Memory & Cognition 1975, Vol. 3 (3), 287 294 What auses the spaing effet? Some effets ofrepetition, duration, and spaing on memory for pitures DOUGLAS 1. HNTZMAN, JEFFERY J. SUMMERS, and RCHARD A. BLOCK University oforegon, Eugene, Oregon 97403 Three experiments examined effets of the spaing of repetitions on memory for pitures. n Experiment, the duration of the first presentation (P, ) was manipulated, as was P.p. spaing. The effet of spaing on judged frequeny was independent of P duration. n Experiment, pitures were given M massed presentations just prior to the P M,P M + 1 spaing interval. The form of the spaing urve was independent of M. Neither experiment onfirmed the predition of "overhabituation," derived from the habituation-reovery explanation of the spaing effet. n Experiment, subjets made both duration and frequeny judgments. The duration judgment results were not onsistent with the notion that subjets remember multiple massed presentations as single ourrenes of espeially long duration. Some evidene from Experiments and suggests that an interrupted stimulus is reognized better than one that is not interrupted-a finding that, if repliable, would support the habituation-reovery aount of the spaing effet. Why do two suessive presentations of a to-be-remembered.stimulus lead to poorer long-term retention than two presentations that are spaed further apart? A number of explanations of this puzzling finding have been proposed (L Hintzman, 1974). This paper reports experiments that were done to test two explanations of the spaing effet. One of these explanations has been proposed before, and the other apparently is new. Briefly, the spaing effet an be haraterized as follows: Let us all the first and seond presentations of an item P and P 2, respetively. As the P -P 2 interval is lengthened, long-term retention of the item inreases, approahing an asymptote at a spaing interval of approximately 15 se. The effet has been obtained in memory for words, nonsense syllables, sentenes, and pitures; in tasks as diverse as paired assoiates, free reall, reognition memory, verbal disrimination, and the distrator task; using measures of reall and reognition auray, reognition lateny, and judged frequeny. When pitures are used as stimuli, empty and filled P -P 2 intervals are found to have about the same effet, suggesting that the proess underlying the spaing effet is primarily a time-dependent one (Hintzman & Rogers, 1973). And the evidene suggests that it is defiient retention of P 2 rather than P that produes the poor performane when the P -P 2 interval is short (Bjork & Allen, 1970; Hintzman, Blok, & Summers, 1973). The present experiments follow those of Hintzman and Rogers (1973) in using seni pitures as This researh was supported by a researh grant from the Offie of Eduation, United States Department of Health, Eduation, and Welfare. and by Grant GB-40360 from the National Siene Foundation. Requests for reprints should be sent to Douglas L. Hintzman. Psyhology Department. University of Oregon, Eugene. Oregon 97403. stimuli and judged frequeny as the primary dependent variable. Complex senes have several advantages for experiments suh as the present ones: They are easily remembered, they maintain subjet's interest through long presentation sequenes, and they are apparently diffiult or impossible to rehearse (Cohen, 1973; Shaffer & Shiffrin, 1972). This last property makes them espeially useful in investigating the spaing effet, beause it gives the experimenter greater ontrol over temporal variables than an be obtained when verbal materials are used. HABTUATON RECOVERY HYPOTHESS The first explanation of the spaing effet onsidered here is the habituation-reovery hypothesis (Hintzman, 1974). Aording to this hypothesis, whenever a stimulus item is studied some internal proess involved in storing a memory trae of that stimulus beomes adapted-that is, its threshold of ativation is raised. This adaptation or habituation ontinues for as long as attention is direted to the stimulus. When the stimulus eases or attention is direted elsewhere, reovery from habituation begins. f P 2 ours before reovery from the habituation to P is omplete, enoding ofp 2 will be less effetive than if P 2 is delayed. Thus the spaing effet ours beause immediately after a trae of P has been enoded the pathway involved is in a refratory state. n order to be maximally effetive, P 2 must our outside the refratory period. Viewed in this way, the spaing urve is seen as traing out the time-ourse of reovery from habituation. f the spaing effet is due to habituation, then it may be possible to affet the time ourse of reovery by produing different degrees of habituation prior to the 287
288 HNTZMAN, SUMMERS, AND BLOCK spaing interval (f. "overhabituation" or "below-zero habituation" in the animal learning literature). n Experiment, this was attempted by manipulating the exposure duration of P and varying the P -P 2 spaing interval, n Experiment, either one, two, or three massed presentations, followed by a varied spaing interval, preeded the final presentation of the piture. n both experiments it was predited that the greater was the exposure of the stimulus just prior to the. spaing interval, the slower would be the reovery traed out by the spaing urve. Thus a partiular form of interation was predited in whih the rate at whih the spaing urve approahes asymptote is inversely related to the amount of exposure to the stimulus immediately preeding the spaing interval. EXPERMENT Method Materials and design. The experimental items were 144 olor pitures of a variety of vaation senes. Eah piture was randomly assigned to 1 of 22 experimental onditions. The 18 pitures in one ondition did not our during the presentation series, but served as distrator items on the frequeny-judgment test. Another 18 pitures ourred one time during the presentation series, 6 at eahof three different levels of exposure duration. The remaining 108 pitures ourred two times eah. Six were assigned to eah of 18 onditions representing the orthogonal ombination of three levels of P duration and six levels of P -P, spaing. For both single and double-presentation pitures, the three levels of duration were set at 3, 6, or 9 se, onset-to-onset time. Sine the time required for slide hanges was 0.8 se, the atual exposure durations were 2.2, 5.2, and 8.2 se. For double-presentation pitures, the six levels of spaing interval between presentations were 0.8, 3.8, 6.8, 9.8, 12.8, and 18.8 se from offset of P to onset of P,. The P, duration of repeated pitures was always 5.2 se (onset to offset). The presentation series was divided into six bloks, and eah of the 21 onditions (distrator ondition not inluded) was represented by one piture in eah blok. The within-blok order of onditions was random, subjet to limitations imposed by spaing requirements. n eah blok, an additional two pitures, drawn from the same pool, ourred three times eah. These pitures, whih were not tested, were repeated at spaing intervals whih varied. The first 10 and last 10 slides in the entire series were also filler items, drawn from the same pool, and some of these were repetitions. Altogether, there were 290 slides (inluding repetitions and fillers) in the presentation series. The frequeny-judgment test series, whih was omposed of one slide of eah of the 144 experimental pitures, was divided into six bloks of 24 pitures. Eah experimental ondition was represented by one piture in eah blok, exept for the distrator ondition, whih was represented by three in eah blok. The within-blok order of onditions was random, and all bloks of the presentation series were represented about equally in eah blok. The test sheet for frequeny judgments onsisted of 144 onseutively numbered blank spaes. Subjets and proedure. The subjets were 63 paid volunteers obtained through an advertisement in the University of Oregon ampus newspaper. They were tested in six sessions of 9 to 12 subjets eah. Aross sessions, pitures were partially rotated through onditions, so that eah piture ourred twie at eah level of duration of the first (or only) presentation, and double-presentation pitures ourred one at eah level of P, -P, spaing. At the start of eah session, subjets were told that a series of 290 slides would be shown, that the series would take 29 min, and that the slides would appear for varying amounts of time. They were told that many of the pitures would be repeated, and that they were simply to study eah slide for as long as it appeared and try to remember it for a later test, the nature of whih was not speified. The slide series was then presented on a wall of the experimental room by a. Carousel projetor ontrolled by a tape reorder and Kodak sound synhronizer. Presentation was paed by pulses whih had been prereorded on the tape by omputer. Following presentation of the slides, test sheets were distributed, and SUbjets were told that a test series of pitures would be shown one at a time, and that they were to judge how many times eah piture had ourred in the previous series. They were told to guess when unertain, and if they did not remember a piture as having ourred, to give a judgment of zero. The test series was then presented at a 9-se rate. The, number of eah test slide 0-144) was announed, so that subjets ould write eah judgment in the appropriate spae. Results The mean frequeny judgment for distrator (frequeny of zero) pitures was 0.23. Mean judgments for the three single-presentation onditions and the 18 double-presentation onditions are presented in Figure 1. Four features of the data should be noted. First, judged frequeny of single-presentation pitures inreased slightly with exposure duration. Seond, the judgments given to double-presentation pitures tended to be higher the longer was the duration of P' This outome is ontrary to what would be expeted if, as the habituation-reovery hypothesis suggests, the inhibitory effet of P is greater the longer the subjet has attended to P' Third, a spaing effet was obtained. And fourth, also ontrary to the predition of "overhabituation," the magnitude and form of the spaing effet were unaffeted by the duration of P. 2.00 11.75 i' u: 1.25 o o j 1.00 '( t Si,""le P, Owation (se) 8.2 r-o 5.2 0--0 2.2.--,,,,.8 3.8 6.8 9B 12.8 Offset to Rz Onset (se.) 18.8 Figure 1. Effets of P duration and P, -P, spaing interval on judged frequeny, Experiment. r
CAUSES OF SPACNG EFFECT 289 The preeding onlusions were all supported by a planned-omparisons analysis of variane (Grant, 1954). The analysis showed an effet of frequeny, F(1,62) = 1219, p<.oo; an effet of duration of single-presentation items, F(l,62) = 7.50, P <.01; and an effet of P duration for double-presentation items, F(l,62) = 10.1, p <.005. To test for the spaing effet, omparison oeffiients were seleted whih approximated the form of the spaing urve obtained in previous experiments using pitures (Hintzman & Rogers, 1973, Figure 2). For spaings of 0.8, 3.8,6.8, 9.8, 12.8, and 18.8 se, respetively, the oeffiients were: -16, -6, 0, 5, 8, and 9. The overall effet of spaing was signifiant, F(l,62) = ll, P <.001; but three tests for pairwise interations of P duration with spaing were all nonsignifiant, F < 1. The outome is lear. f the spaing effet is produed by habituation, then inreases in habituation that will affet the spaing funtion apparently annot be produed by extending the duration of P beyond 2.2 se. There remains the possibility, however, that repeated presentation of a piture might produe "overhabituation" even though inreased duration does not. To test this predition, we performed Experiment. EXPERMENT Method Materials and design. The experimental items were 140 olor vaation pitures. Eah piture was randomly assigned to 1 of 14 onditions. The 28 pitures in one ondition served as distrator items. Another 28 pitures ourred one time eah during the presentation series. Eight-four pitures ourred either two, three, or four times during the presentation series, seven in eah of 12 onditions representing the orthogonal ombination of three levels of massed repetition of the first M presentations and four levels of P M -P M + 1 spaing. The three levels of M (the same piture shown M times in suession) were one, two, or three; the four levels of spaing between PM (the last massed presentation) and PM+l were zero, one, three, and seven intervening items (0.8,3.8,9.8, and 21.8 se, respetively). The presentation series was divided into seven bloks, and eah of the 13 onditions (distrator ondition not inluded) was represented in eah blok. The within-blok order of onditions was random, subjet to limitations imposed by spaing requirements. Eah of the 12 multiple-presentation onditions was represented in eah blok by one piture, and the single presentation ondition was represented by three or four pitures. n addition, there were three or four filler items in eah blok, ourring two, three, or four times at spaing intervals that varied. The first two and last two slides in the entire series were also filler items. Altogether, there were 324 slides in the presentation series. The frequeny-judgment test, whih was omposed of 144 pitures, was divided into four bloks of 36. Eah of the 12 multiple-presentation onditions was represented by two pitures in three of the bloks and by one piture in another. Both the single-presentation ondition and distrator ondition were represented by seven pitures per blok, and the remaining test items were fillers. The within-blok order of onditions was random, and all bloks of the presentation series were about equally represented in eah blok of the test series. Subjets and proedure. The subjets were 41 paid volunteers obtained as in Experiment 1. They were run in four sessions of 10 or 11 subjets eah. Aross sessions, pitures were rotated through spaings, but they remained in the same frequeny ondition and blok. Thus eah multiple-presentation piture ourred at eah level of PM-PM +1 spaing. At the outset of eah session, subjets were told that a series of 324 pitures would be shown for 3 se eah, that the series would take 16 min, that many of the pitures would be repeated several times, and that they were simply to study eah slide and try to remember it for a later test. Presentation and test proedures were essentially the same as in Experiment 1. Results The mean frequeny judgment for distrator items was 0.17. Mean judgments from the other onditions are presented as a funtion of spaing and M, the number of massed repetitions preeding the spaing interval, in Figure 2. As is lear from the figure, mean judged frequeny inreased with M (as it should, sine true frequeny is M + 1), and also inreased with spaing, at all levels of M. Further, the form of the spaing urve does not appear to have been altered systematially by hanges in M. t is puzzling that the urve for M = 2 is loser to that for M = 1 than to that for M = 3 sine frequeny disrimination is typially better between lower than between higher frequenies (Hintzman, 1969). This outome may reflet an inadvertent item seletion problem: Pitures were rotated through all spaings within frequeny levels, but not aross frequenies. Nevertheless, the ruial rotation is the one that was arried out, sine the predition being tested onerns differenes in the form of the spaing urves. An analysis of variane onfirmed the foregoing onlusions. Linear trend on frequeny was signifiant, F(l,40) = 1379, p <.001; as was the overall effet of spaing, tested using omparison oeffiients -6, -1, +3, and +4, F(l,40)=235, p<.ool. Three tests for interations of M and spaing, involving all pairwise 3.0 t 5 (2.0 u:: 1.5 LOT 0""// / o t Single -------...,...-_---e----- M-.1>-...3 / ----0 3 - PM + Spailg Figure 2. Effets of number of prior massed presentations (M) and PM-PM+l spaing interval on judged frequeny, Experiment. 7
290 HNTZMAN, SUMMERS, AND BLOCK ombinations of the three levels of M, all failed to reah signifiane, p >.05. These results are in omplete agreement with those of Experiment. f the spaing effet is aused by a habituation-reovery proess, then apparently neither inreasing the exposure duration of a stimulus beyond 2.2 se nor repeating it several times in suession has any appreiable effet on the ourse of reovery from habituation. The predition of "overhabituation" has not been onfirmed. NTERVAL RETREVAL HYPOTHESS Experiments and, in showing that neither inreased exposure nor repetition just prior to the spaing interval influenes the form of the spaing effet, reveal still another finding that may be important. The spaing effet appears to be a funtion of the time from offset of P to onset of Pz, not of onset-to-onset time. This outome is onsistent with a hypothetial reovery proess only if it is assumed that reovery is independent of the amount of prior exposure. t is also onsistent with an explanation of the spaing.effet that we shall all the interval retrieval hypothesis. This hypothesis makes a rather unorthodox assumption about the nature of the information that is enoded in memory, and so some spae must be devoted to developing the argument. The shortest p. -P 2 interval used in the previous experiments was 0.8 se. Suppose this interval were made suessively shorter, until the offset of the stimulus was so brief that it was diffiult to detet. Under suh irumstanes, it would not be surprising if subjets sometimes reported seeing the stimulus only one, sine in order to know it ourred twie one would have to be aware of the interval separating P and Pz. Now extend this notion to intervals of 0.8 se and longer. Judging from memory that an item ourred twie instead of only one an be thought of as equivalent to remembering that an interval separated P and Pz. Of ourse 0.8 se of time is not likely to go unnotied when it ours. But it is possible that the memory trae of the P -P z interval is retrieved more easily the longer was the duration ofthe interval, just as the memory trae of a piture is retrieved more easily the longer was its exposure duration. This hypothesis requires reversal of the figure-ground relation ustomarily used in thinking about memory experiments. t is assumed that the retrieved information from whih the subjet infers that a stimulus ourred twie is the memory trae of the offset-to-onset interval rather than the trae of the stimulus item itself. A diffiulty with the interval retrieval hypothesis, in addition to its unorthodox nature, is that it does not lend itself as easily to explaining the effet of spaing on reognition or reall as. it does to judged frequeny. Nevertheless, beause of the agreement of the hypothesis with the outomes of Experiments and an attempt was made to test the hypothesis diretly. f subjets remember a massed-repetition item as having ourred only one beause they fail to retrieve the spaing interval, then they should judge the exposure duration of the remembered "single presentation" as having been espeially long. Experiment tested this predition using a task in whih subjets judged both frequeny and exposure duration. Sine exposure duration must be varied in order for a request for duration judgments to seem reasonable, the presentation series inluded not only an orthogonal manipulation of frequeny and spaing, but also an orthogonal manipulation of frequeny and exposure duration. The latter manipulation also allowed a repliation of a previous finding of relative independene of judgments of frequeny and duration (Hintzman, 1970), this time using pitures instead ofwords. EXPERMENT Method Materials and design. The experimental items were olor transparenies of 96 vaation senes. Eah piture was randomly assigned to 1 of 14 onditions. The 18 pitures in one ondition served as distrator items. Another 18 pitures ourred one time eah, six at eah of three levels of exposure duration (2.2, 5.2, and 8.2 se). Thirty pitures ourred two times eah, six in eah of five onditions. n three of these onditions, exposure duration was 2.2 se, and the P.-P 2 spaing was either 0.8, 3.8, or 15.8 se. n the other two, spaing was 15.8 se and exposure duration was either 5.2 or 8.2 se, with p. and P 2 durations the same for a given piture. A further 30 pitures ourred three times eah, six in eah of five onditions varying in spaing and exposure duration exatly as in the ase of the double-presentation pitures. The nature of the inomplete orthogonal design is illustrated in Figure 3. For pitures that ourred two and three times, spaing was varied only for items having an exposure duration of 2.2 se, and for items that varied in exposure duration, spaing was always 15.8 se. The interval from the offset of one slide to onset of the next was always 0.8 se. The presentation series was divided into six bloks, and eah of the 13 onditions shown in Figure 3 was represented by one piture in eah blok. The within-blok order of onditions was random, subjet to the limitations imposed by the spaing requirements. The first 14 and last 10 slides in the entire series were filler items, some of whih were repetitions, and an additional 19 filler items were distributed throughout the presentation series. Altogether, there were 211 slides in the presentation series. The test series, omposed of the 96 experimental pitures, was divided into six bloks of 16 slides eah. Eah of the 13 experimental onditions was represented by one piture and the distrator ondition by three pitures in eah blok. The within-blok order of onditions was random, and all bloks of the presentation series were about equally represented in eah blok of the test series. There were two similar test sheets for eah subjet, the first numbered 1-48 and the seond 49-96. To the right of eah number were (1) a olumn ontaining a 0, headed "Did not our," (2) a olumn ontaining the numerals 1, 2, and 3,
CAUSES OF SPACNG EFFECT 291 headed "Number of times," and (3) a olumn ontaining the numerals, 3, 6, and 9, headed "Exposure duration." A main heading of "Did our" extended over both the seond and third olumns. Subjets and proedure. There were 82 subjets, obtained as in Experiments and. They were run in 10 sessions of 7 to 10 subjets eah. Aross sessions, pitures that ourred two and three times were rotated through all 10 ombinations of two or three presentations, the three levels of spaing, and the three exposure durations. The distrator and single-presentation pitures were rotated through these two onditions and the three levels of exposure duration. The presentation instrutions and proedure were essentially the same as in Experiment. For the test, SUbjets were asked to judge how many times eah test piture had previously ourred and for how long it had been shown, guessing if unertain. f they did not remember a piture as having previously ourred, subjets were told to irle the 0 on the test sheet under the olumn headed "Did not our." f they thought that a piture had ourred in the previous series they were to indiate how many times it was shown by irling either the, 2, or 3 in the olumn headed "Number of times," and the amount of time the piture was shown eah time it ourred by irling either, 3,6, or 9 in the olumn headed "Exposure duration." t was emphasized that every time a piture was shown it ourred for the same amount of time, and that the exposure duration judgment referred to eah individual presentation, disregarding the number of repetitions and total exposure time. Results and Disussion Reognition proportions were first determined for eah subjet in eah ondition, by ounting "Did not our" responses as reognition failures. Then mean judgments of frequeny and duration were determined for eah subjet, ounting reognized items only. Thus judgments of zero frequeny were not inluded in the frequeny-judgment distributions as they were in Experiments and. The proportion of false alarms made to distrator items was.09. For the other onditions, the mean proportions of pitures orretly reognized as old are presented in Figure 4. The left panel of the figure shows effets of frequeny and spaing with duration onstant at 2.2 se, and the right panel shows the joint effets of frequeny and duration with spaing onstant at 15.8 se. (Note that two data points in the left panel are dupliated in the right panel sine the onditions were 8.2 5.2 2.2 DURATON (SEC.) Figure 3. Design of Experiment (6 pitures per ell). 100 "0.95 j =d!.90 j 85 j J /".80 1.75 f!!! }!!.8 3.8 15.8 2.2 5.2 82 SpoinQ (sed Duration (se) Figure 4. Reognition memory as a funtion of spaing, frequeny, and exposure duration. the same.) The data in the right panel learly show that both frequeny and exposure duration affeted reognition performane. Linear trends on frequeny and duration were both reliable, F(1,81) = 141 and 2004, respetively, both p <.001. The data in the left panel show a spaing effet in reognition memory for pitures. Although the urve for frequeny of three was somewhat irregular, the overall spaing effet, tested using omparison oeffiients of -3, +1, +2 for spaings of 0.8,3.8, and 15.8 se, respetively, was signifiant, F(1,81) = 5.72, p<.05. Mean judgments of frequeny, onditionalized on reognition, are presented in Figure 5. The right panel shows that judged frequeny inreased with both frequeny and exposure duration, although frequeny had by far the greater effet. Linear trends on frequeny and duration were both reliable, F(1,81) = 992 and 33.5, respetively, both p <.00. The effet of spaing, shown in the left panel, was also signifiant, F(,81) = 89.5, p <.001. Mean judgments of duration are presented in Figure 6. The right panel shows thatjudgments ofduration, while an inreasing funtion of both frequeny and duration, were influened more by duration than by frequeny. Linear trends of duration judgments on both frequeny and duration were signifiant, F(1,81) =40.8 and 173.5, respetively, both p <.001. The finding that subjets were able to reliably judge duration indiates that the experiment is a valid test of the predition regarding. effets of spaing on remembered duration. Aording to the interval retrieval hypothesis the subjet mistakenly judges a massed-repetition item as having ourred one beause of a failure to retrieve the trae of the P -P 2 interval. Thus, two short presentations should be remembered as one long presentation; and so the funtion relating judged duration to spaing should be the inverse of that for
292 HNTZMAN, SUMMERS, AND BLOCK judged frequeny. As is shown in the left panel of Figure 6, this is not the ase. The effet of spaing on judgments of duration, although slight, was signifiant and in the same diretion as the effet on judged frequeny, F(l,81) =12.5, P <.001. Thus, overall mean duration judgments offer no support for the interval retrieval hypothesis. Although it is reasonable to predit from the hypothesis that overall mean duration judgments will show a reverse spaing effet, a more sensitive test involves only those ases where judged frequeny was below the true frequeny. Three diret-differene t tests were done, seleting for eah only the subjets who ontributed sores to both ategories being ompared. n all three tests, the ontrast was between the 0.8- and 3.8 se spaing onditions, sine it was between these two onditions that the greatest effet of spaing on judged frequeny was found. For true frequeny of two and judged frequeny of one, 60 subjets ontributed. Mean duration judgments for 0.8 and 3.8 se spaings, respetively, were 5.04 and 4.80 t(59) = 0.88. For true frequeny of three and judged frequeny of one, 36 subjets ontributed, and the respetive means were 5.38 and 526 t(35) = 0.37. For true frequeny of three and judged frequeny of two, with 71 subjets ontributing, mean duration judgments were 5.05 and 4.94, t(70) =.48. None of the three tests approahed statistial signifiane Thus even when duration judgments were onditionalized on underestimation of frequeny, there was no evidene that the O.S-se spaing interval resulted in longer remembered durations than the 3.8-se interval. Disomfirmation of the predited relationship between judged duration and spaing seems unequivo!;:al. The joint effets of frequeny and duration on the three dependent variables, illustrated in the right-hand panels of Figures 4, 5, and 6. ae expressed in terms of explained variane in Table 1. This information an be ompared with similar data from the Hintzman (1970) study in whih words were used as stimuli. The omparison should be made with aution, however, as several details-most notably the range of exposure durations-differed in the two studies. Note that while frequeny and duration both ontributed to reognition probability, total time aounted for no more than 69% of the variane, while the additive effets of the two variables aounted for as muh as 93%. Thus a given amount ofexposure time was less effetive as one long presentation than it was as two or three spaed presentations. (When reognition was expressed as d', the outome was essentially the same.) More evidene relevant to this point will be given in the onluding setion of the paper. Frequeny judgments, as Table 1 shows, were determined almost entirely by frequeny. n ontrast, duration judgments were affeted muh less by frequeny than by duration. The additive effets of the 250. 2.25.1:: 8' 2.00 a:. i_> " 1.75 u.!; 1.50 o 1.25 65 8 6.0. " 5.5 o!; 5.0 o 4.5.8 3.8 15.8 2.2 5.2 8.2 Spaing (se.! Duration (se.! Figure S. Mean judgment of frequeny (onditional on reognition), as a funtion of spaing, frequeny, and exposure duration. 70,.----------,-----------, t f,, f 8 3.8 158 Spaing (se),, j 2.2 52 82 Duration (se) Figure 6. Mean judgment of duration (onditional on reognition), as a funtion of spaing, frequeny, and exposure duration. two variabes on judged duration exeeded the effet of total time (97% vs. 91 % of the variane). The overall pattern of frequeny. and duration-judgment results is similar to that obtained using words (Hintzman, 1970). Table 1 Variane Among Means (r 2 ) Aounted for by Frequeny, Duration, and Total Time Dependent Variable Judgment Given Reog- Reognition lndependent Variable nition 10F 10D Frequeny (F).77.96.21 LogF.83.98.20 Duration (D).09.01.76 LogF.10.01.75 Total Time (F x D).54.50.86 LogF X D.69.51.91 Note-lOF '" judgment of frequeny, lod = judgment of duration.
CAUSES OF SPACNG EFFECT 293 However, the effet of duration on duration judgments was muh stronger in the present experiment than in the previous one. This differene may be partly due to the different stimuli used in the two studies, although the wider range of exposure durations used here was undoubtedly a major fator. Taken together, the present frequeny- and duration-judgment data onfirm the onlusion of Hintzman (1970) that frequeny and duration have effets on memory that are disrirninably different to the subjet at the time ofretrieval. CONCLUSONS Neither the predition that "overhabituation" produed by prolonged or repeated exposure of a stimulus would affet the ourse of reovery, and hene the form of the spaing urve, nor the predition that massed presentations would tend to be remembered as one long exposure was onfirmed in the present experiments. Thus, both the habituation-reovery hypothesis and the interval retrieval hypothesis have failed to reeive support in deliberate attempts to find evidene in their favor. We are seemingly left with only one new bit of information about the spaing effet: that it is a funtion of time from the offset of P rather than from its onset. The preditions of three alternative explanations of the spaing effet that were not onsidered in the previous disussion may be related to this finding. The first explanation, onsolidation theory, is vague with respet to the relative importane of the onset and offset of P in produing the effet. This hypothesis holds that the spaing effet is due to an interation between the proessing of P 2 and the ongoing proessing or onsolidation of P in the period following its presentation (Landauer, 1969; Peterson, 1966). Sine the theory does not address the question ofeffets of exposure duration, it is not lear whether onsolidation should be assumed to start at the onset of P or at its offset or whether exposure duration should affet the intensity and duration of the subsequent onsolidation ativity. Thus, different versions of onsolidation theory would have predited quite different outomes of Experiment. A seond explanation of the spaing effet, the ontextual hange hypothesis, begins with the assumption that elements of the stream of onsiousness present when a stimulus ours beome assoiated with the representation of the stimulus, and provide information, at the time of retrieval, that the stimulus ourred in the experimental setting (Anderson & Bower, 1972). The spaing effet is explained by the additional assumption that massed repetitions tend to sample the same ontextual elements, and so they aid retention less than do repetitions that are spaed further apart in time (Bower, 1972; Hintzman, 1974). The problem with this hypothesis, also, is its ambiguity with respet to effets of duration of P. s the turnover of elements in the stream of onsiousness stritly orrelated with time, or are the same elements maintained somehow throughout the exposure ofa given stimulus, from its onset to its offset? Again, different versions of the hypothesis would make different preditions regarding the duration manipulation of Experiment. A third hypothesis is that the enoding ofp 2 is more effetive the greater is the degree offorgetting ofp' An extreme statement of this position, whih holds that the subjet will enode P 2 only if P annot be retrieved (Tzeng, 1973), is ertainly inorret, sine the spaing effet ours with materials for whih reognition auray is very high. A weak version of the hypothesis, relating the effetiveness of enoding of P 2 to the relative diffiulty of P retrieval, is more plausible. However, neither the duration of P, in Experiment, nor the number of prior presentations of the item, in Experiment, influened the form or magnitude of the spaing effet. Sine both manipulations are known to affet reognition, their lak of an effet on the spaing urve appears inonsistent with a hypothesis that relates the spaing effet to retrieval diffiulty. Can anything positive be onluded about the nature of the proess underlying the spaing effet? One unantiipated lue, whih was disovered in the reognition data of Experiment, is presented in Figure 7. At the left of eah row in the figure is shown the temporal distribution of study time for pitures in a given ondition. To the right of this are listed the total presentation time and the probability ofreognition for that ondition. Comparisons an be made between onditions that are mathed with regard to initial onset to final offset times. The top panel ompares one 5.2-se exposure with two massed 2.2-se exposures. The middle Total P(Rn) Time (se.l 4.4.822..J 52 L 52 802 6.6885..Jr-----,8'"'.2=-----,L 8.2.823 4.4.859...Jr---- 8 :".2,..-----.L 8.2.852 Figure 7. Effets of probability of reognition (right olumn) of seleted onditions of Experiment (top two panels) and Experiment (bottom panel). nterrupted pitures are better reognized than omparable uninterrupted pitures.
294 HNTZMAN, SUMMERS, AND BLOCK panel ompares one 8.2-se exposure (in the enter) with three massed 2.2-se presentations and with two 2.2-se presentations having an offset-to-onset time of 3.8 se. n the bottom panel, a omparison of two appropriate onditions from Experiment is made. n all omparisons, reognition memory for the interrupted stimulus was better than that for the ontinuous stimulus with whih it is ompared. This was true despite the fat that the total study time of the interrupted stimulus was less. While only the.885 vs..823 omparison of the middle panel was statistially signifiant t(81) =3.11, P <.01, the diretions of all the differenes favored the interrupted stimuli over the ontinuous ones. Apparently, at least within ertain limits, turning a piture off briefly one or more times produes better retention than simply leaving the piture on. This phenomenon might be alled the interrupted stimulus effet. The present results do not allow us to eliminate possibl trivial explanations of this effet (for example, onset of a piture may ause the subjet to examine it briefly to see whether it has hanged). Nevertheless this finding, together with the evidene that the joint effets of duration and frequeny on reognition memory are not multipliative (Table 1), seems entirely onsistent with the spaing effet itself. t may be that the underlying proess is the same. The failitative effet of stimulus interruption suggests that, despite the lak of evidene of "overhabituation" in Experiments and, some sort of habituation-reovery hypothesis may yet turn out to be the best explanation of the effet of spaing on memory. REFERENCES Anderson, J. R., & Bower, G. H. Reognition and retrieval proesses in free reall. Psyhologial Review, 1972, 79, 97-123. Biork, R. A., & Allen, T. W. The spaing effet: Consolidation or differential enoding? Journal of Verbal Learning and Verbal Behavior, 1970, 9, 567-572. Bower, G. H. Stimulus-sampling theory of enoding variability. n A. W. Melton and E. Martin (Eds.), Coding proesses in human memory. Washington, D.C: Winston, 1972. Pp. 85-123. Cohen, G. How are pitures registered in memory? Quarterly Journal of Experimental Psyhology, 1973, 25,557-564. Grant, D. A. Analysis-of-variane tests in the analysis and omparison of urves. Psyhologial Bulletin, 1956, 53, 141-154. Hintzman, D. L. Apparent frequeny as a funtion of frequeny and the spaing of repetitions. Journal of Experimental Psyhology, 1969, 80, 139-145. Hintzman, D. L. Effets of repetition and exposure duration on memory. Journal of Experimental Psyhology, 1970, 83, 435-444. Hintzman, D. L., Blok, R. A., & Summers, J. J. Modality tags and memory for repetitions: Lous of the spaing effet. Journal of Verbal Learning and Verbal Behavior, 1973, 12, 229-239. Hintzman, D. L. Theoretial impliations of the spaing effet. n R. L. Solso (Ed.), Theories in ognitive psyhology: The Loyola Symposium. Potoma, Md: Lawrene Erlbaum Assoiates, 1974. Pg. 77-99. Hintzman, D. L., & Rogers, M. K. Spaing effets in piture memory. Memory & Cognition, 1973, 1, 430 434. Landauer, T. K. Reinforement as onsolidation. Psyhologial Review, 1969, 76, 82-96. Peterson, L. R. Short-term verbal memory and learning. Psyhologial Review, 1966, 73, 193-207. Shaffer, W.O., & Shiffrin, R. M. Rehearsal and storage of visual information. Journal of Experimental Psyhology, 1972,92, 292-296. Tzeng, O. J. L. Stimulus meaningfulness, enoding variability, and the spaing' effet. Journal of Experimental Psyhology, 1973,99,162-166. (Reeived for Publiation August 16, 1974; aepted September 4, 1974.)