Subcortical neural coding mechanisms for auditory temporal processing

Hearing Research 158 (2001) 1^27 www.elsevier.com/locate/heares

Review

Subcortical neural coding mechanisms for auditory temporal processing Robert D. Frisina * Otolaryngology Division, Surgery, NeurobiologypAnatomy and Biomedical Engineering Departments, University of Rochester School of Medicine and Dentistry, International Center for Hearing and Speech Research, National Technical Institute for the Deaf, Rochester Institute of Technology, 601 Elmwood Avenue, Rochester, NY 14642-8629, USA Received 12 September 2000; accepted 9 April 2001

Abstract Biologically relevant sounds such as speech, animal vocalizations and music have distinguishing temporal features that are utilized for effective auditory perception. Common temporal features include sound envelope fluctuations, often modeled in the laboratory by amplitude modulation (AM), and starts and stops in ongoing sounds, which are frequently approximated by hearing researchers as gaps between two sounds or are investigated in forward masking experiments. The auditory system has evolved many neural processing mechanisms for encoding important temporal features of sound. Due to rapid progress made in the field of auditory neuroscience in the past three decades, it is not possible to review all progress in this field in a single article. The goal of the present report is to focus on single-unit mechanisms in the mammalian brainstem auditory system for encoding AM and gaps as illustrative examples of how the system encodes key temporal features of sound. This report, following a systems analysis approach, starts with findings in the auditory nerve and proceeds centrally through the cochlear nucleus, superior olivary complex and inferior colliculus. Some general principles can be seen when reviewing this entire field. For example, as one ascends the central auditory system, a neural encoding shift occurs. An emphasis on synchronous responses for temporal coding exists in the auditory periphery, and more reliance on rate coding occurs as one moves centrally. In addition, for AM, modulation transfer functions become more bandpass as the sound level of the signal is raised, but become more lowpass in shape as background noise is added. In many cases, AM coding can actually increase in the presence of background noise. For gap processing or forward masking, coding for gaps changes from a decrease in spike firing rate for neurons of the peripheral auditory system that have sustained response patterns, to an increase in firing rate for more central neurons with transient responses. Lastly, for gaps and forward masking, as one ascends the auditory system, some suppression effects become quite long (echo suppression), and in some stimulus configurations enhancement to a second sound can take place. ß 2001 Elsevier Science B.V. All rights reserved. Key words: Amplitude modulation; Forward masking; Hearing; Deafness ; Aging; Ear

1. Introduction 1.1. Organization of the presentation Dynamic features of biologically relevant complex sounds can be analyzed into temporal and spectral features amenable to scienti¢c investigation. Rapid changes in sound frequency (frequency modulation,

* Tel.: +1 (716) 275-8130; Fax: +1 (716) 271-8552. E-mail address: [email protected] (R.D. Frisina).

FM) are important features for speech perception or auditory processing of animal vocalizations. On the temporal side, singular features, such as stops and starts or gaps in acoustic stimuli, may be distinguished from more periodic or steady state temporal features of sound such as recurring envelope £uctuations. A useful example of sound envelope features that occurs in speech are periodic intensity changes that approximate sinusoidal amplitude modulation (AM). This review focuses on mammalian auditory temporal processing mechanisms at the level of single units (neurons), from systems physiology and sensory neuroscience perspectives. Reviews of evoked potentials in response to temporal features of sound and of FM coding mechanisms await other occasions.

0378-5955 / 01 / $ ^ see front matter ß 2001 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 8 - 5 9 5 5 ( 0 1 ) 0 0 2 9 6 - 9

HEARES 3704 1-8-01

2

R.D. Frisina / Hearing Research 158 (2001) 1^27

1.2. Preview Prior to the pioneering studies of MÖller in the 1970s, little was known about how the auditory nerve or central auditory system processed temporal features of sound. Auditory physiologists focused on the tonotopic or cochleotopic organization of the system, investigated frequency response areas, and determined the abilities of auditory neurons to encode the intensity of simple tones (pure tones) and their combinations. Since then we have learned much, particularly at the levels of the auditory nerve, cochlear nucleus, inferior colliculus (IC) and auditory cortex. Consistent with a systems analysis approach to unraveling the physiological processing capabilities of the auditory system, this review starts with what we know about temporal processing in the auditory nerve, then proceeds to the cochlear nucleus and higher levels of the central auditory system. Surprisingly, little is directly known about temporal processing at the level of the cochlear hair cells. Unfortunately, space does not allow the presentation of every article ever published on the subject at hand, so an apology is made at the outset for any relevant work not presented here. The purpose of the current article is to provide the reader with representative ¢ndings and additional citations to works that they may peruse in detail. For AM coding, one of the major principles to be uncovered is that, as determined by methodologies employed to date, as one ascends the central auditory system, a neural encoding shift occurs. An emphasis on synchronous responses for temporal coding exists in the auditory periphery and cochlear nucleus, and more reliance on rate coding of periodicity occurs as one moves centrally to the auditory midbrain and to the cortex (e.g. Langner, 1992, 1997). On another tack, it is likely that AM processing channels are established at the level of the cochlear nucleus and then proceed through neural pathways that are poorly understood, with AM information ascending through the auditory brainstem to the cortex. At the level of the IC and cortex, this organization has been demonstrated topographically (Langner, 1992). In addition, converging evidence suggests that auditory brainstem temporal processing mechanisms become degraded in cases of auditory neuropathy (Zeng et al., 1999) and with age (Frisina and Frisina, 1997), resulting in signi¢cant impairments in speech comprehension, even in instances when very little peripheral change in auditory sensitivity occurs. Another general principle is that depending on signal and noise levels, in absolute terms and relative to each other, background noise can often enhance the coding of temporal features of sound whereas at other times it degrades them. Before commencing with an examination of similarities and di¡erences in AM processing mechanisms at

di¡erent levels of the auditory system, some explanation of di¡erent metrics used to measure neural AM coding is in order. There have been three main quantitative measures of how single neurons (units) encode AM. The ¢rst is the number of spikes that ¢re in synchrony with the modulating waveform (AM frequency). This is referred to as the synchronous ¢ring rate, and is proportional to the ¢rst Fourier coe¤cient of the Fourier analysis of the AM response. The second is a measure of the proportion of total spikes that are synchronized to the AM. This is commonly referred to as the synchronization coe¤cient (SC), the synchronization index (SI), or the vector strength (r or VS), which is equal to half the percent modulation of the response. It is proportional to the ¢rst Fourier coe¤cient over the average ¢ring rate (DC term of the Fourier series). The SC can vary from 0 to 1, and the percent modulation of the response from 0 to 2. The other main quantitative measure of the synchronous response to AM is the AM response gain. This is a dB scale where the percent modulation of the response is divided by the percent modulation of the stimulus (see Frisina et al., 1996, for equations). For this response gain dB scale, positive values indicate that the AM synchronous response exceeds that of the stimulus, i.e. the AM synchronous response is ampli¢ed relative to the modulation depth of the stimulus. Conversely, a negative AM response gain indicates that the modulation of the spike response of a unit to AM is less than that of the stimulus. 2. Auditory nerve AM responses 2.1. Di¡erent classes of nerve ¢bers To a ¢rst approximation, temporal response patterns of auditory nerve ¢bers are similar, except that the frequencies to which they are most sensitive vary depending upon which part of the basilar membrane they originate from. Based upon thresholds and spontaneous ¢ring rates, mammalian auditory nerve ¢bers have been separated into three classes (Liberman, 1978). Auditory nerve ¢bers with high spontaneous activity rates and low thresholds comprise one class, and conversely, nerve ¢bers with low spontaneous activity rates and higher thresholds make up another class. Auditory nerve ¢bers with intermediate spontaneous ¢ring rates and thresholds are referred to as a `medium' class. This classi¢cation scheme has some functional usefulness in that some theories of sound loudness coding take advantage of the fact that high spontaneous ¢bers can encode intensity at low levels, and low spontaneous ¢bers, which often have rate level functions with a sloping saturation, can encode sound intensity at higher levels (when all of the high spontaneous ¢bers are in

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

3

Fig. 1. Auditory nerve ¢bers give strong synchronous responses to AM at low intensity levels and weaker ones as intensity rises. PSTHs are displayed for 75 ms tone bursts with carriers at a unit's BF. Upper row: high sound levels (50^58 dB re threshold). Lower row: low sound intensities (18^22 dB re threshold). AM r values for medium spontaneous rate ¢bers (middle column) exceed those of high spontaneous ¢bers (left column), and r values for low spontaneous units (right column) exceed those of the medium spontaneous ¢bers. The quantitative values in the upper right of each PSTH were measured in a 50 ms steady state time window beginning 25 ms after response onset. Bin width = 500 Ws, AM frequency = 120 Hz, with each cycle of the AM stimulus designated by the arrows below each PSTH. Stimulus onset was delayed by 5 ms re time = 0. Left column: unit 197, th = 15 dB, BF = 2.9 kHz, SR = 30; Rayleigh stats (test for statistical signi¢cance of the synchronous response, NS = no signi¢cant AM response): 52 dB NS, 22 dB P 6 0.001. Middle column: unit 209, th = 30 dB, BF = 1.4 kHz, SR = 4; Rayleigh stats: 50 dB NS, 20 dB P 6 0.001. Right column: unit 135, th = 36 dB, BF = 8.1 kHz, SR = 1; Rayleigh stats: 58 dB NS, 18 dB P 6 0.001. Average: total number of spikes in response to the AM stimulus; gain: AM response gain (see Frisina et al., 1996 for equation); Spike Density: number of spikes/bin; Synch: number of spikes phase-locked to the AM stimulus (¢rst Fourier coe¤cient); th: threshold (dB SPL). Figure reproduced from Frisina et al. (1996, ¢gure 1, p. 478), with permission of the author and publisher.

their saturation region in terms of maximal ¢ring rate). Relevant to the current treatise, these classes have different responses to AM, as delineated below. 2.2. AM transfer functions: lowpass It was discovered early on that at moderate intensity levels (up to about 35 dB above threshold), auditory nerve ¢bers display lowpass AM transfer functions (MÖller, 1976a; Javel, 1980; Palmer, 1982). The high frequency cuto¡s of these functions are typically in the range of 500^800 Hz, and local maxima are sometimes observed at high modulation frequencies for low characteristic frequency and low spontaneous auditory nerve ¢bers (see Langner, 1992, ¢gure 3). Response modulation gains for high spontaneous auditory nerve ¢bers tend to be lower than for medium spontaneous ¢bers, which in turn are below those of low spontaneous ¢bers (Joris and Yin, 1992; Cooper et al., 1993 ; Rhode and Greenberg, 1994a; Frisina et al., 1996). For high best frequency (BF) auditory nerve ¢bers, robust responses to AM can be obtained for low carrier frequencies (i.e. below 1 kHz) residing in the tails of

their tuning curves (Frisina et al., 1996; Henry, 1998). This allows for low frequency sound envelope information, which can be utilized as a neural coding mechanism for sound localization, to be encoded by neural channels that carry high frequency sound carrier information. 2.3. AM coding with sound level Phase locking to envelope £uctuations falls o¡ significantly with increases in sound level above moderate average stimulus intensity levels, as displayed in poststimulus time histogram (PSTH) form in Fig. 1 (MÖller, 1976a ; Evans and Palmer, 1980; Smith and Brachman, 1980a ; Yates, 1981). The declines at high AM frequencies can be less than at low ones (Yates, 1981). At high sound levels, low spontaneous ¢bers give the highest AM response gains, and high spontaneous ¢bers give the poorest responses, with medium spontaneous ¢bers in between (Frisina et al., 1996). At AM frequencies in the 150^300 Hz range, the magnitude of the AM response gain for high spontaneous auditory nerve ¢bers is quantitatively predictable

HEARES 3704 1-8-01

4

R.D. Frisina / Hearing Research 158 (2001) 1^27

from the slope of the rate level function to pure tones (Smith and Brachman, 1980b,c). The onset rate level function, measured during the initial 1 ms of the auditory nerve ¢ber response is a better quantitative predictor then the classic, steady state rate level function. The correspondence between the slope of the rate level function and AM vector strength as a function of sound intensity is poorer for low spontaneous ¢bers, relative to the other groups, as presented in Fig. 2 (Frisina et al., 1996). Lastly, if an AM stimulus is left on for a while, the AM response will decline with time due to adaptation, particularly of the decremental portion of the AM cycle (Smith et al., 1985). 2.4. AM processing in background noise Overall, AM responses of auditory nerve ¢bers are preserved or slightly enhanced under many conditions of background noise (Rhode and Greenberg, 1994a; Frisina et al., 1996), as presented here in Fig. 3 and Table 1. The data presented in Table 1 indicate that preservation or enhancement of auditory nerve AM response gains in the presence of background noise can be achieved by increases in the synchronous ¢ring rate (synch spikes) for high and medium spontaneous rate ¢bers, or by reductions in the average ¢ring rate (average spikes) for all three types of auditory nerve ¢bers. 3. Cochlear nucleus periodicity coding 3.1. Basic organization relevant to temporal processing

Fig. 2. Virtually all auditory nerve ¢bers have monotonic rate^intensity functions for pure tone bursts at BF in quiet. Dynamic operating ranges for pure tone rate level functions (triangles) are similar to the operating ranges for AM (circles). The intensity range of the steeply sloping portion of the rate level functions are indicated by the vertical dotted lines in each graph. Note that the maximum AM coding tends to occur in the same range as the steeply sloping portion of the rate level function, i.e. between the vertical lines. (a) Unit 197, th = 15 dB, BF = 2.9 kHz, SR = 30; (b) unit 209, th = 30 dB, BF = 1.4 kHz, SR = 4; (c) unit 232, th = 28 dB, BF = 1.5 kHz, SR = 0. Figure reproduced from Frisina et al. (1996, ¢gure 2, p. 479), with permission of the author and publisher.

All information from the cochlea is carried via auditory nerve ¢bers to the cochlear nucleus. In the cochlear nucleus, parallel processing channels are established where certain features of the acoustic stimulus are abstracted and preferentially coded at the expense of other acoustic features. Generally speaking, the anatomical complexity and corresponding sound processing sophistication increases as one moves from the rostral pole of the anteroventral cochlear nucleus (AVCN) in a posterior direction through intermediate regions of the ventral cochlear nucleus (VCN), to the caudal pole of the posterior ventral cochlear nucleus (PVCN), and ¢nally proceeding to the dorsal cochlear nucleus (DCN). Through this progression, the number of dendrites on the cochlear nucleus cells and the number and complexity of excitatory and inhibitory inputs to the neurons tend to increase, reaching a maximum level of complexity in the DCN. The basic neurophysiological response properties of cochlear nucleus cells increase in complexity, and become more di¡erent from the responses of auditory nerve ¢bers, along this topographical progression as well.

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

3.2. Pioneering investigations Initial single unit studies of auditory temporal processing were conducted by MÖller utilizing the anesthetized rat as the experimental animal (MÖller, 1972, 1973, 1974a,b, 1975a,b, 1976a,b,c). MÖller's primary goal was to perform a linear systems analysis of single neuron responses to determine the ¢delity with which AM is processed in the cochlear nucleus. He discovered that, except at very high stimulus modulation depths, most cochlear nucleus neurons encode AM with high ¢delity. Units were tuned to di¡erent AM frequencies (80^500 Hz) and their tuning properties were relatively insensitive to changes in stimulus parameters such as modulation depth, sound duration, and whether continuous tones or repetitive tone bursts were used. AM tuning properties of a single unit remained stable for hours, and were the same regardless of which stimulus paradigm was used to obtain them. For instance, in some cases tones or wideband noise amplitude modulated with sinusoids were used and the response modulation was computed from period histograms. In other cases, continuous tones or noise modulated with pseudorandom noise (random noise that is repeated periodically) were employed and cross-correlation functions obtained. This allowed for the computation of AM tuning properties using Fourier transforms. Remarkably, MÖller found that some cochlear nucleus units can amplify the depth of modulation of their responses, relative to that in the stimulus, for sound levels up to 60 dB above their thresholds. In contrast, he found that most auditory nerve ¢bers could do this

5

only over a 30 dB range. Lastly, he pointed out that the range of intensity levels over which signi¢cant responses to AM occurs exceeds the operating range of the rate level functions for most cochlear nucleus units. As is true of all good pioneering work, MÖller's studies raised important questions such as: Which cochlear nucleus unit types encode AM the best? Where are they located ? Are there relations between a cochlear nucleus unit's responses to simple sounds and its ability to process AM ? Are the ¢ndings in rats applicable to other mammals? What are the neural mechanisms for AM coding ? 3.3. Ventral nucleus: hierarchy of enhancement for AM processing Frisina and coworkers sought answers to some of these questions using an anesthetized gerbil single unit preparation (Frisina et al., 1982). In the VCN, they found that a relation exists between a unit's responses to pure tones and its AM coding ability (Frisina et al., 1985, 1990a). Speci¢cally, of the four main unit types found in the VCN, onset units gave the strongest phaselocked AM responses, followed by chopper, primarylike-with-notch, and primarylike units, respectively, as shown in Fig. 4. As sound level is raised, VCN units showed extended dynamic ranges for AM coding relative to auditory nerve ¢bers, following the hierarchy of enhancement just described for moderate sound levels. The AM coding behavior of the major VCN unit types as a function of sound level and AM frequency are given in Fig. 5. Notice that at low average intensities,

Table 1 E¡ects of background noise on AM processing at high sound levels ^ 75 dB SPL Unit type High SR n = 24

Medium SR n=9

Low SR n=8

a b

AM response measure Mean gain (dB) S.E.M. Mean synch spikes S.E.M. Mean average spikes S.E.M. Mean gain (dB) S.E.M. Mean synch spikes S.E.M. Mean average spikes S.E.M. Mean gain (dB) S.E.M. Mean synch spikes S.E.M. Mean average spikes S.E.M.

120 Hz

300 Hz

Quiet

+6a

0b

Quiet

+6a

314.6 0.9 21.2 1.8 642 30.1 39.4 1.6 34.7 8.4 526 83.8 33.5 1.7 48.6 16.0 356 55.4

312.5 1.3 24.2 2.6 537 31.3 36.4 1.7 38.0 6.8 464 70.8 35.4 4.0 35.0 12.5 269 54.9

314.4 1.1 20.9 2.4 536 34.2 35.7 1.2 43.3 10.5 432 61.9 36.7 3.7 29.0 10.6 266 56.8

39.2 1.6 47.6 8.4 611 32.4 311.2 1.9 31.8 5.2 622 74.3 32.9 1.4 41.4 7.1 323 52.0

310.4 2.2 35.3 8.4 473 35.5 39.9 1.5 35.8 8.4 572 89.3 31.5 1.5 37.9 8.0 291 63.2

Designates background noise at +6 dB S/N. Designates background noise at 0 dB S/N.

HEARES 3704 1-8-01

0b 39.1 1.5 37.7 6.7 512 31.1 37.3 1.6 45.8 12.3 490 69.0 33.0 2.9 36.1 11.5 269 65.2

6

R.D. Frisina / Hearing Research 158 (2001) 1^27 Fig. 3. Auditory nerve ¢ber AM r values for tones in quiet (solid circles) are preserved in the presence of a loud, wideband background noise (0 dB S/N, open circles). Note that the AM r values of the low SR unit (bottom) tend to exceed those of the medium SR unit (middle), which in turn exceed those of the high SR ¢ber (top). Data are displayed for 75 and 125 ms AM tone bursts with carriers at the unit's BF. Tone burst sound level = 55 dB SPL. Dotted lines indicate r values between 0 and 0.1 to facilitate comparison between the three graphs. The stimulus modulation depth was 35%. Top: unit 195, th = 20 dB, BF = 4.7 kHz, SR = 56. Middle: unit 209, th = 30 dB, BF = 1.4 kHz, SR = 4. Bottom: unit 135, th = 36 dB, BF = 8.1 kHz, SR = 1. S/N: signal to noise ratio, other abbreviations the same as previous ¢gures. Figure reproduced here from Frisina et al. (1996, ¢gure 11, p. 485), with permission of the author and publisher. 6

the AM transfer functions are lowpass in nature, like auditory nerve ¢ber AM transfer functions. However, as sound level is raised, the AM transfer functions become more bandpass (peaked) in form. Observe that the greatest enhancement of AM response gains in VCN units relative to auditory nerve ¢bers is at high sound levels. Rhode and Greenberg (1992, 1994a) made further

Fig. 4. At high sound levels, the strength of phase-locked responses to AM in the VCN varies for di¡erent unit types. These PSTHs were collected in response to 100 repetitions of a tone burst at a unit's BF (shown schematically at the bottom). Note that the stimulus amplitude remained constant for the initial 50 ms, and was sinusoidally modulated at 150 Hz during the last 50 ms (highlighted by the dotted rectangles). The stimulus intensity is in the upper right of each PSTH (50^60 dB re th). Bin width = 600 Ws, stimulus modulation depth = 35 or 50%, rise/decay time = 2.5 ms. (a) Onset units tended to show the strongest AM phase locking. (b) Chopper units also display vigorous AM coding. AM encoding in some chopper and onset units occurs at sound levels up to 90 dB re threshold. (c) Primarylike-with-notch units show moderate synchrony at high intensities. (d) Primarylike units show the least amount of AM coding in the VCN. On-L: unit 130, th = 33 dB, BF = 12.0 kHz. Chopper: unit 114, th = 32 dB, BF = 11.9 kHz. Primarylike-with-notch: unit 122, th = 25 dB, BF = 14.0 kHz. Primarylike: unit 162, th = 20 dB, BF = 15.2 kHz. Pri, primarylike units. Figure reproduced from Frisina et al. (1985, ¢gure 2, p. 419), with permission of the author and publisher.

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

7

Fig. 5. Analysis of VCN AM response surfaces revealed that a relation exists between a neuron's ability to encode AM and its responses to simple sounds. Onset units produce the strongest AM phase locking, followed in order by chopper, primarylike-with-notch and primarylike units. All four VCN unit types have AM responses that exceed those of high spontaneous auditory nerve ¢bers, particularly at high sound levels. The three-dimensional surfaces display how AM response gains change with sound intensity and AM frequency. The responses of the four VCN unit types are overplotted the response surface of a typical high spontaneous auditory nerve ¢ber (dotted lines). The surfaces delineate volumes that give the viewer an impression of a neuron's overall AM coding ability. On-L: unit 130, th = 33 dB, BF = 12.0 kHz. Chopper: unit 131, th = 9dB, BF = 9.9 kHz. Primarylike-with-notch: unit 160, th = 16 dB, BF = 17.1 kHz. Primarylike: unit 162, th = 20 dB, BF = 15.2 kHz. Auditory nerve ¢ber: unit M8-6, th = 16 dB, BF = 6.5 kHz. Stimulus modulation depth = 35%. Figure reproduced from Frisina et al. (1985, ¢gure 3, p. 420), with permission of the author and publisher.

detailed investigations of the VCN hierarchy of enhancement for AM coding in the anesthetized cat. They classi¢ed their units into slightly di¡erent AM coding groups with primarylike units showing the lowest AM synchrony, followed by chopper and on-L units, and then primarylike-with-notch and onsetchopper units yielding the highest AM response synchrony. Comprehensive group data for all the unit types of this investigation are presented in Fig. 6. Note that the cochlear nucleus unit types tend to have more clearly de¢ned peaks than those of auditory nerve

¢bers. These unit type di¡erences in the AM coding hierarchy between the gerbil and cat could be a true VCN species di¡erence or may be due to unit sampling/classi¢cation discrepancies between these two investigative groups. Rhode and Greenberg (1994a,b) extensively examined AM tuning properties in the VCN. They discovered little relation between the upper cuto¡ of a cochlear nucleus unit's AM transfer function and its BF, unlike the relation that exists in the auditory nerve and for VCN primarylike units. In terms of the distri-

HEARES 3704 1-8-01

8

R.D. Frisina / Hearing Research 158 (2001) 1^27

Fig. 6. Population data from each of the main cochlear nucleus unit types demonstrates that the amplitude modulation transfer functions (MTFs) tend to be more bandpass in nature and have a lower high cuto¡ frequency relative to auditory nerve ¢bers. Modulation depth = 100%, sound levels = 50^60 dB SPL, tone burst carrier = BF, duration = 100 ms. AN, auditory nerve; Cs, sustained chopper; Ct, transient chopper; Oc, onset chopper; Ol, onset-L; PL, primarylike; P/B, pauser/buildup unit of DCN; PLn, primarylike-with-notch; Sync. Coe¡., r, vector strength or synchronization index. Figure reproduced from Rhode and Greenberg (1994a, ¢gure 10, p. 1808), with permission of the author and publisher.

bution of best modulation frequencies (BMFs), primarylike-with-notch units in the VCN had the highest BMFs (488 þ 54 Hz). Primarylike (375 þ 66 Hz) and transient choppers (372 þ 148 Hz) were next, followed by sustained choppers (305 þ 94 Hz). DCN pauser/ buildup units had the lowest BMF range (261 þ 145 Hz) in the cochlear nucleus. When proceeding up through the brainstem auditory system, this decrease in the mean BMFs based upon AM synchrony seen between the VCN and DCN continues.

3.4. Possible neural mechanisms Frisina et al. (1990b, 1993) explored possible neural mechanisms at the level of the VCN that could account for the abilities of these units to extract and amplify sound envelope information. A likely candidate explanation employed the notion that convergence of inputs from auditory nerve ¢bers with di¡erent spontaneous activity rates occurs for VCN units so that AM information at low sound levels comes from high spontane-

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

ous auditory nerve ¢bers and at high sound levels from low spontaneous units. Another mechanism that could contribute to this hierarchy of enhancement utilizes the fact that VCN stellate cells, which are correlated with onset and chopper unit types, have inputs from many auditory nerve ¢bers that are spatially distributed across their long tapering dendrites (Cant and Morest, 1984). Modeling these dendrites as cables, and calculating current spread from many distributed inputs along their length, suggests that low frequency sound envelope information would be transmitted to the stellate cell soma more e¡ectively than high frequencies, and that with appropriate dendritic geometry and membrane characteristics, each VCN cell could be tuned to a particular AM frequency (Banks and Sachs, 1991). Relations between these dendritic properties and membrane characteristics, and response pattern temporal characteristics were explored in the cat VCN for simple sounds by Rhode and Smith (1986a), and their ¢ndings in the cat con¢rmed and extended those of Frisina and coworkers in the gerbil. Rhode (1994) utilizing 200% modulated AM stimuli in the VCN demonstrated that the synchronous response to AM, rather than the average ¢ring rate to AM (total spikes), is the means of encoding AM frequency e¡ectively. This is because he observed that functions representing discharge rate vs. AM frequency were £at (all-pass), whereas the synchrony coe¤cient vs. AM frequency plots were bandpass with noticeable peaks (best modulation frequencies). In terms of sharpness of AM tuning, cochlear nucleus units are more sharply tuned than auditory nerve ¢bers, with choppers displaying the greatest sharpness of tuning, particularly the sustained chopper subgroup. In addition, there was a tendency for many VCN units to increase their sharpness of AM tuning with intensity. This was especially apparent for the on-L unit type. Other ¢ndings by Rhode and colleagues concerning VCN AM coding and the possible neural mechanisms responsible for AM processing were consistent with those found for the gerbil. One question addressed by both investigative groups, concerning the roles of excitatory and inhibitory inputs regarding AM coding mechanisms in the VCN, is whether inhibitory inputs increase AM encoding by simply reducing the average ¢ring rate, or whether they can actually enhance the synchronous AM response. Frisina et al. (1997b) performed an initial investigation of this question utilizing multi-barrel electrodes loaded with the glycine antagonist strychnine, and found that some chopper units showed declines in AM processing due to reductions in their synchronous AM response as shown here in Fig. 7. Although not the focus of the present review, intracellular recordings from auditory brainstem slice prep-

9

Fig. 7. PSTH responses of a VCN chopper unit to AM in quiet for pre- and post-strychnine conditions. Application of this glycine blocker reduces the synchronous response to AM (Synch Sp.) but not the overall ¢ring rate (Total Sp.). The stimulus was a 50 ms BF tone. The abscissae represent time (ms) with stimulus onset at 10 ms, and the ordinates plot spike density (spikes/bin). Gain: AM response gain (see Frisina et al., 1996 for equation). Figure reproduced from Frisina et al. (1997b, ¢gure 5, p. 120), with permission of the author and publisher.

arations by Oertel and coworkers give important insights into cellular mechanisms that may be responsible for temporal processing in the cochlear nucleus (Oertel, 1985, 1997, 1999 ; Trussell, 1997; Ferragamo et al., 1998). Their intracellular recordings reveal that the membrane characteristics and relative contributions of excitatory and inhibitory inputs determine the di¡ering roles that spherical or globular/bushy cells and multipolar/stellate cells play in preserving or extracting VCN temporal information in primarylike

HEARES 3704 1-8-01

10

R.D. Frisina / Hearing Research 158 (2001) 1^27

Fig. 8. AM transfer functions (MTFs) presented in quiet (solid lines) and in background noise (dashed functions) for auditory nerve and main cochlear nucleus response types. Notice that background noise reduces the synchronous response, and produces a slight £attening of many of the MTFs. The background noise intensity was adjusted so that the noise by itself evoked a response that was 50% of the maximal noise driven spike rate. Modulation depth = 100%, sound levels = 50^60 dB SPL, tone burst carrier = BF, duration = 100 ms. AN, auditory nerve; B, buildup unit of DCN; Cs, sustained chopper; Ct, transient chopper; Oc, onset chopper; PL, primarylike; Sync. Coe¡., r, vector strength or synchronization index. Figure reproduced from Rhode and Greenberg (1994a, ¢gure 17, p. 1816), with permission of the author and publisher.

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

and onset/chopper units, respectively. In the AVCN, bushy cells have large synaptic currents with rapid kinetics in response to inputs from big calyceal terminals from a small group of auditory nerve ¢bers. This synaptic input con¢guration results in one postsynaptic spike generated precisely in time for each incoming spike. These neurons are ideal for preserving the temporal information present in their inputs. Multipolar/ stellate neurons, on the other hand, have distributed inputs and membrane characteristics that are conducive to extracting temporal information from incoming spike trains from a larger number of auditory nerve ¢bers. Lastly, recent intracellular ion channel studies of octopus cells by Cai et al. (2000a,b) suggest that these onset units of the PVCN are exquisite for encoding temporal features of sound, such as AM, under a variety of acoustic stimulus conditions. 3.5. Dorsal cochlear nucleus AM coding Kim et al. (1990) found cat DCN units to be quite adept at processing AM across a wide range of sound levels, utilizing autocorrelation/power spectrum Fourier analysis methodologies. Rhode and Greenberg (1994a) followed up this investigation and demonstrated that cat pauser/buildup units, which correspond to the fusiform cell anatomical class (Rhode and Smith, 1986b), fell in the middle group for cochlear nucleus AM coding pro¢ciency, with response synchronies comparable to VCN chopper and on-L units (see Fig. 6). The DCN pauser/buildup units showed the most resistance to changes in AM coding of any cochlear nucleus unit types as the sound level increased, which was also observed by Frisina et al. (1994, 1997b) in the chinchilla. In the cat DCN, Joris and Smith (1998) examined AM processing capabilities of interneurons: onsetchopper, type II, and output neurons that project to the inferior colliculus, type III/IV. They demonstrated that all three unit types show stronger synchronous responses to AM than auditory nerve ¢bers, particularly at high sound levels. They also found that best AM frequencies and high cuto¡s were signi¢cantly lower for DCN units with high BFs than for auditory nerve ¢bers with comparable BFs. 3.6. E¡ects of background noise on AM encoding Frisina et al. (1994, 1997b) and Rhode and Greenberg (1994a) investigated the e¡ects of background noise on cochlear nucleus units. AM coding in all cochlear nucleus unit types remained quite robust in the presence of a wideband background noise, as shown here in Fig. 8. In fact, for cochlear nucleus units, there were certain AM frequencies where the synchronous responses in the background noise exceeded those in

11

quiet, an example of enhanced AM processing in a noisy acoustic environment. The ability of many cochlear nucleus units to encode AM despite the presence of an intense wideband masking noise is no doubt due to the presence of signi¢cant lateral inhibitory response areas for non-primarylike cochlear nucleus units that act to suppress the average ¢ring rate in response to the background noise (Rhode and Greenberg, 1994b). 4. Superior olivary complex 4.1. Binaural enhancement of envelope temporal features Consideration of the interaural processing capabilities of neurons at the level of the superior olivary complex (SOC) suggests that they have the potential to be specialized for coding low frequency envelope information (below 1 kHz). For example, the response of a neuron to a particular combination of interaural phase di¡erences at certain AM frequencies could be enhanced. This would allow SOC cells, speci¢cally high BF neurons in the lateral (LSO) or medial (MSO) superior olives, that are normally unresponsive to low frequency sounds, to encode low frequency sound envelope information useful for sound localization, as has been demonstrated in human psychophysical experiments (Hafter et al., 1990). Joris and Yin (1995) were the ¢rst auditory physiologists to test this hypothesis at the single unit level in the LSO of the barbiturate anesthetized cat. Speci¢cally, they presented AM sounds binaurally, varied the interaural phase of the AM signals to the two ears, and measured the responses of `IE' neurons. These LSO neurons are inhibited by contralateral sounds and excited ipsilaterally, and gave minimal responses when the AM sounds to the two ears were in phase, or nearly so. These neurons give robust responses to AM when the AM signals were out of phase at the two ears, as illustrated in Fig. 9. There was a good correspondence between the best AM interaural phase di¡erence and the best interaural time difference measured with pure tones, consistent with the neural IE operation. Interaural time coding using the AM envelope information was limited by severe declines in response rates with increases in the AM frequency beyond several hundred Hz. This limit is much lower than that of MSO cells, which are capable of analyzing interaural time/phase di¡erences utilizing carrier frequency information up to several kHz. Batra et al. (1997a) undertook studies of single unit AM coding in the SOC utilizing an unanesthetized rabbit physiological preparation. A response pattern, which they termed `trough-like', was found to be characteristic of LSO IE units (Fig. 10). This response to interaural AM phase di¡erences showed a minimal dis-

HEARES 3704 1-8-01

12

R.D. Frisina / Hearing Research 158 (2001) 1^27

Fig. 9. The single unit recording results of the Joris and Yin (1995) investigation support the hypothetical bases of sensitivity to interaural phase di¡erences (IPDs) in LSO neurons for sound envelopes, in this case modulated sinusoidally at low AM frequencies. PSTHs illustrate predicted phase relationships for in-phase (A: IPD = 0 cycles) and out-of-phase (B: IPD = 0.5 cycles) AM signals. CN, cochlear nucleus; MNTB, medial nucleus of the trapezoid body. Figure reproduced from Joris and Yin (1995, ¢gure 1, p. 1045), with permission of the author and publisher.

charge at a particular interaural time di¡erence, across di¡erent AM frequencies. Joris (1996) pursued the SOC line of research further by extending binaural AM recordings into the anesthetized cat MSO `EE' area. Neurons in this region give excitatory responses to sounds presented to either ear. These cells were not as e¡ective as LSO neurons in

coding interaural envelope time di¡erences, although the coding that did occur was predictable from responses to monaural sounds. For MSO EE units, Batra et al. (1997a), in the unanesthetized rabbit, discovered a prominent `peak-type' response pattern, one that discharged maximally at a given interaural AM time difference, even when the AM frequency was varied.

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

13

Fig. 10. Interaural time delay (ITD) sensitivity of four SOC units for sinusoidally modulated AM tones. The left column represents interaural delay functions for AM tones of di¡erent modulation frequencies. The top two units show a `peak' response where a certain interaural time delay gives a maximal response regardless of AM frequency. The bottom two units give a `trough' response, where a minimum in spike ¢ring occurs at the same interaural delay regardless of AM frequency. The functions for the right column are least squares ¢t, the carrier frequencies (unit best frequency) are given above each plot in the left column, and the AM modulation depth was 80%. CD, characteristic delay, equals the slope of the mean interaural phase plot; CP, characteristic phase, the y-intercept of the mean interaural phase plot. Figure reproduced from Batra et al. (1997a, ¢gure 2, p. 1226), with permission of the author and publisher.

Exemplary responses are put forth in Fig. 10. Some units were intermediate between peak and trough types, and showed more variation of the preferred interaural time di¡erence as a function of AM frequency. Batra and coworkers found that the range of interaural time di¡erences encoded by their unit sample ( þ 300 Ws) covered the behavioral free ¢eld, low frequency sound localization range for the rabbit. Batra et al. (1997b) examined the neural mechanisms for these two binaural AM coding unit types in the SOC and concluded that both types employ coincidence detection as a means of e¡ectively processing the interaural low frequency temporal information in the manner just described. More recently, Kuwada and Batra (1999) continued their investigations of AM coding in the SOC perio-

livary regions of the unanesthetized rabbit, examining di¡erent pure tone unit types monaurally. They demonstrated that AM responses are signi¢cantly enhanced in single units with o¡ responses to tone bursts relative to units displaying sustained responses. This enhanced coding of AM, via inhibitory rebound, can improve processing of AM interaural phase di¡erences useful for sound localization, as well as extend the range of coding for sound intensity. More speci¢cally, the sustained neurons responded to AM in a manner similar to what has been reported for most cochlear nucleus neurons : the discharge rate was constant as the AM frequency was varied, but the synchrony changed. Using the SC as a metric, the sustained neurons were weakly bandpass. The strength of synchrony was comparable

HEARES 3704 1-8-01

14

R.D. Frisina / Hearing Research 158 (2001) 1^27

with SCs clustered between 0.8 and 1.0. The o¡ neurons synchronized strongly even to relatively shallow modulation depths. Echolocating mammals, such as bats, navigate and hunt £ying insects utilizing a biosonar system, where many parts of the auditory system are specialized for encoding and analyzing the di¡erences between stereotypical emitted sound pulses and their returning echoes (e.g. Pollak et al., 1972, 1979; Suga and Jen, 1976 ; Suga et al., 1978; Pollak and Bodenhamer, 1981 ; Suga and Manabe, 1982 ; Frisina et al., 1989 ; O'Neill et al., 1989) As explored by Grothe (1994), in some bats that have small heads, such as the well-characterized mustached bat, the MSO functions as a monaural nucleus giving preferential responses to sound envelope information, including AM. Grothe et al. (1997) went on to investigate the more typical binaural MSO of the Mexican free-tailed bat to see how it processes AM signals. They found that most units (55% of their sample) displayed bilateral excitation with a weaker component of inhibition, so called `EI/EI' neurons. The remaining units exhibited reduced inputs, generally lacking ipsilateral inputs. Most units displayed monotonic rate level functions for pure tones, but were highly nonmonotonic for AM signals. Eighty-eight percent of the neurons gave strong, phase-locked responses to sinusoidal AM, and displayed modulation transfer functions that were lowpass in shape. Interaural stimulus di¡erences altered the AM coding in unpredictable ways, sometimes enhancing it and at other times degrading it. It is interesting to note that Kuwada and Batra (1999) found units in the periolivary regions of the rabbit that have similar AM response properties as these specialized bat MSO units of Grothe et al., suggesting that such monaural units may be present in the SOC of non-echolocating mammals, but in di¡erent sublocations. 5. Inferior colliculus AM coding Fig. 11. IC AM modulation transfer functions are robust in that they generally yield the same results using sinusoidal AM or stimuli modulated with noise or pseudorandom noise. In the former case, response characteristics are obtained from period histograms, and in the latter case they are obtained by Fourier transforms of the crosscorrelation function of the stimulus and response. Cf = best frequency. Figure reproduced from MÖller and Rees (1986, ¢gures 7 and 8, p. 212), with permission of the author and publisher.

to what has been observed in the cochlear nucleus. In contrast, responses of o¡ neurons di¡ered substantially from those reported in the cochlear nucleus. The discharge rate of o¡ neurons was modulated by the AM frequency, declining at higher AM frequencies. These neurons synchronized strongly at lower frequencies,

5.1. Basic anatomy and circuitry relevant to temporal processing The largest portion of the auditory midbrain is the central nucleus of the inferior colliculus (ICC). Dorsal and caudal to the ICC lies a cortical strip commonly called the dorsal cortex (DC). Ventrolateral and rostral to ICC resides the external nucleus (E). An area anterior to ICC and E, including the deepest layers of the superior colliculus (SC), comprises a region involved in multimodality (auditory, visual) processing of spatial maps. Much of the general organizational plan and cytoarchitectonics for the mammalian ICC have been worked out in cat by Morest, Oliver and colleagues (Oliver and Morest, 1984; Oliver and Huerta, 1992).

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

Many similarities exist between this basic organization in cat and that of the ICC of mice and other rodents (Meininger et al., 1986). There are two principal cell types in the ICC : disc-shaped neurons, with elongate somata and £attened dendritic ¢elds, and stellate neurons whose dendritic arborizations extend across the ICC laminae. The ICC has a very prominent ¢brodendritic laminar organization. The laminae are composed of the somata and dendrites of the primary principal cells : the discshaped neurons. These neurons have oval to fusiform shaped perikarya and £attened dendritic arbors. These characteristics of the ICC principal neurons, together with the incoming lateral lemniscus (LL) ¢bers, compose the ICC laminae. The laminae in most mammals such as cat and mouse are oriented like a stack of £apjacks, with the main surface of the each pancake aligned approximately in the horizontal plane, with a tilt in the dorsomedial/ventrolateral direction. Therefore, moving from dorsal to ventral through the ICC, one proceeds through the stack of laminae. In terms of tonotopicity, the dorsal laminae encode low frequencies, and the ventral laminae process high frequencies. Orthogonal to these main tonotopic laminae, functional suborganizations exist. Topographic organizations for AM coding (described below) and for pitch analysis that may underlie the psychophysical phenomenon of critical bands (Schreiner and Langner, 1997) coexist as part of the primary cochleotopic topography. Virtually all of the processing in lower centers of the auditory brainstem project onto ICC neurons (e.g. Frisina et al., 1998). Primary projections originate from all three divisions of the contralateral cochlear nucleus, the SOC bilaterally, the ipsilateral ventral nucleus of the LL (VNLL), and the IC bilaterally. Other inputs come from the dorsal nucleus of LL (DNLL) and intermediate nucleus of LL (INLL) bilaterally, the contralateral DC and E, the ipsilateral central gray (reticular formation), and all three divisions of the ipsilateral medial geniculate body of the thalamus (MGB). Major inputs, such as those from the contralateral cochlear nucleus and NLL, are tonotopically organized.

15

5.2. Neuron coding as a function of AM frequency Rees and MÖller (1983, 1987) and MÖller and Rees (1986) were the ¢rst to explore AM processing in single units of the mammalian IC, using the anesthetized rat as the animal preparation. Using both AM tones and noise, and the same carriers modulated with pseudorandom noise, they discovered that AM transfer functions in the IC had much lower high cuto¡ frequencies than for the more peripheral levels of the auditory system, examples of which are given in Fig. 11. Therefore they noted that a trend was occurring for the auditory nerve, cochlear nucleus and IC, in that the higher in the

C Fig. 12. AM transfer functions at the level of the inferior colliculus become more £at as average intensity is decreased (top) and as background noise becomes louder (bottom). The data in the top graph were obtained for unit 822/3 using BF tones modulated with a pseudorandom noise waveform. Intensity levels given are re BF threshold. The data in the bottom graph were obtained for another unit with the same procedures, however the background noise level was varied, as indicated by the signal to noise ratios given for each modulation transfer function. Figure reproduced from Rees and MÖller (1987, ¢gures 2 and 9, pp. 134, 138), with permission of the author and publisher.

HEARES 3704 1-8-01

16

R.D. Frisina / Hearing Research 158 (2001) 1^27

system one goes, the greater the AM coding shifted from a temporal code to a rate code. Rees and MÖller also noted that AM coding as measured in period histograms in response to sinusoidal AM di¡ered in subtle ways from that measured using the cross-correlation of spikes with pseudorandom noise modulated signals. This indicated the presence of even order nonlinearities in IC AM coding, indicating di¡erent responses to the rising and falling phases of the AM. In some units this manifested itself as half-wave recti¢cation. Similar to the Frisina et al. (1985, 1990a) studies in the cochlear nucleus, Rees and MÖller discovered that some IC units could encode AM robustly over a wide range of average sound levels, and that AM transfer functions change from lowpass to bandpass in shape with increases in stimulus intensity, as displayed here in Fig. 12 (top). Lastly, they noted that when a broadband background noise was added to the stimulus con¢guration, modulation transfer functions shifted from being bandpass in nature to lowpass (Fig. 12, bottom). Rees and Palmer (1989) continued this line of experimentation in the anesthetized guinea pig utilizing sinusoidal AM tones at a unit's BF. They analyzed AM coding in terms of the synchronous response (phase locking) and the average response (total spikes) and found that the peak of the AM vs. spike response functions were the same for the synchronous and average responses in about one half of the units of their sample. They also noted that the average intensity level where the greatest correspondence occurred between synchronous and average AM response measures was in the steeply sloping portion of the rate level function. They

Fig. 13. Schematic summary of relations between rate level functions (a) and modulation transfer functions (b) at the level of the inferior colliculus. As background noise is added, rate level functions shift down and to the right (dashed function and arrow in graph a). If sound level is raised in quiet (moving from 1 to 2 in graph a) the AM transfer function becomes more bandpass (moving from 1 to 2 in graph b, transfer function becomes more peaked). If background noise is then added (moving from 2 to 3 in both graphs) the rate level function shifts to the right and the AM transfer function becomes more lowpass in shape. Figure reproduced here from Rees and Palmer (1989, ¢gure 1, p. 1979), with permission of the author and publisher.

also found consistent relations between shifts in the operating ranges of IC units that occur in background noise and AM coding shifts in noise, as conceptualized in summary form in Fig. 13. 5.3. Topographic organization for AM coding In a series of enlightening investigations the IC of cats and guinea fowl (Langner, 1981, 1983), Langner and Schreiner discovered a topographic organization for temporal coding in the central nucleus that is overlaid and somewhat orthogonal to the ICC isofrequency laminae described above. Utilizing an anesthetized cat preparation for single and multi unit recordings, Langner et al. (1987) initially uncovered a correlation between a unit's latency and its best modulation frequency (BMF) for AM. Speci¢cally, units with high BMFs had the shorter latencies. Langner and Schreiner (1988) then went on to measure AM transfer functions, yielding BMFs both in terms of synchrony (sBMF) and total spike rate (rBMF). They found that for units with similar BFs for pure tones, and therefore in the same ICC isofrequency lamina, there was considerable variation in AM BMFs. They also observed that the upper limit for BMFs increased approximately in proportion to the BF/4 up to BFs as high as 1 kHz, with the lower limit on BMFs also increasing slightly with BF. Eightyfour percent of the unit sample had rBMFs below 100 Hz. For units with bandpass modulation transfer functions (MTFs), there was a high correlation of the peak of the rBMF and sBMF (r2 = 0.95), and these BMFs were correlated with response latency to BF tones and BF itself. These ¢ndings, along with the observation that many units displayed intrinsic oscillations (intervals were integer multiples of 0.4 ms) in response to sounds lacking temporal structure, suggested that within an ICC isofrequency lamina, a temporal analysis of sound temporal (periodicity pitch) information takes place. In the companion report (Schreiner and Langner, 1988), the topographic organization of rBMFs was demonstrated, covering a BMF range of 10^1000 Hz. The highest BMFs were clustered in a sector approximately in the middle and lateral third of a BF frequency lamina. Lower BMFs were found as one moved away from the high BMF cluster. Put another way, `iso-BMF contours' were arrayed in concentric rings around the high BF region, as displayed in Fig. 14. Topographic clusters of units with similar latencies, Q10 values and binaural properties were also found within an isofrequency lamina. 5.4. Binaural processing enhances AM encoding Batra and colleagues have made signi¢cant contribu-

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

17

Fig. 14. Topographic organization of AM BMFs within two iso-frequency bands in the inferior colliculus. Three-dimensional projections (left column) and two-dimensional contour plots (right column) are displayed for frequency band laminae of 3 þ 0.5 kHz (n = 36, upper row) and 12 þ 1 kHz (n = 39, lower row). For the three-dimensional plots (left column), the elevation of the gridded surface is proportional to BMF obtained at or near those line intersections marked by a dot. For the two-dimensional plots (right column) the contours connect sites of neuronal responses exhibiting the same BMF. The frequency change between consecutive iso-BMF contours is 100 Hz. c, caudal; l, lateral; m, medial; r, rostral. Figure reproduced from Schreiner and Langner (1988, ¢gure 4, p. 1827), with permission of the author and publisher.

tions to understanding the binaural processing of AM in the IC. Batra et al. (1989) in their initial AM paper dealt primarily with an issue that others have dealt with before, namely the highest frequency to which AM sensitivity occurs, however they were the ¢rst to use an unanesthetized (rabbit) preparation, with emphasis on use of dichotic stimuli. Because they were dealing with interaural time delay (ITD) sensitivity, they were able to compare the (AM) frequencies at which synchrony to the envelope occurred to the frequencies at which sensitivity to ITDs for AM take place. A major discovery was that, like the previous investigations, synchrony to envelopes was concentrated at low AM frequencies, with most responses below about 90 Hz (mean BF), whereas sensitivity to AM ITDs extended to much higher AM frequencies: at least 350 Hz. Lastly, because they were able to extract the synchrony from a binaural stimulus, Batra and colleagues assessed synchrony to both contralateral and ipsilateral AM. Both of these declined at about the same frequency. In their next e¡ort, Batra et al. (1993) dealt more with sound localization issues related to temporal processing in their unanesthetized preparation. They argued that if you examine only responses above V250

Hz in the IC, then you ¢nd the same peak-type and trough-type binaural neurons that are observed in the SOC. In the IC, the peak types favored ipsilateral delays and the trough types showed no preference, and the characteristic delays fell within the behaviorally relevant range ( þ 300 Ws). At lower modulation frequencies, the responses tend to be irregular and suggestive of the convergence of other binaural in£uences from the lower auditory centers. About half of the peak-type neurons are EI, suggesting that these neurons use temporal cues to localize sounds at high modulation frequencies, but intensity cues when the temporal £uctuations are absent, weak or at a low AM frequency. Heil et al. (1995) made a comprehensive study of AM coding during development in the IC for newborn anesthetized gerbils, ranging in age from 13 to 160 days after birth (DAB). They found for single units that thresholds and latencies for BF tones decreased exponentially with age, and that sync and rate BMFs for AM tones increased exponentially with age, reaching adult values at about 40 DAB. Modulation transfer functions were lowpass or bandpass in shape, with the proportion of lowpass declining with age. BMFs not only increased with age, as shown in Fig. 15, but the

HEARES 3704 1-8-01

18

R.D. Frisina / Hearing Research 158 (2001) 1^27

Fig. 15. During the ¢rst month of postnatal development, BMFs at the level of the inferior colliculus have peaks that shift to higher and higher frequencies with age, and show a broader range. Rate BMFs (left column) shift to slightly higher AM frequencies than sync BMFs (right column). Figure reproduced from Heil et al. (1995, ¢gure 5, p. 371), with permission of the author and publisher.

range expanded as well. Lastly, supporting previous ¢ndings in the IC of cat and guinea fowl described above, a topographic arrangement of BMFs was discovered for the gerbil, as displayed in Fig. 16. Although in the present review details of the neural bases of echolocation cannot be covered, it has been found that echolocating mammals such as bats also show robust temporal processing, including AM and gap coding, at the level of the IC, a processing task

that is partly determined by the spatiotemporal interplay of excitatory and inhibitory inputs to IC principal neurons (Lesser et al., 1990; Grothe et al., 1996). Along these lines, Lu et al. (1997, 1998), utilizing the unanesthetized brown bat preparation with multibarrel electrodes, demonstrated that Q-aminobutyric acid (GABA) inhibition plays an important role in shaping IC neuron responses to temporally dynamic sound features such as gaps and stimulus modulations.

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

19

Fig. 16. AM coding, whether measured in terms of rate (left) or in terms of the synchronous response (right), has a topographic organization in the gerbil IC during the ¢rst month of postnatal development and into adulthood. Low BMFs tend to be located medially and higher BMFs laterally. Figure reproduced from Heil et al. (1995, ¢gure 5, p. 371), with permission of the author and publisher.

6. Gap encoding and forward masking 6.1. Auditory nerve The response of an auditory nerve ¢ber is reduced by the presence of a preceding stimulus (masking stimulus) even if a gap in time is present between the two sounds. This e¡ect, which involves a reduction in spike ¢ring rate during the gap between the sounds, is related to the amount of adaptation occurring at the hair cell/spiral ganglion neuron synapse induced by the masking stimulus (Smith, 1979 ; Harris and Dallos, 1979; Westerman and Smith, 1984, 1987). The recovery from adaptation can be described by a time constant measured by varying the gap width, which alters the amount of forward masking (Smith and Brachman, 1982). Relkin and Turner (1988), using a novel dual interval neurophysiological paradigm employing detection theory, demonstrated that when presented with BF tones, chinchilla auditory nerve ¢bers can encode gaps in a way superior to human psychophysical performance. These results implied that human behavioral forward masking (gap processing) abilities are limited by suboptimal processing in the central auditory system. The abilities of mammalian auditory nerve ¢bers to encode temporal gaps in wideband noise stimuli have been investigated in the chinchilla, a rodent with excellent low frequency hearing and possessing speech processing capabilities (Zhang et al., 1990). Taking into account reductions in ¢ring rate during the gap and increases in spike counts immediately after the gap, Zhang et al. found that gap encoding improved with gap duration and sound level (30^80 dB SPL). Many units had gap thresholds as low as 2 ms. Furthermore,

gap thresholds decreased with ¢ber BF. Quantitative agreement was obtained between gap coding by single neurons here and gap encoding for chinchillas measured psychophysically (Giraudi et al., 1980 ; Salvi and Arehole, 1985). A similar study in the avian auditory nerve (starling: Klump and Gleich, 1991) for the most part yielded similar results as seen for mammals. For example, the best minimum gap thresholds in some units were as low as 1.6 ms, and improvements in gap thresholds were seen with elevations in sound level. However, unlike the chinchilla, no relation was observed between gap thresholds and ¢ber BF. Like Salvi's group for the chinchilla, Klump and Gleich found a correspondence between psychophysical minimal gap thresholds, of the order of 1.8 ms as measured by Klump and Maier (1989) for starlings, and the best gap thresholds for single auditory nerve ¢bers. 6.2. Ventral cochlear nucleus Boettcher et al. (1990) examined response of cochlear nucleus units in the anesthetized chinchilla to a pair of BF tones of varying amplitudes and separated by di¡erent gap widths. They found that recovery of the response to the probe tone was exponential in form and the responses of primarylike, primarylike-with-notch and chopper units were similar to auditory nerve ¢bers. The coding of gaps by onset and pauser/buildup units was very di¡erent from their ascending inputs from the auditory nerve, probably due to the more complex excitatory/inhibitory circuitry involved in determining their responses to temporal features of sound. Shore (1995) examined tone burst gap encoding abil-

HEARES 3704 1-8-01

20

R.D. Frisina / Hearing Research 158 (2001) 1^27

Fig. 17. E¡ects of GABAA antagonist (bicuculline, BMI) on probe recovery functions in an AVCN onset-chopper unit. BMI application increased responses to the probe, as exempli¢ed in the PSTHs shown here for the 2, 4 and 8 ms probe delays. Masking returned to normal following cessation of drug delivery. Pa, unmasked response to the probe; Pm, response to the probe when preceded by a masker. Figure reproduced from Backo¡ et al. (1997, ¢gure 3, p. 160), with permission of the author and publisher.

ities (forward masking) of many VCN unit types in anesthetized guinea pigs. Primarylike responses were very similar to those of auditory nerve ¢bers, and displayed the longest time constants of any VCN units for recovery of responses to the second burst (probe: the stimulus following the gap between the two stimuli). Masking of the response to the probe was maximal for the ¢rst 2 ms of the response (onset ¢ring rate), and diminished as spikes were counted over longer time periods (steady state ¢ring rate). Shorter masking

tones produced more di¡erences in the responses of the various VCN unit types to the probe tones. High spontaneous VCN units were more resistant to masking than units with lower spontaneous rates. Non-primarylike units showed nonmonotonic relationships between the ¢ring rate evoked by the masker compared to the probe response decrement, suggesting that both adaptation and inhibition are in play to produce greater probe response decrements in the VCN compared to the auditory nerve for a given gap width. When consid-

Fig. 18. Schematic diagram of the two types of forward masking e¡ects observed in the DCN. The type A e¡ect was characterized by an immediate suppression of the probe tone response (as is present in the auditory nerve). The type B e¡ect was characterized by probe tone suppression that only became apparent after a well-de¢ned time interval following the masker o¡set (not present in the periphery). Each trace of 0^3 spikes represents a cochlear nucleus unit recording during the response window (W), as described in the sketch at the top. vT represents the time delay (gap) between the masking tone (conditioner, 40 ms) and probe tone (10^40 ms). Rise/fall times = 3 ms, sound levels = 72^80 dB SPL, and carrier frequencies were at a unit's BF. Figure reproduced from Kaltenbach et al. (1993, ¢gure 1, p. 36), with permission of the author and publisher.

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

ered along with results reviewed above concerning gap encoding in the auditory nerve, such as those of Relkin and Turner (1988), it appears that gap thresholds in the VCN are closer to human psychophysical gap thresholds than is the case for auditory nerve ¢bers. Shore (1998) went on to test the hypothesis that the descending, e¡erent pathways into the cochlear nucleus can signi¢cantly a¡ect gap processing in the VCN. She did this by recording from single units 60 min following the cutting of the centrifugal pathways lying medial to the cochlear nucleus. She found that in primarylike, sustained choppers and onset units, removing descending inputs caused a reduction in forward masking, i.e. enhanced recovery of the response to the probe. In contrast, e¡erent lesions resulted in increased masking for primarylike-with-notch and other chopper units. Backo¡ et al. (1997) explored the role of inhibitory inputs in shaping gap encoding in the cochlear nucleus using multibarrel electrodes ¢lled with GABA and glycine blockers in the anesthetized chinchilla. They employed identical clicks as the masking and probe stimuli, and varied the gap duration systematically. Blockade of GABAergic and glycinergic receptors by iontophoretic application of bicuculline and strychnine, respectively, shortened forward masking recovery times in two thirds of the units tested, as exempli¢ed in Fig. 17. Conversely, iontophoresis of inhibitory neurotransmitter agonists increased recovery times, reinforcing the importance of inhibitory circuitry in determining temporal processing abilities of cochlear nucleus cells. Burkard and Palmer (1997) examined responses of VCN choppers in the urethane anesthetized guinea pig to click trains in quiet and background noise, where the gap between successive clicks and the intensity levels of the stimuli were varied. They discovered that the latency of the response to the clicks declined with intensity, that suppression of the response to the background noise occurred after each click, that high levels of background noise abolished the click response but low levels of the noise enhanced responses to the clicks, and for very short interclick gaps the PSTH response pattern resembled that for a tone burst, i.e. the natural chopping frequency was unrelated to any periodicity in the stimulus. 6.3. Dorsal cochlear nucleus Kaltenbach et al. (1993) investigated adaptation and inhibition in DCN units of the anesthetized hamster using a forward masking (gap) paradigm where the frequency of the masking tone was varied and that of the probe was held constant at a unit's BF. Two patterns of forward masking suppression were discovered: for type A, the suppression of the response to the probe became apparent immediately following the masker o¡set, and

21

the e¡ective frequency bandwidth for suppression narrowed as the gap between the masker and probe increased. For type B, suppression of the probe response did not appear until masker/probe gaps became quite long, and the relation between the gap duration and e¡ective bandwidth was nonmonotonic. These two types of forward masking responses are displayed in Fig. 18. Type B forward masking is quite di¡erent from that in the auditory nerve, and comes from DCN neural circuitry which could subserve certain characteristics of echo suppression (Kaltenbach et al., 1993), including the precedence e¡ect, as originally investigated in vitro by Wickesberg and Oertel (1990) using electrical stimulation. Utilizing the anesthetized chinchilla, Palombi et al. (1994) also examined responses of DCN units to gaps,

Fig. 19. Gap coding functions for single units at the level of the IC take the same functional form as behavioral data. It is therefore possible that these IC units are part of a neural pathway in the central auditory system subserving gap encoding. Behavioral mean data come from ¢ve animals utilizing an inhibition of the acoustic startle response paradigm. The behavioral data are normalized to the magnitude of the startle response with no inhibiting stimulus. Therefore, as the inhibiting stimulus becomes more perceptible, the startle response declines. The phasic unit (onset, n = 13) and the sustained unit (primarylike, n = 3) functions were measured in the same mice that the behavioral data were obtained from, using an unanesthetized neurophysiological preparation. The normalization point for the physiological responses is the spike rate in response to the probe stimulus when no gap is present. For sustained units, the normalized response is the ¢ring rate decrease in response to the gap, normalized to the response with no gap. For transient units, the normalized response is the reciprocal of the ¢ring rate increase in response to the gap, normalized to the response with no gap. Figure reproduced here from Walton et al. (1997, ¢gure 12, p. 173), with permission of the author and publisher.

HEARES 3704 1-8-01

22

R.D. Frisina / Hearing Research 158 (2001) 1^27

in this case focusing on pauser/buildup units that correspond to the pyramidal/fusiform neurons (a major output cell of the DCN). Their main ¢nding was that many neurons showed facilitation of the probe response at short gap intervals, much like the behavior of the type B units just described for the study by Kaltenbach and colleagues. Palombi and coworkers noted that when the probe response is facilitated by the `masker' at short gap widths, the probe ¢ring pattern showed a reduced interspike interval and more of a chopper temporal response pattern. 6.4. Similarity of behavioral and single unit processing for gaps in the inferior colliculus Walton et al. (1997) combining psychological and electrophysiological measures of auditory temporal processing demonstrated that the gap encoding capabilities of IC single neurons were very similar to the behavioral gap functions. Data derived from an inhibition of startle re£ex behavioral paradigm and single unit physiological recordings from the IC of unanesthetized mice are presented in Fig. 19. Note the similarities in the shapes of the gap functions, suggesting that IC units of these types could subserve a portion of the central auditory system neural circuitry responsible for the behavior. Barsz et al. (1998), using the same unanesthetized mouse preparation for single unit recordings of IC gap processing, extended these ¢ndings to show that increases in stimulus rise/fall time, such as those occurring in natural sounds such as speech, result in : (1) increased gap thresholds, (2) larger ¢rst spike la-

Fig. 21. Temporal processing at the single unit level degrades in aging animals. This ¢gure shows how the strength of single unit gap responses of the unanesthetized mouse IC become weaker in old animals. Neural recovery functions from young adults (open data points) show signi¢cant recovery at short gap widths and often show responses that exceed 100%. Old mice (¢lled data points) need longer gap widths for the response of the probe to approach that of the masker, and very few units show enhanced gap responses that exceed 100%. Percent recovery is the ¢ring rate in response to the probe, divided by the onset response to the masker, both of which were wideband noise bursts. Figure reproduced from Walton et al. (1998, ¢gure 9, p. 2772), with permission of the author and publisher.

tency variations, and (3) signi¢cant changes in response strength for a given gap width. As part of the quest to determine the neural mechanisms for this IC gap processing at anatomical levels, Frisina et al. (1997a, 1998) examined the neural pathway connections of the IC region recorded from by Walton and colleagues. Further details on this are given next when considering how gap processing changes with age. Fig. 20. As animals age, temporal processing at the single unit level degrades, commensurate with behavioral declines in auditory temporal coding. The histogram shows how the distribution of single unit gap thresholds of the unanesthetized mouse IC shift to longer gap widths for old animals. Hatched bars are from young adults (n = 78) and ¢lled bars are for old mice (n = 108). MGT, minimum gap threshold. Figure reproduced from Walton et al. (1998, ¢gure 8, p. 2771), with permission of the author and publisher.

6.5. Temporal coding declines with age in the auditory midbrain Presbycusis ^ age related hearing loss ^ is a ubiquitous decline in auditory sensitivity that occurs in all mammals. The particular rate and nature of age related hearing loss experienced by a particular person or ani-

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27

23

Fig. 22. Calbindin immunoreactivity declines with age in the IC for both CBA and C57 mouse strains (top histograms). However, calretinin upregulates with age in CBA mice, who have a minimal age related hearing loss (bottom left histogram). In contrast, the immunopresence of calretinin remains the same in C57 mice who are profoundly deaf by old age (bottom right histogram). *P 6 0.05; **P 6 0.01; error bars denote S.E.M. CB, calbindin-labeled neurons; count, number of immunoreactive neurons in the dorsomedial IC; CR, calretinin-labeled neurons; DC, dorsal cortex of IC; LN, lateral nucleus of IC; NCO, commissural nucleus of IC. Figure reproduced from Zettel et al. (1997, ¢gure 12, p. 109), with permission of the author and publisher.

mal depends upon many factors including species, genetic complement, history of noise exposure, extent of chemical insults including alcohol, chemotherapeutic agents and antibiotics, and other factors. One of the challenges for auditory neuroscientists interested in examining the etiologies of presbycusis is to separate the e¡ects of age related cochlear pathologies from those of the central auditory system. In one such set of related studies, Walton et al. (1998) examined central aging changes neurophysiologically and Ison et al. (1998) looked at the situation from the behavioral perspective. Using the same unanesthetized CBA mouse preparation mentioned above for single unit studies of the IC and the behavioral paradigms, these investigators demonstrated that auditory temporal processing abilities for gaps decline with age behaviorally and neurophysiologically. Speci¢cally, Walton et al. (1998) discovered that the number of single units with short gap thresholds

declined signi¢cantly in aged animals, as displayed in Fig. 20. In addition, the strength of response to encode the gaps was drastically impaired with age, as presented in Fig. 21. As part of this investigative group's e¡orts to determine the neural bases for these age related temporal processing declines, Frisina et al. (1995, 1998) found that signi¢cant declines occurred in certain pathways to the CBA mouse IC region studied by Walton and colleagues, including inputs from the contralateral cochlear nucleus and the ipsilateral anterolateral periolivary nucleus. Zettel et al. (1997), utilizing immunocytochemical procedures to identify the presence of calcium binding proteins, found that for the IC region studied by Walton and coworkers, signi¢cant age related declines in calbindin took place, whereas there was a noteworthy upregulation of calretinin in this same area of IC with age (Fig. 22).

HEARES 3704 1-8-01

24

R.D. Frisina / Hearing Research 158 (2001) 1^27

7. Conclusions and future research Neural mechanisms for processing of key temporal features of sound reviewed here give some insights into and understanding of how single neurons encode sound timing elements progressively as one ascends the auditory system from the auditory nerve through the auditory brainstem. For example, the fundamental means of coding sound gaps at the auditory nerve level is via a cessation of activity in response to the gap. The neural mechanisms for single neuron gap encoding higher up in the brainstem, as auditory units become more transient in their responses to sounds, switches to an increase in spike ¢ring rate at the conclusion of the gap. From ¢ndings such as these that have been reviewed in this article, it is evident that much basic knowledge has been gleaned from sub-primate mammalian neurophysiological and behavioral experiments in the past 30 years. Application of this knowledge to clinical problems of hearing impairment and presbycusis requires more e¡ort in the areas of multidisciplinary experimental studies utilizing animal models with hearing loss. Our psychoacoustic and research audiology colleagues have accomplished more in terms of understanding the e¡ects of hearing impairment and aging on auditory temporal processing (Moore et al., 1992; Fitzgibbons and Gordon-Salant, 1996; Snell, 1997; Snell and Frisina, 2000). It is time for auditory neuroscientists to emphasize explorations of the e¡ects of hearing loss and aging in their neurophysiological, anatomical, genetic, immunocytochemical and molecular biology studies of temporal sound-feature coding in the peripheral and central auditory systems. Acknowledgements Thanks to Dr. Ranjan Batra and two anonymous reviewers for helpful critiques. Thank you to Patricia Bardeen for administrative support. Work supported by the National Institute on Aging at NIH, Grant P01 AG09524 ; and the International Center for Hearing and Speech Research, National Technical Institute for the Deaf, Rochester (NY) Institute of Technology.

References Backo¡, P.M., Shadduck Palombi, P., Caspary, D.M., 1997. Glycinergic and GABAergic inputs a¡ect short-term suppression in the cochlear nucleus. Hear. Res. 110, 155^163. Banks, M.B., Sachs, M.B., 1991. Regularity analysis in a compartmental model of chopper units in the anteroventral cochlear nuclei. J. Neurophysiol. 64, 606^629. Barsz, K., Benson, P.K., Walton, J.P., 1998. Gap encoding by inferior

collicular neurons is altered by minimal changes in signal envelope. Hear. Res. 115, 13^26. Batra, R., Kuwada, S., Stanford, T.R., 1989. Temporal coding of envelopes and their interaural delays in the inferior colliculus of the unanesthetized rabbit. J. Neurophysiol. 61, 257^268. Batra, R., Kuwada, S., Stanford, T.R., 1993. High-frequency neurons in the inferior colliculus that are sensitive to interaural delays of amplitude-modulated tones: Evidence for dual binaural in£uences. J. Neurophysiol. 70, 64^80. Batra, R., Kuwada, S., Fitzpatrick, D.C., 1997a. Sensitivity to interaural temporal disparities of low-and high-frequency neurons in the superior olivary complex. I. Heterogeneity of responses. J. Neurophysiol. 78, 1222^1236. Batra, R., Kuwada, S., Fitzpatrick, D.C., 1997b. Sensitivity to interaural temporal disparities of low-and high-frequency neurons in the superior olivary complex. II. Coincidence detection. J. Neurophysiol. 78, 1237^1247. Boettcher, F.A., Salvi, R.J., Saunders, S.S., 1990. Recovery from short-term adaptation in single neurons in the cochlear nucleus. Hear. Res. 48, 125^144. Burkard, R., Palmer, A.R., 1997. Responses of chopper units in the ventral cochlear nucleus of the anaesthetized guinea pig to clicksin-noise and click trains. Hear. Res. 110, 234^250. Cai, Y., McGee, J., Walsh, E.J., 2000a. Contributions of ion conductances to the onset responses of octopus cells in the ventral cochlear nucleus: Simulation results. J. Neurophysiol. 83, 301^314. Cai, Y., McGee, J., Walsh, E.J., 2000b. Processing of pitch information in complex stimuli by a model of octopus cells in the cochlear nucleus. In: Greenberg, S., Slaney, M. (Eds.), Computational Models of Auditory Function. IOS Press, pp. 1^12. Cant, N.B., Morest, D.K., 1984. The structural basis for stimulus coding in the cochlear nuclei of the cat. In: Berlin, C. (Ed.), Hearing Sciences. College Hill Press, San Diego, CA, pp. 374^ 422. Cooper, N.P., Robertson, D., Yates, G.K., 1993. Cochlear nerve ¢ber responses to amplitude-modulated stimuli: Variations with spontaneous rate and other response characteristics. J. Neurophysiol. 70, 370^386. Evans, E.F., Palmer, A.R., 1980. Dynamic range of cochlear nerve ¢bers to amplitude modulated tones. J. Physiol. 298, 33P^34P. Ferragamo, M., Golding, N.L., Oertel, D., 1998. Synaptic inputs to stellate cells in the ventral cochlear nucleus. J. Neurophysiol. 79, 51^63. Fitzgibbons, P.J., Gordon-Salant, S., 1996. Auditory temporal processing in elderly listeners. J. Am. Acad. Audiol. 7, 183^189. Frisina, D.R., Frisina, R.D., 1997. Speech recognition in noise and presbycusis: Relations to possible neural sites. Hear. Res. 106, 95^ 104. Frisina, R.D., Chamberlain, S.C., Brachman, M.L., Smith, R.L., 1982. Anatomy and physiology of the gerbil cochlear nucleus: an improved surgical approach for microelectrode studies. Hear. Res. 6, 259^275. Frisina, R.D., Smith, R.L., Chamberlain, S.C., 1985. Di¡erential encoding of rapid changes in sound amplitude by second-order auditory neurons. Exp. Brain Res. 60, 417^422. Frisina, R.D., O'Neill, W.E., Zettel, M.L., 1989. Functional organization of mustached bat inferior colliculus: II. Connections of the FM2 region. J. Comp. Neurol. 284, 85^107. Frisina, R.D., Smith, R.L., Chamberlain, S.C., 1990a. Encoding of amplitude modulation in the gerbil cochlear nucleus: I. A hierarchy of enhancement. Hear. Res. 44, 99^122. Frisina, R.D., Smith, R.L., Chamberlain, S.C., 1990b. Encoding of amplitude modulation in the gerbil cochlear nucleus: II. Possible neural mechanisms. Hear. Res. 44, 123^142. Frisina, R.D., Walton, J.P., Karcich, K.J., 1993. Di¡erential abilities

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27 to extract sound-envelope information by auditory nerve and cochlear nucleus neurons. In: Verrillo, R.T. (Ed.), Sensory Research Multimodal Perspectives. Erlbaum, Hillsdale, NJ, pp. 151^175. Frisina, R.D., Walton, J.P., Karcich, K.J., 1994. Dorsal cochlear nucleus single neurons can enhance temporal processing capabilities in background noise. Exp. Brain Res. 102, 160^164. Frisina, R.D., Walton, J.P., Lynch-Armour, M.A., Byrd, J.D., 1995. Aging-related changes in connections to a functionally-characterized region of the inferior colliculus of the CBA mouse model of presbycusis. Soc. Neurosci. Abstr. 21, 908. Frisina, R.D., Karcich, K.J., Tracy, T.C., Sullivan, D.M., Walton, J.P., 1996. Preservation of amplitude modulation coding in the presence of background noise by chinchilla auditory-nerve ¢bers. J. Acoust. Soc. Am. 99, 475^490. Frisina, R.D., Walton, J.P., Lynch-Armour, M.A., Klotz, D.A., 1997a. E¡erent projections of a physiologically characterized region of the inferior colliculus of the young adult CBA mouse. J. Acoust. Soc. Am. 101, 2741^2753. Frisina, R.D., Wang, J., Byrd, J.D., Karcich, K.J., Salvi, R.J., 1997b. Enhanced processing of temporal features of sounds in background noise by cochlear nucleus single neurons. In: Syka, J. (Ed.), Acoustical Signal Processing in the Central Auditory System. Plenum Press, New York, pp. 109^125. Frisina, R.D., Walton, J.P., Lynch-Armour, M.A., Byrd, J.D., 1998. Inputs to a physiologically characterized region of the inferior colliculus of the young adult CBA mouse. Hear. Res. 115, 61^81. Giraudi, D., Salvi, R.J., Henderson, D., Hamernik, R., 1980. Gap detection by the chinchilla. J. Acoust. Soc. Am. 68, 802^806. Grothe, B., 1994. Interaction of excitation and inhibition in processing of pure tone and amplitude modulated stimuli in the medial superior olive of the mustached bat. J. Neurophysiol. 71, 706^ 721. Grothe, B., Covey, E., Casseday, J.H., 1996. Spatial tuning of neurons in the inferior colliculus of the big brown bat: e¡ects of sound level, stimulus type and multiple sound sources. J. Comp. Physiol. A 179, 89^102. Grothe, B., Park, T.J., Schuller, G., 1997. Medial superior olive in the free-tailed bat: response to pure tones and amplitude-modulated tones. J. Neurophysiol. 77, 1553^1565. Hafter, E.R., Dye, R.H., Jr., Wenzel, E.M., Knecht, K., 1990. The combination of interaural time and intensity in the lateralization of high-frequency complex signals. J. Acoust. Soc. Am. 87, 1702^ 1708. Harris, D.M., Dallos, P., 1979. Forward masking of auditory nerve ¢ber responses. J. Neurophysiol. 42, 1083^1107. Heil, P., Schulze, H., Langner, G., 1995. Ontogenetic development of periodicity coding in the inferior colliculus of the Mongolian gerbil. Audit. Neurosci. 1, 363^383. Henry, K.R., 1998. Cochlear tuning in the gerbil: A comparison of responses to sinusoidal amplitude modulation and di¡erence tone stimuli. Audiology 37, 262^277. Ison, J.R., Agrawal, P., Pak, J., Vaughn, W.J., 1998. Changes in temporal acuity with age and with hearing impairment in the mouse: A study of the acoustic startle re£ex and its inhibition by brief decrements in noise level. J. Acoust. Soc. Am. 104, 1696^1804. Javel, E., 1980. Coding of AM tones in the chinchilla auditory nerve: Implications for the pitch of complex tones. J. Acoust. Soc. Am. 68, 133^146. Joris, P.X., 1996. Envelope coding in the lateral superior olive. II. Characteristic delays and comparison with responses in the medial superior olive. J. Neurophysiol. 76, 2137^2156. Joris, P.X., Smith, P.H., 1998. Temporal and binaural properties in dorsal cochlear nucleus and it output tract. J. Neurosci. 18, 10157^ 10170.

25

Joris, P.X., Yin, T.C.T., 1992. Responses to amplitude-modulated tones in the auditory nerve of the cat. J. Acoust. Soc. Am. 91, 215^232. Joris, P.X., Yin, T.C.T., 1995. Envelope coding in the lateral superior olive. I. Sensitivity to interaural time di¡erences. J. Neurophysiol. 73, 1043^1062. Kaltenbach, J.A., Meleca, R.J., Falzarano, P.R., Myers, S.F., Simpson, T.H., 1993. Forward masking properties of neurons in the dorsal cochlear nucleus: Possible role in the process of echo suppression. Hear. Res. 67, 35^44. Kim, D.O., Sirianni, J.G., Chang, S.O., 1990. Responses of DCNPVCN neurons and auditory nerve ¢bers in unanesthetized decerebrate cats to AM and pure tones: analysis with autocorrelation/ power spectrum. Hear. Res. 45, 95^113. Klump, G.M., Gleich, O., 1991. Gap detection in the European starling (Sturnus vulgaris). III. Processing in the peripheral auditory system. J. Comp. Physiol. A 168, 469^476. Klump, G.M., Maier, E.H., 1989. Gap detection in the starling (Sturnus vulgaris): I. Psychophysical thresholds. J. Comp. Physiol. A. 164, 531^539. Kuwada, S., Batra, R., 1999. Coding of sound envelopes by inhibitory rebound in neurons of the superior olivary complex in the unanesthetized rabbit. J. Neurosci. 19, 2273^2287. Langner, G., 1981. Neuronal mechanisms of pitch analysis in the time domain. Exp. Brain Res. 44, 450^454. Langner, G., 1983. Evidence for neuronal periodicity detection in the auditory system of the guinea fowl: Implications for pitch analysis in the time domain. Exp. Brain Res. 52, 333^355. Langner, G., 1992. Periodicity coding in the auditory system. Hear. Res. 60, 115^142. Langner, G., 1997. Relative and absolute pitch perception explained by common neuronal mechanisms. In: Syka, J. (Ed.), Acoustical Signal Processing in the Central Auditory System. Plenum Press, New York, pp. 57^64. Langner, G., Schreiner, C.E., 1988. Periodicity coding in the inferior colliculus of the cat. I. Neuronal mechanisms. J. Neurophysiol. 60, 1799^1822. Langner, G., Schreiner, C., Merzenich, M.M., 1987. Covariation of latency and temporal resolution in the inferior colliculus of the cat. Hear. Res. 31, 197^202. Lesser, H.D., O'Neill, W.E., Frisina, R.D., Emerson, R.C., 1990. ONOFF units in the mustached bat inferior colliculus are selective for transients resembling `acoustic glint' from £uttering insect targets. Exp. Brain Res. 82, 137^148. Liberman, M.C., 1978. Auditory-nerve response from cats raised in a low-noise chamber. J. Acoust. Soc. Am. 63, 442^455. Lu, Y., Jen, P.H.S., Zheng, Q.Y., 1997. GABAergic disinhibition changes the recovery cycle of bat inferior collicular neurons. J. Comp. Physiol. A 181, 331^341. Lu, Y., Jen, P.H.S., Wu, M., 1998. GABAergic disinhibition a¡ects responses of bat inferior collicular neurons to temporally patterned sound pulses. J. Neurophysiol. 79, 2303^2315. Meininger, V., Pol, D., Derer, P., 1986. The inferior colliculus of the mouse. A Nissl and Golgi study. Neuroscience 17, 1159^1179. MÖller, A.R., 1972. Coding of amplitude and frequency modulated sounds in the cochlear nucleus of the rat. Acta Physiol. Scand. 86, 223^238. MÖller, A.R., 1973. Statistical evaluation of the dynamic properties of cochlear nucleus units using stimuli modulated with pseudorandom noise. Brain Res. 57, 443^456. MÖller, A.R., 1974a. Coding of amplitude and frequency modulated sounds in the cochlear nucleus. Acustica 31, 202^299. MÖller, A.R., 1974b. Responses of units in the cochlear nucleus to sinusoidally amplitude-modulated tones. Exp. Neurol. 45, 104^ 117.

HEARES 3704 1-8-01

26

R.D. Frisina / Hearing Research 158 (2001) 1^27

MÖller, A.R., 1975a. Dynamic properties of excitation and inhibition in the cochlear nucleus. Acta Physiol. Scand. 93, 442^454. MÖller, A.R., 1975b. Latency of unit responses in cochlear nucleus determined in two di¡erent ways. J. Neurophysiol. 38, 812^821. MÖller, A.R., 1976a. Dynamic properties of primary auditory ¢bers compared with cells in the cochlear nucleus. Acta Physiol. Scand. 98, 157^167. MÖller, A.R., 1976b. Dynamic properties of excitation and 2-tone inhibition in the cochlear nucleus studied using amplitude modulated tones. Exp. Brain Res. 25, 307^321. MÖller, A.R., 1976c. Dynamic properties of the responses of single neurones in the cochlear nucleus of the rat. J. Physiol. 259, 63^82. MÖller, A.R., Rees, A., 1986. Dynamic properties of the responses of single neurons in the inferior colliculus of the rat. Hear. Res. 24, 203^215. Moore, B.C.J., Peters, R.W., Glasberg, B.R., 1992. Detection of temporal gaps in sinusoids by elderly subjects with and without hearing loss. J. Acoust. Soc. Am. 92, 1923^1932. Oertel, D., 1985. Use of brain slices in the study of the auditory system: spatial and temporal summation of synaptic inputs in cells in the anteroventral cochlear nucleus of the mouse. J. Acoust. Soc. Am. 78, 328^333. Oertel, D., 1997. Encoding of timing in the brain stem auditory nuclei of vertebrates. Neuron 19, 959^962. Oertel, D., 1999. The role of timing in the brain stem auditory nuclei of vertebrates. Annu. Rev. Physiol. 61, 497^519. O'Neill, W.E., Frisina, R.D., Gooler, D.M., 1989. Functional organization of mustached bat inferior colliculus: I. Representation of FM frequency bands important for target ranging revealed by 14 C-2-Deoxyglucose autoradiography and single unit mapping. J. Comp. Neurol. 284, 60^84. Oliver, D.L., Huerta, M.F., 1992. Inferior and superior colliculi. In: Webster, D.P., Popper, A.N., Fay, R.R. (Eds.), The Mammalian Auditory Pathway, Neuroanatomy. Springer-Verlag, New York, pp. 268^221. Oliver, D.L., Morest, D.K., 1984. The central nucleus of the inferior colliculus in the cat. J. Comp. Neurol. 222, 237^264. Palmer, A.R., 1982. Encoding of rapid amplitude £uctuations by cochlear-nerve ¢bers in the guinea pig. Arch. Otol. Rhinol. Laryngol. 236, 197^202. Palombi, P.S., Backo¡, P.M., Caspary, D.M., 1994. Paired tone facilitation in dorsal cochlear nucleus neurons: A short-term potentiation model testable in vivo. Hear. Res. 75, 175^183. Pollak, G.D., Bodenhamer, R.D., 1981. Specialized characteristics of single units in inferior colliculus of mustache bat: frequency representation, tuning, and discharge patterns. J. Neurophysiol. 46, 605^620. Pollak, G.D., Henson, O.W., Novick, A., 1972. Cochlear microphonic audiograms in the pure tone bat Chilonycteris parnellii parnellii. Science 176, 66^68. Pollak, G.D., Henson, O.W., Johnson, R., 1979. Multiple specializations in the peripheral auditory system of the CF-FM bat, Pteronotus parnellii. J. Comp. Physiol. A 131, 255^266. Rees, A., MÖller, A.R., 1983. Responses of neurones in the inferior colliculus of the rat to AM and FM tones. Hear. Res. 10, 301^ 330. Rees, A., MÖller, A.R., 1987. Stimulus properties in£uencing the responses of inferior colliculus neurons to amplitude-modulated sounds. Hear. Res. 27, 129^143. Rees, A., Palmer, A.R., 1989. Neuronal responses to amplitudemodulated and pure-tone stimuli in the guinea pig inferior colliculus, and their modi¢cation by broadband noise. J. Acoust. Soc. Am. 85, 1978^1994. Relkin, E.M., Turner, C.W., 1988. A re-examination of forward masking in the auditory nerve. J. Acoust. Soc. Am. 84, 584^591.

Rhode, W.S., 1994. Temporal coding of 200% amplitude modulated signals in the ventral cochlear nucleus of cat. Hear. Res. 77, 43^68. Rhode, W.S., Greenberg, S., 1992. Physiology of the cochlear nuclei. In: Popper, A.N., Fay, R.R. (Eds.), The Mammalian Auditory Pathway: Neurophysiology. Springer-Verlag, New York, pp. 94^ 152. Rhode, W.S., Greenberg, S., 1994a. Encoding of amplitude modulation in the cochlear nucleus of the cat. J. Neurophysiol. 71, 1797^ 1825. Rhode, W.S., Greenberg, S., 1994b. Lateral suppression and inhibition in the cochlear nucleus of the cat. J. Neurophysiol. 71, 493^ 514. Rhode, W.S., Smith, P.H., 1986a. Encoding timing and intensity in the ventral cochlear nucleus of the cat. J. Neurophysiol. 56, 261^ 286. Rhode, W.S., Smith, P.H., 1986b. Physiological studies on neurons in the dorsal cochlear nucleus of cat. J. Neurophysiol. 56, 287^307. Salvi, R.J., Arehole, S., 1985. Gap detection in chinchillas with temporary high-frequency hearing loss. J. Acoust. Soc. Am. 77, 1173^ 1177. Schreiner, C.E., Langner, G., 1988. Periodicity coding in the inferior colliculus of the cat. II. Topographical organization. J. Neurophysiol. 60, 1823^1840. Schreiner, C.E., Langner, G., 1997. Laminar ¢ne structure of frequency organization in auditory midbrain. Nature 388, 383^386. Shore, S.E., 1995. Recovery of forward-masked responses in ventral cochlear nucleus neurons. Hear. Res. 82, 31^43. Shore, S.E., 1998. In£uence of centrifugal pathways on forward masking of ventral cochlear nucleus neurons. J. Acoust. Soc. Am. 104, 378^389. Smith, R.L., 1979. Short-term adaptation in single auditory nerve ¢bers: Some poststimulatory e¡ects. J. Neurophysiol. 49, 1098^ 1112. Smith, R.L., Brachman, M.L., 1980a. Response modulation of auditory-nerve ¢bers by AM stimuli: E¡ects of average intensity. Hear. Res. 2, 123^144. Smith, R.L., Brachman, M.L., 1980b. Operating range and maximum response of single auditory-nerve ¢bers. Brain Res. 184, 499^505. Smith, R.L., Brachman, M.L., 1980c. Dynamic response of single auditory-nerve ¢bers: some e¡ects of intensity and time. In: van den Brink, G., Bilsen, F.A. (Eds.), Psychological, Physiological and Behavioural Studies in Hearing. Delft University Press, Delft, pp. 312^319. Smith, R.L., Brachman, M.D., 1982. Adaptation in auditory nerve ¢bers: A revised model. Biol. Cybern. 44, 107^120. Smith, R.L., Brachman, M.D., Frisina, R.D., 1985. Sensitivity of auditory-nerve ¢bers to changes in intensity: a dichotomy between decrements and increments. J. Acoust. Soc. Am. 78, 1310^1316. Snell, K.B., 1997. Age-related changes in temporal gap detection. J. Acoust. Soc. Am. 101, 2214^2220. Snell, K.B., Frisina, D.R., 2000. Relationships among age-related differences in gap detection and word recognition. J. Acoust. Soc. Am. 107, 1615^1626. Suga, N., Jen, P.H.S., 1976. Disproportionate tonotopic representation for processing CF-FM sonar signals in the mustache bat auditory cortex. Science 194, 542^544. Suga, N., Manabe, T., 1982. Neural basis of amplitude-spectrum representation in the auditory cortex of the mustached bat. J. Neurophysiol. 47, 225^255. Suga, N., O'Neill, W.E., Manabe, T., 1978. Cortical neurons sensitive to combinations of information-bearing elements of biosonar signals in the mustached bat. Science 200, 778^781. Trussell, L.O., 1997. Cellular mechanisms for preservation of timing in central auditory pathways. Curr. Opin. Neurobiol. 7, 487^492. Walton, J.P., Frisina, R.D., Ison, J.R., O'Neill, W.E., 1997. Neural

HEARES 3704 1-8-01

R.D. Frisina / Hearing Research 158 (2001) 1^27 correlates of behavioral gap detection in the inferior colliculus of the young CBA mouse. J. Comp. Physiol. A 181, 161^176. Walton, J.P., Frisina, R.D., O'Neill, W.E., 1998. Age-related alteration in processing of temporal sound features in the auditory midbrain of the CBA mouse. J. Neurosci. 18, 2764^2776. Westerman, L.A., Smith, R.L., 1984. Rapid and short-term adaptation in auditory nerve response. Hear. Res. 15, 249^260. Westerman, L.A., Smith, R.L., 1987. Conservation of adapting components in auditory-nerve ¢bers. J. Acoust. Soc. Am. 81, 680^691. Wickesberg, R.E., Oertel, D., 1990. Delayed frequency-speci¢c inhibition in the cochlear nuclei of mice: A mechanism for monaural echo suppression. J. Neurosci. 10, 1762^1768.

27

Yates, G.K., 1981. Frequency dependent e¡ects in auditory nerve response to amplitude modulated pure tones. J. Physiol. 320, 87P. Zeng, F.-G., Oba, S., Garde, S., Sininger, Y., Starr, A., 1999. Temporal and speech processing in auditory neuropathy. NeuroReport 10, 3429^3435. Zettel, M.L., Frisina, R.D., Haider, S., O'Neill, W.E., 1997. Agerelated changes in calbindin D-28k and calretinin immunoreactivity in the inferior colliculus of CBA/CaJ and C57Bl/6 mice. J. Comp. Neurol. 386, 92^110. Zhang, W., Salvi, R.J., Saunders, S.S., 1990. Neural correlates of gap detection in auditory nerve ¢bers of the chinchilla. Hear. Res. 46, 181^200.

HEARES 3704 1-8-01