AUDIO/MARCH 198736

Early on around this magazine, there was a sort of Occam’ sAudio Razor: “If a piece of gear measures well but soundsbad, it is bad, but if it sounds good and measures poorly, it’s agood piece of gear.” The idea behind the quasi-motto was tofree one’s ears, and perception, from the tyranny of meters.Over the last four decades, lots of midnight oil has beenburned trying to make measurements mean something, whichusually wound up as an attempt to make gear which alreadysounded good also measure well. On the hi-fi end of things,at least, not much ef fort has been spent on what it means to“sound good,” or “hear well,” or simply “hear .”

How we hear is an endlessly fascinating subject for somefew audiophiles, most of whom know how easily any of thesenses—and hearing is no exception—can be fooled. Indeed,stereo sound is an illusion. If, however, you have ever listenedto a discussion of how one hears at a hi-fi store, audio club oreven a learned society convention, you have already foundout how few people are truly knowledgeable in this area.

In an ef fort to free us from the tyranny that the ear is an infallible judge of sound, I am proud and pleased to present an article, with an illustrative Eva-Tone SoundSheet, by one ofthe few true authorities in this field, Dr . Diana Deutsch, onwhat a curious thing it is, our sense of hearing. — E.P.

In increasingly large numbers, peo-ple are choosing to listen to musicthrough stereo headphones. This

development has occurred despite thefact that most recordings are notdesigned for headphone listening, butrather to be played through loudspeak-ers. It is a happy coincidence thatstereo recordings sound acceptableeither way. Yet the creative opportuni-ties provided by headphone listeninghave only just begun to be explored.

One highly successful use of head-phones involves binaural recording.Two microphones are placed at the

ears of a dummy, and two very similarrecordings are produced from these,differing only as would the sound sig-nals arriving at the ears of a listenersituated in the same position. Whenthese recordings are played backthrough stereo headphones, remark-able realism is obtained.

There is, however , another use ofstereo headphones which takes us inthe direction opposite that of increasedrealism, to an unexpected and para-doxical world of illusion. Rather thanpresenting highly similar signals to thetwo ears, entirely dif ferent signals are

Illusions For StereoHeadphones

DR. DIANA DEUTSCH

AUDIO/MARCH 1987 37

presented. Ef fects obtained with thistechnique are not only startling to expe-rience, but also demonstrate certainproperties of the hearing mechanismwhich might otherwise have passedunrecognized.

Let us begin with a very simplesound pattern, which is illustrated inFig. 1. A 400-Hz sine-wave tone isdelivered to one ear , and at the sametime an 800-Hz sine-wave tone is deliv-ered at equal amplitude to the otherear. When this combination lasts for

Dr. Diana Deutsch has been a member of the research faculty of the University ofCalifornia, San Diego, since 1970, the year she was awarded a Ph.D in psycholo-gy from that institution. She is the founding editor of Music Perception, a journalpublished by the University of California Press; coauthor (with J. A. Deutsch) ofPhysiological Psychology (Dorsey Press, 1966; Second Edition, 1973), and editorof The Psychology of Music (Academic Press, 1982). In addition, she is an activemember of the Acoustical Society of America and has served on the AdvisoryCouncil of the International Association for the Study of Attention and Performance.She is a Fellow of the American Association for the Advancement of Science andof the Society of Experimental Psychologists. Recently made a Fellow of the AudioEngineering Society, she was guest editor for a special issue, entitled "AuditoryIllusions and Audio," of the Society’s journal (Vol. 31, No. 9, Sept. 1983).

Illus

tratio

n: P

hilip

And

erso

n

AUDIO/MARCH 198738

several seconds, most people hearboth the high tone and the low one, andcan localize them correctly.

Now let us consider a variant of thispattern which I devised. (See Deutsch,D., “An Auditory Illusion,” Nature, Vol.251, 1974, pgs. 307-309.) For the first250 mS, the 800-Hz signal is presentedto the right ear and the 400-Hz signal tothe left. The tones then interchangepositions, so that for the next 250 mSthe 400-Hz signal is presented to theright ear and the 800-Hz signal to theleft. The tones then switch back to theiroriginal positions, and the procedure isrepeated. So, as illustrated in Fig. 2A,each ear receives a pattern that con-sists of two tones presented in alterna-tion. Yet when the right ear receivesthe high tone, the left ear receives thelow tone, and vice versa. This patternis given in Sound Example 1. (Be sure,when listening to this and the other

examples, that the loudspeakers onyour system are turned off, and that thechannels are carefully balanced forloudness.)

Surprisingly, this simple pattern isalmost never heard correctly , andinstead gives rise to a number of illu-sions. Most people obtain a perceptsuch as illustrated in Fig. 2B. This con-sists of a single tone which switchesfrom ear to ear; as it switches, its pitchsimultaneously shifts back and forthbetween high and low. In other words,the listener hears a single high tone inone ear which alternates with a singlelow tone in the other ear.

Clearly, there can be no simple basisfor this illusion. We can explain the per-ception of alternating pitches by sup-posing that the listener hears the tonespresented to one ear and ignores theothers. But then we cannot explain whythese tones should appear to be switch-ing between ears. Alternatively, we canexplain the perception of a single tonewhich alternates from ear to ear by sup-posing that the listener is constantlyshifting his attention between left andright. But then the pitches of the tonesshouldn’t change with changes in theirperceived locations. The illusion of asingle tone that alternates simultane-ously both in pitch and in location pres-ents us with a paradox.

The ef fect becomes even strangerwhen we consider what happens whenthe earphone positions are reversed.The ear that had been hearing the hightone continues to hear the high tone,and the ear that had been hearing thelow tone continues to hear the low tone!This creates the peculiar impressionthat the high tone has migrated fromone earphone to the other, and that thelow tone has also migrated in analo-gous fashion. The best way to experi-ence this ef fect is to switch the ear-phones around several times while thepattern is playing, and ask yourselfeach time which ear is hearing the hightone. Most people find that the hightone appears to stay in one ear and thelow tone in the other ear , regardless ofhow the earphones are positioned.

Another interesting thing to try at thispoint is to begin by listening to the illu-sion in stereo, and then change the set-ting to mono, so that both ears nowreceive both channels. At this pointyour percept should change dramatical-

ly: You should hear a single complextone coming simultaneously from bothearphones, together with clicks occur-ring four times per second. (The clicksare due to the transients produced byswitching the signals between 400 and800 Hz). Then change the settingback to stereo, and the illusion shouldreappear. Sound Example 2 presentsthe pattern in stereo, then in mono,and then in stereo again.

How can we account for this illu-sion? Clearly , there is no simpleexplanation. But if we assume thatseparate brain mechanisms exist fordeciding what sound we hear and fordeciding where the sound is comingfrom, we are in a position to advancean explanation. The model is illustrat-ed in Figs. 3 and 4. To obtain the per-ceived pitches, the frequencies arriv-ing at one ear are attended to, andthose arriving at the other ear are sup-pressed. However, each tone is local-ized at the ear receiving the higherfrequency signal, regardless ofwhether a pitch corresponding to thehigher or the lower frequency is in factperceived.

Figure 3 illustrates the model forthe case of a listener who perceivedthe pitches corresponding to the fre-quencies delivered to his right ear .When a high tone is delivered to hisright and a low tone to his left, hehears a high tone, because this isdelivered to his right ear . He alsolocalizes the tone in his right ear ,because this ear is receiving the high-er frequency. But when a low tone isdelivered to the right ear and a hightone to the left, he now hears a lowtone, because this is delivered to hisright ear, but he localizes the tone inhis left ear instead, because the leftear is receiving the higher frequency .So he hears the entire sequence as ahigh tone to the right which alternateswith a low tone to the left. You cansee that reversing the earphone posi-tions wouldn’t change this basic per-cept (the sequence would simplyappear to be of fset by one tone).However, Fig. 4 illustrates the samemodel for the listener who perceivesthe pitches corresponding to the fre-quencies d elivered to h is l eft e arinstead, using the same localizationrule. You can see that the identicalpattern is now heard instead as a high

Mus

ic C

opyi

st: B

ert K

osow

Fig. 1—400-Hz sine-wave tone delivered to left ear and 800-Hz toneto right ear.

Fig. 2—Octave illusion pattern, with400- and 800-Hz tones first deliveredto left and right ears, respectively,and then interchanging positionsevery 250 mS (A); and the most common percept resulting from thatpattern (B).

AUDIO/MARCH 1987 39

placed both ways. But left-handers did-n’t show this tendency.

In a more extensive study , I dividedthe population of listeners into threegroups on the basis of handedness,using the V arney and Benton handed-ness questionnaire shown in Fig. 7.People scoring at least nine out of 10“rights” were designated right-handers,those scoring at least nine out of 10“lefts” w ere d esignated l efthanders,and those with eight or fewer “lefts” or“rights” were designated mixed-han-ders. Each group was then furtherdivided into two on the basis of whether

or not the listener had a left- or mixed-handed parent or sibling.

This six-way division was found tocorrelate with how the octave illusionwas perceived. Right-handers weremore likely to hear the high tone on theright than were mixed-handers, andmixed-handers were more likely to do so than were left-handers. And for allthree handedness groups, those with-out left- or mixed-handed parents orsiblings were more likely to hear thehigh tone on the right than were thosewith left- or mixed-handed parents orsiblings. (See Deutsch, D., “The

tone to the left alternating with a lowtone to the right.

In order to test this hypothesis, Idevised a new pattern, illustrated inFigs. 5 and 6. You can see that one earreceives three high tones followed bytwo low tones, while simultaneously theother ear receives three low tones fol-lowed by two high tones. This basicpattern is repeatedly presented, with-out pause. It was found that, indeed,most people perceived a pattern ofpitches corresponding to the frequen-cies presented either to the right or tothe left. That is, they heard a repeatingpattern that consisted either of threehigh tones followed by two low tones,or of three low tones followed by twohigh tones. Also in confirmation of themodel, each tone was localized in theear receiving the higher frequency ,regardless of whether a pitch corre-sponding to the higher or lower fre-quency was in fact perceived.

So when a low tone was heard, itappeared to be coming not from theearphone which was in fact deliveringit, but from the opposite earphone.When a listener who heard the pitchesdelivered to his right ear was present-ed with channel A to his right and chan-nel B to his left, as shown in Fig. 5, heheard three high tones to his right alter-nating with two low tones to his left.When the earphone positions werereversed, as shown in Fig. 6, this listen-er now heard two high tones to his rightalternating with three low tones to hisleft! So the procedure of reversing theearphone positions appeared to causethe channel to the right to mysteriouslydrop a high tone and the channel to theleft to mysteriously add a low tone!(See Deutsch, D. and P . L. Roll,“Separate ‘What’ and ‘Where’ DecisionMechanisms in Processing a DichoticTonal Sequence,” Journal ofExperimental Psychology: HumanPerception and Performance , V ol. 2,1976, pgs. 23-29.

There is yet another surprisingaspect to this illusion: Right-handersand left-handers dif fer statistically interms of where the high and the lowtones appear to be localized. In onestudy, I had people listen to this patternwith earphones positioned first one wayand then the other. Most right-handersheard the high tone on the right and thelow tone on the left, with earphones

Right-handers and left-handers differ statistically in terms of where high and low tones appear to be localized.

Fig. 4—Same as Fig. 3 but for “left-eared” listener .

Fig. 3—Perceived pitch and localization for “right-eared” listener .

AUDIO/MARCH 198740

Octave Illusion in Relation to Handed-ness and Familial Handedness Back-ground,” Neuropsychologia, Vol. 21,1983, pgs. 289-293.)

How do these findings relate to theorganization of the brain in relation tohandedness? In the large majority ofright-handers, the left hemisphere of

the brain is dominant (i.e., speech isprocessed primarily in this hemi-sphere). But this is true of only abouttwo-thirds of left-handers, the remain-ing one-third being right-hemispheredominant. W e also know that peoplewith left- or mixed-handers in theirimmediate family are less likely to have

a pattern of dominance typical of right-handers than those with only right-handers in their family. So this patternof results indicates that we tend tolocalize the tones in this illusion inaccordance with our patterns of hemi-spheric dominance.

Now, the perception of a single hightone in one ear which alternates with asingle low tone in the other ear is mostcommonly obtained. But some peopleexperience quite dif ferent illusions.Some hear a single tone which switch-es from ear to ear , and whose pitcheither remains the same or changesonly slightly as the tone appears toshift in location. Other people obtain anumber of dif ferent complex percepts,two of which are illustrated in Fig. 8.For example, one person might hear alow tone which alternates from ear toear and whose pitch shifts back andforth by a semitone, together with anintermittent high tone in one ear .Another person might hear a high tonealternating from ear to ear , with anintermittent low tone in one ear . Yetother people find that the pitches of thetones appear to change with continuedlistening. Large dif ferences in timbreor sound quality are sometimesdescribed; for example, the high tonesmay have a flute-like quality and thelow tones a gong-like quality.

Complex percepts of the illusion aretypically unstable, so a person maypass from one to another within a fewseconds and describe the pattern asconstantly changing its character . Aconsiderably higher proportion of left-handers obtain complex percepts thando right-handers. This second hand-edness correlate is probably based onanother relationship between handed-ness and brain organization. It con-cerns the degree to which one hemi-sphere of the brain is dominant overthe other. In right-handers, there tendsto be a pronounced dominance of theleft hemisphere, but in the left-han-ders, patterns of dominance tend to beless pronounced.

The illusion is sometimes perceivedin a way that is analogous to the per-ception of ambiguous figures in vision.As illustrated in Fig. 9, the high tonemay first be heard on the right and thelow tone on the left. Then after a fewseconds, the high tone will switch tothe left and the low tone to the right.

The octave illusion may be heard in analogous fashion to the way that we see ambiguous figures.

Fig. 6—Same as Fig. 5 but with earphone positions reversed.

Fig. 5—Three high tones followed by two low tones delivered to right ear , simul-taneous with three low tones followed by two high tones delivered to left ear .

Fig. 9—Possible instability of percept obtained from octave illusion pattern.

AUDIO/MARCH 1987 41

After a few more seconds, the toneswill interchange positions again, and soon. In a similar way, if we scrutinize theNecker cube of Fig. 10, it will appear toswitch back and forth in orientation, sothat the front face periodically changesplaces with the back one.

If you consistently hear the hightone on the right and the low tone onthe left when the stereo channels are inbalance, you might find that you canachieve a “Necker cube” perceptinstead by gradually altering balanceso as to increase the amplitude of thesignal to the left ear relative to the right.At some stage, the high tone will sud-denly appear to switch to the left andthe low tone to the right. Havingreached this stage, shift the balanceback a little so as to reduce the ampli-tude of the signal to the left ear , untilthe tones appear to return to their orig-inal locations. By shifting the balanceback and forth in this way, you may finda point of equilibrium at which the toneswill appear to spontaneously inter-change locations in space.

Playing w ith t he o ctave i llusion i nthis f ashion i s r ather l ike s crutinizingsome of Escher ’s woodcuts. Take, forexample his “Regular Division of thePlane III,” shown in Fig. 1 1. In theuppermost portion of this picture, theblack horsemen clearly provide the fig-ure and the white horsemen theground. In the lowermost portion, thissituation is reversed. But in the mid-dle, there is a region of ambiguity inwhich your perception alternatesbetween these two interpretations.

What happens when the pattern isplayed at dif ferent speeds? SoundExample 3 presents the pattern firstat the original tempo of four tones persecond. Then the tempo is graduallyincreased to 20 tones per second,and finally it is slowed down to onetone every four seconds. You canhear the illusion sharpen as thetempo is increased, and graduallydeteriorate as the tones are playedmore slowly . At the slowest tempo,both of the simultaneously soundedtones may be heard.

We may next ask what happenswhen the alternating tones are not inoctave relation. Sound Example 4presents the pattern with tones relatedby a minor third. You can hear that theimpression is quite dif ferent, though anillusion is still produced.

What happens when the sounds arepresented through loudspeakers ratherthan earphones? One experiment toinvestigate this question was performedin an anechoic chamber . The listenerwas first positioned so that one speakerwas exactly on his right and the otherexactly on his left, as shown in Fig. 12.

When the octave illusion was played, ahigh tone appeared to be coming fromthe speaker on the right, and itappeared to alternate with a low tonecoming from the speaker on the left. Asthe listener turned slowly, the high toneremained on his right and the low toneon his left. When, however , the listen-

Fig. 8—Two alternative percepts obtained from octave illusion pattern.

Fig. 7—Varney and Benton handedness questionnaire.

Fig. 11—Escher woodcut, “Regular Division of the Plane III,” illustrating ambiguityof visual percept.

er came to face one speaker , with theother exactly behind him, the illusionabruptly disappeared; a single com-plex tone was heard instead, asthough coming simultaneously fromboth speakers. But as he continued toturn, the illusion abruptly reappeared,with the high tone still on his right andthe low tone on his left. In otherwords, after he had turned 180°, itappeared as though the speaker thathad been producing the high tone wasnow producing the low tone, and thatthe speaker that had been producingthe low tone was now producing thehigh tone!

The ef fect also works in certainnon-anechoic environments, thoughthe acoustics of normal rooms candegrade the illusion considerably .The following demonstration is, how-ever, generally very successful:Begin by listening to the pattern withearphones in their usual position.Then, while the pattern is playing,slowly remove the earphones andbring them out in front of you, as illus-trated in Fig. 13. If you obtain a clearand consistent illusion in the firstplace, you will probably find that youcan bring the earphones out a con-siderable distance before the ef fectdisappears. There is another point ofinterest here. Once the illusion islost, it is necessary to return the ear-phones considerably closer (if notright back onto your ears) before it isrecaptured.

What happens if, instead of twoalternating tones, we present a moreelaborate pattern? To examine thisquestion, I devised the pattern shownin Fig. 14A and given in SoundExample 5. You can see that this con-sists of a major scale whose succes-sive tones alternate from ear to ear .The scale is played simultaneously inboth ascending and descending form;when a tone from the ascending scaleis in one ear, a tone from the descend-ing scale is in the other ear . Figures14B and 14C show the ascending anddescending components separately ,and you can see that the patternshown in Fig. 14A is produced by thesuperposition of the patterns shown inFigs. 14B and 14C. This sequence isplayed repeatedly without pause.(See Deutsch, D., “T wo-ChannelListening to Musical Scales,” Journal

AUDIO/MARCH 198742

Fig. 10—Necker cube, illustrating instability of visual percept.

©M

. C. E

sche

r Hei

rs, c

/o C

ordo

n A

rt, B

aarn

, The

Net

herla

nds

AUDIO/MARCH 1987 43

of Acoustical Society of America, V ol.57, 1975, pgs. 1156-1160.)

This scale pattern also produces anumber of dif ferent illusions. The onemost commonly experienced is illustrat-ed in Fig. 14D. A perceptual reorgani-zation occurs such that a melody corre-sponding to the higher tones appears tobe coming from one earphone, and amelody corresponding to the lowertones appears to come from the other .When the earphone positions arereversed, the higher and lower tonesusually maintain their apparent loca-tions. So again, the procedure ofreversing the earphone positionsappears to cause the higher tones tomigrate from one earphone to the other,and the lower tones to migrate in anal-ogous fashion.

The ways in which the higher andlower tones are heard again correlatewith handedness. Right-handers tendto hear the higher tones on the rightand the lower tones on the left, but left-handers don’t show this tendency .Some peo ple hear on ly the highertones, and little or nothing of the lowertones. Interestingly, among those whohear only the higher tones, a larger

number are able to localize them cor-rectly.

The scale illusion often works wellwith sounds presented through stereo-phonically separated loudspeakers innormal room environments. You maywant to listen to Sound Example 5 thisway, making sure you are situatedroughly equidistant from the two loud-speakers. Whether or not the spatialeffect works convincingly in your envi -ronment, you should certainly experi-ence a perceptual reorganization of themelodic lines, such that when the chan-nels are played together in stereo, themelodies that you hear are quite dif fer-ent from those that you hear when eachchannel is played separately.

Variants of the scale illusion can eas-ily be produced. For instance, SoundExample 6 presents a two-octave majorscale pattern, switching from ear to ear(or from loudspeaker to loudspeaker) inthe same way as before. This pattern isillustrated in Fig. 15. When the twochannels are played together in stereo,most people hear a higher scale whichmoves down an octave and back, andthey simultaneously hear a lower scale,which moves up an octave and back,with the two meeting in the middle. Butwhen you play each channel separately,the tones are instead heard to be jump-ing around over a large pitch range.Sound Example 7 presents another vari-ation, a one-octave chromatic scalewhich alternates from ear to ear in thesame fashion, as shown in Fig. 16. Asyet another variant, Sound Example 8presents a two-octave chromatic scalewhich alternates in the same fashion.This example is illustrated in Fig. 17.For all these variants (as well as for theoriginal illusion), it is interesting to listento each channel separately, and then togradually equalize the balance of thechannels and experience the two melod-ic patterns transforming into dif ferentones.

Similar effects can even occur in lis-tening to live music. Figure 18 shows apassage from the last movement ofTchaikovsky’s Sixth Symphony. As youcan see, the theme is formed of noteswhich alternate between the first andsecond violins, while the second voicealternates in converse fashion. A similararrangement holds for the viola andcello parts. However , the voices aregenerally heard instead as illustrated onthe right side of Fig. 18. It remains a

Reversing the earphone positions does not usually reversethe apparent left/right location of tones.

Fig. 14—Scale illusion using one-octave major scale. Sound patterndelivered to right and left ears (A),based on ascending and descendingscales (B and C), produces an illusorypercept (D).

Illus

tratio

n: P

hilip

And

erso

n

Fig. 12—Octave illusion using loud-speakers; with speakers exactly to leftand right of listener (A), with speakersexactly in front of and behind listener(B), and after listener has turned 180 °(C).

Fig. 13—Octave illusion can be sustained even with earphones infront of listener.

AUDIO/MARCH 198744

mystery whether Tchaikovsky intend-ed to create an illusion here, orwhether he expected listeners to hearthis passage as in the written score.

Why should we experience this illu-sion? Because of the complexity ofour sound environment, we cannotrely on classical localization cuesalone (such as dif ferences in ampli-tudes and arrival times at each ear) todetermine the locations of simultane-ously presented sounds. Therefore,other cues must also be taken intoconsideration. One such cue is simi-larity of frequency spectrum: Similarsounds are likely to be coming fromthe same source, and dif ferentsounds from dif ferent sources. Sowith patterns such as we have beenconsidering, it makes sense to con-clude that tones in one frequencyrange are coming from one source,and that tones in another frequencyrange are coming from a dif ferentsource. W e therefore perceptuallyreorganize the tones on the basis ofthis interpretation.

There is an interesting visual ana-log of this effect. In Fig. 19, we see aphotograph of a hollow mask, takenfrom the inside. Although the featuresof the face, such as the nose, are pro-jecting inward, away from us, we per-ceive the face as projecting outward,towards us. Our expectations thatfaces should project outward are sostrong that we perceive this picturequite incorrectly. Further, we continueto do so despite our conscious knowl-edge of the illusion.

So far, we have been consideringcases where the sounds presentedthrough the two earphones (or loud-speakers) are simultaneous. Whathappens when time dif ferences areintroduced? In one experiment, Idevised two simple melodic patternsand asked listeners to identify on eachtrial which one they had heard. Thepatterns are shown in Figure 20.

In one condition, the tones com-prising the patterns were presented tothe two ears simultaneously , asshown in Fig. 21A. Under these cir-cumstances the patterns were easy toidentify, and performance on the taskwas very good. In a second condition,the tones were switched haphazardlybetween the ears, as shown in Fig.21B. As can be heard in Sound

Despite our conscious knowledge of an illusion, we mayoften continue to perceive what we hear incorrectly .

Fig. 16—Same as Fig 15 but using one-octave chromatic scale.

Fig. 15—Scale illusion using two-octave major scale.

AUDIO/MARCH 1987 45

Example 9, the switching proceduremade the task much more dif ficult.Most people found that their attentionwas directed to the sounds coming fromone earphone or the other , and it wasvery dif ficult for them to integrate thetwo into a coherent melody.

A third condition (Fig. 21C) wasexactly as the second, except that themelody was accompanied by a drone.Whenever a tone from the melody wasin the right ear, the drone was in the leftear, and whenever a tone from themelody was in the left ear , the dronewas in the right ear . So sounds wereagain presented to both ears simultane-ously, even though the melody was stillswitching from ear to ear , exactly asbefore. As can be heard in SoundExample 10, the presence of the dronein the opposite ear caused the soundsto merge perceptually , so that themelody could easily be identified.Performance in this condition was againvery good. In a fourth condition, shownin Fig. 21D, a drone again accompaniedthe melody, but it was presented to thesame ear as the melody component.This meant that input was again to oneear at a time. As you can hear in SoundExample 11, it was again very difficult tointegrate the dif ferent sounds. (SeeDeutsch, D., “Binaural Integration ofMelodic Patterns,” Perception andPsychophysics, Vol. 25, 1979, pgs. 399-405.)

This experiment shows that whensignals are coming from two dif ferentlocations, temporal relationshipsbetween them are important determi-nants of how they are perceptuallygrouped together . When both earsreceive input simultaneously , integra-tion of patterns is easy . But whensounds arriving at the two ears areclearly separated in time, we insteadfocus attention on one ear or the other ,and find it much more dif ficult to com-bine the two into a single perceptualstream.

What happens in the intermediatecase, where the signals to the two earsare not strictly synchronous, but insteadoverlap in time? In a further experi-ment, I found that this intermediate caseproduced intermediate results.Identification of the melody with a strict-ly synchronous drone in the oppositeear was easiest. Next easiest identifi-cation of the melody was with an asyn-

Fig. 17—Same as Fig. 15 but using two-octave chromatic scale.

Fig. 18—Passage from last movement of Tchaikovsky’s Sixth Symphony, showingseparate parts for first and second violin, viola, and cello (left), and how theseparts are usually perceived (right).

Fig. 19—A visual example of perceptu-al rearrangement: Hollow maskappears to project outward.

Pho

togr

aph:

Ben

Ros

e

chronous drone, while the worst resultswere with no drone.

Why should the perceptual systemfunction in this fashion? Temporal rela-tionships between sound signals pro-vide important cues as to whether theyare coming from the same source orfrom dif ferent sources. So we shouldexpect that the more clearly signals atthe two ears are temporally separated,the more we should treat them as com-ing from separate sources, and so themore we should tend to group them byspatial location. If such grouping werestrong enough, it should prevent usfrom linking together sounds arisingfrom these different sources.

To place these findings in a moregeneral context, we may note that thecomposer Berlioz has argued for thecompositional importance of spatialarrangements. As he wrote in hisTreatise on Instrumentation:

I want to mention theimportance of the dif ferentpoints of origin of the tonalmasses. Certain groups of anorchestra are selected by thecomposer to question andanswer each other; but thisdesign becomes clear andeffective only if the groupswhich are to carry on the dia-logue are placed at a sufficientdistance from each other. Thecomposer must therefore indi-cate in his score their exactdisposition. For instance, thedrums, bass drums, cymbals,and kettledrums may remaintogether if they are employed,as usual, to strike certainrhythms simultaneously. But ifthey execute an interlocutoryrhythm, one fragment of whichis given to the bass drums andcymbals, the other to kettle-drums and drums, the ef fectwould be greatly improvedand intensified by placing thetwo groups of percussioninstruments at the oppositeends of the orchestra, that isat a considerable distancefrom each other.

The experiments that we have beendescribing indicate that spatial arrange-ments of instruments should indeed

AUDIO/MARCH 198746

Fig. 21—Patterns of Fig. 20, with tones presented to two ears simultaneously(A), switching haphazardly between ears (B), switching haphazardly and accom-panied by a drone in the opposite ear (C), and switching haphazardly andaccompanied by a drone in the same ear (D).

Fig. 20—Simple melodic patterns usedto examine effects of time differenceson perception.

AUDIO/MARCH 1987 47

have profound ef fects on how music isperceived. When a rapid pattern oftones is distributed between two sets ofinstruments, and these tones are clearlyseparated in time, we may be unable tointegrate them so as to form a coherentmelody. If, however , the tones overlapin time, such integration is more readilyachieved. But there is a trade-of f: Asthe temporal overlap is increased, ourability to identify the locations of differentsounds decreases, and when the tonesare simultaneous, spatial illusions tendto occur.

Let us finally return to the question ofhow perception of simultaneous tones isaffected by whether the higher tone is tothe right and the lower to the left, or viceversa. As we have seen in the octaveand scale illusions, right-handers tend tohear the higher tones on the right andthe lower tones on the left, regardless oftheir actual locations. So combinationsof the “high-right/low-left” type tend to becorrectly localized, and combinations ofthe “high-left/low-right” type tend to bemislocalized. Other recent experimentshave shown this to be true in more gen-eral settings also. And in further study Ifound that, in addition, there is anadvantage to the “high-right/low-left” dis-position in terms of how well the pitchesof the tones are perceived. (SeeDeutsch, D., “Dichotic Listening toMelodic Patterns and Its Relationship toHemispheric Specialization of Function,”Music Perception , V ol. 3, 1985, pgs.127-154.)

Now, to the extent that ef fects of thissort occur in listening to live music, wemay advance the following line of rea-soning. In general, seating arrange-ments for contemporary orchestras aresuch that, from the performers’ point ofview, instruments with higher registerstend to be to the right, and instrumentswith lower registers to the left. As anexample, Fig. 22 shows a seating planfor the Chicago Symphony, viewed fromthe rear of the stage. In the string sec-tion, the first violins are to the right of thesecond violins, which are to the right ofthe violas. These are, in turn, to the rightof the cellos, which are to the right of thebasses. In the brass section, the trum-pets are to the right of the trombones,which are to the right of the tuba. Noticealso that the flutes are to the right of theoboes, and the clarinets to the right ofthe bassoons. The same general princi-

ple holds for choirs and other singinggroups. Since it is important that the dif-ferent performers in an ensembleshould be able to hear each other aswell as possible, we may conjecture thatthis type of arrangement has evolved bytrial and error because it is conducive tooptimal performance.

But this presents us with a paradox.Since the audience sits facing theorchestra, as shown in Fig. 23, this left-right disposition is, from their point of

view, mirror-image reversed: Instru-ments with higher registers are now tothe left, and instruments with lower reg-isters to the right. So from the audi-ence’s standpoint, this arrangement issuch as to cause perceptual dif ficulties.In particular, instruments with low regis-ters which are to the audience’ s rightshould tend to be poorly perceived andlocalized.

It is not all clear what can be doneabout this. We can’t simply mirror-image

The spatial arrangements of instruments should indeed have profound effects on how music is perceived.

Fig. 23—Same as Fig. 22 but as viewed from audience.

Fig. 22—Seating plan for Chicago Symphony, as viewed from rear of stage.

AUDIO/MARCH 198748

reverse the orchestra, because then theperformers wouldn’t be able to heareach other so well. Suppose, then, thatwe turned the orchestra 180°, as awhole, so that the players had theirbacks to the audience. This wouldn’tprovide a solution, because then thebrasses and percussion would be clos-est to the audience, and so woulddrown out the strings. Suppose, then,that w e “ retrograde-inverted” t heorchestra so that they had their backs tothe audience, with the brasses and per-cussion farthest away and the stringsthe closest. This wouldn’t provide asolution either , because then the con-ductor wouldn’t be able to hear thestrings, and so wouldn’t be able to con-duct efficiently.

One solution (suggested by my col-league Robert Boynton) would be toleave the orchestra as it is, but havethe a udience h anding u pside-downfrom the ceiling! (See Fig. 24.) Thissolution is, however, unlikely to be pop-ular with concert-goers! On the otherhand, for the case of sounds repro-duced in stereo, an obvious suggestionpresents itself: Try reversing the chan-

nels on your system. This solution isnot without its drawbacks; the musicwon’t sound the same as in concerthalls, and the arrangement will be unfa-miliar even as a reproduction. But youmay want to try the experiment anyway.

Finally, I should mention that most ofthe perceptual ef fects described hereoccur even though the listener has fullinformation as to what the sound pat-tern really is. There are other cases inlistening to music, however , in whichprior knowledge of the music has a pro-found influence on how it is perceived.One such ef fect, which I originallydemonstrated using the tune “Y ankeeDoodle,” is particularly striking. If youplay a well-known melody, but displaceits individual notes at random into dif-ferent octaves, people will be unable torecognize the melody unless they aregiven clues on which to base a hypoth-esis (such as its rhythm, its contour ,and so on). But if you give the listenerthe name of the melody beforehand,this problem essentially disappears.(See Deutsch, D., “OctaveGeneralization and Tune Recognition,”Perception and Psychophysics, Vol. 11,

1972, pgs. 411-412.)Sound Example 12 presents anoth-

er well-known melody , with its tonesplaced haphazardly in different octavesin this fashion. Listen to this example,and try to identify the tune. Then listento Sound Example 13, which presentsthe same melody without the octave-randomizing transformation. Finally ,listen to Sound Example 12 again, andyou will find that the melody is nowmuch easier to follow.

This little experiment can also easi-ly be performed by anyone with accessto a musical instrument. Make sure,though, that you don’t give your sub-jects any hints as to what the melodyis, and that you scramble the octavesvery well, or they might recognize themelody on the basis of a small part thatwas left intact. Also, choose a melodythat is as free of rhythmic cues as pos-sible, or they might be able to make theright guess on the basis of the rhythmalone. If you follow this procedure, it’ spretty sure to work!

Additional ReadingBerlioz, H., Treatise on Instru-

mentation, I. Strauss, editor , and T.Front, translator (E. F . Kalmus,1948).

Butler, D., “Melodic Channeling in aMusical Environment,” presented atthe Research Symposium on thePsychology and Acoustics of Music,Kansas, 1979.

Deutsch, D., “Auditory Illusions,Handedness, and the SpatialEnvironment,” Journal of the AudioEngineering Society, Vol. 31 (1983),pgs. 607-622.

Deutsch, D., “Musical Illusions,”Scientific American , V ol. 233(1975), pgs. 92-104.

Deutsch, D., “The Processing of PitchCombinations,” The Psychology ofMusic, D. Deutsch, editor(Academic Press, New York, 1982).

Machlis, J., The Enjoyment of Music,Fourth Edition (Norton, New Y ork,1977).

Varney, N. R. and A. L. Benton, “TactilePerception of Direction in Relationto Handedness and FamilialHandedness Background,”Neuropsychologia, Vol. 13 (1975),pgs. 449-454.

Zangwill, O. L., Cerebral Dominance

If a familiar melody is played with its notes displaced in different octaves, people will be unable to recognize it.

Fig. 24—One way to optimize the left-right arrangement of an orchestra, both forthe players and the audience.

Illus

tratio

n: P

hilip

And

erso

n