Lateral asymmetry in the perceived sequenceof speech and nonspeech stimuli*

PAUL BERTELSON and FRAN~OISETISSEYREtUniversity of Brussels, Brussels, Belgium

The task was to estimate the position where a click had been superimposed ina spoken sentence. Experiment 1 confirmed Fodor and Bever's observation of anear-asymmetry effect: the click is located earlier when it is presented to the leftear and the sentence to the right ear than with the opposite arrangement, InExperiment 2, combinations of monaural and binaural presentations wereconsidered. They made it possible to eliminate interpretations which link thelaterality effect to the fact of reaching or not reaching a particular ear andshowed that the relevant factor is the relative position of the stimuli in acousticspace. Experiments 3 and 4 explored the relation between spatial separation andperceived sequence in greater detail. The relation involves a plateau: when theclick comes to the left of the speech, it is preposed to a degree which depends onthe amount of spatial separation; but, when it comes to the right of the speech,separation is irrelevant and the mean error is of the same order of magnitude asin a control condition without separation.

The present study was stimulatedby an observation reported by Fodorand Bever (1965) in the well-knownpaper where they analyzed theinfluence of grammatical structure onthe apparent location of an extraneoussignal in a spoken sentence. Contraryto Ladefoged and Broadbent (1960),who had used diotic presentationsthrough loudspeakers in open space,Fodor and Bever used dichoticpresentations through earphones, thespeech being delivered to one ear andthe signal (a click) to the other ear.They mentioned among theiraccessory resul ts that negativedisplacements [i.e., Iocating the signalearlier than its objective position)prevailed for the Sa who heard thesignal in the left ear and the speech inthe right ear and positivedisplacements for those who had theopposite arrangement. This remarkablefinding has been confirmed in furtherexperiments by Bever and hisassociates (Bever, Lackner, & Stolz,1969; Bever, Lackner, & Kirk, 1969).It has, however, not been subjected toa systematic experimental analysisand, as a result, has gone somewhat

*Tbis wodl: was camed out underContract 612 between the Belgian "Fondsde Ja Recherche fondllJllentale collective,"the University of BrusseIs, P. BerteIson(Laborato~de PsYchologie ex~rimentale),and M. WlQskop (Institut de PhontStique).Help was received from the "Institut dePhonetique" at an stages: MoniqueWlÜskop-Hianne constructed and recordedthe sentences; A. Landercy developed theprocedure for positioning the clicks,designed the voice key, and calibrated thec1icks and the sentences: and M. WlÜskopmade the remazkable faci1ities of bisJaboraOOry available.

t'Requests for reprints should be sent 00:P. BerteIson. LaboraOOire de Psychologieex~rimentale, 117 avenue Ad. Buyl, 1050Bruxelles, Ilelglum.

unnoticed.! despite the wide interestwhich has developed recently forlaterality effects in dichotic listening(Kimura, 1961, 1964) and intachistoscopic reeognition (see thereview by White, 1969). The followingexperiments were run to analyze thephenomenon further.

MATERIAL AND APPARATUSThe material was the same as in a

previous study by the authors(Bertelson & Tisseyre, 1970). Briefly,it consisted of French sentences, 11 to15 syllables long (mean: 13.6),recorded at a very fast pace (7syllables/sec) by a female speaker.There were four lists of 10 sentenceseach, matched for intonation pattern.The sentences were recorded on onetrack of the tape, and 1-msec signalswere positioned on the other track, atlocations always corresponding to themiddle of a vowel. Positions rangingfrom the 4th to the 12th syllable ofthe sentenee were used, with roughlyequal probabilities. The mean positionwas 0.4 syllable after the middle of thesentence.

A set of Sharpe HA-8 paddedearphones was used for delivering thematerial to the Sa. Each earphone wasconnected to one output channel of astereophonie amplifier (MBLEType 660/845). The tapes were playedon a Revox A77 tape recorder. Theoutput of the signal track wasconnected to a voice key (Landercy,1970) which, when activated,produced a 1-msec rectangular pulse ofadjustable amplitude. This was fedeither to one or to both low-Ievelinputs of the amplifier, while theoutput from the other track of thetape recorder was aimilarly connectedto the high-level inputs. The sentenceand the click could thus each be sent

to one particular ear or to both ears,making all combinations of monauraland binaural presentations possible.

The loudness levels of the sentencesand of the click were measured usingan artificial ear (Brüel & Kjaer,Type 4152) with a 6-cc coupler, aBrüel & Kjaer 2603 microphoneamplifier, and a Hewlett-PackardLoudness Analyzer, Model 8501 A.The peak Ioudnese levels for thesentences ranged from 25 to 35 sones.The clicks used in Experiments 2, 3,and 4 were 35 sones, Those ofExperiment 1 were 35 (weak) and 55(strong) sones.

EXPERIMENT 1The main purpose of this

experiment was to replicate Fodor andBever's observation of anear-asymmetry effect. The twopresentation modes used by theauthors, i.e., sentence right/click leftand sentence left/click right, werecompared. Simultaneously, thepossible role of the intensity of theclick was analyzed. This was animportant precaution for thecombinations of monaural andbinaural presentations which wereplanned for later experiments.

MethodFour conditions were compared,

resulting from the combination of twopresentation modes-SRCL (sentenceto right ear, click to left ear) andSLCR (opposite arrangement)-and oftwo loudness levels for theclick-strong (55 sones) and weak(35 sones). The latter was equal to thepeak level of the loudest sentences.Each S had one list of 10 sentencesunder each of the four conditions.Half the Ss started with the twoconditions with the strong click, halfwith the weak click. Within eachgroup, the order of the twopresentation modes was similarlybalanced. Four Sa were tested undereach of the four resulting orders, andthe order of the four lists wasorganized in Latin square among them.

'Ille 16 Sa were students and werepaid for their participation.

They were told that a click wouldbe heard during each sentence, that itcould occur in any position, and thattheir task was to report their bestestimation. They wrote down thepassage around the place where theythought the click had oceurred, with avertieal bar indieating that position.

Errors were scored in number ofintersyllable intervals separating theposition whieh was reported from theposition on whieh the cliek actuallyoeeurred, a half unit being countedwhen the reported position fell exactlyin the interval between two syllabies.'Illey were called negative when the

356 Copyright 1972, Psychonomic Society, Inc., Austin, Texas Perception ~ Psychophysies, 1972, Vol. 11 (5)

*ln one-syllable units. See text (or description o( scorinll procedure,tNellative means that the reporteä position precedes the actual position o( the click.

Table 1Re_la of Experiment 1

(lb)

112518

81813

922

>0

202221

252826

2225

=0

Percent Errorst

695361

675461

6853

<0

8MCL ;; SLCM and SRCM < SMCR;

(la)

Following the other version, called theright-ear-delay hypothesis, materialreaching the right ear is perceived ascoming later than similar materialreaching only the left ear. A binauralstimulus is delayed to the same extentas a right monaural one, in comparisonto a left monaural one. If it is positionin auditory space that influencesperceived order, a binauraI stimuluswill be perceived as coming later thana left monaural one, but earlier than aright monaural one.

Peculiarly convenient for testing theear-of-entry hypotheses are the pairsof conditions SMCL-SLCM andSR CM-SMCR. In the two conditionsof the first pair, the click and thespeech both reach the left ear; hencethe left-ear-precedence hypothesispredicts the same mean error. On theother hand, since only the speechreaches the right ear in Condition8MCL and only the click does inCondition SL CM, the right-ear-delayhypothesis predicts a smaller (or morenegative) mean error in SMCL than inSL CM· For the pair SR CM-SMCR, theopposite predictions are made: left-earprecedence predicts a smaller meanerror in SR CM than in SMCR, andright-ear-delay predicts no difference.To sum up: left-ear precedencepredicts

8MCL < SL~ and SRCM < 8MCR

(lc)

(symbols like 8MCL will be used

right-ear-delay predicts

Finally, sinee in each pair the twoconditions have the components inopposite relative positions, thespatial-position hypothesis predicts

Condiüon

Cliek Presentation MeanIntensity Mode Error*t SD*

SRCL -.95 1.20Weak SLCR -.68 1.35

Mean -.82 1.28

SRCL -.99 1.15Strong SLCR -.59 1.09

Mean -.79 1.12

Mean ofBoth SRCL -.97 1.18lntensities SLCR -.64 1.22

is ambiguous from at least two pointso f vi e w : (1) Material presentedthrough an earphone to the left ear isfeIt as coming from a source in the leftpart of auditory space, and vice-versafor the right ear. The relevant factor inear asymmetry might thus be the factof reaching or not reaching a particularear or the apparent location of thesource in space, (2) The side on whichit is presented might be critical forboth the click and the speech or foronIy one of the two components.Other ear-asymmetry effects have beenfound to be specific of particularmaterial, as in the well-known case ofdichotic memory where speechreceived in the right ear is betterrecalled than speech receivedsimultaneously in the left ear (Kimura,1961), while the opposite is true ofmelodies (Kimura, 1964) and ofemotional tone (Haggard & Parkinson,1971). In Experiment 1, the speechand the click changed sidessimultaneously and the possible effectsof these two changes wereconfounded.

The question of ear of entry vsposition in auditory space can beanswered by using binauralpresentations. AB is weil known, astimulus presented binaurally throughearphones is feit as originating from asource in the middle of the head. Itsapparent loeation is thus in betweenthose of left and right monaural ones(hence it will be labeled as comingfrom the "middle" and thecorresponding symbols will be 8M andCM). For such a stimulus, ear-of-entryand spatial-position interpretationsgive different predictions. There are, infact, two different ear-of-entryinterpretations which are consistentwith the observed ear asymmetry. Forone, which we shall call the left-earprecedence hypothellis, materialreaching the left ear is perceived ascoming earlier, relative to someconstant reference stimulus, thansimilar material reaching only the rightear. A binaural stimulus obtains, then,the same precedence as a left monauralstimulus over a right monaural one.

reported position preeeded the actualposition. If, e.g., the bar on the S'sprotocol fell on any of the letten ofthe syllable following the actualposition, an error of +1 syllable wascounted. If it fell between the twosyllabies, the score was +0.5 syllable.In our previous study (Bertelson &Tisseyre, 1970), the SB were told thatthe click was always superimposed onvowels, and so no provision had to bemade for responses falling betweensyllabies. It was decided not to providethat information any Ionger for fearthat it might reduce the sensitivity ofthe responses, and the scoring systemhad to be adjusted accordingly.

The Ss had no prior knowledge ofthe sentences which were presentedonly onee, with the click. Before eachlist began, the way in which thematerial would be distributed betweenthe ears was explained.

Results and DiscussionIn Table 1, the responses have been

broken down into the three eategoriesof negative errors, correct responses,and positive errors. At both clickintensities, there are more negativeerrors and less positive ones inCondition SR CL than in SLCR. Theproportion of correct responses is little"lffected. On the other hand, thebreakdown is really not influenced bythe intensity of the cllck, except for avery slightly larger proportion ofcorrect responses with the strong click.

The same general pattern appears inmean errors, also shown in Table 1,which are not affected by clickintensity but are very strongly affectedby presentation mode, Analysis ofvariance showed a significant effect ofpresentation mode (F;; 15.1,df > 1,15, P < .01) and no effect ofclick intensity (F < 1).

The apparent temporal position ofthe click is clearly affected by the waythe two components of the task areallotted to the two ears, Fodor andBever's ear-asymmetry effect has thusbeen confirmed.

On the other hand, the results differfrom Fodor and Bever's in that astrong tendency towards negativedisplacement is observed under bothpresentation mod es , Even inCondition SL CR, where the tendencyis less pronounced, 68% of erroneousresponses are on the negative side(against 43% in Fodor and Bever'sexperiment) and the mean error isnegative in 12 of the 16 Ss (p ;; .08 bya sign test) and for 36 of the 40sentences. The problem will beconsidered again in the generaldiscusslon.

EXPERIMENT 2The nature of the ear-asymmetry

effect as demonstrated by Fodor andBever and confirmed in Experiment 1

Perception & Psychophysics, 1972, Vol. 11 (5) 357

Table 2Result. ot Experiment 2

Percent ErrorsMean

Condition Error SD <0 =0 >0

SMCL -.86 1.21 63 24 13SMCR -.40 1.06 47 32 21SRCM -.87 1.06 66 26 8SLCM -.53 1.18 50 35 15

indifferently to denote a condition orthe mean error under that condition).

Other pairs of conditions (such asSR CL -SR CM or SL CL -SMCL) wouldalso provide tests of ear-of-entryinterpretations, but the present pairshad the interesting property, for a firstexamination, that each predictionholds whether laterality affects theperception of the sentence only, of theclick only, or of both.

The question of specificity can beanswered by two other comparisonsinvolving the same four conditions.The predictions are straightforwardhere. If presentation mode mattersonly for speech, we have

SRCM < SLCM and SMCL = ~CR'

(2a)

If it matters only for the click, wehave

(2b)

If it matters for both,

SRCM < SLCM and SMCL < ~CR'

(2c)

AB it is possible that the effectwould in some way be related tolateral dominance, only right-handedSs were used in this and the followingexperiments. The results of Ss whoeither wrote with the left hand ordeclared being left-handed orambidextrous, when questioned afterthe experiment, were discarded fromthe present analysis. No suchprecaution had been taken inExperiment 1.

MethodThe responses of 32 inexperienced

right-handed Ss will be analyzed. Eachheard the four lists, one under each ofthe four conditions: binaural speechand click to the left ear (SMCL ) or theright ear (~CR)' binaural click andspeech to the left ear (SL CM) or theright ear (SR~). Sixteen Ss startedwith the two conditions with binauralspeech, and the other 16 with thosewith binaural click. Within each group,a subgroup of 8 Ss had the monauralstimulation first left then right and the

other subgroup in the opposite order.'Ibe order of the lists was organized infour different Latin squares. 'Ibe levelof the speech and of the click were leftat the same intensity levels (click,35 sones; speech peak level, about35 sones) during binaural andmonaural presentations. This meansthat the binaurally presentedstimulation was felt to be louder thanthe monaurally presented one. Sinceno effect of click intensity had beenfound in Experiment 1, it seemedunnecessary to compensate for thesedifferences.

ResultsThe mean errors appear in Table 2.

Variance analysis revealed a conditioneffect significant at p = .001 (F = 14.3,df = 3,93). Subsequent application ofScheffe's test (Hays, 1963, p. 484)permitted the formation of twogroups, one with Conditions SR CMand ~CL and the other withConditions SLCM and SMCR , suchthat each condition of one group givesresults significantly different at p = .05from each condition of the othergroup and no difference within agroup is -significant. In other words,each condition with the clickoriginating from a location to the leftof the speech produces a morenegative mean displacement than eachof the conditions with the click to theright of the speech. If proportions ofthe different response categories areconsidered instead of mean errors, thesame general picture emerges.

DiscussionThe two ambiguities described in

the introduction have been eliminated.(1) The results are inconsistent withthe predictions of the two ear-of-entryinterpretations (la and Ib), and theysupport those (1c) of thespatial-position hypothesis. (2) Theysupport predictions (2c) derived fromthe hypothesis that the laterality ofboth the click and the speech affectperceived order.

If we combine both results, itappears that the relevant factor inlateral asymmetry is the relativeapparent location of the sources ofstimulation in auditory space,

AB usual, the negative aspects of theconclusions, those regarding theinterpretations that have beenrejected, can be stated more firrnly

than the positive aspects. Forspecificity, we can say that the spatialposition of each component can affectperceived order. The conclusion,however, rests on the particular caseswhere one component is shifted fromthe left to the right, while the othercomponent is kept in the middle. Onecannot affirm that, whatever theposition of one component, movingthe other one will affect perceivedorder.

Similarly, the support for aspatial-position interpretation comesfrom the sole consideration of whathappens when the two componentsexchange positions between themiddle and one side. This does notpermit us to specify the relationbetween perceived order and spatiallocation in a detailed way. Thesimplest possible relation would bethat the apparent time of occurrenceof the click relative to the speech isprogressively delayed when it is movedfrom the farthest position to the leftof the speech to the farthest one tothe right. But a difficulty for thissimple hypothesis arises immediatelyfrom a comparison of the results ofExperiments 1 and 2. Lateralityproduced an effect of similarmagnitude (.33 and .39 syllabies),despite the fact that left-rightseparation of the components wasinvolved in Experiment 1 and onlyleft·middle or middle-right inExperiment 2. 'Ibere is thus asuggestion that, although perceivedsequence depends on relative positionof the components in acoustic space, itmight not depend on the distancebetween them, The next experimentwill be devoted to that problem.

EXPERIMENT 3This experiment was run to

compare the lateral asymmetrybetween conditions with full-headseparation of the components andbetween conditions with half-headseparation. Since there were fourhalf-head separation conditions toconsider, and only four lists ofsentences were available, thecomparison was made in two steps, ontwo separate groups of Ss.

MethodTwo groups of 16 right-handed

inexperienced Ss were tested. 'Ibe Ssof Group 1 heard one list under eachof Conditions ~CL, SMCR, SR CL,and SL CR and those of Group 2 underConditions SR CM, SLCM, SR CL, andSLCR, The order of presentation ofthe four conditions was organized inLatin squares.

Results and DiscussionThe mean errors appear in Table 3.

In each group, the laterality effect due

358 Perception &; Psychophysics, 1972, Vol. 11 (5)

Table 4Experiment 4: Mean Erron and Standard Deviations

Table 3Experiment 3: Mean Errors and Standard Deviations

Results and DiscussionThe mean errors are given in

Table 4. In Group L, the negativedisplacement increases with increasingseparation of the components.Variance analysis revealed a significanteffect of conditions (F =4.8,df = 3,69; p < .01). The differenceswere further analyzed through plannedcomparisons: (1) The mean error overConditions ~CL and SR~ wasfound to be significantly more

MethodTwo groups of 24 inexperienced

right-handed Ss were tested. Each Sheard the four lists, one under each offour presentation modes. The Ss ofGroup L had Conditions SRCL,SR CM, SMCL, and SMCM. Those ofGroup R had Conditions SLCR,SL CM, SMCR, and SMCM. For eachgroup, the order of the conditions wasorganized using six Latin squares. Thefour lists were always presented in thesame order,

negative than the mean for ControlCondition SMCM (t = 2.07, df = 69;p< .025). (2) Tested in the same way,the difference between ConditionSR CL and the mean of SMCL andSR CM fell just short of the .05 level (t

1.63; t.05 1.67). (3) Thedifference between ~ CL and SR CM,for which there was no prediction, wassubmitted to a two-tailed test andfound to be nonsignificant (t = 1.17;t. 1 0 = 1.67).

In Group R, the control condition,~ CM, produces a slightly morenegative mean error than the threeother conditions, wh ich givepractically identical results. Varianceanalysis yielded a nonsignificant effectof conditions (F< 1). Plannedcomparisons, run according to thesame principle as in Group L, gave nosignificant outcome: the onlynoticeable difference, that betweenSMCM and the mean of ~ CR andSL CM, did not reach the .10 level.

This picture of the relation betweenapparent order and spatial separationis in agreement with the suggestionbrought forth by Experiment 3.Conditions where the click is to theleft of the speech give larger negativedisplacements than does the controlcondition with both components inthe middle. Among these conditions,the negative displacement seems toincrease with increasing spatialseparation of the components. On theother hand, for the conditions wherethe click comes to the right of thespeech, degree of separation seems tomake no difference and the negativedisplacement is not, or is only slightly,smaller than in the control condition.

The affirmation that an effect ofdegree of separation was shown inGroup L might be criticized, not somuch in view of the low significancelevel at which the null hypothesis wasrejected, but mainly because itinvolves considering as meaningful the.08 syllable difference betweenConditions SMCL and SR CL, whilethe larger difference (.12 syllable)between ~ CL and SR CM is discardedas nonsignificant. The conclusion wasmade possible by the pooling of thelatter two eonditions, There was goodjustification for pooling: the twoconditions had yielded virtuallyidentical results in Experiment 2. But

Group R

Meansn Condition Errot sn1.16 SLCR -.49 1.08

1.15 SLCM -.48 1.151.05 SMCR -.49 1.22

1.05 SMCM -.58 1.19

sn1.241.10

1.071.10

Group L

Degree of MeanSeparation Condition Error

Full-head SRCL -.85

Half-head SRCM -.65SMCL -.77

No Separation SMCM -.54

present experiment was performed tocompare (1) conditions with full-headseparations with conditions withhalf-head separation, and(2) conditions with half-headseparation with a control conditionwithout separation, with separatecomparisons being run for the caseswhere the click is to the left of thespeech and where the click is to theright of the speech. For the sake ofhomogeneity, the experiment was runwith the same set of four sentences asthe former ones, which involved usingtwo separate groups of Ss and limitingoneself to one only of the threepossible control conditions, The onewith both components presentedbinaurally was chosen.

to full-head separation (SL CR -SR CL )is larger than the one due to half-headseparation: .56 against .23 in Group 1and .52 against .34 in Group 2. In eachgroup, the corresponding plannedcomparison (Hays, 1963, p. 462ff) wasperformed as folIows: Group 1,(SLCR - SRCL)- (SMCR - ~CL) =.33; t = 2.0, df = 45; p < .05; Group 2,(SLCR -SRCL)-(SLCM -SRCM) =.19; t = 1.36, df = 45; p < .10.

If one limited oneself to thequestion asked in the introduction, theconclusion would be that, despite therather low significance levels, theresults run counter to the suggestionderived from the two previousexperiments: full-head separationproduces a larger lateral asymmetrythan does half-head separation. ButTable 3 shows clearly that thedifference is due wholly to theconditions where the click comes tothe left of the speech: SR CL producesa larger negative displacement thaneither SMCL or SR CM (the differencesare .28 and .30 syllable, respectively).When the click comes to the right ofthe speech, only small and erraticdifferences are observed betweenSLCR and either SMC R or SLCM (.05and -.11 syllable, respectively). Theseobservations suggest strongly thatamount of separation might influenceapparent order when the click is to theleft of the speech, not when it is to theright. The suggestion, which had notheen anticipated, can, however, not betested on the present data,

Proportions in the differentresponse categories provide no furtherinformation and have been omitted.

EXPERIMENT 4All the previous analyses have been

based on differences betweenconditions with opposite lateralities:conditions with the click to the left ofthe speech have been compared toconditions with the click to the rightof the speech. Control conditionswithout spatial separation have notbeen considered. Now thatExperiment 3 has suggested that theeffects of spatial separation might notbe equally strong for all relativepositions of the components, itbecomes necessary to study therelation in a more detailed way. The

Group 1 Group 2

Degree of Mean MeanSeparation Condition Ezror' SD Condition Errot

Full-head SRCL -1.07 1.23 SRCL -1.19SLCR -.51 1.16 SLCR -.67

Half-head SMCL -.79 1.12 SRCM -.90SMCR -.56 1.23 SLCM -.56

Perception & Psychophysics, 1972, Vol. 11 (5) 359

~04er

I lerw

- '5 _-----rJzCl

Ll "' ............

~ - '6 ... '

- '7

·'8SPEECH,

0-9 " LEFT _.- LEFT

"" " MIDOLE - MIDDLE

-1-0 T,/RIGHT ----- RIGHT'!

1

Fig. 1. Mean error per condition as afunction of spatial separation betweenclick and speech. All responsescollected during the four experimentshave heen pooled (for Experiments 2,3, and 4, right-handed Ss only). Eachpoint is based on 480 responses forCondition SMCM, 880 for ConditionsSRCL and SLCR, and 720 for all otherconditions. Separation is negativewhen the click is to the left of thespeech: -1 and +1 correspond toleft-right separation, -.5 and +.5 toleft·middle or middle-right separation.Brackets indicate ±1 standard errorintervals.

of ear-of-entry Interpretations, forboth share with spatial-positioninterpretations the prediction that:SLCL =SMCM =SRCR•

GENERAL DISCUSSIONApart from confirming the

existence of an ear-asymmetry effectin the click location task, ourexperiments have brought severalprecisions regarding the origin of thephenomenon. It now appears as oneparticular manifestation of a moregeneral relation hetween perceivedsequence and the relative positions inauditory space of the apparent sourcesof the speech and of the click. Therelation, however, is not a simple one:perceived order is influenced by spatialseparation only when the click is tothe left of the speech.

The expression "relative positions inauditory space of the apparentsources" should be read keeping inmind that only one way of creating animpression of spatiallocation, namely,manipulating the distribution of thecomponents between the twoearphones, has been used throughout.Other ways should be tried before theconclusion can be generalized. Oneshould, for instance, have the speechand the click delivered throughloudspeakers with different angularseparations, A comparison of thepresent results with those of a previousstudy using the same material

even if based on sound statisticalp r o c e d u r e , the result iscounterintuitive. For that reason, itmay be worth mentioning that in apreliminary experiment where 16 Sswere run under exactly the sameconditions as Group L (but with a lesssatisfactory control of list effects-thereason why the data were not includedin the analysis), the mean errors forConditions SMCL and SR CM were,respectively, -.55 and -.70 syllable.The difference was thus in theopposite direction to the one observedin Group L, and, in fact, the twoconditions give exactly the same mean(-.67) when the results of the twogroups of Ss are pooled,

The fact that perceived order is notinfluenced by separation when theclick is to the right of the sentencedoes not discard the previousconclusion that the relevant factor inlateral asymmetry is relative positionof the components. AB was noted inthe discussion of Experiment 2, theconclusion was compatible withseveral relations between perceivedsequence and spatial position.Experiments 3 and 4 have now shownthat the relation is more complex thana simple constantly increasing one. Tohelp visualizing the relation, the meanerrors per condition have heencalculated over all the responsescollected in all four experiments andare shown in Fig. 1. Such data, wheredifferent groups of Ss havecontributed to different but partiaIlyoverlapping sets of means, could notbe submitted to statistical tests; but inview of the rather high variability ofthe data, it was feit worthwhile topresent estimates based on the largestavailable sampies. The pattern whichemerges is rather clear.

In Experiment 2, rejection of thetwo ear-of-entry interpretations wasbased on the respective findings thatSLCM < SMCL and SRCM < ~CR'It was mentioned in the discussionthat other comparisons could haveheen used to the same purpose, andsome of them have now becomeavailable. Each ear-of-entry hypothesispredicts one particular ordering of themean errors under the sevenconditions which have been used, andthese orderings can be compared withthe data appearing in Fig. 1. Theleft-ear-precedence hypothesis predictsSRCL = SRCM < SMCL = SMCM =SLCM < ~CR = ~CR and theright-ear-delay hypothesis predictsSRCL = SMCL < SRCM = ~CM =SMCR < SLCM = SLCR, Each of thesepredictions contains numerousdiscrepancies with the data.

The omission of Control ConditionsSL CL and SR CR, which might appearas a weak point in our evidence, is ofno consequence regarding the rejection

01'0 0'5 o ·'5 ·1'0SPATIAL SEPARATION

presented through two loudspeakers(Bertelson & Tisseyre, 1970) suggeststhat the two situations do notnecessarily yield identical results. Inthe former study, where there was noseparation between the loudspeakers,the mean error was about -1 syllable,which is about twice the negativedisplacement observed inExperiment 4 under the controlcondition without separation and ofthe same order of magnitude as inCondition SR CL' The measures arenot strictly comparable, in view of thedifferences in instructions between thetwo studies, but it would be surprisingif these differences could bythemselves produce such large shifts inmean errors.

The relation of the presentphenomenon to other perceptualasymmetries is far from clear. This isespecially the case of the earasymmetry observed in dichoticspeech recognition (Kimura, 1961,1966; Bryden, 1963). Kimura'sinterpretation of that phenomenon interms of localization of the speechcenter and occlusion of ipsilateralspeech messages by contralateral oneswould have difficulties explaining aneffect linked to spatial position, not toear of entry. Kinsboume's (1970)more sophisticated idea of a relationbetween central dominance and thespatial distribution of attention wouldavoid that difficulty but wouldpresumably predict some precedenceof material presented to the right.Irrespective of the particularexplanation which is considered, it isdoubtful whether a single mechanismmight explain both right-ear advantagein dichotic speech recognition andprecedence of a click presented to theleft of the speech in the click locationtask.

Regarding the possibility of arelation to cerebral dominance, thepooled results of the 18 Ss whoseresults have heen discarded from themain analysis in Experiments 2, 3, and4 because they wrote with the lefthand or reported left-handed orambidextrous habits have beenpresented in Table 5. Their meanerrors are on the whole slightly lessnegative than those of theright-handed SB, but the differencesbetween conditions displayalateralasymmetry of same size and direction.The observation offers no support tothe notion of some relation to cerebraldominance but Is, of course, notsufficient to discard it: handedness is arather poor predictor of cerebraldominance (Kimura, 1966), and ourway of assessing it was obviously notoptimal.

Another possible relation whichmight be worth exploring is to culturalstereotypes linked with the direction

360 Perception & Psychophysics, 1972, V01. 11 (5)

Table I)Experimentt 2, 3, and .: Mean Euora of Left-Handed and Ambidexizous Ss Compared

to ThoN of Ri&ht·Handed Ss

*Experiment 2-Bix SB; Experiment 3. Group I-three SB, Group 2-01le S; Experiment 4,Group Lrrfour SB, Group R-(ive SBtNumber ot reBponBeB

Left-Handed andRight-Handed Ambidextrous*

Mean MeanConditiona Nt Error Nt Error

SRCL 660 -1.01 80 -.80

SRCM,SMCL 1.40 -.81 220 -.56

SMCM,SLCM 2480 -.52 .20 -.31SMCR,SLCR

of reading. Time runs from left toright on most g r a p h i c a lrepresentations, and people oftenmake a gesture to the left or to theright when saying "before" or"after.,,2 A companion paper isdevoted to that problem.

In the authors' opinion, it is tooearly to speculate on the nature of themechanism which might produce theeffect. The only conclusion that canbe stated firmly for the moment isthat, at sorne stage of speechprocessing, temporal sequence must berepresented in a code which isinfluenced by spatial position. Nosuggestion can yet be offered regardingthe reason why the influence is feitonly when the click is to the left ofthe speech.

Before more detailed interpretationscan be attempted, more information isneeded on the possibility of lateralityeffects in other situations involvingjudgments of temporal order. Veryfew data are available. Efron (1963)has examined the effect of spatialseparation on judgments of temporalorder in pairs of visual flashes and ofcutaneous shocks. He found a verysmall (± 6 msec) precedence of stimulipresented on the right side; The effectis thus in the opposite direction to theone studied here and of much smallermagnitude: at the mean rate of 7syllables/sec, the mean value of thelaterality effect between SR CL and8M CM observed in Fig. 1 representsabout 70 msec. In our laboratory,Bauwens (1970) has studied the effectof spatial separation of the sources onthe perceived temporal order of a clickand a flash. In one condition, the flashcame always from a centrally locatedtube and the click from one of threeloudspeakers situated to the left, inthe middle, or to the right. In theother condition, the click always camein the middle and the flash in themiddle or on one side. In no conditionwas any effect of spatial positionapparent. It is tempting to hypothesizethat one of the components submittedto the judgment of order must be atemporally organized pattern for

lateral asymmetry to occur, Importantin this respect is the preliminary resultmentioned by Bever (1971), that earasymmetry is not observed when thespeech component consists of randomword lists.

A problem without direct relevanceto lateral asymmetry concerns thegeneral direction of the errors. Underall conditions, a clear majority ofprepositions has been observed. Evenin the four conditions where thetendency is less pronounced, thosewhere the click was not to the left ofthe sentence (SMCM, SMCR, SLCM,and SL C R ), negative responsesrepresent 52% of all responses, 7~% ofall errors. The tendency emerges fromthe important variations which areobserved between sentences.3 InConditions SMCR and Sz.~ ofExperiment 2, for instance, wheremean error per sentence ranges from-3.20 to +.75 syllabIes, 34 sentencesout of 40 produce a negative meanerror. A similar tendency to preposean extraneous signal superimposed onan unknown string of speech has beenreported by Ladefoged and Broadbentand by Bertelson and Tisseyre (1970)but not by Fodor and Bever (1965),Bever, Lackner, and Kirk (1969), orBever, Lackner, and Stolz (1969).(There was a small overall tendencytowards positive errors in the laststudy and a negligibly small onetowards negative errors in the othertwo.) Since the two studies mentionedfirst were conducted with dioticpresentations in open space and thelast three with dichotic presentations,one could wonder if that difference inpresentation procedure was not at theorigin of the discrepancy. The presentresults seem to rule out thatpossibility. There is for the time beingno convincing explanation for thediscrepancy. More recently, sizablenegative displacements have beenobserved in an experiment by Wanenand Obusek (1971), using a slightlydifferent situation and only oneposition of the extraneous signal inone single sentence, and also in anexperiment by Bever (1971, p. 236),

u8ing monaural preaentationa(Conditiona 8J,CL and SR CR ). Inexperiments reported recently byReber and AnderBOn (1970), which areremarkable for a careful contral ofcllck positions, a majority of negativeerrors was observed early in the aeaaionand gave way progreaaively to positiveones, so that no tendency prevailedover the whole session. Examinationof the procedure auggests, however,that in both experiments the Ba hadthe poaaibility to gain knowledgeabout several syntactic and semanticfeatures of the material, and it hasbeen shown (Bertelson & Tisseyre,1970; Warren & Obusek, 1971) thatprior knowledge can eliminate negativedisplacements.

REFERENCESBAUWENS, M. Entree prioritaire et

jugement d'ordre temporeL UnpubUshedthesis, University of BrusseIs, 1970.

BERTELSON, P.. &. TISSEYRE, F.Perceiving the sequence of speech andnon-speech stimuli. Tbe QuarterlyJournal of Experimental Psychology,1970,22,653-662.

BEVER. T. G. Tbe nature of cerebraldominance in speech behaviour of thechild and adult. In R. Huxley and E.Ingram (Eds.). LangUßge aequisition:ModelB and methods; London: AcademicPress, 1971. Pp. 231-254.

BEVER. T. G., KIRK, R., &. LACKNER, J.An autonomic reßection of syntacticstructure. Neuropsycholo&ia, 1969. 7.23-28.

BEVER, T. G •• LACKNER, J •• &. KIRK. R.Tbe underlying strueture of sentences Arethe primary units of immediate speechprocessing. Perception &. Psychophysics,1969.6. 225-23••

BEVER, T. G •• LACKNER. J., &. STOLZ.W. Transitional probability is not ageneral mechanism for the segmentationof speech. Journal of ExperimentalPsychology. 1969,70, 387-394.

BRYDEN, Mo P. Ear preference in auditorypereeption. Journal of ExperimentalPsychology, 1963, 65. 103-106.

EFRON, R. Tbe effect of handedness on thepereeption of simultaneity and temporalorder. BDin. 1963.86. 261-28•.

FODOR. J., &. BEVER. T. G. Thepsychological reality of lin&uisticsegments. Journal of Verbal Learning &.Verbal Behavior. 1966••••14-420.

HAGGARD. M. P., &. P ARKINSON. A. M.Stimulus and task factors as determinantsof ear advantages. Tbe Quarterly Journalof Experimental Psychology, 1971, 23,168-177.

HAYS. W. SfatiatiCB_ New Yom: Holt,Rinehart &. Winston. 1963.

KIMURA, D. Cerebral dominance and thepereeption of verbal stimuli. CanadianJournal of Psychology, 1961. 15.166-171.

KIMURA. D. Left-right differences in theperception of melodies. Tbe QuarierlyJournal of Experimental Psychology,1964, 16. 355-356.

KIMURA, D. Functional asymmetry of thebraln indichotic listening. Cortex. 1966,3,163-178.

KINSBOURNE. M. Tbe cerebral basis oflateral asymmetries in attention. I~. A. F.Sanders (Ed.), Attention andperformance. III. Amsterdam:Nonh-HoUand. 1970. Pp. 193-201.

LADEFOGED, P •• &. BROADBENT, D. E.Perception of sequence in auditoryevents. The Quarterly Journal ofExperimental Psychology. 1960. 12.162-170.

Perception &IPsychophysics, 1972, Vol. 11 (5) 361

LANDERCY. A. Etude et realisation d'unecl~ vocale. Bruxelles: Rapport d'activitesde I'Institut de Phonetlque, 1969. 2.115-120.

REBER. A. 5 •• &: ANDERSON. J. R. Theperception of clicks in linguistic andnonlinguistic messages, Perception &Psychophysics. 1970. 8. 81-89.

WARREN. R. M•• &: OBUSEK. C. J. Speechperception and phonemic restorations.Perception & Psychophysics. 1971. 9.358-362.

WHITE. M. J. Laterality differences inperception. Psychological Bulletin. 1969.72. 387-.05.

NOTES1. Since this paper was written. Bever

(1971) has discussed the relationship of theear-asymmetry phenomenon to cerebraldominance and given a preliminarydescription of some experiments specificaliydesigned to analyze it.

2. We are indebted to P. A. Ostenieth forthis observation.

3. It· has been suggested that theprevalence of negative errors might resultfrom a combination of (1) the fact that inour material the mean location of the clickswas later (•• syllable) than the middle of thesentence, and (2) a negative correlationbetween mean error and position of theclick in the sentence, as reported bvLadefoged and Broadbent (1960) and byReber and Anderson (1970). The alightasymmetry in the distribution of click

positions may have led to a smalloverestimation of the negative displacementbut cannot be responslble for the wholeeffect. In Experiment 2, for instance, therewas a -.22 produet-moment correlationbetween mean error per sentence anddistance of the click from the middle of thesentence, which means that moving themean click position exactly to the middlemight increase the mean error by about 0.02syllable.

(Accepted (or publication December 8,1971.)

362 Perception & Psychophysics, 1972, Vol. 11 (5)