for the middle unit, and 6.4 kHz for the unit with the lowest BF. Figure 3D shows a population of d-versus-bandwidth curves. These are all the cases of units whose BFs lie in the octave range between 8 and 16 kHz, and for which at least five bandwidths were tested. Whereas every unit shows a critical bandwidth, the actual value of the critical bandwidth is between 0.2 and 6.4 kHz, a much wider range than the range of BFs. Even excluding the atypical extreme case of a unit whose critical bandwidth was 0.2 kHz, the range of critical bandwidths is still between 0.8 and 6.4 kHz. Note that this is not the behavior expected of the simple model mentioned at the beginning of this section, if the linear tuning stage is assumed to be peripheral auditory nerve tuning.
1 dB) and the noiseband rate is higher than the tone rate in the inhibitory regime at high levels (at 9 dB).
View larger version (10K):
[in a new window]
FIG. 4.
Analysis of noiseband-widening experiments. Solid lines: responses of type IV unit (BF = 16.6 kHz) to noisebands at 3 spectrum levels as functions of bandwidth; spectrum level given as noise power in 1-Hz band, in units of dB re 20 µPa. Dashed lines: predicted responses, computed as BF tone response at same total stimulus energy. A: at lowest level and smallest bandwidths (<1.5 kHz), responses to noiseband are extremely well predicted by BF tone. B and C: agreement is less good; disinhibition is seen at higher bandwidths (>3 kHz), which is not predicted by responses to BF tones.
). Figure 5A shows the responses of a type IV unit to a BBN with a notch centered on BF (S2 in Fig. 1A), and the response to a noiseband whose bandwidth (0.8 Hz) matches the notch width, also centered on BF (S1 in Fig. 1A). The sum of these two stimuli is a BBN (S3 in Fig. 1A), so, within the linear region, the sum of the responses to the individual stimuli should equal the response to the BBN. Figure 5B shows the responses of the unit to BBN together with a linear prediction of this response, computed by adding the driven rates to the noiseband and the notch (Eq. 1).
View larger version (21K):
[in a new window]
FIG. 5.
Filling notch as test for linearity; each column shows data from 1 type IV unit. A and D: responses to narrow noiseband ( 
, S1 in Fig. 1A) and to notch of similar width (- - -, S2 in Fig. 1A). BFs of units and bandwidths/notch widths given in key. B and E: responses to BBN(, S3 in Fig. 1A) and linear prediction of Eq. 1. C and F: differences between BBN responses and predictions from B and E, in terms of d (Eq. 2) plotted as function of upper limit L of summation used to produce d. Note corner level beyond which d increases rapidly. Distance between threshold of unit and this corner attenuation is linearity range of unit. In all plots, stimulus level is given as spectrum level.
10 dB. Figure 5, D and E, are another demonstration of the same phenomenon in a second unit; this time the response to the notch is inhibitory at low levels, and the response to the BBN is smaller than the response to the narrow noiseband by exactly the inhibition supplied by the notch; the prediction superimposes on the BBN data up to about 25 dB. In both examples, at higher levels the prediction is much smaller than the actual measured response. The high-level behavior of the prediction reflects the strong inhibitory response to noisebands at high levels. Presumably the inhibition is due to type II inhibition of type IV units; with broadband stimuli, like BBN and the notch noise, type II units are not active or only weakly activated (Nelken and Young 1994; Young and Brownell 1976; Young and Voigt 1982), and the inhibition is not seen. Thus the predicted and measured data diverge when the type II inputs become dominant.
0]. The maximum of a type IV unit's rate-level function, of course, occurs at the threshold of the type II inhibitory input to the type IV unit (Young and Voigt 1981), so the region of linearity corresponds roughly to sound levels below the threshold of type II inhibitory input.
View larger version (17K):
[in a new window]
FIG. 6.
Distribution of median linearity ranges for 34 type IV units, equal to range from threshold to rapid increase in d in plots such as Fig. 5, C and F. Median of this distribution: 15 dB.
22 dB SPL) and bandwidth (1.5 kHz), as a function of the center frequency of the band. The response shows an excitatory region when the noiseband is centered at BF (arrow at 6.3 kHz). Figure 7B shows the response to a dual noiseband stimulus (S3 in Fig. 1B) consisting of two bands, each of the same bandwidth and spectrum level as the band in Fig. 7A (); the two bands were separated by 3.2 kHz in center frequency (NW in key). This figure shows response versus the center frequency of the whole complex as it is moved in frequency; the abscissa is the center frequency of the notch between the two noisebands. The dashed line shows the linear prediction of the response to the two-band stimulus computed by adding the driven rates to the single-band stimuli (Eq. 1). The similarity between the predicted and the measured responses is clear.
View larger version (25K):
[in a new window]
FIG. 7.
Linear prediction of response to notch, as in Fig. 1B. Each row shows data from different type IV unit. Responses to single noisebands are shown at left (S1 or S2 in Fig. 1B); noisebands had fixed spectrum level and bandwidth and were moved in frequency; center frequency of noiseband is plotted on abscissa. Arrows: BFs of units. BW in key: noise bandwidth. Response rates to a pair of noisebands (S3 in Fig. 1B) are shown at right ( ); these noisebands have same characteristics as single band, and are displaced in center frequency by NW + BW, giving notch of width NW between the 2 bands; NW is given in key. Rate is plotted against center frequency of interband notch as whole complex is moved in frequency. Dashed lines: predicted responses to a pair of noisebands computed by adding driven rates for individual noisebands (as in Eq. 1), where relationship of center frequencies of noisebands and BF is preserved. Examples are ordered by their goodness of fit, emphasizing fit to notch. A-D: good fits. E and F: medium fit. G-J: poor fits. Values of d (Eq. 2) for comparisons in right column are as follows: (B) 1.20; (D) 7.45; (F) 12.5; (H) 8.21; (J) 16.4.
the prediction is dominated by inhibition, whereas the unit almost did not change its mean firing rate in response to the two-band stimuli at most frequencies.
View larger version (15K):
[in a new window]
FIG. 8.
A: scatter plot of d, for comparisons like those in right column of Fig. 7, against spectrum level of noisebands, where spectrum level is measured re spectrum level at rate maximum of reference (200 Hz) noiseband rate-level function. Note that above 5 dB re rate maximum, predictions are roughly evenly divided between good (d < 6.2, its median value) and poor (d > 6.2) values, whereas below 5 dB most predictions are good. B: box plot of rank of d as function of rank of notch width. Values of d were ranked within each unit, and ranks were normalized to 1. They are plotted against rank of notch width, again ranked within each unit. For each notch width, distribution of d ranks is plotted as box. Short midline of box: median d rank. Whole box covers 50% of data range, arranged symmetrically around median (interquartile range). Length of each dashed line: 1.5 times interquartile range. Plus signs: single values beyond range covered by dashed lines. Note evident decrease of d rank with notch width rank.
0], this effect was lost when plotting d as a function of notch width for the whole population. To demonstrate the effect despite this large variability, the d values for each unit were normalized by using ranks. For each unit tested with the use of this paradigm, the d value at each notch width was ranked (there were between 2 and 6 notch widths tested for each unit) and the ranks were normalized between 0 and 1; notch widths were also ranked, and the ranks were compared (Fig. 8B). The decrease of the rank of d with increase in bandwidth is readily apparent in this figure. There was a significant effect of the notch width on the rank of d [1-way ANOVA, F(5,120) = 11.9, P 
0.001].
DISCUSSION
Abstract
Introduction
Methods
Results
Discussion
References
suggested a model of DCN type IV units that accounts qualitatively for most of the known response characteristics of these cells (Fig. 9). In the model, two sources of inhibition differentially affect the responses of type IV units for narrowband and broadband stimuli. Inhibition caused by type II units is dominant for narrowband stimuli at high levels, whereas for broadband stimuli a wideband inhibitor (WBI) inhibits type II units strongly and type IV units weakly. In the following discussion, we interpret the results of this paper in terms of this model.
View larger version (19K):
[in a new window]
FIG. 9.
Working model for generation of type IV response characteristics in dorsal cochlear nucleus (DCN). Horizontal line at bottom: tonotopic array of inputs to DCN from auditory nerve fibers (Osen 1970 ) and perhaps from multipolar neurons of ventral cochlear nucleus (Oertel et al. 1990; Smith and Rhode 1989). Inputs to DCN are shown originating at points along line corresponding to their BFs and terminating on type IV units (IV) and on 2 populations of inhibitory interneurons, type II units (II) and postulated wideband inhibitor (WBI). Type II units respond to narrowband stimuli (tones and narrow noisebands), but not to BBN. They inhibit type IV units strongly and produce nonmonotonic BF responses shown in Fig. 2, A, C, and E; nonmonotonicity occurs because type II units have higher thresholds than excitatory inputs to type IV units (Young and Brownell 1976; Young and Voigt 1981). WBI responds preferentially to broadband stimuli; it strongly inhibits type II units, preventing them from responding to broadband stimuli, and weakly inhibits type IV units, producing their inhibitory responses to notch noise (Nelken and Young 1994). Characteristics inferred for WBI match those of onset-C units of ventral cochlear nucleus (Winter and Palmer 1995), which are known to project inhibitory terminals to DCN (Smith and Rhode 1989).
). At low levels, around the maximum of the BF rate-level function, type II units are minimally active. As a result, the weak inhibition of type IV units by the WBI is revealed, so that the type IV units are slightly more inhibited by narrowband noise than by BF tones. At high levels, type IV responses are dominated by type II inhibition. The major effect of the WBI in this case is to inhibit type II units more strongly with narrow noisebands than with tones, and the net result is a weak disinhibition of type IV units with narrow noisebands. This explanation works only if type II units respond slightly more weakly to high-level noisebands than to high-level BF tones. This is, in fact, the case: among 12 type II units studied with both BF tones and 200-Hz noisebands, the median ratio of the maximum noiseband response to the maximum tone response was 0.91, with a range of 0.64-0.98 (e.g., Fig. 13 in Nelken and Young 1994).
11 dB, the WBI is excited only by stimuli with bandwidth larger than ~1 kHz, whereas at 1 and 9 dB it is already excited by stimuli narrower than 200 Hz.
with the use of noiseband-widening methods similar to those of Figs. 2 and 4. The resulting bandwidths were highly correlated with BF. Thus the critical bandwidths of DCN type IV units, as measured in Fig. 2, must be determined by factors in addition to peripheral tuning. These factors might include variation in the BF range over which auditory nerve fibers converge on DCN type IV units and variation in the integration bandwidths and thresholds of DCN inhibitory interneurons.
). Therefore both linearity regimes probably reflect mostly interactions between the excitatory input on type IV units and the WBI. In fact, whenever type II units are activated, the linearity prediction fails in such a way that the prediction does not include the effects of the type II inhibition of type IV units. This was discussed in RESULTS for the case shown in Fig. 5. In cases such as that in Fig. 7, most failures in the prediction of notch-noise responses by the sum of the responses to noisebands were caused by the predicted rates being too small. A likely cause is the bandwidth behavior of the WBI. The WBI is postulated to respond more strongly to wideband than narrowband stimuli and the onset-C neurons show this behavior over a very wide bandwidth (Palmer et al. 1996). Because the notch is embedded in a noiseband with bandwidth more than twice the width of the single bands used in the prediction, the notch noise should activate the WBI more strongly and the type II units less strongly than the single bands; the net effect should be larger responses to the notch noise than to the single noisebands.
; Reed and Blum 1994). The argument is based on modeling studies in which much of the variability in type IV unit properties is explained by variations in the connectivity between type II and type IV units. However, this model does not explain either the bandwidth of the effects of notch noise on type IV units or the low threshold of the broadband inhibition, as discussed by Nelken and Young (1994). The decrease in rate at the peak of BF rate-level functions, shown in Figs. 2A and 3B, is also difficult to explain with the model of Blum et al. In our model, it is explained by activation of the WBI; in the model of Blum et al., it would have to be explained by type II units responding more strongly to narrow noisebands than to tones at levels near threshold; such a difference is not observed (e.g., Fig. 13 of Nelken and Young 1994).
), and should decrease to zero as the linear range is approached. In this sense, this paper imposes boundary conditions on any realistic model of DCN type IV units.
| |
ACKNOWLEDGEMENTS |
|---|
The comments of our colleagues in the Hearing Sciences Center were helpful in preparing this manuscript.
This work was supported by National Institute of Deafness and Other Communications Disorders Grant DC-00115 and by a grant from the Israeli Science Foundation.
| |
FOOTNOTES |
|---|
Address for reprint requests: I. Nelken, Dept. of Physiology, Hebrew UniversityHadassah Medical School, PO Box 12272, Jerusalem 91120, Israel.
Received 30 August 1996; accepted in final form 15 April 1997.
| |
REFERENCES |
|---|
|
Abstract
Introduction
Methods
Results
Discussion
References
|
|---|
This article has been cited by other articles:
|
|
 |
|
  S. Bandyopadhyay, L. A. J. Reiss, and E. D. Young Receptive Field for Dorsal Cochlear Nucleus Neurons at Multiple Sound Levels J Neurophysiol, December 1, 2007; 98(6): 3505 - 3515. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. A. J. Reiss, S. Bandyopadhyay, and E. D. Young Effects of Stimulus Spectral Contrast on Receptive Fields of Dorsal Cochlear Nucleus Neurons J Neurophysiol, October 1, 2007; 98(4): 2133 - 2143. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Fujino and D. Oertel Bidirectional synaptic plasticity in the cerebellum-like mammalian dorsal cochlear nucleus PNAS, January 7, 2003; 100(1): 265 - 270. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Sen, F. E. Theunissen, and A. J. Doupe Feature Analysis of Natural Sounds in the Songbird Auditory Forebrain J Neurophysiol, September 1, 2001; 86(3): 1445 - 1458. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. J. Yu and E. D. Young Linear and nonlinear pathways of spectral information transmission in the cochlear nucleus PNAS, October 24, 2000; 97(22): 11780 - 11786. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. A. Spirou, K. A. Davis, I. Nelken, and E. D. Young Spectral Integration by Type II Interneurons in Dorsal Cochlear Nucleus J Neurophysiol, August 1, 1999; 82(2): 648 - 663. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. X. Joris and P. H. Smith Temporal and Binaural Properties in Dorsal Cochlear Nucleus and Its Output Tract J. Neurosci., December 1, 1998; 18(23): 10157 - 10170. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. Nelken, P. J. Kim, and E. D. Young Linear and Nonlinear Spectral Integration in Type IV Neurons of the Dorsal Cochlear Nucleus. II. Predicting Responses With the Use of Nonlinear Models J Neurophysiol, August 1, 1997; 78(2): 800 - 811. [Abstract] [Full Text] [PDF]
|
|
| ||||||
