Abstracts of Presentations at ARO 1999

Abstracts of Presentations at ARO 1999 Midwinter Meeting by HRC Faculty and Students

February 14, 1999

Abstract #88, Date 2/14/99, Session E3, Poster (J56)
Role of apparent signal movement in reducing masking in a sound field
*P.S. Deliwala, G. Kidd, Jr. (Boston University, Boston, MA)

This study examined the effect of apparent movement on signal detection. Signal thresholds were measured in quiet and in the presence of an energetic and an informational masker in a quiet but non-anechoic room. The sounds were presented via 7 speakers separated by 30 degrees, arranged in a semicircle around the subject. Thresholds were measured for 500-Hz and 3000-Hz signals using a 2AFC task.
A multiple-burst paradigm was used in which the signal and the masker were gated on/off synchronously in a sequence of 7 bursts. The informational masker was comprised of 7 tones, the frequencies, phases and levels were chosen randomly on every interval. Masker tones were chosen from a frequency range 3-octaves wide centered around the signal frequency. In order to minimize "energy-based" masking masker tones were not selected from a region near the signal frequency. The 7-tone masker was presented by sending one tone to each of the 7 speakers. Each speaker played the same tone on every burst throughout the sequence. The energetic masker was a Gaussian noise 3-octaves wide and centered around the signal frequency. A different sample of the noise was played from each of the 7 speakers. Each speaker played the same sample on every burst throughout the sequence. The signal was presented in 2 spatial patterns: stationary and moving. In the stationary condition the signal was presented from one of the 7 speakers chosen randomly on every presentation. In the moving condition the signal location was varied - starting from the right and moving to the left or vice versa - during the burst sequence. These two signal conditions were presented as 2 interleaved adaptive tracks. The task of the listener was to identify the interval that had the signal.
The results for 4 subjects show that signal movement did not significantly improve detection in quiet. However, signal movement improved detection in the presence of the energetic masker by 4 dB and in the presence of the informational masker by 12 dB, on average. Results suggest that apparent movement is more effective in reducing informational masking than energetic masking.
Supported by a Dudley Allen Sargent grant from Boston University and the ONR/MURI Z883402 and NIH/NIDCD grants DC03281 and DC00100


Abstract #103, Date 2/14/99, Session E4, Poster (B71)
Localization of near-field sources in a reverberant room
S.G. Santarelli, N. Kopco, *B.G. Shinn-Cunningham (Boston University Department of Cognitive and Neural Systems)

Subjects were asked to point to the position of sound sources within one meter of the head while seated in a medium-sized, echoic classroom. These localization results are compared to results from a previous study that used an identical procedure (but a different set of subjects) in anechoic space. Overall, localization in the reverberant room is worse than observed in the anechoic conditions (a surprising result, given that, for sources near the head, the reverberant energy is weak relative to the direct sound reaching the ears).
In order to understand these results more fully, gross interaural time difference (ITD) and interaural level difference (ILD) cues were estimated for near-field sources in anechoic space using a simple model. This analysis is especially interesting since iso-ITD and iso-ILD surfaces in the near field differ from the "cone-of-confusion" (iso-binaural cue) surface that occurs for sources relatively far from the head. To analyze response errors for each subject, the estimated ITD and ILD values are found for both source and response positions. From this analysis, localization errors are estimated in units of binaural cues (ITD and ILD) and non-binaural cues. This analysis implies that reverberation does not affect interaural time and interaural intensity localization errors in the same way. It also appears that there are consistent subject differences in both binaural abilities and in non-binaural localization abilities, and that these differences capture much of the intersubject variability observed.
Supported by AFOSR grant F49620-98-1-0108


Abstract #105, Date 2/14/99, Session E4, Poster (B73)
Investigation of the relationship between three common measures of precedence: Fusion, localization dominance, and discrimination suppression
*R.Y. Litovsky, B.G. Shinn-Cunningham, A.D. Stein, D.A. Singer (Boston University)

In reverberant environments listeners have a remarkable ability to localize and identify sound sources. The precedence effect (PE) refers to a group of auditory phenomena that are thought to be related to this ability. Traditionally, three measures have been used to quantify the PE: (1) Fusion; at short delays (1-5 ms for clicks) the lead and lag are perceptually fused into one auditory event. (2) Localization dominance; the leading source dominates the perceived location of the fused image. (3) Discrimination Suppression: at short delays, changes in the location or interaural parameters of the lag are difficult to discriminate compared with change in the lead. Little is known about the relation between these aspects of the PE, since they are rarely studied in the same listeners. In the present study, extensive measurements of these phenomena were made in six normal-hearing listeners. Results from all three measures are consistent with a simple model in which lead and lag information interacts perceptually and the strength of the interaction decreases with spatiotemporal separation of the lead and lag. At short delays, lead and lag both contribute to spatial perception, but the lead dominates (to the extent that only one position is even heard). At intermediate delays, two spatial locations are and reverse traveling waves have similar delays for frequencies below CF.
Supported by NSF and NIDCD


Abstract #106, Date 2/14/99, Session E4, Poster (B74)
The effect of expected and unexpected auditory signal location on identification performance in a multi-source environment
*T.L. Arbogast, G. Kidd, Jr. (Boston University, Boston, MA)

Auditory spatial attention is one mechanism which may contribute to a listener's ability to identify a particular sound source in a multi-source environment. The role auditory spatial attention plays in such an environment was investigated using a variation of the probe-signal method (Greenberg and Larkin, 1968, J. Acoust. Soc. Am. 44:1513). The experiment took place in a quiet room with 7 speakers at 30 degree intervals, arranged in a semi-circle in front of the listener. The signal was comprised of 8 contiguous, 60-ms pure-tone bursts arranged in either a rising or falling frequency pattern. The maskers were also comprised of 8 contiguous pure-tone bursts but with durations which varied randomly from 20 to 100 ms. The six maskers were played simultaneously with the signal and were constructed to produce informational, rather than energetic masking. For each masker, frequencies were chosen randomly on each burst from a narrow frequency band, independent from the signal frequency band. The task was 1I-2AFC fixed-level identification. The listener was instructed to focus attention on a speaker specified by the experimenter (expected location) and to identify the signal as quickly as possible, but without sacrificing accuracy. Percent correct identification and response time were measured in two conditions: 1) the signal presented at the expected location in 100% of trials; and 2) the signal presented at the expected location in 75% of trials (Signal trials), and at one of the remaining speakers, chosen randomly, in 25% of trials (Probe trials). Accuracy and response time were compared for expected and unexpected signal location at each speaker. Mean data indicate significantly better identification performance and faster response times for expected signal location, than for unexpected signal location. This suggests that listeners can tune auditory spatial attention to aid in signal perception in a multi-source environment.
Supported by the Dudley Allen Sargent Research Fund from Boston University, ONR/MURI grant Z883402, and NIH/NIDCD grants DC03281 and DC00100


Abstract #110, Date 2/14/99, Session E4, Poster (B78)
Discriminating harmonicity
*G. Kidd, Jr., C.Y.P. Chiu, C.R. Mason (Boston University, Boston, MA)

Simultaneous tones that are harmonically related tend to be grouped perceptually to form a unitary auditory image. A single partial that is mistuned stands out from the other tones, and two harmonic complexes with different fundamental frequencies can readily be perceived as separate auditory objects. These phenomena are often cited as evidence for the strong role of 'harmonicity' in perceptual grouping and segregation of sounds. This study was an attempt to measure the discriminability of harmonicity. Here, the degree of harmonicity refers to how near components fall to expected harmonic intervals. In a 2I2AFC paradigm, the listener's task was to choose the most harmonic sound. Signals and foils were comprised of eleven tones with the lowest component's expected frequency, Fl, chosen randomly on every trial from 200 to 480 Hz.
In the first experiment, the signal was varied from perfectly harmonic (all eleven tones at the expected integer multiples of Fl) to less harmonic by adding frequency jitter to each component (10% of Fl to 40% of Fl). The components in the foils always had the maximum jitter (50% of Fl, referred to as 100% jitter). Performance decreased from near perfect for 0% jitter to near chance for 80% jitter and was influenced by Fl. The decline was not abrupt: for example, with 40% jitter, listeners were still 80% correct on average.
In the second experiment, a masker was presented simultaneously with signals (0% jitter) and foils. The maskers were comprised of 4, 8 or 12 tones having frequencies that were randomly chosen on each presentation from 100 to 5520 Hz. In addition to a monaural condition (SmMm), a dichotic condition (SmMo) was tested. Adding a masker significantly interfered with discrimination. Dichotic presentation of these audible stimuli improved performance by 5-10 dB.
The results of these experiments suggest that the current approach is a useful way to measure the discriminability of an acoustic variable thought to underlie a perceptual property important in grouping and segregation.
Supported by ONR/MURI Z883402, NIH/NIDCD DC03281 and DC00100


Abstract #118, Date 2/14/99, Session E4, Poster (B86)
Failure to train away the precedence effect
*R.Y. Litovsky, M.L. Hawley, B.F. Fligor (Boston University); P.M. Zurek (M.I.T.)

In reverberant environments listeners are able to localize and identify a sound source, regardless of an abundance of reflections. This ability is often studied using the precedence effect. Under headphones, stimuli commonly consist of two pairs of clicks, each with a given ITD. Their onset is separated by several msec to represent a direct sound and a reflection. A common finding is that with short separations between the two click pairs (1-3 msec) just-noticeable differences (JNDs) in the ITD of the lagging stimulus (lag JND) are significantly higher than JNDs for the leading stimulus (lead JND). This difference in ITD JND is thought to reflect the fact that the auditory system attributes greater perceptual weight to directional information contained in the leading source, and reduced weight to information contained in the lagging source. It has been suggested [Saberi and Perrott, J. Acoust. Soc. Amer., 87, 1732-1737, 1990] that listeners can train away high lag JNDs by listening to the task for 12-20 hours. Saberi and Perrott reported that after training with a delay of 2.35 msec, lag JNDs were as good as lead JNDs for all delays tested (0.38 to 10 msec), and thus suggested that precedence may not be "hard wired". Given the importance of this learning effect for understanding the precedence effect, we sought to replicate it.
Eight subjects were trained using similar procedures to those of Saberi and Perrott. One group was trained for 12-22 hours using a method of constant stimuli to vary ITD. A second group was trained for 20 to 31hours using an adaptive procedure to vary ITD. Results suggest that none of our listeners showed sign of training away precedence. After 12- 31 hours of listening, all subjects maintained high lag JNDs and low lead JNDs. Our failure to train away the precedence effect suggests that directional information contained in the lagging source is not always easily accessed, and that learning to utilize that information may depend either on particular experimental procedures that are yet to be identified, or on subject differences. We currently have no explanation for the discrepancy between our results and those of Saberi and Perrott.
Supported by NIH and ONR


Abstract #203, Date 2/14/99, Session E9, Poster (B171)
EarLab: A virtual hearing laboratory
*D.C. Mountain, H.M. Hoffman, V. Vajda, S.G. Deligeorges, A.E. Hubbard (Boston University Hearing Research Center)

EarLab is a web-based resource that combines an on-line data warehouse containing a wide variety of auditory information with a set of integrated computational tools for the manipulation and visualization of this data. The long range goal for EarLab is to increase understanding of the hearing process through improved collaboration between engineers, biologists and behavioral scientists.
A major design objective for EarLab is to create an environment that will enable researchers from all disciplines who work on hearing to share and compare their results electronically. The intent is not to duplicate existing methods for publishing scientific results, but rather to augment these methods by providing tools to compare published and unpublished data from different laboratories, species, and techniques as well as to provide a means to disseminate and integrate large data sets that can only be summarized in conventional publications. Additional design objectives are to provide online computational models that can be run over the web and to integrate EarLab's auditory database system with other bioinformatic and neuroinformatic resources on the world wide web. Examples of these resources include anatomical, bibliographic, genetic, molecular, and Human Brain Project web sites.
Demonstrations of EarLab in action will presented along with some examples of research projects using EarLab.
Supported by ONR, NIDCD, and the Boston University Hearing Research Center

February 15, 1999

Abstract #252, Date 2/15/99, Session J, Symposium , 7:45p
Binaural processing of speech by hearing impaired listeners
*H.S. Colburn, M.L. Hawley (Hearing Research Center and Biomedical Engineering, Boston University)

Binaural processing is known to improve speech intelligibility in noisy environments and to aid in localization of sounds, at least for listeners with normal hearing. The abilities of listeners with hearing impairments to take advantage of binaural processing in these circumstances have been studied much less, and the picture is not completely clear. This presentation reviews what we know about these abilities and attempts to relate performance in these tasks to binaural abilities in isolated tests of binaural interaction such as interaural time discrimination. As a part of this analysis, abilities in intelligibility and localization that are possible with monaural listening, bilateral listening (two ears used separately), and true binaural listening are described. Zurek [1993, chap. in "Acoustical factors affecting hearing aid performance", Allyn and Bacon] has studied the effect of separating a target and a single masker on the intelligibility threshold for speech masked by noise. He showed that the improvement with separation comes from two sources, the head shadow effect that leads to improved signal-to-noise ratio at the more favorably located ear and the "binaural squelch effect" that comes from binaural interaction. With a single masker and target, the head shadow effect for intelligibility threshold can be up to 15 dB and the binaural squelch effect is about 2 to 4 dB. Since the intelligibility of sentence material improves by 10-15% for each decibel of increase in signal-to-noise ratio, small decibel improvements can lead to large intelligibility improvements. In recent studies using multiple competing sources, Hawley et al. [1998, JASA 103(5 Pt. 2):3051] showed that even though overall thresholds were elevated for the listeners with hearing impairments, the overall advantage of separating the location of the target and competing sources is similar for listeners with normal and impaired hearing. In localization studies, good ability requires the comparisons of signals at each ear in order to separate stimulus characteristics from source location characteristics. The cues that are used for localization are different depending on the situation, with interaural time and intensity cues dominating in the horizontal plane and spectral cues dominating in the vertical plane. The abilities of listeners to perform these tasks depend on a variety of factors, including their hearing losses, their binaural abilities, the hearing aids that they wear, and the nature of the interference.
Supported by National Institutes of Deafness and Other Communicative Disorders: DC00100


Abstract #331, Date 2/15/99, Session K4, Poster (B78)
Validating the measured stiffness map: A biophysical model
*R.C. Naidu, D.C. Mountain (Boston University Hearing Research Center, Boston, MA)

Recent experimental evidence indicates that: 1) the cochlear stiffness gradient is a 100 times smaller than that assumed by most cochlear models (Naidu and Mountain, Hear.Res., 124, 1998); and 2) the organ of Corti (OC) is longitudinally coupled (Naidu and Mountain, ARO 1998, 21:718). This study investigates the effect of incorporating this lower stiffness gradient and coupling within the OC on the results generated by a passive biophysically based long wave model.
The model is a modified 1-D transmission line model of the gerbil cochlea with the displacement of each section dependent on that of its neighbours due to longitudinal coupling. Distributed representations of the stapes and round window are included. Almost all of the parameters are based directly on available physiological and anatomical data.
Model results were compared to experimental data obtained at high SPL. In the frequency range over which the long wave assumption is valid, the model correctly predicts both the magnitude and phase of basilar membrane velocity (Xue et al, JASA, 97, 1995) and intra-cochlear pressure (Olson, JASA, 103, 1998). Further, the good agreement between the predicted and experimentally estimated travel times (Schmeidt and Zwislocki, JASA, 61, 1977) to several locations along the cochlea validates the range of the experimentally determined stiffness gradient incorporated in this model. Longitudinal coupling did not affect the passive velocity response of the OC although in a pseudo-active situation (achieved by reducing the damping of each resonant section), the peak of the traveling wave was broadened. In the pseudo-active model, there was also an increased accumulation of phase around the characteristic frequency. The former result is consistent with the conclusions of Wickesberg and Geisler (In: Peripheral Auditory Mechanisms, Springer-Verlag, 1985).
Supported by NIDCD


Abstract #332, Date 2/15/99, Session K4, Poster (B79)
Experimental evidence does not support feed-forward outer hair cell forces
*K.D. Karavitaki (Massachusetts Institute of Technology, Cambridge, MA and Boston University Hearing Research Center, Boston, MA ); D.C. Mountain (Boston University Hearing Research Center, Boston, MA)

Recent models (Steele et al., in Biophysics of Hair Cell Sensory Systems, world Scientific, 1993, pp. 207-215, Geisler and Sang, Hearing Res., 1995, pp. 132-146) refer to the outer hair cells (OHC) as being part of a 'feed-forward' system where forces produced at the basolateral end of the cells cause hairbundle deflections at a more apical region. This hypothesis is based on anatomical observations (Voldrich, in Mechanics of Hearing, Delft University Press, 1983, pp. 163-167) showing that the OHC are situated in an oblique angle with respect to the longitudinal axis running from the base to the apex of the cochlea. We investigated this hypothesis by estimating the radial and longitudinal motion of OHC in an excised gerbil cochlea. If the 'feed-forward' hypothesis is correct we expect that the longitudinal motion would be comparable to the radial motion.
Data for this study were collected from the apical turn of excised gerbil cochleas using a video stroboscopy system. AC current was delivered between the scala vestibuli of the apical turn and scala tympani of the second turn at frequencies of 20 to 4 kHz. Electrically-evoked motion was estimated using a 2-dimensional cross correlation technique. Plots of the frequency response of both the radial and the longitudinal components of motion were generated.
The experimental results indicate that the longitudinal component of OHC motion in the apex is about ten times smaller than the radial component. These results are inconsistent with the prediction from the 'feed-forward' hypothesis.
Supported by NIDCD


Abstract #333, Date 2/15/99, Session K4, Poster (B80)
Forward and reverse traveling waves in the cochlea
*H.H. Nakajima, R.C. Naidu, A.E. Hubbard, D.C. Mountain (Boston Univ. Hearing Research Center, Boston, MA)

Classical cochlear models predict that forward and reverse traveling waves have the same velocity at a given location. Therefore the delay for a wave propagating from the tympanic membrane to a fixed point along the cochlear partition should be the same as the delay for a wave originating from that point and propagating back to the tympanic membrane.
To test this prediction, we estimated the forward traveling wave delay from the cochlear microphonic (CM) phase and the reverse traveling wave delay from the electrically-evoked otoacoustic emission (EEOE) phase. The group delays of this system were obtained from the slopes of the phase curves. A glass microelectrode with a 2 to 3 micrometer tip was introduced into the scala media at different locations along the cochlea. This electrode was used to record the CM and to stimulate with current to produce EEOEs in the same cochlea. The characteristic frequency (CF) of the electrode location was estimated by summating potential and CM recordings. The location of the electrode was determined from digitized images of the cochlea. Our estimate of the CF of the electrode location was consistent with the gerbil frequency-place map obtained by Muller (1996, Hear. Res. 94, 148-156).
We calculated group delays for frequencies below and above the CF of the electrode location. Above CF, the EEOE group delay decreases while the CM group delay remains approximately the same. Because the CM and EEOE are distributed responses, they may not accurately reflect basilar membrane motion for frequencies near or above CF. For frequencies below CF, we found that the EEOE and CM group delays are approximately the same. Therefore we conclude that the forward and reverse traveling waves have similar delays for frequencies below CF.
Supported by NSF and NIDCD

February 16, 1999

Abstract #502, Date 2/16/99, Session R1, Poster (J2)
Intralabyrinthine infusion of BrdU potentiates gentamicin damage in the chick auditory epithelium
*D.W. Roberson (Boston Children's Hospital and Harvard Department of Otology and Laryngology); E.P. Messana, J.A. Alosi (Boston Children's Hospital); D.A. Cotanche (Boston Children's Hospital and Harvard Department of Otology and Laryngology)

A number of laboratories have used systems to continuously infuse the inner ear with various substances (Roberson et al., Auditory Neuroscience, 2:195; Rubel et al., Science, 267:701). Previous work has shown that intralabyrinthine infusion of trititated thymidine does not itself cause hair cell death in the chick. This study was designed to determine whether infusion of BrdU alters the effect of gentamicin toxicity on the chick cochlea.
Nine chicks had a pump implanted which continuously infused the left inner ear with 0.5 microliters/hour of BrdU (20 mg/ml, in saline). Ten chicks did not have pumps placed. Two days after pump placement, half the chicks in each group were given a single dose of 200 mg/kg of gentamicin; the other half were given 300 mg/kg of gentamicin. Four days after gentamicin injection, all chicks were sacrificed and all ears harvested. Cochleae were stained with phalloidin and examined with immunofluorescent microscopy. Extent of total and partial hair cell loss was expressed as a percent of the length of the cochlea.
In both pump ears and non-pump ears 300 mg/kg of gentamicin caused greater damage than 200 mg/kg. Within each dosage group, however, damage in the right ear was similar regardless of whether the left ear had pump placement or not. Therefore data from all ears which did not have pump placement (right ears of pump animals, and both ears of non-pump animals) were pooled.
In chicks injected with 200 mg/kg of gentamicin, damage was greater in the pump ears (N = 5; total HC loss = 43 7%, partial HC loss = 55 7%) than in the non-pump ears (N = 13; total HC loss = 22 5%; partial HC loss = 31 6%).
In chicks injected with 300 mg/kg of gentamicin, damage was also greater in the pump ears (N = 4; total HC loss = 63 26%, partial HC loss = 76 17%) than in the non-pump ears (N = 13; total HC loss = 32 4%; partial HC loss = 43 5%).
Continuous intralabyrinthine infusion of BrdU in saline potentiates hair cell loss due to systemic gentamicin injection. Whether this potentiation is due to BrdU toxicity or to perilymph disruption from the saline infusion is unclear.
NIDCD DC-01689, NIDCD DC-00152, The National Organization for Hearing Research & The Children's Hospital Otolaryngology Foundation Research Fund


Abstract #505, Date 2/16/99, Session R1, Poster (J5)
Hair cell loss and supporting cell rearrangement in the chick cochlea after gentamicin treatment
*S. Goklani (WT Clarke High School, Westbury, LI, NY); E.P. Messana (Children's Hospital); D.A. Cotanche (Children's Hospital and Harvard Medical School)

Many studies have used aminoglycoside treatment to induce hair cell regeneration in the avian cochlea. While details of these studies have differed, the basic result has generally been the same, i.e. loss of both tall and short hair cells from the proximal (high frequency) end of the cochlea and a gradual replacement of these missing cells by regeneration. SEM and LM studies of damaged cochleae have led to the hypothesis that the dying hair cells were ejected up into the scala media and overlying tectorial membrane. Thus, the surviving cells in the damaged region were thought to be almost exclusively supporting cells. However, it has recently been proposed that the hair cells eject only their sensory bundles in response to aminoglycosides and that the remaining portion of the hair cell rebuilds a new bundle in order to regain its functional capacity. We have set out to test this idea in the two-week-old chick cochlea damaged by a single injection of gentamicin at 300 mg/kg. Whole-mount preparations of the cochlea were made at 3, 4, and 5 days after the gentamicin injection and were labeled with phalloidin to detect the actin cytoskeleton, pan-cadherin antibodies to outline the cells apical junctional complexes, and myosin VIIa antibodies (kindly provided by Dr. Tama Hasson, Yale University) to specifically identify hair cells. On day 3 all of the hair cells were ejected from ts on the behavior of models with linear somas. This is motivated by a key finding in Agmon-Snir et al., that soma excitation is facilitated when synaptic inputs occur coincidentally on distinct four-compartment dendrites compared to when they occur directly on a single dendrite. This behavior is also observed when inputs are on simplified, single-compartment dendrites. Our analyses are used to evaluate the effects and importance of the number and strength of synaptic inputs, the number of compartments, and dendritic length. Dendritic lengths are intriguing in NL neurons because of the observations (Smith and Rubel, 1979) that the lengths of dendrites are adjusted during development to progress from long to short as the characteristic frequencies of cells progress from low to high frequency. The number and strength of inputs are especially interesting in the analysis because varying these parameters leads to an enhancement or reduction of the advantage of binaural over monaural coincidence detection. Results from models with linear and nonlinear cell bodies will be compared with in vitro brain-slice experiments from the chick (Reyes et al., 1996). Specifically, the empirical results obtained by applying current-clamped and conductance-clamped inputs at the soma will be related to modeling results where inputs are placed at the soma and/or dendrites.
Supported by NIDCD and ONR.


Abstract #571, Date 2/16/99, Session R3, Poster (B71)
Simple models of nucleus laminaris neuron
*V.K. Dasika (Hearing Research Center and Biomedical Engineering, Boston University); J.A. White (Biomedical Engineering, Boston University); H.S. Colburn (Hearing Research Center and Biomedical Engineering, Boston University)

Physiologically abstracted and simplified models of nucleus laminaris (NL) neurons are described and tested. These models are closely related to models of MSO neurons (e.g., Han and Colburn, Hearing Research 1993: Brughera et al., Auditory Neuroscience, 1996) and to general models of coincidence detection (Agmon-Snir et al., Nature 1998). The underlying goal of this study is to use simple models to isolate the sources of various effects that are observed physiologically and in complex models. We study the effects of dendrites on the behavior of models with linear somas. This is motivated by a key finding in Agmon-Snir et al., that soma excitation is facilitated when synaptic inputs occur coincidentally on distinct four-compartment dendrites compared to when they occur directly on a single dendrite. This behavior is also observed when inputs are on simplified, single-compartment dendrites. Our analyses are used to evaluate the effects and importance of the number and strength of synaptic inputs, the number of compartments, and dendritic length. Dendritic lengths are intriguing in NL neurons because of the observations (Smith and Rubel, 1979) that the lengths of dendrites are adjusted during development to progress from long to short as the characteristic frequencies of cells progress from low to high frequency. The number and strength of inputs are especially interesting in the analysis because varying these parameters leads to an enhancement or reduction of the advantage of binaural over monaural coincidence detection. Results from models with linear and nonlinear cell bodies will be compared with in vitro brain-slice experiments from the chick (Reyes et al., 1996). Specifically, the empirical results obtained by applying current-clamped and conductance-clamped inputs at the soma will be related to modeling results where inputs are placed at the soma and/or dendrites.
Supported by NIDCD and ONR.

February 17, 1999

Abstract #847, Date 2/17/99, Session Y5, Poster (B112)
Performance limits for frequency and intensity discrimination based on a computational auditory-nerve model
*M.G. Heinz (Speech and Hearing Sciences Program, MIT, Cambridge, MA); L.H. Carney, H.S. Colburn (Hearing Research Center and Department of Biomedical Engineering,Boston University, Boston, MA)

Limits on psychophysical performance for intensity and frequency discrimination have been previously evaluated using optimal decision theory (ODT) (Siebert, 1968, in: Recognizing Patterns, and 1970, IEEE); however, the use of an analytical AN model limited the parameter space over which predictions were valid. Heinz et al. (ARO 1998) combined ODT with a computational AN model to evaluate frequency discrimination over a wide parameter space. This analysis has been extended to include intensity discrimination and random-amplitude frequency discrimination.
Our AN model included linear gamma-tone filters, saturating rate-level curves, low-threshold high-spontaneous-rate fibers, rolloff in phase-locking, neural adaptation, and realistic onsets and offsets. Independent Poisson processes were used to model 30,000 AN fibers spaced according to a human cochlear map. Performance limits based on AN discharge times (all-information) were compared to limits based on discharge counts (rate-place).
Rate-place predictions for frequency discrimination were not consistent with human performance ({uc delta}f) versus frequency at high frequencies. However, all-information {uc delta}f was smaller than and matched the shape of human {uc delta}f versus frequency up to at least 8 kHz. All-information {uc delta}f improved much too rapidly versus duration, and continued to improve with level above 30 dB SPL, unlike humans. Intensity discrimination ({uc delta}L) based on rate-place was within a factor of two of all-information {uc delta}L. Both schemes were roughly consistent with human {uc delta}L, i.e., "near-miss" to Weber's Law, shallow slope versus duration, and flat with frequency. Random-amplitude variation did not affect optimal frequency discrimination based on either all-information or rate-place.
There is more than sufficient rate-place information to explain human performance in all tasks studied. However, our results suggest temporal information is used for frequency discrimination at all frequencies, but with restrictions in the temporal extent of comparisons between AN discharges. Such a decision process is physiologically realistic and could likely account for both frequency and intensity discrimination.
Supported by NIH Grant T32DC00038 and ONR Grant Agmt #Z883402


Abstract #848, Date 2/17/99, Session Y5, Poster (B113)
Spatiotemporal coding of sound level: Quantifying the information provided by level-dependent phase cues
*L.H. Carney (Hearing Research Center & Dept. of Biomedical Engineering, Boston Univ., Boston, MA); M.G. Heinz (Speech and Hearing Sciences Program, MIT, Cambridge, MA); H.S. Colburn (Hearing Research Center & Dept. of Biomedical Engineering, Boston Univ., Boston, MA)

This study represents an extension of previous work related to understanding the significance of auditory-nerve (AN) response properties in level coding (Siebert, 1968, in: Recognizing Patterns, MIT Press; Colburn, 1981, RLE Report, MIT). This study focused on changes in the phase of phase-locked action potentials as a function of level and how these changes in phase may contribute to the encoding of level. Most previous studies of level coding based on the properties of AN discharges have focussed on discharge rates and synchronization of discharges to the stimulus waveform. These studies typically have not included the nonlinear tuning of the auditory periphery and the concomitant effects on the timing of phaselocked AN responses (Anderson et al., 1971, JASA 49:1131-1139) associated with the compressive nonlinearity. These nonlinear changes in phase vary with characteristic frequency (CF) and thus introduce level-dependent cues in the relative timing across AN fibers. These cues can be detected by comparing discharge times across fibers with different CFs.
A simple analytical model for AN fibers included rate- and synchrony-level functions, a simple description of the changes in phase with level, and Poisson statistics. The available information from nonlinear phase changes was quantified to allow their potential contribution to level discrimination to be directly compared to the contributions of rate and synchrony measures. Changes in the phase of AN responses provide cues for level discrimination that are comparable or greater than those provided by discharge rate and synchrony, and the phase cues persist over a wider dynamic range. Heinz et al. (ARO 1998, 1999) have shown that temporal cues can contribute significantly over a wide frequency range. Evaluating the potential significance of the nonlinear response properties of AN fibers associated with the compressive nonlinearity is important for understanding neural mechanisms involved in level coding. The compressive nonlinearity, which is associated with the cochlear amplifier, is a fragile property of the ear that is believed to be affected in common forms of hearing impairment.
Supported by: NSF Grant IBN9601215 (LHC), and NIH Grant T32DC00038 (MGH)

HRC Home Page|What's New|Faculty|Laboratories|Departmental Links
Research Areas|Graduate Students|Courses|Seminars

Comments to : Ann Green agreen@enga.bu.edu