Feb 1, 1996 - list items were re-presented prior to recall so that only their order had to be ... Earley Gate, Whiteknights Road, Reading, RG6 2AL, U.K. (email: ...
THE QUARTERLY JOU RNAL O F EXPERIM ENTAL PSYCHOLOGY, 1996, 49A (1), 140± 158
Direction of Report in Spatial and Verbal Serial Short-term M em ory Paul Farrand and Dylan Jones Cardiff University of Wa les, Ca rdiff, U.K.
Four experiments examined the role played by item and order information in deter mining the effects of order of report of a sequence from short-term memory. Experim ents in which list items were re-presented prior to recall so that only their order had to be reported showed no differences in performance between the forward and backw ard directio n of report. This result was found w ith lists of auditory± verbal, visual± verbal, and spatial stimuli. W hen the list items were not re-presented, so that recall of both items and order was required, recall in the b ackw ard directio n of report was signi® cantly worse than in the forward direction of report, both in spatial and verbal tasks. The results point to the symmetry of inter-item associations, though only equivocally so, b ut they suggest stro ngly that the processes of spatial and verbal serial recall share many functional characteristics.
Everyday experience suggests that once a sequence of verbal items has been learned, recalling it in a reversed order is rather dif® cult. For example, ® rst attempts to report the alphabet in its reversed order are inept and error-prone. Even an un fam iliar sequence, su ch as a teleph one num ber, is dif® cult to recall in the reversed order. Formal tests of the phenomenon w ithin various laboratory tests of serial order are relatively comm on. From the very earliest em pirical investigations in experimental psychology, observations suggest that recall of a reversed sequence is poor, and that only relatively little inform ation about the reversed ord er is represented as th e person learns the sequence in the forward direction. So, re-learning a sequence in a reversed order takes less tim e than learning it de novo, but the saving is not app reciable and certainly not enough to suggest that the representation of order is symm etrical (Ebbin ghaus, 1885/1913). Even with appreciable savings when reversed order is called for, asymm etry of representation is logically not a necessary conclusion. G iven that serial order tasks com prise two major functional componentsÐ item in form ation and associative inform ationÐ it is logically possible th at the representation of associative information is indeed symm etrical, but
Requests for reprints should be sent to Dylan Jones, School of Psychology, Cardiff University of Wales, PO Box 901, Cardiff, CF1 3YG, U.K. Paul Farrand is at the Department of Psychology, University of Reading, 3 Earley Gate, Whiteknights Road, Reading, RG6 2AL, U.K . (email: P.A.Farrand@ Reading.ac.uk) Thanks are due to Phil Beaman, Andy Bridges, and Bill Macken for comments on an earlier draft of this paper. Alison Murray made crucial observations about the design of the spatial memory task. q
1996 The Experimental Psychology Society
D IR E C T IO N O F R E P O R T
141
that in the process of retrieval in reversed order item inform ation is som ehow less accessible. If nothing else, th is would be convenient, as it would allow the con siderable parsim ony and predictive power of symm etry as a global proposition that covers several facets of serial recall performance to prevail, with the subsidiary proposition of item memory being called upon only for the case of reversed recall. The TODAM model (see M urdock, 1982, 1983; M u rdock & Lewandowsky, 1986) embodies such sym metry and is successful both at furnishing a general account of serial order performance and at the same tim e providing a m eans by which poor perform ance in reversed recall together with detailed predictions about the serial order interactions can be enshrined (L ewandowsky & M urdock, 1989). The suggestion that associative information is bi-directional has, however, been criticized on the basis that this con clusion can only be reached in a sam pling without replacement imp lem entation of TODAM , in w hich previously selected item s are still available for recall (M ewhort, Popham & James, 1994; Nairne & Neath, 1994). In this simulation it is the item that sh ould have been re called that is removed rather than the item actually recalled, as would be usual in experimental studies (e.g. see Lewandowsky & M urdock, 1989). If an incorrect item is selected, it rem ains in the set of to-be-recalled item s and is available for later selection, and the item that should have been recalled is rem oved. U sing a simulation that assum ed that reported items are removed from the su bset of still to-b e-recalled items, the TODAM simulation could only generate a serial position curve that exhibited primacy, but n o recency (Nairne & Neath, 1994; but see Lewandwosky & Li, 1994). T he implementation of such a sam pling without replacement system within th e TO DAM model has been w id ely criticized on the grou nds that it does not re¯ ect the ways in which hum an subjects perform serial recall tasks (for a discussion see M ewhort et al., 1994; N airne & Neath, 1994). To overcome several of the criticisms levelled at the TO DAM m odel with respect to the direction of rep ort, an alternative class of theory assum es distinct qualitatively different processes for forward and reversed recall. This analysis is based on the effects of different kinds of interference on forward and reversed recall. Tasks that interfere with the formation of inter-item associations at stu dy disru pt forward but not backward recall, w hereas tasks that alter the visuo-spatial characteristics of the study m aterial impaired backward recall but n ot forward recall (Li & Lewandowsky, 1993, 1995). The pattern of resu lts suggests that forward recall is based on inter-item associations and backward recall is based on visuo-spatial representations. The presen t paper addresses th e issue of the symm etry of represe ntation via the following logic. P rovid ing the su bject at retrieval with all the item s from the to-berem embered setÐ effectively restricting the subject’ s resp onse to th e ordering of elem entsÐ should yield a relatively pure measure of th e strength and symm etry of inter-item associations. During recall for the order of th e elem ents, th e selected items will be excluded from the list of to-be-tem s, not the items that should have been selected, as in the sampling with replacement im plem entation of the TO DAM simulation. This will allow a judgement of whether in ter-item associations are sym metric, as is prop osed by th e TO DAM m odel, whilst circumventing m any of the criticisms of the TODA M simulation. If the inter-item associations are indeed symm etric in a task in which item information is provided, perform ance in the reversed order sh ould be the sam e as for the forward order, whereas in circumstan ces in which th e individual h as to
142
FARR AN D AN D JON ES
generate both the item s and their order, the u sual result of in ferior recall in the reversed direction should be observed. Aside from addressing the general issue of order of report, th e experim ents that follow address particular issues associated with presentation modality. Speci® cally, signi® cance is attached to the ® ndin gs of Isaacs and Vargha-K hadem (1989). They found the u sual result with verbal recallÐ retrieval in reversed order is worse than in forward orderÐ bu t with a serial spatial task performance was roughly the same for forward and reversed recall. Two possible interpretations can be set upon these results. Th e ® rst is to suggest that spatial representation plays a pivotal role in the ef® ciency of reversed recall. If, as Li and Lewandowsky (1993, 1995) suggest, spatial strategies underpin retrieval in reversed order, the relative superiority of reversed spatial recall can be explained in this w ay. A n alternative explanation is that the results stem from a peculiarity of the balance in recall for item and order in form ation in the verbal and the spatial tasks u sed by Isaacs and Vargha-K hadem (1989), not from the differences in spatial and verbal codingÐ th at is, the m odality of presentation. Speci® cally, we suggest th at in their study the verbal task called u pon the recall of both item and order inform ation but the sp atial memory task required only order inform ation, and it is this difference that determ ines the effect of direction of report, n ot th e difference in class of representation used fo r the verbal and sp atial tasks. The verbal task used by Isaacs and Vargha-K hadem (1989) required the spoken recall of auditorily presented digits, so that the subject w as required to recall vocally the items in either the order in which they were p resen ted or in reverse order. Although the d igits were drawn from the sam e ® xed set, they were not re-presented to the subject at retrieval, and subjects had to produce the item and report its ord er. In contrast, in the spatial task, known as the Corsi B locks task (see e.g. De Renzi & N ichelli, 1975), subjects were asked to base their rep ort on the order in which a series of wooden blocks were to uched by the experimenter. The blocks rem ained in full view of the subject thro ughout th e presentation and retrieval phase. At retrieval, the subject was required to poin t to the blocks in their order of presentation or in reversed order. Because the items were available to the su bject throughout the trial, the retrieval test was arguably one of m emory for the order of the item s aloneÐ the subject did not also have to recall the item s. The demands made by retrieval seem to be quite different for the verbal and spatial task. If the differences in the direction of report shown by Isaacs and Vargha-Khadem (1989) arose from qualitatively different demand s im posed on m emory at retrieval by the spatial and verbal task, then their ® ndings cannot be taken as evidence for functionally d istinct dom ains of spatial and verbal representation. Generally, therefore, the following experiments centre up on the relative contribution of item and order inform ation to forward and backward serial recall. Tasks for which order inform ation only is required should reveal the degree of sym metry in in ter-item associations ind ependent of the con tam inatin g effect of item inform ation. In term s of performance, th erefore, we m ay expect th at in this case reversed retrieval will approach the perform ance of forw ard retrieval. From the two-p rocess view of Li and Lewandowsky (1993, 1995) it is possible to predict that the balance of item and order inform ation should be irrelevant; the extent to which spatal strategies may be employed is the primary determ inant of the effect of retrieval direction. A spatial task should naturally invoke
D IR E C T IO N O F R E P O R T
143
sp atial strategies and thus have characteristics of perform ance th at resemble reversed verbal recall. For the present series of experim ents, a set of tasks was d eveloped that called for the retrieval of either order inform ation alone or a combination of item and order information. Th ese were prod uced in versions in which th e to-be-recalled m aterial was either verbal or spatial. The aim of the series w as to use these tasks to dem onstrate w hether the differences in the direction of report found betw een the verbal and sp atial tasks depend upon retrieval requirem ents (the retrieval hypoth esis) or the modality of th e stimuli (the modality hypothesis). E X P E R IM E N T 1 Three versions of a serial recall task were developedÐ using auditory± verbal, spatial, or visual± verb al to-b e-rem embered stimuliÐ w ith which to test the prediction that when only order has to be reported, forward and reversed directions of report yield roughly sim ilar levels of performance. Generally, the task involved mem ory for a sequence of seven stimuli that were presented serially. After the presentation of the last stimulus in the sequence, the screen went blank for a 10-sec retention interval, after which the seven stimuli were re-presented simultaneously. The subject w as required to report the order of presentation of th e stimuli by usin g a mouse to mark the items, as indicated by instructions, either in the order in wh ich they h ad been presented , or in th e reverse of the order of presentation. At retrieval, imm ediate feedback was given about which items in the represented list had been selected. Subjects undertook tests of recall with each of the three types of task and knew in advance of each trial w hich direction of report was required. Given that in each of the tasks retrieval called for order inform ation alone, it was predicted by the retrieval hypothesis that the effect of recall direction should be the sam e in the three types of taskÐ namely, that there should be n o difference between forward and backward recall. O n the m odality hyp othesis it wou ld be predicted th at backward recall should be worse, but only with the verbal tasks. The spatial task employed in Experiment 1 was designed to minimize the likelihood of verbal recoding, and therefore to be, as far as possible, a pure measure of serial spatial memory. Features of the task designed to achieve this aim were (a) a high degree of spatial uncertainty (the items could appear at any point in a 350 3 350 matrix), so that a simple set of verbal co-ordinates could not be employed to re-code the stimuli; (b) the to-be-recalled items were not drawn from a ® xed set; rather, they changed from trial to trial, so that verbal labels could not become associated with particular locations; and (c) there were no extraneous cues (such as grid lines) that could serve as a basis for recoding the material in terms of verbal coordinates (see Jones, Farrand, Stuart, & M orris, 1995, for a further discussion).
M e th o d S u b je c ts Eighteen m ale and fem ale undergraduates, reporting normal or corrected-to-norm al vision, volunteered to participate in this experiment and were paid an honorarium.
144
FARR AN D AN D JON ES
M a te ria ls Three versions of a serial recall task were employed. The spatial task and the visual± verbal task had been used previously by Jones et al. (1995). Spa tia l Ta sk. A sequence of seven black dots 0.5 cm in diameter was presented randomly, one dot at a time, within a 350 3 350 pixel matrix (taking up a white area of the screen 16.5 cm 3 16.5 cm). During stimulus preentatio n there was o nly one dot present on the screen at any time. For each trial, the co-ordinates for positio ning the dots were randomly generated in software, but constrained by the rule that if any randomly generated dots were to appear closer than 60 units (2 cm) in either axis of the matrix, another set of dot-co-ordinates w as generated. After the dots were presented, the screen was blank for 10 sec, during which the subjects were expected to rehearse. To prom pt retrieval, all seven dots were re-presented simultaneously upon a white screen in the sam e spatial locations in w hich they had originally been presented. T he subject responded by using the mouse to locate and click on the dots in the order speci® ed in advance by instructio nsÐ that is, either forw ard or backward. To indicate when a dot had been selected, the shading of the dot lightened appreciably and rem ained that way until the w hole response sequence had been produced. Visua l± Verba l Ta sk. On each trial a different random sequence of seven co nsonants, drawn from the set F, K, L, M , Q , R , and Y, w as presented in 30-point Geneva bold upper-case typ e in the centre of the com puter screen. Items were presented individually, and at the end of the sequence the screen went blank. D uring stimulus presentation only one co nsonant was present at any tim e. At retrieval, follow ing a 10-sec retention interval, all seven consonants were re-presented simultaneously, each within a drawn box, in a random sequence displayed left to right on the screen. The sequence of consonants displayed at retrieval was random iz ed o n each trial. As with the spatial task, the subject used the mouse to select each consonant in the dir ection of report speci® ed by the experimenter. Upon selection, the shading w ithin each box changed and remained that way until all the items in the respo nse sequence had b een selected. Auditory± Verba l Ta sk. The same seven co nso nants as those used for the visual± verbal task were recorded in a male voice and digitized to eight-bit resolution at a sampling rate of 22 kH z using a M acRecorder analogue-to-digital co nverter. Utterances were edited using SoundEdit Pro digital editing software and stored for use in a SuperCard environment. The randomiz ed sequences were played back by the computer via Sony Cd 250 headphones at approxim ately 65 dB(A), as measured by a sound-level m eter from an arti® cial ear. After a 10-sec retention interval, the co nso nants were represented in visual form in random order to prompt recall, and, as for the visual± v erbal task, the random order in this display changed on each trial. Subjects m arked the order of the item s using a mouse, just as for the visual± verbal version of the task. In the spatial and visual± verb al versions of the task, each item w as presented for a duration of 1 sec, followed by an interval of 1 sec during w hich the screen was blank. Auditory± verbal item s were presented for a duration of 750 msec, followed by 1250 msec of silence. T his resulted in a rate of presentation of o ne item per 2 sec for each of the three versio ns of the task. All materials were presented on an Apple M acintosh IIvi microcomputer using SuperCard prog rams.
P ro c e d u re Subjects were seated approxim ately 50 cm from the computer screen upo n w hich w ritten instructions and the tasks were presented. Subjects were told about the general form of the task and the requirement for serial recall in the order speci® ed by the instructio ns displayed during the task. Prior to the beginning of the experiment proper, subjects were practised (for 6 trials) in the use
D IR E C T IO N O F R E P O R T
145
of the m ouse. The three versions of the serial recall taskÐ visual± verbal, auditory± verbal, and spatialÐ were presented in blocks of 32 trials. W ithin each block, forward recall was required for half the trials and backward recall for the other half. The order of blocks and halves within blocks was randomized. The presentation of trials was self-paced; sub jects clicked o n a green ``butto n’ ’ o n the computer screen to indicate when they were ready to be gin each trial. In each trial, after the presentation of the seventh (last) stimulus, the screen w as blank for 10 sec, during which subjects were expected to rehearse. After the retention interval, items in the sequence were re-presented simultaneously. T his served as a prompt for the subject to be gin retrieval by marking with a mousedriven pointer on the screen. The w ritten instructions encou raged subjects to recall as quickly and accurately as possible and informed them that once an item had be en selected, it could not be altered. Subjects were tested individually, and the experimental session lasted 55± 60 min. T he experimenter was present throughout the session and ensured that subjects followed the instructions.
R e s u lts a n d D isc u s s io n The results were scored w ith respect to serial position: Only a correct response in its appropriate serial position was scored as correct. A 2 (direction of report: forward and backward) 3 3 (task typ e: spatial, visual± verbal, auditory± verb al) 3 7 (serial position) repeated-m easures A NOVA w as performed on the error data. There was no effect of the direction of rep ort, F(1, 714) = 0.43, M S e = 8.05, p > 0.05; backward direction of report was perform ed as well as th e forw ard direction of report (see Figure 1, in which ``serial position’ ’ is used to refer to the direction of report, not th e order of presen tation, in order to facilitate the comparison of fo rward and backw ard direction of report). T here were highly signi® cant main effects of task type, F(2, 714) = 72.19, M S e = 8.05, p < 0.0001, and serial position , F(6, 714) = 25.55, MS e = 8.05, p < 0.0001, but, imp ortantly for the hypothese s u nder test, th ere was no signi® cant Task Type 3 Direction of Report interaction, F(2, 714) = 0.73, M S e = 8.05, p > 0.05. S im ilarly, interactions with serial position were also non-signi® cant: Task Type 3 Serial Position, F(1 2, 714) = 1.58, M S e = 8.05, p > 0.05, D irection of Report 3 Serial Position, F(6, 714) = 0.65, M S e = 8.05, p > 0.05. Planned contrasts between the forward and backward direction of report for each task type support the conclusion that the backward direction of report w as perform ed as accurately as the forward direction of report, ir resp ective of the m odality of presentation. There was no difference between the direction of report on the spatial, F(1 , 714) = 1.6 5, MS e = 8.05, p > 0.05, visual± verbal, F(1, 714) = 0.06, M S e = 8.05, p > 0.05, or auditory± verb al, F(1, 714) = 0.16, M S e = 8.05, p > 0.05, tasks. F urther planned contrasts carried out between d ifferent task types failed to reveal signi® cant differen ces between recall performance on the visual± verb al and auditory± verb al task, F(1, 714) = 0.21, MS e = 8.05, p > 0.05. H owever, both the visual± verb al and auditory± verbal tasks were performed signi® cantly better than was the spatial task, F(1, 714) = 112.92, M S e = 8.05, p < 0.0001, and, F(1, 714) = 103.45, M S e = 8.05, p < 0.0001, respectively. The results of Experiment 1 suggest that if the to-b e-rem embered items were represented, so that the subject was requ ired to recall only their order, backward recall was performed as accurately as forward recall. Th is ® nding supports the retrieval hypothesis and suggests, amongst other things, that the differences reported w ith the spatial and verbal d irection of report in the Isaacs and Vargha-K hadem (1989) study arose from a
146
FARR AN D AN D JON ES
F IG . 1 .
Experiment 1. Manipulation of the direction of report of spatial, auditory± verbal, and visual± verb al order information. Maximum error = 16. Serial position is plotted for the direction of report, not the order of presentation.
change in the balance of item and order in form ation in the two tasks. On the assumption that the tasks are a relatively pure measure of order information , the results also suggest strongly that the prediction that arises from the T ODAM model of symmetry in interitem association is correct. O f further signi® cance is that the assum ption of sym metry in inter-item associations is supported by a task in which the selecte d items were eliminated from the subset of still to-be-recalled items, therefore overcoming m any of th e criticisms levelled at th e sampling w ithout replacement TO DAM simulation (M ewhort et al., 1994; Nairne & N eath , 1994). M odality, and hence the degree to which spatial represen tation can be d eployed to aid reversed retrieval, does not, however, appear to be an im portant determ inant of perform ance (Li & Lewandowsky, 1993, 1995). Before concluding that the outcom e supports the re trieval hypothesis prima fa cie, it is important to acknowledge that there are several differences between the procedure adopted by Isaacs and Vargha-Khadem (1989) and that of Experim ent 1 which could account for the different outcom e, am ong them that for their auditory± verbal task the response m ode was spoken rather than by selection using a m ouse. The possible role of
D IR E C T IO N O F R E P O R T
147
response m odality is among th e factors explored in Experim ent 2. Further experimental evidence bearing on the role of item and order information within this paradigm w as also needed in view of the fact that Experiment 1 con® rm ed the retrieval hypoth esis by means of a null effectÐ an outcome that could have been the result of a lack of experim ental power. E X P E R IM E N T 2 Based on th e ® ndings of Experim ent 1, it was predicted that when the to-be-remembered item s were not re-presented prior to retrieval, so that subjects were required to recall both item and order information, the results would be of the sam e general form as those of Isaacs and Vargha-K hadem (1989). The focus of Experim ent 2 was restricted to verbal recall in which a m odi® ed version of the auditory± verb al task used in Exp eriment 1 was used. Instead of re-presentin g th e to-be-remembered item s at retrieval, subjects were obliged to generate their responses, there by requiring the retrieval of both item and order information. Also, in Experiment 2 the role of response modality was exp lored by contrasting the effects of written with spoken recall. M e th o d S u b je c ts Eighteen m ale and fem ale undergraduates, reporting normal or corrected-to-norm al vision, volunteered to participate in this experiment and were paid an honorarium.
M a te ria ls The same seven consonants as those used for the verbal tasks in E xperiment 1 were employed.
P ro c e d u re As for Experiment 1, subjects were required to recall, in either a forward or backward order, seven auditorily± presented consonants. Although the mode of presentation and retention interval were exactly the same as for the auditorily presented verbal material in Experiment 1, the method of response production was different. Subjects were required either to speak their responses aloud or to write their responses down, from left to right, on a response blank. As before, they were instructed to recall the sequence as quickly and accurately as possible and that once a response had been made, it should not be altered. The instructions encouraged subjects not to leave any items out and to guess if necessary. A number of precautionary measures were set in place with the aim of making the dem ands of response production as similar as possible for written and spoken recall (see M orris & Jones, 1987). For written recall, subjects moved a card over previous responses within a recalled sequence to prevent their use in supporting subsequent recall, a facility not available with spoken recall. To prevent any possible artefactual effects of the motor aspects of this method of report being con® ned to written recall, subjects were instructed to mim ic this movement during spoken recall by moving a card over blank response boxes after each item was spoken. This provided subjects with additional feedback about their prog ress in producing the response sequence, as well as slowing down spoken response speed to a rate approxim ating that of the written responses. Each subject undertook four blocks of 16 trials. Conditions were blocked, and the order of presentation of the blocks was randomiz ed. Subjects were informed of the
148
FARR AN D AN D JON ES
required direction of report prior to the beginning of each trial block. The experimenter was present throughout the experimental session, which typically lasted 40± 45 min.
R e s u lts a n d D isc u s s io n A strict serial recall criterion was adopted for the scoring of resp onses, as in Experiment 1. A 2 (direction of rep ort: forward and backw ard) 3 2 (modality of retrieval: spoken and writte n) 3 7 (serial position) repeated-measures ANOVA was performed on the serial order error data. As predicted, there was a signi® cant m ain effect of the direction of report, F(1, 476) = 25.16, MS e = 14.25, p < 0.0001, re¯ ecting the superior level of recall performance w ith a forward than w ith a backward direction of report (see F igure 2, in which serial position is plotted for the direction of report, not the order of p resentation). Furthermore, spoken recall was inferior to written recall, as shown by the highly signi® cant effects of the m odality of retrieval, F(1, 476) = 12.29, M S e = 14.25, p < 0.0001. There was additionally a signi® cant effect of serial position, F(6, 476) = 12.13, M S e = 14.25, p < 0.0001; however, the S erial Position 3 Direction of Report, F(6, 476) = 1.59,
F IG . 2 .
Experiment 2. M anipulation of the direction of report and mode of recall of auditory± verbal item and order information. Maximum error = 16. Serial position is plotted for the direction of report, not the order of presentation.
D IR E C T IO N O F R E P O R T
149
MS e = 14.25, p > 0.05, and the Serial Position 3 M odality of Retrieval, F(6, 476) = 0.75, MS e = 14.25, p > 0.05, interactions were n on-signi® cant. Furthermore, the Direction of Report 3 M odality of Retrieval interaction w as non-signi® cant, F(6, 476) = 0.02, M S e = 14.25, p > 0.05. Planned contrasts between the direction of report for each modality of retrieval su pport the prediction that recall performance w ith backward recall w as signi® cantly worse than with forward, both with sp oken, F(1, 476) = 11.03, M S e = 14.25, p < 0.01, and w ritten, F(1, 476) = 14.23, M S e = 14.25, p < 0.001, recall. As anticipated by the retrieval hypothesis, the results of Experiment 2 show that if a su bject was required to recall both item and order information, backward recall performance was signi® cantly worse than forw ard recall perform ance. The results do not sup port the possibility that the crucial feature of the procedure in the stu dy of Isaacs and VarghaKh adem (1989) was the spoken recall used in the verbal task: In Experiment 2 of the current series, backward recall w as worse with both sp oken and written re call. Instead, Experiment 2 provides further support for the notion th at th e differences in d irection of report (Isaacs & Vargha-K hadem, 1989) arose from differences in th e dem and for item and order information in the verbal and spatial tasks.
E X P E R IM E N T 3 Although it is possible to be relatively certain that the effect of direction of report does not depend on th e m odality of retrieval, we can be less su re whether the effect of direction of report is one con® ned to auditory presen tation or is a general characteristic of verbal material and could therefore also be found with visu ally p resented verb al material. T hat a reversed direction of report w as particularly easy with visually presented m aterial is su ggested ind irectly by som e ® nd ings using free recall tasks. W ith visual presentation of verbal material, it has been claimed that a backward direction of report is in fact ``preferred’ ’ (Nilsson, Wright, & M urdock, 1979). W hen free recall of visual± verbal material is required, subjects begin recall w ith the last item presented and proceed in the reverse order to that of presentation, w hich results in a better level of perform ance with a backward than with a forward direction of report (M adig an, 1971; Nilsson et al., 1979). A s a check on the possibility that the damaging effect of reversed recall was con ® ned to auditory presentation, in Experiment 3 the sam e design was adopted as for Experiment 2, w ith the difference that recall of visu al± verbal sequences was required instead. M e th o d S u b je c ts Eighteen m ale and fem ale undergraduates, reporting normal or corrected-to-norm al vision, volunteered to participate in the experiment as part of the requirement for course credit.
M a te ria ls The same seven consonants as those used in Experiment 1 were em ployed.
150
FARR AN D AN D JON ES
P ro c e d u re As for Experiment 1, after a 10-sec retention interval, subjects were required to recall seven visually presented consonants in a forward or backward direction of report. T he m ethod of presentation was exactly the same as for the visual± verbal task used in Experiment 1, and the requirement to recall item and order inform ation w as the same as for Ex periment 2. Subjects were required to recall the visually presented consonants either by writing the respo nses down or by speaking them aloud. Experiment 3 employed the sam e technique of concealing previous responses. As with Experiment 2, each subject received four blocks of 16 trials. Conditions were blocked, and the order of presentatio n of the blocks w as randomiz ed. Subjects were informed of the required direction of report prior to the be ginning of each trial block. T he experimental session lasted 40± 45 min, during which the experimenter w as present to observe the subject’ s perfor mance.
R e s u lts a n d D isc u s s io n As before, responses were scored in terms of a strict serial order criterion: A response was scored as being correct when it appeared in its correct serial p osition. A 2 (direction of report: forward and backward) 3 2 (modality of retrieval: spoken and written ) 3 7 (serial position) repeated-m easures ANOVA was perform ed on the error data. T here was a signi® cant m ain effect of th e direction of report, F(1, 476) = 49.91, M S e = 9.54, p < 0.0 001, indicating the better level of recall performance with a forward than with a backward direction of report (see F igure 3). There was a signi® cant effect of serial positio n, F(6, 476) = 17.59, M S e = 9.54, p < 0.0001, but no difference in recall performance dependent upon th e m odality of retrieval, F(1, 476) = 0.7 7, M S e = 9.54, p > 0.05. The interaction of Serial Position 3 M odality of Retrieval, F(6, 476) = 0.41, M S e = 9.54, p > 0.05, and Serial Position 3 Direction of Report, F(6, 476) = 0.40, M S e = 9.54, p > 0.0 5, were non-signi® cant. Furthermore, the Direction of Report 3 M odality of Retrieval interaction was non-signi® cant, F(1, 476) = 0.15, M S e = 9.54, p > 0.05. In Figure 3, serial position is plotted for the direction of report, not the order of presen tation. Planned contrasts carried out between the two directions of report for each m odality of retrieval support the suggestion that the backward direction of report was signi® cantly worse than th e forward direction of report, with both spoken, F(1, 476) = 22.33, M S e = 9.5 4, p < 0.0001, and written, F(1, 476) = 27.72, M S e = 9.54, p < 0.0001, recall. T he method of retrieval, however, had no effect upon recall perform ance, w ith either a forward , F(1, 476) = 0.80, MS e = 9.54, p > 0.05, or backward, F(1, 476) = 0.12, M S e = 9.5 4, p > 0.05, direction of report. The effect of d irection of report was similar for those visual± verbal and auditory± verbal tasks in wh ich recall for item and order information was required. T he only difference of n ote was th at written recall of auditory± verbal material was better than written or spoken recall of visual± verbal materialÐ a ® nding that has precedent (see e.g. M orris & Jones, 1987). That there were no other signi® cant differences between the direction of report of auditory± verbal and visual± verbal inform ation suggests th at the effect d oes not depend on the modality of input, at least in the verbal domain. The outcome of Experiment 3 showed that wh en verb al m aterial was presented visually, forward recall was su perior to backward recallÐ a result that fails to sup port the notion that backward direction of report was ``preferred’ ’ for visual± verbal m aterial
D IR E C T IO N O F R E P O R T
151
F IG . 3 .
Experiment 3. Manipulation of the direction of report and mode of recall of visual± verbal item and order information. Maximum error = 16. Serial position is plotted for the direction of report, not the order of presentation.
(M adig an, 1971). T he ® ndings of Experim ent 3 reinforce the view that th e disruption that arose with th e backward recall of auditory± verbal material in the study of Isaacs and Vargha-K hadem (1989) was not a product of the m odality of presentation nor the modality of retrieval, but arose instead from the obligation to recall item as well as order inform ation. Experim ent 4 was a fu rth er test of the idea th at th e requirem ent to recall item and order inform ation results in the disrup tion of backward recall, this time on a test of spatial recall. If the demands of retrieval dictate the effect of direction of report, then it sh ould be possible to assem ble a test of mem ory for spatial m aterial that requires retrieval of item and order information and to show signi® cant differences in the direction of report, sim ilar in kind to those already shown for verbal m aterial.
E X P E R IM E N T 4 In terms of the form of the to-be-remembered m aterial, the task used in Experiment 4 was sim ilar to that used in Experim ent 1, however, in Experiment 4 the to-b e-rem embered item s were not re-presented p rior to recall, instead, subjects, in addition to report-
152
FARR AN D AN D JON ES
ing the order of items, also had to generate their position. Subjects recalled the stimulus sequence by using the mouse to m ark on a blank screen the spatial location in which each stimulus was presentedÐ either in the order in which they were presented or in the reverse ord er. If the hypothesized effect of retrieval demands is correct, this version of the task should produce effects of direction of report closely resem bling those found with verbal tasks, nam ely inferior perform ance for reversed recall relative to forw ard recall. M e th o d S u b je c ts Eighteen m ale and fem ale undergraduates, reporting normal or corrected-to-norm al vision, volunteered to participate in this experiment and were paid an honorarium.
M a te ria ls The same spatial stimuli as those presented in Experim ent 1 were used.
P ro c e d u re As with Experiment 1, subjects were required to retrieve, after a 10-sec retention interval, the order of presentation of the seven spatial stimuli in a forward or backward direction of report. Although the presentation of the stimuli w as exactly the same as for the spatial task used in Experiment 1, the method of recall was different. Instead of re-presenting the dots, at retrieval the subjects were presented with a blank screen. In this case the recall sequence w as produced by using a mouse to point and mark the position in which each stimulus was originally presented on the blank screen. Subjects were instructed to recall either forward or backward, in blocks of trials. M arking of locatio n on the screen was achieved by pressing the butto n on the mouse, upon w hich the computor returned a dot of the same dim ensions as a dot in the stimulus set. On each trial, at retrieval the dots stayed o n the screen until the whole sequence was shown, and then the screen went blank before the next trial. Following a 10-sec retentio n interval, subjects were prom pted for recall by the appearance of a vertical display consisting of seven panels, one panel for each of the dots to be recalled, along the right-hand edge of the otherw ise blank screen. As the subject placed each dot on the screen, so another panel changed in colour (to g reen), to indicate the num ber of items that had been recalled, and accordingly the display provided information about the number still to be recalled. Subjects were instructed to recall as quickly and accurately as possible and infor med that o nce an item had been marked, it could not be altered. Subjects received two blocks of 16 trials, and the ordering of the blocks w as randomiz ed. Prior to the beginning of each block, subjects were instructed on the required directio n of report. T he experimental session lasted 30± 35 min, and the experimenter was present to check o n com pliance with instructions.
R e s u lts a n d D isc u s s io n The scoring of subjects’ responses was based on the ``city block’ ’ deviation of the response co-ordinates from the co-ord inates of the stimulus appropriate for that point in the response sequence. T he displacement error was taken as the sum of the deviation of units in azimuth and in elevation of the retrieved item from the co-ordinates for the
D IR E C T IO N O F R E P O R T
153
position of the correct item. Values were retu rn ed in term s of the position within a 350 3 350 u nit m atrix that represented the screen on which responses were produced. In the forward direction of report the ® rst response was m easured in relation to the position of presentation of the ® rst target in the sequence, whereas for a backward report the ® rst response was m easured in relation to the position of presentation of the ® nal target in the sequence. A 2 (direction of report: forward and backward ) 3 7 (serial positions) repeated measures ANOVA was perform ed on the city block error data. As predicted by the retrieval hypothesis, in this task backward recall was signi® cantly worse than forward recall, F(1, 17) = 21.50, M S e = 2868.5, p < 0.001. There was also a signi® cant main effect of serial position, F(6, 102) = 94.35, M S e = 780.15, p < 0.0001, and a sign i® cant D irection of Report 3 Serial Position interaction, F(6, 102) 3 23.8, M S e = 922.98, p < 0.0001. This interaction was probably due to a better level of recall perform ance at Serial Positions 1 and 2 when recalling in a backward order, F(1, 102) = 27.94, M S e = 922.98, p < 0.0001. (See Figure 4, in which serial position is plotted for the direction of report, not the order of presen tation.) Forward
F IG . 4 .
Experiment 4. M anipulation of the order of report of spatial item and order infor mation (city block error). Serial position is plotted for the direction of report, no t the order of presentation.
154
FARR AN D AN D JON ES
recall was, however, signi® cantly better than backward recall between Serial Positions 3 and 7, F(1, 102) = 38.88, MS e = 922.98, p < 0.0001. The results of Experiment 4 indicate that w hen subjects were required to recall item and order inform ation in a test of serial spatial m emory, backward recall was signi® cantly worse than forward recall. The superiority of recall for the ® rst two items reported in a backward order was probably due to these item s being free of overwriting from subsequent items, by bein g presented last and recalled ® rst (for similar results see Broadbent & B roadbent, 1981; Ph illips & Ch ristie, 1977). The results of Experim ent 4 indicate that when subjects have to recall both item and order inform ation in a test of serial spatial memory, results sim ilar to those found with verb al tasks may be obtained. That is, a spatial task shows inferior recall with a backward direction of reportÐ a result that is very similar to th at obtained with verbal tasks with similar retrieval characteristics. An important reservation needs to be m ade in connection with the in terpretation of the city block metric. U nlike the more straightforw ard serial error score, city block error is an ir resolvable composite of ordinal and spatial error components. Hence, the m easured error m ay arise from producing the correct item in the sequence but displaced from its correct position on the screen; however, the same m argin of error could arise because the response m ay be out of sequence, so th at the item may be displaced both from its correct sp atial position and from its ordinal position. Although in principle it would be u seful to understand the probable functional signi® cance of th e interplay of these sources of error, there is no feasible way of separating them, and therefore th e city block score should always be consid ered to be an amalgam of their effects. H owever, this dif® culty is relatively m inor and should n ot detract from the general force of the results using the city block metric in resolving con¯ ict between m odality and retrieval hypotheses. T he results of Experiment 4 are complementary to and entirely consistent with the others of the set; their signi® cance is setting in place a key elem ent in the logic of the experimental seriesÐ namely, that poor perform ance in backward recall, hitherto a characteristic con® ned to verbal recall tasks, m ay also be exhib ited with a sp atial task. Given the pattern of results from the series as a whole, it seem s reasonable to conclude th at the retrieval demandsÐ rather than the modality of the to-be-rem embered material dictatesÐ the effect of direction of report. G E N E R A L D IS C U S S IO N The resu lts of the foregoing series can be taken as support for the symm etry of in ter-item associations, although th is is not the only le gitim ate interpretation. By showing that the reliance on item information is critical in the symm etry of recall, th e results also suggest that qualitatively different strategies are not used for forward and reversed recall. T hat the effect of direction of report fails to interact with m odality of pre sentation further points to the fact that spatial strategies do not underpin reversed recall (pa ce Li & Lewandowsky, 1993, 1995). At a m ore speci® c level, the results of th e series provide su bstantial support for the notion that the differences in the stu dy of Isaacs and VarghaKh adem (1989) between the spatial and auditory± verbal direction of report arose not from an effect of th e m odality of the stimuli, bu t from the requirem ent to recall item and order information. W hen the same task demands are made upon m emory retrieval, similar
D IR E C T IO N O F R E P O R T
155
results are obtained with the recall of both spatial and verbal tasks. If th e to-b e-rem embered spatial and verbal items were re-presented p rior to recall, so that only their order of presentation had to be recalled, the backward direction of report w as sim ilar to the forward direction of report (Experim ent 1). W hen the su bject h ad to generate an item as well as report its order, backward recall was signi® cantly worse than forward recall (Experim ents 2, 3, and 4). There are a number of possible problem s of interpretation in the fo re going series. One is the p ossibility that the spatial task behaves rather like the verbal task only because the su bjects recode the items using a verbal strategy. In general, therefore, it could be argued that the spatial task used by us does not rep resent processing in the spatial domain. The sp atial task used in the current series was used previously in a study by Jo nes et al. (1995), and th e results of that study offer corroboration to the assumption that the spatial tasks were verbally recoded. Evidence takes th e form of the pattern of cross-m odal interference of sp atial tasks by verbal second ary tasks (irrelevant speech, mouthed articulatory suppression). Irrelevant speech disrupted serial recall of spatial m aterial, whereas a spatial secondary task (M oar box tracking) disrupted serial verbal recall (Jones et al. 1995)Ð a pattern of results that can be explained by supposing that spatial and verbal dom ains of memory share m any fu nctional characteristics. G iven also that spatial and verbal shortterm memory are subject to the sam e limitations in term s of list length and retention interval and exhibit sim ilar-shaped serial position curves (Jones et al., 1995), this supports the notion that both spatial and verbal short-term mem ory share a comm on rehearsal mechanism . Given th at the subject has to generate item and ord er information, wh at is the critical feature that makes backward recall more dif® cult in a verbal task? One feature m ay be the need to deploy executive resources for the verbal task when the direction of report is backward (Scho® eld & Ashm an, 1986). Smyth and Scholey (1992) su ggested a view rather sim ilar to thisÐ that whereas executive resou rces were required to reverse the order of presentation of the verbal items, additional executive resou rces were not required to reverse the order of presentation of th e spatial C orsi Blocks items. In the Corsi B locks task, they argue, the spatial item s are encoded m erely in relation to th e position of the other items present in the display (S myth & Scholey, 1992) rather than in terms of som e stored representation of a letter or word. This rather suggests that the d ifferences between the verbal and sp atial m aterials arise because of differences in encoding. H owever, the evidence available from the curren t series of stu dies suggests that the d ifferences between spatial and verbal tasks are conn ected with factors at retrieval, not at encoding. Certainly, the effect suggested by Smyth and S choley (1992), that the subject stores relative position inform ation, is likely to be m ore germane to the Corsi B lo cks tasks in which th e items remain in view during the presentation phase than in the tasks used in the current paper in which only single item s were presented at a tim e, for which coding of su ch relative judgements is made less likely. As the fo regoing discussion suggests, it is assumed implicitly that the conditions of encoding are equivalent for all experim ental cond ition s, particularly for the two directions of report. In each case, cueing of order of report was done before the list was presented, and this may have prom pted the subject to adopt distinctly different rehearsal strategies for the different retrieval cond itions. Typical subjective reports of task strategies suggest
156
FARR AN D AN D JON ES
that this is unlikely, however. T he dominant strategy w as serial rehearsal, regardless of instruction condition or task type. Th ere is evidence to suggest that differences do arise in the direction of report when the required recall order is either p re-cued or post-cued (H in richs, 1968; N ilsson et al., 1979). It is on ly the recall of visually presen ted verbal material th at is affected by the change from p re-cueing to post-cueing, however, leaving the direction of report of auditory± verbal m aterial largely unaffected (Nilsson et al., 1979). If strategies are used to reverse the order of report du ring encoding, therefore, it would appear th ey are used only w hen the verbal material is visually presented. Differences in the order of report of auditory± verb al m aterial and spatial m aterial would therefore still need to be accounted for. If the important distinction between the effects of backward and forward recall is at retrieval, then what precisely is the factor resp onsible for the effect? Although up to this point we have sustain ed the functional distin ction between m emory for items and their order and, further, mapped these onto our various manipulations of th e task, from another viewpoint this might be seen as questionable. U ndoubtedly, in the spatial task, the representation of the items at retrieval (Experim ent 1), drawn as they are from a virtu ally in ® nite set, w ill minimiz e the need to retrieve item inform ation. But, in the verbal task the set is ® xed and can be learned very easily w ithin the duration of a typical experimental session. If this is tru e, the claim in the verbal case that re-presenting the item s at retrieval gives the subject item information that would otherwise be absent m ight be seen as dubious. Such considerations might be se iz ed upon by those who are rather more agnostic about the distinction between item and order inform ation. However, the empirical evid ence for such a distinction is strong (e.g. see Bjork & Healy, 1974; D etterm an & B rown, 1974; Devine, B urke, & Rohack, 1979; Donald son & G lathe, 1969; Estes, 1972; Healy, 1974, 1982; Rohrm an & Jahn ke, 1965; Zimm erm an & Un derwood, 1968). Evidence for this distinction takes two m ajor form s: (1) that item and order inform ation decay at different rates in bo th human (Bjork & H ealy, 1974; Shiffrin & Cook, 1978) and animal p opulations (M aki, Beatty & Clouse, 1984), and (2) that there are differences in the sh ape of th e serial position curves obtained when recalling either item or order information (B attacchi, Pelam atti & U miltaÁ, 1990; H ealy, 1974, 1982). An alternative view of the effect of direction of recall and m odality of m aterial may be couched in term s of a retrieval strategy. O ne pivotal feature of the versions of the task in which the stimulu s set was re-presented at retrieval is that shading was used to indicate which item s had been reported. This meant that as the respon se sequence was generated, the set of response alternatives became progressively sm aller. So the effect of re-presenting the item s w as not one of reducing the uncertainty about the stimulu s setÐ rather, it had som ething to do with the availability of in form ation about item s in a response sequence th at had already been selected. One possibility is that even when subjects are asked to recall in a backward d irection, the list is rehearsed in the forw ard order and, when p romp ted to recall, subjects assemble the reversed list. This is a relatively dif® cult undertakin g, p articularly in knowing wh ich response has already been pro duced. A rguably, this dif® culty is exacerbated when the set from which the stimuli are drawn is small, as th e proactive interferen ce from previous trials will be relatively high. T he pattern of data suggests that having some means of knowing which responses have been pro duced is relatively more helpful for verbal material when recalling backwards. However, why th is is
D IR E C T IO N O F R E P O R T
157
a particular dif® culty for verb al and not for spatial sequences is still a m atter for conjecture. Symm etry of inter-item associations is suggested by the results, but rehearsal strategies could also explain the outcome, albeit post hoc. M odality of the task is not crucial to the effect of ord er, only the balance of memory for item and order inform ation. T he experiments support the notion that both sp atial and verbal information share a com mon rehearsal mechanism and are isom orph ic within short-term mem ory. Although the data presented here leave the issue of the precise mechanism responsible for the effect unresolv ed, at th e empirical level the claim of functional fractionation into separate verbal and spatial system s is question ed by them. Such ® ndings challenge those m odels of short-term m emory based u pon a function al fractionation of verbal and spatial shortterm memory, which is a m ajor characteristic of working mem ory models (see e.g. Baddeley, 1986, 1992; Baddeley & Hitch, 1974), and tentatively sup port a model based upon the equivalence of such representations (see e.g. Jo nes, 1993; Jones, Beaman, & M acken, in press; Jo nes et al., 1995).
R EFER EN C ES Baddeley, A.D. (1986). Working memory. Oxford: Oxford University P ress. B ad deley, A.D. (1992 ). Is working m em ory workin g? T he ® fteenth B artlett lecture. The Qua rterly J ourna l of Experimenta l P sychology, 44A , 1± 31. Baddeley, A.D., & H itch, G. (1974). Working m em ory. In G.H . B ow er (Ed.), The psychology of lea rning a nd motiva tion: Adva nces in resea rch a nd theory, Vol. 8 (pp. 47± 89). L on do n: Academ ic P ress. Battacchi, M .W., Pe lamatti, G.M ., & Um iltaÁ , C. (199 0). Is there a mod ality effect? Evidence for visual recency an d suf® x effects. M emory a nd Cognition, 18 , 651± 658. Bjork, E.L ., & Healy, A.F. (19 74). S hort-term order an d item retention. J ourna l of Verba l Lea rning a nd Verba l B eha vior, 13, 80 ± 97. Broadbe nt, D.E., & Broadbent, M .H .P. (1981) . Recency effects in visual mem ory. Qua rterly J ourna l of Experimenta l P sychology, 33A , 1± 15. De Renzi, E., & Nichelli, P. (1975). Verbal and no n-ve rbal short-ter m m emory impairment following hem ispheric dam age. Cortex, 11, 341± 35 4. Detterman, D.K ., & Brow n, J. (1974) . O rder infor mation in short-ter m m emory. J ourna l of Experimenta l P sychology, 103, 740± 75 0. Devine, J.V., Burke, M .W., & Roh ack, J.J. (1979) . S timulus similarity an d order as facto rs in visual shortterm memo ry in no n-hum an primates. J ourna l of Experimenta l P sychology : Anima l B eha v ior P rocesses, 5, 335± 354. Do naldson, W., & Glathe, H . (19 69). Recognitio n m emory for item an d order infor mation. J ourna l of Experimenta l P sychology, 82, 557± 560. È ber da s GedaÈ chtnis. L eipzig : D un ke r. (Translation b y H . Ru yer & C.E. Eb bing haus, H . (1 88 5). U Bussenius, M emory. New York: Teachers College, Columbia University, 1913 . Estes, W.K . (1972) . An associative basis for coding an d org an ization in m em ory. In A.W. M elton and E. M artin (E ds.), Coding processes in huma n memory (pp. 161± 190) . Washington, DC: W inston. Healy, A.F. (197 4). Separating item from order infor mation in short-term m emory. J ourna l of Verba l Lea rning a nd Verba l B eha vior, 13, 644± 655. Healy, A.F. (1982) . Short-term m emo ry for order infor m ation. In G.H. B ow er (E d.), The psychology of lea rning a nd motiva tion: Adva nces in resea rch a nd theory, Vol. 8 (pp. 19 1± 238). New York: Acad emic Press. Hinrichs, J.V. (1968). P restimulus an d poststimulus cueing of recall order in the m em ory span . P sychonomic S cience, 12 , 261± 262.
158
FARR AN D AN D JON ES
Isaacs, E .B., & Vargha-K hadem, F. (1989) . D ifferential course of development of spatial an d ve rbal mem ory span: A norm ative study. B ritish J ourna l of Developmenta l P sychology, 7, 377± 380 . Jon es, D.M . (1 99 3). O bjects, stream s and th reads o f a uditory attentio n. In A.D. Badd eley & L . Weiskran tz (Eds.), Attention: S election, a wa reness a nd control (pp. 87± 104). Ox ford: O xford University Press. Jones, D.M ., B eaman, C.P., & M acken, W.J. (in press). T he ob ject-oriented episodic record m odel. In S. Gathercole (E d.), M odels of short-term memory. L ondo n: Erlbaum. Jones, D.M ., Farrand, P., S tu art, G.P., & M o rris, N. (1995) . Functional equivalence of verbal an d spatial information in serial short-term memo ry. J ourna l of Experimenta l P sychology : Lea rning, Memory, a nd Cognition, 21, 1008 ± 1018 . L ewan dowsky, S., & L i, S.-C. (19 94). M em ory for serial order revisited. P sychologica l R eview, 101, 539 ± 547. L ewan dowsky, S., & M urdock, B.B. (1989) . M emory for serial order. P sychologica l R eview, 96, 25± 57. L i, S.-C., & L ewan dow sky, S. (199 3). Intralist distractors an d recall direction: Con straints o n m od els of mem ory for serial order. J ourna l of Experimenta l P sychology : Lea rning, Memory, a nd Cognition, 19, 895± 908. L i, S.-C., & L ewandow sky, S. (19 95). Forward and backward recall: Different retrieval processes. J ourna l of Experimenta l P sychology : Lea rning, Memory, a nd Cognition, 21, 837± 847 . M adigan, S.A. (197 1). M od ality an d recall order interactio ns in short-term mem ory for serial order. J ourna l of Experimenta l P sychology, 87, 294± 296. M aki, W.S., Beatty, W.W., & Clou se, B.A. (19 84). Item an d order infor matio n in spatial m em ory. J ourna l of Experimenta l P sychology : Anima l B eha vior P rocesses, 10, 437± 452. M ewho rt, D.J.K ., Poph am , D., & Jam es, G. (199 4). O n serial recall: A critiq ue in th e th eory of distributed associative m emory. P sychologica l R eview, 101, 534 ± 538. M orris, N., & Jones, D.M . (1987). Reporting words from th e eye or the ear: To write or to speak? Ergonomics, 30, 665± 674. M urdo ck, B.B . (198 2 ). A theory for th e storage an d retrieval of item and associative info rm ation, P sychologica l R eview, 89, 609± 626 . M urdock, B.B. (1983) . A distributed mem ory m od el for serial-order infor mation. P sychologica l Review, 90, 316± 338. M urdock, B.B., & L ewandowsky, S. (1986) . Chaining 100 years later. In F. K lix & H . Hagendorf (E ds.), Huma n memory a nd cognitiv e ca pa bilities: M echa nisms a nd performa nces. Am sterdam : Elsevier. Nair ne, J.S., & N eath, I. (1994) . Critique of the retrieval/deblurring assumptions of the theory of distributed associative m emory. P sychologica l R eview, 101, 528 ± 533. N ilsson, L .G., Wright, E., & M urdock, B.B. (1979) . O rder o f recall, ou tput inte rference, and the mod ality effect. P sychologica l R esea rch, 41, 63± 78. Phillips, W.A., & Christie, D.F.M . (1977). Interference with visualisation. Qua rterly J ourna l of Experimenta l P sychology, 29, 637± 650. Roh rm an, N.L ., & Jahnke, J.C. (1965) . Effect of recall co nditio ns, presentation rate and retentio n interval o n short-term m emo ry. P sychologica l R eports, 16, 877± 883. Scho® eld, N.J., & Ashm an, A.F. (1986). T he relatio nship between digit span an d cognitive processing across ab ility g roups. I ntelligence, 10, 59± 73. Shiffrin, R .M ., & Cook, J.R. (1978 ). Short-term forgetting of item an d order information. J ourna l of Verba l Lea rning a nd Verba l B eha vior, 17 , 189± 218. S myth, M .M ., & S choley, K.A . (1992) . Deter mining spatial span: T he role of m ovem ent tim e an d articulation rate. Qua rterly J ourna l of Experimenta l P sychology, 45A , 479± 501. Zim merman, J., & U nderwood, B.J. (1968 ). O rdinal position kn owledge within an d across lists as a functio n of instructio ns in free-recall learning. J ourna l of Genera l P sychology, 79, 30 1± 307. M a nuscript receiv ed 12 October 1994 Accepted revision received 26 Ma y 1995