Acta Psychologica 0 North-Holland
41 (1977) 4199432 Publishing Company
ISOLATION EFFECT: OVERALL SHORT-TERM MEMORY
LIST
FACILITATION
IN
Richard S. CIMBALO, Rosemary A. CAPRIA, Linda L. NEIDER, and Mary Ann C. WILKINS Daemerz College, Amherst,
Received
November
U.S.A.
1976
The classical literature argued that when attention-getting items are placed in the middle of a to-be-learned list of items, the enhanced performance on the isolate is obtained at the expense of the other list items. Three experiments are presented which are not consistent with this tradeoff formulation and argue rather for an improvement in overall list performance. It is argued that this effect: (a) seems related to serial memory tasks wherein a repetition strategy is employed, (b) requires that the isolated item be centrally located in the list, and (c) is facilitated by a simultaneous presentation of list items.
The great bulk of the research findings dealing with the isolation effect (IE) tend to support Wallace’s (1965) observation that the advantage accruing to the isolated item is obtained at the expense of other list items. This ‘rob Peter to pay Paul’ explanation would predict no difference in overall performance between isolated and unisolated lists. However, since Wallace’s review article at least three conventional learning studies have appeared which demonstrate an overall list facilitation effect (OLFE) (Raskin et al. 1967; Turner and Lippman 1969; Simon 1971). There have also been at least three short-term memory (STM) studies which have produced the OLFE (Cimbalo and Pelonero 1970; Johansson 1970; Stock 1970). Prior to Wallace’s review a STM (Young and Supa, 1941) and a classical verbal learning task (Wishner et al. 1957) had reported an OLFE. Wallace dealt with the Young and Supa article as a curious anomaly and the Wishner et al. article was not cited. These studies pose a serious problem for a simple attentional type of theory which conforms to the total time hypothesis, i.e., the ‘rob Peter
420
R. S. Cimbalo et al./Isolation and list facilitation in STM
to pay Paul’ explanation, but are very compatible with theories which posit a reduction in interference and/or change in organization which results in enhanced processing efficiency. The isolation methods used to obtain the OLFE have varied. They include shock (Raskin et al. 1967), more than one isolated item (Simon 1971; Stock 1970; Wishner et al. 1957; Young and Supa 1941), and simultaneous list presentation (Cimbalo and Pelonero, 1970; Johansson 1970). The three experiments that follow are attempts to converge on the necessary and sufficient conditions for the occurrence of the OLFE.
Experiment
1
In the first experiment the number of dimensions along which the isolated item differed was made to differ. Size, color, spacing or all three variations in combination were used to achieve isolation. It was hypothesized that the combination of all three variations should yield the greatest isolation effect and provide a clearer picture of any effect of isolation on surrounding items. The great majority of studies which have dealt with the IE have used multi-trial presentations. The determination of the IE and the spread of effect in these studies involved an averaging technique. Inasmuch as we seem to be dealing with attentive processes, which can change with time and number of exposures, it seemed legitimate to question whether similar results would be obtained with a single isolation trial per subject. Earlier experiments have used the single trial technique (Green 1958; Saltz and Newman 1959) but they were not designed to assess the effect of isolation on other list items and overall performance differences. Harcum (1965) has argued that pre-knowledge of the isolation is a prerequisite for obtaining the IE. This argument would preclude the IE in this study. Method Size (isolated = 7x 10 mm; unisolated = 4x 7 mm), color (isolated = red; unisolated = black), and spacing (isolated = 14 mm; unisolated = 7 mm) differences for the fifth consonant in a lo-consonant list were the characteristics used to achieve isolation. These conditions considered separately along with a condition wherein these three characteristics were combined and a homogeneous group comprised the five conditions of the study. There were 24 Ss in each of the five conditions. The Ss served in the experiment as part of a course requirement. The stimulus material consisted of ten randomly selected consonants (y considered a vowel) centered on index cards with 7 mm between each consonant. The lists were presented for 10 set in one channel of a Scientific Prototype 3-channel tachistoscope. Preceding the stimulus presentation by 2 set was a small cross in the center of the field of vision which subjects were asked to fixate. Succeeding the stimulus was the printed word ‘record’.
421
R. S. Cimbalo et al/Isolation and list facilitation in STM
Ss were instructed to learn the list of consonants and upon presentation of the record signal they were to write down the consonants in the order in which they appeared on the card. They were also told that their recall would be paced and that they would be required to recall the items in beat to an electronic metronome which sounded once each second. 5s were shown five lists in all. The first list was a practice list after which the Ss were asked if they had any questions. They were also to read the letters of the practice list aloud to serve as a check that the letters could be seen. The next three lists were unisolated and the same three were used for all Ss with the order randomized. The fifth list was one of the five experimental conditions. There were two randomly designed lists for each experimental condition. Results An ANOVA of the total number of items correct wherein both item and order information were required for the four isolated conditions displayed an F < 1. When the unisolated condition was compared with the average of the four isolated conditions, a significant difference in overall performance was obtained, F( 1,46) = 8.01, p < 0.01. Because of the very similar shapes of the serial position curves for the various isolated conditions, the data for these isolated conditions were averaged and compared with the unisolated condition. The results, which appear in fig. 1, indicated that the differences in performance can be attributed to differences in the isolated item and the two items which precede it. The differences in performance at positions 3, 4, and 5 were significant at the 0.01 level with z = 2.58, z = 3.79 and z = 6.07 respectively (Bruning and Kintz 1968). None of the other positions displayed significant differences. The proportions of Ss correctly recalling the isolated item for the size, combination, color, spacing and unisolated conditions were 0.96, 0.92, 0.83, 0.67, and 0.33 respectively.
3 t; #
8z $b
-MEAN
OF THE 4 ISOLATEQ CONDITIONS
A--iiMEAN
OF UNISOLATED CONDITION
ca
a$ $d .=g 8= a.
02.
g 2
3
4
SERIAL
5
6
7
0
9
IO
POSITION
Fig. 1. The combined means of the four isolated bination) and the means for the unisolated condition
conditions (size, spacing, color, and complotted as a function of serial position.
422
R. S. Cimbalo et al./Isolatiorl and list facilitatiorl irz STA4
Performance was significantly better for all types of isolation when compared to the unisolated condition. Size differences produced significantly better performance than either spacing or color with z = 4.06 and z = 2.03 respectively. Performance for the combination isolation condition was significantly better than the spacing condition with z = 3.2 1. Contrary to expectations the combination did The failure to find an enhancement for the not yield the best performance. combination condition could be attributed to the near perfect performance in the size and combination conditions, i.e., a ceiling effect. If color and spacing or some less potent isolation characteristics were compounded one would have a more sensitive test of the combination effect (Erickson 1963).
Experiment
2
Cimbalo and Pelonero (1970) and Johansson (1970) both found the OLFE and these studies used the simultaneous presentation method. To the best of our knowledge none of the earlier studies which assessed the OLFE used a simultaneous presentation technique. This experiment sought to compare the simultaneous and successive methods of presentation with the prediction that the OLFE would be found with the simultaneous but not with the successive presentation method, i.e., is the simultaneous method of presentation a necessary condition for the OLFE? Johansson (1970) found that the particular strategy reported by the S encoding the isolated list affects the spread of effect. Experiment 2 also attempted to determine whether the OLFE was contingent upon a particular encoding strategy. Method The Ss in this experiment were 50 undergraduate students who served to fulfill a course requirement. The stimulus material consisted of 32 lists of ten randomly selected consonants. The consonants were typed on transparencies and mounted on 35 mm slides. The isolation effect was achieved by placing a 4x 8 mm consonant in the fifth position on 16 of the 32 lists. All other consonants were 2 mm*. One-half of the 32 lists contained all ten items on a slide (simultaneous) and the remaining half contained one item per slide (successive). The slides were shown on a Kodak Carousel projector. A separate 2x 2x 2 completely within-subject ANOVA was performed for each of two rates of presentation conditions. One-half of the Ss received items at 60 and 30 items/min (60-30 condition) and the remaining half at 60 and 15 items/min (60- 15 condition). Isolation (isolated vs. unisolated) type of presentation (simultaneous vs. successive) and rate (60 and 30 items/min or 60 and 15 items/min) were the three factors. Before each list the S was cued by a slide as to whether the presentation would be ‘fast’ or ‘slow’. At the end of each list the Ss were allowed fifteen seconds to record the items. The Ss were instructed to record the items from left to right in the order in which they appeared.
R. S. Cimbalo et aLlIsolation and list facilitation in STM
423
After the learning task was completed and the response sheets were handed to the E the Ss were asked to complete a questionnaire indicating which of four strategies they utilized or was most like the dominant strategy utilized to learn the lists. The strategies were (1) repeating; that is simply repeating the letters to oneself in order to retain them, (2) association; that is inserting vowels to make words or making sentences, (3) rhythm; that is repeating the items in a rhythmical pattern and (4) visual imagery; that is retaining a picture of the items in one’s head. Results Overall performance The three main effects for the 60-30 presentation conditions were found to be significant. Performance was superior for isolated, simultaneously presented, and slower lists of items. The significant Type of Presentation x Rate (F(1,24) = 9,85, ~~0.01) and the Isolation x Type of Presentation x Rate (F(1,24) = 13.42, p
424
150%
i $W200% ASSOCIATION (60-15)y- <
ASSOCIATION (60-30) 140-
I20
B .-. A-A
,oo_ti
I SIMULTANEOUS-FAST SUCCESSIVE-FAST SIMULTANEOUSSLOW SUCCESSIVE-SLOW
I
\ \
/
I
\ I \
:
\ \
? i
;
’ j 234567
L
i
’
I
L 8
9
’ IO
-60’ ;
j
2
SERIAL POSITION
*-.SIMULTANEOUS-FAST
-
SUCCESSIVE
a
5
1
6
b
7
hIO
*
8
9
REPETITION (60-15)
8_240”
C I
4
SERIAL POSITION
REPETITION (60-30) 140
’
b
3
140
tl
t -*
D
k-4
120
t-t-
SIMULTANEOUS-FAST SUCCESSIVE-FAST SIMULTANEOUS-SLOW SUCCESSIVESLOW
-SLOW
-40 -60
1,
,
/
2
t 11
3
4
I
IA,
5
6
I
7
SERIAL POSITION
8
9
IO
-60
I
I
I1
I
2
3
I
I1
4
5
6
I
I
7
‘3
L
I
9
IO
SERIAL POSITION
Fig. 2. Serial position curves for the two dominant strategies for each between-subject rate condition (60-30, 60-15) with type of presentation and within-subject rate conditions as the parameters. Positive and negative percentage differences on the ordinate refer to isolated item superiority and inferiority, respectively, as computed from the equation below. Isolated Item Performance - Unisolated Item Performance x loo Percentage Difference = Unisolated Item Performance
R. S. Cimbalo et al./Isolation and list facilitation in STM
425
for the isolated list from the unisolated list item performance for the two strategies selected most often by Ss, i.e., repetition and association. The results are plotted in fig. 2. For the associative strategy the slower rate in both the 60-30 and 60-15 conditions produced the greater IE (fig. 2a-b). Although there was some enhancement for items prior to the isolated item the largest improvement was for items succeeding the isolated item. With the repetition strategy the faster rate in both the 60-30 and 60-15 conditions produced the greatest IE (fig. 2c-d). There does not appear to be a unilateral spread of effect with this strategy. Separate ANOVAs were performed on the two major strategies, viz., association and repetition. The OLFE for both the 60-30 and 60-15 conditions was found to be significant with F(l,lS) = 6.60, p
Experiment
3
This experiment will attempt to clarify the OLFE findings by presenting the Ss with the same list a number of times and noting developmental differences in performance between isolated and unisolated lists. It is hypothesized that when a S is permitted only brief exposure to a serial list the overall performance between the isolated and unisolated lists should not differ because the organizational processes necessary for this effect have not had time to be effected. As Ss are given more exposures of the same list, isolation should facilitate the formation of various strategies which promote overall list facilitation. Thus, on latter trials the overall performance should be better for those lists containing an isolated item.
426
R. S. Cimbalo et al. /Isolatiorz and list facilitation in STM
Method Ninety students from Introductory Psychology classes served to fulfill a course requirement. The design was basically a 2x 3x 3 factorial. List Type (isolated vs. unisolated) and Duration (1 set, 2 set and 5 set) were between-subject variables while Exposure (1, 6 and 12) was a within-subject variable. Three IO-item lists were constructed using the 20 consonants of the alphabet. The items in the list were randomly generated from the 20 consonants without replacement. All letters were printed in black on a 35 mm slide with an electric typewriter (IBM Adjutant type-face). Each letter was approximately 1x2 mm. Isolation was achieved by making the fifth letter in the list approximately 3x 8 mm. The three isolated lists contained the same letters as the three unisolated lists. Ss were presented the same slide 12 times. They were told that after each slide exposure they must record the items on the given answer sheets in the same order as those letters they had been presented. When the Ss had finished recording their responses they were to turn the page and wait for the next presentation. The Ss were instructed not to look back at earlier responses, The time between exposures was 15 set and guessing was encouraged.
Results
A 2x 3x 3 ANOVA on the number of items correct per list showed the main effects of the three variables to be significant. In order to be correct an item had to be recorded in the proper position. There was a significant difference in performance between list types indicating better performance for isolated lists with F( 1,84) = 4.65, p
427
R. S. Cimbalo et al./Isolation and list facilitation in STM
-UNlSOLATEO -ISOLATED
0.4Oh6 EXPOSURES J
040l$
12 MFQSURES ,, ,, 12345
Fig. 3. Proportion
,
,
,
,
6 7 8 SERIAL POSITION
of Ss who got an item correct
,
,
9
ID
in the various serial positions for 1, 6, and 12
exposures.
(Newman-Keuls, pccO.01) showed for a single exposure there was not a significant IE, but it was significant for six and 12 exposures. Two additional ANOVAs were performed with isolation and duration as between-subject variables. On the first of these, the dependent variable was the exposure on which item 5 was gotten correct. No significant difference was found between the isolated and the unisolated item. The second analysis noted the total number of trials that item 5 was correct in the two different list types with F( 1,84) = 21.50, p < 0.001 indicating that the isolated items were correct on a greater number of trials.
Discussion All three experiments offer additional support to the existence of the OLFE. The OLFE occurred for single presentations and was stronger for simultaneously presented lists. Its seeming current pervasiveness in view of its classical rarity (Wallace 1965) is of some interest. In his
428
R. S. Cimbalo et al. fIsolatior1 and list facilitation in STM
review paper of much of the extant literature up to that time, Wallace concludes “. . . that isolating an item in a list has little, if any, influence on the speed of learning that list. It appears that the isolated item will be learned more rapidly as a result of its isolation, but the improvement occurs at the expense of other items in the list” (p. 417). The three experiments presented here and many of those cited in this article leave little doubt that this conclusion must be revised. Of crucial importance to understanding the OLFE is the specification of the contingent conditions governing the effect. All of the studies cited and presented here which have found the OLFE have had lists longer than the immediate memory span. Another necessary condition for the OLFE which seems to hold up well under scrutiny is that the isolated item be centrally located in the list. Cimbalo (1969, 197 1) and Johansson (1970) both experimentally manipulated position of isolation and the OLFE was significant only for lists in which the isolated item occupied a central position. Experiment 2 directly manipulated the method of presentation and found evidence that the OLFE was facilitated by simultaneously presenting the list of items. However, only the Isolation X Type of Presentation X Rate interaction in the 60-30 condition was significant although the interactions of isolation in the 60- 15 condition are in the same direction. The simultaneous presentation condition also enhanced the IE. Here, however, the Isolation X Type of Presentation interaction was significant in the 60-15 but not in the 60-30 condition although the direction of the difference was the same. Subjects who reported using the association strategy displayed little evidence for the OLFE. Significant main effects and interactions in both major conditions for the repetition strategy argue strongly for its enhancement of the OLFE. STM studies, because of their faster presentation rates, rely more heavily on a repetition strategy. Subjects’ reports of dominant strategies in Experiment 2 would support this reliance. Gibbons and Leicht (1970) attempted to control the strategies used by subjects and found that instructions to repeat the items produced the OLFE but the reverse effect was found when subjects were allowed to freely choose their own strategy. The increased use of the STM paradigm since Wallace’s review (1965) may explain the increase in the reported OLFEs. Turner and Lippman (1969) present evidence which shows that both a high-intensity stimulus (tone) and information about the nature of
R. S. Cimbalo et al./Isolation and list facilitation in STM
429
the isolate were necessary to produce the OLFE. The variance between their findings and those reported in Experiment 1 may be attributable to task and resultant strategy differences. They used a serial anticipation task which would seem to rely more heavily on an associative strategy, whereas the STM task relies more on a repetition strategy. The fact that the OLFE and IE were obtained in Experiment 1 and not in the single exposure condition of Experiment 3 is of some interest. It might be argued that presentation rates between 120 items/ min and 60 items/min are necessary to produce the effect but this would not be in agreement with earlier research which found both effects at 240 items/min (Cimbalo and Pelonero 1970). Another possible explanation would be in terms of strategy differences resulting from the different task requirements. The differences between the isolated and unisolated conditions were due to differences in performance for the first five items in the unisolated condition. As can be seen in figs. 1 and 3 the performance on these items was much poorer in Experiment 1 than in Experiment 3. Perhaps more subjects adopted the more effective grouping strategy (Wickelgren 1964) and the left to right encoding when they knew they would receive multiple presentations and the use of this strategy in the unisolated condition in Experiment 3 eliminated any differences between the isolated and unisolated conditions. The foregoing argument greatly reduces any special significance attached to the absence of the OLFE and IE in the single exposure condition of Experiment 3. However, the data of Experiment 3 for the six and 12 exposure conditions argue strongly for the subdividing function of the isolated item in the list. The list clearly appears to have been subdivided by the subject. Initially the subjects approach the serial learning task in what appears to be the left-to-right strategy which characterizes reading in our culture (Harcum 1965). For the single exposure, fastest presentation rate condition none of the last five items were gotten correct by any of the subjects. In the same condition, however, perfect performance was achieved by all of the subjects for the first three serial positions. With further processing time the isolated item serves to subdivide the isolated list into two sets of five items. Primary emphasis seems to be placed on the first five items and when performance there is fairly high the emphasis switches to the later items. In contrast, as can be seen in fig. 3, the unisolated lists follow the more traditional characteristics of the serial position curve viz., the
430
R. S. Cimbalo et al./Isolatiorz and list facilitation in STM
bowed shape, and the minimum performance point beyond the midpoint of the curve. The IE, holding true to form as one of the more robust effects in the verbal learning literature, was found in all three experiments. The variables which affected the OLFE also affected the IE in a similar way. The consistent covariation of these two effects suggests a single underlying causal process. The developing performance characteristics of Experiment 3 suggest that this causal process is the subdivisional or organizational role played by the isolated item in its central location in the list. Of the several theories used to try to account for the IE only the Surprise Theory of Green (1958) or the simple attentional theory proposed by Wallace (1965) seem most easily discredited in view of the findings presented here. Surprise theory or the simple attentional theories maintain that the enhancement of the isolated item is obtained at the expense of the other items in the list. Other prominent theoretical explanations which would be compatible with the OLFE would be in terms of a reduction of interference, the orienting reflex (Simon 197 l), list structure (Stock 1970), effective list-shorterning (Cimbalo and Pelonero 1970), Gestalt concept of cortical traces, and anchor points in rehearsal (Feigenbaum and Simon 1962). The interference and orienting reflex explanations would predict a positive spread of effect to both preceeding and succeeding items. The serial position curves in all three experiments indicate the lack of any consistent spread effects and would argue strongly against such hypotheses. If Gestalt theory views the effect of the isolated item as a redefinition of the figure, i.e., the figure consists of two parts or the two halves of the list then it squares well with the serial position data and the OLFE. However, it would have difficulty with the strategy differences in Experiment 2. The three remaining explanations, viz., list structure, effective list-shortening and anchor points in rehearsal are all information processing theories with different focal points. They stress the subjects’ ability to organize and simplify complex arrays of information. They view the isolated item(s) as the crucial element in the organization and simplification of informational complexity. The data seem most consistent with an information processing theory of both the Von Restorff effect and the serial position effect. The isolated item in the center of the list seems to provide another anchor point which restructures the information into more manageable
R. S. Cimbalo et al./Isolation and list facilitation in STM
431
units for rehearsal. This would be in agreement with Wickelgren’s (1964) research which shows better overall performance for certain size units, viz., three or four items. Failures of some studies to obtain the classical Von Restorff effect can be attributed to different response demands imposed on the subject. If the subject is to recall as many items as possible and in serial order as in the experiments reported here, then a seemingly intelligent strategy in the light of Wickelgren’s findings would be to learn a set of items with clear bounds that can be rehearsed and recalled as a unit. For example, in a serial probe task as used by Lively (1972) the units may be pairs and this may explain the curious anomaly that an IE was not obtained. An information processing theory would relate the strategies of the subject, which are to a large extent determined by the demands of the task, in any explanation of the isolation effect. Murdock (1974) seems guilty of a gross overgeneralization and a serious neglect of the task demands when he states that for the Von Restorff effect “. . . the only effect here seems to be an increased tendency for the salient item to be given as a response” (p. 149). More appropriately, the only effect here seems to be an increased tendency for the salient item to determine the response dimensions. The once very elusive OLFE accompanying an IE (see Wallace’s (1965) review article) now seems reliably obtainable. The conditions which appear related to producing and/or enhancing the OLFE would be a list of items at or greater than the immediate memory span, simultaneous presentation, an isolated item or structural break in the central portion of the list, the subject engaged in a serial task and the utilization of a repetition strategy. Which of these conditions are clearly and empirically shown to be necessary and more importantly necessary and sufficient awaits more comprehensive theories and the requisite research.
References Bruning, J. L., and B. L. Kintz, 1968. Computational handbook of statistics. Glenview, Ill.: Scott Foresman. Cimbalo, R. S., 1969. Effects of item isolation upon the processing of serially presented information. (Unpubl. doct. diss.) Buffalo: State Univ. of New York. Cimbalo, R. S., 1971. Short-term memory: List length and the isolation effect. Proceedings of the 79th Annual Convention of the American Psychological Association 6, 35-36.
432
R. S. Cimbalo et
al. fholatiou
and list facilitatiolz ir?STM
Cimbalo, R. S., and K. C. Pelonero, 1970. The isolation effect and mechanisms in short- and long-term memory. Psychonomic Science 21, 69-70. Erickson, R. L., 1963. Relational isolation as a means of producing the Von Restorff effect in paired-associate learning. Journal of Experimental Psychology 66, 11 l-l 19. Feigenbaum, E. A., and H. A. Simon, 1962. A theory of the serial position effect. British Journal of Psychology 53, 307-320. Gibbons, D. E. and K. L. Leicht, 1970. Effects of meaningfulness and learning instructions on the isolation effect. Psychonomic Science 18, 353- 354. Green, R. T., 1958. Surprise, isolation and structural change as factors affecting recall of a temporal series. British Journal of Psychology 49, 21-30. Harcum, E. R., 1965. Pre-knowledge of isolation as a prerequisite for the isolation effect. Psychonomic Science 3, 571-572. Johansson, B. S., 1970. Attention and the Von Restorff effect. British Journal of Psychology 61,163-170. Lively, B. L., 1972. The Von Restorff effect in short-term memory. Journal of Experimental Psychology 93, 361-366. Murdock, B. B., 1974. Human memory: Theory and data. Potomac, Md.: Erlbaum. Neale, J. M., and R. M. Liebert, 1967. Science and behavior. Englewood Cliffs, N.J.: PrenticeHall. Raskin, D. C., M. Hattle and E. W. Rubel, 1967. The effects of electric shock isolation in serial learning. Psychonomic Science 8,413-414. Saltz, E., and S. E. Newman, 1959. The Von Restorff isolation effect: Test of the intra-list association assumption. Journal of Experimental Psychology 58,445-451. Simon, E. B., 1971. A method for producing overall list facilitation with non-noxious isolatton stimuli. Paper presented at the Western Psychological Association Meeting, San Francisco, April, 1971. Stock, W. A., 1970. Structure and short-term memory. Psychonomic Science 18, 2133214. Turner, K. D. and L. G. Lippman, 1969. Serial isolation effect as a function of instructions and stimulus intensity. Paper presented at the Western Psychological Association Meeting, Vancouver, April, 1969. Wallace, W. P., 1965. Review of the historical, empirical, and theoretical status of the Von Restorff phenomenon. Psychological Bulletin 63,410-424. Wickelgren, W. A., 1964. Size of rehearsal group and short-term memory. Journal of Experimental Psychology 68,413-419. Wishner, J., T. E. Shipley and M. S. Hurvich, 1957. The serial position curve as a function of organization. American Journal of Psychology 70, 258-262. Young, C. W., and M. Supa, 1941. Mnemic inhibition as a factor in the limitation of the memory span. American Journal of Psychology 54, 546-552.