RESEARCH REPORT FREE RECALL IN WILLIAMS SYNDROME: IS THERE A DISSOCIATION BETWEEN SHORT- AND LONG-TERM MEMORY? Jon Brock, Gordon D.A. Brown and Jill Boucher (Department of Experimental Psychology, University of Bristol, Bristol, UK; University of Warwick, Coventry, UK)

ABSTRACT Two experiments used the free recall paradigm to investigate verbal memory abilities in Williams syndrome (WS) – a rare genetic disorder. In an earlier free recall study, Vicari et al. (1996a) reported that, unlike TD controls, children with WS showed a recency effect but failed to show a primacy effect. These authors interpreted their findings as evidence for a dissociation between relatively strong verbal short-term memory and relatively impaired verbal long-term memory. In Experiment 1 of the current study, children with WS and TD controls showed comparable improvements in performance with repeated testing of the same material, indicating similar long-term learning of the test items. Neither group showed evidence of primacy effects. However, the extent of primacy effects in free recall is known to depend on the rehearsal strategy that participants adopt. In Experiment 2, therefore, participants were encouraged to engage in overt cumulative rehearsal. This manipulation resulted in significant and comparable primacy effects in both groups, although neither group demonstrated a significant change in overall performance. There was therefore no evidence from either experiment for a dissociation between short- and long-term verbal memory in WS. Key words: Williams syndrome, verbal short-term memory, verbal long-term memory, language development

INTRODUCTION Williams syndrome (WS) is a rare genetic disorder that is associated with mild to moderate mental retardation. The cognitive profile in WS is uneven, with verbal abilities typically reported as being superior to visuo-spatial abilities, particularly in older children and adults (e.g., Bellugi et al., 2000; Jarrold et al., 2001). However, there is also evidence for relative strengths and weaknesses within the linguistic domain, suggesting that language development follows an atypical trajectory (c.f., Karmiloff-Smith, 1998). In particular, it has been argued that there is a dissociation between relatively strong phonological abilities and impaired or atypical semantic processing (e.g., Rossen et al., 1996; Temple et al., 2002; Thomas and Karmiloff-Smith, 2003), although the precise nature of this dissociation remains unclear. The current study investigated the verbal memory abilities of individuals with WS using the free recall paradigm. In particular, it investigated a claim made by Vicari et al. (1996a) who proposed that WS is associated with relatively strong phonological short-term memory (STM) but relatively impaired semantic long-term memory (LTM). This claim was based on the results of a study of free recall in which participants were presented with lists of 12 words and were required to immediately recall as many list items as possible in any order. TD five-year-old children showed a bow-shaped serial position curve with significant Cortex, (2006) 42, 366-375

primacy and recency effects (i.e., relatively good recall of words presented early and late in the list respectively). However, children with WS who had similar vocabulary scores to the TD controls performed significantly more poorly overall, and showed a recency effect but no primacy effect. Vicari et al. (1996a) interpreted their results in terms of ‘dual-store’ models of memory (e.g., Atkinson and Shiffrin, 1968; Waugh and Norman, 1965). According to such models, verbal material is initially maintained in a limited capacity short-term store that holds mainly phonological information (c.f., Craik, 1968; Shallice, 1975). However, items can pass from the short-term store into a relatively stable long-term store that codes material in a more durable semantic form (c.f., Glanzer, 1972). The recency effect in free recall is assumed to reflect the output of the short-term store, because it is removed by the addition of a short delay between presentation and recall (e.g., Glanzer and Cunitz, 1966; Postman and Phillips, 1965). In contrast, primacy effects are assumed to arise because earlypresented items are rehearsed more often (Rundus, 1971; see also Tan and Ward, 2000) and this increases their chances of entering the long-term store. Vicari et al. (1996a) therefore proposed what we will refer to as the ‘LTM account’, arguing that the normal recency effect and reduced primacy effect in their WS group were indicative of “a dissociation between normal short-term and deficient long-term verbal memory in WS” (p. 510). They went on to suggest that “our work supports recent observations of a dissociation

Free recall in Williams syndrome

between spared phonological and impaired lexicalsemantic aspects of language” (p. 511). This apparent dissociation between STM and LTM in WS has interesting parallels in the neuropsychological literature. In particular, amnesic patients have been found to demonstrate intact recency effects but absent primacy effects in free recall, as well as showing patterns of performance across a range of verbal memory tasks that are consistent with selective impairments in LTM (e.g., Baddeley and Warrington, 1970; Delis et al., 1991). Obviously, such findings are consistent with the idea that the reduced primacy effect in WS is a consequence of impaired LTM. However, there are two major concerns with this LTM account. First, as noted above, dual-store models assume that primacy effects arise because participants rehearse early items more than later items. Indeed, when participants are prevented from rehearsing, primacy effects are reduced or even disappear (e.g., Glanzer and Cunitz, 1966; Richardson and Baddeley, 1975). Consequently, an alternative interpretation of the reduced primacy effect in WS is simply that individuals with WS do not engage in rehearsal (henceforth referred to as the ‘rehearsal account’). In other words, Vicari et al.’s (1996a) findings may reflect strategic differences between groups, rather than any underlying structural memory differences. Second, the LTM account relies on the validity of dual-store models, and these are not without their critics (see Crowder, 1993). In particular, it has been argued that both primacy and recency effects can be explained in terms of a single memory system (c.f., Brown et al., 2002; Crowder, 1993; Glenberg et al., 1980; Neath, 1993) by assuming that the probability of retrieval of an item depends on the interval between the most recent encounter with that item and its retrieval (Tan and Ward, 2000). Recency effects arise because the interval between presentation and retrieval is relatively small for later items in the list. Primacy effects can be accounted for in similar terms because primacy items are typically rehearsed later than midlist items, so the interval between the last rehearsal of these items and their retrieval is also relatively small. Clearly, if such ‘single-store’ models are correct, then individuals with WS cannot have a selective deficit in LTM. Instead, the absence of primacy effects must be interpreted in terms of rehearsal strategy (i.e., the rehearsal account). On the other hand, if this rehearsal account can be eliminated, then WS may be taken as evidence against single-store models and in favour of separable memory systems. The two experiments reported in this paper therefore set out to distinguish between the LTM account and the rehearsal account. Experiment 1 attempted to replicate the reduced primacy effect reported by Vicari et al. (1996a). In addition, it was possible to investigate long-term learning effects

367

whereby overall performance improves when participants are repeatedly tested using the same material. These effects are substantially reduced in amnesia (Delis et al., 1991; Lezak, 1995), so the LTM account predicted that individuals with WS would also show reduced long-term learning effects. In contrast, the rehearsal account assumes that LTM in WS is normal, and so predicted that people with WS would show similar long-term learning effects to controls. Experiment 2 again pitted the two accounts against each other. Participants were instructed to rehearse overtly, so it was possible to ensure that they really were rehearsing. The rehearsal account predicted that participants with WS would now demonstrate primacy effects, whereas the LTM account predicted that individuals with WS would still fail to show a primacy effect – depending on the locus of the deficit, rehearsed items would either fail to enter LTM or would not be retrievable from LTM. EXPERIMENT 1 Experiment 1 investigated long-term learning effects and serial position curves in the free recall of children with WS. Two methodological issues are important here. The first concerns the procedure for presenting stimuli. In order to obtain reliable serial position curves it is necessary to test participants on multiple trials. Studies with adults (on which models of free recall performance are based) typically use completely different words on each trial (here termed the ‘new words’ procedure), whereas studies of free recall in TD children have usually employed the same words in each list but in a different order each time (‘repeated words’ procedure). However, Vicari et al. (1996a) used a procedure (here termed the ‘repeated list’ procedure) whereby participants were presented with the same word list (i.e., the same words in the same order) on each of five trials. A potential disadvantage of this procedure is that it appears to encourage participants to produce items in the same order as they were presented (Lezak, 1995). This makes interpretation of serial position effects problematic because mechanisms involved in encoding serial order are likely to contribute to primacy effects (c.f., Page and Norris, 1998). Consequently, in Experiment 1, we tested participants using all three procedures to enable comparisons with the Vicari et al. (1996a) study as well as with studies of free recall in TD populations. The second issue concerns the measure on which control participants were matched. Cross-sectional studies of free recall in TD children show that developmental improvements in overall performance between the ages of six and ten years reflect large increases in the recall of primacy items and only relatively modest increases in recall of

368

Jon Brock and Others

recency items (Cole et al., 1971; Cuvo, 1975; Ornstein et al., 1977). This may, at least in part, reflect the increasing use of rehearsal strategies (c.f., Ornstein et al., 1977). However, another interpretation of such findings is that LTM develops later than STM (Cole et al., 1971). Arguably, therefore, if groups are not matched on overall verbal memory ability, a relative deficit in LTM in WS could simply be a reflection of a non-specific delay in verbal memory abilities relative to controls. To avoid such concerns, the performance of children with WS was compared with that of a group of younger TD children who were individually matched on a baseline measure of overall memory performance – the total number of words recalled on the first trial. If the LTM account is correct, then comparable levels of overall performance in the two groups would in fact reflect a relatively small contribution from LTM but a relatively large contribution from STM in the WS group. However, groups should still be differentiable in terms of learning effects. More specifically, individuals with WS would be expected to resemble patients with amnesia and show relatively modest improvements in performance with repeated testing of the same stimuli in the repeated words and repeated list conditions. Moreover, when looking at overall performance collapsed across trials, the two groups should be matched on performance in the new words condition where no long-term learning is possible, but the WS group would be expected to show relatively poor performance in the repeated words and repeated list conditions compared with controls. In contrast, if STM and LTM are equally delayed in WS then there should be no group differences in the extent of learning effects. Method Participants The WS group were 11 children (6 males, 5 females) with a clinical diagnosis of WS who were recruited through the Williams Syndrome Foundation. The age range of the WS group was 11.1 to 18.0 years. Four of the group had taken a FISH test for deletion of the ELN gene (Lowery et al., 1995), and this had proved positive in each case. The control group were selected from a group of 24 TD control children, ranging in age from 5.1 to 8.8 years, who were recruited from a mainstream primary school and were tested on all measures. Stimuli One hundred and forty-four one-syllable nouns were taken from the MRC Psycholinguistic database (Coltheart, 1981). All words had early age of acquisition ratings (four years or less), high

concreteness ratings (300 or greater), high familiarity ratings (greater than 400) and high written frequencies (at least 30 words per million; Kucera and Francis, 1967). For each participant, the words were randomly assigned to 12 lists of 12 words. Five lists were used for the new words condition so that a different list could be presented on each trial. One list was used for the repeated words condition, but the order of the items was changed for each trial. One list was used for the repeated list condition, with the items being presented in the same order on each trial. The remaining five lists were used in Experiment 2. Procedure Children with WS were tested at home in a quiet room with a parent present if required. TD controls were tested in the library of their school. The three conditions were always completed within the same session with a filler task between each condition. The order of testing for the three conditions was counter-balanced such that an approximately equal number of participants within each group were allocated to each of the six possible orders of testing. Following Vicari et al. (1996a), the experimenter read each word aloud at a rate of approximately one word every 2 seconds. In order to be certain that participants were attending to the stimuli, they were instructed to repeat each word aloud upon hearing it. At the end of each list, the experimenter said “Ok, go” and participants were required to try and remember as many of the words as possible. Results The first trial of each condition was the same insofar as participants had not previously been tested on the words in that list. For group-matching purposes, a measure of overall performance on the first trial was therefore obtained by calculating the total number of words recalled on the first trial of each block, across the three conditions. A subset of 11 TD children (4 males, 7 females) were then selected who were individually matched to children in the WS group on this measure to within one word. Participant details are shown in Table I. Vocabulary mental ages, assessed using the British Picture Vocabulary Scale (BPVS II; Dunn et al., 1997), were significantly higher in the WS group than in the TD group [t(14.3) = 5.52, p < .001]. However, there were no significant group differences in nonverbal ability [t(19) = 1.42, p = .173] measured using the Ravens Coloured Progressive Matrices (Raven, 1993). Table I also shows overall performance in each of the three conditions in terms of the total number of words recalled in the five trials. These results were analysed by performing an ANOVA with group as a between-subjects factor and condition as

Free recall in Williams syndrome

369

TABLE I

Participant details for Experiment 1. Free recall performance measured in terms of total number of words recalled WS N = 11 Chronological age (years, months) BPVS age (years, months) Ravens matrices score* First trial New words condition Repeated words condition Repeated list condition

TD N = 11

Mean

SD

Mean

SD

13.8 9.3 20.4 8.27 14.5 18.1 20.7

2.6 1.9 5.5 1.85 3.1 5.0 6.5

6.1 6.1 17.5 8.27 12.5 18.7 19.1

0.5 0.1 3.7 1.56 2.8 5.4 6.7

Note. *Mean Ravens matrices score for WS group based on N = 10; SD = Standard Deviation.

a repeated measure. Group differences in learning effects would lead to an interaction between group and condition because learning effects should influence overall performance on the repeated words and repeated list conditions but not the new words condition. There was a significant effect of condition [F (2, 40) = 15.90, p < .001] resulting from inferior performance in the new words condition compared with the other two conditions (Bonferroni correction applied). However, there was no main effect of group [F (1, 20) = .32, p = .578] and, crucially, the interaction between group and condition was non-significant [F (2, 40) = .69, p = .507]. Learning effects were also investigated by looking at the changes in the total number of items recalled across the five trials in each condition. Figure 1 shows overall performance in the new words, repeated words and repeated list conditions for each of the five trials within a block. For each condition, results were subjected to two-way ANOVA with group as a between-subjects factor

and trial as a repeated measure using linear contrasts (this increased statistical power but precluded the use of a three-way ANOVA). Thus, the extent of group differences in learning rates should be reflected in the size of the interactions between group and trial. In the new words condition, there was no significant main effect of group [F (1, 20) = 2.31, p = .144], or of trial [F (1, 20) = 1.81, p = .193], and no significant interaction between group and trial [F (1, 20) = 1.81, p = .193]. In the repeated words condition, there was a significant effect of trial [F (1, 20) = 13.12, p = .002], but no significant effect of group [F (1, 20) = .08, p = .778], and no interaction between trial and group [F (1, 20) = .67, p = .423]. In the repeated list condition, there was again a significant effect of trial [F (1, 20) = 29.44, p < .001], no significant effect of group [F (1, 20) = .34, p = .567], and no significant interaction between group and trial [F (1, 20) = .01, p = .937]. Figure 2 shows the performance in each condition as a function of serial position.

Fig. 1 – Learning effects in the new words (left), repeated words (middle), and repeated lists (right) conditions in Experiment 1. Error bars represent ± one standard error.

370

Jon Brock and Others

Fig. 2 – Serial position curves for the new words (left), repeated words (middle), and repeated list (right) conditions in Experiment 1.

Following Vicari et al. (1996a), results were analysed by comparing three serial positions – primacy (mean of the first three items), midlist (mean of the middle six items), and recency (mean of the last three items). To avoid problems of floor effects as much as possible, results were also collapsed across the three conditions. ANOVA was then performed with group and serial position as between- and within-subjects factors respectively. The effect of serial position was significant [F (1.43, 28.60) = 160.98, p < .001; Greenhouse Geisser correction applied], reflecting a significant advantage for recall of recency items over primacy and midlist items (Bonferroni correction applied). However, there was no significant effect of group [F (1, 20) = .11, p = .741], and the group by serial position effect was also non-significant [F (2, 40) = 1.35, p = .267; sphericity assumed]. Qualitatively similar patterns of results were also observed when performance was analysed for the three conditions separately. In particular, the interaction terms were non-significant in each analysis (p > .1 in all cases). Discussion Experiment 1 investigated the free recall performance of children with WS and younger TD controls matched on initial levels of overall performance. The LTM account predicted that individuals with WS would show reduced longterm learning effects compared with controls. Moreover, based on the results of Vicari et al. (1996a), it was predicted that individuals with WS would demonstrate smaller primacy effects in their serial position curves. In fact, neither prediction was supported.

Looking first at learning effects, in the repeated words condition and the repeated list condition, both groups showed significant improvements in performance with repeated testing of the same material. Crucially, there was little evidence for group differences in the rate at which performance improved in the two groups. As such, there was no evidence to suggest that individuals with WS resemble amnesic patients in showing reduced long-term learning effects. Because the two groups were matched on initial levels of performance, this also ensured that overall performance in these two conditions was similar in the two groups. This in turn meant that there was no significant group by condition interaction when performance in the repeated words and repeated lists conditions were compared with performance in the new words condition in which no learning occurred1. The results concerning serial position effects are less clear cut. Given the findings of Vicari et al. (1996a), it was expected that there would be group by position interactions, with the WS group recalling relatively fewer primacy items than controls. Because the two groups in the current study had comparable levels of overall performance, this should have been mirrored by relatively good recall of recency items in the WS 1In

a recently published study, Nichols et al. (2004) reported that, compared with TD children, individuals with WS showed significantly slower rates of long-term learning on the California Verbal Learning Test (Delis et al., 1994) – a free recall test that has a procedure similar to that in the repeated list condition of the current study. However, learning rates were investigated by calculating for each individual the slope of the least-squares regression line fitted to the plot of performance across the five trials. Unfortunately, regression slopes fitted to individual data are notoriously unreliable (see Carter et al., 1986). Moreover, in contrast to the current study, the groups were not closely matched on performance on the first trial, so slower rates of improvement may simply have been a consequence of poorer initial performance in the WS group.

Free recall in Williams syndrome

group. Instead, the two groups showed similar serial position curves in all three conditions. It is important to note that, in both studies, children with WS demonstrated recency effects but not primacy effects. The difference lies in the serial position curves for the TD children – those tested by Vicari et al. (1996a) showed both primacy and recency effects, whereas the TD children in Experiment 1 showed only recency effects. This was true even in the repeated list condition where the procedure closely resembled that used by Vicari et al. (1996a). The reason for this discrepancy is unclear. As noted in the Introduction, developmental improvements in free recall performance largely reflect increased recall of primacy items. However, the TD children in Experiment 1 were older than those tested by Vicari et al. (1996a) so, if anything, one might expect them to show relatively large primacy effects. Nevertheless, the TD children in the Vicari et al. (1996a) study did show much better overall performance than the current control group (as well as outperforming both WS groups), so it may be that the discrepancy between the studies reflects differences in overall levels of performance. Alternatively, it is possible that the primacy effects reported by Vicari et al. (1996a) reflected the fact that these authors used exactly the same list for all participants, so serial position was confounded with the intrinsic memorability of the particular items. This potential confound was avoided in the current study because, in all three conditions, different participants received different lists. A final possibility concerns the influence of rehearsal. In the current study, participants were instructed to repeat aloud each word as it was presented, and this requirement may have effectively prevented them from rehearsing. In the Vicari et al. (1996a) study, participants were not required to repeat the items, and this may have enabled TD children to rehearse other items in between presentations. Given the importance attributed to rehearsal in generating primacy effects, this could potentially explain why Vicari et al. (1996a) reported primacy effect whereas we found no such effects. However, the TD children tested by Vicari et al. (1996a) were all five years of age, and the available evidence suggests that children do not typically start rehearsing until the age of around seven (Cuvo, 1975; Hagen and Kingsley, 1968). The important point, however, is that it is impossible to determine conclusively whether or not the children with WS in the Vicari et al. (1996a) study or those in the current study were rehearsing, and therefore impossible to separate the LTM account and the rehearsal account on the basis of their serial position curves. Logically, in order to differentiate between these two accounts, it is necessary to look at individuals with WS who are rehearsing. One approach might be to test older or more able individuals with WS who have mental ages well above the seven-year level (the age at

371

which TD children are thought to begin rehearsing spontaneously). However, simply having a mental age above seven years is no guarantee that an individual is rehearsing. Instead, therefore, the approach adopted in Experiment 2 was to teach participants from Experiment 1 how to rehearse, and then examine the effect of such instruction on their serial position curves. EXPERIMENT 2 In Experiment 2, participants performed free recall but were instructed to rehearse aloud the previously presented stimuli in the intervals between presentation of the list items. Previous studies using this ‘overt rehearsal’ paradigm show that adults tend to engage in what is termed ‘cumulative rehearsal’ – that is, rehearsing as many of the previously presented items as possible between each presentation of a new item (Rundus, 1971; Tan and Ward, 2000). In contrast, TD sixyear-old children spontaneously adopt a ‘simple rehearsal’ strategy – rehearsing only the mostrecently-presented item (Ornstein et al., 1977). The type of rehearsal adopted appears to be crucial in determining the extent of primacy effects. When older children and adults are instructed to use a simple rehearsal strategy, primacy effects are substantially reduced (e.g., Glanzer and Meinzer, 1967; Ornstein et al., 1977). Similarly, six-year-old children do not typically show primacy effects unless they are taught how rehearse in a cumulative fashion (Ornstein et al., 1977). According to dual-store models, the type of rehearsal strategy adopted is critical because the probability of an item entering LTM depends on the number of times it is rehearsed – in cumulative rehearsal, early items are rehearsed more than midlist items, whereas simple rehearsal entails that all items are rehearsed approximately equally. In contrast, single-store models assume that cumulative rehearsal is important because it allows primacy items to be rehearsed after midlist items are presented, whereas in simple rehearsal the items are rehearsed in the same order as they are presented. In Experiment 2, therefore, participants were taught to use a cumulative rehearsal strategy. The rehearsal account predicted that this manipulation would lead to significant primacy effects in the WS group. In contrast, the LTM account predicted that individuals with WS would fail to show a primacy effect or would at least show a significantly reduced primacy effect compared with TD controls. Method Participants The 11 children with WS and 24 TD controls who took part in Experiment 1 also took part in Experiment 2.

372

Jon Brock and Others TABLE II

Participant details for Experiment 2. Free recall performance measured in terms of total number of words recalled WS N = 11 Chronological age (years, months) BPVS age (years, months) Ravens matrices score* New words condition Overt rehearsal condition

TD N = 11

Mean

SD

Mean

SD

13.3 9.3 20.2 14.5 13.8

2.2 1.1 5.8 3.1 4.5

6.6 6.1 19.4 15.1 15.5

0.3 1.1 3.0 2.8 2.8

Note. *Mean Ravens matrices score for WS group based on N = 9; SD = Standard Deviation.

Stimuli For each participant, the stimuli were the remaining 60 words that had not been used in Experiment 1. These words were randomly split into 5 lists of 12. Procedure All participants were tested in Experiment 2 after they had performed the three conditions in Experiment 1. In most cases, Experiment 2 was performed later within the same session as Experiment 1, but for some of the TD children the constraints of the school timetable meant that this was not possible and testing was conducted during a later session. The procedure was similar to that for the new words condition of Experiment 1. However, before testing, participants were told that it would help them to remember more words if they practised remembering as many words as they could (in any order) after the experimenter read each word. Ornstein et al. (1977) used the overt rehearsal paradigm with TD children and presented words at a rate of one every 5 seconds. However, pilot studies showed that the children with WS were unable to rehearse overtly at this rate. The experimenter therefore gave up to 10 seconds for rehearsal but continued if the child indicated that they could not recall any more items. Results Only 10 of the children with WS (5 boys, 5 girls) and 11 of the TD controls (4 boys, 7 girls) were able to consistently engage in cumulative rehearsal (the remaining individuals could only use a simple rehearsal strategy and were therefore excluded from further analyses). Details of the remaining participants are shown in Table II. As in Experiment 1, the WS group had superior vocabulary knowledge [t(19) = 3.76, p = .001], but there was no significant difference in performance of the two groups on the Ravens matrices [t(18) = .43, p = .673]. Table II also shows measures of overall free recall performance. There were no significant group differences in performance in the new words condition of Experiment 1 [t(19) = –

.45, p = .654] or in the overt rehearsal condition in the current experiment [t(19) = –1.08, p = .296]. In addition, paired samples t-tests showed that neither the WS group [t(9) = – .65, p = .531] nor the TD group [t(10) = .41, p = .692] showed any significant effect of condition on overall performance (i.e., rehearsal did not lead to a significant improvement in performance). Figure 3 shows the serial position curves for the two groups in the overt rehearsal condition in Experiment 2 (right panel), together with the serial position curves for the same participants on the new words condition of Experiment 1 (left panel). As in Experiment 1, results were collapsed into three serial positions – primacy (mean of the first three items), midlist (mean of the middle six items), and recency (mean of the last three items). The results were then subjected to three-way ANOVA with group as a between-subjects factor and condition and serial position as repeated measures. The main effects of group [F (1, 19) = 2.03, p = .170] and condition [F (1, 19) = .00, p = .991] were non-significant. There was a significant main effect of serial position [F (1.29, 24.47) = 58.73, p < .001; Greenhouse-Geisser correction applied], with all pairwise differences being significant (Bonferroni correction applied), but the interaction between serial position and group was nonsignificant [F (2, 38) = 1.12, p = .337; sphericity assumed]. Of greater interest, there was a significant interaction between serial position and condition [F (1.53, 28.98) = 63.58, p < .001; Greenhouse-Geisser correction applied], indicating that the shape of the serial position curves was different in the two conditions. Planned comparisons demonstrated that recall of primacy items was significantly better in the overt rehearsal condition than in the new items condition of Experiment 1 [t(20) = –7.17, p < .001], whereas the reverse was true for recency items [t(20) = 8.10, p < .001]. The effect of condition was nonsignificant for midlist items [t(20) = .45, p = .655]. Crucially, however, there was no significant threeway interaction [F (2, 38) = .35, p = .706; sphericity assumed], indicating that the effect of condition on the serial position curves was similar for both groups.

Free recall in Williams syndrome

373

Fig. 3 – Serial position curves for the new words condition of Experiment 1 (left) and the overt rehearsal condition of Experiment 2 (right).

Discussion In Experiment 2, participants were encouraged to use a cumulative rehearsal strategy whereby they rehearsed a number of different items in the intervals between presentations of list members. According to both single- and dual-store models, cumulative rehearsal should lead to primacy effects. However, the LTM account predicted that the WS group would have shown a reduced primacy effect (if any at all). In fact, both groups showed significant and comparable primacy effects. As such, the results of Experiment 2 provide further support for the rehearsal account as opposed to the LTM account. A further point worth noting is that teaching participants to engage in rehearsal did not result in significant improvements in performance in either group. Thus, while rehearsal led to increased recall of primacy items, there were corresponding decreases in recall of recency items. In fact, this finding is consistent with the results of previous studies in neurotypical adults (Tan and Ward, 2000) and children (Ornstein et al., 1977) showing that, although primacy effects are increased when participants engage in cumulative as opposed to simple overt rehearsal, the type of rehearsal strategy has little effect on overall performance. A subtle difference here is that, in the current study, we were comparing an overt cumulative rehearsal condition with a ‘no instructions’ condition in which it is impossible to determine the type of rehearsal strategy employed, if any. Nevertheless, our findings suggest that teaching rehearsal

strategies to individuals with WS is unlikely to improve their everyday memory abilities, although, of course, it remains possible that other mnemonic strategies may prove useful. GENERAL DISCUSSION Vicari et al. (1996a) argued that individuals with WS demonstrate a dissociation between relatively strong verbal STM and relatively impaired verbal LTM. This claim was based on the finding that, unlike TD controls, children with WS failed to show a primacy effect in their serial position curve. In this paper, we challenged this ‘LTM account’ and proposed an alternative ‘rehearsal account’, according to which, there is no selective deficit in LTM and the absence of reliable primacy effects in WS is a consequence of a failure to rehearse. The current study tested the contrasting predictions of these two accounts. Taken together, the current results would appear to provide strong support for the rehearsal account as opposed to the LTM account. In Experiment 1, both groups showed comparable long-term learning effects. This finding is contrary to the predictions of the LTM account because amnesic patients (i.e., individuals with verbal LTM deficits) do show impaired long-term learning on similar tasks (Delis et al., 1991; Lezak, 1995). As such, comparisons between individuals with WS and such patients would appear to be unwarranted. Analysis of serial position curves in Experiment 1 proved inconclusive as neither group showed significant

374

Jon Brock and Others

primacy effects – a finding explicable in terms of both the LTM account and the rehearsal account. However, in Experiment 2, when participants engaged in overt cumulative rehearsal, both groups showed significant and comparable primacy effects. This findings was predicted by the rehearsal account, whereas the LTM account predicted absent or reduced primacy effects. It is worth noting that the overall performance of the WS group on the free recall tasks was comparable to that of TD children who had significantly lower vocabulary scores, although the two groups had similar scores on the Ravens matrices. In fact, complementary results were reported in the study by Vicari et al. (1996a), where groups were matched on receptive vocabulary but the WS group performed significantly more poorly on the free recall task. Similarly, studies assessing verbal memory abilities in WS using serial recall tasks have, in general, found that individuals with WS perform more poorly than vocabulary-matched TD controls (Brock et al., 2005; Jarrold et al., 2004; Laing et al., 2001; see also Vicari et al. 1996a), but perform at a comparable level to TD controls matched on nonverbal intelligence (Vicari et al., 1996b). Somewhat surprisingly, therefore, verbal memory abilities in WS would appear to be below the level predicted by verbal mental age, but at approximately the same level as that predicted by nonverbal mental age. One possible explanation for this pattern of results is that performance on recall tasks relies on metacognitive strategies that may be linked to more general levels of intellectual functioning. Alternatively, it may simply reflect the use of receptive vocabulary knowledge as a measure of verbal mental age. The performance of individuals with WS on receptive vocabulary tests is consistently found to be superior to performance on other verbal and nonverbal abilities (e.g., KarmiloffSmith et al., 1997; Thomas et al., 2001; Volterra et al., 1996). For example, in a recent study, Robinson et al. (2003) tested 39 children with WS and a TD control group matched on performance on the Test for Reception of Grammar (Bishop, 1989). The two groups demonstrated comparable verbal memory abilities and nonverbal intelligence (measured by the Matrices subtest of the Kaufman Brief Intelligence Test; Kaufman and Kaufman, 1990), but the children with WS showed superior receptive vocabulary knowledge to controls. This suggests that the apparent discrepancy between verbal memory and verbal mental age in WS would probably disappear if other measures of verbal mental age were employed2. 2There

is clear evidence that, in WS, receptive vocabulary is a strength relative to non-verbal abilities, and that performance on certain visuospatial tasks is impaired relative to verbal abilities in general. However, there is little evidence for a more general dissociation between verbal and non-verbal abilities. As such, it is not surprising that verbal memory abilities should be in line with non-verbal mental age.

We conclude by discussing the wider implications of the current findings. First, the results of the current study have important implications for theoretical accounts of the cognitive and linguistic profile in WS. As noted above, the dual store account posits that, in free recall, the long-term store contributes primarily semantic (as opposed to phonological) information (c.f., Atkinson and Shiffrin, 1968; Glanzer, 1972). Vicari et al. (1996a) therefore argued that the apparent deficit in verbal LTM in WS was consistent with evidence for a more general deficit in semantic processing. Similarly, Thomas and Karmiloff-Smith (2003) have, more recently, cited Vicari et al.’s (1996a) findings as evidence consistent with the hypothesis that language acquisition in WS is over-reliant on phonological as opposed to semantic processing. By rejecting the LTM account, the current study serves an important purpose in narrowing the scope of any semantic processing deficit in WS. A final issue concerns the implications of the current results for the debate regarding the relative merits of single- versus dual-store models of memory. Clearly, Vicari et al.’s (1996a) assertion that WS is associated with a selective deficit in LTM is incompatible with a single-store account. Had the current study supported such a claim, it would have provided evidence against single-store models, and in favour of a dissociation between LTM and STM. However, because the current study supported the rehearsal account rather than the LTM account, there is no evidence from WS at present that cannot be explained in terms of singlestore models. Of course, the current findings can also be interpreted (rather less parsimoniously) in terms of dual-store models, but only if one assumes that long- and short-term verbal memory abilities are equally impaired in WS. Acknowledgements. This research was supported by a Ph.D. studentship awarded to Jon Brock by the Williams Syndrome Foundation, and by grants 88/S15050 from BBSRC (UK) and grants R000239002 and R000239351 from ESRC (UK). We thank the families from the Williams Syndrome Foundation, and the staff and pupils of Woodloes Infant School (Warwick) for their co-operation in this work. We are also grateful to Alan Baddeley, Chris Jarrold, and Geoff Ward for helpful comments. REFERENCES ATKINSON RC and SHIFFRIN RM. Human memory: A proposed system and its control processes. In Spence KW and Spence JT (eds), The Psychology of Learning and Motivation (vol. II). New York: Academic Press, 1968. BADDELEY AD and WARRINGTON EK. Amnesia and the distinction between long- and short-term memory. Journal of Verbal Learning and Verbal Behavior, 9: 176-189, 1970. BELLUGI U, LICHTENBERGER L, JONES W, LAI Z and ST GEORGE M. The neurocognitive profile of Williams syndrome: A complex pattern of strengths and weaknesses. Journal of Cognitive Neuroscience, 12: 7-29, 2000. BISHOP DVM. Test for Reception of Grammar (2nd Ed.). University of Manchester: The Author, Age and Cognitive Performance Research Centre, 1989.

Free recall in Williams syndrome BROCK J, MCCORMACK T and BOUCHER J. Probed serial recall in Williams syndrome: Lexical influences on phonological shortterm memory. Journal of Speech, Language, and Hearing Research, 48: 360-371, 2005. BROWN GDA, NEATH I and CHATER N. A ratio model of scaleinvariant memory and identification. Manuscript submitted for publication, 2002. CARTER RC, KRAUSE M and HARBESON MM. Beware the reliability of slope scores for individuals. Human Factors, 28: 673-683, 1986. COLE M, FRANKEL F and SHARP D. Development of free recall learning in children. Developmental Psychology, 4: 109-123, 1971. COLTHEART M. The MRC psycholinguistic database. Quarterly Journal of Experimental Psychology, 33A: 497-505, 1981. CRAIK FIM. Types of error in free recall. Psychonomic Science, 10: 353-354, 1968. CROWDER RG. Short-term memory: Where do we stand? Memory and Cognition, 21: 142-145, 1993. CUVO AJ. Developmental differences in rehearsal and free recall. Journal of Experimental Child Psychology, 19: 265-278, 1975. DELIS DC, KRAMER JH, KAPLAN E and OBER BA. The California Verbal Learning Test – Children’s Version. San Antonio: Psychological Corporation, 1994. DELIS DC, MASSMAN PJ, BUTTERS N, SALMON DP, CERMAK LS and KRAMER JH. Profiles of demented and amnesic patients on the California Verbal Learning Test: Implications for the assessment of memory disorders. Psychological Assessment, 3: 19-26, 1991. DUNN LM, DUNN LM, WHETTON C and BURLEY J. British Picture Vocabulary Scale. Windsor: NFER-Nelson, 1997. GLANZER M. Storage mechanisms in recall. In Bower GH (ed), The Psychology of Learning and Motivation: Advances in Research and Theory (vol. V). New York: Academic Press, 1972. GLANZER M and CUNITZ AR. Two storage mechanisms in free recall. Journal of Verbal Learning and Verbal Behavior, 5: 351-360, 1966. GLANZER M and MEINZER A. The effects of intralist activity on free recall. Journal of Verbal Learning and Verbal Behavior, 6: 928-935, 1967. GLENBERG AM, BRADLEY MM, STEVENSON JA, KRAUS TA, TKACHUK MJ, GRETZ AL, FISH JH and TURPIN BM. A twoprocess account of long-term serial position effects. Journal of Experimental Psychology: Human Learning and Memory, 6: 355-369, 1980. HAGEN JW and KINGSLEY PR. Labeling effects in short-term memory. Child Development, 39: 113-121, 1968. JARROLD C, BADDELEY AD, HEWES AK and PHILLIPS CA. Longitudinal assessment of diverging verbal and non-verbal abilities in the Williams syndrome phenotype. Cortex, 37: 423-431, 2001. JARROLD C, COWAN N, HEWES AK and GUNN DM. Speech timing and verbal short-term memory: Evidence for contrasting deficits in Down syndrome and Williams syndrome. Journal of Memory and Language, 51: 365-380, 2004. KARMILOFF-SMITH A. Development itself is the key to understanding developmental disorders. Trends in Cognitive Sciences, 2: 389-398, 1998. KARMILOFF-SMITH A, GRANT J, BERTHOUD I, DAVIES M, HOWLIN P and UDWIN O. Language and Williams syndrome: How intact is “intact”? Child Development, 68: 246-262, 1997. KAUFMAN AS and KAUFMAN NL. Kaufman Brief Intelligence Test. Circle Pines: American Guidance Service, 1990. KUCERA H and FRANCIS VW. Computational Analysis of Presentday American English. Providence: Brown University Press, 1967. LAING E, HULME C, GRANT J and KARMILOFF-SMITH A. Learning to read in Williams syndrome: Looking beneath the surface of

375

atypical reading development. Journal of Child Psychology and Psychiatry, 42: 729-739, 2001. LEZAK MD. Neuropsychological Assessment (3rd Ed.). Oxford: Oxford University Press, 1995. LOWERY MC, MORRIS CA, EWART A, BROTHMAN LJ, ZHU XL, LEONARD CO, CAREY JC, KEATING M and BROTHMAN AR. Strong correlation of elastin deletions, detected by FISH, with Williams syndrome: Evaluation of 235 patients. American Journal of Human Genetics, 57: 49-53, 1995. NEATH I. Distinctiveness and serial position effects in recognition. Memory and Cognition, 21: 689-698, 1993. NICHOLS S, JONES W, ROMAN M, WULFECK B, DELIS DC, REILLY J and BELLUGI U. Mechanisms of verbal memory impairment in four neurodevelopmental disorders. Brain and Language, 88: 180-189, 2004. ORNSTEIN PA, NAUS MJ and PRINCE-STONE B. Rehearsal training and developmental differences in memory. Developmental Psychology, 13: 15-24, 1977. PAGE MPA and NORRIS D. The primacy model: A new model of immediate serial recall. Psychological Review, 105: 761-781, 1998. POSTMAN L and PHILLIPS LW. Short-term temporal changes in free recall. Quarterly Journal of Experimental Psychology, 17: 132-138, 1965. RAVEN JC. Coloured Progressive Matrices. Oxford: Information Press Ltd, 1993. RICHARDSON JTE and BADDELEY AD. The effect of articulatory suppression in free recall. Journal of Verbal Learning and Verbal Behavior, 14: 623-629, 1975. ROBINSON BF, MERVIS CB and ROBINSON BW. The roles of verbal short-term memory and working memory in the acquisition of grammar by children with Williams syndrome. Developmental Neuropsychology, 23: 13-31, 2003. ROSSEN M, KLIMA ES, BELLUGI U, BIHRLE A and JONES W. Interaction between language and cognition: Evidence from Williams syndrome. In Beitchman JH, Cohen N, Konstantareas M and Tannock R (eds), Language Learning and Behaviour. New York: Cambridge University Press, 1996. RUNDUS D. Analysis of rehearsal processes in free recall. Journal of Experimental Psychology, 89: 63-77, 1971. SHALLICE T. On the contents of primary memory. In Rabbitt PMA and Dornic S (eds), Attention and Performance V. London: Academic Press, 1975. TAN L and WARD G. A recency-based account of the primacy effect in free recall. Journal of Experimental Psychology: Learning Memory and Cognition, 26: 1589-1625, 2000. TEMPLE CM, ALMAZAN M and SHERWOOD S. Lexical skills in Williams syndrome: A cognitive neuropsychological analysis. Journal of Neurolinguistics, 15: 463-495, 2002. THOMAS MSC, GRANT J, BARHAM Z, GSÖDL MK, LAING E, LAKUSTA L, TYLER LK, GRICE S, PATERSON S and KARMILOFFSMITH A. Past tense formation in Williams syndrome. Language and Cognitive Processes, 16: 143-176, 2001. THOMAS MSC and KARMILOFF-SMITH A. Modeling language acquisition in atypical phenotypes. Psychological Review, 110: 647-682, 2003. VICARI S, BRIZZOLARA D, CARLESIMO GA, PEZZINI G and VOLTERRA V. Memory abilities in children with Williams syndrome. Cortex, 32: 503-514, 1996a. VICARI S, CARLESIMO G, BRIZZOLARA D and PEZZINI G. Short term memory in children with Williams syndrome: A reduced contribution of lexical-semantic knowledge to word span. Neuropsychologia, 34: 919-925, 1996b. VOLTERRA V, CAPIRCI O, PEZZINI G, SABBADINI L and VICARI S. Linguistic abilities in Italian children with Williams syndrome. Cortex, 32: 663-677, 1996. WAUGH NC and NORMAN DA. Primary memory. Psychological Review, 72: 89-104, 1965. Jon Brock, Department of Experimental Psychology, University of Oxford, South Parks Road, Oxford OX1 3UD, UK. e-mail: [email protected]

(Received 3 October 2003; reviewed 1 December 2003; revised 28 February 2004; accepted 4 May 2004)