The serial position effect in short-term memory under E- and S-paced conditions

The serial position effect In short-term memory under E- and S-paced conditionsl NORMAN R. ELUS AND JEANNE DUGAS, DEPARTMENT OF PSYCHOLOGY, UNIVERSITY OF ALABAMA, University, Alabama 35486 In a STM task Sa were presented visually nine CVCs in a row. A probe followed, requiring S to remember the position of one of the CVCs. SPE were found for both E- and S-paced conditions, with S-pacing facilitating primacy performance only, Latencies showed an SPE for 7 of 15 Ss. Long latencies resulted in more accurate performance in primacy, suggesting that Ss use rehearsal strategies. 100 90 ;-paced - - . S-paced ... - ... 80 II> <U II> 5 70 c. II> Cl> The study of serial position effects (SPE) in human learning and memory has a long history and a tedious recount seems inappropriate here. No single theoretical account of the phenomenon is entirely adequate. From the viewpoint of information theory, the processes involved in learning and memory, particularly short-term memory (STM), are encoding, storage, and retrieval, and the SPE in STM could be attributed to anyone or a combination of these theoretical constructs. Most studies of SPE have presented stimulus items at a constant rate, apparently assuming that items throughout a list are encoded with equal facility. Thus, SPE in such studies are usually attributed to storage or retrieval processes. If determinants of SPE occur in encoding, this should be reflected in item latencies, time spent viewing each item or interitem intelVals in an S-paced task. On the other hand, if the SPE are due entirely to storage or retrieval processes, these measures would not mow any differential effects. MElHOD SubjecH ud Apparatus The Sa were 30 undergraduates who participated to fulfill a psychology coune requirement. The apparatus has been described previously (Ellis & Hope, in press). In brief, a row of nine miniature projectors successively presented Yisualltimuli from left to right, and then a tenth projector located above the row "probed" memory for the position of a "Stimulus. That ii, S attempted to respond to one of the positions which was identical to the probe by pressing a Plexiglas key over the position he thought was correct. The stimuli were 12 CVCs with" values ranging from .30 to .40 selected from Noble's lilt (Noble, 1961). All eves began withll different letter. A doorbell dlime signaled a correct response. Interstimulus intervals, the interval between 'the ninth stimulus and the probe, and the interval between probe offset and response were recorded with a print-out counter and a .S-sec pulse a:: 60 Cl> u I 0 ~ u I I> 50 ~ I / I L4: / I I~ \ \ \ \ \ \ \ \ U ~ 0 u 1: Cl> u ~ 10 Serial A:lsition Fig. I. Pel'Clent correct tint choice rapoIIIeI for E- ..... S-pec:ed aroupiIu • function of ..... position. 10 9 meam generator. 8 1'nN:e4_ The 30 Sa were assigned equally and without bias to E-paced and 8-paced groups. In both groups the duration ofeach eve was .75 sec. For E-paced Ss all events occurred automatically. The offset of one stimulus was contiguous widt the onset of the next, and the probe immediately followed offset of the ninth CVC. For the S-paced condition S was instructed to present stimuli to himself by pressing each position, one after the other from left to right, and then the probe. He was further instructed that he could proceed at his own pace and that this was not a "timed" task. In both E- and 8-paced conditions, a correction procedure was employed and E manually recorded wrrect and incorrect responses. The E preset one of 18 randomly detennined sequences of nine eves and a probe for a trial. Each of the nine positions was probed twice in an I &-trial block, and one replication of the HI sequences pmvided four "tests" for each position in the 36 triah. 7 u ~ 6 c: ~ 5 u c: IV ~ 4 c: '" IV ::E 3 RESULTS AND DISCUSSION Figure I depicts per cent correct responses for E- and S-paced groups over the nine positions. An analysis of variance of these data yielded an F(l,28) = S.7, p < .01 for E-versus S-paced, F(8,224) = 15.S, p < .001 for serial position, and F(S,224) =3.1, p < .005 for the interaction. No explanation of the coincidence of performance at Position 2 seems obvious. Figure 2 presents mean interitem latencies for the S-paced group. Figure 3 showing individual latencies in four 9-trial blocks for six selected Ss, indicates that the averaged culVe in Fig. 2 is misleading. Five of these S8 show substantial changes in latency as a function of serial Psychon.~i.,1968, Vol. 12 (2) 2 23456789 Serial Position Fig. 2. Mean latencies of the S-paced aroup u. function of teriII potition. 55 5' 10 - I ....... 2 ~ 4 <:>-<1 3 , 13 52 .-. BLOCK 1 6 2 J. 4 .5 6 7 8 9 Senal FbSIlion 15 Fig. 3. Mean latencies of 6 individual Ss in the S·paced group in four 9·triaI blocks. position, i.e., SPE similar to those usually found for accuracy indices. The perfonnance of two other Ss was similar to that of these five. The performance of S four is quite representative of that of seven additional" Ss. For these eight Ss latencies are short, 100 90 80 If) ~ c: Short Latencies . - Long Latencies ..... - ..... ~ \ 8. 70 & ~ 60 :g \ "'--"""", \ \ \ \\ / / \ / \ ~ 50 ¥ \ \ I I I I ~ -V-"'" \ \ l.: I I I I \ \ / u { \ / I t) \ I I I \ ~ 40 8 20 10 in fact, hardly more than reaction time to the offset of the preceding stimulus. Moreover, the changes over serial position are negligible. In order to evaluate the effects of latency strategy, Fig. 4 contrasts accuracy scores of the seven Ss with long latencies with those of the eight Ss with minimal latencies. It is apparent that primacy perfonnance is markedly facilitated for the Ss showing long interitem latencies. On the Qther hand, recency is invariate for these different strategies. Figure 3 demonstrates the high intrasubject consistency which occurred in latency strategy. It was anticipated that Ss would develop a rehearsal strategy over trials which would be reflected as a changing latency pattern. This did not prove to be the case. In the main, Ss adopted a strategy initially and retained it throughout the task. Also, those Ss with minimal latencies initially continued in a similar manner over the four 9-trial blocks. Response latencies (interval between probe and choice response) for incorrect responses were appreciably longer than those for correct responses for all IS Ss in the S-paced group. These data indicate that performance in a sequentially presented STM task yields SPE which agree with previous findings (Ellis & Hope, in press). The fonn of SPE is dependent upon pacing conditions with an S·pacing condition facilitating primacy perfonnance. Recency is uninfluenced by type of pacing. Roughly half of the Ss under an S-paced procedure show an SPE in interitem latencies resembling the SPE typically found in accuracy data. Primacy performance of these Ss is subs (...truncated)