Perception & Psychophysics1973. Vol. 14. No.3. 518-530

Conservation of temporal information by perceptual systems*ROBERT EFRONt

Neurophysiology-Biophysics Research LaboratoryVeterans Administration Hospital, 150 Muir Road, Martinez, California 94553

Experiments in the auditory, visual, and vibratory modality employed two brief discriminably different stimuli,which were presented with temporal asynchronies unequivocally below the Hirsh-Sherrick threshold for performingtemporal order judgments. A pair of such stimuli, experienced as a unitary perceptual event, is referred to as a"micropattern" composed of two "stimulus elements." Ss could readily distinguish between two such micropatterns inwhich the temporal order of the stimulus elements was reversed. The discrimination was based on the perceptualdominance of the second stimulus element of each micropattern. This perceptual dominance was studied as a functionof stimulus-element onset and offset asynchrony, the duration and intensity of the stimulus elements, and thedifference (in frequency or wavelength) between the stimulus elements of a micropattern. The results suggest theexistence of an operation of all perceptual systems in the time domain that acts to conserve information concerning thetemporal order of the two stimulus elements at the expense of discriminatory acuity of the first element.

Despite the striking anatomical and physiologicaldifferences found in the various perceptual systems(with respect to the structure and function of thespecialized receptor organs as well as in their centralana t omical connections), all perceptual systems,nevertheless, perform certain identical operations on theincoming data. The phenomena of temporal summation,masking, and persistence (Efron, 1973) are examples ofsuch operations in the time domain. The experiments tobe described in the present report reveal the existence ofanother operation common to all perceptual systems inthe temporal sphere-one which conserves informationconcerning the temporal order of two rapidly successivestimuli.

The experiments to be reported are concerned withthe perceptual effects produced by two discriminablydifferent, successive, brief stimuli when they arepresented within a time span that is below the thresholdat which temporal order judgments can be correctlyperformed. The threshold for correct judgments of thetemporal order of two successive stimuli presented in thesame sensory modality has been studied intensively inthe past decade in trained and untrained normal, as wellas in brain-damaged, Ss (Hirsh, 1959; Broadbent &Ladefoged, 1959; Hirsh & Sherrick, 1961; Efron, 1963;Hirsh & Fraisse, 1964; Edwards & Auger, 1965; Holmes,1965; Lowe & Campbell, 1965; Hornick, Elfner, &Bothe, 1969; Jerger et al, 1969; Gengel & Hirsh, 1970;Swisher & Hirsh, 1972). In these various experiments,the two stimuli to be temporally ordered were presentedeither to different groups of receptors (in which casethey could be physically identical) or to the samereceptors (in which case they must be discriminablydifferen t ). In either case, Hirsh and Sherrick (1961)found in trained psychophysical Ss that temporal order

*From the Veterans Administration Hospital and theDepartment of Neurology, University of California, School ofMedicine, Davis, California.

tI would like to thank M. Corder and M. Strachan for beingpatient and good-humored Ss for these tedious experiments.

judgments could no longer be made correctly more than75% of the time when the interval between the onsets oftwo stimuli (stimulus onset asynchrony-SOA) wasreduced to 20 msec. The same value was found for theauditory, visual, and tactile modality and also forcross-modality temporal order judgments. Correspondingvalues of approximately 60 msec were obtained byEfron (1963) and Hirsh and Fraisse (1964) in untrainednormal Ss, and values of 150 to 600 msec have beenobtained with adult patients suffering from lesions ofthe dominant temporal lobe by Efron (1963), Edwardsand Auger (1965), Holmes (1966), and Swisher andHirsh (1972).

It should be noted that in the works cited, thestimulus parameters, the psychophysical methods, andthe criteria used by the Ss varied widely. The strikingeffect of training and criteria on the measured thresholdsfor temporal order judgments in audition was reportedby Broadbent and Ladefoged (1959), who found in theirexperiment that the threshold, which was initially over150 msec, fell precipitously as Ss had more prolongedexposure to the task. The final thresholds were of theorder of those reported by Hirsh (1959) and Hornicket al (1968)-which are the lowest values found in theliterature (17 and 14 msec, respectively, in the auditorymodality).

The specific purpose of the present study was todetermine how temporal order information is processedby the nervous system when the two stimuli arepresented at stimulus onset asynchronies which areunequivocally below the threshold for temporal orderjudgments. If all temporal order information is not lost,then in what form is it retained?

EXPERIMENTAL PARADIGM

Two brief, discrirninably different stimuli aredelivered to a sense organ within such a short intervalthat the 0 is unable to report the correct temporal orderand experiences the two stimuli as a single perceptual

518

Fig. 1. Representation of the experimental paradigm. Incenter of figure are two identical rnicropattems, each composedof Stimulus Elements X and Y. The radial vectors illustrate fourtypes of alterations to which the identical micropatterns weresubjected. For each type of alteration, the stimulus elementasynchrony increases in the direction of the vector. ED indicatesthe element duration; MPD, the micropattern duration;ON-ASYN, the asynchrony between the onsets of the twoelements of a micropattem; and OFF-ASYN, the asynchronybetween the offsets.

event. For example, the two stimuli might be a lO-msecburst of a 1,OOO-Hz signal followed immediately(stimulus onset asynchrony = 10 msec) by a lO-msecburst of a 2,000-Hz signal of the same energy). The SwiIl hear a single complex sound, equivalent to atwo-note chord, in which he can identify both the high­and low-pitch components-but he will be unable toreport (at any level above chance) the actual temporalorder of the two tones. The two questions to which thefollowing experiments are addressed are: (a) Can the Sdistinguish between the perceptual event just describedand a second event in which the same two tones arepresented but in reversed temporal order, that is, whenthe 2,OOO-Hz tone precedes the 1,OOO-Hz tone? Ifdiscrimination between these two perceptual experiencesis possible, then temporal order information isconserved, despite the fact that temporal orderjudgments cannot be made. (b) If discrimination ispossible, on what perceptual feature of the stimuli is thedistinction based?

Essentially identical experiments were performed inthree sensory modalities-audition, vibration, and vision.In each modality, the S was required to discriminatebetween two perceptual events, each of which wascomposed of two stimuli having asynchronous onsets oroffsets which are below the threshold for temporal orderjudgments. To avoid confusion when speaking of"e vents" or "pairs of stimuli." the following

nomenclature will be adopted: Each pair of stimuliproducing a unitary perceptual event will be referred toas a "micropattern." (The term "micropattern" has beenused to distinguish sharply between the temporalparameters of the pattern discrimination tasks describedin this report and the temporal parameters used in themany studies of pattern discrimination in man andanimals where sequences of discriminably differentstimuli are employed in which the intervals between thestimuli are far above the threshold for temporal orderjudgrnents.) Each micropattern is composed of twodiscriminably different stimulus "elements"-X and Y.Expressed in these terms, the Ss are required todiscriminate, using "same" or "different" replies,between two micropatterns (presented 0.5 sec apart),each of which is composed of the same two elements.

The duration of a stimulus element will be referred toas "element duration" (ED). The duration of themicropattern, measured from the onset of the firstelement to the termination of the second element will becalled the "rnicropattern duration" (MPD). The timerelationship between Elements X and Y will be describedin terms of their "element onset asynchrony"(ON-ASYN) and "element offset asynchrony"(OFF-ASYN). In the experiments to be described, thealterations of ED, ON-ASYN, OFF-ASYN, and MPD onmicropattern discrimination were studied.

Figure I illustrates schematically the basic alterationsof these four temporal parameters that were employedin the present experiments. In the center of the figureare two micropatterns, each composed of twodiscriminably different stimulus elements (X and V). Forschematic convenience, Stimulus Element X is alwaysdrawn above the line and Stimulus Element Y is drawnbelow the line. The ON-ASYN and OFF-ASYN are bothzero. and the EDs and MPDs are equal. These twomic ropatterns are thus physically identical andperceptually indistinguishable. In each of the four typesof alteration (denoted by the four radial vectors), theasynchrony between the stimulus elements is increased.In Type I alterations, the ED is held constant and theON-ASYN and OFF-ASYN are increased equally. TheMPD must necessarily increase since the EDs are ofconstant duration. In Type II alterations, the MPD isheld constant and the ON·ASYN and OFF-ASYN areagain increased equally. In this case, the EDs mustnecessarily decrease in duration since the MPDs are ofconstant duration. In Type lIlA alterations, theOFF-ASYN is always zero and only the ON-ASYN isincreased. This necessarily results in an increase of theduration of one of the elements, and this will bereflected in an increase of the MPD. In Type IIIBalterations, the ON-ASYN is always zero and only theOFF -ASYN is increased. One ED and the MPD must alsoincrease, as was the case for Type IlIA alterations.

As the element asynchrony in each type of alterationis increased, the two micropatterns must becomediscriminable: Type I and Type II alterations will

519CONSERVATION OF TEMPORAL INFORMATION

UPI 1111

.',.,E "

,,.,, iliA

~)r

~hL~"'---.J'--_-___

10----- _MPO-- __~ON -A5TN ------..OfF-ASTN --. .~--~

~ ~ "'PI I

520 EFRON

Instrumental

Fig. 2. Representation of the four possible modes ofpresenting two micropatterns composed of identical stimuluselements, X and Y.

2nd MICRO'ATTERNht MICRO'ATTERN

A. CC?u II c:£O (DIFFERENT)

--0.5 Ie'-

I.~ II ~ (SAME)

C. c::ffl If ~(DIFFERENT)

Element duration (ED), element onset asynchrony(ON-ASYN), element offset asynchrony (OFF-ASYN),micropattern duration (MPD), the 0.5-sec interval betweenmicropatterns, and the 3.0-sec interval between trials werecontrolled with a digital logic system (lconix 6257-6010-6171),which was used for all experiments. Its accuracy was ±.Ol msec.

presented with no feedback. For every change in stimulusparameters, the Ss were again given eight training trials withpreknowledge of the correct reply, followed by 100 trialswithout feedback.

METHODS

ultimately result in discriminable micropatterns whenthe Hirsh-Sherrick threshold is exceeded. That is to say,the Ss will be able to distinguish different micropatternsby identifying the temporal order of the two elements:Type IlIA and Type I1IB alterations will also ultimatelybecome discriminable, since one of the elements of eachmicropattern will have a greater subjective intensity asits duration increases in relation to the other. It isimportant to note that the magnitude of the elementasynchronies employed in the present experiments werealways sufficiently small so that the micro patterns usedcould never be discriminated on the basis of perceivedtemporal order or perceived intensity differences.

For any set of values of ED, ON-ASYN, OFF-ASYN,and MPD illustrated in Fig. I, four modes ofpresentation of the two micropatterns are possible andall were employed in random presentation. In Fig. 2, thefour modes of presentation are illustrated only for thetwo micropatterns in the upper left corner of Fig. 1.

The nontemporal parameters of themicropatterns-Le., the parameters which madeElements X and Y discriminably different-were thedifference in frequency (~n between the two elements(in the auditory and vibratory experiments) and thedifference in wavelength (~A) between the two stimuluselements (in the visual experiments). The effect ofvariation of ~f and ~A on micropattern discriminationwas studied for selected values of the temporalparameters (ED, ON-ASYN, OFF·ASYN, and MPD).

Psychophysical Auditory Experiments

The method of forced choice ("same"-"different") was usedfor all determinations of Experiments I, II, IlIA, and IIIB. Foreach data point, 100 trials were used, 25 of each of the fourpossible modes of presentation illustrated in Fig. 2, which werepresented in pseudorandom order. These 100 trials werecompleted in a single session. The percentage of correct replieswas calculated for each 100 trials. (A method of adjustment wasused for Experiments IV and V and will be described later.)

The two micropatterns (separated by an interval of 0.5 sec)were presented every 3 sec, and a verbal response was requiredwithin that interval. In the visual experiments, the use of adental bite for head fixation precluded verbal responses and ahand signal was substituted.

Two highly trained psychophysical Ss were used for allexperiments. With one exception, all of the experiments werecompleted by the same two Ss, The exception was the singleexperiment in the vibratory modality which could not beperformed by 5 M.e., who had had no previous experience withvibratory stimuli. Another 5 (the author) was used for thisexperiment.

For each experimental session (which consisted of a single setof stimulus parameters), a brief training period was given. The Sswere presented with two trial exposures of each of the fourpossible modes of presentation of the two micropatterns (seeFig. 2). Prior to each of these eight training trials, the 5 wasgiven advance knowledge of the correct reply. The 8 was nottold the temporal order of Elements X and Y; he was informedonly that the two micropattems would be same or different.Following the eight training exposures, which were intended toestablish the criterion of judgment, the 100 test trials were

The audio signals were generated by two independentcontinuous sine-wave generators, the frequency of which wasalways monitored by separate counter timers. These signals werepassed through two audio gates (lconix 0137), which shaped therise and decay times of the stimulus elements. An exponentialrise and decay time of approximately 7 msec was used O/e =5.0 msec). The timing of gate opening and closing was controlledby the Iconix digital logic system, which was not synchronizedto the frequency of either of the two sine-wave generators. Theoutput from each audio gate was passed through an attenuator(Hewlett-Packard 350D), electrically added (Philbrick P-85AU),passed through a filter (Krohnhite 3500-R), with a bandpass of1,000-2,000 Hz, and finally sent to the S's right ear viaheadphones (Koss-Pro 4A) in an audiometric testing chamber(lAC 401-A). For the experiments of Type V (to be describedlater), which were performed only in the auditory modality, anadditional sine-wave generator was placed in the testing chamber.The output of this third sine-wave generator, which was adjustedby the S, was passed through a third audio gate and mixed withthe outputs of the other two oscillators. The third oscillatorproduced only one of the elements in the second micropattern(denoted by the symbol V in Fig. 6).

The steady-state amplitude of each stimulus element was1.0 V peak-to-peak, measured across the coil of the earphone.This voltage resulted in a sound level of 80 ± 0.5 dB (re .0002dynes/em"), measured at the surface of the earphones with a6-cc coupler (Bruel & Kjaer, Type 2603 amplifier, Type 4134microphone). The frequencies used were between 1,300 and1,940 Hz-the essentially flat portion of the auditory sensitivitycurve. In those experiments, in which stimulus frequencies were

CONSERVAnON OF TEMPORAL INFORMAnON 521

Audition

Visual Experimen ts

experiments of Type I (Fig. 1, upper left group ofschematics), where the ON-ASYN was always equal tothe OFF-ASYN. Since the data of both Ss showed nosignificant differences, the results were collapsed acrossSs. The ordinate is the percentage of correct replies (50%being the chance level); the abscissa is the difference infrequency of the two elements.

For the experiments reported in Fig. 3, five values ofelement asynchrony (2, 4, 6, 10, and 17 msec) wereused, and the results for each value of asynchrony aredenoted by a common symbol. The element durationswere 17 msec (including the 7-msec rise and decaytimes). Since the EDs were of equal duration, the MPDfor each of the above five values of element asynchronywere 19, 21, 23, 27, and 34 msec, respectively. Thesound levels of the two elements were equal (80 dBSPL). The frequency of Element X was always 1,300 Hz;Element Y was 1,320, 1,340, 1,380, 1,460, 1,620, or1,940 Hz in six different experiments. These values giverise to frequency differences (Af) of 20, 40, 80, 160,320, and 640 Hz, which are noted on the abscissa. Eachpoint in the graph is obtained from 200 trials (25 ofeach of the four combinations illustrated in Fig. 2 foreach of the two Ss). A horizontal line, drawn at the 75%level, defines the threshold value used in theseexperiments.

As can be seen in Fig. 3, the Ss were always abovethreshold for discriminating differences between the firstand second micropatterns when the difference infrequency between the two elements and the elementasynchrony was large, but fell below threshold when thedifference in frequency between the two elements wasmade smaller and when the element asynchrony wasreduced. A chance level of performance was found withthe smallest frequency differences and shortest elementasynchronies. For any specific Af, the Ss generally hadhigher discrimination performances for large-elementasynchronies, and their performance deteriorated as theelement asynchrony was decreased.

From these results, it was quite obvious that the SScould distinguish the difference between MicropatternsXY and YX, even with small differences in frequencybetween Elements X and Y. With larger differences infrequency, the Ss remained above threshold, even withelement asynchronies of 2 msec-a value which is farbelow the Hirsh threshold of 17 msec for perceivedtemporal order judgments in audition. In additionalType I experiments (not illustrated), the same two Sswere still always above threshold, with an ED of12 msec, an element asynchrony of 2 msec, a MPD of14 msec, and a Af of 5,000 Hz. Informal studies on 10other Ss revealed the same phenomena. From theseresults, it was concluded that temporal orderinformation is conserved, -even at relativelyshort-element asynchronies-despite the fact thattemporal order judgments could not be made. On whatperceptual cue. then. was the rnicropatterndiscrimination performed?

ON-.nN . O.. ·.UN

.UDITIONUPI J fXPUIMfN,

&---0 ASYIlii • """.ec...---.. An",. to''''MD-'-c ASYN '''',ee•. - .. ASYN· of "",ee..--. ASYN. 2.",.c

I6ilOiff.,e",. i ... ".,,".ftol_{6,I·

10

II,

100

outside this narrow range, the bandwidth of the filter wassuitably increased.

The identical apparatus was used, except that the output ofthe filter (bandwidth of 50-500 Hz) was fed to a power amplifier(Ling Model 25). which operated a vibrator (Ling Model 102).The rise and decay times of the stimulus elements was 10 msec.A probe, attached to the vibrator, was applied to the tip of theright third finger. The skin deformation was 0.04 in.peak-to-peak (see Efron, 1973, for details). The frequencies ofvibration were between 100 and 400 Hz.

Fig. 3. Discrimination of micropatterns, Experiment I(audition). For all mieropattems, the ON-ASYN was equal to theOFF-ASYN (see Type I alterations in Fig. I). Each set ofsymbols (0, 6, 3, ., .) represents a different elementasynchrony. Data of two Ss,

RESULTS

Figure 3 presents the data obtained from a series of

Experiment I

The visual stimuli were derived from two glow modulators(Sylvania RI131C), driven by the Iconix digital logic system.The light output from each glow modulator was passed throughneutral density filters and then narrow-band (9-mJ.L halfbandwidth) interference filters (Bausch & Lomb). The twomonochromatic light beams were combined. using a prism. andfell upon a disk of ground glass. The ground glass then served asa secondary source (of mixed light) which was focused on the S'scornea for a Maxwellian view. In other experiments, wide-bandWratten colored filters were used. The S's head was fixed by useof a dental bite, and he observed the visual stimuli through anaperture placed on the surface of the lens used to focus the lighton the cornea. The visual target was 1 deg 45 min of visual anglein diam and had a luminance of approximately 15 fl. The Sviewed this target in a darkened, but not lightproof, room.Sufficient ambient light was available to enable the S, after aperiod of dark adaptation, to see the aperture without a fixationlight.

_Vibratory Experiments

522 EFRON

Experiments lIlA and IIlB

deferred until after the results of Experiments lIlA andIIlB are presented,

Fig. 4. Discrimination of micropattems, Experiment II(audition). For all micropatterns, the ON-ASYN was equal to theOFF-ASYN (see Type II alterations in Fig. 1). Each set ofsymbols (0, 6, 8, ., -> represents a different elementasynchrony. Data of two Ss. Curves marked 0-0 in this figureare repeat studies of similarly marked curves in Fig. 3.

ON-ASYN .OfF-ASYN

AUOITIONTY" D IE X"IIMENT

o---c ASYN ·17 .....e~-. ASYH • to""••<0--.-0 ASYN • 6 ",aft

.-- .. .l5YN ... "nee

....-- ASYH • 2 "".e

160 320 I ',~o.ff.,.._. ift ',eqwncJ (~. I

Ill'

10

.ell

liCl40

70

I_ i!ttO~.. so

In Type I and Type II alterations, the micropatternshad equal element onset and offset asynchronies. InType III alterations, however, the effects of the elementonset and offset asynchrony were studied separately. InExperiment IlIA, the two elements (of bothmicropatterns) terminate synchronously, but commenceasynchronously (see Fig. 1, lower left schematics). InExperiment IlIB, the two elements commencesynchronously, but terminate asynchronously (seeFig. 1, lower right schematics).

The results of some of these experiments are listed inTable 1. The upper half of this table contains the resultsof five IlIA experiments with different stimulusparameters. The lower half of the table contains theparallel results obtained in each of the five IIIBexperiments with identical parameters. In all 10experiments, the two elements were tone bursts of 1,300and 1,500 Hz of equal amplitude (M::: 200). The datahas been collapsed across Ss as there were no significantdifferences.

The results (Column 1) indicate that the twomicropatterns cannot be discriminated when the elementasynchrony of 2 msec is at the onset or leading end ofthe micropattern (the values are below 75% thresholdlevel for Experiment lIlA), but are almost perfectlydiscriminated when there is a 2-msec offset asynchrony(Experiment IIlB).

Despite the fact that the same amount of temporalorder information (2.0 msec) is provided by the elementonset asynchrony (Experiment IlIA) as is provided bythe element offset asynchrony (Experiment IIIB), the

90

100All Ss observed that they were basing their reports("same" or "different") on an obvious difference in thedominant pitch of the two micropatterns. For example,if the Ss were asked to compare a rnicropattern in whichthe sequence was 1,300-1,940 Hz with a secondmicropattern in which the sequence was 1,940-1,300 Hz,they all reported that the first rnicropattern had a higherdominant pitch than the second. The Ss' replies thuscorrelated perfectly with the temporal order of thestimulus elements: The stimulus element in amicropattern which was presented last dominated theperceptual experience of that rnicropattern.

It should be stressed that the Ss had no knowledge ofthe temporal order of the elements or that theirperceptual experience of each micropattern wascorrelated with the temporal order of its elements. Whenthey were ultimately informed of the correlation (at theend of the series of experiments), the Ss could then inferthe temporal order of the stimulus elements. However,during the course of the experiments, this informationwas not available to them and this inference could nothave been made.

In addition to the use of micro patterns in which theelements were tone bursts of two different frequencies,Experiment I was also performed using micropatternsconsisting of a tone and a white noise burst andmicropatterns consisting of two pink noise bursts (ofdifferent bandwidths). Again, the Ss reported that theymade their discriminations on a readily perceptibledifference in the dominant pitch of the twomicropatterns-the spectral characteristics of the secondelement always dominating the pitch of thernicropattern.

Experiment II

Figure 4 presents the pooled data of both Ss from aseries of Type II alterations (Fig. 1, upper rightschematics), where the ON-ASYN was always equal tothe OFF-ASYN. The ordinate is the percentage ofcorrect replies; the abscissa is the difference in frequencybetween the two elements. The same values for thefrequency of the elements was employed as inExperiment I, and the same values of elementasynchrony (2,4, 6, 10, 17 msec) were also repeated. Inall of these experiments, the MPD was held constant at34 msec. Since the MPDs were of constant duration foreach of the above five values of element asynchrony, theEDs were 32, 30, 28,24, and 17 msec, respectively. Thesound levels of the two elements were equal (80 dBSPL). The top curve of Fig. 4 obtained with an ASYN of17 msec was identical with the stimulus parameters ofthe top curve of Fig. 3 (ASYN ::: 17) and is merely thesame experiment repeated several months later.

A comparison of Fig.4 with Fig. 3 reveals thatType II alterations resulted in a deterioration ofdiscriminatory performance for element asynchronies of2, 4, and 6 msec. The explanation of this result will be

CONSERVATION OF TEMPORAL INFORMATION 523

Table I

Column I Column 2 Column 3 Column 4 Column 5

Element Durations 17-19 msec 27-29 msec 57-59 msec 57-67 msec 17-27 msecElement Asynchrony 2.0 msec 2.0 msec 2.0 msec 10.0 msec 10.0 msecElement Onsets Asynchronous (Experiment lIlA) 61% 43% 51% 57% 80%Element Offsets Asynchronous (Experiment IIlB) 96% 85% 66% 100% 100%

perceptual system appears to make very little, if any, useof the former.

In the Type III alterations (unlike those of Types Iand II), the energy in each element was unequal. In thefirst pair of Type III alterations (Column 1), the elementdurations were 17 and 19 msec (including rise and decaytimes). The longer element would be expected (by virtueof temporal summation) to sound louder than theshorter element, and its pitch would be expected todominate the perception of that micropattern. Forexample, in the trials in which the l,sOO-Hz element was19 msec and the I,300-Hz element was 17 msec, the1,500-Hz element would be expected to be slightlylouder. When this micropattern was compared to one inwhich the I,300-Hz element was 19 msec and the1,500-Hz element was 17 msec, the small difference inloudness of the different elements might make the twomicropatterns discriminably different. However, this wasnot the case. In the lIlA experiments of Column 1, thisdifferential loudness factor was apparently not ofsufficient magnitude to be used as a discrimination cue.since the Ss' performance was near the chance level. Itcan be concluded, therefore, that the high level ofdiscriminatory performance in the parallel IIIBexperiment (Column I) could not be attributed to adifferential loudness factor, but must be attributed tothe element offset asynchrony per se. The results ofExperiment 1II thus indicate that the ON-ASYN of themicropatterns of Experiment I played little or no role inimproving the discrimination performance.

The second and third columns of Table 1 show theconsequences of increasing the duration of bothelements while keeping the ON-ASYN and OFF-ASYNat 2.0 msec. Again, Experiment IIlB yields a betterdiscriminatory performance than the equivalentExperiment IlIA. However, it is also quite clear thatincreasing the duration of the elements whilemaintaining a 2-msec offset asynchrony deteriorates theperformance in the IIIB experiment (Columns 2 and 3).This is exactly what occurred in Experiment II ascompared to Experiment I (compare Fig. 4 with Fig. 3),where performance deteriorated when the elementdurations were made longer (in Experiment II).

When the element durations are made longer in bothExperiment II and Experiment Ill, the 2-msecOFF·ASYN represents a smaller proportion of theoverall micropattern duration. This decreased ratio couldaccount for the decreased performance. In the fourthpair of experiments (Table 1, Column 4), the elementdurations were 57 and 67 msec (i.e., with a IO-msecasynchrony), reproducing the ratio used in the

experiments of Column 1. Both Ss obtained a 100% levelof discrimination in Experiment IlIB, but were still atchance level of performance for Experiment lIlA. Fromthe pattern of results in the first four pairs of theseexperiments, it appears that the trailing end of themicropattern becomes a less effective discrimination cueas the element durations are prolonged. Theeffectiveness can be restored, however, with an increaseof the element offset asynchrony.

The first four pairs of these experiments indicate thatthe pertinent information used by the S to discriminatemicropatterns lies at the end of the pattern (the elementoffset asynchrony), and that the element onsetasynchrony provides little, if any, information. However,a very large element onset asynchrony will also improvediscriminatory performance. In the fifth pair ofexperiments (Column 5), element durations of 17 and27 msec were used. In this case, the Ss achieved a higherlevel of performance in Experiment lIlA than wasobtained in Experiment lIlA of Column 1. Theimproved performance in this last experiment(Column 5) can be easily accounted fdr by the fact thatthe longer element had nearly twice the energy of theshorter element. With twice the energy, the longerelement would be expected to have a discriminableeffect on the dominant pitch of the micropattern (byvirtue of its increased loudness) and the micropatterndiscriminating performance should be improved. Despitethis improvement, the Ss achieved only an 80% correctdiscrimination in Experiment IlIA compared to a 100%performance in the equivalent Experiment lIIB(Column 5)-indicating once again that the elementoffset asynchrony provides the more salient cue formicropattern discrimination.

It should be noted that, while Hirsh's (1959)experiment on temporal order judgments employed theparadigm of Type IlIA alterations (see Fig. 1), heemployed element durations of 500 msec. His Ss were atthreshold with an ON-ASYN of 17-20 msec. With such alarge ratio of ED to ON-ASYN, the dominance effectdescribed in the present report is not observed, andwould not have interfered with his results or conclusions(as suggested by Broadbent and Ladefoged). It wouldappear that the ON·ASYNs (when of sufficientmagnitude) will enable the Ss to make temporal orderjudgments, but do not give rise. to the perceptualdominance effect.

It is difficult to explain the failure of the elementonset asynchrony (in the IlIA experiments) to give riseto any perceptual dominance of the leading (longer)element or to any other perceptual cues which

524 EFRON

-t2m"c---t

A~._~;t::,~y --._-- 0.5 I.e ---< Y

TISTMICRO'AnUN

STANDARD

MICRO'AnUN

Fig. S. Paradigm of Experiment IV (audition). The 8 is alwayspresented with two micropatterns in which the temporal orderof the elements is reversed, Line A represents the initial valuesset by E, Line B represents one step in the adjustment by the 8of the test micropattern. Line C represents the final step of theS's adjustment of the test rnicropattem (see Table 2 for results).

of Y, and X, were recorded by the E. The durations ofY, and X, were again made equal by the E, and the 5repeated the adjustment. This adjustment procedure wasrepeated 30 times. Two stages in the adjustmentprocedure which altered the relative durations of Y, andX, in the test micropattern are illustrated in Fig. 5, Band C.

The S altered the durations of Y, and X, in Ovl-msecsteps by turning a set of switches. The equipment wasprogrammed so that any increase in the duration of Ytwas exactly counterbalanced by a reduction in theduration of Xt. The sum of the durations of the twoelements in the test micropattern was thus constant andalways equal to the sum of the durations of the twoelements in the standard micropattern.

Table 2 contains the results of this experiment. At thenull point of the Ss' adjustment, the duration of the firstelement of the test micropattern was 15.25 msec; thesecond element was 8.75 rnsec. At these settings, the Ssfound the test and standard micropattern to beindiscriminable.

The results of the experiment thus indicate that thedominating effect of the second element of the testmicropattern could be reversed by a reduction in itsduration (energy) coupled with an increase in theduration (energy) of the first element.

When the Ss adjusted the relative durations of the twoelements of the test micropattern, a complex set of changesoccurred in other stimulus parameters as well. These additionalchanges came about as a consequence of the interaction of anumber of unavoidable instrumental, physical, andpsychophysical constraints which will be discussed briefly.

In order to prevent click transients from interfering with thepsychophysical matching task, the rise and decay times of theelements could not be instantaneous. The same rise and decaytime (7.0 msec) were used as were used in all previousexperiments. In order to keep the duration of each micropattern

facilitated discrimination between micropatterns. Apossible explanation may be a delay in an attentionswit ching mechanism suggested by Kristofferson(I967a, b) or a delay in a channel switching mechanismpostulated by Broadbent (1957). By the time thenervous system has "switched" to the micropattern, thesalient information (in a 2-msec ON-ASYN) may be lost.To test this possibility, each micro pattern was precededby a click-the intended purpose of which was to"switch" or "cue" the auditory system just before themicropattern was delivered. Despite a wide range of clickintensities and time relationships to the micropattern, noimprovement in discriminatory performance was notedin experiments of the IlIA paradigm by the introductionof a click. This negative finding does not, of course,completely exclude a channel switching explanation,since the "correct" parameters of the switching stimulusmay not have been employed. It does, however,undercut this explanation.

Experiment IV

In Experiments I, II, and III, the Ss indicated thatthey based their replies on the dominant pitch of themicropatterns. The second, or trailing, element of eachmicropattern dominated the perceptual experience ofthat micropattern. The object of Experiment IV was anattempt to reduce the dominating perceptual effect ofthe second element of a micropattern by a reduction ofits energy and to increase the perceptual effect of thefirst element by an increase of its energy-the totalenergy in the micropattern to be unchanged.

The S was presented with two micropatterns. Thefirst, or "standard" micropattern, consisted of theelement sequence X, - Y, (where the subscript denotesthat these elements are part of the standardmicropattern). X, was 1,500 Hz and Ys was 1,300 Hz,and their amplitudes and durations were equal-theformer being 80 dB SPL and the latter being 12.0 msec(including the 7.0-msec rise and decay time). The second(or "test") micropattern (presented 0.5 sec later) alwaysconsisted of the reverse temporal sequence (Y t - Xt ) ,

where Ytwas 1,300 Hz and x, was 1,500 Hz.This paradigm is illustrated by the two micropatterns

of Fig. SA. It will be noted that in Fig. 5, whichrepresents the temporal parameters of this auditoryexperiment, the rise and decay times of each element areillustrated. When presented with these twomicropatterns, the S could discriminate a difference atthe 100% level of performance. Using a method ofadjustment, he was then required to alter the ratio ofenergies of the two elements of the test micropattern~y adjusting their relative durations) until he could nolonger distinguish this micropattern from the standardmicropattern (which was never altered). The S waspermitted to examine as many pairs of micropatterns ashe wished until he was satisfied with his adjustment. Atthe completion of this task, the values of the durations

CONSERVAnON OF TEMPORAL lNFORMAnON 525

Table 2

MicropattemParameters

Standard Micropattern(Set by E)

Test Micropattern(Bold Face Values Set by S)

Element SequenceElement FrequencyElement Duration

Energy Ratio (Y :X)

Element X s1500 Hz

12.0 msec

1:1

Element Y s1300 Hz

12.0 msec

Element Y t1300 Hz

15.27 msec15.24 msee

Element X t1500 Hz

8.73 msee ± 0.13 (MBS)8.76 msec ± 0.36 (MC)

14.6:1

below the Hirsh-Sherrick threshold of 20 msec, the duration ofthe stimulus elements had to be brief. For the standardmicropattern, this was 12 msec. Finally, the time constants ofthe audio gates operated in such a fashion that elements shorterthan the time constant were markedly attenuated with respect toamplitude and became asymmetrical with respect to rise anddecay times.

Figure 5A illustrates that the exponential rise of each element(reaching lie of the 1.0-V signal in 5.0 msec) was permitted tolast for only 5.0 msec. The audio gate then was instructed toclose, which initiated the exponential decay of the tone burst,which required 7.0 msec for completion. The duration of therust element was thus 12 msec. The second element wasprogrammed to begin 5.0 msec after the onset of the ftrstelement-vthat is, at the same instant the rust element wasbeginning to decay. The rise of the second element was alsoterminated after 5.0 msec and took 7.0 msec to decay fully.Thus, both elements were of the same duration (12 msec) andthe same peak amplitude. Figure SA also illustrates the initialparameters of the test micropattern also set by the E. They areidentical with those of the standard micropattern except for thetemporal order of the two frequencies.

Figure SB illustrates one step of the adjustment proceduremade by the S on the test micropattern. The S merely altered aswitch which increased the duration of the rust element by2.0 msec and decreased the duration of the second element bythe same amount. However, this change in duration had thefollowing consequences, which can be seen in the tracing: Therise of the first element was permitted to last for 7.0 msec, thusincreasing its amplitude slightly. Since the decay time wasunaffected and remained at 7.0 msec, the duration of the firstelement was 14.0 msec. However, the rise of the second element(which was always initiated at the same moment the firstelement began to decay) was permitted to continue for only3.0 msec. This diminished its peak amplitude. Since its decaytime was unaffected, its duration was 10.0 msec. It will be notedthat this reduction in duration of the second element, with itsassociated decrease in amplitude. actually resulted in a moremarked reduction of its energy than would be the case if only itsduration had been reduced.

Figure 5C illustrates the effects of an additional 1.25-msecincrease in the duration of the rust element and an equaldecrease in the duration of the second element of the testmicropattern. The increase in duration of the rust elementnecessarily increased its amplitude slightly (because of theexponential rise-time characteristics of the audio gate). Theparallel decrease in the duration of the second element by1.25 msec cut short its rise time and thus markedly affected itsamplitude. It is clear that a reduction of the duration of thesecond element to 8.75 msec from 12.0 msec has a markedeffect on its am plitude, as well as the symmetry of rise anddecay times. Figure 5C represents the actual shape of the twostimulus elements at a time when the two Ss reported that thetwo micropatterns of Fig. 5C were indistinguishable. Theduration of the rust element of the test micropattern is15.25 msec, and the duration of the second element is8.75 msec. To ascertain the ratio of energies in the two elementsat the null point, the element wave forms of Fig. 5C weredigitized, squared. and integrated, using a computer of averagetransients. At the experimentally determined null point. the

ratio of durations of the rust to the second element was 1.75: 1.However, the ratio of energies was 14.6: 1 (see Table 2). Thisratio of energies conveys a more meaningful measure of thedegree to which the second element dominates the perception ofa micropattem than does the ratio of durations.

Experiment V

Experiment V was intended to determine in moredetail the nature of the perceptual interactions betweenthe two elements of a micropattem. The S was presentedwith a standard micropattern in which the EDs wereboth 12 msec and of equal amplitude (80 dB SPL). Thefrequencies of the two elements of the standardmicropattern were 1,500 and 1,300 Hz (see Fig. 6,Line A). A half-second later, the S was presented with atest micropattern in which one of the elements (denotedby the letter V) had a different frequency from its"mate" in the standard rnicropattern. The S wasrequired to adjust the frequency of this variable element(V) until the two micropatterns were perceptuallyindistinguishable. The S was permitted to deliver tohimself as many pairs of micropatterns as he desireduntil he was satisfied with his adjustment. The frequencyof the variable element (V) was recorded and anothertrial was instituted. Twenty null measures were made foreach of 10 conditions, illustrated in Fig. 6, A·E andA'·E'. In Conditions E and E', the S simply made afrequency match for an isolated element. The results ofExperiment V on the two Ss are found in Table 3, wherethe value of (V) is the mean of 30 trials.

A comparison of Table 3 with Fig. 6 reveals that thedifferential limen (dl) for frequency is more than

Table 3

Mean MeanFre- Fre-

Con- quency Con- quencydition (V) SD <lioon (V) SD

SM.C.A 1513 ±17.6 A' 1320 ±24.3B 1498 ± 9.9 B' 1307 ±12.0C 1517 ±20.8 C 1312 ±3S.8D 1501 ± 8.4 D' 1306 ±13.5E 1519 ±10.4 E' 1304 ±11.8

S M.B.S.A 1519 ±14.5 A' 1307 ±27.6B 1501 ± 3.8 B' 1304 ± 3.5C 1508 ±17.4 C' 1327 ±24.4D 1510 ± 6.2 D' 1301 ± 4.9E 1500 ± 5.1 E' 1304 ± 2.7

526 EFRON

The experiments in the visual modality were

Vision

The experiments in the vibratory modality wereessentially identical with those in the auditory modality.The stimulus parameters for these experiments areillustrated in the upper-left schematic of Fig. 1. Themicropatterns were composed of stimulus elementshaving different frequencies. The frequency ofElement X was always 100 Hz; Element Y was 400, 300,or 150 Hz. These values give rise to t:.f values of 300,200, and 50 Hz. The two micropatterns were presentedin the same four combinations indicated in Fig. 2, andthe Ss were required to report if the two micropatternswere the same or different. The element durations were20 msec, which included a lO-msec rise and decay time.(The use of faster rise and decay times producedperceptual transients which made the experimentsvirtually impossible to perform.) The ON-ASYN andOFF-ASYN were both 20 msec.

The Ss reported much more difficulty in making thediscriminations in the vibratory modality than they didin the auditory modality, and one of the two Ss wasunable to perform at above chance levels with theseshort vibratory pulses. Although good performancecould be achieved by this S with longer elementdurations, the need to keep the element asynchroniesunequivocally below the threshold for explicit temporalorder judgments resulted in the exclusion of this S. Theauthor served as the second S.

The pooled results on both Ss are displayed in Fig. 7.The ordinate is the percentage of correct discriminations(out of 200 trials) and the abscissa is the t:.f. As was thecase in the auditory modality, a reduction in thefrequency difference between the two elementsdegraded the performance.

Both Ss based their replies on the perceived differencein dominant frequency of vibration. As was the case inthe auditory modality, the perceived dominantfrequency of each micropattern was the frequency ofthe second element. Given the subjective difficulty inperforming these judgments in the vibratory modality,further experiments such as those performed in theauditory modality were not pursued. However, theresults which were obtained with these Type Ialterations closely resembled those found in the auditorymodality.

Vibration

The results of Experiment V indicate that theperceptual acuity of the first element (as measured bythe ell) is diminished. This finding, when coupled withthe results of Experiment IV, suggests that thedegradation of the perception of the first elementinvolves not only an effective intensity attenuation but adespecification of its frequency as well.

u

I. ~~

'1 ~~L \.Y

C.

~ 11. *'1500

D. ~ 11 ~tsoc

E. II\!:o/ ""!7

-12_-

~1500

rI ~ ----;r-.---1'5~OO-V~_.. ~----

ST"'NDARD TEST

MICROP...TTERN MICROPATTERN

.... ~ --~~~300~~ .. v

""'J ~.-- O.s..., -

,QE. ....11300 '"- . _'_

Fig. 6. Paradigm of Experiment V (audition). Tenexperiments, A-E, A' -E', are illustrated. In each case, the S isrequired to match the text micropattern to a standardmicropattem by adjusting the frequency of Stimulus Element Vin the test micropattern. Experiments E and E' are controlexperiments using only single elements of 1,500 and 1,300 Hz,respectively (see Table 3 for results).

doubled in Conditions A, A', C, and C' for both Sscompared to their ell in Conditions B, B', D, D', E, andE'. There is no intra-S significant difference in the dl ofthese last six conditions. In contrast, in Conditions A,A', C. and C' (where the dl was markedly elevated), thevariable element was the first element in the testrnicropattern.

The increase in the dl for the first, but not for thesecond element, in the test micropattern indicates thatthe perceptual experience of the first element isdegraded by the second but that the reverse is not thecase. The dl of the second element (Conditions B, B', D,and D') did not differ from the dl measured with thatelement in isolation (Conditions E and E').

CONSERVAnON OF TEMPORAL INFORMA nON 527

so 100 00 300Diff.rence in ,,.eque,,,c..,! 1'. Hz \

Fig. 7. Discrimination of micropatterns (vibration). Data oftwo Ss (see text for parametric details),

70

- 60v

•::0

u~

50

40

90

80

100

perceived a yellow flash and were unable to determinethe temporal order of the red and green elements. Theybased their judgments on the difference in hue (or lackthereof) between the two micropatterns. The dominanthue reported by the S for each micropattern againcorrelated perfectly with the temporal order of theelements in that micropattern.

Additional studies in the visual modality wereperformed using wide-band colored filters. Experimentsusing 20- or 10-msec element durations passed through aWratten No. 29 (red) and a Wratten No. 58 (green) filtershowed the same dominance effect. To demonstrate thatthe phenomenon was not unique to the red-greencomponent-color system, the same experiments wererepeated using Wratten No.47 (blue) and WrattenNo. 15 (yellow) filters, In this last case, eachmicropattern gave rise to a white flash. However, asexpected, the sequence blue-yellow produced ayellowish-white flash and the sequence yellow-blue gaverise to a bluish-white flash-hue differences whichpermitted the micropattems to be readily discriminated.

Experiments identical to Experiments lIlA and IIIB inthe auditory modality were also performed in the visualmodality. Once again, the element offset asynchronyproved to be the effective parameter in permittingmicropattern discrimination. Type III alterations wereperformed with red-green as well as with blue-yellowmixtures, and gave the same results.

The adjustment experiments (Experiments N and V)

5 10 15 20Element onset osynchrony (msec)

Fig. 8. Discrimination of rnicropatterns (vision). Data of twoSs (see text for parametric details).

'0

90

10

70

~60

•v.50

.0

essentially identical to those previously described inaudition and the vibratory sense.

In these experiments, each micropattern consisted oftwo elements of different wavelength-620 mn (red) and510 m,u (green). The ED was 20 msec for each element,and the rise and decay times were less than 0.4 msec.The Ss were required to report if the micropatterns werethe same or different. The four types of temporalalterations illustrated in Fig. 1 and the same set of fourcombinations (Fig. 2) were used as in the auditoryexperiments. Identical psychophysical procedures wereemployed.

When exposed to a rapid sequence of these twoelements, all Ss reported perceiving a yellow flash.However, they could readily discriminate between twomicropatterns (in which temporal order of elements wasreversed) and reported that they performed thisdiscrimination by using a hue difference. All Ss reportedthat the appearance of the red-green sequence wasslightly greenish-yellow, while the appearance of thegreen-red sequence was slightly orange-yellow.

As the element asynchrony was decreased (in Type Ialterations), discriminatory performance deteriorated(see Fig. 8). The experiment was then repeated with twoelements having wavelengths of 600 and 528 mn. Sincethe results did not differ (at the 0.05 level), the data ofboth experiments were pooled. Figure 8 shows thepooled results for both AAs. The ordinate is thepercentage of correct discriminations. The abscissa is theelement asynchrony. The experiment was attempted forthe third time with elements having wavelengths of 570and 545 mil. Both Ss found it impossible to discriminatebetween the micropatterns and were at a chance level ofperformance at all values of element asynchrony.

The results of these experiments parallel those inaudition and vibration (where the elements had differentfrequencies). With the element parameters used, the Ss

5~R EFRON

were not performed in the visual modality due toeq uipment limitation.

DISCUSSION

The results of these experiments demonstrate thattemporal order information is conserved by perceptualsystems even when the asynchrony between the stimuluselements is below the threshold for temporal orderjudgments. This conservation of temporal orderinformation is achieved by a transformation into anotherperceptual dimension or quality-one in which there isno subjective experience of time. This is not a newobservation. It is well known. for example, that clicksdelivered via earphones to the right and left ears with aninterstimulus interval of a few hundred microsecondsgive rise toa single perceptual experience, which isIateralized (in the head) toward the ear receiving theleading stimulus. In the visual modality, thephenomenon of apparent motion is observed when tworapidly successive stimuli are presented to spatiallyseparated retinal loci. In both cases, the interval betweenthe two stimuli is too short to allow for any judgment oftemporal order-the temporal information is conservedby means ofa transformation into another (spatial)percept ual dimension. In the present experiments, thetemporal information is transformed into the perceptualdimensions of "pitch" (in both the auditory andvibrotactile senses) and hue (in the visual modality).

TIle experimental results reveal the existence ofanother aspect of this process of conservation oftemporal order information-that it is accomplished atthe expense of a degradation of the perceptual acuity ofthe first of the two rapidly sequential stimuli. Thedegradation of the perceptual experience of the firststimulus element. with respect to its subjective intensity(Experiment IV) and with respect to its dl(Experiment V). is associated with the subjectivedominance of the second element of the micropattern.In contrast. the perceptual experience of the secondelement appears to be unaffected by the presence of thefirst. The discrimination of a difference betweenmicropatterns thus depends on a retroactive effect of thesecond stimulus element on the first within eachmicropattern. What is the nature of this retroactiveeffect')

TIle two most general concepts that deal with theperceptual degradation of the first of two rapidlysuccessive stimuli are (a) the concept of (retroactive)masking demonstrated in all perceptual modalities and(b) the concept of metacontrast, which has beenconsidered, heretofore. to be a property of the visualsystem. The relation of each of these concepts to thepresent findings will be discussed in the following twosections. The third section of the discussion will beconcerned with some recent experiments which mightprovide an alternative approach. The final section willdescribe some experiments in the auditory modality by

other investigators which relate to the phenomenon ofrnicropa ttern discrimina tion.

Masking

There are two aspects of the results that stronglyundercut any explanation in terms of retroactivemasking: The first relates to the energy parameters andthe second to the frequency parameters of themicropattern .

In the first place, the phenomenon of retroactivemasking is usually found only when the energy and/orduration of the masking stimulus is much larger thanthat of the target or probe stimulus. In contrast, theperceptual degradation of the first stimulus element ofthe micropattern is readily demonstrated with twostimulus clements' of equal energy (Type I, II, and Vexperiments) or nearly equal energy (Type IIIexperiments). The energy parameters of discriminablemicropatterns are th us inconsistent with those whichgenerate retroactive masking effects.

Secondly, the phenomenon of masking is mostmarked (in all modalities) when-the masking stimulusactivates the same group of receptors as the probe (first)stimulus. [In auditory theory, the concept of criticalbands deals quantitatively with the observation thatmasking is most marked if the first stimulus is centerednear the middle of the bandwidth of the maskingstimulus. The amount of masking decreases markedly asthe frequency of the probe stimulus is moved outsidethe bandwidth of the masking stimulus (Hornick et al,1969; Elliott, 1962, 1969)]. In contrast, the presentexperiments show unequivocally that the interactionbetween the two stimulus elements of each micropatternis more marked (and discrimination performanceconsequently improved). when the frequency differencebetween the clements is increased, i.e., when the criticalbands of the two elements are least likely to overlap. Inan extreme test of this argument, Ss were presented withmicropatterns consisting of stimulus elements of 1,000and 6,000 Hz. Such stimuli are so far removed from eachother's critical bands that retroactive masking of the firststimulus element would be negligible, the perceptualdominance of the trailing element would disappear, andthe micropatterns should be nondiscriminable. This wasnot the case: The perceptual dominance of the trailingstimulus element was powerful, and perfectmicropattern discrimination was achieved.

The same considerations pertaining to energy andwavelength parameters are applicable to the visualmodality, and the results are once again inconsistentwith any explanation in terms of retroactive masking.

Metacontrast

Unlike the phenomenon of masking, which istypically produced by two rapidly successive stimuli ofmarkedly unequal energy delivered to the same receptor

CONSERVATION OF TEMPORAL INFORMATION 529

elements, the phenomenon of metacontrast is typicallygenerated by two successive stimuli of equal energywhich fallon spatially separated populations ofreceptors (see Lefton, 1973, for a recent review ofparametric issues). In the rnicropattern experiments, thestimulus elements also activated two spatially separatedpopulations of receptors. In the auditory modality, thetwo tone elements (even when their frequencies werenot markedly different) undoubtedly stimulatedspatially separate, but overlapping, populations offrequency-specific elements on the basilar membrane. Inthe visual modality, the two colored flashes (althoughfalling on the same region of the retina) surelydischarged overlapping populations ofwavelength-specific cells. In the vibratory modality, thetraveling wave in the skin resulting from two stimuluselements of different frequency probably stimulatedspatially overlapping populations of receptor elements.The activation of different populations of receptors bythe stimulus elements is thus analogous to the spatiallyseparated stimuli which give rise to visual metacontrasteffects.

In contrast to the similarity of spatial factors, thetemporal parameters of the stimuli that give rise tometacontrast effects are quite unlike the temporalparameters of the rnicropattems. In those metacontrastparadigms employing two stimuli of equal or nearlyequal energy, the maximum metacontrast effect occurswhen the second stimulus follows the first by a delay ofapproximately 100 msec. A severe attenuation and/ortotal loss of the metacontrast effect occurs when theinterstimulus interval is appreciably shorter or longerthan this value-the much-discussed U-shaped function.Although the present experiments did demonstrate amore marked perceptual interaction between thestimulus elements as the element asynchrony increased,Ss could discriminate auditory micropatterns reliably(>75%) even when the element asynchrony was only2 msec (see Figs. 3, 4, and Table 1). In the visualmodality, the interaction effect was often detectablewith element asynchronies below 5 msec. While thesevalues are far shorter than those described in anymetacontrast experiment, they are not necessarilyinconsistent with such an explanation.

Iconic Storage-Processing Time

Since the work of Sperling (1963) on short-termstorage effects in vision, a number of investigators haveadvanced a variety of models-the goal of which hasbeen to integrate such time-dependent perceptualprocesses as short-term (iconic) storage, masking,metacontrast, and apparent movement (Haber &Standing, 1969; Kahneman, 1967; Liss, 1968; Massaro,1972; Efron, 1973). The central feature of these modelsis the idea that the onset of the second of two rapidlysuccessive stimuli terminates the central processing ofthe information contained in the first stimulus. This

premature termination of the processing of the firststimulus reduces the amount of information that can beextracted from it and thus degrades the perceptualacuity. It should be noted that these models have beengenerated to deal with experimental paradigms in whichthe two stimuli are immediately successive or followeach other with a short interstimulus interval and havenot been applied to stimuli that temporally overlap.

In those micropatterns in which the two stimuluselements are consecutive (see Type I and Type IIalterations of Fig. 1), these models would predict thatthe onset of the trailing stimulus element would arrestthe central processing of the initial stimulus element.The consequent perceptual degradation of the firstelement in each micropattem would make the twornicropatterns discriminable when the temporal order ofelements is reversed. This explanation is consistent witha number of experiments performed in the visual (Haber& Standing, 1969) and auditory (Massaro, 1972)modalities using successive stimuli without temporaloverlap. For the stimuli of Experiment InA (see Fig. 1),the argument could possibly be extended by assumingthat the onset of the second stimulus element interferedwith the central processing of the initial (leading) part ofthe first element. This process-which would degradeonly that information which makes the micropatternsdistinguishable-might account for the Ss' inability toutilize the element onset asynchrony information thatwas characteristic of the Type IlIA experiments (seeTable 1). However, it is not clear how this argumentcould be extended further to deal with the results of theType II1B experiments, where the element onsets aresimultaneous and where there is no salient signal toterminate the processing of the first element. Even moredifficult to explain by these models is the importanteffect of the ratio of element offset asynchrony toelement duration, which was characteristic of theType II1Bexperiments (see Table 1).

Related Phenomena in Auditory Modality

Several recent experiments in the auditory modalitywhich may bear a close relation to the phenomenon ofmicropattern discrimination warrant particularattention.

Patterson and Green (1970), using Huffman sequences(a pair of brief wave forms which differ only in phasespectra) have shown that such stimuli are discriminable.Although Ss report merely hearing two clicks, theseclicks can be discriminated by virtue of a characteristicpitch difference. They concluded that "the ear issensitive to small differences in the arrival time of energyat different frequencies and that these differences allowthe observer to discriminate among the differentHuffman sequences." The Patterson and Greenexperiment showed that temporal asynchronies as smallas 1.5-2.0 msec can be utilized by the auditory

530 EFRON

system-values almost identical to those obtained withmicropatterns.

Nabelek and Hirsh (I969) and Nabelek, Nabelek, andHirsh (1970), using brief tone bursts of increasing ordecreasing frequency, have shown that the terminalfrequency of the tone burst is more significant for pitchdiscrimination than the initial frequency. This finding soclosely resembles the phenomenon of perceptualdominance of the trailing element of micropatternscomposed of two elements that it prompted furtherinformal experiments using two micropatterns, eachconsisting of three 5-msec stimulus elements of differentfrequency. Ss could discriminate easily between twosuch micropatterns when the second and third element(or the first and third) had reversed temporal orders-afinding completely consistent with the Nabelek,Nabelek, and Hirsh studies using stimuli of continuouslychanging frequency. A similar result was also obtained inthe visual modality using micropatterns with threedifferent hues.

The existence of virtually identical psychophysicalcharacteristics involving micropattern discrimination inthree modalities presents an unusual problem to besolved. On the one hand, an explanation is requiredwhich is sufficiently general to be equally applicable toall three modalities, which have strikingly differentreceptor organs, central neural connections, andfunctional properties. On the other hand, an explanationis required which is sufficiently specific to accountquantitatively for the phenomenon in each modality(considered in isolation) in terms of its knownpsychophysical and neurophysiological characteristics.

REFERENCES

Broadbent, D. E. Mechanical model for human attention andimmediate memory. Psychological Review, 1957, 64,205-215.

Broadbent, D. E., & Ladefoged, P. Auditory perception oftemporal order. Journal of the Acoustical Society of America,1959,31, 1539.

Edwards, A. E., & Auger, R. The effect of aphasia on theperception of precedence. Procedings of the AmericanPsychological Association, 1965, 207-208.

Efron, R. Temporal perception, aphasia and deja vu. Brain,1963,86,403-424.

Efron, R. An invariant characteristic of perceptual systems in thetime domain. In S. Kornblum (Ed.), Attention andperformance IV. New York: Academic Press, 1973.

Elliott, L. L. Backward and forward masking of probe tones ofdifferent frequency. Journal of the Acoustical Society ofAmerica, 1962,34,1116-1117.

Elliott, L. L. Masking of tones before, during and after briefsilent periods in noise. Journal of the Acoustical Society ofAmerica, 1968,45, 1277-1279.

Gengel, R. W., & Hirsh, I. 1. Temporal order: The effect of singleversus repeated presentations, practice and verbal feedback.Perception & Psychophysics, 1970, 7, 209-211.

Haber, R. N., & Standing, L. G. Direct measures of short-termvisual storage. Quarterly Journal of Experimental Psychology,1969,21,43-54.

Hirsh, I. J. Auditory perception of temporal order. Journal ofthe Acoustical Society of America, 1959,31,759-767.

Hirsh, I. J., & Fraisse, P. Simultaneite et succession de stimuliMterogenes. Annee de Psychologie, 1964,64, 1-19.

Hirsh, I. 1., & Sherrick, C. E. Perceived order in different sensemodalities. Journal of Experimental Psychology, 1961, 62,423-432.

Holmes, H. Disordered perception of auditory sequences inaphasia. Unpublished PhD thesis, Harvard University, 1965.

Hornick, J. L., Elfner, L. F., & Bothe, G. G. Auditory temporalmasking and the perception of order. Journal of theAcoustical Society of America, 1969,45,712-718.

Jerger, J., Weikers, N. J., Sharbrough, F. W., & Jerger, S.Bilateral lesions of the temporal lobe. A case study. ActaOtolaryngologica, 1969, Supplement 258.

Kahneman, D. An onset-onset law for one case of apparentmotion and metacontrast. Perception & Psychophysics, 1967,2,577-584.

Kristofferson, A. B. Successiveness discrimination as a two-state,quantal process. Science, 1967a, 153, 1337-1339.

Kristofferson, A. B. Attention and psychophysical time. ActaPsychologica, 1967b, 27, 93-100.

Lefton, L. A. Metacontrast: A review. Perception &Psychophysics, 1973, 13, 161-171.

Liss, P. Does backward masking by visual noise stop stimulusprocessing? Perception & Psychophysics, 1968,4,328-330.

Lowe, A. D., & Campbell, R. A. Temporal discrimination inaphasoid and normal children. Journal of Speech & HearingResearch, 1965,8,313-314.

Massaro, D. W. Stimulus information vs processing time inauditory pattern recognition. Perception & Psychophysics,1972, 12, 50-56.

NabeIek, I., & Hirsh, I. J. On the discrimination of frequencytransitions. Journal of the Acoustical Society of America,1969,45,1510-1519.

Nabelek, I. Y., Nabelek, A. K., & Hirsh, I. J. Pitch of tone burstsof changing frequency. Journal of the Acoustical Society ofAmerica, 1970,48,536-553.

Patterson, J. H., & Green, D. M. Discrimination of transientsignals having identical energy spectra. Journal of theAcoustical Society of America, 1970,48, 894-905.

Sperling, G. A model for visual memory tasks. Human Factors,1963,5, 19-31.

Swisher, L., & Hirsh, I. J. Brain damage and the ordering of twotemporally successive stimuli. Neuropsychologia, 1972, 10,137-152.

(Received for publication March 16, 1973;revision received June 7,1973.)