binaural pitch perception in normal... listener

www.elsevier.com/locate/heares

Hearing Research 223 (2007) 29–47

HearingResearch

Research paper

Binaural pitch perception in normal-hearing andhearing-impaired listeners

Sebastien Santurette, Torsten Dau *

Centre for Applied Hearing Research, Ørsted-DTU, Technical University of Denmark, DTU Bygning 352, Ørsteds Plads, 2800 Kgs. Lyngby, Denmark

Received 31 March 2006; received in revised form 25 September 2006; accepted 26 September 2006Available online 14 November 2006

Abstract

The effects of hearing impairment on the perception of binaural-pitch stimuli were investigated. Several experiments were performedwith normal-hearing and hearing-impaired listeners, including detection and discrimination of binaural pitch, and melody recognitionusing different types of binaural pitches. For the normal-hearing listeners, all types of binaural pitches could be perceived immediatelyand were musical. The hearing-impaired listeners could be divided into three groups based on their results: (a) some perceived all types ofbinaural pitches, but with decreased salience or musicality compared to normal-hearing listeners; (b) some could only perceive the stron-gest pitch types; (c) some were unable to perceive any binaural pitch at all. The performance of the listeners was not correlated withaudibility. Additional experiments investigated the correlation between performance in binaural-pitch perception and performance inmeasures of spectral and temporal resolution. Reduced frequency discrimination appeared to be linked to poorer melody recognitionskills. Reduced frequency selectivity was also found to impede the perception of binaural-pitch stimuli. Overall, binaural-pitch stimulimight be very useful tools within clinical diagnostics for detecting specific deficiencies in the auditory system.� 2006 Elsevier B.V. All rights reserved.

Keywords: Binaural pitch; Hearing impairment; Melody recognition; Spectral and temporal resolution; Auditory modeling

1. Introduction

Pitch is a very important auditory sensation that plays amajor role in auditory object formation and speech percep-tion. It is also responsible for the musical perception ofsounds. Understanding and modeling pitch perceptionhas been historically challenging because pitch is a subjec-tive attribute. Binaural pitch, also termed dichotic pitch,was first studied in detail by Cramer and Huggins (1958).Binaural pitch can be produced by presenting a randomnoise stimulus to one ear, while a slightly different randomnoise stimulus is simultaneously presented to the other ear.The difference between the noise stimuli is most commonlya phase shift in some specific frequency components. The

0378-5955/$ - see front matter � 2006 Elsevier B.V. All rights reserved.

doi:10.1016/j.heares.2006.09.013

* Corresponding author. Tel.: +45 4525 3936.E-mail addresses: [email protected] (S. Santurette), tda@oersted.

dtu.dk (T. Dau).URL: http://www.oersted.dtu.dk/cahr (T. Dau).

introduction of an interaural phase-difference spectrum,from the combination of the left and right noise stimuli,creates an additional auditory object. This new auditoryobject manifests as a pitch sensation at a specific frequency,i.e. the listener is able to hear a more or less distinct tone inthe noise. If the noise stimuli are presented separately toeach ear, only noise can be perceived. A simple visual anal-ogy to binaural pitch is the phenomenon of stereopsis(Julesz, 1971). Stereopsis occurs when two slightly differentrandom sets of dots are presented to each eye, and a three-dimensional image is perceived.

Several studies have investigated different aspects of bin-aural-pitch perception in normal-hearing listeners. Thesehave used and compared different pitch types, such as theHuggins’ pitch (Cramer and Huggins, 1958), the binauraledge pitch (Klein and Hartmann, 1981), and the binauralcoherence edge pitch (Hartmann, 1984b), further describedin the following. Akeroyd et al. (2001) showed that, innormal-hearing listeners, the perception of these three pitch

mailto:[email protected]

mailto:tda@oersted.

http://www.oersted.dtu.dk/cahr

1 All experiments were approved by the ethics committee of Copenhagencounty (Den Videnskabsetiske Komite for Københavns Amt).

30 S. Santurette, T. Dau / Hearing Research 223 (2007) 29–47

types was immediate. They also compared the salience ofthese pitch types using a melody recognition task. Severalattempts have been made to account quantitatively for bin-aural-pitch perception. A number of models have been sug-gested for the processing of binaural pitch. The centralactivity pattern (CAP) model (Raatgever and Bilsen,1986), and models based on the equalization–cancellation(EC) model by Durlach (1960), all use temporal cues toprocess binaural stimuli. The CAP model uses a tuningof the signals, both in frequency and interaural time differ-ence in the peripheral channels, to derive a two-dimen-sional central activity pattern. The EC models, on theother hand, assume that the binaural system first equalizesthe signals in corresponding channels. This is achieved byusing amplitude adjustment and phase shifting, before sub-tracting the signals to obtain a residual activity spectrum(e.g. Durlach, 1960; Culling et al., 1998). Comparingresults from psychophysical experiments to model predic-tions, Hartmann and Zhang (2003) found that the CAPmodel did not predict the salience of different versions ofthe Huggins’ pitch and the binaural coherence edge pitchcorrectly. However, the results were seen to favor an ECprocess. Culling et al. (1998) also demonstrated that theCAP model did not correctly predict the range of transitionbandwidths over which the Huggins’ pitch and binauraledge pitch are audible, contrary to their modified versionof the EC model (mEC). However, the CAP model accu-rately predicted the lateralization of binaural pitch.

The aim of the present study was to investigate theeffects of hearing impairment on binaural-pitch perception.As binaural-pitch signals are easy to generate and their per-ception seems to be immediate (Akeroyd et al., 2001), itwould be of interest to find a link between the ability ofhearing-impaired listeners to perceive binaural-pitch sig-nals and the deficiency of a particular auditory function.If this can be achieved, binaural-pitch signals could becomeuseful test-stimuli for specific clinical diagnostics. In orderto identify and understand the mechanisms underlying bin-aural-pitch perception, and how these mechanisms areaffected by hearing impairment, some fundamental ques-tions first need to be addressed: Are hearing-impaired lis-teners able to perceive binaural-pitch signals at all? If yes,is that perception immediate as in normal-hearing listeners?Are different binaural-pitch types perceived and affected inthe same way in a hearing-impaired listener? Finally, in thecase that hearing-impairment affects binaural-pitch percep-tion, are hearing-threshold elevation, reduced frequencyselectivity, and reduced temporal resolution among thekey factors in the impaired listener’s difficulties with binau-ral pitch? These fundamental questions were the main con-cerns of the present study.

The influence of hearing impairment on binaural-pitchperception was investigated in a set of psychophysicalexperiments, including a melody recognition test similarto that performed by Akeroyd et al. (2001). The detectionand discrimination of Huggins’ pitch, as a function of fre-quency, was also tested in experiments similar to those of

Hartmann and Zhang (2003). Hearing-impaired listenersparticipated in additional experiments evaluating severalauditory abilities with which patients with a hearing lossare commonly found to have difficulties: frequency dis-crimination, frequency selectivity, and temporal resolu-tion. The purpose was to investigate if a link could bemade between the subjects’ performance in each of thesetasks, and their performance in the binaural-pitch experi-ments. Frequency discrimination abilities are expected tobe directly correlated to the performance of the subjectsin pitch discrimination and melody recognition tasks. Thisis because these experiments require the listener to have amusical perception of sounds, and thus to be able to dis-criminate sounds producing close pitch frequencies. Animpaired frequency selectivity, associated with a broaden-ing of the auditory filters at the level of the basilar mem-brane, can also give rise to changes in the internalrepresentation of the interaural-correlation spectrum ofbinaural signals. A listener with a sensorineural hearingloss, involving a deficit in frequency selectivity, mighttherefore be expected to have more difficulties perceivingbinaural pitch than a normal-hearing listener. As binau-ral-pitch perception involves an analysis and comparisonof the temporal fine structure of the left and right signalsby the binaural system, one may furthermore expect thatdeficits in temporal resolution might affect binaural-pitchperception. However, because binaural pitch is a low-fre-quency phenomenon, and becomes very difficult to per-ceive for pitch frequencies above about 1600 Hz(Cramer and Huggins, 1958), one can expect an elevationof the listener’s hearing threshold at high frequencies notto be responsible for potential difficulties in perceivingbinaural pitch.

In the following, results from three experimentsinvolving binaural pitch will be presented first: theyinclude Huggins’ pitch detection, Huggins’ pitch discrim-ination, and melody recognition with different types ofbinaural-pitch stimuli. The relation of these results withmeasures of spectral and temporal resolution will thenbe investigated. Implications of the results will finallybe discussed.

2. Method

2.1. Subjects

Fourteen normal-hearing (NH) and 10 hearing-impaired(HI) subjects participated in the experiments.1 A pure-toneaudiogram was obtained for each of the subjects prior tothe experiments. All NH subjects’ hearing thresholds werefound to be below 20 dB HL at all test frequencies, whilethe HI subjects obtained the various audiogram shapesshown in Fig. 1. NH subjects were aged between 22 years

S. Santurette, T. Dau / Hearing Research 223 (2007) 29–47 31

and 40 years. HI subjects reported different signs of hearingimpairment, and their age was between 25 years and68 years. The origins of the HI subjects’ hearing losses werediverse, and the different categories of impairment encoun-tered are briefly presented in the following:

� Subjects 1 and 2 showed audiograms that were similar tothose of NH listeners. However, both subjects reportedthat they had problems understanding speech in noisyenvironments or situations with several concurrentspeakers.

.125 .25 .5 1 2 3 4 6 8−10

0102030405060708090

100

Right ear

.12

Frequency

−100

102030405060708090

100−10

0102030405060708090

100

Hea

rin

g t

hre

sho

ld (

dB

HL

)

−100

102030405060708090

100

.125 .25 .5 1 2 3 4 6 8

−100

102030405060708090

100.12

Frequency

Subject 1 [1.5|+3] S

Subject 3 [4.0|–8] S

Subject 5 [17.5|+15] S

Subject 7 [20.5|–3] S

Subject 9 [36.0|+4] S

Fig. 1. Audiograms of the 10 hearing-impaired subjects. In the bottom-left cosubject in dB HL between 125 Hz and 2000 Hzjan asymmetry factor corresponin dB HL between 125 Hz and 2000 Hz].

� Subjects 3, 4, and 5 reported that they had been exposedto loud sounds, thus one may assume cochlearouter-hair-cell damage. Those subjects all showed ahigh-frequency hearing loss, especially around the reso-nance frequency of the ear canal (3–4 kHz), but hadnormal hearing at low frequencies. Subject 5 had themost severe hearing loss while subject 3 showed a mildimpairment. All three subjects could understand speechvery well in quiet environments, but indicated that theyhad difficulties with speech intelligibility in noisyenvironments.

Left ear

5 .25 .5 1 2 3 4 6 8100102030405060708090100

(kHz)

−100102030405060708090100− 100102030405060708090100

Hea

rin

g t

hre

sho

ld (

dB

HL

)

−100102030405060708090100

5 .25 .5 1 2 3 4 6 8

−100102030405060708090100

(kHz)

ubject 2 [3.0|–2]

ubject 4 [16.0|–8]

ubject 6 [19.5|–1]

ubject 8 [29.5|+11]

ubject 10 [44.5|–15]

rner of each audiogram are also indicated: [the average hearing loss of theding to the average difference between the right and left hearing thresholds


� Subject 6 had normal hearing at low frequencies, andstarted suffering from hearing problems progressivelyafter the age of 60, which suggests the sensorineuralhearing loss was age-induced.� Subject 7 suffered from Meniere’s disease. Symptoms

involved in Meniere’s disease typically include vertigoand a low-frequency hearing loss, which can vary withtime, combined with tinnitus and sensations of fullnessin the ear. Subject 7’s audiogram showed only a mildhearing loss at low frequencies, and a progressively moresevere hearing loss at frequencies above 2 kHz.� Subjects 8, 9, and 10 had sensorineural hearing losses,

with audiograms showing a slope falling down towardshigh frequencies. Subject 8 had been operated followingan otosclerosis infection. Subject 9 had been sufferingfrom diabetes for her whole life. She suggested her hear-ing problem was probably linked to the disease, whichhad also progressively affected other perceptual abilitieslike her eyesight. The hearing loss of subject 10 was aconsequence of cranial trauma following a car accident,which had affected other perceptual abilities like olfac-tion. This suggests that the hearing loss may be due todamage of central areas of the auditory system.

2.2. Stimuli

Six binaural-pitch types with different frequency-depen-dent interaural phase-difference patterns were used in theexperiments, indicated in Fig. 2.

In the Huggins’ pitch (HP) phase-difference pattern(Cramer and Huggins, 1958), the signals in both channelsare in phase at all frequencies except a narrow frequencyrange around the boundary frequency, fb (Fig. 2a). In thetransition area around fb, a linear phase shift is introducedsuch that, if the transition bandwidth is b, the phase differ-ence between both channels varies linearly from 0 to 2p inthe frequency interval [fb � b/2, fb + b/2]. The phase differ-ence is maximally p at the boundary frequency. Detectionof HP is optimal for transition bandwidths between 4%and 32% of the boundary frequency (Culling et al., 1998).

In the binaural edge pitch (BEP) phase-difference pattern(Klein and Hartmann, 1981), the signals in both channelsare in phase below the boundary frequency (or edge fre-quency) fb, and in antiphase above fb (Fig. 2b). In theBEP configuration, no transition is needed for optimaldetection of the pitch, but a linear phase shift from 0 top may as well be used without altering the pitch detection,provided the transition bandwidth is below 8% of theboundary frequency (Culling et al., 1998).

In the binaural coherence edge pitch (BICEP) phase-dif-ference pattern (Hartmann, 1984b), the signals in bothchannels are in phase below the boundary frequency (oredge frequency) fb, and a random phase difference is intro-duced at all frequencies above fb (Fig. 2c). In other words,the left and right noises are coherent below fb and incoher-ent above fb, thus creating a coherence edge. Reversing the

coherent and incoherent parts of the noise, i.e. introducingthe random phase difference at all frequencies below fb andkeeping the noises in phase above fb, produces a similarpitch sensation (Hartmann, 1984b). This configuration willbe referred to as reversed binaural coherence edge pitch orRBICEP (Fig. 2d).

The four configurations presented above (HP, BEP,BICEP, and RBICEP) are the most common patterns forgeneration of a binaural-pitch signal using a frequency-dependent phase difference between both channels. Twoadditional configurations were also investigated in the pres-ent study. One was used by Yost (1991) and Zhang andHartmann (2004), and resembles the HP configurationexcept that the phase in a narrow bandwidth around fb isshifted by a fixed amount in one channel. Thus, the phasedifference between both channels is constant as a functionof frequency, and there are two sharp edges similar to theedge of the BEP configuration. A phase shift of p will beused in this study (Fig. 2e), and the configuration will becalled binaural band pitch (BBP). The second additionalconfiguration is similar to the BBP configuration exceptthat the edges are coherence edges, which means that thephase in a narrow bandwidth around fb is randomized inone channel (Fig. 2f). This configuration, similar to thestimulus used in Experiment 1 of Akeroyd and Summer-field (2000), will be called binaural incoherent-band pitch

(BIBP).The binaural-pitch signals were generated as follows:

(1) Random noise with the desired duration was gener-ated in the spectral domain, using a sampling rate of44,100 Hz. (2) The amplitude was adjusted such that thenoise had a constant amplitude and only the phase wasrandomized, i.e. white noise was created. (3) All frequencycomponents above 4000 Hz were set to zero. (4) The stim-ulus obtained after step 3 was left unchanged and wastransformed back to the time domain using the inverseFourier transform algorithm. It was then fed to the leftchannel. (5) Some of the phase components of the stimu-lus obtained after step 3 were modified in order to createthe desired phase-difference pattern between the two ears.The stimulus was transformed back to the time domainusing the inverse Fourier transform algorithm. It was thenfed to the right channel.

2.3. Apparatus

During all tests, subjects were sitting in a soundproof lis-tening booth equipped with a PC. All stimuli were gener-ated in MATLAB and converted to analog signals usinga 16-bit D/A converter with a sampling rate of44,100 Hz. They were fed through the PC soundcard(RME DIGI96/8) which was connected to SennheiserHD580 headphones. The level of the signals was calibratedby connecting the headphones to an artificial ear (IEC 318)which was itself linked to a B&K2607 sound level meter.The presentation level of all stimuli was 65 dB SPL forthe NH subjects and was adjusted for the HI subjects

fb0

180

360

90

270

Frequency

Inte

raur

al p

hase

diff

eren

ce (

deg)

(a) Huggins’ pitch (HP)

fb0

180

360

90

270

Frequency

Inte

raur

al p

hase

diff

eren

ce (

deg)

(b) Binaural edge pitch (BEP)

fb0

180

360

90

270

Frequency

Inte

raur

al p

hase

diff

eren

ce (

deg)

(c) Binaural coherence edge pitch(BICEP)

fb0

180

360

90

270

FrequencyIn

tera

ural

pha

se d

iffer

ence

(de

g)

(d) Reversed BICEP (RBICEP)

fb0

180

360

90

270

Frequency

Inte

raur

al p

hase

diff

eren

ce (

deg)

(e) Binaural band pitch (BBP)

fb0

180

360

90

270

Frequency

Inte

raur

al p

hase

diff

eren

ce (

deg)

(f) Binaural incoherent-band pitch(BIBP)

Fig. 2. Six phase-difference patterns producing a binaural-pitch sensation: (a) Huggins’ pitch: linear phase shift from 0 to 2p around the boundaryfrequency fb, (b) binaural edge pitch: constant phase shift of p above fb, (c) binaural coherence edge pitch: random phase shift above fb, (d) reversedbinaural coherence edge pitch: random phase shift below fb, (e) binaural band pitch: constant phase shift of p around fb, (f) binaural incoherent-bandpitch: random phase shift around fb.


depending on their hearing thresholds. As reported byCramer and Huggins (1958), variations in level do not seemto affect binaural-pitch perception much at comfortablepresentation levels, as long as the signals are not too weak.In the present study, the levels were chosen such as to pro-duce comfortable loudness in both ears.

3. Binaural pitch experiments

Before the different binaural-pitch experiments werecarried out, the subjects were screened to ensure theycould hear binaural pitch, in a preliminary test. Each lis-tener was presented a 10-second signal, consisting of 10noise intervals of 1 s duration each. A 30-ms cosine-rampwas applied to the onset and offset of each interval. The

first and the last interval contained no binaural pitch, i.e.diotic noise, while intervals 2–9 each contained a Hug-gins’ pitch with values of fb ranging from 523.25 Hz(C5) to 1046.50 Hz (C6), so that the musical C-scalewas played. The transition bandwidth was equal to 16%of fb. The task of the listener was to listen to the noisestimulus and tell the experimenter if something else thannoise could be perceived. After listening to the stimulusonly once, all 14 NH subjects and HI subjects 1–8reported they could hear a tone rising progressively. Thisconfirms that binaural pitch, if perceivable, is an immedi-ate sensation and requires no learning in order to be per-ceived (Akeroyd et al., 2001). When asked where theperceived pitch was located, 13 NH and 7 HI subjectsclearly mentioned that the pitch was heard inside the


head and lateralized towards the right ear, while the tworemaining subjects heard the pitch more towards theback of the head. HI subjects 9 and 10 reported theycould not hear anything but noise, even by listening tothe stimuli several times. For these two subjects, the stim-uli were also played at higher levels and with other pitchtypes, but in none of the configurations could the sub-jects perceive a pitch. This suggests that the ability of ahearing-impaired person to detect HP is either immediateor non-existing under these test conditions. It is of inter-est to mention that the two subjects who could not per-ceive binaural pitch at all (subjects 9 and 10) were thosefor whom it was most likely that central areas of theauditory system had been damaged. Those two subjectsdid not participate further in the experiments involvingbinaural pitch.

3.1. Huggins’ pitch detection

This first experiment aimed at measuring the detectabil-ity of HP as a function of frequency. The experiment wassimilar to Experiment 1 of Hartmann and Zhang (2003),but the upper frequency limit was extended.

80 125 200 315 500 800 1250 2000

0102030405060708090

100

Normal hearing average

80 125 200 315 5

Frequenc

2 2.5 3 2 2.5 3 2 2.50

102030405060708090

100

Co

rrec

t re

spo

nse

s (%

)

2 2.5 3 2 2

80 125 200 315 500 800 1250 20000

102030405060708090

100

80 125 200 315 5

Frequenc

Subject 1 Subject 2

Subject 4 Subject 5

Subject 7 Subject 8

Fig. 3. Detectability of HP for eight different boundary frequencies: results froover 14 normal-hearing subjects are indicated in each plot by empty circles. ThHartmann and Zhang (2003) with normal-hearing listeners.

3.1.1. Method

The test consisted of 80 trials testing eight differentboundary frequencies, with 10 trials for each value of fb.The tested boundary frequencies were 80 Hz, 125 Hz,200 Hz, 315 Hz, 500 Hz, 800 Hz, 1250 Hz, and 2000 Hz.The order of the trials was randomized, and a two-alterna-tive forced-choice (AFC) procedure was used. In each trial,the listener was presented two noise intervals of 500 mseach, separated by a silence of 500 ms. A 30-ms cosine-ramp was applied to the onset and offset of each interval.One of the intervals contained a Huggins’ pitch (withb = 0.16fb) whereas the other interval contained no binau-ral pitch, i.e. the noise was the same in both ears. The inter-val containing the pitch was random. The task of thesubject was to indicate through a computer interface whichinterval contained the pitch. No feedback was provided.

3.1.2. Results and discussion

Detectability was measured as the percentage of correctresponses for each boundary frequency. Chance level cor-responded to 50% of correct responses. The average resultsover all 14 NH subjects and the individual results of the 8HI subjects are given in Fig. 3.

Individual data

00 800 1250 2000

y (Hz)80 125 200 315 500 800 1250 2000

0102030405060708090100

.5 3 2 2.5 2 2.5 3 2 2.5 3 2 2.50102030405060708090100

Co

rrec

t re

spo

nse

s (%

)

00 800 1250 2000

y (Hz)80 125 200 315 500 800 1250 2000

0102030405060708090100

Subject 3

Subject 6

Hartmann & Zhang (2003) NH data

m eight hearing-impaired subjects. The average results and standard errore ninth subplot in the lower-right corner indicates the results obtained by


For NH listeners, the average detectability of HPappeared to be best for frequencies between 200 Hz and1250 Hz, where it lies above 94%. The most salient fre-quency was 500 Hz, and this was also the frequency wherethe smallest standard deviation was observed. At low fre-quencies, the results are consistent with those obtained byHartmann and Zhang (2003), showing a declining detect-ability with decreasing frequency, reaching chance level at80 Hz. The large standard deviation at this particular fre-quency is due to the fact that, although most subjectsscored around chance level or below, some of them couldperceive the pitch even at such a low frequency. The pre-sentation level being the same at all frequencies, the lowerdetectability of HP at low boundary frequencies may bedue to the higher hearing threshold at these frequencies,as suggested by Cramer and Huggins (1958). Culling(1999) suggested that the use of a transition bandwidth b

proportional to the bandwidth of the auditory filter cen-tered on fb could be better than using a fixed percentageof fb. In this experiment, the fixed value of b = 16% of fb

appears to be close to the equivalent rectangular band-width (ERB), as defined by Glasberg and Moore (1990),except at frequencies below 200 Hz where the ERB pro-gressively becomes broader than b. Whether this has aninfluence on HP detectability at low frequencies, and isresponsible for the results obtained, was not further inves-tigated in the present study. At the upper boundary fre-quency of 2000 Hz, a large standard deviation was againobtained. Here, nine subjects showed a very good detect-ability (80–100%), while five of them obtained a score inthe range 20–60%, reflecting that their detection of thepitch was non-existing or very weak. This suggests a highvariability of the decline of the pitch at high frequenciesamong listeners. Interestingly, subjects who easily detectedHP at 2000 Hz did not necessarily detect the pitch at 80 Hz,and vice versa, suggesting that performance in binaural-pitch detection at low versus high frequencies might notbe linked.

The HI subjects can be divided into two groupsaccording to their results: (1) Subjects 2, 3, 7, and 8obtained similar results as the NH subjects, with highscores (above 80%) in the frequency range between200 Hz and 1250 Hz, and a declining detectability below200 Hz and above 1250 Hz. The hearing impairment ofthese subjects did not seem to affect HP detectability.(2) The other HI subjects (1, 4, 5, and 6) obtained highscores in a narrower mid-frequency range than NH listen-ers, and a detectability that declined faster outside thismid-frequency range. For subjects 5 and 6, both lowand high frequencies were affected. For subject 4, onlyhigh boundary frequencies were significantly affected, withchance level reached at 1250 Hz and 2000 Hz. For subject1, a poorer detectability than the NH average wasobtained only for low values of fb (200 Hz and below).Thus, for the subjects from the second group, detectabilityof HP appeared to be reduced either for low boundaryfrequencies only, high boundary frequencies only, or both

low and high boundary frequencies. Thus, as in NH lis-teners, there is no link between performance in HP detec-tion at low and high frequencies. Finally, HP detectabilitywas not related to the shape of the audiogram or to theorigin of the subject’s hearing loss, with a correlationcoefficient r = 0.14 between the average score of the sub-jects in the HP detection test and their average hearingthreshold between 125 Hz and 2000 Hz.

3.2. Huggins’ pitch discrimination

The aim of this second experiment was to measure theability of the subjects to discriminate two Huggins’ pitcheswith close boundary frequencies (separated by 1.5 musicalsemitones). This was designed to test the musicality of thepitch for the listeners. The experiment was similar toExperiment 2 of Hartmann and Zhang (2003), but withan extended high-frequency limit. For comparison pur-poses, the test was also performed with pure-tone signals.

3.2.1. Method

The test consisted of two runs: a first run with pure-tonesignals, and a second run with HP stimuli. Each run con-sisted of 80 trials testing eight different frequencies, with10 trials for each frequency value. The tested frequencieswere 80, 125, 200, 315, 500, 800, 1250, and 2000 Hz. Theorder of the trials was randomized, and a 2-AFC procedurewas used. On each trial, two intervals of 500 ms each, sep-arated by a silence of 500 ms, were presented to the listener.A 30-ms cosine-ramp was applied to the onset and offset ofeach interval. In the first run, pure-tone stimuli were used,and on each trial the frequency of the second pure tone waseither 9% (or 1.5 musical semitones) lower than, 9% higherthan, or the same as that of the first one, which was equalto one of the values listed above. In the second run, bothintervals contained a Huggins’ pitch (with b = 0.16fb).The boundary frequency in the first interval was equal toone of the values listed above, while the boundary fre-quency in the second interval had one of the following val-ues, chosen randomly: 9% lower than, 9% higher than, orthe same value as in the first interval. The task of the sub-ject was to indicate through a computer interface whetherthe pitch frequency in the second interval was lower, thesame, or higher than the pitch frequency in the first inter-val. No feedback was provided.


Discrimination was measured as the percentage of cor-rect responses for each frequency. Chance level corre-sponded to 33% of correct responses. The average resultsand standard error over all 14 NH subjects and the individ-ual results from the 8 HI subjects are given in Fig. 4.

For NH subjects, pure-tone discrimination was verygood, with average scores above 91%, for all frequenciesexcept the lowest test-frequency of 80 Hz. At such a fre-quency, most subjects showed a reduced ability to discrim-inate the tones, with scores in the range of 50–80%. One

80 125 200 315 500 800 1250 2000

0102030405060708090

100

Pure tone NH average HP NH average Pure tone individual data HP individual data

80 125 200 315 500 800 1250 2000

Frequency (Hz)80 125 200 315 500 800 1250 2000

0102030405060708090100

2 2.5 3 2 2.5 3 2 2.50

102030405060708090

100

Co

rrec

t re

spo

nse

s (%

)

2 2.5 3 2 2.5 3 2 2.5 2 2.5 3 2 2.5 3 2 2.50102030405060708090100

Co

rrec

t re

spo

nse

s (%

)

80 125 200 315 500 800 1250 20000

102030405060708090

100

80 125 200 315 500 800 1250 2000

Frequency (Hz)80 125 200 315 500 800 1250 2000

0102030405060708090100

Subject 1 Subject 2 Subject 3


Subject 7 Subject 8 Hartmann & Zhang (2003) NH data

Fig. 4. Discrimination of a 1.5-semitone interval with pure-tone and HP stimuli for eight different frequencies: average results and standard error over 14normal-hearing subjects and results from 8 hearing-impaired subjects. The ninth subplot in the lower-right corner indicates the results obtained byHartmann and Zhang (2003) with normal-hearing listeners.


might expect that the results for HP discrimination reflect acombination of those obtained in the HP detection experi-ment (Section 3.1) and the pure-tone discrimination exper-iment. The task required the listener to be able to detect HPand to be able to discriminate between two close pitch fre-quencies. In fact, at low frequencies (80, 125 Hz), thedeclining detectability of HP reduces HP discrimination.Discrimination appeared to be nearly impossible at80 Hz. At 125 Hz, discrimination was very difficult orimpossible for some listeners, only two subjects obtainingmore than 70% correct responses. The results at low fre-quencies are consistent with those obtained by Hartmannand Zhang (2003) only showing a slightly higher value at200 Hz. At high frequencies, HP discrimination becomesmore difficult with increasing frequency, but the frequencyat which the decay starts varies among listeners. Results inthe mid-frequency range clearly show that Huggins’ pitchsignals are as musical as pure tones between 200 Hz and800 Hz, and up to 1250 Hz for about half of the test sub-jects, the average upper-limit probably lying between800 Hz and 1250 Hz.

Some HI subjects obtained results close to the NHaverage in the pure-tone condition. Only subjects 3, 6,and 8 showed a different behavior. Subject 3 had scores

about 25–30% lower than the normal-hearing average atall frequencies. Subjects 6 and 8 showed more variableresults with frequency and obtained the best discrimina-tion at a frequency of 315 Hz. In the HP condition, sub-jects 3, 6, and 8 were also the ones obtaining the lowestoverall scores, never reaching more than 80% of correctresponses at any frequency. Discrimination of HP gener-ally appeared to be best in a narrow mid-frequency range,which was different for each subject but always between200 Hz and 800 Hz. When comparing the results to thosefrom the HP detection experiment (Section 3.1), one cannotice that subjects 3 and 8 had difficulties in HP discrim-ination although they showed a very good detectability ofHP. This suggests that some subjects with difficulties infrequency discrimination may detect binaural pitch verywell, but still find it difficult to hear the musical aspectsof the pitch.

3.3. Melody recognition

The third experiment was similar to that carried out byAkeroyd et al. (2001). It was aimed to measure the melodyrecognition capabilities of the subjects, using pure tones aswell as the six different binaural-pitch types.


3.3.1. Method

The experiment consisted of 70 trials testing seven differ-ent types of stimulus: pure tone, HP, BEP, BICEP, RBI-CEP, BBP, and BIBP (see Fig. 2). For the HP, BBP, andBIBP configurations, the bandwidth of the phase shiftedarea was equal to 16% of the boundary frequency. Forthe BEP configuration, no transition range was used. Tendifferent melodies, each consisting of 16 equal-durationnotes (Table 1a), were played using each type of stimulus.The frequency of the pure tone or the boundary frequencyof the binaural-pitch signal was equal to the frequency ofeach note (Table 1b). The order of the trials was random-ized in such a way that neither the same stimulus type, northe same melody was played in two successive trials. Oneach trial, the melody consisted of 16 notes of 300-ms dura-tion, each separated by a silence of 100 ms. A 30-ms cosine-ramp was applied to the onset and offset of each note. Inorder to have equal-duration notes, the rhythm of the mel-odies was modified so that long-duration notes were splitinto several equal-duration notes. A 10-AFC procedurewas used. After the stimulus was presented once in eachtrial, the task of the subject was to indicate through a com-puter interface which melody was played. No feedback wasprovided. The experiment was repeated twice, i.e. each sub-ject participated in a total of three runs.

Before the experiment, it was ensured that the subjectshad a sufficient knowledge of the melodies. The purposewas to avoid strong bias effects because they knew someof the melodies better than others. Some of the chosen mel-odies (no. 4, 5, 7, 9, and 10 in Table 1a) were different fromthose used by Akeroyd et al. (2001) so that they were morelikely to be known by the listeners. Before they participated

Table 1The 10 melodies used in the melody recognition experiment, with correspondi

Melody Notes

1 2 3 4 5 6

(a) The 16 notes forming each of the 10 melodies

1 Au Clair de la Lune C5 C5 C5 D5 E5 E52 Frere Jacques C5 D5 E5 C5 C5 D53 The Alphabet Song C5 C5 G5 G5 A5 A54 Old MacDonald Had A Farm F5 F5 F5 C5 D5 D55 London Bridge G5 A5 G5 F5 E5 F56 Yankee Doodle F5 F5 G5 A5 F5 A57 Sur le pont d’Avignon F5 F5 F5 F5 G5 G58 Chimes of Big Ben A5 F5 G5 C5 C5 G59 Eurovision Theme C5 F5 F5 F5 G5 A5

10 Jingle Bells E5 E5 E5 E5 E5 E5

Note

(b) Note frequencies

C5D5E5F5G5A5A5#

B5C6

in the experiment, the subjects went through a learningprocedure of the 10 chosen melodies. The learning processcontained two phases: (1) The aim of phase 1 was to famil-iarize the subjects with the melodies themselves. The sub-jects could listen to the pure-tone version of the melodiesas many times as they wished using a computer interface,through which they could enter their own title or commentsin order to help them identify each melody. (2) The aim ofphase 2 was to familiarize the subjects with hearing themelodies played with binaural-pitch signals. Still using acomputer interface, the subjects went through a fixed pro-cedure where they were instructed to listen to the melodiesa limited number of times. Melodies 1–10 were all played infive different conditions with pure-tone, HP, BEP, andBICEP stimuli. At the end of phase 2, the subjects hadthe opportunity to listen to the pure-tone melodies again,as in phase 1.


The melody recognition ability of the subjects was mea-sured as the percentage of correct responses for each stim-ulus type. Chance level corresponded to 10% of correctresponses. The average results and standard error over 13NH subjects as well as individual results from all 8 HI sub-jects are shown in Fig. 5. The results from seven of the NHtest subjects, selected for their very good knowledge of thepure-tone melodies after the learning phase, have also beenaveraged. The results shown in Fig. 5 represent the averageof the three runs performed by the subjects. In order toevaluate the statistical significance of the differences inthe results for the seven stimulus types, a set of Wilcoxonmatched-pairs signed-ranks tests (Fisher and van Belle,

ng notes and their frequencies

7 8 9 10 11 12 13 14 15 16

D5 D5 C5 E5 D5 D5 C5 C5 C5 C5E5 C5 E5 F5 G5 G5 E5 F5 G5 G5G5 G5 F5 F5 E5 E5 D5 D5 C5 C5C5 C5 A5 A5 G5 G5 F5 F5 F5 F5G5 G5 D5 E5 F5 F5 E5 F5 G5 G5G5 C5 F5 F5 G5 A5 F5 F5 E5 E5G5 G5 A5 A5# C6 F5 E5 F5 G5 C5A5 F5 A5 F5 G5 C5 C5 G5 A5 F5A5 F5 F5 C6 C6 C6 A5 A5 A5 A5E5 E5 E5 G5 C5 D5 E5 E5 E5 E5

Frequency (Hz)

523.25587.32659.26698.46783.99880.00932.33987.77

1046.50

Tone HP BEP BICEPRBICEP BBP BIBP

0102030405060708090

100Tone HP BEP BICEPRBICEP BBP BIBP

Pitch typeTone HP BEP BICEPRBICEP BBP BIBP

0102030405060708090100

0102030405060708090

100

Co

rrec

t re

spo

nse

s (%

)

0102030405060708090100

Co

rrec

t re

spo

nse

s (%

)

Tone HP BEP BICEPRBICEP BBP BIBP0

102030405060708090

100

Tone HP BEP BICEPRBICEP BBP BIBP

Pitch typeTone HP BEP BICEPRBICEP BBP BIBP

0102030405060708090100

Normal hearing average (all 13 subjects) Normal hearing average (7 selected subjects) Individual data



Subject 7 Subject 8 Akeroyd et al. (2001) NH data

Fig. 5. Melody recognition of pure-tones and six types of binaural pitches: average results and standard error over 13 normal-hearing listeners andindividual results from eight hearing-impaired listeners. The mean value of the three runs for each subject was used in the calculations. The ninth subplotin the lower-right corner indicates the results obtained by Akeroyd et al. (2001) with normal-hearing listeners.


1993) was performed. Results from the statistical tests aresummarized in Table 2.

In their study, Akeroyd et al. (2001) found that, inNH listeners, the three binaural-pitch types they tested(HP, BEP, and BICEP) were less salient than pure-tonesignals. They also found that HP was the most salientbinaural-pitch configuration, followed by BEP andBICEP. The average results over the four runs performedin their experiment are indicated in the lower-right sub-plot in Fig. 5. Average results from the current studyseem to follow the same trend as theirs, although melodyrecognition using HP and BEP does here not significantlydiffer from that obtained with pure tones (Table 2). HP

Table 2Results from a set of Wilcoxon matched-pairs signed-ranks tests for eachexperiment

Tone HP BEP B

Tone – No (18/9) No (19/12) YHP No (10.5/8) – No (4.5/8) YBEP Yes (0/8) Yes (1.5/8) – YBICEP Yes (0/8) Yes (0/8) No (6/8) –RBICEP Yes (0/8) Yes (0/8) No (3/7) NBBP No (4/7) No (6/7) Yes (0/7) YBIBP Yes (0/8) Yes (1.5/8) Yes (1/8) Y

Comparisons for the 13 normal-hearing subjects are given above the diagonaldiagonal. For a given pair, ‘Yes’ indicates that the results differ significantlyindicate the corresponding T and N values (T/N). A significance level of 0.05

actually seems to produce the same clear pitch perceptionas the pure tones, which is consistent with the results ofthe HP discrimination experiment (Section 3.2), whichsuggested that HP was as musical as pure tones in thefrequency range considered here. BICEP was found tobe significantly less salient than BEP, which itself is onaverage less salient than HP. Results obtained with thethree additional pitch types showed a clear pattern. RBI-CEP was found to be significantly less salient than allother pitch types. The subjects generally found the melo-dies more difficult to perceive with RBICEP than withBICEP. Hartmann and McMillon (2001) found thatRBICEP was easier to perceive than BICEP for low

pairwise comparison of the pitch types used in the melody recognition

ICEP RBICEP BBP BIBP

es (0/11) Yes (0/13) No (5/7) No (17/9)es (2/11) Yes (0/13) No (24/10) No (17/9)es (4/12) Yes (0/13) No (15/11) No (10.5/8)

Yes (7.5/13) Yes (1.5/11) Yes (6/13)o (14.5/8) – Yes (0/13) Yes (0/13)es (0/8) Yes (0/8) – No (24.5/10)es (0/8) Yes (0/8) No (10/8) –

. Comparisons for the eight hearing-impaired subjects are given below the, ‘No’ that no significant difference was found. Numbers in the bracketswas used.

2 The average of the results with all six binaural-pitch stimuli in themelody recognition test and the average of the ERB values from both earsin the frequency selectivity test were used to derive the correlationcoefficients. Correlation significance was determined with a significancelevel of 0.05.


boundary frequencies (below about 600 Hz), and that itwas less salient than BICEP for high boundary frequen-cies (above about 600 Hz). Frequencies of the notes usedin the 10 melodies were between 523.25 Hz and1046.50 Hz, which means that most notes, and thereforemost values of fb, lie in the ‘high-frequency’ binaural-pitch region. Therefore, in this test, RBICEP should thenbe slightly weaker than BICEP, which would explain thedifference between both configurations in the results. Therecognition obtained with the BBP and BIBP configura-tions appeared to be as good as with the pure-tone andHP configurations, although the phase-difference patternsin HP (linear phase shift), BBP (constant phase shift),and BIBP (random phase shift) are very different. Thissuggests that the pattern of the phase shift along fre-quency used in the area around fb may not be of muchimportance for the auditory system when detecting bin-aural pitch, as long as the left and right noises are inphase outside this area. The number of edges (one ortwo) in the interaural phase-difference pattern appearsto be what mostly determines the salience of the pro-duced binaural pitch, with either simple phase-shift edges(sharp or smooth) or coherence edges. HP (two smoothedges), BBP (two sharp edges), and BIBP (two coherenceedges), all created by a pattern with two edges, are signif-icantly more salient than BEP (one simple edge), BICEP(one coherence edge), and RBICEP (one coherence edge),which are created by a pattern with only one edge.Whether the edges are coherence edges or not appearsto be the secondary factor influencing the salience ofthe pitch.

For none of the HI subjects were the results similar tothe NH results for all pitch types. Thus, all types of hearingimpairment encountered in this study affected melody rec-ognition produced by binaural pitch in some way. In orderto make the results easier to read, the perception of eachbinaural-pitch type can at first be evaluated in a simpleway by dividing the vertical axis of Fig. 5 in three intervals:

(1) 70–100% of correct responses (good): the pitch is eas-ily detected by the subject and is musical, i.e. can beused to produce recognizable melodies;

(2) 30–70% of correct responses (intermediate): the pitchis detected by the subject but with a reduced salienceor musicality;

(3) 0–30% of correct responses (poor): the pitch cannotbe detected by the subject or has a very weak salienceor musicality.

With the HP configuration, all subjects obtained agood recognition, except subjects 3 and 6 whose scoreswere intermediate. The finding that none of the subjectshad poor recognition is consistent with the finding thatall of them were able to detect HP immediately in thepreliminary test and obtained a good detectability ofHP at least in the mid-frequency range in the HPdetection experiment (see Section 3.1). With BEP, no

HI subject obtained a good recognition, and BEP wassignificantly less salient than HP for the HI group (Table2). This shows an important difference between HI andNH subjects, for whom BEP almost had the same strongsalience as HP. Whereas five subjects were able to detectBEP, the other three subjects (2, 6, and 8) obtainedscores that were very close to chance level. With BICEPand RBICEP, subjects 1, 4, and 5 obtained intermediaterecognition for both pitch types, subject 7 obtained anintermediate recognition for BICEP and a poor recogni-tion for RBICEP, and all other subjects had a poor rec-ognition for both pitch types. Among the three subjectshaving intermediate scores, three of them (subjects 4, 5,and 7) found BICEP easier to perceive than RBICEP,which shows the same trend as observed with the NHsubjects. The scores obtained with BICEP and RBICEPwere significantly lower than with HP for all subjects,and lower than with BEP for all subjects except subject4. BICEP and RBICEP clearly appeared to be the twomost difficult pitch types to perceive for the HI subjects,all scores lying below 50% except for subject 4. The threesubjects who supposedly could not perceive BEP all hada poor recognition, suggesting that they also have a veryweak or no perception of the binaural coherence edgepitches. Results obtained with BBP are very similar tothose obtained with HP, as was also found for the NHsubjects. However, while the results for HP and BIBPwere similar for the NH listeners, BIBP appeared to besignificantly less salient than HP for the HI group. Theresults were not correlated to the average value of thesubjects’ pure-tone hearing thresholds between 125 Hzand 2000 Hz, with a correlation coefficient2 r = �0.09.A patient with a normal audiogram may indeed still suf-fer from some deficiency of the auditory system: subjects1 and 2 had normal-hearing thresholds and could theneasily have been considered as being NH subjects, butthe results they obtained in this experiment show thatthere must be deficits in their binaural processing ofsounds.

The main results for the HI listeners can be summarizedby dividing the subjects into three groups:

(1) Listeners who can perceive all types of pitches clearly,but for whom binaural pitch generally is less salientthan for NH persons. For these listeners, as in NHlisteners, the most salient pitches are HP, BBP, andBIBP, followed by BEP, BICEP, and RBICEP. Sub-jects 1, 4, and 5 belong to this group.

(2) Listeners who can perceive binaural pitches clearlyexcept when coherence edges are used to generatethem. These listeners perceive HP, BBP, and BEP


as subjects from group 1, but find it very difficult tohear BICEP and RBICEP. Subjects 3 and 7 belongto this group.

(3) Listeners who can perceive binaural pitches as thosefrom group 2, except with the BEP configuration,which they find almost impossible to hear. Subjects2, 6, and 8 belong to this group. These three subjectsall mentioned that they could not hear the melodies atall when they were played with BEP and BICEPstimuli.

4. Relation with measures of spectral and temporal resolution

The three following tests were aimed at characterizingmore precisely the hearing impairment of each HI subject.They could also be used to investigate the correlationbetween the performance in the binaural-pitch experi-ments, and that in basic tasks of frequency discrimination,frequency selectivity, and temporal gap detection.

4.1. Frequency discrimination

This first additional test was aimed at measuring thesubjects’ just noticeable difference (JND) in frequency ofpure tones, HP, and BIBP, at a frequency of 500 Hz. All10 HI listeners as well as 8 NH listeners participated in thisexperiment.

Table 3(a) Frequency just noticeable difference, (b) equivalent rectangular bandwidth

(a) Freq. JND at 500 Hz (%)

Tone HP BIBP

NH subjects

Average 0.6 2.3 4.3Subject A – – –Subject B – – –Subject C – – –Subject D – – –Subject E – – –Subject F – – –Subject G – – –

HI subjects

Subject 1 2.5 12.4 22.2Subject 2 0.8 3.2 15.5Subject 3 23.1 28.2 30.9Subject 4 1.2 5.4 15.1Subject 5 0.6 2.2 3.0Subject 6 9.1 15.4 11.5Subject 7 0.7 6.4 25.9Subject 8 8.4 12.3 11.0Subject 9 29.9 – –Subject 10 17.4 – –

(a) Frequency just noticeable difference (JND) of pure-tone, HP, and BIBP saverage over eight normal-hearing subjects and individual results from 10 hea(b) Equivalent rectangular bandwidth (ERB), in Hz, of the auditory filter cent(both ears). The ERB value suggested by Glasberg and Moore (1990) is 78.7 H(c) Broadband noise gap detection threshold values, in ms, measured for nine helevel of 75 dB SPL. The normal-hearing gap detection threshold as measured

4.1.1. Method

The frequency JND was measured using a three-inter-val, adaptive 3-AFC procedure. On each trial, two intervalshad an identical frequency of 500 Hz, whereas the otherrandomly chosen interval had a frequency equal to500(1 + x/100) Hz, where x is the tracking variableexpressed as a percentage of the test-frequency. The stimulihad a duration of 500 ms and were separated by 500 ms ofsilence. The task of the listener was to indicate through acomputer interface in which interval the pitch frequencywas higher than in the two other intervals. Feedback wasprovided to the subject after each response. A 1-up, 2-downtracking rule converging at the 70.7% point of the psycho-metric function (Levitt, 1971) was used. The starting valueof x was 25%, and the step size was decreased after eachupper reversal, with successive reduction factors of 2, 1.5,and 1.125. One run stopped after 12 reversals, and theJND was calculated as the average peak value of x overthe last eight reversals. A total of 6 runs were performed,two with each type of stimulus.

4.1.2. Results and discussionThe frequency JND was measured for each stimulus

type as a percentage of the test frequency. The averageresults over the 8 NH subjects as well as individual resultsfrom the HI subjects are given in Table 3a.

The frequency JND of the 500 Hz pure tone was onaverage found to be below 1% (5 Hz) for all NH subjects.

and (c) broadband noise gap detection

(b) ERB at 500 Hz (Hz) (c) Gap detection

Right ear Left ear Threshold (ms)

97.6 97.6 4.6 ± 0.692.8 87.8 4.8 ± 0.698.8 95.8 –

110.8 122.6 –116.0 112.2 4.5 ± 0.778.1 76.4 4.9 ± 0.685.3 94.2 4.8 ± 0.5

– – 3.8 ± 0.6

– – –102.3 105.8 4.6 ± 1.1104.5 115.5 4.4 ± 0.7138.8 116.8 6.7 ± 0.9100.5 105.0 6.2 ± 0.9109.3 104.6 6.8 ± 0.9129.4 123.0 3.9 ± 0.6171.5 187.6 6.2 ± 0.6157.8 >200.0 12.9 ± 2.1187.1 188.1 10.7 ± 1.1

timuli expressed as a percentage of the (boundary) frequency of 500 Hz:ring-impaired subjects.ered at 500 Hz for six normal-hearing and nine hearing-impaired subjects

z.aring-impaired subjects and five normal-hearing subjects at a presentation

by Zeng et al. (2005) at a presentation level of 50 dB SPL is 3 ms.


The average result over all subjects was 0.6%, correspond-ing to 2.9 Hz. This is slightly higher than the value of 1–2 Hz generally obtained at 500 Hz (Moore, 2003). Withthe HP stimulus, the average JND increased to 2.3%(11.7 Hz), and reached an average of 4.3% (21.4 Hz) forthe BIBP stimulus. The JND value obtained for HP is rel-atively large compared to that found by Hartmann (1993)in his pitch matching experiment, probably due to therather large width of the HP boundary region used here(16% of fb). These results suggest that the pitch frequencyof binaural pitches is not as precise as that of pure tones,which is consistent with the perception reported by somelisteners that binaural pitches sound more like narrow-band noise signals than tones (especially BEP, BICEP,and RBICEP stimuli). The presence of coherence edges inthe BIBP stimulus spread the pitch frequency around fb

to a greater extent. However, in all three cases, theobtained frequency JNDs were sufficiently small, i.e. lessthan 9% or 1.5 musical semitones as used in the HP dis-crimination experiment (see Section 3.2). This suggests thatall three pitches are musical around 500 Hz.

For the HI subjects, four subjects showed results similarto the NH average (subjects 2, 4, 5, and 7) in the pure-tonecondition. The other six subjects obtained higher JNDsthan NH listeners, ranging from 2.5% for subject 1 to 8–10% for subjects 6 and 8, and up to 23% for subject 3. Notethat the two subjects who could not perceive binaural pitch(9 and 10) obtained considerably higher JNDs for puretones than the NH average. In the HP condition, only sub-jects 2 and 5 showed similar results to NH listeners, and allother subjects obtained higher JND values, ranging from5% for subject 4 to 28% for subject 3. In the BIBP condi-tion, only subject 5 showed a very good frequency discrim-ination, even compared to the NH average. All othersubjects obtained much higher JND values, ranging from11% for subject 8 to 31% for subject 3. For subjects 1, 2,3, 4, 5, and 7, frequency discrimination was easiest forpure-tone stimuli. Their JNDs were higher in the HP con-dition than in the pure-tone condition, and higher in theBIBP condition than in the HP condition. For subjects 6and 8, the HP stimuli were the most difficult to discrimi-nate, followed by BIBP and pure-tone stimuli.

The frequency discrimination abilities of the subjectswere not related to audibility as reflected in their audio-gram. For example, subject 3 had an audiogram showinga very mild impairment, whereas their abilities to discrim-inate between frequencies appeared to be severely affected.The opposite is true for subject 5, who has very high hear-ing thresholds at high frequencies (above 80 dB HL), but isactually the only subject who showed a normal frequencydiscrimination for the three stimuli. Moreover, these twosubjects have a noise-induced hearing loss, suggesting thattwo hearing impairments from the same origin can lead tovery different perceptual problems. The two subjects withnormal audiograms (1 and 2) obtained different resultsfor frequency discrimination. Subject 2 obtained compara-ble results to NH listeners with the pure-tone and HP stim-

uli, but had more difficulties in the BIBP condition. Subject1 had a slightly higher JND than NH listeners in the pure-tone condition, but had obvious difficulties in discriminat-ing both HP and BIBP.

The results are consistent with those obtained in the HPdiscrimination and melody recognition tests (Sections 3.2and 3.3), in which subjects 3, 6, and 8 showed the largestdifficulty in discriminating pure-tone and HP stimuli, aswell as recognizing pure-tone melodies. These three sub-jects also have the highest frequency JNDs for pure tones.This confirms our expectation that problems with fre-quency discrimination potentially affect the musical abili-ties of some subjects.

4.2. Frequency selectivity

The second additional test measured the frequency selec-tivity abilities of the subjects. The bandwidth of the audi-tory filter centered at 500 Hz was evaluated monaurallyin both ears of 9 HI listeners as well as 6 NH listeners. Thisfrequency was chosen as it is the most salient for all listen-ers (Fig. 3) for the detectability of HP.

4.2.1. Method

The notched-noise method (Patterson, 1976) was usedfor determining auditory filter shapes. The signal was a500 Hz tone and the spectrum level of the noise was40 dB. The detection threshold for the signal tone was mea-sured at six relative notch widths g = Df/fc: 0, 0.05, 0.1, 0.2,0.3, and 0.4, where Df is the notch width and fc the signalfrequency. The lower and upper cut-off frequencies of thenoise were equal to fc(0.6 � g) and fc(1.4 + g), respectively.A three-interval, adaptive 2-AFC procedure was used foreach threshold measurement. In each trial, the three inter-vals contained the notched noise, one randomly choseninterval also contained the tone signal. The subject’s taskwas to detect the interval containing the pure tone. The ini-tial level of the pure tone was 70 dB SPL. A 1-up, 2-downtracking rule with four step sizes of 8 dB, 4 dB, 2 dB, and1 dB was used. The step size was halved after each secondreversal. Using the average value over the last six reversals,the mean threshold value was calculated as a function ofthe relative notch width g for each subject. A rounded-exponential filter (Patterson et al., 1982) was fitted to theexperimental data using a least-square fit. The equivalentrectangular bandwidth (ERB), of the subject’s auditory fil-ter at 500 Hz, was then derived from the parameters of thefitted filter.


The ERB values obtained for both ears of each subjectare given in Table 3b. Although the ‘normal’ ERB valueat 500 Hz, as suggested by Glasberg and Moore (1990), isequal to 78.7 Hz, the results obtained with the NH listenersof the present study show a high variability among sub-jects, with ERB values ranging from 76 Hz to 122 Hz. Itis therefore not possible to consider the HI subjects who


obtained ERB values in that interval in both ears (subjects2–6) as having impaired frequency selectivity compared tothe NH listeners of the present study. However, all otherHI subjects do show a reduced frequency selectivity. Theright-ear auditory filters of HI subjects 4 and 7(ERB = 139 Hz and 129 Hz) appear to be slightly broaderthan normal, while HI subjects 8–10 obtained ERB valuesthat were considerably higher than those of all other sub-jects. The two subjects who could not perceive any binauralpitch at all (HI subjects 9 and 10) are among those with thehighest ERB values. However, the correlation between theERB values obtained for HI subjects and their perfor-mance in the melody recognition experiment is not signifi-cant, with a correlation coefficient2 r = �0.63.

4.2.3. Effect of broadening auditory filters

The differences in interaural correlation as a function offrequency, inherent to binaural-pitch stimuli, are essentialto create binaural pitch. Because the bandwidth of theauditory filters has a direct influence on the internal repre-sentation of the interaural-correlation spectrum, animpaired frequency selectivity should limit the ability ofthe listener to perceive binaural pitch. For instance, forthe BICEP stimulus, the waveform interaural correlation(before auditory filtering) is equal to 1 below fb and 0above fb (Fig. 6, upper left panel). Once passed through a

Fig. 6. Modification of the BICEP stimulus using interaural correlation in o

set of auditory filters, this external waveform interaural-correlation spectrum is transformed into an internal inter-aural-correlation spectrum in which the decay around fb isnot sharp anymore but progressive (Fig. 6, lower leftpanel). The edge in this internal interaural-correlationspectrum becomes smoother if the auditory filters arebroadened. If one assumes that the internal interaural-cor-relation representation is a key element used by the binau-ral system in the processing of binaural signals, it mightthen be possible to further investigate the influence of audi-tory-filter broadening on binaural-pitch perception bymodifying the external waveform interaural-correlationspectrum of the stimulus itself. The external interaural-cor-relation spectrum of the BICEP stimulus was modified here(Fig. 6, upper right panel) such that the internal interaural-correlation spectrum it produced when fed through normalauditory filters matched the internal interaural-correlationspectrum produced by the original BICEP stimulus whenfed through a set of impaired auditory filters (Fig. 6, lowerright panel). This was done by introducing a transition inthe external interaural-correlation spectrum instead of asharp edge. In this way, it was possible to simulate theeffect of broadened auditory filters and to measure the con-sequences of this on binaural-pitch perception in NH lis-teners. The gammatone filterbank (Patterson et al., 1995)provided in the auditory toolbox by Slaney (1998) was used

rder to simulate its perception by a listener with broader auditory filters.


to simulate auditory filtering, and the broadening factor b

was varied to simulate different filter bandwidths.The melody recognition ability of four of the NH sub-

jects was measured with pure tones, BICEP stimuli, andmodified versions of the BICEP with broadening factorsof 1.2, 1.6, and 2.0. The procedure was identical to thatdescribed in Section 3.3.1. The experimental results aregiven in Fig. 7. The melody recognition abilities of the sub-jects were found to decrease progressively with increasingvalues of b, which suggests that the bandwidth of the audi-tory filters might play an important role in the ability of alistener to perceive BICEP in a musical way. The strengthof this effect was different for the different subjects. Inter-estingly, a similar variability was also found when estimat-ing the frequency selectivity of the same four subjects at500 Hz (Table 3b). The two subjects whose melody recog-nition abilities decreased the most rapidly with increasingb (subjects C and D) were also those with the highestERB values at 500 Hz, which supports the idea that perfor-mance with frequency selectivity and melody recognition ofBICEP might be related.

In an additional stimulus configuration, reduced audi-bility at high frequencies was also simulated by attenuatingthe amplitude of the high-frequency (HF) components ofthe original BICEP stimulus. The amplitude of frequencycomponents above 500 Hz was attenuated progressivelywith increasing frequency, with reductions of 5 dB at

Tone b=1.0 b=1.2 b=1.6 b=2.0 HF loss

0

10

20

30

40

50

60

70

80

90

100

Pitch type

Per

cen

t co

rrec

t

Subject ASubject BSubject CSubject D

Fig. 7. Melody-recognition of pure-tones, BICEP and four modifiedversions of the BICEP stimulus in four normal-hearing listeners.b = 1.0:BICEP; b = 1.2–1.6–2.0: modified BICEP (simulating broaderauditory filters); HF-loss: BICEP with reduced amplitude at highfrequencies (simulating a higher threshold). The equivalent rectangularbandwidths (ERB) of the left and right auditory filters centered at 500 Hzfor the same four subjects are indicated in Table 3b.

1 kHz, 35 dB at 2 kHz, 65 dB at 3 kHz, and 80 dB at4 kHz. This modification did not appear to have any effect(‘HF-loss’ in Fig. 7), i.e. the results obtained with the ‘HF-loss’ configuration were similar to those obtained with theoriginal BICEP.

4.3. Temporal resolution

The third additional test measured the temporal resolu-tion of 9 HI and 5 NH listeners, using a gap detectionthreshold measurement.

4.3.1. Method

Temporal resolution was measured by estimating thedetectability of temporal gaps in broadband noise. A fixedlevel of 75 dB SPL was used in this experiment. A three-interval, adaptive 3-AFC procedure was used. All intervalscontained white noise with a bandwidth of 3000 Hz and anupper cut-off frequency of 4000 Hz. One random intervalcontained a gap in the noise. A 1-up, 2-down tracking rulewas used. The initial gap duration was 120 ms, and threestep sizes corresponding to a reduction factor of 2, 1.5,and 1.125 were used. The step size was reduced after eachsecond reversal. The gap detection threshold was calcu-lated as the average over the last eight reversals. A totalof three runs was performed by each subject.


The gap detection thresholds obtained for each subjectare given in Table 3c. Plomp (1964) and Penner (1977) pro-posed a typical threshold value of 2–3 ms for NH listeners,while Zeng et al. (2005) found the NH threshold to beabout 3 ms. The values obtained in the present study wereslightly higher, with an average value of 4.6 ms for NH lis-teners. HI subjects 2, 3, and 7 obtained results close to theNH average, while HI subjects 4, 5, 6, and 8 obtainedslightly higher thresholds between 6 and 7 ms. The twoHI subjects who could not perceive binaural pitch at all(HI subjects 9 and 10) obtained considerably higher thresh-olds than all other subjects. Overall, there is a significantcorrelation between the gap detection thresholds obtainedfor HI subjects and their performance in the binaural-pitchexperiments, as indicated by a correlation coefficient2

r = �0.79. Furthermore, comparing the gap detectionresults with the pure-tone hearing threshold of the subjects,averaged over all audiometric test frequencies, gives a sig-nificant correlation coefficient2 r = 0.74. This seems consis-tent with earlier studies on gap detection. HI subjects oftenshow reduced temporal resolution as a result of the lowsensation level of the stimuli and the reduced audible band-width of the stimuli (e.g. Florentine and Buus, 1984; Salviand Arehole, 1985).

As indicated earlier, reduced frequency selectivity couldaffect binaural-pitch perception already at a cochlear stage.It is therefore difficult to separate the effects caused by defi-ciencies in temporal processing from those caused byimpaired frequency selectivity if the psychophysical


experiments are performed in subjects who also havebroadened auditory filters. It would be interesting to inves-tigate binaural-pitch perception in a group of subjects withsimilar frequency selectivity (e.g. subjects with similarouter-hair-cell functionality) but with different aspects oftemporal deficiencies at cochlear or more central stagesof the auditory system.

5. General discussion

5.1. Summary of the main results

First, results for the normal-hearing listeners are sum-marized. A preliminary test with the Huggins’ pitch(HP) showed that it produced an immediate sensation.In the first experiment, the detectability of HP was mea-sured as a function of frequency, and the pitch was foundto be most salient for boundary frequencies between200 Hz and 1250 Hz, with a declining detectability below200 Hz and above 1250 Hz. The second experiment was apitch discrimination test and aimed at measuring the musi-cality of HP. The pitch was found to be as musical as thatproduced by pure tones for boundary frequencies between200 Hz and at least 800 Hz. Discrimination of HPappeared to be progressively more difficult below 200 Hzand above 800 Hz. In the third experiment, the melody

recognition abilities of the listeners were measured withpure tones and six different types of binaural pitches. Itwas found that the salience of binaural pitch was largestfor HP, followed by the binaural edge pitch (BEP) andthe binaural coherence edge pitch (BICEP). Using a con-stant or random phase shift around the boundary fre-quency instead of the linear phase shift of HP did notaffect the salience and musicality of the produced pitch.In the case of BICEP, the pitch strength was found tobe higher if coherent noise was used above the edge thanbelow the edge, for edge frequencies between 520 Hz and1050 Hz. The just noticeable difference in frequency (JND)at 500 Hz was measured for pure-tone, HP, and BIBPstimuli. The average JND was found to be higher forHP than for pure tones, and higher for the binaural inco-herent-band pitch (BIBP) than for HP. The JNDs werefound to be low enough to allow the identification of mel-odies played with those three stimuli for pitch frequenciesaround 500 Hz. Estimates of auditory filter bandwidth atthe same frequency using the notched-noise methodshowed a surprisingly large variability among thenormal-hearing listeners. A test simulating the effect of

broadening of the auditory filters by modifying the inter-aural-correlation spectrum showed that binaural-pitchperception became more difficult as the simulatedbandwidth of the filters was increased. The strength ofthe effect was not the same among subjects, and subjectswho already had relatively broad auditory filtersperformed worse if the smoothed interaural-correlationspectrum was introduced. The simulation of reducedaudibility did not affect binaural-pitch perception.

The same set of experiments was also carried out by 10hearing-impaired (HI) listeners with different types of audi-tory deficiencies. It was found out that the two HI listenerswho reported central auditory processing deficits could notperceive binaural pitch at all. However, for most HI listen-ers, binaural pitch was perceivable, and in all cases pro-duced an immediate sensation as in the normal-hearing(NH) listeners. Hearing impairment had a minor effect onthe detectability of HP. The range of boundary frequenciesin which the pitch is easily detectable was the same forsome HI listeners as for NH listeners. However, for severalHI listeners the frequency region in which the pitch wasmost salient was reduced on the low-frequency side orthe high-frequency side, or on both sides. Discriminationof HP stimuli with close boundary frequencies was foundto be typically more difficult for HI listeners than for NHlisteners. In contrast to NH listeners for whom all binau-ral-pitch types were shown to be musical, the melody recog-

nition abilities of HI listeners appeared to be very differentdepending on the pitch type used. Despite the fact that theycould detect HP easily, some HI listeners only showed avery weak or non-existing perception of BICEP. Amongthese, some also showed a very weak or non-existing per-ception of BEP. The JND in frequency at 500 Hz was foundto be considerably higher for some HI subjects than theNH average, and the same general trend was observedfor both subject groups: for most subjects BIBP was moredifficult to discriminate than HP, which was itself more dif-ficult to discriminate than pure tones. The difficulties of HIsubjects with frequency discrimination were consistent withtheir discrimination and melody recognition performancesin the binaural-pitch experiments. Results from the fre-

quency selectivity test were not significantly correlated tothose from the melody recognition experiment. However,subjects who could not hear binaural pitch at all were alsothose who had the broadest auditory filters at 500 Hz. Thelargest gap detection thresholds were obtained by the twosubjects who could not perceive binaural pitch at all. Per-formance in the gap detection experiment was overall cor-related to melody recognition with binaural-pitch stimuli.However, it appeared to be influenced by the reduced sen-sation level or the reduced audible bandwidth of the stimuliby the subjects.

5.2. Modeling binaural pitch

The experimental results of the present study can be dis-cussed in the light of the modified equalization–cancella-tion (mEC) model, as described by Culling et al. (1998).In this model, the left and right stimuli are fed through abank of gammatone filters (Patterson et al., 1995) and amodel simulating hair-cell transduction (Meddis et al.,1990). Each frequency channel is then treated indepen-dently during an EC process in which the RMS values ofthe left and right signals are first equalized and an interau-ral time delay s giving an optimal cancellation is intro-duced. A spectrum of residual activation is obtained after


the cancellation process. Using the auditory toolbox bySlaney (1998), the binaural-pitch signals used in the melodyrecognition experiment (Fig. 2) were fed through the modelfor a boundary frequency of 500 Hz. Signals were 300 mslong and 80 frequency channels between 100 Hz and22,050 Hz were used. For the HP, BBP, and BIBP config-urations, the bandwidth of the phase shifted area was equalto 16% of the boundary frequency. For the BEP configura-tion, no transition was used, as was indicated in Fig. 2b.The obtained residual-activation spectra for five averagediterations of the model are plotted in Fig. 8.

In the case of HP, BBP, and BIBP, where the left andright noises are in phase outside the frequency range closeto fb, the cancellation can be performed almost perfectlyusing a zero time delay (s = 0 ms), except in the areaaround fb, where a complete cancellation cannot bereached for any value of s. This gives rise in all three casesto a peak in the residual-activation spectrum (Figs. 8a, 8e,and 8f). It is this peak, centered on the boundary fre-quency, which is assumed to account for the presence of

125 250 500 20001000–20

–15

–10

–5

0

5

10

Frequency (Hz)

Res

idu

al a

ctiv

atio

n (

dB

) b=1b=2

(a) Huggins’ pitch (HP)

125 250 500 20001000–20

–15

–10

–5

0

5

10

Frequency (Hz)

Res

idu

al a

ctiv

atio

n (

dB

) b=1b=2

(c) Binaural coherence edge pitch(BICEP)

125 250 500 20001000–20

–15

–10

–5

0

5

10

Frequency (Hz)

Res

idu

al a

ctiv

atio

n (

dB

) b=1b=2

(e) Binaural band pitch (BBP)

Fig. 8. Residual activation spectra recovered by the modified equalization–cafrequency of 500 Hz. b = 1: normal auditory filters; b = 2: broader auditory filtcoherence edge pitch, (d) reversed binaural coherence edge pitch, (e) binaural

the binaural pitch according to the mEC model. The modelpredicts almost no difference in perception between thesethree stimuli, because the residual-activation spectra arealmost identical, a similarity which was also observed byAkeroyd and Summerfield (2000) for HP and BIBP. Thisis consistent with the perceptual results from the melodyrecognition experiment (Section 3.3.2), where scoresobtained by the listeners were very similar for HP, BBP,and BIBP. For the BEP signal (Fig. 8b), cancellation canonly occur almost completely below the boundary fre-quency, where the left and right signals are in phase,whereas cancellation is only partially achieved above fb,where the signals are in antiphase. This is due to the factthat, for noise signals, a single interaural phase delay cor-responds to slightly different interaural time delays at sev-eral frequencies close to each other. Within the mECmodel that uses a single value of the time delay s per fre-quency channel, this means that cancellation cannot beobtained for all frequencies within a single channel (Cullinget al., 1998). The presence of a peak is due to the fact that

125 250 500 20001000–20

–15

–10

–5

0

5

10

Frequency (Hz)

Res

idu

al a

ctiv

atio

n (

dB

) b=1b=2

(b) Binaural edge pitch (BEP)

125 250 500 20001000–20

–15

–10

–5

0

5

10

Frequency (Hz)

Res

idu

al a

ctiv

atio

n (

dB

) b=1b=2

(d) Reversed BICEP (RBICEP)

125 250 500 20001000–20

–15

–10

–5

0

5

10

Frequency (Hz)

Res

idu

al a

ctiv

atio

n (

dB

) b=1b=2

(f) Binaural incoherent-band pitch (BIBP)

ncellation model for six types of binaural-pitch signals with a boundaryers by a factor of 2. (a) Huggins’ pitch, (b) binaural edge pitch, (c) binauralband pitch, (f) binaural incoherent-band pitch.


frequency channels close to fb admit frequency componentswith a wide range of different time delays, thus making can-cellation more difficult (Culling et al., 1998). Here again, itis this peak which is assumed to account for the presence ofa binaural pitch. The fact that cancellation is poorer on oneside of fb with the BEP than with the HP stimulus couldexplain why the salience of BEP is generally found to beweaker than that of HP. Because BBP is in fact similarto the combination of a BEP stimulus and a reversedBEP stimulus with close boundary frequencies, the peakobtained with the BBP stimulus looks like the combinationof two neighboring peaks which would be produced byBEP stimuli, but are not resolved here due to the narrowbandwidth of the phase-shifted area of BBP. In the caseof the BICEP and RBICEP configurations (Figs. 8c and8d), cancellation occurs almost perfectly on the coherentside of fb, but becomes very difficult on the incoherent side,due to the random phase shifts introduced. The perceptionof a binaural pitch in these cases can only be explained bythe presence of an edge in the residual-activation spectrum,which means that pitch detection could be produced by theenhancement of such an edge by the auditory system. Thecontrast enhancement at the edge could be due to a processof lateral inhibition similar to that possibly responsible forperception of a weak pitch near the edge frequency of ahigh-pass or low-pass noise (Klein and Hartmann, 1981;Culling et al., 1998). Akeroyd et al. (2001) demonstratedthat adding a simple contrast enhancement step to such amodel could predict the differences in salience betweenHP, BEP, and BICEP. However, such a process is not con-sistent with the conclusions of Hartmann (1984a), whofound no evidence for the existence of central lateralinhibition.

The presence of edges in the interaural phase-differencepattern could be what mostly influences the salience of theproduced binaural pitch. A parallel may be drawn betweenthe importance of the number of edges in the phase-differ-ence pattern and in the residual-activation pattern resolvedby the mEC model. The most salient pitches are indeedthose which give rise to a peak of residual activation(HP, BBP, BIBP). These are followed by the BEP, whichcreates a peak but with a major decay only on one side,and the least salient are those which only create an edgein the residual-activation spectrum (BICEP, RBICEP).Moreover, if an EC process is performed by the auditorysystem, the results obtained with HI subjects in the melodyrecognition experiment then suggest that subjects whocould not perceive BICEP and RBICEP (groups 2 and 3,see Section 3.3.2) had difficulties detecting an edge, butnot a peak, in the residual-activation spectrum. Assumingthat the auditory system performs a contrast enhancementprocess, in order to detect the edges in the spectrum,implies that these subjects could not perceive BICEP andRBICEP because of some deficiency in the mechanismunderlying lateral inhibition.

It can be seen in Fig. 8 that the width of the auditory fil-ters affects the sharpness of the peak or edge, but hardly

affects its amplitude (dashed curves in Fig. 8). Peak ampli-tude does therefore not seem to account correctly for thesalience of the binaural pitch in this model, which is notconsistent with current hypotheses that pitch salience isdetermined by the amplitude of the residual activation(Culling et al., 1998). Besides its amplitude, the sharpnessof the peak and the amount of cancellation achievable bythe system, on both sides of the peak, seem to be twoparameters related to the salience of the binaural pitch.This favors a pitch detection process in which the differencein residual activation between neighboring frequencychannels would play the most important role. A processof contrast enhancement, which involves such anacross-frequency comparison, has been suggested in earlierstudies (e.g. Akeroyd et al., 2001). As an alternativeapproach to the purely temporal processing assumed bythe mEC model, a model that evaluates the spatial cross-correlation of non-corresponding left and right frequencychannels could be useful when interpreting binaural pitchdata (Loeb et al., 1983; Shamma et al., 1989; Shamma,2001).

5.3. Implications and conclusions

Binaural pitch perception in hearing-impaired listenerswas found not to be related to the shape of their audio-grams. This confirms that additional tests are essentialto better characterize hearing impairment, and to deter-mine which functions of the auditory system are affected.Frequency selectivity proved a very interesting measure:The width of the auditory filters was found to have animportant influence on the perception of binaural stimuli,and to explain some differences observed even amongnormal-hearing listeners. The fact that different hearing-impaired listeners were found to react differently to binau-ral-pitch stimuli, and that their perception was eitherimmediate or non-existing, makes binaural-pitch stimulipotentially useful for clinical diagnostics. For example,if it was shown that only listeners with central auditorydeficiencies were unable to perceive binaural pitch, thena short binaural-pitch detection test could be designedto test deficiencies in the upper stages of the auditory sys-tem. This would be consistent with the suggestion byDougherty et al. (1998) that binaural-pitch signals couldpossibly be used to detect cognitive disorders such as dys-lexia. Further studies with groups of hearing-impairedsubjects, suffering from similar impairments, would be rel-evant in order to evaluate if such clinical applications canbe implemented. Experiments with homogeneous subjectgroups would also be essential to study the influence ofa deficiency in a specific area of the auditory system onbinaural-pitch perception.

Acknowledgement

We would like to thank our colleagues at the Centre forApplied Hearing Research for fruitful discussions and two


anonymous reviewers for their helpful and constructive re-marks and suggestions.

Appendix A. Supplementary data

Supplementary data associated with this article can befound, in the online version, at doi:10.1016/j.heares.2006.09.013.

References

Akeroyd, M.A., Summerfield, A.Q., 2000. The lateralization of simpledichotic pitches. J. Acoust. Soc. Am. 108 (1), 316–334.

Akeroyd, M.A., Moore, B., Moore, G.A., 2001. Melody recognition usingthree types of dichotic-pitch stimulus. J. Acoust. Soc. Am 110 (3, Pt. 1),1498–1504.

Cramer, E.M., Huggins, W.H., 1958. Creation of pitch through binauralinteraction. J. Acoust. Soc. Am. 30, 413–417.

Culling, J.F., 1999. The existence region of Huggins’ pitch. Hear. Res. 127,143–148.

Culling, J.F., Summerfield, A.Q., Marshall, D.H., 1998. Dichotic pitchesas illusions of binaural unmasking – I – Huggins’ pitch and thebinaural edge pitch. J. Acoust. Soc. Am. 103 (6), 3509–3526.

Dougherty, R.F., Cynader, M.S., Bjornson, B.H., Edgell, D., Giaschi,D.E., 1998. Dichotic pitch: a new stimulus distinguishes normal anddyslexic auditory function. Neuroreport 9 (13), 3001–3005.

Durlach, N.I., 1960. Note on the equalization–cancellation theory ofbinaural masking level differences. J. Acoust. Soc. Am. 32, 1075–1076.

Fisher, L.D., van Belle, G., 1993. Biostatistics: A Methodology for theHealth Sciences. Wiley Science.

Florentine, M., Buus, S., 1984. Temporal gap detection in sensorineuraland simulated hearing impairments. J. Speech. Hear. Res. 27 (3), 449–455.

Glasberg, B.R., Moore, B.C.J., 1990. Derivation of auditory filter shapesfrom notched-noise data. Hear. Res. 47, 103–138.

Hartmann, W.M., 1984a. A search for central lateral inhibition. J. Acoust.Soc. Am. 75 (2), 528–535.

Hartmann, W.M., 1984b. Binaural coherence edge pitch. J. Acoust. Soc.Am. 75, S22.

Hartmann, W.M., 1993. On the origin of the enlarged melodic octave. J.Acoust. Soc. Am. 93 (6), 3400–3409.

Hartmann, W.M., McMillon, C.D., 2001. Binaural coherence edge pitch.J. Acoust. Soc. Am. 109 (1), 294–305.

Hartmann, W.M., Zhang, P.X., 2003. Binaural models and the strength ofdichotic pitches. J. Acoust. Soc. Am. 114 (6, Pt. 1), 3317–3326.

Julesz, B., 1971. Foundations of Cyclopean Perception. University ofChicago Press, Chicago, IL.

Klein, M.A., Hartmann, W.M., 1981. Binaural edge pitch. J. Acoust. Soc.Am. 70 (1), 51–61.

Levitt, H., 1971. Transformed up–down methods in psychoacoustics. J.Acoust. Soc. Am. 49, 467–477.

Loeb, G.E., White, M.W., Merzenich, M.M., 1983. Spatial cross-correlation: a proposed mechanism for acoustic pitch perception.Biol. Cybern. 47, 149–163.

Meddis, R., Hewitt, M.J., Shackleton, T.M., 1990. Implementation detailsof a computational model of the inner hair-cell/auditory-nervesynapse. J. Acoust. Soc. Am. 87, 1813–1816.

Moore, B.C.J., 2003. An Introduction to the Psychology of Hearing, fifthed. Academic Press.

Patterson, R.D., 1976. Auditory filter shapes derived with noise stimuli. J.Acoust. Soc. Am. 59, 640–654.

Patterson, R.D., Nimmo-Smith, I., Weber, D.L., Milroy, R., 1982. Thedeterioration of hearing with age: frequency selectivity, the criticalratio, the audiogram, and speech threshold. J. Acoust. Soc. Am. 72,1788–1803.

Patterson, R.D., Allerhand, M.H., Giguere, C., 1995. Time-domainmodeling of peripheral auditory processing: a model architecture and asoftware platform. J. Acoust. Soc. Am. 98, 1890–1894.

Penner, M.J., 1977. Detection of temporal gaps in noise as a measure ofthe decay of auditory sensation. J. Acoust. Soc. Am. 61, 552–557.

Plomp, R., 1964. The rate of decay of auditory sensation. J. Acoust. Soc.Am. 36, 277–282.

Raatgever, J., Bilsen, F.A., 1986. A central spectrum theory of binauralprocessing. Evidence from dichotic pitch. J. Acoust. Soc. Am. 80, 429–441.

Salvi, R.J., Arehole, S., 1985. Gap detection in chinchillas with temporaryhigh-frequency hearing loss. J. Acoust. Soc. Am. 77 (3), 1173–1177.

Shamma, S.A., 2001. On the role of space and time in auditory processing.Trends Cog. Sci. 5 (8), 340–348.

Shamma, S.A., Shen, N., Gopalaswamy, P., 1989. Stereausis: binauralprocessing without neural delays. J. Acoust. Soc. Am. 86 (3), 989–1006.

Slaney, M., 1998. Auditory toolbox: a MATLAB toolbox for auditorymodeling work, version 2. Technical Report No. 1998-010 (IntervalResearch Corporation).

Yost, W.A., 1991. Thresholds for segregating a narrow-band from abroadband noise based on interaural phases and level differences. J.Acoust. Soc. Am. 89, 838–844.

Zeng, F.G., Kong, Y.Y., Michalewski, H.J., Starr, A., 2005. Perceptualconsequences of disrupted auditory nerve activity. J. Neurophysiol. 93,3050–3063.

Zhang, P.X., Hartmann, W.M., 2004. Lateralization of the Huggins pitch.J. Acoust. Soc. Am. 115 (abst), 2534 (4aPP6).

http://dx.doi.org/10.1016/j.heares.2006.09.013

http://dx.doi.org/10.1016/j.heares.2006.09.013

binaural pitch perception in normal... listener

Documents

binauralpitch perception

pitch sensation

binauralpitch stimulimight

dichotic pitch

modeling pitch

binaural pitch hearing

dierent pitch types

perception of binaural