Permanent City Research Online URL:

Transcription

1 Meyer, B. J., Gardiner, J. M. & Bowler, D. M. (2014). Directed Forgetting in High-Functioning Adults with Autism Spectrum Disorders. Journal of Autism and Developmental Disorders, 44(10), pp doi: /s y City Research Online Original citation: Meyer, B. J., Gardiner, J. M. & Bowler, D. M. (2014). Directed Forgetting in High- Functioning Adults with Autism Spectrum Disorders. Journal of Autism and Developmental Disorders, 44(10), pp doi: /s y Permanent City Research Online URL: Copyright & reuse City University London has developed City Research Online so that its users may access the research outputs of City University London's staff. Copyright and Moral Rights for this paper are retained by the individual author(s) and/ or other copyright holders. All material in City Research Online is checked for eligibility for copyright before being made available in the live archive. URLs from City Research Online may be freely distributed and linked to from other web pages. Versions of research The version in City Research Online may differ from the final published version. Users are advised to check the Permanent City Research Online URL above for the status of the paper. Enquiries If you have any enquiries about any aspect of City Research Online, or if you wish to make contact with the author(s) of this paper, please the team at

2 Running head: DI RECTED FORGETTI NG I N ADULTS WI TH ASDs Direct ed forget t ing in high- functioning adults w ith Autism Spectrum Disorders Brenda J. Meyer 1, John M. Gardiner 2 and Derm ot M. Bowler 2 1 Developm ental Brain-Behaviour Laboratory, Psychology, Southam pton Universit y, Southam pton 2 Autism Research Group, Departm ent of Psychology, City University London Corresponding author Em ail: B.J.m eyer@soton.ac.uk Telephone: + 44 (0)

3 Abstract Rehearsal strategies of adults with aut ism spectrum disorders and dem ographically m atched typically developed adults were strategically m anipulated by cueing part icipants to either learn, or forget each list word prior to a recognit ion task. Part icipants were also asked to dist inguish between autonoetic and noetic states of awareness using the Rem em ber/ Know paradigm. The ASD group recognised a sim ilar num ber of to-be-forgotten words as the TD group, but significant ly fewer to-be-learned words. This deficit was only evident in Rem em ber responses that reflect autonoetic awareness, or episodic m em ory, and not Know responses. These findings support the elaborat ive encoding deficit hypothesis and provide a link between the previously established m ild episodic m em ory im pairm ents in adults with high functioning aut ism and the encoding strategies em ployed. Key w ords: autism, autonoetic awareness, elaborat ive rehearsal, episodic m em ory 2

4 I ndividuals with aut ism spectrum disorders (ASDs) typically display spared and im paired perform ance on m em ory tasks. For exam ple, adults and children with ASDs show m ild deficits in the recall of related lists of words, or word-strings, com pared with dem ographically-matched com parison part icipants, although the recall of unrelated words between ASD and com parison sam ples, is sim ilar (Bowler et al. 2010; Bowler et al. 2007; Bowler et al. 2000; Bowler et al. 1997; Gaigg et al. 2008; Herm elin and O'Connor 1967; Mottron et al. 2001; Sm ith et al. 2007; Toichi and Kam io 2003), but see (Beversdorf et al. 1998; Lopez and Leekam 2003; Ram ondo and Milech 1984). These deficits have been explained in term s of the week central coherence account of autism (Frith and Happe 1994; Happe 1994) that holds that individuals wit h ASDs show a pattern of neuropsychological deficit s consistent with an im paired ability to place inform at ion within a wider context. That is, they display a local rather than a global bias (Frith and Happe 1994; Happe 1994). However, the recall of related m aterial by individuals wit h ASDs can m atch that of com parison sam ples when provided with appropriate support, such as cues at test (Boucher 1981; Boucher and Warrington 1976b; Bowler et al. 2004; Bowler et al. 1997). This would im ply that the m ild m em ory im pairm ents in ASD populat ions result from im paired recall, although it has also been hypothesised that recall deficits of part icular m aterial results from an encoding deficit, or m ore specifically, that individuals with ASDs engage in less associative elaborat ive rehearsal than typically developed (TD) individuals (Bowler et al. 2000; Mottron et al. 2001; Sm ith et al. 2007; Toichi and Kam io 2003). 3

5 Two types of encoding were ident ified in early m em ory research that have subsequently becom e known as m aint enance, or rote, rehearsal and elaborat ive rehearsal (Craik and Lockhart 1972; Woodward et al. 1973). Maintenance, or rote rehearsal can be defined as the technique whereby m aterial is m erely repeated or m aintained either by internal or external speech. Elaborat ive rehearsal can be defined as the technique by which m aterial is extended upon, usually by m aking associat ive links with other stim uli, or prior knowledge. Studies com paring the effects of m aintenance and elaborat ive rehearsal have established that m aintenance rehearsal strengthens the m em ory trace wit hin the context of the rehearsed item. This has been shown to im prove recognit ion, but appears to have little im pact upon freerecall. Elaborat ive rehearsal creates additional m em ory traces between the rehearsed item and other item s, thus increasing the num ber of cues for retrieval. Consequently, the effects of elaboration im prove the chances of both free-recall and recognit ion (Craik and Tulving 1975; Geiselm an and Bj ork 1980; Greene 1987; Rundus 1977). The advantageous effects of elaborat ive rehearsal have been offered to explain well established effects within m em ory research. For exam ple, concrete nouns have been shown to be easier to recall than abst ract nouns, due to the im aginat ive and contextual qualities of concrete words that, in com parison to abstract words, m ore readily lend them selves to elaboration (Paivio and Csapo 1969). Toichi and Kam io ( 2003) replicated this concretenoun effect in a sam ple of TD adults using Japanese words. I n a further study by the sam e authors, a group of adults with High-Functioning Autism (HFA) recalled fewer concrete nouns, but a sim ilar num ber of abstract nouns to that of a m atched sam ple of adults without aut ism. As the free-recall of concrete 4

6 nouns is considered to be enhanced because they increase the potential to m ake associat ive links in com parison to abstract nouns, the authors concluded that the individuals with aut ism engaged in less associat ive elaborative encoding than the adults without aut ism ( Toichi and Kam io 2003). A further well established phenom enon in m em ory research also linked to rehearsal strategies is the levels of processing fram ework. Asking part icipants to attend to sem antic features ( m eaning) of words increases the efficacy of recall in com parison to asking participants to attend t o phonological features (sound), which in turn is m ore effective than asking them to attend to graphic, or visual feat ures. Consequently, t he efficacy of recall is linked to the depth of encoding (Craik and Lockhart 1972; Craik and Tulving 1975). Toichi and Kam io ( 2002) found that in a levels of processing recognit ion task, a group of adults with HFA displayed a reduced levels-of-processing effect in com parison to TD adults, even though overall recognit ion rates were sim ilar between the two groups. As the adults with HFA recognised significant ly m ore words with graphic cues than the com parison group and, unlike com parisons, they did not recognise significant ly m ore words with sem ant ic cues than those with phonological, or graphic cues, the authors suggested a possible superior rote m em ory in their sam ple of individuals with HFA which m ay com pensate to a certain extent for a potential elaborat ive encoding deficit (Toichi and Kam io 2002). The evidence that the individuals wit h HFA m ay have em ployed different strategies from those of the TD individuals in this sam ple was also supported by the finding that the perform ance at sem antic level significant ly correlated with non-verbal reasoning abilit y (based on Raven s coloured progressive m atrices task) in the HFA group, but not in the com parison group (Toichi and Kam io 2002). 5

7 Differences between adults with HFA and com parison groups in t he recollective experience of recognition m em ory have also been established using the Rem em ber/ Know paradigm (Bowler et al. 2007; Bowler et al. 2000; Gardiner et al. 2003). Within this paradigm part icipants are asked, following recognit ion of an item, to dist inguish between two states of awareness ( Rem em ber and Know), with Rem em ber responses reflecting autonoetic awareness and Know responses reflecting noetic awareness (Gardiner and Richardson-Klavehn 2000; Tulving 1985). Tulving ( 2002) defined autonoetic awareness as that which allows us to be aware of subj ective t im e in which events happened (p. 2) and it reflects episodic m em ory (Tulving 2002). I n contrast, noetic awareness of an item / event lacks the associat ion with the t im e or place that the m em ory was acquired while m aintaining the quality, or strength, of the recognit ion and reflects sem ant ic m em ory. Consequently, if part icipants can link t heir recognit ion of an item to the t im e of learning ( e.g. they rem em ber what they thought about when they first saw the word, or rem em ber whereabouts in the list the word fell in relat ion to the other words), they are asked to give a Rem em ber response. I f they are unable to offer this addit ional inform at ion but st ill have a strong feeling of recognit ion, they are asked to give a Know response. A group of adults with Asperger s Syndrom e (AS), while correctly recognising a sim ilar num ber of item s overall, offered significant ly fewer Rem em ber responses than a com parison group m atched for age and I Q, thus dem onstrating reduced autonoetic awareness, or a m ild episodic m em ory im pairm ent (Bowler et al. 2000). Subsequent studies have established reduced Rem em ber responses in individuals with HFA com pared with a m atched com parison group in conscious awareness tasks (Bowler et al. 2007) and in a sensory/ perceptual source m em ory task (Souchay et al. 2012). 6

8 As the Rem em ber/ Know paradigm has also been used to establish a dissociat ion of the effects of rehearsal, wit h elaborat ive rehearsal select ively affecting Rem em ber responses and m aintenance rehearsal selectively affect ing Know responses (Dobbins et al. 2004; Gardiner et al. 1994; Gardiner and Richardson-Klavehn 2000), these findings also are consistent wit h the notion that the adults with AS engage in less elaborative rehearsal than TD adults. As yet, there is no conclusive evidence to link the episodic m em ory im pairm ents in this populat ion specifically to the rehearsal st yles em ployed. An earlier experim ent conducted by us aim ed to direct ly test the rehearsal strategies of a group of adults wit h AS by m eans of an overt rehearsal experim ent ( in which part icipants were asked to try and learn lists of words by repeating them out loud). While t his study failed to show a significant difference in the natural rehearsal strategies em ployed, it did reveal a t rend for the adults with AS to engage in slight ly less elaborat ive rehearsal and slight ly m ore rote rehearsal than a m atched com parison group (Sm ith et al. 2007). However, the overt rehearsal technique places addit ional task dem ands that m ay m ask the subt le m em ory im pairm ents of individuals with ASDs. The m ain aim of the present study, therefore, was t o find a m ore sensit ive technique than the overt rehearsal m ethod that m easures covert encoding strategies and avoids addit ional task dem ands. A sensit ive way to covertly m anipulate rehearsal strategies is the directed forgetting paradigm. Using t his established m em ory paradigm m ay effectively reveal potential subtle differences between natural rehearsal strategies of adults with ASDs and a dem ographically m atched com parison sam ple. I n the directed forgetting paradigm, part icipants are instructed to either learn or forget words, either after the presentation of each item, or after the 7

9 entire list has been displayed (Basden and Basden 1996; MacLeod 1999). Typically, in a subsequent recall or recognit ion test in which part icipants are unexpectedly asked t o recall or recognise words from a list regardless of the type of cue ( learn/ forget), part icipants show a dist inct bias to recall or recognise m ore to-be-learned words than t o-be-forgotten words. I n the case of the item m ethod, the 'directed forgetting effect' was suggested to occur because participants only engage in act ive elaborat ive rehearsal following the cue to learn, but not following the cue to forget (Basden and Basden 1996; Bj ork and Woodward 1973; Woodward and Bj ork 1971). The rehearsal strategy of the part icipants using the directed forgetting task can be further m anipulated by introducing a cue-delay (Woodward et al. 1973). Maintaining the overall presentat ion tim ing of the word to-be-learned or to-be-forgotten, while int roducing a delay prior to the learn/ forget cue being presented ( cuedelay) was hypothesised by Woodward et al (1973) to increase the use of rote rehearsal (while part icipants awaited the cue) and decrease the opportunity for elaborat ive rehearsal. Using t he cue-delay in a directed forgetting recognit ion experim ent that also incorporated the Rem em ber/ Know procedure, Gardiner et al., ( 1994) found im proved recognit ion of to-be-forgotten words after a long cue-delay ( 5 sec) in com parison to a short cue-delay (0 sec) (Gardiner et al. 1994). Overall, Rem em ber responses ( regardless of cue delay) were m uch higher for to-belearned words, while Know responses were slight ly higher for to-be-forgotten words. These results are consistent with the view that the effects of effortful, elaborat ive learning strategies increase autonoetic awareness and less effortful, rote strategies increase noetic awareness. This assum ption was further supported by the effect of the cue delay. The im pact of the increased 8

10 m aintenance rehearsal after the long cue-delay (hypothesised to increase rote rehearsal) was evident in increased Know responses versus Remem ber. Conversely, the im pact of the short cue-delay, ( hypothesised to m axim ise the potential for elaborat ive rehearsal), was evident in increased Rem em ber responses versus Know. Consequently, while both elaborat ive and rote rehearsal strategies im prove recognit ion, t he effects of each strategy result in a dissociat ive effect on the type of recognition (Rem em ber or Know). Thus, the Rem em ber/ Know paradigm appears to reflect the extent to which each rehearsal strategy (rote/ elaborative) is effectively em ployed. The study reported here replicated the recognition experim ent by Gardiner et al., (1994) with the aim of establishing whether the m ild episodic m em ory im pairm ents previously established in adults with ASDs, m ay result from the use of less elaborat ive rehearsal strategies than that of TD adults m atched for age and I Q. I t was hypothesised that adults with ASDs would display a dim inished directed-forgetting effect in com parison to a m atched com parison group, as a result of less effective learning strategies for the tobe-learned words. Consistent with this hypothesis, this potent ial deficit should be evident in Rem em ber responses, not Know ( as elaborat ion only affects R) and should be m ost apparent in a short cue-delay condit ion (when the potential to engage in elaborat ive rehearsal is m axim ised). 9

11 Method Part icipants A total of 32 adults with and without HFA participated in the experim ent, 16 (12 m en and 4 wom en) of whom had a diagnosis of AS, supported by ADOS assessments, where possible. All except two participants in the ASD group had received their diagnosis from experienced clinicians within the UK health services. ADOS assessm ents were not system atically carried out but were available for 9 of the individuals for whom algorithm scores supported the clinical diagnosis according to I CD-10 criteria (World Health Organizat ion 1992) and DSM-I V-TR criteria (Am erican Psychiatric Association 2000), excluding the m easure for language developm ent, as this inform ation was often unavailable. There were no evident language im pairm ents in any of the part icipants during int eractions with the experim enter. For two AS individuals, no written confirm at ion of their AS diagnosis was available but our research experience with this clinical group left us in no doubt that the int eractive style of these individuals were consistent with a clinical diagnosis of ASD. Since rem oving these two participants from the analyses did not alter t he trend of the overall observat ions reported below, we took the conservative decision to retain them in our sam ple and report the m ain results of interest with and without these two participants and their appropriately m atched TD com parison pair. I n total, sixteen part icipants from the overall sam ple ( 10 m en and 6 wom en) were TD adults, selected to m atch individuals wit hin the ASD group on the grounds of age and verbal I Q scores, as m easured on the WAI S I I I ( Wechsler Adult I ntelligence Scale, Third Edit ion, The Psychological Corporation, 1997). All part icipants had a VI Q score of 84, or above and volunteered to take part following recruitm ent cam paigns, either placed in the 10

12 local press (for TD adults), or leaflets distributed to Asperger support groups (for ASD participants). All part icipants gave inform ed consent prior to taking part in the study. Procedures regarding consent and part icipat ion rights m et prior ethical approval from the inst itut ion and were therefore perform ed in accordance with the ethical standards laid down by the Declarat ion of Helsinki ( 1964) and its later am endm ents. Design and Materials The experim ent used a m ixed 2 x 2 x 2 design. The between group variable was ASD group vs. TD group. The other experim ental design factors were the directed forgetting instruct ions ( learn vs. forget) and t he period of cue delay (short vs. long). The m aterials were 176 concrete nouns taken from (Paivio et al. 1968) lists, each with im agery and concreteness ratings of 5 or above and m eaningfulness rat ings of 4.5 or above and thus all contained highly elaborat ive qualit ies. These were divided into 2 lists of 88 words, each m atched on rating scores for im agery, concreteness, m eaningfulness and word length overall. One of the two lists was presented as study item s, and the other was presented as distracter item s during the recognit ion trial. These were switched between part icipants in order to counterbalance study item s vs. distracters. The study list of 88 words was subdivided into 44 words with a short cue-delay and 44 words with a long cue-delay. Within each of these 44 words, half were designated to-be-learned item s and half were designated tobe-forgotten item s. The presentation of the words was blocked according to cue-delay. Half of each part icipant group received the words with a long cuedelay first and half received the words with a short cue-delay first. Within these 2 blocks, the words and types of cue were random ised for each 11

13 part icipant, so that they were unable to predict which words were to-belearned and which were to-be-forgotten. The type of cue was counterbalanced between each half of the list, so that words that were to-be-learned for half of each participant group becam e to-be-forgotten words for the other half of each group. Consequently, all list words were rotated across each condit ion. Procedure Prior to the recognit ion trial, part icipants were given standard directed forgetting instruct ions (see Appendix 1). Participants were then shown a practice trial so that t hey experienced both the visual presentat ion of the words and cues, as well as the change in t he tim ing of the cue and were not taken by surprise during the actual study phase. The practice trial incorporated a total of 10 words drawn from the sam e source as the study and distracter item s, but not included in either experim ental phase. The practice item s were followed random ly by either LLLL (the cue to learn) or FFFF (the cue to forget). The first 5 pract ice item s were followed by a short cue-delay and the latter half were followed by a long cue-delay. Once participants had indicat ed that they understood the task, the study phase began. As in the pract ice trial, words were displayed individually in black text on a laptop com puter screen with a white background. Each word was centred horizontally and vert ically on the screen and rem ained on display for 1 s. The cues were also centred horizontally and vertically. I n the short delay condit ion, the cue appeared directly ( 0 s.) after the st udy-word. The cue (LLLL or FFFF) was then displayed for 1 s., followed by a 5 s. interval before the next studyword appeared. I n the long-delay condit ion, the cue appeared after a 5 s. delay following the st udy word. The next st udy word im m ediately followed (0 12

14 s.) the cue s 1 s. presentation period. Consequently, the total presentation tim e was 7 s. for both the short and long delay condit ions. At the end of the study phase, part icipants were inform ed that they would take part in a recognit ion test in which they were unexpectedly asked to ident ify ALL of the studied item s (regardless of the cue). They were also given detailed instruct ions on the Rem em ber/ Know procedure (see Appendix 2). The term s Type A and Type B were used instead of Rem em ber and Know, to ensure that the subsequent decisions were not influenced by the use of these words in everyday language. The dist inct ion between Type A and Type B awareness was based on the instruct ions used in previous experim ents with sim ilar sam ples who did not show any difficulty with the dist inct ion ( Bowler et al. 2000). To further ensure that there were no between-group biases with the understanding between these two states of awareness, part icipants were then given exam ples of hypothetical descriptions of m em ory, which they were asked to ident ify as either Type A, or Type B responses (Appendix 2). Once a part icipant had dem onstrated an ability to dist inguish between the two types, they took part in a recognition test of the 88 study item s plus 88 distractor item s (176 words in t otal). The presentation of the words was random ised, so that the distracter and study item s ( whether to-be-learned, to-be-forgotten, or short-cued or long-cued) were all interm ingled. These item s were displayed individually on the sam e laptop com puter in the sam e font type and size as at test. Part icipants were rem inded that they should give a yes response if they recognised the test it em, even if they knew it was a word they were instructed to forget. I f the part icipant responded yes, they were then asked to indicate whether they had a Type A or a Type B recollection. Responses were selfpaced and the experim enter recorded the responses via the keyboard. From 13

15 tim e to t im e the experim enter would ask the participant to clarify the choice of response ( A/ B) by asking them to describe what they rem em bered about the word, to check that these descript ions fitted with the A/ B dist inct ion. Each part icipant was instructed not to guess and to only indicate a posit ive recognit ion if they were sure that they had seen the word at test ( Appendix 2). 14

16 Results The dem ographic characterist ics of the ASD and TD participants were sim ilar and are sum m arised in Table 1. The m ean age of all part icipants was 37 years and the m ean VI Q score was 105. The VI Q and FI Q of the two part icipants without a form al diagnosis of an ASD were 111 and 135 (VI Q) and 119 and 140 ( FI Q), respectively. When the two part icipants without confirm at ion of a form al diagnosis of AS, t ogether with their paired TD com parisons were excluded, the overall m ean age was 36 years and the m ean VI Q score was 103. (I nsert Table 1). The average proportion of correctly recognised words was calculated for each participant group according to the t ype of response ( Rem em ber/ Know) for each condit ion, as well as words incorrectly ident ified. The m ean proportion (SD) of unstudied words reported as having been seen were very low for both the ASD participants ( 0.03 [ 0.03] and 0.07 [ 0.08] ) for Rem em ber and Know responses, respectively) and TD part icipant s ( 0.01 [ 0.03] and 0.07 [ 0.11] ) for Rem em ber and Know responses, respectively), thus indicat ing that response criteria were sim ilarly stringent for each group. Com parison of the overall recognit ion data between groups indicate that the proportion of to-be-learned item s correctly recognised by the ASD group is considerably lower than the TD group following either a short, or a long cuedelay (Table 2). I n contrast, the proport ion of to-be-forgotten words is sim ilar for both groups regardless of the delay. 15

17 (I nsert Table 2). These data were analysed by a 2 (ASD vs. TD groups) x 2 ( learn vs. forget instruct ion) x 2 ( short vs. long cue-delay) x 2 ( Rem em ber vs. Know response) ANOVA. Pearson r was calculated as an estim ate of effect size (.2 = recom m ended m inim um effect size;.5 = m oderate effect;.8 = strong effect) as it is reported as a good indicator of strength of association for psychological data that is norm ally distributed (Ferguson 2009). The results of the all part icipant analysis show a significant effect of the type of instruct ion (F ( 1,30) = 80.29; p <.01; r =.85). Overall, the recognition of to-be-learned words was higher than the recognit ion of to-beforgotten words. There was also a significant m ain effect of response type (F ( 1.30) = 11.55; p <.01; r =.53), with there being m ore Rem em ber responses on average, than Know responses ( Table 2). There was a significant interact ion between the type of instruct ion and the length of cue delay ( F ( 1.30) = 11.02; p <.01; r =.52). The recognition of to-be-learned words was greater following a short cue-delay than a long cue-delay; whereas recognition of to-be-forgotten words was sim ilar for both cue condit ions ( Table 2). There was also a significant interact ion between the type of instruct ion and the type of response (F ( 1,30) = 32.44; p <.01; r =.72). Overall, m ore to-be-learned item s were recognised with a Rem em ber response (M = 0.48, Se = 0.04) than a Know response (M = 0.20, Se = 0.02) and, conversely, m ore to-be-forgotten item s were recognised with a Know response (M = 0.23, Se = 0.20) than a Rem em ber response (M = 0.20, Se = 0.03). Each of the results reported above rem ained significant, with p <.01 when the analysis was conducted 16

18 excluding the two participants without confirm at ion of a form al diagnosis of AS and their paired TD com parisons. Of note, there was a significant three-way interact ion between the type of instruct ion, the length of cue-delay, and response type in the all part icipant analysis (F ( 1,30) = 6.39; p =.02; r =.42), and in the analysis excluding the two part icipants without confirm at ion of a form al diagnosis of AS and their paired TD com parisons (F ( 1,26) = 6.87; p =.01; r =.46). This interaction therefore indicates that the m anipulated factors of the experim ent each had a different im pact on the type of response of the part icipants overall. The between participant ANOVA results of the all part icipant analysis showed a significant interaction between instruction and group (F ( 1,30) = 6.90; p =.01; r =.43). This stat ist ically significant interact ion was m aintained in the analysis excluding the two part icipant s without confirm at ion of a form al diagnosis of AS and t heir paired TD com parisons (F ( 1,26) = 5.33; p =.03; r =.41). As we predict ed that the ASD group would be im paired in their recognit ion of to-be-learned words and not in their recognit ion of to-beforgotten words and that this difference would be m ost notable in the short cue-delay condit ion, independent t-tests were perform ed on the m ean recognit ion of to-be-learned and to-be-forgotten words. I n the short cue-delay condit ion, the Levene s test for equalit y of variances was significant for the tobe-learned words indicating that the assum ption of equal variances between groups in this condit ion had been violated. The t value with adj usted df to account for the inequalit y of variance was t (23) = 2.52; p = 0.19; Table 2. Levene s test was not significant for either the to-be-forgotten words in this condition, or for the long cue delay condition. The difference in recognition of to-be-learned words between the ASD and TD participants in the long cue- 17

19 delay condit ion was nearing significance (t ( 30) = 1.79; p =.08); whereas the difference in recognit ion between the two groups for to-be-forgotten words was not significant ( Table 2). When this analysis was conducted excluding the two part icipants without confirm at ion of a form al diagnosis of AS and their paired TD com parisons, there was a significant difference in recognit ion of to be learned words following a short-cue delay (t ( 18) = 2.82; p =.01) and following a long cue delay (t ( 26) = 2.48; p =.02). The results of the all part icipant analysis ANOVA showed a significant three-way interact ion between group, type of instruct ion, and Rem em ber/ Know response (F ( 1,30) = 4.65; p =.04; r =.37). The pattern of results in Table 2 shows that the instruct ion to learn or forget had a dissociat ive effect on the Rem em ber versus Know responses in the TD group (i.e. the instruct ion to learn increased Rem em ber responses com pared with Know and the instruct ion to forget increased Know responses com pared with Rem em ber). I n the ASD group, however, the instruct ion to learn increased Rem em ber responses com pared with Know, but the instruct ion t o forget resulted in a sim ilar num ber of Rem em ber and Know responses. However, when the data were analysed excluding the two part icipants without confirm at ion of a form al diagnosis of AS and their paired TD com parisons this three way interaction was no longer stat ist ically significant (F ( 1,26) = 3.79; p =.06; r =.36). Post hoc analysis The SDs for to-be-learned words were larger for the ASD group than for the TD with regard to Rem em ber responses (Table 2). The range of scores with a Rem em ber for the ASD group was.05 to.95 com pared wit h.23 to.95 18

20 for the TD group for the to-be learned words following a short cue-delay and.14 to 1.0 (ASD) versus.32 to 1.0 ( TD) following a long cue-delay. Only 8 of the ASD participants ( 50% ) m et or exceeded the first 25 th percentile score of the TD group for Remem ber responses following both a long-, and a short-cue delay. However, the AS participants with the lowest recognition scores in the short cue-delay condition were not the sam e participants with the lowest scores in the long cue delay condit ion, indicating that greater variability in the scores was not m erely due to part icular outliers. I n order to investigate this further, and to consider the role of I Q in relation to the deficit in recall of tobe-learned words for the ASD group, Pearson s correlational analysis were conducted to establish a potential relat ionship between VI Q and recognit ion Rem em ber outcomes for to-be-learned words following a short and long cuedelay for both ASD and TD groups. There was a significant correlat ion between VI Q and the recognition of to-be-learned words with a Rem em ber response following a short cue-delay in the TD sam ple (r =.55; p =.03), but not for the ASD sam ple (r = -.05; p =.85). There was no correlat ion between VI Q and recognit ion of to-be-learned words with a Rem em ber response following a long cue-delay in the TD sam ple (r =.25; p =.35), nor in the ASD sam ple, although this was nearing significance ( r =.46; p =.07). Sim ilar analyses were conducted with the PI Q scores and all revealed non-significant results, although PI Q and the recognition of to-be-learned words with a Rem em ber response for the ASD part icipants following a long cue-delay was also nearing significance (r =.44; p =.09). 19

21 Discussion The overall recognit ion of to-be-learned words was greater than to-beforgotten words in both cue-conditions, regardless of part icipant group. These results indicate that t he adults with ASDs and the TD group were subj ect to the directed forgetting effect (Basden and Basden 1996; Bj ork and Woodward 1973; Woodward and Bj ork 1971). The statist ically significant three-way interact ion between instruction, delay and response shows that the m anipulat ion of the cue tim ing im pacted differently upon the instructions to learn or forget and, in turn, on the recollective experience of Rem em ber versus Know. A com parison of the m eans presented in Table 2 shows that both groups recognised proportionately m ore to-be-learned words following a short cue-delay ( when potential to engage in elaborative rehearsal is m axim ised), than a long cue-delay. Conversely, the recognition of to-be-forgotten words was proportionally higher following a long cue-delay (when use of rote rehearsal is increased while await ing for instruction to learn or forget) than a short cue-delay. This pattern of results replicates the previous findings of Gardiner et al., (1994) and Woodward et al., (1973) and is consistent with the theory that the cue delay m anipulates the rehearsal strategies em ployed. Com parison of the data in term s of Rem em ber and Know in Table 2 reveals that the advantageous recognit ion of to-be-learned words in bot h groups of part icipants was only evident in Rem em ber responses, not Know. As Rem em ber responses reflect autonoetic awareness, and as the increased recognition of to-be-learned words was evident in Rem em ber responses but not in Know responses for both groups, these results are consist ent with previous findings that elaborat ive rehearsal independently influences Rem em ber responses (Dobbins et al. 2004; Gardiner et al. 1994; Gardiner and 20

22 Richardson-Klavehn 2000). I t is also consistent with previous findings that both the effortful learning strategies and the effects on subjective awareness are qualitat ively sim ilar for part icipants wit h ASDs and com parisons (Bowler et al. 2007). We proposed that the adults with ASDs would display a dim inished directed-forgetting effect com pared with TD adults as a result of less effective learning strategies for the to-be-learned words. Consistent with t his hypothesis, the proportion of to-be-learned words correctly recognised by the ASD group was lower, on average, than those correctly recognised by the TD group, whereas the proportion of to-be-forgotten words was sim ilar between the two groups, regardless of the cue-delay. These results support previous findings that adults with ASDs engage in less effective learning st rategies, on average, than TD adults, at least when the words have been selected for their potential to engage in elaborat ion, as is the case for concrete nouns (Toichi and Kam io 2003). These results are also consistent with the encoding deficit hypothesis previously offered to explain m ild m em ory im pairm ents in sim ilar sam ples (Bowler et al. 2000; Sm ith et al. 2007; Toichi and Kam io 2003). Based on previous findings that adults with HFA display m ild episodic m em ory im pairm ents (Bowler et al. 2000; Gardiner et al. 2003), we hypothesised that any potential deficit in the recognition of to-be-learned words for the ASD group would be evident in Rem em ber responses, but not Know. The results in Table 2 show that the ASD group recognised fewer to-belearned words with a Rem em ber response, on average, following both a short and long cue-delay than the TD group; whereas the Know responses for each group were sim ilar in both cue-delay condit ions. As Rem em ber responses have been previously shown to be increased by elaborat ive rehearsal and Know 21

23 responses have been shown to be increased by rote rehearsal ( Dobbins et al. 2004; Gardiner et al. 1994; Gardiner and Richardson-Klavehn 2000), the dim inished recognit ion of study item s in Rem em ber responses and not Know is consistent wit h our hypothesis of an elaborative encoding deficit and indicates that the effortful learning strategies em ployed by the adults wit h ASDs were typically less effect ive than those em ployed by the TD group, either because the ASD participants engaged in less elaborative rehearsal, or the elaborat ive rehearsal they did engage in did not enhance their recollect ive experience to the sam e extent as that of the TD group. I t was also hypothesised that this effect of cue delay m ight have greater im pact on the TD group than the ASD group, part icularly if the ASD part icipants were not able to take full advantage of the addit ional t im e available to engage in elaborat ive rehearsal. The m eans in Table 2 show that the recognition of both ASD and TD participants was im proved by the addit ional t im e available following the short cue-delay and both these increased responses are reflected in Rem em ber responses, not Know. However, there was a statist ically significant deficit in the overall recognition of to-be-learned words for the ASD group following a short cue-delay in com parison to the TD group. ASD part icipants also recognised fewer to-belearned words, on average, than the TD group following a long cue-delay, but not to a significant extent. Previous research has indicated that task support at the retrieval stage (Boucher 1981; Boucher and Warrington 1976b; Bowler et al. 2004; Bowler et al. 1997) can im prove perform ance for individuals with ASDs, so establishing ways to support and encourage m ore consistent and effective learning strategies in individuals with ASDs m ay equalize perform ance in m em ory tasks 22

24 in which som e of this population are currently under-perform ing. However, it is im portant to note that not all of the part icipants in the ASD group were outperform ed by the TD group in this current t ask. The larger SDs for the ASD participants com pared with the TD part icipants for Rem em ber responses of to-be-learned words shows that the perform ance of the ASD participants was m ore inconsistent than the com parison sam ple ( as the TD participants individual perform ance was m ore stable across the two condit ions). Other trends in the data support the hypothesis that different strategies were being em ployed, at least by som e of the ASD participants. For exam ple, there was a significant three-way interact ion between instruct ion, response and group in the analysis including all part icipants, indicat ing that the instruction to learn, or forget, im pacted differently on each group in relat ion to Rem em ber and Know responses. This level of significance was not m aintained when the analysis was conducted excluding t he two part icipants without a confirm ed form al diagnosis of AS and their appropriately m atched TD pair. However, this m ay have been due to a lack of power in t he data excluding these part icipants as the results were in the sam e direct ion and j ust m issed significance, whereas the effect sizes of each analysis with and without these part icipants were very sim ilar. There was also a very slight trend for the ASD participants to recognise m ore to-be-forgotten words with a Rem em ber response than the TD part icipants and for t he TD participants to recognise m ore to-be-forgotten words with a Know response than the ASD part icipants, at least in the short cue-delay condit ion. The results of the post-hoc analysis was also of interest, as the relat ionship wit h VI Q on the recognition perform ance for to-be-learned 23

25 words with Rem em ber response differed between the two groups. Specifically, VI Q was related to the recognit ion of to-be-learned words with a Rem em ber response for the TD group following a short-cue delay, but not following a long-cue delay; whereas VI Q was not related to the recognit ion of to-belearned words with a Rem em ber response for the ASD group following a shortcue delay, but neared significance following a long cue-delay. Alt hough not a robust finding in this experim ent, this trend for atypical patterning of Rem em ber/ Know responses for the to-be-learned versus to-be-forgotten words, together with the atypical correlat ional trends in VI Q/ PI Q and episodic m em ory perform ance m ay provide further j ustificat ion to assess the potential for individuals with HFA to engage in com pensatory learning strategies (Toichi and Kam io 2002). Recent EEG studies during recognit ion m em ory tasks in adults with and without ASDs revealed functional neurological differences between the two groups, regardless of task perform ance (Massand and Bowler 2013; Massand et al. 2013). Specifically, in a context m em ory task, TD adults showed different ERP patterns according to the type of task, while adults with ASDs showed sim ilar ERP patterns across the tasks (Massand and Bowler 2013). I n a yes-no recognit ion task, ASD adults showed a pariet al focus for old-versus new recognit ion com pared with a m id-frontal bias for sim ilar tasks in TD adults (Massand et al. 2013). Together with the inconsistency in perform ance across condit ions for som e individuals with ASDs in this current study, these results are suggestive of an executive dysfunct ion account of autism (Ozonoff et al. 1991; Russell 1997), or if considered in term s of failing to generalise strategies across condit ions is consistent with the week central coherence account (Frith and Happe 1994; Happe 1994). Our st udy did not direct ly test these skills, so no furt her inferences can be m ade in this regard, 24

26 but our results provide j ustificat ion to explore and direct ly com pare learning strategies between ASD and com parison sam ples. I n part icular, further research to investigat e the role of VI Q and PI Q on verbal m em ory tasks in adults with ASDs appears warranted. Conclusion The m ild deficits in the ASD group in com parison to the TD group evident in this study were specific to their episodic recollect ion of words that they were instructed t o learn. These findings not only refine our understanding of the m ild m em ory im pairm ents previously established in individuals with ASDs, but provide an im portant link between the deficits in free-recall and im paired recognit ion in this populat ion, as elaborat ive rehearsal has been shown to im pact upon both (Craik and Tulving 1975; Geiselm an and Bj ork 1980; Greene 1987; Rundus 1977). Our findings indicate that adults with ASDs engage in less effective elaborat ive rehearsal than TD adult s. Elaborat ive rehearsal has been shown to enhance recall of part icular m aterial, m ost notably concrete nouns, or words with associat ive qualit ies (Paivio and Csapo 1969; Toichi and Kam io 2003), so the hypothesis of an elaborat ive encoding deficit can feasibly be offered to explain previous findings of recall deficits in adults with ASDs, such as the reduced levels-of-processing and concrete noun effects (Boucher and Warrington 1976a; Bowler et al. 2000; Bowler et al. 1997; Sm ith et al. 2007; Tager-Flusberg 1991; Toichi and Kam io 2003). Elaborat ive rehearsal im proves autonoetic awareness and is dem onstrated by increased Rem em ber responses in com parison to Know responses (Dobbins et al. 2004; Gardiner et al. 1994; Gardiner and Richardson-Klavehn 2000). Consequently, our findings also establish a link between the im paired episodic 25

27 m em ory in adults wit h ASDs (Bowler et al. 2007; Bowler et al. 2000; Gardiner et al. 2003; Souchay et al. 2012) and the encoding deficit hypothesis. As yet, there is no evidence to suggest that individuals with ASDs cannot engage in elaborative rehearsal as effectively as TD individuals, m erely that they do not appear to naturally em ploy elaborat ive strategies to the sam e extent. Our results suggest that som e individuals with ASD m ay have a norm al capacity and efficiency to em ploy elaborat ive rehearsal strategies. I ndeed, the fact that, like the TD group, the individuals with ASDs also benefited from the addit ional t im e available in the short cue-delay condition does offer som e support for the notion that they are engaging in elaborat ive rehearsal, but in som e cases, they appear to do so less effectively. 26

28 Conflict of I nterest The authors declare that they have no conflict of interest. References Am erican Psychiatric Association (2000). Diagnost ic and stat ist ical m anual of m ental disorders (4th ed., text rev.). Washington, DC. Basden, B.H. & D.R. Basden (1996). Directed forgetting: further com parisons of the item and list m ethods. Mem ory 4, Beversdorf, D.Q., et al. ( 1998). The effect of sem antic and em otional context on written recall for verbal language in high funct ioning adults with autism spectrum disorder. Journal of neurology, neurosurgery, and psychiatry 65, Bj ork, R.A. & A.E. Woodward ( 1973). Directed forgetting of individual words in free recall. Journal of Experim ental Psychology 99, Boucher, J. (1981). I m m ediate free recall in early childhood aut ism : another point of behavioural sim ilarity with the am nesic syndrom e. Br J Psychol 72, Boucher, J. & E. Warrington ( 1976a). Mem ory deficits in early infantile aut ism : som e sim ilarities to am nesic syndrom e. British Journal of Psychology 67, Boucher, J. & E.K. Warrington (1976b). Mem ory deficits in early infant ile autism : som e sim ilarities to the am nesic syndrom e. Br J Psychol 67, Bowler, D.M., S.B. Gaigg & J.M. Gardiner (2010). Mult iple list learning in adults with aut ism spectrum disorder: parallels with frontal lobe dam age or 27

29 further evidence of dim inished relat ional processing? J Autism Dev Disord 40, doi: / s x Bowler, D.M., J.M. Gardiner & N. Berthollier ( 2004). Source m em ory in adolescents and adult s with Asperger's syndrom e. J Autism Dev Disord 34, Bowler, D.M., J.M. Gardiner & S.B. Gaigg ( 2007). Factors affecting conscious awareness in the recollect ive experience of adults with Asperger's Syndrom e. Consciousness and cognition 16, Bowler, D.M., J.M. Gardiner & S. Grice ( 2000). Episodic m em ory and rem em bering in adult s with Asperger Syndrom e. Journal of Autism and Developm ental Disorders 30, Bowler, D.M., N.J. Matthews & J.M. Gardiner (1997). Asperger's Syndrom e and m em ory: Sim ilarity to aut ism but not am nesia. Neuropsychologia 35, Craik, F.I.M. & R.S. Lockhart ( 1972). Levels of processing: A fram ework for m em ory research. Journal of Verbal Learning and Verbal Behavior 11, Craik, F.I.M. & E. Tulving (1975). Depth of processing and the retention of words in episodic m em ory. Journal of Experim ental Psychology: General 104, Dobbins, I.G., N.E.A. Kroll & A.P. Yonelinas ( 2004). Dissociat ing fam iliarit y from recollection using rote rehearsal. Mem ory & Cognition 32, Ferguson, C.J. (2009). An effect size prim er: A guide for clinicians and researchers. Professional Psychology: Research and Pract ice 40,

30 Frith, U. & F. Happe ( 1994). Aut ism : beyond "theory of m ind". Cognition 50, Gaigg, S.B., J.M. Gardiner & D.M. Bowler ( 2008). Free recall in autism spectrum disorder: the role of relat ional and item -specific encoding. Neuropsychologia 46, doi: / j.neuropsychologia Gardiner, J.M., D.M. Bowler & S.J. Grice (2003). Further evidence of preserved prim ing and im paired recall in adults with Asperger's Syndrom e. Journal of Autism and Developm ental Disorders 33, Gardiner, J.M., B. Gawlik & A. Richardson-Klavehn ( 1994). Maintenance rehearsal affects knowing, not rem em bering; elaborat ive rehearsal affects rem em bering, not knowing. Psychonom ic Bulletin & Review 1, Gardiner, J.M. & A. Richardson-Klavehn ( 2000). Rem em bering and knowing. I n: Tulving, E., F.I.M. Craik ( eds) Handbook of m em ory.( pp ). Oxford: Oxford University Press. Geiselm an, R.E. & R.A. Bj ork ( 1980). Prim ary versus secondary rehearsal in im agined voices: differential effects on recognition. Cognitive Psychology 12, Greene, R.L. ( 1987). Effects of m aintenance rehearsal on hum an m em ory. Psychological Bulletin 102, Happe, F.G. ( 1994). An advanced test of theory of m ind: understanding of story characters' thoughts and feelings by able aut ist ic, m entally handicapped, and norm al children and adults. J Autism Dev Disord 24,

31 Herm elin, B. & N. O'Connor (1967). Rem em bering of words by psychotic and subnorm al children. Br J Psychol 58, Lopez, B. & S.R. Leekam ( 2003). Do children with aut ism fail to process inform at ion in context? Journal of child psychology and psychiatry, and allied disciplines 44, MacLeod, C. ( 1999). The item and list m ethods of directed forgetting: test differences and the role of dem and charact erist ics. Psychonom ic Bulletin & Review 6, Massand, E. & D.M. Bowler ( 2013). Atypical Neurophysiology Underlying Episodic and Sem ant ic Mem ory in Adults with Autism Spectrum Disorder. J Autism Dev Disord, doi: / s Massand, E., D.M. Bowler, L. Mottron, A. Hosein & B. Jem el ( 2013). ERP correlates of recognit ion m em ory in Aut ism Spectrum Disorder. J Autism Dev Disord 43, doi: / s x Mottron, L., K. Morasse & S. Belleville (2001). A study of m em ory funct ioning in individuals with aut ism. Journal of Child Psychology and Psychiatry 42, Ozonoff, S., B.F. Pennington & S.J. Rogers ( 1991). Executive function deficits in high-funct ioning autist ic individuals: relat ionship to theory of m ind. Journal of child psychology and psychiatry, and allied disciplines 32, Paivio, A. & K. Csapo ( 1969). Concrete im age and verbal m em ory codes. Journal of Experim ental Psychology 80, Paivio, A., J.C. Yuille & S.A. Madigan ( 1968). Concreteness, im agery, and m eaningfulness values for 925 nouns. Journal of Experim ental Psychology Monograph Supplem ent 76,