Neural Encoding of Single-Formant Stimuli in the Cat. II ...

JOURNALOF NEUROPHYSIOLOGY Vol. 71, No. 1, January 1994. Printed in U.S.A.

Neural Encoding of Single-Formant Stimuli in the Cat. II. Responses of Anteroventral Cochlear Nucleus Units

XIAOQIN WANG AND MURRAY B. SACHS Department of Biomedical Engineering, Center for Hearing Sciences, and W. AI. Keck Center for the Study of the Special Senses, Johns Hopkins University, Baltimore, Maryland 2120.5

SUMMARY AND CONCLUSIONS

1. We have studied responses of anteroventral cochlear nucleus (AVCN) units to single-formant stimuli (SFS), in an effort to make quantitative comparisons with responses observed in auditory-nerve fibers ( ANFs) to the same stimuli (Wang and Sachs 1993 ) and to reveal some of the signal processing mechanisms at the AVCN. Single-unit recordings and subsequent analyses were performed on each type of commonly recorded units, namely primarylike (Pri), primarylike with notch (PN), sustained chopper (ChS), transient chopper (ChT), and onset chopper (OnC), as well as a few onset (On) units, from the AVCN in anesthetized cats. The responses were obtained at a wide range of sound levels and at a frequency range of 1- 10 kHz. Modulation in the envelopes of discharge patterns was quantified by a measure called modulation depth.

2. At moderate to high sound levels, most AVCN units were found to have enhanced modulation depth compared with that of ANFs, although the degree of enhancement varies among different types. All AVCN units, except Pri type, showed an enhancement in modulation depth over that of the highest of ANFs at moderate to high sound levels in the order of (from the highest to the lowest) On, OnC, ChT/PN, and ChS. Specifically, 1) modulation depth in Pri units was comparable to that of high spontaneous rate (SR) ANFs at low sound levels and to that of low/ medium SR ANFs at high sound levels (in dB SPL). When sound level was normalized by unit threshold, Pri units, on average, exhibited only limited enhancement in envelope modulation at high sound levels (>80 dB re threshold); 2) PN units showed substantially enhanced modulation depth over that of all SR groups of ANFs at moderate to high sound levels in dB SPL or dB re threshold scales; 3) significant enhancement in modulation depth was seen in both ChS and ChT units, with a slightly higher modulation depth in ChT type across sound levels (in dB SPL or dB re threshold); 4) modulation depth of OnC units was higher than those of primarylike (Pri and PN) and chopper (ChS and ChT) units at a wide range of sound levels; 5) responses from a limited sample of On units showed the highest modulation depth among all types of AVCN units.

3. Detailed analysis revealed that the enhanced modulation depth in the responses of AVCN units is the result of increased envelope peak height and decreased envelope minimum, relative to those of ANFs. In particular, 1) PN units achieved this enhanced modulation depth mainly by a reduction in the envelope minimum and a small increase in the envelope peak height; 2) the enhancement in ChS and ChT units resulted from decreased envelope minimum and substantially increased envelope peak height; 3) the greatly reduced envelope minimum in OnC units is the crucial factor in the observed enhancement. Some of OnC units also showed significantly increased envelope peak height; 4) high modulation depth in On units was caused by an extremely low envelope minimum. Data presented in this study also showed that AVCN units can have different average discharge rates and yet still exhibit enhanced modulation depth.

4. The envelope latency of responses to SFS in AVCN units exhibited properties similar to the first spike latency determined by tonal stimuli. ChS/ChT units were found to have longer latency than Pri/ PN units do at all frequencies studied. For both Pri / PN and ChS / ChT units, envelope latency decreased as sound level increased. However, there was a greater reduction in latency from low to high sound level in ChS/ChT units than in Pri/PN units.

5. The results of the present study suggest convergence of multiple ANFs on an AVCN cell as a mechanism for the observed increment in envelope peak height. The effect of convergence is stronger at high sound levels if inputs from low/ medium SR or off-best frequency (BF) ANFs are present. Two mechanisms are suggested to explain the reduction in envelope minimum: 1) a threshold effect due to temporal integration of subthreshold excitatory postsynaptic potentials (EPSPs) and 2) somatic inhibition that changes the discharge threshold in an AVCN cell. The above three mechanisms are proposed as potential candidates responsible for the observed modulation enhancement.

INTRODUCTION

Understanding the roles that cochlear nucleus (CN) neurons have in auditory information processing has been one of the key objectives of auditory research in recent years. The neural spike trains carried by the auditory-nerve fibers (ANFs) enter the brain stem through the CN, where the central processing of auditory signals begins. One important question about signal processing in the CN concerns how different features of acoustic stimuli are processed by different cell types. In this paper we focus on the processing of stimulus envelope modulation, a feature important in the perception of speech (de Boer 1976; Flanagan 1972; Plomp 1976).

It has long been proposed that auditory information is processed by parallel pathways arising within the CN (Evans and Nelson 1973a,b; Kiang et al. 1973; Morest et al. 1973; Poljak 1926). The CN has traditionally been divided into three subdivisions according to its anatomic structures, the anteroventral ( AVCN), posteroventral (PVCN), and dorsal (DCN). Cell types with distinctly different morphol- ogies (Lorente de No 198 1) and membrane properties (Hirsch and Oertel 1988; Oertel 1983; Wu and Oertel 1984) have been found within each subdivision. Different morphological cell types are characterized by different responses to tonal stimuli [reviewed by Young ( 1984)]. The present report focuses on the AVCN.

Two types of large neurons, bushy and stellate cells, are intermingled with small cells throughout the AVCN

0022-3077/94 $3.00 Copyright 0 1994 The American Physiological Society 59

60 X. WANG AND M. B. SACHS

(Brawer et al. 1974; Cant and Morest 1979b; Tolbert and Morest 1982). Bushy cells commonly have a small number of short, stubby dendrites with many branches (Brawer et al. 1974; Cant and Morest 1979a). They are further divided into two morphological classes: spherical and globular bushy cells (Brawer and Morest 1975; Brawer et al. 1974; Cant and Morest 1984; Fekete et al. 1984; Lorente de No 198 1; Rouiller and Ryugo 1984; Ryugo and Fekete 1982). Evidence suggests that spherical bushy cells receive one or a few ANF inputs through large somatic terminals (i.e., the end bulbs of Held; Brawer and Morest 1975; Lorente de No 198 1; Rouiller et al. 1986; Ryugo and Fekete 1982; Ryugo and Sento 199 1 ), whereas globular bushy cells receive a larger number of ANF inputs through smaller somatic terminals (i.e., modified end bulbs; Brawer and Morest 1975; Lorente de No 198 1; Fekete et al. 1984; Rouiller et al. 1986). Stellate cells usually have a large number of long, tapering, and less-branched dendrites (Brawer et al. 1974). On the basis of patterns of synaptic innervation, AVCN stellate cells are divided into two morphological types (Cant 198 1). One type of stellate cell is sparsely innervated on its somata and receives most synaptic inputs on its dendritic tree. The other group of cells has synaptic inputs con- tacting >80% of the area of their somata and proximal dendrites. A large portion of these inputs are noncochlear and may be inhibitory, on the basis of vesicle morphology (Cant 198 1; Smith and Rhode 1989; Tolbert and Morest 1982).

Studies in isolated brain slice preparations reveal distinctly different membrane properties in AVCN stellate and bushy cells. Oertel and colleagues (Oertel 1983; Wu and Oertel 1984) found that stellate cells have a linear membrane voltage-current relation, whereas bushy cells have a nonlinear voltage-current relation. Moreover, stellate cells discharge regularly when depolarized by steady current, but bushy cells give only one discharge. Besides differences in their morphology and membrane properties, bushy and stellate cells differ in their responses to pure tone stimuli. Bushy cells exhibit poststimulus time histograms (PSTHs) to best frequency (SF) tones that are similar to those observed in the ANFs and thus are called primarylike units, whereas stellate cells have PSTHs with equally spaced peaks and are therefore referred to as chopper units (Pfeiffer 1966). In addition, analysis of interspike intervals of responses to short tone bursts (Bourk 1976; Young et al. 1988a) shows that primarylike units have larger coefficients of variation (CV, defined as the standard deviation of interspike intervals divided by the mean interval) than those of chopper units. This implies that chopper units discharge more regularly than do primarylike units. The chopper- PSTH is the result of this regular firing pattern. Phase-locking in the responses of primarylike units is as strong as that of ANFs; phase-locking is much weaker in the responses of chopper units (Blackburn and Sachs 1989; Bourk 1976).

Recent studies of responses to speech sounds indicate that different aspects of the ANF representation of speech spectra are processed by populations of AVCN chopper and primarylike units (Blackburn and Sachs 1990). The study by Blackburn and Sachs ( 1990) showed that the spectrum of a vowel is well represented by rate-place profiles (plots of discharge rate vs. unit BF) of AVCN chopper units over the entire range of stimulus intensities tested, but temporal-

place profiles (plots of some measure of phase-locking vs. unit BF) observed in ANFs are degraded in chopper units. On the other hand, primarylike units preserve temporal- place but not rate-place profiles from the ANF. These results suggest that populations of chopper units may function as a rate processor of stimulus spectrum, whereas primarylike units may have a role in preserving temporal information. Although details of the processing leading to these differences are not known, Blackburn and Sachs suggest that their results are consistent with the hypothesis that chopper units perform a “selective listening” operation on their inputs (Winslow et al. 1987). By selective listening we mean that responses of choppers reflect their inputs from high spontaneous rate (SR) ANFs at low sound levels and their low SR inputs at high sound levels. High SR ANFs provide a good rate-place representation of vowel spectra at low sound levels and not at high sound levels because of rate saturation; low SR ANFs provide a good representation at high sound levels because of their higher thresholds and wider dynamic ranges (Liberman 1978; Sachs and Young 1979; Schalk and Sachs 1980).

Envelope modulation is another feature that carries important information in the speech signal (de Boer 1976; Flanagan 1972; Plomp 1976). A study of neural representation of the envelope modulation across sound levels by AVCN cells will provide us not only with information on signal processing mechanisms in the CN but also with an important data base for constructing better models of AVCN cells. In a previous study with single-formant stimuli (SFS), we showed that the representation of envelope modulation is preserved at high sound levels by low/ medium SR ANFs but not by high SR ANFs (Wang and Sachs 1993 ) . In the present report, we will describe how the ANF representation of envelope modulation in SFS is trans- formed by AVCN units. Because of the differences in the behavior of low and high SR ANFs, our results are important in evaluating the selective listening hypothesis.

METHODS

Preparation

Techniques used in preparations are essentially identical (including anesthesia) to those used in our previously reported ANF experiments (Wang and Sachs 1993), except the exposure of the AVCN and the use of platinum-iridium microelectrodes. Briefly, healthy adult cats weighing 2-4.5 kg and free from external ear obstructions or middle ear infections were injected intramuscu- larly with ketamine hydrochloride ( lOO- 120 mg). Atropine sul- fate was administered to reduce mucous secretion. A tracheotomy was performed. Anesthesia was maintained by intravenous injection of pentobarbital sodium throughout the experiment. Cats were placed in a soundproof chamber (Industrial Acoustic 1204A). Rectal temperature was maintained at -38OC. The external meatus was exposed and transected. A several-feet-long tube ( 1.4 mm in diam) was glued to a hole drilled into the bulla to prevent the buildup of negative pressure (Guinan and Peake 1967). The cranium between the tentorium and the nuccal ridge was removed, and the AVCN was exposed by partial aspiration of the cerebellum. Platinum-iridium microelectrodes with platinized tips were used to record from AVCN units and were placed in the AVCN under direct visual observation: Electrodes were advanced dorsoventrally with the use of a hydraulic microdrive.

ENCODING SINGLE-FORMANT STIMULI IN AVCN 61

A Single-Formant-Stimulus (SFS)

Waveform

8 msec

B Envelope Features

Period Histogram

WI

1 - - - - -_

E

Y

o-

cn”

8 b -I--- 4-- I J-1

t msec

Amplitude Spectrum

FIG. 1. A: example of single-formant stimuli (SFS). Waveform and amplitude spectrum of an SFS with carrier frequency equal to 1 .O kHz are illustrated. Period of stimulus shown is - 8 ms; 2 periods are plotted. B: features of envelope of a period histogram and definition of measure of envelope modulation, modulation depth. Period histogram of neural response to SFS is plotted with its computed envelope superimposed. Envelope features used in present study are marked. Modulation depth is defined as ratio of envelope peak height to envelope maximum, as shown.

a envelope maximum (spikedsec)

b envelope minimum (spikedsec)

c envelope peak height (spikedsec)

t envelope peak location (msec)

Modulation Depth of C

Period Histogram =

a

Acoustic stimuli

Acoustic stimuli were delivered via a closed acoustic system utilizing an electrostatic earspeaker ( Sokolich 1977 ) . Acoustic calibrations at the eardrum were performed for each cat with the use of a probe tube. Examples of acoustic calibrations for this system have been published previously (Winslow and Sachs 1987). Stim- uli used included asynchronous short BF tone bursts (50-ms duration, 1.6-ms rise and fall time, 500-ms repetition rate) to collect data for PSTHs and regularity analysis and synchronous bursts of SFSs (400-ms duration, IO-ms rise and fall time, 1 ,OOO-ms repetition rate) at a number of sound levels to collect data for constructing period histograms. The detailed description of SFS is given in Wang and Sachs ( 1993). Briefly, SFS used in this study is a narrowband stimulus, digitally synthesized to approximate the response of an resistor-inductor-capacitor (RLC) circuit to a peri- odic impulse train. The 3-dB bandwidth of the stimulus was set equal to the average 3-dB bandwidth of vowel formants at the stimulus carrier frequency for frequencies ~3.0 kHz and set to a constant thereafter up to a high frequency limit of 10.0 kHz. The formant bandwidths were taken from the study by Dunn ( 196 1). As a result, the amount of amplitude modulation in SFS increases from 1 .O to 3.0 kHz and remains constant for higher carrier frequencies (Wang and Sachs 1993). A set of 128 stimuli were pre- synthesized and stored; their carrier frequencies were logarithmi- cally equally spaced over the frequency range between 1 .O and 10.0 kHz. The fundamental frequency of the stimulus, equal to 1 /stimulus period, is the subharmonic of the carrier frequency closest to 125 Hz. The period of an SFS is -8.0 ms, digitized at 256 samples per period. An SFS with carrier frequency of 1 .O kHz is shown in Fig. 1A. Two periods of the SFS and its amplitude spectrum are plotted in this figure. For each AVCN unit to be studied, one stimulus was chosen from the set such that its carrier frequency was as close as possible to the unit BF or to some other

desired frequencies; with 128 stimuli, carrier frequency was al- ways within 1.0% of BF or the desired frequency, the limitation being imposed by the number of digital waveforms that were stored.

Data recording Discharges of single AVCN units were recorded in the form of

spike times for off-line analysis. Time resolution was 10 ,US for responses to SFS and 100 pus for all other stimuli. During an experiment, an AVCN unit was first isolated with noise bursts as search stimuli. Then BF and threshold of the unit were determined from a tuning curve derived using an automated tuning curve proce- dure (Kiang et al. 1970). Responses to short tone bursts at the BF, 30 dB above unit threshold, were collected to allow computation of PSTHs for the purpose of unit classification; usually 300 short tone bursts were presented. Samples of spike waveform were digitized and stored to identify prepotentials in the waveform (Pfeiffer 1966). Spontaneous discharges were recorded over a long period, usually over 50 s, to obtain an accurate estimate of SR. A rate-level function for long tone bursts at the BF in 1 -dB increments was also obtained. Finally, long bursts of SFS (400 ms in duration) chosen from the set of 128 were presented once per second until ~4,000 spikes were collected or 100 SFS bursts were presented. Data were collected for SFSs over a range of sound levels. For data presented in this report, the carrier frequency of SFSs was set to the unit BF (within I % ) .

Histological reconstruction Histological reconstructions of the AVCN were performed in

most AVCN experiments to verify the locations where recordings were made (Wang 199 1). At the end of each experiment, one or two electrolytic lesions were made, usually at the posterior edge of all recording sites. The animal was then perfused transcardially


with isotonic saline, followed by 10% formaldehyde (Formalin). After fixation, the brain was removed and the CN was blocked. Frozen sagittal sections were stained with cresyl violet and Luxol fast blue. Usually a few recording tracks could be identified by the darkly staining inflammatory response produced by the microelectrodes and the electrolytic lesions. None of the identified tracks or lesion sites was outside the AVCN. Not all electrode tracks were reconstructed in histology; however, because all penetrations were anterior to the lesions that were found, we are confident that our recordings were all made within the AVCN.

Data analysis

AVCN data presented in this report are from 13 experiments. ANF data used for the purpose of comparison are from a previously reported study (Wang and Sachs 1993) in which the same SFSs were used. Unless indicated, ANF data were pooled from seven experiments. In some cases, a smaller group of ANF data is used in comparison, and this is indicated in the figure caption. These ANF experiments were chosen because they were conducted closest in time to the AVCN experiments reported here and had a fairly large number of units that distribute across SR and BF. ANFs are divided into three populations in the present study on the basis of their SR: low SR ( SR < 1 .O spikes/s), medium SR ( 1 .O < SR 5 18.0 spikes/s) and high SR (SR > 18.0 spikes / s) .

UNIT CLASSIFICATION. AVCN units were classified using the same scheme as that reported by Blackburn and Sachs ( 1989) except that an additional type, onset-chopper, is added. The classification was based on 1) analysis of PSTH patterns of responses to 50-ms tone bursts at unit BF, 30 dB above threshold; 2) analysis of interspike interval regularity within the 12- to 20-ms interval after stimulus onset of responses to 50-ms tone bursts; 3) for units with “chopping” PSTH, spike-per-peak (the fraction of stimulus trials during which a spike occurred in each peak) values were computed for the first few peaks in the PSTH. AVCN units were classified into six types, primarylike (Pri), primarylike with notch (PN), sustained chopper (ChS), transient chopper (ChT), onset chopper (OnC), and onset (On), according to the following criteria [see Blackburn and Sachs ( 1989) for details]. A few units that did not fit any of the above six types were classified as “unusual” and were not included in the data analysis. Briefly, units were first classified into “primarylike, ” “onset,” or “multimodal” category on the basis of their PSTH shape. Units in the “primarylike” category were further classified into Pri, PN, or On type using the criteria described by Blackburn and Sachs ( 1989). Units in “onset” category were classified as On type. Units in “multimodal” category were examined to see whether they had significant “chopping.” This was done by calculating spike-per-peak values for the first few peaks in the PSTH (Young et al. 1988a). Units that had sufficiently high spike-per-peak values (>0.9 in the 1 st 3 peaks) were put in the chopper category and were then subject to the analysis of their regularity. I f a unit had a nearly constant mean interspike interval (ISI) or had a linearly increasing mean IS1 and a CV < 0.3 (averaged in 12-20 ms after tone onset), it was classified as ChS. Otherwise, the unit was classified as ChT. Units that did not have 20.9 spikes-per-peak values in the first three peaks of their PSTHs were classified as OnC type. In this study, we defined the OnC type solely on the basis of the PSTH shape. The tuning characteristic and dynamic range, which were used by Rhode and Smith (1986), were not included in the definition adopted in this study. However, many of our OnC units had monotonically increasing rate-level functions in response to BF tones (data not shown). Throughout the rest of this paper, each of the six AVCN unit types will be referred by its abbreviation. For the sake of brevity, we will sometimes use the word “primarylike” for

Pri and PN types, “chopper” for ChS and ChT types, and “onset” for OnC and On types.

COMPUTINGPERIODHISTOGRAMSANDENVELOPES. Themeth- ods used to compute period histograms and their envelopes are described in detail in Wang and Sachs ( 1993 ) . A period histogram is constructed by folding a data record about one period of the stimulus. The envelope of a period histogram is constructed on the basis of the spectrum of the histogram. The spectra of period histograms of responses to SFSs in AVCN units usually consist of from one to a number of isolated peaks, depending upon the unit type and BF value; these peaks are centered at dc and at frequencies equal to unit BF and its harmonics. Each of the peaks can be considered as a narrowband signal, whose waveform in time do- main is equivalent to a sinusoid multiplied by a slowly changing waveform. This slowly changing waveform is referred to as the fractional envelope of the spectral peak. Fractional envelopes from all spectral peaks are summed linearly to obtain the envelope of the period histogram. An example of a period histogram and its envelope is shown in Fig. 1 B.

QUANTIFICATION OF ENVELOPE FLUCTUATION. Several features are used to characterize the envelope. They include envelope maximum, envelope minimum, and their difference, envelope peak height. Figure 1 B illustrates the definitions of envelope features. Fluctuation in the envelope of a period histogram is quantified by a measure called modulation depth, defined as the ratio of envelope peak height to envelope maximum. The value of the modulation depth ranges from 0.0 (flat envelope) to 1.0 (fully modulated; i.e., envelope minimum equals 0).

RESULTS

Envelope modulation in responses to SFS

PRIMARYLIKE POPULATIONS. Period histograms. Primary- like units include two subtypes: Pri and PN. These units are found to have regularity and phase-locking to BF tones similar to those of ANFs (Blackburn and Sachs 1989; Bourk 1976; Young et al. 1988a). The period histograms of their responses to SFS shown in Fig. 2 are also similar to those of ANFs (Wang and Sachs 1993). In this figure, we plot period histograms and their envelopes from a Pri (BF 3.026 kHz) and a PN (BF 2.634 kHz) unit at a number of sound levels. Both units show phase-locking to their BF (equal to stimulus carrier frequency) in the period histograms. The PN unit shown maintains high envelope modulation at 95 dB SPL, a level at which envelope modulation nearly disap- pears in the responses of high SR ANFs (Wang and Sachs 1993 ) . Envelope fluctuation in period histograms of the Pri unit shown decreases faster as sound level increases than does that of the PN unit. Notice that in both the Pri and PN units shown in Fig. 2, envelope maximum increases as sound level increases and then decreases after achieving a maximum value, whereas envelope minimum increases monotonically as sound level increases.

Modulation depth. We examine first modulation depth as a function of sound level in dB re threshold. Because we have shown previously that such functions are similar for ANFs of all SRs (Wang and Sachs 1993 ), a systematic difference between these functions of one type of AVCN unit and those of ANFs would be a clear indication of a transformation, rather than preservation, of the representation of modulation. Modulation depth is plotted versus sound level in dB re threshold in Fig. 3 for both Pri and PN units in three BF ranges. In each panel the shaded area represents


Pri

5/l/91, Unit 2.01, BF 3.026 kHz, 8/14/90, Unit 3.02, BF 2.634 kHz, SR 65.5 spikeshec, TH 5.5 dB SPL SR 1.2 spikeshec TH 13.2 dB SPL

800 26 dB SPL

PN

1 35 dB SPL

1

8oo 46 dB SPL

-

Oi 11 I I I i I I

800

1 86 dB SPL

00 0 8

msec

1 75 dB SPL

1 95 dB SPL

0 8 msec

FIG. 2. Examples of period histograms with envelopes superimposed from a primarylike (Pri) and a primarylike-with-notch (PN) unit at various sound levels. Parameters of units are indicated on plot. Note that sound levels in right column are 9 dB higher than those in left column.

the distribution of the same functions from ANFs with all SRs in the same BF range. The shape of modulation depth versus sound level functions from Pri units (solid lines) is similar to that of ANFs. Except in a few cases, modulation depth of Pri units falls into the area of ANFs at all BFs we studied. PN units (dashed lines), however, behave quite differently from ANFs. Although a few PN units are similar to ANFs, the majority clearly show higher modulation depth at moderate to high sound levels above threshold. Because the falling limb of curves in Fig. 3 is roughly linear, we summarize the above observations by computing sound levels in dB re threshold corresponding to a 0.75 modulation depth and plotting them in Fig. 4A against unit BF. Figure 4B, inset, illustrates how the computation is carried out. Nearly all Pri units fall within the range of ANFs, whereas the PN units lie above that range.

We now examine envelope modulation in primarylike units with respect to sound level in dB SPL. In Fig. 4B, sound levels required to produce a 0.75 modulation depth are plotted versus unit BF for both primarylike units and ANFs. The separation of low/ medium SR from high SR ANFs (Wang and Sachs 1993), with low/medium SR ANFs maintaining modulation at higher sound levels, is

evident in this plot. Pri units are mixed with both high and low/medium SR ANFs. Nearly all PN units are shown to maintain 0.75 modulation depth at sound levels (in dB SPL) comparable to or higher than those of low/medium SR ANFs. Other aspects of the responses will be examined later to further explain the similarities and differences between primarylike units and ANFs reflected in envelope modulation.

The analysis of synchronization index of individual spectral peaks of period histograms and their contributions to overall envelope fluctuation in primarylike units (Wang 199 1) shows distributions similar to those of ANFs reported in Wang and Sachs ( 1993). CHOPPER POPULATIONS. Period histograms. Period histograms and their envelopes for chopper unit responses to SFS are shown in Fig. 5. It is well known that synchronized responses to BF tones degrade in chopper units (Bourk 1976 ) . However, synchronized responses to stimulus envelope are maintained in chopper units over a wide range of sound levels, as shown in Fig. 5. Period histograms of the ChS unit shown in this figure have wider peaks than those

- Pri ----_ PN

ANF (all SR)

BF l-2 kHz

BF 2-3 kHz

0 100 Sound Level (dB re Threshold)

FIG. 3. Comparison of modulation depth vs. sound level (dB re threshold) functions between primarylike units and auditory nerve fibers ( ANFs). Units are separated into 3 best frequency (BF) groups and plotted in each panel, respectively ( l-2,2-3, and 3- 10 kHz) . Pri units are plotted in solid lines, PN units in dashed lines. Shaded area on background indicates distribution of same functions from ANFs, including low, medium, and high spontaneous rate (SR) populations (Wang and Sachs 1993).


modulation depth over that of ANFs are mostly high BF ChT units (shown in dashed lines). Sound levels in dB re threshold corresponding to a 0.75 modulation depth are computed and plotted in Fig. 7A to summarize these observations. All chopper units are shown to be close to or higher than the average value of ANFs across BF, indicating that individual chopper units can code modulation at higher sound levels than do ANFs. Comparison between chopper units and ANFs when sound level is in dB SPL is made in Fig. 7 B. Chopper units are shown to maintain 0.75 modulation depth at higher sound levels than do high SR ANFs. Roughly one-half of the chopper units shown lie in the range of low/medium SR ANFs. The remaining choppers keep 0.75 modulation depth at higher sound levels than do all ANFs. Five ChT units and two ChS units extend 0.75 modulation depth to 100 dB SPL or over, a phenomenon rarely observed in ANFs. ONSET POPULATIONS. Period histograms. Figure 8 shows period histograms and their envelopes from an OnC and an On unit. The shape of these histograms is very different

El3 x Cl

AOx A

0 Low SR ANF A bled SR ANF

x High SR ANF

0 Pri

q PN

Cl q

0 0

EfPAX q O& A

A A

9( x” x Ag x

x

X

x x ChS Ch-6

X 1 l/14/90, Unit 2.01, BF 5.744 kHz 8/7/90, Unit 6.03, BF 4.667 kHz SR 0.0 spikes/set, TH 16.8 dB SPL SR 0.0 spikes/set, 4.5 dB SPL TH

XX

1000 45 dB SPL 1 38 dB SPL AI zi Ls s -------- --- .- ;;;i f-3

0.75

ii Sound Level

0 2 4 6 8 10 1000 55 dB SPL 7 58 dB SPL

BF (kHz)

FIG. 4. Comparison between primarylike units and ANFs in maintaining modulation depth at high sound levels. Sound levels corresponding to a 0.75 modulation depth are plotted vs. unit BF. Sound levels 2 100 dB are plotted as 100 dB. ANF data are pooled from 2 experiments. Symbols for both plots are defined in legend ofA. A : sound levels in dB re threshold. B: sound levels in dB SPL. Inset illustrates how computation is carried out.

g y 1000 c2-

75 dB SPL 4 78 dB SPL of the ChT unit, which is typical of many units we studied. Envelope peak height is generally greater in chopper units than in ANFs (see Fig. 12). Some chopper and a few OnC units gave period histograms that exhibit “multiple peak” patterns in response to SFS, i.e., with two or three peaks in envelopes. The intervals between these multiple peaks were found to be highly correlated with chopping intervals of PSTHs for BF tones, indicating that the “multiple peak” patterns are due to the regular firing of chopper units (Wang 1991).

Modulation depth. Modulation depth of chopper units is plotted versus sound level in dB re threshold in Fig. 6. Also shown is the range of modulation depths of ANFs as the shaded area. For all BF ranges studied, most chopper units have higher modulation depths at moderate to high sound levels than do ANFs. The average position of the falling limbs of curves in Fig. 6 for all chopper units within a BF range lies near or above the upper boundary of ANF data. Notice that choppers that show the largest enhancement in

1000 3

95 dB SPL 4 98 dB SPL

0 0 8

me6 0 8

msec

FIG. 5. Examples of period histograms with envelopes superimposed from a sustained (ChS) and a transient chopper (ChT) unit at various sound levels. Parameters of units are indicated on plot.


- ChS _____ ChT

ANF (all SR)

BF 1-2 kHz

Oi 1 f I I 1 I I I I 1

BF 2-3 kHz

0; I I I I 1 I I I I 1

0 100

Sound Level (dB re Threshold)

FIG. 6. Comparison of modulation depth vs. sound level (dB re threshold) functions between chopper units and ANFs. Units are separated into 3 BF groups and plotted in each panel, respectively. ChS units are plotted in solid lines, ChT units in dashed lines. Shaded area in each panel indicates distribution of same functions from ANFs, including low, medium, and high SR populations (Wang and Sachs 1993 ) .

from that of ANFs. Period histograms of the OnC unit shown in Fig. 8 are characterized by a narrow envelope peak and complete absence of spikes in a portion of period. This OnC unit also shows a monotonically increasing envelope peak height and decreasing envelope peak width with increasing sound level, which is not seen in ANFs, primarylike or chopper units. A large number of OnC units were observed to have monotonically increasing rate-level functions for BF tones across a broad range of sound levels (data not shown). The period histograms of the On unit shown in Fig. 8 are pulse-like with fully modulated envelope at all sound levels tested, which is typical of On units.

Modulation depth. Only a few On units were encountered in the experiments conducted. The percentage of On units encountered in previous studies was similarly small (Blackburn and Sachs 1989; Bourk 1976; Shofner and Young 1985 ) . On units usually show extremely high modulation depth even at the highest sound level tested. This can be seen in Fig. 9, where modulation depth of On (dashed lines) and OnC (solid lines) units is plotted versus sound level in dB re threshold. Every On unit we studied maintained modulation depth >0.9 at all sound levels. Because

the number of On units we studied was small, observations regarding On units should be viewed as being preliminary. The performance of OnC units is diverse, as seen in Fig. 9 B. The diversity of OnC units in Fig. 9 reflects the nature of the unit classification scheme we adopted in this study. The class of OnC units is more heterogeneous than the primarylike and chopper classes. Some OnC units have PSTHs that clearly exhibit “chopping” patterns and significant sustained rates but with smaller “spike-per-peak” values than those of ChT/ChS units. These units behave more like choppers than On units. The other extreme of OnC units has PSTHs which exhibit some degree of “chopping” but have low sustained rates, which makes them closer to On than to ChT/ChS units.

Finally, sound levels in both dB re threshold and dB SPL at which modulation depth is equal to 0.75 are computed for onset units and plotted in Fig. 10. In Fig. 1OA we see that three out of seven OnC units studied are similar to ANFs; the other four show 75% modulation at levels higher than those of ANFs. Because On units often have very high

A 100

1 80 -

60 -

40 -

20 -

0. A A A

A&A A

A:* e A

A )eXA X XA

9 x Oxx x AX

A

o Low SR ANF 0 ChS A Med SR ANF A ChT

1 x High SR ANF

AeoA A A A

0

0 0

b a. l

A

e 0

@@LAX

0 4

XX

40 “co<X xx

0 A x 3

0 A

A

OAt ’ 0

A A

O” Ao.a A

0 A

A A

A

X

X

n A A x

0 i I I I I I I I I ,

0 2 4 6 8 10

BF (kHz)

FIG. 7. Comparison between chopper units and ANFs in maintaining modulation depth at high sound levels. Sound levels corresponding to a 0.75 modulation depth are plotted vs. unit BF. Sound levels 2 100 dB are plotted as 100 dB. ANF data are pooled from 2 experiments. A : sound levels in dB re threshold. B: sound levels in dB SPL. Symbols for both plots are defined in legend of A.


BF thresholds and the highest sound level used in our experiments is - 110 dB SPL for all units, data available from On units are usually within a small range of sound levels above threshold. This is why On units presented in Fig. 10 B (analysis in dB SPL) do not show up in Fig. 1OA (analysis in dB re threshold). In Fig. 10 B, we see that On and OnC units are as good as or better than the best low /medium SR ANFs in maintaining envelope modulation at high sound levels (in dB SPL). COMPARISON OF ENVELOPE MODULATION ENCODING AMONG

AVCN TYPES. Through the above analyses, we have compared modulation depth of each AVCN unit type with that of ANFs. We now turn to more quantitative comparisons between AVCN and ANF units in an effort to determine an order of a unit’s ability to code envelope modulation. Be- cause of the different functional implications, comparisons

QnC

8/7/90, Unit 5.02, BF 4.576 kHz SR 0.0 spikeshec, TH 26.5 dB SPL

On

8/7/90, Unit 7.02, BF 2.996 kHz SR 0.0 spikeshec, TH 33.5 dB SPL

600 1 58 dB SPL 3ooo 1 51 dB SPL

600 1 68 dB SPL 3000 1 61 dB SPL

OL,

600 1 88 dB SPL 3000 1 81 dB SPL

0

600 1 98 dB SPL 3000 1 91 dB SPL

0

3000 1 71 dB SPL

0 I

FIG. 8. Examples of period histograms with envelopes superimposed from an onset chopper (OnC) and an onset (On) unit at various sound levels. Parameters of units are indicated on plot.

l- - - - OnC ----- On

ANF(all SR)

BF 2-3 kHz

BF 3-l 0 kHz

0 Sound Level (dB re Threshold)

FIG. 9. Comparison of modulation depth vs. sound level (dB re threshold) functions between onset units and ANFs. Units are separated into 2 BF groups and plotted in each panel, respectively. No onset units were recorded in BF range of l-2 kHz. OnC units are plotted in solid lines, On units in dashed lines. Shaded area on background indicates distribution of same functions of ANFs, including low, medium, and high SR populations (Wang and Sachs 1993).

will be made with respect to sound level both in dB re threshold and in dB SPL.

Comparison with respect to sound level in dB relative to threshold. Because the modulation depth of SFSs varies with carrier frequency (set equal to unit BF in this study), considerations need to be taken in grouping units for comparisons. Units with BFs between 3 and 10 kHz are grouped together because the modulation depth of SFS stays constant in this frequency range (see METHODS). For this group of units, we compare their modulation depth versus sound level functions, which are shown in Figs. 3, 6, and 9. In Fig. 11, least-square fits of modulation depths at different sound levels are shown for ANFs and each type of AVCN units with BFs between 3 and 10 kHz. Each curve is obtained by fitting the function y = 1 - log [ 1 + ( x/a)b] to modulation depths at various sound levels from one type of unit. The functional form of the fit is chosen on the basis of the general shape of modulation depth versus sound level functions in the frequency range under consideration. Figure 11A shows the results when sound level is in dB re threshold. Low/medium and high SR ANFs are grouped together in this figure because their modulation depths do not show systematic differences with respect to sound level in dB re threshold (Wang and Sachs 1993). From moderate to high sound levels (>50 dB above threshold) we see the following relationships among unit types: 1) the modulation depth of Pri units is similar to that of all ANFs at all but very high sound levels, where they appear to show some enhancement over ANFs; 2) ChS units show modulation depth con- siderably higher than that of ANFs as well as Pri units but lower than that of all other AVCN types at most sound


0 Low SF? ANF + OnC

A Med SR ANF 0 On x High SR ANF

1 I I I I I 1 I I

0

0

0 x

x A

A x x

XX

O&X xx

0 A x # A

2 x

xp8

0 A

A A A

A #Xx x

xx

0

AAx

x x X

x x

0 2 4 6 8 10

BF (kHz)

FIG. 10. Comparison between onset units and ANFs in maintaining modulation depth at high sound levels. Sound levels corresponding to a 0.75 modulation depth are plotted vs. unit BF. Sound levels 2 100 dB are plotted as 100 dB. ANF data are pooled from 2 experiments. A: sound levels in dB re threshold. B: sound levels in dB SPL. Symbols for both plots are defined in legend of A.

levels; 3) the On units have the highest modulation depth, followed by OnC units, whose modulation depth is higher than ANFs and all other AVCN types at high sound levels; 4) there is little difference between modulation depths of PN and ChT units at most sound levels, except at very high sound levels (>9OdB) where modulation depth in ChT units exceeds that of PN units. Quantitative comparisons made for units with BFs of l-3 kHz, based on the data shown in Figs. 4, 7, and 10, are in general consistent with those stated above (data not shown).

We can thus summarize each AVCN type’s ability to code envelope modulation at moderate to high sound levels (in dB re threshold) by the relationship

On > OnC > ChT - PN > ChS > Pri - ANF (all SR)

where the symbols ‘5” and “ m” denote “higher than” and “comparable to,” respectively.

Comparison with respect to sound level in dB SPL. Figure 11 B shows least-square fits of modulation depths at various

sound levels (in dB SPL) for ANFs and each type of AVCN units, all with BFs between 3 and 10 kHz. Low/medium and high SR ANFs are fitted separately in this figure. There is a monotonic increase in modulation depth at moderate to high sound levels in the order of high SR ANF, low/medium SR ANF, PN/ChS/ChT, OnC, and On unit. The difference among PN, ChS, and ChT units is small. Modula- tion depth of Pri units is comparable to that of high SR ANFs at low sound levels and to that of low/medium SR ANFs at high sound level. At medium sound levels, modulation depth of Pri units is between those of low/medium and high SR ANFs. Similar relationships are also found among units with BFs of l-3 kHz (data not shown). One excep- tion is that low-BF ChS units (BF < 3 kHz) seem to be less able than PN or even ChT units to code envelope modulation at moderate to high sound levels.

In summary, the relationship among unit types with re-

5 0.6 .- 5

z r” 0.4

---- ChT

- OnC

n-- On - * - - ANF (all SR)

‘\ \ I I I I I I

0 20 40 60 80 100 120

dB re Threshold

I3 1.2

1 I-

5 0.6 - - Pri .- 5 --

z

RN - ChS

s 0.4 - - - -- ChT

- OnC

-0 On 0.2 - “.:- - ANF (low/med SR)

\ ‘. \ **

- - - - ANF (high SR) \ I \ ,

\ ‘\ 0 I I I I I I

0 20 40 60 80 100 120

dB SPL

FIG. 1 1. Quantitative comparison of modulation depth vs. sound level functions between ANFs and anteroventral cochlear nucleus ( AVCN) units with BFs of 3- 10 kHz. Modulation depths measured at different sound levels are fit to function y = 1 - log [ 1 + (~/a)~]. A : curves are least-square fits of modulation depths at various sound levels in dB re threshold. B: same as A, but sound level is in dB SPL. Symbols are indicated on plot. ANF data are pooled from 2 experiments.


gard to their ability to maintain modulation depth at moderate to high sound levels (in dB SPL) appears to be

On > OnC > PN - ChT - ChS > Pri

- low/medium SR ANF > High SR ANF

Analysis of components of modulation depth

The above analysis demonstrates the enhancement in modulation depth among most AVCN units at moderate to high sound levels. A question one might ask next is, What are the mechanisms underlying such enhancement? Be- cause modulation depth is a normalized measure, it does not directly point to sources of the enhancement. Because envelope maximum is the sum of envelope minimum and envelope peak height, modulation depth can also be ex- pressed as a function of envelope peak height and envelope minimum, i.e.

Modulation Depth = 1

1+ envelope minimum

envelope peak height

Thus a decrease in envelope minimum or an increase in envelope peak height will lead to an increase in modulation depth. We shall see below that the above expression of modulation depth is more convenient for the discussion of possible underlying mechanisms because envelope minimum is typically a monotonic function of sound level, whereas envelope maximum and envelope peak height are not. As will be discussed later, changes in envelope peak height and envelope minimum can result from different mechanisms. In the following, we will examine these two variables at different sound levels and their effects on modulation depth. All analyses in this section are based on data from units with BFs between 3 and 10 kHz; the parameters of SFSs are constant within this frequency range. ANALYSIS OF ENVELOPE PEAK HEIGHT. Envelope peak height, defined as the difference between the maximum and the minimum of an envelope, is a measure of absolute fluctuation in an envelope. As reported previously (Wang and Sachs 1993), the envelope peak height of ANFs first rises as sound level increases and then decreases at high sound levels. When plotted versus unit threshold, envelope peak height from both low/medium and high SR ANFs peaks at about the same sound level in dB re threshold. However, because of their higher thresholds, low/medium SR ANFs maintain higher envelope peak height at high sound levels (in dB SPL) than do high SR ANFs.

Figure 12 gives an overview of envelope peak height as a function of sound level in dB SPL for each of the six AVCN types with ANF data plotted on the background. One characteristic of ANFs is that envelope peak height of low/ medium SR ANFs peaks at a higher sound level (dB SPL) than does that of high SR ANFs. Thus envelope peak height is higher in high SR ANFs at low sound levels and higher in low/medium SR ANFs at high sound levels (all in dB SPL). As shown in Fig. 12A, the envelope peak height of Pri units is similar to that of ANFs in that 1) envelope peak height of Pri units is confined to the range of that of ANFs and 2) envelope peak height of most individual Pri units seems to follow that of ANFs in one SR group. PN units show some degree of increase in envelope peak height at

r] High SR ANF ___

cj:z;: Low/med SR ANF

2000 1 D. ChT

Sound Level (dB SPL) FIG. 12. Comparison of envelope peak height between ANFs and

AVCN units. Envelope peak height is plotted vs. sound level in dB SPL for AVCN units with BF between 3 and 10 kHz. Shaded and outlined areas on background indicate distributions of same functions from low/ medium and high SR ANFs, respectively. A-F: Pri, PN, ChS, ChT, OnC, and On units, respectively.

moderate to high sound levels over that of ANFs (Fig. 12B). In contrast, chopper units exhibit considerable differences from ANFs (Fig. 12, C and D). Most chopper units have envelope peak heights greater than the highest envelope peak height of all ANFs over a range of sound levels. A large portion of OnC units shows much increased envelope peak height over that of ANFs (Fig. 12 E). The maximum envelope peak height in OnC units is, in some cases, more than twice that of the maximum envelope peak height of all ANFs. Interestingly, our limited data show that there is not much deviation of envelope peak height in On units from that in ANFs (Fig. 12 F).

Because there is a fairly large variation in the distribution of envelope peak height within each SR group of ANFs and each type of AVCN unit, values of envelope peak height at a low and a high sound level, 40 and 80 dB SPL, are com-


40 dB SPL A 2000

1 o Low SR ANF cl Pri A Med SR ANF q PN x High SR ANF

1000

= B % 2j 2000 2

. ChS

‘CJ A ChT

u A

1000

0

C 2000 1 # OnC

i

4 4 0 On

4

1000

I’ Ax’ x ’ xx x

x #@# sg ,z”%k yxts;:

0 b&a

A0 b A4l

a” A bag A

A

0 O 4

I I I I I I I 3 4 5 6 7 8 9 10

BF (kHz)

80 dB SPL

2000 0 Low SR ANF III Pri

A Med SR ANF 83 PN

X High SR ANF

A *A o Ax @

>o( XxXxX x I I I

E 2000 0 CRS

A ChT

A

A A

F 2000

1 4 OnC 4

0 On

1 I 4

1000 1

4 4

A A 40

00 4 0 %O

t&b *x & ,"~~xf;x#oo I I I 1

3 4 5 6 7 8 9 10

BF (kHz)

FIG. 13. Envelope peak heights at 40 (A-C) and 80 (D-F) dB SPL are plotted vs. unit BF for AVCN units and ANFs. Units included in this figure have BFs between 3 and 10 kHz. A and D: Pri and PN vs. ANF. B and E: ChS and ChT vs. ANF. C and F: OnC and On vs. ANF. Symbols are indicated on plot.

puted and plotted versus BF in Fig. 13 to give a more quantitative view of the above observations. Notice that high SR ANFs have greater envelope peak height than that of low/ medium SR ANFs at 40 dB SPL (Fig. 13A). The opposite is true at 80 dB SPL (Fig. 13 0). Pri units do not show significant differences from ANFs as a group at either sound level (Fig. 13, A and D). At 40 dB SPL, PN units have envelope peak heights approximately equal to those of high SR ANFs but higher than those of low/ medium SR ANFs (Fig. 13A). At 80 dB SPL, PN units show envelope peak height comparable to or higher than low/medium SR ANFs and higher than that of high SR ANFs (Fig. 13 0). Figure 13, B and E, shows that the majority of chopper units have significantly higher envelope peak height at both low and high sound levels. The increment in envelope peak height over that of ANFs is higher at 80 than at 40 dB SPL.

OnC units seem to divide into two groups (Fig. 13, C and F). About one-third of OnC units have extremely high envelope peak height. The rest of the OnC units have their envelope peak heights equal to those of ANFs. ANALYSIS OF ENVELOPE MINIMUM. In our earlier report (Wang and Sachs 1993), we showed that envelope minimum is a monotonically increasing function of sound level at moderate to high sound levels. Furthermore, it was shown there that, at the same sound level (dB SPL), high SR ANFs have higher envelope minima than do low/medium SR ANFs. Given an envelope peak height, a higher envelope minimum means a lower modulation depth.

In Fig. 14, A and B, envelope minimum is plotted versus sound level in dB SPL for Pri and PN units (BF 3- 10 kHz) along with ANF data. Envelope minimum measured at 80 dB SPL for all primarylike units and ANFs is plotted in Fig.


c

A ‘Ooo 1 I-1 High% ANF

-I Low/med SR ANF

Pri (BF 3-l OkHz)

PN (BF 3-l OkHz)

Sound Level (dB SPL)

0 Low SR ANF

A Med SR ANF

X High SR ANF

0 Pri

q PN

1 BF (kHz) 10

FIG. 14. Comparison of envelope minimum between ANFs and primarylike units. A and B: envelope minimum is plotted vs. sound level in dB SPL for Pri and PN units, respectively (BF 3- 10 kHz). Shaded and outlined areas on background mark range of same functions from low/ medium and high SR ANFs. C: envelope minimum at 80 dB SPL is plotted vs. unit BF for all primarylike units and ANFs. Symbols are indicated on plot.

14C. Note that low/medium SR ANFs as a group show lower envelope minima than do high SR ANFs. Pri units are shown to have their envelope minima spread over the whole range of both low/medium and high SR ANFs (Fig. 14A). Some Pri units even show envelope minima higher than those of ANFs. PN units, on the other hand, have their envelope minima confined entirely to the range of low/medium SR ANFs with a few of them in the overlapping area of low and high SR ANFs at high sound levels (Fig. 14B). At 80 dB SPL, the majority of PN units have envelope minima comparable to or lower than those of low / medium SR ANFs (Fig. 14C). At this sound level, PN units have enhanced modulation depth with a value close to that of chopper units (Table 1 ), and their envelope peak height is higher than that of ANFs but lower than that of chopper units (Fig. 13, D and E). Data in Fig. 14C indicate that the reduced envelope minimum is a significant factor in giving the enhanced modulation depth seen in PN units.

Envelope minima of chopper units are analyzed in Fig. 15. Both ChS and ChT units have their envelope minima distributed within the range for low/medium SR ANFs (Fig. 15, A and B). At 80 dB SPL, envelope minima of the

majority of chopper units are equal to or smaller than those of low/medium SR ANFs (Fig. 15C). Several ChT units have nearly zero envelope minima at 80 dB SPL, whereas their envelope peak height can be as high as 500 spikes/ s or above at the same sound level (Fig. 13 E), more than twice those of ANFs (Table 1). The mean envelope minimum of chopper units (49.96 spikes/s, ChS; 57.45 spikes/s, ChT), for BF 3- 10 kHz at 80 dB SPL, is slightly higher than that of PN units (46.35 spikes/ s) in the corresponding group ( see Table 1). It is noted that the variation in envelope minimum among chopper units is large. The higher envelope minimum reduces modulation depth in chopper units to values close to those of PN units though chopper units have higher average envelope peak height at 80 dB SPL (see Ta- ble 1).

The lowest average envelope minimum in all AVCN units is seen in On units. As shown in Fig. 16B, the few On units we studied have a nearly zero envelope minimum across all sound levels. This extremely low envelope minimum results in an exceptionally high modulation depth in On units, although their envelope peak heights are not greater than those of ANFs (Fig. 12 F). OnC units exhibit envelope minima with values in between those of chopper and On units, as shown in Fig. 16A. At 80 dB SPL, the majority of OnC units have envelope minima comparable to or lower than the lowest envelope minima seen in ANFs (Fig. 16C). Their average envelope minimum (24.12 spikes/s) is much smaller than that of low/medium SR ANFs ( 66.36 spikes/ s) , which is the lowest in SR groups of ANFs (see Table 1). A small envelope minimum and a large envelope peak height at 80 dB SPL give rise to a high modulation depth in OnC units at this sound level.

RELATIONSHIP AMONG ENVELOPE PEAK HEIGHT, ENVELOPE

MINIMUM, AND MODULATION DEPTH. We have seen thus far that both envelope peak height and envelope minimum in- fluence modulation depth. The relationship of these three measures is summarized in Fig. 17. In this figure, envelope peak height is plotted versus envelope minimum for ANFs and AVCN units with 3- 10 kHz BFs, all measured at 80 dB SPL. In Fig. 17, A-D, a point on the abscissa has modulation depth equal to 0; a point on the ordinate has modulation depth equal to 1. Because modulation depth is determined by the ratio of envelope peak height to envelope minimum, all points along a straight line passing through the origin have the same modulation depth. Two such lines, representing 0.64 and 0.9 modulation depth, are plotted in all plots in dashed lines. Thus points with high modulation depth are located at the top left on the plots, whereas those with low modulation depth are at the bottom right.

We see from Fig. 17A that low/medium and high SR ANFs distribute in two clusters, which can be approximately separated by the line representing 0.64 modulation depth or a pair of lines representing envelope minimum of 90 spikes/ s and envelope peak height of 160 spikes/s, respectively. These lines are plotted in Fig. 17, A-D. The separation of two ANF populations in Fig. 17 A means that low/ medium SR ANFs have higher envelope peak height and lower envelope minimum than do high SR ANFs at the sound level analyzed (80 dB SPL).

Figure 17 B shows that most Pri units (7 out of 10) scatter


TABLE 1. Characteristics of responses to SFS at 80 dB SPL

Unit Type N Envelope Minimum

Envelope Peak Height

Modulation Depth

Average Discharge Rate

High SR ANF 36 Low/med SR ANF 31 Pri 11 PN 11 ChS 9 ChT 19 OnC 15 On 3

135.25 + 40.29 66.36 t 39.94

169.74 & 101.12 46.35 AI 42.68 57.45 iz 42.65 49.96 + 52.54 24.12 + 31.31 0.18 Z!Z 0.13

113.45 + 49.08 255.83 k 127.51 238.45 -t 110.53 399.34 AI 222.96 454.94 AI 182.04 536.09 f 272.85 575.88 L 541.31 377.17 XL 266.92

0.44 t 0.14 0.78 + 0.13 0.58 I!I 0.25 0.88 z!I 0.11 0.88 t- 0.08 0.90 xk 0.11 0.93 2 0.09 1.00 * 0.00

182.09 IL 28.13 160.53 AI 50.13 177.72 + 34.60 162.72 AI 76.15 2 14.06 f 73.74 220.74 I!Z 87.88 134.71 Ifr 90.77 47.49 I!I 35.91

Data are from units with BF between 3 and 10 kHz. Numbers are mean I~I SD. SFS, Single-formant stimuli; SR, spontaneous rate; ANF, auditory-nerve fiber; Pri, primary-like; PN, primarylike with notch; ChS, sustained chopper; ChT, transient chopper; OnC, onset chopper; On, onset; BF, best frequency.

below the line representing 0.64 modulation depth, i.e., on the side of high SR ANFs. The envelope peak height for most Pri units is higher than the highest value for high SR ANFs. The Pri units whose modulation depth is in the range of that of high SR ANFs (below 0.64) differ from the main body of high SR ANFs by having higher envelope peak height, an indication that they receive converging ANF inputs at 80 dB SPL (see DISCUSSION). PN units are located where low/medium SR ANFs scatter and they all show modulation depth higher than 0.64 (Fig. 17 B) . Chop-

per units occupy a larger area than do PN units (Fig. 17C). All but one of the chopper units are located above the 0.64 modulation depth line. The higher modulation depths of the onset units at high sound levels result from lower envelope minimum (Fig. 17 D) and, for some units, from large envelope peak height at the same time. It is noted that, except for Pri type, nearly all of the AVCN units shown in Fig. 17 are located above 0.64 modulation depth line, which separates low/ medium and high SR ANFs. Further- more, the majority of AVCN units stays outside the area

E Low/med SR ANF E

Low/med SR ANF

L

:- 8

is--

r$ ChS :- g OnC

g2z (BF 3-10 kHz)

z\ co g2

(BF 3-10 kHz)

0 'zi 5z

g '$ -

z z W W

0 0

B B

1000 1 1000 1 E E

& ‘E g

ia .- ChT :c Q)

r$ On 2. r/l

x2? (BF 3-10 kHz)

g.? (BF 3-10 kHz)

9 'E 0 '5 a>_cn_ z

l5z 2

W W

0 0 I

120 0 120 Sound Level (dB SPL) Sound Level (dB SPL)

C 400

sf EZ zJ* E22

300 .- .- .r 5f-

z-1; 200

8% Zm 2= WiFi 100

?ii

0

P ax X

x x A A

A 2

0 Low SR ANF

A Med SR ANF

X High SR ANF

X X x xx ‘k

AiiX

a ChS

A ChT

0 Low SR ANF l OnC

A MedSRANF 0 On

X High SR ANF

Ax X

x x XX

0 A AAX D 00 AA

A+x +A A

A A :AA

2” 0 A

y@p * I

1 BF (kHz) 10 1 BF ikHz) 10

FIG. 15. Comparison of envelope minimum between ANFs and chop- FIG. 16. Comparison of envelope minimum between ANFs and onset per units. Format of this figure is same as that in Fig. 14. Symbols are units. Format of this figure is same as that in Fig. 14. Symbols are indicated indicated on plot. on plot.


A 500 1

0.9 0 Low SR ANF

I i / / A Med SR ANF

x High SR ANF

B

1 . 0.9 0 Pri

/ H PN /

i

i / 000 i / ;I -7 convergence

A /i

A / i threshold-effect

l ChS

A ChT

-I i / i /

q ! /

’ 0.9 * OnC

0 On

0

Envelope Minimum (spikeslsec)

where low/ medium SR ANFs cluster, which indicates that processing of envelope information carried by ANFs takes place in these units.

Envelope latency in responses to SFS

We have seen how envelope modulation varies as a function of sound level. In this section, the timing of the envelope peak is analyzed. We define envelope latency as the lag between the envelope peak location of a period histogram (see Fig. 1 B) and that of the stimulus used. A 0.55-ms acoustic delay in our sound delivery system is subtracted from the neural data. As one can see from period histograms of responses to SFS (Figs. 2, 5, and 8), the envelope peak location generally decreases as sound level increases. As a result, envelope latency decreases with increasing sound level as well, because envelope location of SFS does not vary with sound level. This is shown in Fig. 18, A and B, in which envelope latency is shown as a function of sound level in dB SPL for primarylike and chopper units with BFs between 2 and 3 kHz. The ranges of envelope latency from low/medium and high SR ANFs are shown by the shaded and outlined areas, respectively, in both figures. PRIMARYLIKE UNITS. Envelope latency of primarylike units, like that of ANFs, decreases as sound level increases up to -60 dB SPL and stays relatively constant thereafter. Note that the envelope latency of primarylike units is roughly parallel to that of ANFs. In other words, the delay in envelope peak between primarylike units and ANFs is approximately uniform across sound levels. Figure 18C compares envelope latency measured at 50 dB SPL between ANFs and primarvlike units. Both Pri and PN units

200 300

FIG. 17. Analysis of relationship among modulation depth, envelope peak height, and envelope minimum. Envelope peak height is plotted vs. envelope minimum, both measured at 80 dB SPL, for ANFs and AVCN units with BF between 3 and 10 kHz. A-D: ANFs, primarylike, chopper, and onset units, respectively. Each of the 2 dashed lines cross- ing origin on A represents a constant modulation depth given by various combinations of envelope peak height and envelope minimum. Number near end of each line indicates the value of the constant modulation depth. Low/ medium SR ANFs on A are approximately separated from high SR ANFs by dashed line representing modulation depth of 0.64. Low/medium SR ANFs are also separated from high SR ANFs by 2 dotted lines parallel to axes along both horizontal (envelope peak height = 160 spikes/s) and vertical (envelope minimum = 90 spikes/s) direc- tions. The 2 dashed and dotted lines on A are also reproduced on B-D to outline distribution of ANFs. Symbols are indicated on plot.

show longer envelope latency than that of ANFs at all BFs tested. CHOPPER UNITS. Chopper units have been shown to have longer minimum first spike latencies than do primarylike units (Blackburn and Sachs 1989; Young et al. 1988a). In Fig. 18 B envelope latency is plotted versus sound level in dB SPL for chopper units with BFs between 2 and 3 kHz. Envelope latencies of ANFs are shown by shaded areas. The average envelope latency of choppers is - 1- 1.5 ms longer than that of high SR units at most sound levels, which is close to the delay measured from the minimum first spike latencies by Young et al. ( 1988a). As shown in Fig. 18 B, there is a larger difference between the envelope latency of ANFs and that of chopper units at low sound levels than at high sound levels. The comparison between chopper units and ANFs in envelope latency at 50 dB SPL across BF is shown in Fig. 18D. Figures 18, C and D, show that the envelope latency in chopper units is significantly longer than that in primarylike units. The difference is greater at BFs above -3 kHz than below this frequency. A similar relationship in minimum first spike latency was reported previously by Young et al. ( 1988a).

DISCUSSION

Possible neural mechanisms for modulation enhancement

Response properties of an AVCN cell depend on the nature of inputs (e.g., excitatory or inhibitory, at or off unit BF, etc.), synaptic configurations (e.g., ANFs synapsing at somata or along dendritic trees) and biophysics of the cell membrane. Specifically, we will consider below these issues: 1) convergence of excitatory ANF inputs, 2) conver-


A r-1

E% 8 1 BF 2-3 kHz

High SR ANF 1-1 High SR ANF : _._.:. _.: ._. .:_ ..:.: :. .iil.i..i!:.j5’:iiiii’.’ Low/med SR ANF ,... .,.. .._._ .:. 1 BF 2-3 kHz ;::::;y;::::_: :;.: :.._ ,. ii’l’iiii.ii’:i’~ ~~~~~~~ SR ANF

.:_. .;... ..:. Pri, PN ChS, ChT

11 I 0 120 0 120

Sound Level (dB SPL) Sound Level (dB SPL)

0 Low SR ANF 0 Pri 0 Low SR ANF 0 ChS

A Med SR ANF 83 PN 1 0

x High SR ANF

A MedSRANF A ChT

x High SR ANF

BF (kHz) IO

FIG. 18. Comparison of envelope latency between ANFs and primarylike and chopper units at different sound levels and BFs. A : envelope latency is plotted vs. sound level in dB SPL for primarylike units with BF between 2 and 3 kHz. Shaded and outlined areas on background indicate distributions of same functions from low/medium and high SR ANFs, respectively. B: same format as in A for chopper units. C: envelope latencies of primarylike units and ANFs at 50 dB SPL are plotted vs. unit BF. D: same format as in C for chopper units. Symbols of unit types are indicated on each plot.

gence of off-BF ANF inputs, 3) the effect of temporal integration, and 4) the role of inhibitory inputs. Quantitative statements about the effect of these mechanisms on envelope modulation are, at least to some extent, model depen- dent. In a future paper (now in preparation) we will present a detailed model analysis of the issues.

CONVERGENCE OF INPUTS. Convergence of excitatory ANFs with the same BFs on an AVCN cell can be of three kinds: 1) all high SR ANFs, 2) all low/medium SR ANFs, and 3) both high and low/medium SR ANFs. Although not much is known about how ANFs converge on AVCN cells, there is strong evidence that the convergence is a common feature in most AVCN cells. Spherical bushy cells, the source of Pri response type, were shown to receive one or a few ANFs on their somata (Cant and Morest 1979b). Evi- dence suggests that ANFs from different SR groups of ANFs tend not to converge on the same spherical bushy cell (Liberman 199 1; Ryugo and Sento 199 1). The number of ANFs synapsing on the somata of globular bushy cells, from which PN units are recorded, was recently suggested to be much larger ( -40) than previously thought, and they may include both low/medium and high SR ANFs (Liber- man 199 1). Stellate cells, which give chopper responses, have been shown to receive ANF inputs on or near cell bodies as well as on distal dendrites (Liberman 199 1; Ryugo et al. 1993; Smith and Rhode 1989). Both low/medium and high SR ANFs were shown to synapse on stellate cells (Ryugo et al. 1993), with the number of input ANFs at

the somata estimated at - 10 (Liberman 199 1). There has been no direct evidence regarding the convergence of inputs from two SR groups of ANFs on the same stellate cell.

To separate the effect of convergence from other neural mechanisms, we assume in our discussions below that all input spikes are suprathreshold, i.e., that each individual spike is able to drive the postsynaptic cell. The situation when input spikes are subthreshold will be discussed later. If every input spike results in an output spike, then it is easy to show that the modulation depth of the output spike train will be no greater than the largest modulation depth among the converging inputs. In any event, convergence cannot therefore explain the gain in modulation depth illustrated by the data in Figs. 3, 6, and 9, where output modulation depth of some AVCN units is greater than the largest ANF modulation depth.

Although the convergence of excitatory inputs does not increase modulation depth, a normalized measure, in an AVCN cell over that of its inputs, it does increase nonnormalized measures like envelope peak height and envelope minimum. The magnitude of nonnormalized measures of output response should increase with the number of converging inputs, though not necessarily linearly, until it is limited by refractoriness or other saturation mechanisms of discharge rate in an AVCN cell. If all converging inputs come from one SR group, and have similar BF thresholds, the increase in envelope peak height relative to that in the input should be the same at all sound levels if the cell were a linear system. For AVCN cells, which are clearly nonlinear,


such convergence should generate less increment at high than at low sound levels because rate saturation and/or refractory properties are likely to reduce envelope peak height at higher sound levels. On the other hand, if a cell receives inputs from both high and low/medium SR ANFs, the number of activated inputs will increase as sound level increases because of recruitment of low SR/ high threshold ANFs. It was shown in our earlier report (Wang and Sachs 1993) that low/medium SR ANFs maintain higher envelope peak height at high sound levels in dB SPL than do high SR ANFs. Thus the convergence of both high and low SR ANFs can result in a larger increment in envelope peak height at high sound levels than at low sound levels above unit threshold. In fact, comparison between chopper units and ANFs shows that there is a larger increase in the envelope peak height at 70 dB re threshold (mean envelope peak heights: ChS/ChT = 487 spikes/s, ANF = 215 spikes/s; ratio of 2.3) than at 30 dB re threshold (mean envelope peak heights: ChS/ChT = 569 spikes/s, ANF = 407 spikes/ s; ratio of 1.4). Such a difference in envelope peak height is also seen in other AVCN units, except Pri type. This evidence suggests that convergence of low/ medium and high SR ANFs may occur in most AVCN units. Nonetheless, convergence of suprathreshold inputs above can generate modulation depths no higher than that of the low/medium SR ANFs.

As we showed in our earlier report, ANFs maintain con- siderably higher modulation depth at high sound levels if the stimulus is centered away from unit BF (Wang and Sachs 1993). In some cases, the modulation depth induced by off-BF stimuli can be significantly higher than even that of low/medium SR ANFs at high sound levels. The off-BF excitatory inputs thus provide potential sources of modulation at high sound levels. It is conceivable that an AVCN unit can achieve enhanced modulation depth by receiving excitatory off-BF ANFs. In fact, there is evidence that some AVCN cells may receive excitatory off-BF ANFs (Carney 1990; Young and Sachs 1988).

TEMPORAL INTEGRATION. We now consider the effects on envelope modulation of subthreshold ANF inputs, i.e., inputs that generate individual EPSPs too small to produce an output spike. In this case, temporal integration of EPSPs from one or more inputs is needed to produce an output spike. In vitro studies have shown that temporal integration of subthreshold EPSPs occurs in stellate cells but does not occur in bushy cells ( Oertel 1983; Oertel et al. 1988 ) . This finding is consistent with the observation that phase-locking to BF tones is preserved in primarylike units but de- grades in chopper units (Blackburn and Sachs 1989; Wang 1991).

If input ANFs produce subthreshold EPSPs, the input discharge rate, summed over all inputs, must be higher than a threshold value to drive an AVCN cell to discharge. The magnitude of this threshold value depends on properties of the spike generator of an AVCN cell. The effect of such a threshold is to shift the output rate versus input rate function to the right; the threshold effect is thus a reduction in output rate at any input rate (ignoring saturation effects). When discharge patterns of the ANF inputs are envelope- modulated, both the envelope minimum and maximum of

the output will be reduced. This reduction can occur with- out a reduction in envelope peak height unless envelope minimum of the input is 0. Such a threshold effect could thus result in an increase in modulation depth in an AVCN cell. INHIBITORY INPUT. There has been considerable evidence showing the existence of inhibitory inputs to ventral cochlear nucleus (VCN) bushy (Adams and Mugnaini 1987; Altschuler et al. 1986; Saint Marie et al. 1989; Wenthold et al. 1987) and stellate cells (Cant 198 1; Saint Marie et al. 199 1; Wenthold 1987; Wickesberg and Oertel 1988) although the nature of the inhibitions (e.g., source, on- or off-BF, etc.) is largely unknown. Inhibitory membrane potentials are frequently seen in intracellular recordings of VCN cells both in vivo (Smith and Rhode 1989) and in vitro (Oertel 1983; Wu and Oertel 1984). Inhibitory inputs can effectively increase discharge threshold of an AVCN cell, reduce the envelope minimum, and thus increase modulation depth. One prediction is that the strength of inhibition should be proportional to sound level, at least in those units that showed extremely high modulation depth at high sound levels. Such inhibitions could be driven by on-BF/ low SR or off-BF ANFs. In fact, there are some anatomic substrates that may support such configurations. For example, Liberman ( 199 1) recently showed that the small cell cap in the AVCN is innervated exclusively by low/medium SR ANFs. These small cells have the potential to give AVCN cells inhibitory inputs that are stronger at high than at low sound levels.

Implicationsfiv signal processing mechanisms in dQ’@rent A VCN cell types

Results of the present study, taken together with available anatomic evidence and studies using pure tone stimuli, offer us some clues to the information processing circuitry at the AVCN. We discuss below some implications of the findings and analysis above for signal processing mechanisms in AVCN major cell types. PRI UNITS. Pri units have been shown to exhibit phase- locking to BF tones similar to that of ANFs (Blackburn and Sachs 1989; Bourk 1976) and to be innervated by one or a few ANFs synapsing on their somata through the endbulb of Held (Liberman 199 1; Lorente de No 198 1; Ryugo and Sento 199 1). This evidence suggests that Pri units receive suprathreshold ANF inputs, because convergence of a small number of subthreshold inputs will certainly result in a reduction in BF phase-locking because of the effect of temporal integration (Oertel et al. 1988). As we have discussed above, such a configuration will lead to no increase in modulation depth and, depending on the number of the converging inputs, equivalent or increased envelope peak height and minimum. In fact, our data showed that Pri units did not exhibit much change in their modulation depth compared with all ANFs (Figs. 3 and 11). There is only small elevation of envelope peak height (Fig. 13 D) and envelope minimum (Fig. 14A) in Pri units over the average values of ANFs. Moreover, average discharge rates of Pri units in response to SFS are comparable to those of ANFs (Fig. 19 A). Observations from responses to SFS are consistent with the notion that Pri units are predominantly


A 1.0 7

0.8 -

0.6 -

o LowNed SR ANF

k 1.0 A b cn B 000 $ 0.8 @ 0

g 0

% 0.6 A

0 ChS A ChT

1.0 o- l cl ooo+ l 8 .’

O 0 0.8 $ $ 00 00

0 +o 0 ?I00

0.6 “XV&

X xxx & X % ox x 0.4 x x

s xx x

X xx zF*

+ OnC 0.2 X WC 0 On

Average Discharge Rate at 80 dB SPL (spikes/set)

FIG. 19. Modulation depth is plotted vs. average discharge rate, both measured at 80 dB SPL, for ANFs and AVCN units from responses to SFS. A-C: comparison between ANFs and primarylike, chopper, and onset units, respectively. Symbols for low/medium (0) and high ( X ) SR ANFs are same in A-C. Symbols for AVCN units are indicated on each plot.

influenced by a small number of suprathreshold ANF inputs. PN UNITS. Evidence indicates that globular bushy cells, the source of PN responses, are innervated by a large number of ANFs ( - 17, Spirou et al. 1990; -40, Liberman 199 1). To achieve the observed BF phase-locking, which is similar to that of ANFs (Blackburn and Sachs 1989; Bourk 1976), the ANF inputs should be either suprathreshold or subthreshold but with nearly identical BFs to allow spatial summa- tion (Oertel et al. 1988). The argument against subthreshold inputs is that they will produce outputs that are more regular than those seen in real PN responses (Young et al. 1993). For responses to SFS, convergence of suprathreshold inputs results in both increased envelope minimum and peak height, as we discussed above. Our data indicate that PN units have higher envelope peak height (Fig. 13 D) but lower envelope minimum (Fig. 14C) at high sound level

than those of ANFs. This suggests that if PN units receive suprathreshold excitatory inputs, they must receive inhibition as well to reduce the envelope minimum. It is possible that ANF inputs to a PN unit are a mix of both supra- and subthreshold events, so that its outputs would not be too regular and its envelope minimum would not be too high. CH-S AND CH-T UNITS. Because of the degradation seen in BF phase-locking in chopper units (Blackburn and Sachs 1989; Bourk 1976), their inputs are unlikely to be dominated by suprathreshold ANF inputs at or near somata. On the other hand, inputs on distal dendrites are subject to attenuation by low-pass filtering (Young et al. 1988b) and are thus likely subthreshold when they reach somata. As a result, temporal integration is necessary to bring a chopper unit to discharge. There is evidence that the membrane characteristics of stellate cells, from which chopper responses are recorded, supports temporal integration (Oertef 198 3 ) . The available data also suggest convergence of multiple ANFs on stellate cells (Liberman 199 1). In response to SFS, many chopper units show a significant increase in envelope peak height (Figs. 12, C and D, and 13, B and E), suggesting a large number of converging inputs. However, data show that envelope minimum in chopper units, on average, is lower rather than higher than that of ANFs at most sound levels (Fig. 1X?), as would be predicted from the convergence. Both temporal integration and inhibition could contribute to the reduced envelope minimum.

An interesting characteristic of chopper responses to SFS, in contrast to primarylike units, is that the difference in envelope latency between chopper units and ANFs is larger at low levels than at high sound levels (Fig. 18 B). The delay between temporal discharge patterns of ANFs and AVCN units can result from synaptic delay, dendritic conduction time, and postsynaptic integration time. Among these, the synaptic delay is likely to remain relatively constant at different sound levels. A longer envelope latency at low sound level suggests two possibilities: 1) inputs to chopper units are dominated by those inputs further away from the somata at low sound levels and by those closer to the somata at high sound levels (e.g., low SR/ high threshold ANFs), so that conduction time is shorter at high sound levels, which leads to a shorter envelope latency; and 2) the integration time is more significant than dendritic conduction time in chopper units, if we assume their inputs are subthreshold. Higher input rate at high sound levels makes temporal integration occur faster and thus leads to a shorter envelope latency. ON-C UNITS. Among all AVCN types, OnC units showed the largest increment in envelope peak height over that of ANFs (Fig. 12 E), an indication of convergence of significant number of ANFs. The envelope minimum in OnC units is lower than that of ANFs (Fig. 16C). In contrast to Pri and chopper units, average discharge rates of many OnC units are lower than that of high SR ANFs at high sound level (Fig. 1 SC). These observations suggest a stronger influ- ence of inhibition in OnC units than in chopper units.

Comparison with previous studies

The SFSs used in this study are similar to AM stimuli in that both have amplitude modulation. However, an SFS has a series of sidebands surrounding the carrier frequency


component, whereas an AM stimulus has only one side- band on each side of the carrier. This difference may affect a CN unit’s response property if it integrates off-BF excitatory or inhibitory ANF inputs. Study of the encoding of envelope modulation in AM or similar stimuli by CN cells was initiated more than two decades ago (e.g., Glattke 1969; Mailer 1972a, 1976) and conducted more recently with better knowledge of recording locations within CN divisions and identification of physiological response types of the recorded units (AM: Frisina et al. 1985, 1990a; Kim et al. 1990; two-tone complex: Greenberg and Rhode 1987). Previous studies with AM sounds found that modulation transfer functions of most CN units, except for Pri type, have band-pass characteristics at moderate to high sound levels (Frisina et al. 1990a; Kim et al. 1990; Molller 1972b), compared with the low-pass characteristics of ANF (Joris and Yin 1992; Merller 1976; Palmer 1982). The peaks of band-pass modulation transfer functions were found in the range of SO-520 Hz in VCN of anesthetized gerbils (Frisina et al. 1990a) and of 50-500 Hz in PVCN and DCN of decerebrate cats (Kim et al. 1990). The fundamental frequency of the SFSs used in the present study ( - 125 Hz) is within the range of frequencies where AM sounds were shown to generate strong phase-locked responses in VCN units.

In the study of Frisina et al. ( 1990a), On-L units, a sub- type of onset category (Rhode and Smith 1986) and similar to our On type, showed the strongest enhancement in modulation gain compared with that of ANFs, followed by chopper, PN, and Pri units. The above relationship was based primarily on characterization of modulation by the magnitude of the fundamental component in the spectra of PSTHs at 50 dB re threshold. The results of the present study agree with their observations that Pri units show the least amount of modulation and On-L units show the most. However, our data showed little enhancement of modulation depth over ANFs in Pri units with respect to sound level re threshold. In addition, by separating units with “chopping” PSTH into three subtypes (ChS, ChT, and One), we found that ChS units show less enhancement than do PN units, which have as much enhancement as ChT units at moderate to high sound levels, whereas OnC units show greater modulation enhancement than do both ChT and PN units. The “chopper” category in the study of Frisina et al. ( 1990a) includes ChS, ChT, and possibly OnC types. This, as well as the fact that some choppers were recorded from PVCN, may explain why the chopper category was found to have higher modulation gain than the PN category in their study. OnC is a predominant response type in PVCN (Rhode and Smith 1986)) but is much less frequently recorded in the AVCN. Kim et al. ( 1990) observed higher maximum vector strength (a measure similar to the& synchronization index) at high sound levels (60-85 dB SPL) in PVCN and DCN units of decerebrate cats than in ANFs.

Greenberg and Rhode ( 1987) studied responses of cat VCN neurons to a two-tone complex and suggested that chopper and On units function as envelope and pitch period encoders, respectively. In the present study, we showed

This evidence suggests an important role of PN units as carriers of envelope or periodicity information to higher centers. PN units also are known to encode fine structures of tones or complex stimuli (Blackburn and Sachs 1989, 1990; Bourk 1976). In this sense, PN units are better encoders of temporal information than all other AVCN types.

Frisina et al. ( 1990b) suggested that two mechanisms may be responsible for the amplification of modulation in ANF discharge patterns; one mechanism would result in lowering the average response to AM stimuli, and the other would cause an increase in the fundamental component of the response. We found that AVCN units can have comparable, increased, or decreased average discharge rate in response to SFS, compared with ANFs, and yet achieve enhanced envelope modulation. In Fig. 19, modulation depth is plotted versus average discharge rate for ANFs and AVCN units with BFs between 3 and 10 kHz, both measured from response to SFS at 80 dB SPL. As we can see from Fig. 19A, the mean of average discharge rates of ANFs is between 150 and 200 spikes/s for both low/medium and high SR groups. For the majority of PN, ChS, and ChT units, their average discharge rates are comparable to or higher than those of ANFs, whereas their modulation depths are comparable to or higher than that of low/medium SR ANFs. Only On and OnC units show lowered average discharge rate at high modulation depth. There is a tendency for some units with higher average discharge rate to show lower modulation depth, but the correlation is weak within each AVCN type. In the responses to AM stimuli, the change in transient response (e.g., measured by modulation depth) as a function of sound level can be very different from that in the average response (e.g., measured by average discharge rate), as was shown in ANF (Smith and Brachman 1980). It is thus more appropriate to state that modulation enhancement is due to dynamic adjust- ment by AVCN cells (i.e., changes in the response envelope), which may, but do not necessarily, result in the reduction in average response rate.

Implications for other auditory functions

The finding of enhancement of envelope modulation in AVCN cells supports the notion that periodicity pitch of complex sounds with missing fundamentals is extracted by the central systems from neural discharge patterns origi- nated from ANFs (de Boer 1976; Plomp 1976) although there is an alternative spectral mechanism at low frequencies (Goldstein 1973; Srulovicz and Goldstein 1983; Ter- hardt 1972; Wightman 1973) a The use of a temporal code of envelope information by the CNS at high frequencies, where frequency components are not resolved at the auditory periphery, has been suggested by a number of psycho- physical studies. For example, envelope fluctuation in the stimulus waveform was shown to affect the pitch perceived from AM wide-band noise (Burns and Viemeister 1976) and harmomc complexes centered at high frequency (Houtsma and Smurzynski 1990). Another possible usage of the envelope cues is in sound localization and lateralization at high frequency. It has been shown that listeners can detect interaural time delays conveyed via the envelope of

that PN units can code envelope modulation just as well as the stimuli centered at high frequencies (Henning 1974; ChS and ChT units, or even better in some cases (Fig. 11). McFadden and Pasanen 1976; Nuetzel and Hafter 1976).


Obviously, for the auditory system to use temporal codes of envelope information, they need be preserved, further enhanced, or even separated from other auditory information. Such processing seems to take place at higher centers beyond the CN [see Langner ( 1992) for review]. There is evidence that some neurons in the inferior colliculus are specialized in representing stimulus periodicity reflected by the envelope of AM sounds (Langner and Schreiner 1988; Schreiner and Langner 1988 ) . Sensitivity to interaural time difference of the envelopes of AM sounds from two ears was found in lateral superior olive (LSO) cells (Caird and Klinke 1983; Joris et al. 1990), which receive inputs from AVCN globular bushy cells (the source of PN responses), and in high-BF medial superior olive (MSO) cells (Yin and Chan 1990) which are binaurally innervated by AVCN spherical bushy cells (the source of Pri responses). Further- more, cells in the inferior colliculus were shown to be sensi- tive to the interaural time difference of the envelopes of AM stimuli centered at high frequency (Batra et al. 1989; Yin et al. 1984).

The authors thank colleagues in the Johns Hopkins Center for Hearing Sciences, Drs. C. Schreiner, P. Leake, and W. Jenkins for helpful com- ments on the manuscripts, and P. Taylor for graphical assistance.

This research was funded by National Institute of Deafness and Other Communications Disorders Grants DC-00 109 and DC-00979 and in part by the W. M. Keck Foundation.

Present address and address for reprint requests: X. Wang, Coleman Lab., University of California at San Francisco, Box 0732, San Francisco, CA 94 143-0732.

Received 11 March 1993; accepted in final form 3 1 August 1993.

REFERENCES

ADAMS, J. C. AND MUGNAINI, E. Patterns of glutamate decarboxylase im- munostaining in the feline cochlear complex studied with silver enhancement and electron microscopy. J. Comp. Neural. 262: 375-401, 1987.

ALTSCHULER, R. A., BETZ, H., PARAKKAL, M. H., REEKS, K. A., AND WENTHOLD, R. J. Identification of glycinergic synapses in the cochlear nucleus through immunocytochemical localization of the postsynaptic receptor. Brain Res. 369: 3 16-320, 1986.

BATRA, R., KUWADA, S., AND STANFORD, T. R. Temporal coding of envelopes and their interaural delays in the inferior colliculus of the unanesthetized rabbit. J. Neurophysiol. 6 1: 257-268, 1989.

BLACKBURN, C. C. AND SACHS, M. B. Classification of unit types in the anteroventral cochlear nucleus: poststimulus time histograms and regularity analysis. J. Neurophysiol. 62: 1303- 1329, 1989.

BLACKBURN, C. C. AND SACHS, M. B. The representations of the steady- state vowel sound /E/ in the discharge patterns of cat anteroventral cochlear nucleus neurons. J. Neurophysiol. 63: 119 l-l 2 12, 1990.

BOER, E. DE. On the residue and auditory pitch perception. In: Handbook of Sensory Physiology. Auditory System. Clinical and Special Topics, edited by W. D. Keidel and W. D. Neff. New York: Springer-Verlag, 1976, vol. V, pt. 3, p. 479-583.

BOURK, T. R. Electrical Responses of Neural hits in the Anteroventral Cochlear Nucleus of the Cat (PhD thesis). Boston, MA: Massachusetts Inst. of Technology, 1976.

BRAWER, J. R. AND MOREST, D. K. Relations between auditory nerve endings and cell types in the cat’s anteroventral cochlear nucleus seen with the Golgi method and Nomarski optics. J. Comp. Neural. 160: 491-506, 1975.

BRAWER, J. R., MOREST, D. K., AND KANE, E. C. The neuronal architecture of the cochlear nucleus of the cat. J. Comp. Neural. 155: 25 l-300, 1974.

BURNS, E. M. AND VIEMEISTER, N. F. Nonspectral pitch. J. Acoust. Sot. Am. 60: 863-869, 1976.

CAIRD, D. AND KLINKE, R. Processing of binaural stimuli by cat superior olivary complex neurons. Exp. Brain Res. 52: 385-399, 1983.

CANT, N. B. The fine structure of two types of stellate cells in the anterior division of the anteroventral cochlear nucleus of the cat. Neuroscience 6: 2643-2655,198l.

CANT, N. B. AND MOREST, D. K. Organization of the neurons in the anterior division of the anteroventral cochlear nucleus of the cat: light-mi- croscopic observations. Neuroscience 4: 1909- 1923, 1979a.

CANT, N. B. AND MOREST, D. K. The bushy cells in the anteroventral cochlear nucleus of the cat: a study with the electron microscope. Neuro- science 4: 1925- 1945, 1979b.

CANT, N. B. AND MOREST, D. K. The structural basis for stimulus coding in the cochlear nucleus of the cat. In: Hearing Science, edited by C. I. Berlin. San Diego, CA: College-Hill, 1984, p. 37 l-422.

CARNEY, L. H. Sensitivities of cells in anteroventral cochlear nucleus of cat to spatiotemporal discharge patterns across primary afferents. J. Neuro- physiol. 64: 437-456, 1990.

DUNN, H. K. Methods of measuring vowel formant bandwidths. J. Acoust. Soc.Am.33: 1737-1746,1961.

EVANS, E. F. AND NELSON, P. G. The responses of single neurones in the cochlear nucleus of the cat as a function of their location and anesthetic state. Exp. Brain Res. 17: 402-427, 1973a.

EVANS, E. F. AND NELSON, P. G. On the functional relationship between the dorsal and ventral divisions of the cochlear nucleus of the cat. Exp. Brain Res. 17: 428-442, 1973b.

FEKETE, D. M., ROUILLER, E. M., LIBERMAN, M. C., AND RYUGO, D. K. The central projections of intracellularly labeled auditory nerve fibers in cats. J. Comp. Neural. 229: 432-450, 1984.

FLANAGAN, J. Speech Analysis, Synthesis and Perception. New York: Springer-Verlag, 1972.

FRISINA, R. D., SMITH, R. L., AND CHAMBERLAIN, S. C. Differential encoding of rapid changes in sound amplitude by second-order auditory neurons. Exp. Brain Res. 60: 4 17-422, 1985.

FRISINA, R. D., SMITH, R. L., AND CHAMBERLAIN, S. C. Encoding of amplitude modulation in the gerbil cochlear nucleus. I. A hierarchy of enhancement. Hear. Res. 44: 99-122, 1990a.

FRISINA, R. D., SMITH, R. L., AND CHAMBERLAIN, S. C. Encoding of amplitude modulation in the gerbil cochlear nucleus. II. Possible neural mechanisms. Hear. Res. 44: 123-142, 1990b.

GLATTKE, T. J. Unit responses of the cat cochlear nucleus to amplitude- modulated stimuli. J. Acoust. Sot. Am. 45: 4 19-425, 1969.

GOLDSTEIN, J. L. An optimum processor theory for the central formation of the pitch of complex tones. J. Acoust. Sot. Am. 54: 1496- 15 16, 1973.

GREENBERG, S. AND RHODE, W. S. Periodicity coding in cochlear nerve and ventral cochlear nucleus. In: Auditory Processing of Complex Sounds, edited by W. A. Yost and C. S. Watson. New York: Hillsdale, 1987, p. 225-236.

GUINAN, J. J. AND PEAKE, W. T. Middle-ear characteristics of anesthetized cats. J. Acoust. Sot. Am. 41: 1237-1261, 1967.

HENNING, G. B. Detectability of interaural delay in high-frequency complex waveforms. J. Acoust. Sot. Am. 55: 84-90, 1974.

HIRSCH, J. A. AND OERTEL, D. Intrinsic properties of neurons in the dorsal cochlear nucleus of mice, in vitro. J. Physiol. Land. 396: 535-548, 1988.

HOUTSMA, A. J. M. AND SMURZYNSKI, J. Pitch identification and discrimi- nation for complex tones with many harmonics. J. Acoust. Sot. Am. 87: 304-310, 1990.

JORIS, P. X. AND YIN, T. C. T. Responses to amplitude-modulated tones in the auditory nerve of the cat. J. Acoust. Sot. Am. 9 1: 2 15-232, 1992.

JORIS, P. X., YIN, T. C. T., AND SMITH, P. H. Mechanisms of azimuthal sound localization in the central nervous system of the cat. Nederlands Akoestisch Genootschap 104: 23-35, 1990.

KIANG, N. Y. S., MOREST, D. K., GODFREY, D. A., GUINAN, J. J., AND KANE, E. C. Stimulus coding at caudal levels of the cat’s auditory nervous system. I. Response characteristics of single units. In: Basic Mecha- nisms in Hearing, edited by A. R. Moller. New York: Academic, 1973, p. 455-478.

KIANG, N. Y. S., MOXON, E. C., AND LEVINE, R. A. Auditory-nerve activ- ity in cats with normal and abnormal cochleas. In: Sensorineural Hear- ing Loss, edited by G. E. W. Wolstenholme and J. Knight. London: Churchill, 1970, p. 24 l-272.

KIM, D. O., SIRIANNI, J. G., AND CHANG, S. 0. Responses of DCN-PVCN neurons and auditory nerve fibers in unanesthetized decerebrate cats to AM and pure tones: analysis with autocorrelation / power-spectrum. Hear. Res. 45: 95-l 14, 1990.

LANGNER, G. Periodicity coding in the auditory system. Hear. Res. 60: 115~142,1992.

LANGNER, G. AND SCHREINER, C. E. Periodicity coding in the inferior colliculus of the cat. I. Neuronal mechanisms. J. Neurophysiol. 60: 1799-1822,1988.

LIBERMAN, M. C. Auditory-nerve responses from cats raised in a low-noise chamber. J. Acoust. Sot. Am. 63: 442-455, 1978.


LIBERMAN, M. C. Central projections of auditory-nerve fibers of differing spontaneous rate. I. Anteroventral cochlear nucleus. J. Comp. Neural. 313: 240-258, 1991.

LORENTE DE No, R. The Primary Acoustic Nuclei. New York: Raven, 1981.

MCFADDEN, D. AND PASANEN, E. G. Lateralization at high frequencies based on interaural time differences. J. Acoust. Sot. Am. 59: 634-639, 1976.

MBLLER, A. R. Coding of amplitude and frequency modulated sounds in the cochlear nucleus of the rat. Acta Physiol. Stand. 86: 223-238, 1972a.

MBLLER, A. R. Coding of sounds in lower levels of the auditory system. Q. Rev. Biophys. 5: 59-155, 1972b.

MBLLER, A. R. Dynamic properties of primary auditory fibers compared with cells in the cochlear nucleus. Acta Physiol. Stand. 98: 157-167, 1976.

MOREST, D. K., KIANG, N. Y. S., KANE, E. C., GUINAN, J. J., AND GOD- FREY, D. A. Stimulus coding at caudal levels of the cat’s auditory nervous system. II. Patterns of synaptic organization. In: Basic Mechanisms in Hearing, edited by A. R. Moller. New York: Academic, 1973, p. 479-509.

NUETZEL, J. M. AND HAFTER, E. R. Lateralization of complex waveforms: effects of fine structure, amplitude, and duration. J. Acoust. Sot. Am. 60:1339-1346,1976.

OERTEL, D. Synaptic responses and electrical properties of cells in brain slices of the mouse anteroventral cochlear nucleus. J. Neuroscience 3: 2043-2053,1983.

OERTEL, D., WV, S. H., AND HIRSCH, J. A. Electrical characteristics of cells and neuronal circuitry in the cochlear nuclei studied with intracellular recording from brain slices. In: Auditory Function, edited by G. M. Edel- man, W. E. Gall, and W. M. Cowan. New York: Wiley, 1988, p. 3 13- 336.

PALMER, A. R. Encoding of rapid amplitude fluctuations by cochlear- nerve fibres in the guinea pig. Arch. Otolaryngol. 236: 197-202, 1982.

PFEIFFER, R. R. Classification of response patterns of spike discharges for units in the cochlear nucleus: tone burst stimulation. Exp. Brain Res. 1: 220-235,1966.

PLOMP, R. Aspects of Tone Sensation. London: Academic, 1976. POLJAK, S. The connections of the acoustic nerve. J. Anat. 60: 465-469,

1926. RHODE, W. S. AND SMITH, P. H. Encoding timing and intensity in the

ventral cochlear nucleus of the cat. J. Neurophysiol. 56: 26 l-286, 1986. ROUILLER, E. M., CRONIN-SCHREIBER, R., FEKETE, D. M., AND RYUGO,

D. K. The central projections of intracellularly labeled auditory nerve fibers in cats: an analysis of terminal morphology. J. Clomp. Neural. 249: 261-278,1986.

ROUILLER, E. M. AND RYUGO, D. K. Intracellular marking of physiologically characterized cells in the ventral cochlear nucleus of the cat. J. Comp. Neurol. 225: 167-186, 1984.

RYUGO, D. K. AND FEKETE, D. M. Morphology of primary axosomatic endings in the anteroventral cochlear nucleus of the cat: a study of the endbulbs of Held. J. Comp. Neural. 210: 239-257, 1982.

RYUGO, D. K. AND SENTO, S. Synaptic connections of the auditory nerve in cats: relationship between endbulbs of Held and spherical bushy cells. J. Comp. Neurol. 305: 35-48, 199 1.

RYUGO, D. K., WRIGHT, D. D., AND PONGSTAPORN, T. Ultrastructural analysis of synaptic endings of auditory nerve fibers in cats: correlations with spontaneous discharge rate. In: The Mammalian Cochlear Nuclei: Organization and Function, edited by M. A. Merchan, J. M. Juiz, D. A. Godfrey, and E. Mugnaini. New York: Plenum, 1993, p. 65-74.

SACHS, M. B. AND YOUNG, E. D. Encoding of steady-state vowels in the auditory nerve: representation in terms of discharge rate. J. Acoust. Sot. Am. 66: 470-479, 1979.

SAINT MARIE, R. L., BENSON, C. G., OSTAPOFF, E.-M., AND MOREST, D. K. Glycine immunoreactive projections from the dorsal to the anteroventral cochlear nucleus. Hear. Res. 5 1: 1 l- 18, 199 1.

SAINT MARIE, R. L., MOREST, D. K., AND BRANDON, C. J. The form and distribution of GABAergic synapses on the principal cell types of the ventral cochlear nucleus of the cat. Hear. Res. 42: 97- 112, 1989.

SCHALK, T. B. AND SACHS, M. B. Nonlinearities in auditory-nerve fiber responses to bandlimited noise. J. Acoust. Sot. Am. 67: 903-9 13, 1980.

SCHREINER, C. E. AND LANGNER, G. Periodicity coding in the inferior

colliculus of the cat. II. Topographical organization. J. Neurophysiol. 60:1823-1840,1988.

SHOFNER, W. P. AND YOUNG, E. D. Excitatory/inhibitory response types in the cochlear nucleus: discharge patterns and responses to electrical stimulation of the auditory nerve. J. Neurophysiol. 54: 9 17-939, 1985.

SMITH, P. H. AND RHODE, W. S. Structural and functional properties dis- tinguish two types of multipolar cells in the ventral cochlear nucleus. J. Comp. Neural. 282: 595-616, 1989.

SMITH, R. L. AND BRACHMAN, M. L. Response modulation of auditory- nerve fibers by AM stimuli: effects of average intensity. Hear. Res. 2: 123-133,198O.

SOKOLICH, W. G. Improved acoustic system for auditory research. J. Acoust. Sot. Am. 62 Suppl. 1: S 12, 1977.

SPIROU, G. A., BROWNELL, W. E., AND ZIDANIC, M. Recordings from cat trapezoid body and HRP labeling of globular bushy cell axons. J. Neuro- physiol. 63: 1169- 1190, 1990.

SRULOVICZ, P. AND GOLDSTEIN, J. L. A central spectrum model: a synthesis of auditory-nerve timing and place cues in monaural communication of frequency spectrum. J. Acoust. Sot. Am. 73: 1266-1276, 1983.

TERHARDT, E. Zur Tonhohenwahrnehmung von Klangen. II. Ein Funk- tionsschema. Acustica 26: 187- 199, 1972.

TOLBERT, L. P. AND MOREST, D. K. The neuronal architecture of the antero-ventral cochlear nucleus of the cat in the region of the cochlear nerve root: electron microscopy. Neuroscience 7: 3053-3067, 1982.

WANG, X. Neural Encoding of Single-formant Stimuli in the Auditory- Nerve and Anteroventral Cochlear Nucleus ofthe Cat ( PhD thesis). Bal- timore, MD: Johns Hopkins Univ., 199 1.

WANG, X. AND SACHS, M. B. Neural encoding of single-formant stimuli in the cat. I. Responses of auditory-nerve fibers. J. Neurophysiol. 70: 1054- 1075,1993.

WENTHOLD, R. J. Evidence for glycinergic pathway connecting the two cochlear nuclei: an immunocytochemical and retrograde transport study. Brain Res. 415: 183-187, 1987.

WENTHOLD, R. J., HUIE, D., ALTSCHULER, R. A., AND REEKS, K. A. Gly- tine immunoreactive localized in the cochlear nucleus and superior olivary complex. Neuroscience 22: 897-9 12, 1987.

WICKESBERG, R. E. AND OERTEL, D. Tonotopic projection from the dorsal to the anteroventral cochlear nucleus of mice. J. Comp. Neural. 268: 389-399,1988.

WIGHTMAN, F. L. The pattern-transformation model of pitch. J. Acoust. Sot. Am. 54: 407-416, 1973.

WINSLOW, R. L., BARTA, P. E., AND SACHS, M. B. Rate coding in the auditory nerve. In: Auditory Processing of Complex Sounds, edited by W. A. Yost and C. S. Watson. Hillsdale, NJ: Erlbaum, 1987, p. 2 12-224.

WINSLOW, R. L. AND SACHS, M. B. Effect of electrical stimulation of the crossed olivocochlear bundle on auditory nerve response to tones in noise. J. Neurophysiol. 57: 1002- 102 1, 1987.

WV, S. H. AND OERTEL, D. Intracellular injection with horseradish peroxi- dase of physiologically characterized stellate and bushy cells in slices of mouse anteroventral cochlear nucleus. J. Neurosci. 4: 1577- 1588, 1984.

YIN, T. C. T. AND CHAN, J. C. K. Interaural time sensitivity in medial superior olive of cat. J. Neurophysiol. 64: 465-488, 1990.

YIN, T. C. T., KUWADA, S., AND SUJAKU, Y. Interaural time sensitivity of high-frequency neurons in the inferior colliculus. J. Acoust. Sot. Am. 76: 1401-1410, 1984.

YOUNG, E. D. Response characteristics of neurons in the cochlear nuclei. In: Hearing Science, edited by C. I. Berlin. San Diego, CA: College-Hill, 1984, p. 423-460.

YOUNG, E. D., ROBERT, J. M., AND SHOFNER, W. P. Regularity and latency of units in the ventral cochlear nucleus: implications for unit classification and generation of response properties. J. Neurophysiol. 60: l-29, 1988a.

YOUNG, E. D., ROTHMAN, J. S., AND MANIS, P. B. Regularity ofdischarge constrains models of ventral cochlear nucleus bushy cells. In: The Mam- malian Cochlear Nuclei: Organization and Function, edited by M. A. Merchan, J. M. Juiz, D. A. Godfrey, and E. Mugnaini. New York: Ple- num, 1993, p. 395-410.

YOUNG, E. D. AND SACHS, M. B. Interactions of auditory nerve fibers and cochlear nucleus cells studied with cross-correlation. Sot. Neurosci. Abstr. 14: 646, 1988.

YOUNG, E. D., SHOFNER, W. P., WHITE, J. A., ROBERT, J. M., AND VOIGT, H. F. Response properties of cochlear nucleus neurons in relationship to physiological mechanisms. In: Auditory Function, edited by G. M. Edel- man, W. E. Gall, and W. M. Cowan. New York: Wiley, 1988b, p. 277- 312.

Neural Encoding of Single-Formant Stimuli in the Cat. II ...

Documents

Transcript of Neural Encoding of Single-Formant Stimuli in the Cat. II ...