Journal of Experimental Psychology:Human Perception and Performance2004, 30, 355-364

The Octave Illusion Revisited Again

Diana DeutschUniversity of California, San Diego

The octave illusion (D. Deutsch, 1974) occurs when 2 tones separated by an octave are alternatedrepeatedly, such that when the right ear receives the high tone, the left ear receives the low tone,and vice versa. Most subjects in the original study reported hearing a single tone that alternatedfrom ear to ear, whose pitch also alternated from octave to octave, and D. Deutsch (1975a) proposedan explanation in terms of separate what and where auditory pathways. C. D. Chambers, J. B.Mattingley, and S. A. Moss (2002) argued that the perceived pitch difference generally correspondsmore to a semitone and proposed an alternative explanation in terms of diplacusis. This articleargues that Chambers et al. used problematic procedures and reports a new experiment on theoctave illusion. The findings confirm that an octave difference is generally perceived, and theyagree with the model of Deutsch (1975a) but are at variance with the diplacusis hypothesis.

I thank Trevor Henthorn for his valuable assistance in set-ting up the instrumentation for the experiment. Correspondence concerning this article shuld beaddressed to Diana Deutsch, Department of Psychology,University of California, San Diego, La Jolla, CA 92093.E-mail: dddeutsch@ucsd.edu

The octave illusion, which was originallydescribed by Deutsch (1974), is a paradoxicalauditory phenomenon that is characterized bylarge individual differences in perception. Thepattern that was first used to create this illusionconsisted of two tones that were spaced anoctave apart and that were repeatedly presentedin alternation. The identical sequence was pre-sented via headphones to both ears simultane-ously; however, when the right ear received thehigh tone, the left ear received the low tone, andvice versa. This pattern gave rise to a number ofdifferent illusory percepts, the most commonone (termed octave) being of a single tone thatalternated from ear to ear, whose pitch alsoalternated from one octave to the other in syn-chrony with the localization shift.

Deutsch (1975a) proposed a model toaccount for the octave percept, based on a

hypothesized separation between what andwhere decision mechanisms in the auditorysystem. The model, hereafter referred to as thetwo-channel model, assumes that (a) to pro-duce the perceived pitches, the frequenciesarriving at one ear are perceived, while thosearriving at the other ear are suppressed fromconscious perception and that (b) each per-ceived tone is localized at the ear receiving thehigher frequency signal, regardless of whethera pitch corresponding to the higher or lowerfrequency is in fact perceived. The modeltherefore assumes that the octave illusionresults from illusory conjunctions of pitch andlocation values.

Chambers, Mattingley, and Moss (2002)argued from a series of experiments that thephenomenology of the octave illusion differsfrom that originally described by Deutsch(1974). More specifically, they asserted that,on listening to the octave illusion, the per-ceived difference between the alternating tonesgenerally corresponds more to a semitone thanto an octave. On this basis, the authors hypoth-esized that the tones at the two ears fuse har-

1

monically to produce a pitch that corresponds tothe low tone,1 and that the slight pitch differencebetween the two alternating tones that is per-ceived is the result of diplacusis.2 They alsoreported that some of their subjects lateralizedeach tone to the ear receiving the higher fre-quency and that some lateralized each tone tothe ear receiving the lower frequency, thoughthey did not offer an explanation for the lateral-ization patterns they obtained.

In this article, I first review early findingsconcerning the phenomenology of the octaveillusion and describe the two-channel modelthat was proposed to explain the octave percept.There follows a critique of the study byChambers et al. (2002), which questions thevalidity of their observations. Because theirhypothesis could hold only if these observa-tions were valid, this hypothesis is also chal-lenged. Finally, a new experiment is reported inwhich the phenomenology of the octave illusionis documented more explicitly than in previousstudies. The findings from this new experimentconfirm that an octave difference between thealternating tones is generally perceived, andthey are in accordance with the two-channelmodel, but cannot be explained on the diplacu-sis hypothesis.

Previous Findings Concerning thePhenomenology of the Octave Illusion and the

Two-Channel Model

In the original experiment of Deutsch(1974), subjects were presented with dichoticsequences consisting of 250-ms tones. As

shown in Figure 1, the tones alternatedbetween 400 Hz and 800 Hz, such that whenthe right ear received 400 Hz, the left earreceived 800 Hz, and vice versa. The toneswere sine waves, at equal amplitude, with nogaps between tones. To minimize transients,there were no amplitude drops at the frequencytransitions, and phase continuity was preservedat the transitions.

Eighty-six subjects (53 right-handers and33 left-handers), all naive concerning theoctave illusion, were presented with a 20-ssegment of the illusion and asked to reportwhat they heard. The positions of the ear-phones were then reversed, and the procedurewas repeated. A number of different illusorypercepts were obtained, and these were divid-ed into three categories. The first, termedoctave, consisted of a single tone that alternat-ed from ear to ear, whose pitch also alternatedfrom one octave to the other in synchrony withthe localization shift. For most subjects whoobtained this percept, when the positions of

2 DEUTSCH

1 When two sine-wave tones that stand in octave relationare presented simultaneously, they fuse perceptually sothat a single tone is heard, whose pitch corresponds to thelow tone (the fundamental frequency). This effect of har-monic fusion has also been found to occur when the tonesare presented dichotically (Houtsma & Goldstein, 1972).2 The term diplacusis refers to a slight difference in per-ceived pitch (of a small fraction of a semitone) that canoccur when the same tone is presented to the left and rightears (van den Brink, 1975a, 1975b). This effect is usual-ly attributed to the structural characteristics of the ear.

Figure 1. The stimulus pattern used in the original exper-iment of Deutsch (1974) describing the octave illusion,together with the percept most commonly obtained.Black boxes indicate tones at 800 Hz and white boxesindicate tones at 400 Hz. From “An Auditory Illusion,”by D. Deutsch, 1974 (September 27), Nature, 251, p.307. Copyright 1974 by Macmillan Publishers Ltd.Adapted with permission.

OCTAVE ILLUSION REVISITED 3

3 The earphone positions were reversed, rather thanreversing the signals delivered to the earphones, in orderto control for possible differences in earphone character-istics.4 In the experiment described by Deutsch (1974), thepitch differences perceived on listening to the octave illu-sion were documented by verbal reports, which werebacked up either by musical notation or by marking theperceived pitch difference on a scale from zero to over anoctave. Because of space limitations, these methods werenot described in this article.

the earphones were reversed, the apparentlocations of the high and low tones did notreverse with them.3 The octave category of per-cept was described in 58% of right-handersand 52% of left-handers. Furthermore, thoseright-handers who obtained this perceptshowed a strong tendency to hear the high toneon the right and the low tone on the left; how-ever, the lefthanders as a group did not prefer-entially localize the high and low tones eitherway.

The second category of percept, termedsingle pitch, consisted of a single tone thatalternated from ear to ear, whose pitch eitherdid not change or changed only slightly withthe shift in the tone’s perceived location. Thispercept was described in 25% of right-handersand 9% of left-handers. The third category,termed complex, comprised a number of differ-ent complex percepts, often involving at leastthree different pitches. This category of per-cept was obtained in 17% of right-handers and39% of left-handers. The two handednessgroups differed significantly, both in terms oftype of percept obtained (with a higher propor-tion of left-handers reporting complex per-cepts) and also in terms of patterns of lateral-ization for the octave percept (with a higherproportion of right-handers reporting the hightone on the right).4

Deutsch (1983b) further studied handed-ness correlates with perception of the octaveillusion in 250 subjects. The tendency to per-ceive the high tone on the right was found to behigher among right-handers than mixed-han-ders, and it was higher among mixed-handers

than left-handers.5 Furthermore, for all threehandedness groups, the tendency to hear thehigh tone on the right was lower among thosewith left- or mixed-handed parents or siblingsthan among those with only right-handed par-ents and siblings. This pattern of results is inaccordance with the neuropsychological litera-ture relating patterns of cerebral dominance tohandedness and familial handedness back-ground (Herron, 1980) and leads to the conjec-ture that perception of the octave illusion mightserve as a reflection of the direction and degreeof cerebral dominance in most individuals.

Other work on the octave illusion wasdirected toward understanding the basis of thetype of percept termed octave. Deutsch (1975a)hypothesized that this percept results from adissociation between the what and where path-ways in the auditory system. The proposedtwo-channel model is shown in Figure 2. Toproduce the perceived pitches, the listener fol-lows the frequencies that are presented to thedominant ear and suppresses from consciousperception those that are presented to the non-dominant ear. However, the listener lateralizeseach perceived tone to the ear receiving thehigher frequency signal, regardless of whetherhe or she perceives a tone corresponding to thehigher frequency or the lower one. Because thepitch and lateralization decision mechanismshere use different rules, illusory conjunctionsof pitch and location result.

Take, as an example, a listener whose pitchperceptions correspond to the frequencies pre-sented to his or her right ear. When the hightone is presented to the right ear and the lowtone is presented to the left ear, this listener per-ceives the high tone, because this tone is pre-sented to the right ear. This listener also lateral-izes the tone to his or her right ear, because thisear is receiving the higher frequency signal.However, when the low tone is presented to theright ear and the high tone to the left ear, the lis-

5 Handedness and familial handedness background wereevaluated using the questionnaire and procedure ofVarney and Benton (1975).

tener hears the low tone, because this is present-ed to the right ear, but the listener lateralizes thetone to his or her left ear instead, because thisear is receiving the higher frequency signal. Sothe listener hears the entire sequence as a hightone to the right alternating with a low tone tothe left. However, now take a listener whosepitch perceptions correspond to the frequenciespresented to his left ear, and hold the lateraliza-tion rule constant. This listener hears the samestimulus pattern as a high tone to the left alter-nating with a low tone to the right.6

To test this hypothesis, Deutsch and Roll(1976) presented 44 right-handers with repeat-ing sequences consisting of 400 – 800-Hzdichotic tone pairs. One ear received a repeat-ing pattern consisting of three high (800-Hz)tones alternating with two low (400-Hz) tones,while simultaneously the other ear received arepeating pattern consisting of three low (400-Hz) tones alternating with two high (800-Hz)tones. The tones were 250 ms in duration andwere separated by 250-ms pauses. The subjects

made two judgments: (a) how many high tonesthey heard in sequence and how many lowtones they heard in sequence– so indicatingwhich ear they followed for pitch–and (b) howmany tones they heard in sequence in the rightear and how many in the left ear – so indicatingto which ear each tone was lateralized. Theresults were as predicted from the two-channelmodel. Concerning the pitch component, mostsubjects reported hearing the patterns of highand low tones that were presented to their rightear rather than their left. However, each tonewas lateralized to the ear that received the high-er frequency signal, regardless of whether apitch corresponding to the higher or lower fre-quency was perceived.

Further experiments (Deutsch, 1978, 1980,1981, 1988) used the two-alternative forced-choice (2AFC) method to investigate perceptionof the octave illusion under parametric manipu-lation. Here, subjects were selected who showedstrong and stable octave percepts. To explore theear dominance component, I presented subjectswith segments of the illusion and asked them toreport on each trial whether they heard a patternthat began with the high tone and ended with thelow tone (i.e., a high-low-high-low pattern) or apattern that began with the low tone and endedwith the high tone (i.e., a low-high-low-high pat-tern). In this way, the subjects indicated whichear they were following for pitch. The relative

4 DEUTSCH

Figure 2. Illustration showing how the outputs of two decision mechanisms, one determining perceived pitch and theother determining perceived location, can combine to produce the octave illusion. Black boxes indicate tones at 800Hz, and white boxes indicate tones at 400 Hz. R = right; L = left. From “Auditory Illusions, Handedness, and theSpatial Environment,” by D. Deutsch, 1983, Journal of the Audio Engineering Society, 31, p. 608. Copyright 1983by the Audio Engineering Society. Adapted with permission.

6 Deutsch (1981) elaborated on this model to provide amore concrete explanation of the octave illusion in termsof its neurophysiological underpinnings. The elaboratedmodel also accommodates findings from later parametricstudies of the illusion, including those showing adependence on the durations between onsets of succes-sive tones. Because of space limitations, this model isnot described here.

amplitudes of the tones at the two ears were var-ied, and the percentages of judgments that corre-sponded to the frequencies presented to the non-dominant ear were plotted as a function of theseamplitude relationships. The strength of the eardominance effect was then measured by the sizeof amplitude difference between the tones at thetwo ears required to counteract it. The effect wasfound to be strong for sequences in which thetwo ears received the same frequencies in suc-cession (i.e., when both the 400-Hz and the 800-Hz tones were presented in immediate succes-sion to the left and right ears, as in the originaloctave illusion pattern), but it was weaker orabsent for sequences in which this pattern ofrelationship did not hold (Deutsch, 1980). Theeffect was also found to be weaker when theinteronset interval between successive tones waslengthened to 3,000 ms, regardless of whetherthis was achieved by inserting silent gapsbetween the tones or by increasing the durationsof the tones themselves (Deutsch, 1981).

To explore the lateralization component, Iagain presented subjects with segments of theillusion but now asked them to report on eachtrial whether they heard a pattern that began atthe left ear and ended at the right ear (i.e., a left-right-left-right pattern) or a pattern that beganat the right ear and ended at the left ear (i.e., aright-left-right-left pattern). In this way, thesubjects indicated whether the tones were later-alized to the ear receiving the higher frequencyor the lower one. The relative amplitudes of thehigher and lower tones were varied, and thepercentages of judgments that corresponded tothe ear receiving the lower frequency were plot-ted as a function of these amplitude relation-ships. It was found that the subjects lateralizedthe tones to the ear receiving the higher fre-quency until the amplitude of the lower tonesexceeded those of the higher ones by a criticallevel. This effect occurred with tones present-ed in rapid repetitive sequence (as in the origi-nal illusion pattern) but was significantly weak-er when only two dichotic tone pairs were pre-sented (Deutsch, 1978) or when more complexpitch configurations were used (Deutsch,

1988).Zwicker (1984) explored the phenomenolo-

gy of the octave illusion in several experiments.Concerning the issue of perceived pitch differ-ences between the alternating tones, the authorreported, from the judgments of 15 subjects,that “mostly octaves, but also often smallerintervals were perceived [italics added]” (p.129). He also confirmed that, for 400-Hz and800-Hz signals, there was a strong tendency tolateralize the tones to the ear receiving the high-er frequency. In a further experiment, he pre-sented 8 subjects with octave illusion patternsat tone durations (and, so, interonset intervals)ranging from 0.01 s to 2.00 s. He wrote,

The observers’ certainty in perceivingDeutsch’s illusion …. showed a clear maxi-mum with tone durations around 200 ms;with decreasing tone durations other acousticillusions appear, while with durations greaterthan about 1 s, the presentation can be per-ceived correctly. (p. 128)

Figure 3 plots the percentages of different per-ceptions of the illusion pattern at the differenttone durations (and, so, interonset intervals) inZwicker’s study, and it can be seen that, atinteronset intervals of 2,000 ms, over 80% ofreports were of no illusion.

Study of Chambers et al. (2002)

Chambers et al. (2002) challenged the reportof Deutsch (1974) concerning the octave illu-sion. From their experiments, they concludedthat the perceived pitch difference between thealternating tones generally corresponds more toa semitone than to an octave. They further con-cluded that listeners do not uniformly lateralizeeach tone toward the ear receiving the higher fre-quency. On the basis of these claims, the authorsproposed an alternative explanation of the octaveillusion. This explanation assumes that

1. listeners perceptually fuse the dichoti-cally presented high and low tones, soas to perceive a pitch that correspondsroughly to the fundamental frequency;

OCTAVE ILLUSION REVISITED 5

2. the slight pitch difference that is heardbetween the alternating tones is theresult of diplacusis; and

3. the perceived tones are sometimes later-alized to the ear receiving the higherfrequency and sometimes to the earreceiving the lower frequency.

The present section presents a critique ofthe study by Chambers et al. (2002) and ques-tions the validity of their observations. Giventhat their model could hold only if these obser-vations were valid, their theoretical proposal isalso challenged. The study consisted of fourexperiments, and these are examined below.

Experiment 1

Experiment 1, titled “Subjective Report,”was intended to be preliminary; as Chambers et

al. (2002) wrote, “Subjective reports were pure-ly qualitative and were not statistically ana-lyzed [italics added]” (p.1292). Fifteen subjectsparticipated; 3 of these were the authors, and noinformation was given concerning whether ornot the remaining subjects were naive concern-ing the octave illusion. The subjects first lis-tened extensively to single dichotic tone pairs at400 Hz and 800 Hz, at several tone durations,which ranged from 200 to 800 ms. The dichot-ic 400 – 800 Hz tones were then presented inalternating sequence. The tone durations againranged from 200 to 800 ms. Furthermore, onhalf the trials the tones were separated by paus-es that were equal in duration to the tones them-selves, so the interonset intervals between suc-cessive tones varied from 200 ms to 1,600 ms.

The issue of tone duration. Chambers et al.(2002) stated that the subjects’ judgmentsexhibited no dependence on tone duration noron the presence of regular silent intervals, andthey claimed that this was in accordance withfindings obtained by others. However, theauthors also stated that the subjects’ reportswere not statistically analyzed, and they pre-sented no data from patterns at specific tonedurations; neither did they present any otherevidence to support their assertion that judg-ments were indeed independent of tone dura-tion. Further, the authors were incorrect inasserting that their observations were in accor-dance with previous ones. As outlined above,Deutsch (1981) compared interonset intervalsof 250 ms with those of 3,000 ms and obtaineda highly significant effect of interonset interval(see also Deutsch, 1983a). This study used sub-jects who had been selected for obtaining astrong octave percept in the first place, whichleads to the surmise that unselected subjectsmight show an even stronger effect of interon-set interval. Such a result was obtained in thestudy by Zwicker (1984) referenced above. Asshown in Figure 3, from Zwicker’s data, onewould expect that at interonset intervals of 800ms, no illusion would be perceived roughly30% of the time, and one would expect that atinteronset intervals of 1,600 ms, no illusion

6 DEUTSCH

Figure 3. Percentage occurrence of different percepts ofthe octave illusion, as a function of tone duration andthus of interonset interval. The data were averagedacross 8 subjects. Percept 1: Two tones of identical pitchalternating from ear to ear. Percept 2: A higher tone inone ear alternating with a lower tone in the other ear (i.e.,the octave percept). Percept 3: A higher tone alternatingfrom ear to ear, together with a lower tone alternatingfrom ear to ear (i.e., no illusion). Percept 4. None of theabove. From “Experimente zur dichotischen Oktav-Tauschung,” by T. Zwicker, 1984, Acustica, 55, p.135.Copyright 1984 by S. Hirzel Verlag, Stuttgart, Germany.Adapted with permission.

would be perceived roughly 80% of the time.Yet, these values of interonset interval wereamong those used by Chambers et al. (2002).

The issue of lateralization. In Experiment1, Chambers et al. (2002) used the followingmethod to evaluate how each tone was lateral-ized. The subjects were presented with alternat-ing 400 – 800 Hz sequences at the variousinteronset intervals described above and wereinstructed to indicate their judgments by tap-ping (e.g. “Tap in time with the higher pitch”or “Tap in time with the left tone”). The exper-imenter then evaluated, by visual observationof the subjects’ tappings, to which ear the high-er and the lower tones were lateralized. Theauthors reported, on the basis of these observa-tions, that 9 subjects lateralized the tonestoward the ear receiving the lower frequency,whereas 6 subjects lateralized the tones towardthe ear receiving the higher frequency.

However, the authors presented no actualdata to support this assertion. They did notstate that the experimenter monitored the sig-nals that were presented to the subjects, and itis unclear how, without such monitoring, hecould determine how the subjects’ tappingscorresponded to their perceptions. Evenassuming that the experimenter had monitoredthe sound signals, no evidence was providedthat he could reliably synchronize his visualperceptions of the subjects’ tappings with thesesignals, nor that the subjects were able to syn-chronize their tappings with these signals. Thisproblem of validity is particularly severe forthe fast rates of presentation needed to producethe octave illusion; yet, at slow rates of presen-tation, the illusion becomes degraded and mayeven disappear (Figure 3). It is important tonote that these informal observations concern-ing lateralization in Experiment 1 were notconfirmed elsewhere in the article byChambers et al. (2002). In contrast, a numberof other experiments, which were reviewedabove, have indicated that, on listening to theoctave illusion, there is a strong tendency forsubjects to lateralize each perceived tonetoward the ear receiving the higher frequency

signal (Deutsch, 1978, 1988; Deutsch and Roll,1976; Zwicker, 1984).

The issue of the size of the perceived pitchdifference. To evaluate the size of the per-ceived pitch difference between the alternatingtones, Chambers et al. (2002) presented thesubjects, for comparison, with sequences inwhich tones alternated between 400 Hz and 800Hz (an octave) and between 400 Hz and 424 Hz(a semitone). The authors reported that 2 of thesubjects perceived a pitch difference of anoctave, 8 (2 of whom were authors) perceived apitch difference of a semitone, 4 (1 of whomwas an author) perceived a pitch difference ofbetween an octave and a semitone (this differ-ence not being further specified), and 2 subjectsperceived no pitch difference. However, sever-al points should here be made.

1. The subjects’ judgments were based onsequences in which the interonset inter-vals varied substantially, includingsome in the range where Zwicker(1984) had reported that sometimes noillusion was obtained. Comparison can-not, therefore, be made directly betweenthese results and those of Deutsch(1974).

2. The subjects had earlier been givenextensive experience with dichotic 400– 800 Hz chords and patterns at thesedifferent tone durations, and this priorexperience may have influenced theirjudgments.

3. At least 3 of the subjects (i.e., theauthors) had prior knowledge of theillusion, and this could have influencedtheir judgments.

4. Twenty-five percent of right-handers inthe original large-scale experiment ofDeutsch (1974) reported little or nopitch difference between the alternatingtones (i.e., their percepts fell into thesingle pitch category). In Experiment 1of Chambers et al. (2002), the authorsreported on the responses of only 12

OCTAVE ILLUSION REVISITED 7

subjects (excluding themselves), so thatsampling bias could have contributed totheir results.

However, as the authors pointed out,Deutsch (1974) did not publish the means bywhich the subjects’ reports of the octave illu-sion were obtained (see Footnote 4). This is rec-tified in the new experiment, which is describedlater in this article.

Experiment 2

In Experiment 2, Chambers et al. (2002) didnot investigate the octave illusion itself, butrather a different effect, and they related theirresults to the informal observations reported inExperiment 1. The experiment used 8 subjectswho had all participated in Experiment 1,including 2 authors (the 2 subjects who hadreported hearing an octave difference betweenthe tones in Experiment 1 were not tested). Thesubjects were presented with tones at 400 Hz orat 800 Hz to one ear, and they were asked tomatch the tone that they heard to a tone of vari-able pitch that was presented to the other ear,thereby obtaining a measure of diplacusis. Thesubjects’ matches varied between no pitch dif-ference to a difference of less than half a semi-tone, with the majority of matches being in thelower part of this range. None of the matchesinvolved a pitch difference that approached asemitone.

On the basis these findings, Chambers et al.(2002) claimed that pitch differences obtainedon hearing the octave illusion are a reflection ofdiplacusis. However, the very slight pitch dif-ferences they found in Experiment 2 wereinconsistent with this interpretation, given theirreports in Experiment 1 of an octave differenceby 2 subjects and of differences that weregreater than a semitone but smaller than anoctave by 4 additional subjects. So even settingaside the problems of interpretation outlinedabove, and the previous reports of an octavedifference between the alternating tones byother researchers (Deutsch, 1974; Zwicker,

1984), diplacusis could only account for someof the reports of the octave illusion obtained byChambers et al. in their own Experiment 1.

Experiment 3

In Experiment 3, subjects were presentedwith dichotic sequences of various types, andwere asked on each trial to report which ear wasreceiving the higher pitch. The subjects werenot asked to report their actual percepts but,rather, to infer what signal configurations werebeing presented. Eight subjects were tested.These had all participated in Experiment 1 andso had received extensive experience withoctave illusion patterns of differing durations,some of which may well have been perceivedveridically (Zwicker, 1984). Chambers et al.(2002) wrote, “The most surprising result fromthis experiment was the capacity listenersdemonstrated to correctly segregate the octaveillusion sequence by ear, despite reporting astandard single-image percept” (p. 1297).However, this result is unsurprising given thesubjects’ prior experience with variants of theillusion in Experiment 1, some of which wouldhave enabled them to infer what signals werebeing presented to each ear.

Experiment 4

Seven subjects participated in Experiment4. These had all participated in Experiment 1,though their percepts of the octave illusion inthis experiment were not given. The subjectswere presented with dichotic 400 - 800 Hzsequences, in which tones at 400 Hz, 800 Hz, or2000 Hz were embedded as deviants. Thedeviant tones were presented diotically (i.e.,simultaneously to both ears) but offset in timeso that they were perceptually displaced to theside of the midline. Averaging across subjects,reaction times were shorter for detecting 800-Hz than 400-Hz deviants. Chambers et al.(2002) interpreted this finding as indicatingthat, on hearing the standard octave illusion, lis-teners would perceive both of the alternating

8 DEUTSCH

tones as closer to 400 Hz than to 800 Hz.However, the finding by Deutsch (1980) thatperception of the octave illusion is sensitive tosequential context leads to the possibility thatthe diotically presented tones may have affect-ed perception of the illusion here also.Furthermore, the data were averaged across all7 subjects, so that the inclusion of even 1 or 2subjects whose percepts fell into the singlepitch category would have skewed it in thedirection reported by Chambers et al.

In sum, taking together the four experi-ments by Chambers et al. (2002), can benoted:

1. Chambers et al.’s (2002) conclusionsrelied heavily on the informal observa-tions from Experiment 1, which theauthors themselves stated were “purelyqualitative and were not statisticallyanalyzed [italics added]” (p.1292).These findings were based on percep-tions of sequences of tones with widelydiffering interonset intervals, includingsome in the range where Zwicker(1984) had reported that sometimes noillusion was obtained. Other aspects ofthe procedures used in this study (suchas the tapping procedure for establish-ing lateralization patterns) were prob-lematic, and many of the observationswere at variance with those of Deutsch(1974, 1978, 1980, 1981, 1988),Deutsch and Roll (1976) and Zwicker(1984).

2. Chambers et al.’s (2002) interpretation ofthe pitch differences heard in the octaveillusion in terms of diplacusis cannotexplain the pitch differences of an octavethat were reported by others (Deutsch,1974; Zwicker, 1984) – or even a size-able proportion of the pitch differencesthat were reported by the authors them-selves in their Experiment 1.

However, Chambers et al. (2002) havechallenged the claim made by Deutsch (1974)

that the majority of subjects, on listening tothe octave illusion, perceive a pitch differenceof an octave between the alternating tones.Instead, they have argued that the most com-mon percept of this illusion involves a verysmall pitch difference (i.e., of a semitone orless). Their argument was based on the infor-mal observations of a small number of sub-jects, some of whom were the authors them-selves, and all of whom had experiencedintensive prior training with dichotic chords atdifferent durations before listening to the illu-sion. However, given that the authors haveraised this issue, an experiment was carriedout to examine the size of the perceived pitchdifference between the alternating tones on lis-tening to the octave illusion, using the stimu-lus parameters that had originally been usedby Deutsch (1974). This new experiment useda procedure that enabled more explicit conclu-sions to be drawn than those from earlierexperiments.

Experiment

In this experiment, musically trained sub-jects were asked to listen to the octave illusionpattern and to write down in musical notationwhat they heard. The subjects were furnishedwith the note name of the low tone in the pat-tern (G4) and were told that this was one of thetones they would hear. They then used relativepitch to notate the other tones that they per-ceived. If the octave illusion simply reflecteddiplacusis, one should expect that, on hearingthis pattern, the subjects would notate only asmall pitch difference between the alternatingtones. To control for the possibility that thesubjects might mistakenly attribute an octavedifference between tones of the same pitch thatwere presented to the left and right ears, twofurther patterns were presented. The first con-sisted of the high tone of the octave illusion pat-tern alone (G5) alternating from ear to ear, andthe second consisted of the low tone of theoctave illusion pattern alone (G4) alternatingfrom ear to ear.

OCTAVE ILLUSION REVISITED 9

Method

Subjects. Twelve individuals with normal hearing (5male and 7 female, ages 18 to 32 years) participated inthe experiment. They had all received at least 4 years offormal musical training and could read and write in sim-ple musical notation. On the basis of the handednessquestionnaire of Varney and Benton (1975), they com-prised 8 right-handers (3 with left- or mixed-handed rel-atives), 3 mixed-handers and 1 left-hander. All subjectswere naive concerning the octave illusion.

Apparatus and stimuli. Three sequences of sine-wave tones were created. The first, the octave illusionpattern, was as in the experiment of Deutsch (1974).This constituted a sequence of tones that alternatedbetween 400 Hz and 800 Hz (corresponding to approxi-mately G4 and G5 in the musical scale). The tones wereof equal amplitude, and 250 ms in duration. To minimizetransients, there were no amplitude drops between tones,and the frequency transitions occurred at zero crossing.7

The identical sequence was presented to both ears simul-taneously; however, when the right ear received the hightone, the left ear received the low tone, and vice versa.The second, the alternating high tone pattern, consistedof tones at 800 Hz (corresponding approximately to G5)that were presented in alternation at the two ears. Alltones were at equal amplitude and 250 ms in duration.The third, the alternating low tone pattern, was identicalto the second, except that the tones were at 400 Hz (cor-responding approximately to G4).

Tones were generated on a NeXTStation Turbo(NeXT Computers, Inc., Redwood City, CA) using thecmusic sound synthesis system (F. R. M. Moore, 1982).The signals were transferred to a MacIntosh G4 comput-er, passed through a mixer (Mackie CR1604; LOUDTechnologies, Inc., Woodinville, WA) and then through anamplifier (NAD 304; NAD Electronics, Sharon, MA),and were presented to subjects via headphones (Grason-Stadler TDH-49, calibrated and matched; Grason-Stadler,Inc., Madison WI) at an amplitude of 70-dB SPL. Thesubject was seated in front of a Keystation 61 synthesizerkeyboard (M-Audio, Irwindale, CA) that was interfacedwith the computer, so that the subject was able to play onthe keyboard and hear the output through earphones whileat the same time listening to the test patterns.

Procedure. All subjects were tested individually, andthey listened to the patterns through earphones. Theywere told that on each trial, they would hear a repeatingsequence of tones, and that they should notate both thesequence of pitches they heard and the perceived ear ofinput for each tone. There were given no initial practicesequences.

The subjects first listened to the octave illusion pat-tern for as long as they wished. They were informed thatone of the tones approximated G4 but were given no fur-ther information. They were enabled to confirm their per-ceptions by matching the tones in the pattern with tonesthey played on the synthesizer keyboard. When theywere certain of their judgments, they notated thesequence of pitches they perceived, together with the per-ceived locations of the tones. Following this, the subjectswere asked to place their earphones in reverse position(see Footnote 3), to repeat the procedure, and to notateagain the sequence of pitches they perceived, togetherwith the perceived locations of the tones. Next, the alter-nating high tone pattern was presented, and the same pro-cedure was followed. Then, the alternating low tone pat-tern was presented, and the same procedure was fol-lowed. Finally, the octave illusion pattern was again pre-sented, and the subjects were asked to report the pitchdifferences that they heard.

Results

Figure 4 presents, as examples, the nota-tions of three of the subjects. Subjects SY andRR notated the octave illusion pattern as a tonecorresponding to G5 on the right, alternatingwith a tone corresponding to G4 on the left,with earphones placed both ways. In contrast,Subject JP notated the same pattern as a tonecorresponding to G5 on the left, alternatingwith a tone corresponding to G4 on the right,with earphones placed both ways. When pre-sented with patterns consisting of single tones(i.e., G5 alternating from ear to ear, and G4alternating from ear to ear) all these subjectsnotated the tones correctly.

The data from all 12 subjects are summa-rized in Table 1. On listening to the octave illu-sion pattern, 7 of the 12 subjects notated thestandard octave percept described by Deutsch(1974), that is, a single tone that alternated fromear to ear, that simultaneously alternatedbetween G4 and G5, with earphones placedboth ways. Of these, 4 subjects heard the hightone on the right and the low tone on the left,with earphones placed both ways; 2 heard thehigh tone on the left and the low tone on theright with earphones placed both ways; and 1heard the high tone on the right and the low

10 DEUTSCH

7 The tones were generated in phase at the two channels.

OCTAVE ILLUSION REVISITED 11

Figure 4. Percepts of the different patterns in the experiment, notated by 3 of the subjects (SY, RR, and JP). Thenotations for Track 1 (a) are of the octave illusion pattern, and for Track 1 (b) of the octave illusion pattern with ear-phone positions reversed. The notations for Track 2 are of a single tone at G5 (high tone), which alternated from earto ear, and the notations for Track 3 are of a single tone at G4 (low tone), which alternated from ear to ear.

tone on the left on one presentation, with theopposite lateralization pattern on the other. Onrelistening to the octave illusion pattern at theend of the session, all these subjects againreported hearing tones an octave apart thatalternated from ear to ear. Four additional sub-jects notated complex percepts that changedwith continued listening, all of which involvedG4 and G5, and so involved an octave differ-ence between the tones. (One of these subjectsnotated G4 alternating from ear to ear, inter-spersed with other notations involving both G4and G5.) The percepts of these 4 subjects there-fore fell into the complex category described inDeutsch (1974). The final subject notated thesame pitch (G4) alternating from ear to ear,

with the tone in the right ear having a sharpertimbre. The percept of this subject therefore fellinto the single pitch category described byDeutsch (1974).

On listening to the alternating high tone pat-tern, 11 subjects correctly notated G5 alternat-ing from ear to ear, whereas 1 subject notated asemitone difference between the tones at thetwo ears (i.e., G5 alternating with G#5). On lis-tening to the alternating low tone pattern, 10subjects correctly notated G4 alternating fromear to ear; 1 notated G3 alternating from ear toear; and the subject who had notated G5 alter-nating with G#5 for the alternating high tonepattern notated G3 alternating with G#3 for thealternating low tone pattern. All subjects except

OCTAVE ILLUSION REVISITED 13

1 therefore notated these patterns correctly asconsisting of the same pitch in both ears, andthe one exception notated a semitone differencebetween the tones at the two the ears.8 The nota-tions of an octave difference in listening to theoctave illusion pattern could not, therefore,have been due to the subjects mistakenlyattributing an octave difference between thetones at the two ears.

Discussion

This experiment provided more explicitdocumentation of percepts of the octave illu-sion than have so far been obtained. It should benoted that, because only 12 subjects were test-ed, the experiment did not provide a measure ofthe statistical distribution of the various per-cepts of the octave illusion, such as had beenprovided earlier in the large scale study ofDeutsch (1974). However, the subjects in thepresent experiment were selected at randomwith the only constraints being that they shouldbe able to read and write in musical notation, tohave normal hearing, and to be naive concern-ing the octave illusion. Eleven of the 12 sub-jects, including those who notated complex per-cepts, notated tones at G4 and G5, and so notat-ed them as separated by an octave; and 7 ofthese notated the standard octave percept. Onesubject notated a single pitch alternating fromear to ear, so that her percept fell into the singlepitch category described by Deutsch (1974).

In sum, 7 of the 12 subjects in this experi-ment notated the octave percept of the octaveillusion, and the notations of 4 more subjectsalso included an octave difference between thetones, with only 1 subject notating a singlepitch alternating from ear to ear. The results ofthis experiment were therefore at variance with

the claim made by Chambers et al. (2002) thatthe octave percept of the octave illusion is rare,and that the perceptions of most subjectsinvolve pitch differences so small as to beamenable to an explanation in terms of diplacu-sis. Rather, they supported the finding byDeutsch (1974) that the majority of subjectsperceived an octave difference between thealternating tones - a finding that was also inaccordance with the report of Zwicker (1984).

The notated perceptions of the single tonesof the same pitch alternating from ear to earprovided a control for the possibility that thesubjects, on listening to the octave illusion pat-tern, might have mistakenly attributed anoctave difference between the tones at the twoears. The finding that the subjects notated thesame pitch at the two ears, with the exceptionof 1 subject who notated a semitone differencebetween the alternating tones, is as expectedfrom other findings on diplacusis, in which thesize of this effect was generally found to be asmall fraction of a semitone (see, e.g., van denBrink, 1975a, 1975b).

No correlate was here obtained betweenperception of the illusion and the subjects’handedness. This result was as expected fromthe small number of subjects tested, given thatlarge groups of subjects are generally requiredfor handedness correlates to emerge at the per-ceptual level (Herron, 1980). Zwicker (1984)also obtained no handedness correlates on com-paring the perceptions of 3 right-handers with 3congenital lefthanders.

General Discussion

In this article it has been argued that thestudy of Chambers et al. (2002) used problem-atic procedures, so that their conclusions con-cerning the lateralization of tones in the octaveillusion were called into question, as were theirconclusions concerning the most frequentlyperceived pitch differences between the alter-nating tones. In addition, an experiment wasreported that used a new procedure that provid-

8 The pitch difference perceived by this subject whenthe same tone was presented to the left and right earsmay have reflected diplacusis. The direction of thispitch difference did not correspond to the subject’s pat-terns of localization for the higher and lower tones inthe octave illusion.

ed more explicit documentation concerning thephenomenology of the illusion than in previousstudies. This experiment confirmed the obser-vations in the original study of Deutsch (1974)on the octave illusion, and its findings wereconsistent with the two-channel model of theoctave percept, which invokes a separation ofthe what and where pathways in the auditorysystem. The finding that the subjects notated anoctave difference between the alternating tonescannot be explained on the proposal byChambers et al. (2002) that the illusion resultsfrom diplacusis. In addition, the diplacusishypothesis cannot explain the dependence ofthe illusion on tone duration or on sequentialcontext; neither can it explain the handednesscorrelates with the type of percept obtained,which indicate that the illusion serves as areflection of brain organization, rather thancharacteristics of the auditory periphery.

The two-channel model of the octave illu-sion has as its core the supposition that the deci-sion mechanisms underlying pitch and lateral-ization are, at some point in the auditory sys-tem, distinct and separate. This supposition is inaccordance with recent findings by auditoryneurophysiologists. In particular, Rauschecker,Tian, and colleagues have obtained evidencethat the lateral belt area of the auditory cortexof the rhesus monkey is subdivided into regionsthat are specialized for the processing of eitherwhat or where information: Neurons in theanterior belt are tuned specifically to type ofmonkey call, whereas neurons in the caudal beltare instead tuned to the spatial location of thesignal (Rauschecker & Tian, 2000; Tian, Reser,Durham, Kustove, & Rauschecker, 2001).Furthermore, it appears that information fromthese two regions forms separate streams, thatproject to spatial and nonspatial areas of thefrontal lobe (Ramonski, et al., 1999).

The two-channel model is in accordancewith other perceptual research indicating thatdifferent attributes of sound are processedalong separate pathways that are at some stageindependent, and so can arrive at inconsistentconclusions (Carlyon, Demany, and Deeks,

2001; Darwin and Carlyon, 1995; Gardner,Gaskill, and Darwin, 1989; Hukin and Darwin,1995; B. C. J. Moore, Glasberg, & Peters,1986). Furthermore, Odenthal (1963); Efronand Yund (1974); Hall, Pastore, Acker, andHuang (2002); Thompson (1994); and Deutsch(1975b) have shown that illusory conjunctionsof different attribute values can occur withother sound configurations also.

The question then arises as to why peopleshould have evolved the two decision mecha-nisms that are hypothesized to produce theoctave illusion. This question can be addressedfor the ear dominance and lateralization com-ponents separately. Note that the ear dominancecomponent has two characteristics: First, itbecomes weaker as the duration between onsetsof successive tones is increased; second, itoccurs in configurations in which the two earsreceive the same frequencies in succession, andis weaker or absent when this condition doesnot hold. Given these characteristics, it wasconjectured (Deutsch, 1981) that this effectreflects the operation of a mechanism that nor-mally helps to counteract misleading effects ofechoes and reverberation. In normal listening,when the same frequency emanates successive-ly from two different regions of space, the sec-ond occurrence may be an echo. This interpre-tation becomes less probable as the delaybetween these two occurrences is lengthened,and it becomes less probable when other fre-quencies intervene between such two occur-rences. On this line of reasoning, the octaveillusion falls into the class of phenomena, ofwhich the precedence effect is another example(Haas, 1951; Wallach, Newman, Rosenzweig,1949), which reflect the activity of mechanismsthat have evolved to counteract unwantedeffects due to the acoustics of the environment.

Concerning lateralization to the higher fre-quency signal, it was conjectured (Deutsch,1981) that this might reflect the action of amechanism designed to handle head shadoweffects. When a complex tone is presented innatural situations, the relative amplitudes of thepartials arriving at the two ears may differ con-

14 DEUTSCH

OCTAVE ILLUSION REVISITED 15

9. An anonymous reviewer raised the question of whatwould be expected on the two-channel model when thesuppression component is weak or absent. In the case ofshort tones separated by pauses, it is expected that the nor-mal process of harmonic fusion would take over and thatthe listener would perceive a pitch that corresponds to thefundamental. This could also explain the single pitch per-cepts obtained by some listeners. However, when thetones themselves are of long duration, the process offusion also breaks down, and both the high and the lowtones may be perceived and lateralized correctly. A partialbreakdown of both suppression and fusion might also beresponsible for the complex percepts obtained by somelisteners. In addition, depending on sequential context,both tones may be perceived, but they may be incorrectlylocalize4d, as in the scale illusion reported by Deutsch(1975b).

siderably, owing to the filtering action of thehead. For example, when a complex tone is pre-sented to the listener’s right, then the higher fre-quency components at the left ear are attenuat-ed relative to the lower frequency components.Assuming that the auditory system interpretsthe pattern that produces the octave illusion asthe first and second harmonic of a complextone, then it would make sense to interpret thesignal as coming from the ear receiving thehigher frequency – in this case, from the listen-er’s right9.

Finally, it should be emphasized that, inorder to evaluate the octave illusion, there is nosubstitute for listening to it as it was originallygenerated. Furthermore, given the large differ-ences between listeners in the way the illusionis perceived, it is important to have others listento it also. The illusion occurs as a sound demon-stration in a number of publications, includingDeutsch (1983a, 1995); Houtsma, Rossing, andWagenaars (1987); Kubovy and Pomerantz(1981); and Pierce (1983).

References

Carlyon, R. P., Demany, L., & Deeks, J. (2001).Temporal pitch perception and the binaural system.Journal of the Acoustical Society of America, 109,686-700.

Chambers, C. D., Mattingley, J. B., & Moss, S. A. (2002).The octave illusion revisited: Suppresson or fusion

between ears? Journal of Experimental Psychology:Human Perception and Performance, 28, 1288-1302.

Darwin, C. J., & Carlyon, R. P. (1995). Auditory group-ing. In B. C. J. Moore (Ed.) Hearing (pp. 387-424).San Diego, CA: Academic Press.

Deutsch, D. (1974, September 27). An auditory illusion.Nature, 251, 307-308

Deutsch, D. (1975a, October). Musical illusions.Scientific American, 233, 92-104.

Deutsch, D. (1975b). Two-channel listening to musicalscales. Journal of the Acoustical Society of America,57, 1156-1160.

Deutsch, D. (1978). Lateralization by frequency forrepeating sequences of dichotic 400 Hz and 800 Hztones. Journal of the Acoustical Society of America,63, 184-186.

Deutsch, D. (1980) Ear dominance and sequential inter-actions. Journal of the Acoustical Society ofAmerica, 67, 220-228.

Deutsch, D. (1981). The octave illusion and auditory per-ceptual integration. In J. V. Tobias & E. D. Schubert(Eds.) Hearing research and theory (Vol 1, pp. 99-142). New York: Academic Press.

Deutsch, D. (1983a). Auditory illusions, handedness, andthe spatial environment. Journal of the AudioEngineering Society, 31, 607-618.

Deutsch, D. (1983b). The octave illusion in relation tohandedness and familial handedness background.Neuropsychologia, 21, 289-293.

Deutsch, D. (1988). Lateralization and sequential rela-tionships in the octave illusion. Journal of theAcoustical Society of America, 83, 365-368.

Deutsch, D. (1995). Musical illusions and paradoxes[CD]. La Jolla, CA: Philomel Records.

Deutsch, D. & Roll, P. L. (1976). Separate ‘what’ and‘where’ decision mechanisms in processing adichotic tonal sequence. Journal of ExperimentalPsychology: Human Perception and Performance,2, 23-29.

Efron, R., & Yund, E. W. (1974). Dichotic competition ofsimultaneous tone bursts of different frequency: I.Dissociation of pitch from lateralization and loud-ness. Neuropsychologia, 12, 249-256.

Gardner, R. B., Gaskill, S. A. & Darwin, C. J. (1989).Perceptual grouping of formants with static anddynamic differences in fundamental frequency.Journal of the Acoustical Society of America, 85,1329-1337.

Haas, H. (1951). Uber den Einfluss eines Einfachechosauf die Horsamkeit von Sprache [Concerning theinfluence of simple echoes on the intelligibility ofspeech]. Acustica, 1, 49-52.

Hall, M. D., Pastore, R. E., Acker, B. E., & Huang, W.(2000). Evidence for auditory feature integration

with spatially distributed items. Perception &Psychophysics, 62, 1243-1257.

Herron, J. (Ed.). (1980). Neuropsychology of lefthanded-ness, New York: Academic Press.

Houtsma, A. J. M., & Goldstein, J. M. (1972). The cen-tral origin of the pitch of complex tones: Evidencefrom musical interval recognition. Journal of theAcoustical Society of America, 51, 520-529.

Houtsma, A. J. M., Rossing, T. D., and Wagenaars, W. M.(Eds.). (1987). Auditory demonstrations. New York:Acoustical Society of America; Eindhoven, theNetherlands: Institute for Perception Research.

Hukin, R. W. & Darwin, C. J. (1995). Comparison of theeffect of onset asynchrony on auditory grouping inpitch matching and vowel identification. Perceptionand Psychophysics, 57, 191-196.

Kubovy, M. & Pomerantz, J. R. (Eds.). (1981).Perceptual Organization. Hillsdale, NJ: Erlbaum.

Moore, B. C. J., Glasberg, B. R., & Peters, R. W. (1986).Thresholds for hearing mistuned partials as separatetones in harmonic complexes. Journal of theAcoustical Society of America, 80, 479-483.

Moore, F. R. M. (1982). The computer audio researchlaboratory at UCSD. Computer Music Journal, 6,18-19.

Odenthal, D. W. (1963). Perception and neural represen-tation of simultaneous dichotic pure tone stimuli.Acta Physiological Pharmacologica Neerlanda, 12,453-456

Pierce, J. R. (1983) The science of musical sound. NewYork: Freeman.

Rauschschecker, J. P., & Tian, B. (2000). Mechanismsand streams for processing of “what” and “where” inthe auditory cortex. Proceedings of the NationalAcademy of Sciences, 97, 1180-1186.

Romanski, L. M., Tian, B., Fritz, J., Mishkin, M.,Goldman-Rakic, P.S., & Rauschecker, J. P. (1999).Nature Neuroscience, 2, 1131-1136.

Thompson, R. F. (1994). Sensitivity to combinations ofmusical parameters: Pitch with duration, and pitchpattern with durational pattern. Perception andPsychophysics, 56, 363-374.

Tian, B., Reser, D. Durham, A., Kustove, A., &Rauschecker, J. P. (2001, April 13). Functional spe-cialization in rhesus monkey auditory cortex.Science, 292, 290-293.

van den Brink, G. (1975a). Monaural frequency-pitchrelations as the origin of binaural diplacusis for puretones and residue sounds. Acustica, 32, 166-174.

van den Brink, G. (1975b). The relation between binaur-al diplacusis for pure tones and for complex soundsunder normal conditions and with induced monauralpitch shift. Acustica, 32, 159-165.

Varney, N. R., & Benton, A. L. (1975). Tactile perception ofdirection in relation to handedness and familial handed-ness background. Neuropsychologia, 13, 449-454.

Wallach, H., Newman, E. B., & Rosenzweig, M. R.(1949). The precedence effect in sound localization.American Journal of Psychology, 62, 315-336.

Zwicker, T. (1984). Experimente zur dichotischen Oktav-Tauschung [Experiments on the dichotic octave illu-sion]. Acustica, 55, 128-136.

16 DEUTSCH