Abstract

Objective: There appears to be a common network of brain regions that underlie the ability to recall past personal experiences (episodic memory) and the ability to imagine possible future personal experiences (episodic future thinking). At the cognitive level, these abilities are thought to rely on "scene construction" (the ability to bind together multimodal elements of a scene in minddependent on hippocampal functioning) and temporal "self-projection" (the ability to mentally project oneself through time-dependent on prefrontal cortex functioning). Although autism spectrum disorder (ASD) is characterized by diminished episodic memory, it is unclear whether episodic future thinking is correspondingly impaired. Moreover, the underlying basis of such impairments (difficulties with scene construction, self-projection, or both) is yet to be established. The current study therefore aimed to elucidate these issues. Method: Twenty-seven intellectually highfunctioning adults with ASD and 29 age-and IQ-matched neurotypical comparison adults were asked to describe (a) imagined atemporal, non-self-relevant fictitious scenes (assessing scene construction), (b) imagined plausible self-relevant future episodes (assessing episodic future thinking), and (c) recalled personally experienced past episodes (assessing episodic memory). Tests of narrative ability and theory of mind were also completed. Results: Performances of participants with ASD were significantly and equally diminished in each condition and, crucially, this diminution was independent of general narrative ability. Conclusions: Given that participants with ASD were impaired in the fictitious scene condition, which does not involve self-projection, we suggest the underlying difficulty with episodic memory/future thinking is one of scene construction. Keywords: autism spectrum disorder, episodic memory, episodic future thinking, scene construction, self-projection Recently, an important link has been made between the ability to mentally re-experience past episodes (episodic memory) and the ability to imagine episodes that one might plausibly experience in the future. This latter ability to mentally pre-experience possible future events has been termed "episodic future thinking" Theories of the Underlying Link Between Episodic Memory and Episodic Future Thinking According to one prominent theory, episodic memory and episodic future thinking are linked because both involve elements of self-awareness. In particular, A second prominent theory, put forward by Hassabis and colleagues (e.g., Hassabis et al. accept that temporal self-projection and selfrelated processing play a role in episodic memory and episodic future thinking. However, these processes are considered "addons" to the basic contribution to episodic memory and episodic future thinking of the ability to construct multimodal scenes in one's mind. Specifically, Hassabis, Kumaran, and Maguire (2007, p. 14372) argue that episodic memory and episodic future thinking involve "at least two components with dissociable neural bases: a network centered on the hippocampus responsible for scene construction, with the amPFC [anterior medial prefrontal cortex], PCC [posterior cingulate cortex] and precuneus mediating selfprojection in time, sense of familiarity, and self-schema" (see Furthermore, argue that each of the multiple cognitive functions described by Buckner and Carroll (2007; episodic memory, episodic future thinking, theory of mind, and navigation) may rely to a greater or lesser extent on these subcomponents. For example, whereas navigation might rely exclusively upon the hippocampal (scene construction) system, theory of mind might rely exclusively on the frontal (self-related processing) system, and episodic memory and episodic future thinking might rely on both systems. Autism spectrum disorder (ASD) refers to a set of developmental disorders diagnosed on the basis of significant behavioral impairments in social interaction, communication, and behavioral flexibility (American Psychiatric Association, 2000; World Health Organization, 1993). At the cognitive level of description, ASD is characterized by a selective diminution of episodic memory, leaving semantic memory undiminished (e.g., To our knowledge, only two full studies have been published on episodic future thinking in ASD. 1 Lind and Bowler (2010) adopted a method originally devised by D'Argembeau and Recently, Crane, Lind, and Bowler (2013) sought to assess episodic memory and episodic future thinking in ASD using a different method from that employed by Although the results of On the one hand, it is possible that both episodic memory and episodic future thinking deficits in ASD are explained by an underlying difficulty with basic scene construction (associated with hippocampal dysfunction). Certainly, this idea dovetails certain existing theories of the causes of episodic memory deficits in ASD. In particular, Bowler and colleagues (e.g., Bowler, Gaigg, & Lind, 2011) have suggested that a difficulty with "binding," which involves encoding the relations between elements that comprise an episode into a single representation and which is associated with (anterior) the hippocampal functioning, plays a central role in producing the memory profile characteristic of ASD On the other hand, difficulties with episodic memory and episodic future thinking in ASD could be explained instead by a selective deficit in self-projection (associated with prefrontal cortex dysfunction). It may be that individuals with ASD are fully capable of forming coherent, multimodal representations of atemporal, non-self-related fictitious scenes (i.e., that people with ASD have intact scene construction ability), but have difficulty mentally projecting themselves through time to "identify with" a past state of self or an anticipated future state of self. In other words, the difficulty in ASD could be with self-projection/self-related processing, rather than with scene construction. This idea is consistent with the notion that ASD involves diminished awareness of aspects of self (e.g., Williams, 2010), as well as with theories that explicitly implicate diminished self-awareness as a contributory cause of the specific profile of memory that characterizes ASD (e.g., Lind, 2010). To explore these issues, we employed a version of the experimental task developed by Hassabis, . To assess episodic future thinking ability, participants were asked to provide detailed descriptions of imagined specific events that 1 An additional pilot study of future thinking in ASD was published by 57 EPISODIC COGNITION IN AUTISM they might plausibly experience in the future (possible future Christmas event, possible event over next weekend, possible future meeting with a friend/relative). To assess episodic memory ability, participants were also asked to describe memories of specific previously experienced past events (last birthday, last week, and last time they went shopping). Finally, to assess the ability to imagine atemporal, non-self-relevant fictitious scenes (i.e., scene construction ability), participants were asked to provide detailed descriptions of imagined commonplace settings (beach, market, ship, pub, forest, and museum). Our rationale was that whereas all conditions of the task required basic scene construction ability, only the past and future events conditions of the task additionally required self-projection, because only these conditions involved imagining the self-relevant scenarios requiring mental time travel. As such, if episodic memory and episodic future thinking deficits in ASD are primarily due to a diminution of self-projection, then we should expect to see impaired performance among ASD participants in the past and future events conditions only. Alternatively, if episodic memory and episodic future thinking deficits in ASD are primarily the result of diminished scene construction ability, then the performance of participants with ASD should be equally impaired across all conditions. Of course, a final possibility is that difficulties with scene construction and difficulties with self-projection independently contribute to episodic future thinking and episodic memory deficits in ASD. In this case, participants with ASD should be impaired in all conditions, but relatively more so in the past and future events conditions. We also included a narrative control task, which involved participants providing an ongoing narrative of a 24-page picture book, Frog, Where Are You? Finally, we also employed the animations task 2 On this basis, episodic future thinking ability and episodic memory ability might be associated significantly with theory of mind ability, whereas basic scene construction ability should not be. Notably, the majority of the evidence on which these theories are based is derived from neuroimaging studies. As far as we know, no study has explicitly investigated the association between episodic future thinking (or the processes of scene construction and self-projection that arguably underlie episodic future thinking) and theory of mind using cognitiveexperimental methods. As such, this was an important aim of our study. Method Participants Ethical approval for this study was obtained from the appropriate university ethics committee. Twenty-seven adults with highfunctioning ASD (21 male; 6 female) and 29 neurotypical comparison adults (22 male; 7 female) took part in this experiment, after giving written, informed consent to take part. All participants were paid standard university fees for their participation. Participants with ASD were recruited (a) via an advertisement on the "Research Projects: Be a Participant" page of The National Autistic Society U.K. Web site (www.autism.org.uk), (b) through local ASD support groups, (c) through the Durham University Service for Students with Disabilities, and (d) through word of mouth. The majority of comparison participants were recruited through advertisements in local newspapers. However, a small number took part in order to receive course credits in partial fulfillment of their undergraduate psychology degrees. Inclusion criteria included having a full-scale IQ of at least 85, being aged 16 to 65 years, and having no neurological or psychiatric disorders other than ASD (no participants needed to be excluded for failing to meet these criteria). Participants in the ASD group had all received formal diagnoses of autistic disorder (n ϭ 5) or Asperger's disorder (n ϭ 22), according to conventional criteria (American Psychiatric Association, 2000; World Health Organization, 1993). All documented diagnostic information was checked thoroughly and provided sufficient information to ensure diagnostic criteria were met in each case. To assess severity of current ASD features among participants with ASD and the presence of ASD-like features among comparison participants, several measures were taken. First, participants 2 It should be noted that there is some ambiguity in the position that Hassabis and colleagues take with respect to the necessity of scene construction for theory of mind. Of the diverse cognitive functions underpinned by self-projection according to Second, a relative or long-standing friend of each participant completed a prepublication version of the Social Responsiveness Scale, Second Edition (SRS-2; Constantino & Gruber, 2012). Scores on this detailed questionnaire provide a valid and reliable indicator of participants' social and communicative abilities. Only three participants with ASD missed the ASD cutoff on the SRS-2 (raw score Ն 60). All comparison participants scored below the ASD cutoff on the SRS-2. Thus, none showed any sign, according to relatives/friends, of manifesting significant ASD-like traits. In addition, 19 of 27 participants with ASD (the remaining 10 participants in the group were unwilling to take part in the assessment) also completed the Autism Diagnostic Observation Schedule-Generic (ADOS; Using the Wechsler Abbreviated Scale of Intelligence (WASI; Wechsler, 1999), the groups were matched closely for verbal and nonverbal ability. The groups were also matched closely for chronological age. Importantly, all effect sizes associated with group differences in baseline characteristics of age and IQ were negligible/small. Participant characteristics are presented in Test and Procedures Experimental (episodic memory, episodic future thinking, and scene construction) task. Each participant was tested individually in a quiet room, and sat opposite the experimenter. Participants were instructed that they would be asked to imagine or remember vivid scenes in their mind, based on a cue card that would set the scene. They would then have to describe this mental representation to the experimenter in as much detail as possible. Before commencing the task, an example cue card ("Imagine you're sitting on a bench in a park. Describe the scene in as much detail as possible") was given by the experimenter, who also provided a model answer. It was highlighted to participants that the description given by the experimenter was multimodal, containing not only visual details but also smells, sounds, and so forth. Each participant was asked to produce a description for 12 scenarios, split into three separate conditions: past events (last week, last birthday, last time they went shopping), future events (this weekend, next Christmas, next time they see a friend or relative), and fictitious scenes (beach, museum, pub, ship, market, forest). Trials were blocked by condition and, across participants, the three conditions were presented in counterbalanced order. For all descriptions in the future events and fictitious scenes conditions, participants were explicitly instructed not to recount an actual memory, or any part of one, but rather to generate a specific novel episode/scene in their mind. In contrast, for the past events condition, participants were told they must recall and describe a real personally experienced episode. Participants' descriptions were audio recorded for later transcription and coding. On each trial, a cue card was placed on the table in front of the participant, detailing a short description of the scenario to be described (e.g., "Standing by a small stream, somewhere deep in a forest"). This card remained on the table throughout each trial, to act as a cue to the participant and remind them of the scenario if necessary. The experimenter read aloud this scenario and asked the participant to produce a vivid multimodal description of the experience and surroundings, using all of their senses. A probing protocol was followed (as outlined in Hassabis, , such that general prompts were given if a description could not be provided or lacked detail (e.g., "Tell me more about the scene"). If a participant became fixated on one aspect of the scene, they were encouraged to move on, and if they provided poor detail, they were asked to elaborate further. Only such general prompts were given, and the experimenter was careful not to lead the participant or introduce any aspect or detail that had not been mentioned by the participant previously. Participants were encouraged to continue with their descriptions until the Following each trial, participants completed a questionnaire rating their response to the cue on a series of elements including how salient the imagined/remembered scene/episode was and how much of a sense of presence they had when imagining/remembering the scene/episode. They were also presented with a series of 12 statements that were designed to gauge how integrated or fragmented their description was thought to be (e.g., "It was not so much a scene as a collection of images"; "I could see it as one whole scene in my mind's eye"). Participants were asked to select those statements from the series that most applied to their imagined/remembered scene/episode. Scoring. Each description was transcribed from audio recordings by an independent transcriber (who was blind to group status and to the aims of the study) and the transcription was subsequently coded according to the detailed guidelines provided by Hassabis, . This coding was carried out by an independent rater who was also blind to group status and to the aims of the study (she had access only to written transcripts-she was not involved in any testing and did not have access to any audio recordings). To assess the reliability of the judgments provided by the main rater, a second coder rated a randomly selected subset (n ϭ 30; 54%) of the transcripts (see reliability values below in the "Description content" and "Independent quality ratings" subsections). For each description, a composite "experiential index" score was calculated, ranging from 0 to 60. This provides the key overall measure of how rich and detailed each description was. The composite score was calculated by combining the following four subcomponent scores. Description content. Within each description, statements were coded as belonging to one of four categories: spatial references, entity presence, sensory description, or thought/emotion/action. On the basis of pilot studies, Hassabis, argued that the production of seven instances per category in each description should be considered a reflection of optimal performance. Thus, the total for each category was a maximum score of 7. This yielded a score out of 28 for each description, as an indication of content quality. Interrater reliability for description content across the four categories was high, Cronbach's ␣ ϭ .97. Participant questionnaire ratings. Ratings from two of the questions that participants completed in the postdescription questionnaires were included in the experiential index. First, their sense of presence in the description was rated on a scale of 1 to 5 (did not feel like I was there at all to strongly felt like I was really there). Second, the perceived salience of the imagined scene was also rated on a scale of 1 to 5 (couldn't really see anything to extremely salient). Each of the scores was on a scale of 1 to 5, later rescaled to scores from 0 to 4 (following Hassabis, . Spatial coherence index. This score was calculated from the responses participants made to the 12 statements on the postdescription questionnaires. Participants were required to tick as many of the statements as they thought applied to the imagined/remembered scene/episode they had just generated. Eight of the statements were "integrated" and suggested that the description was a continuous whole (e.g., "I could see it as one whole scene in my mind's eye"), and four indicated that the scene was more "fragmented" (e.g., "I could see individual details, but it didn't all fit together as a whole scene"). For each integrated statement that was selected, one point was awarded. For each fragmented statement that was selected, one point was subtracted. When totaled, these scores ranged between Ϫ4 and ϩ8. This score was then normalized, to give a spatial coherence index between Ϫ6 and ϩ6, with the coherence of the description increasing as the score increased. Independent quality ratings. Each description was rated out of 10 for its general quality, based on the extent to which it reflected a specific and detailed idea, and to which it reflected a vivid picture of the experience for the rater themselves. A score of 0 indicated that the description lacked any detail or vivid experience, and a score of 10 was assigned if the description was richly detailed and evoked a vivid sense of experiencing. These scores were then rescaled to a score of between 0 and 18, by multiplying by a factor of 1.8. Interrater reliability for the independent quality ratings was high, Cronbach's ␣ ϭ .94. The experiential index was calculated by adding up each of these subcomponent scores: description content (between 0 and 28) ϩ sense of presence (between 0 and 4) ϩ perceived salience (between 0 and 4) ϩ spatial coherence index (between 0 and 6) ϩ independent quality rating (between 0 and 18). The final experiential index score thus ranged between 0 (representation lacked detail and vivid experiencing) and 60 (richly detailed and experienced). In addition to coding the elements included in the Hassabis, content score, we also recorded the number of temporal terms (e.g., yesterday, tomorrow, times of year) used by participants in their descriptions. This allowed us to assess the extent to which participants really were engaging in mental time travel during the past and future event conditions of the task. If participants were engaging in episodic memory to remember experiences from the past in the past event condition, and engaging in episodic future thinking to imagine events that may occur in the future in the future event condition, but imagining atemporal scenes in the fictitious scene condition, then significantly fewer temporal references should be made in the fictitious scene condition than in either of the other conditions. Narrative Control Task The book Frog, Where Are You? The participant was shown the front cover of the book and asked to confirm that they had not encountered the book before. No participant was familiar with the book. They were told that it was a picture book and were instructed to look at the pictures and tell the story. Participants were informed that the experimenter had never seen the book before, and therefore they needed to be as clear as possible while telling the story. To eliminate memory demands, participants looked at each picture and turned the pages as they told the story aloud. The room set up was such that the experimenter could not see the pictures as the story was being told, 60 LIND, WILLIAMS, BOWLER, AND PEEL to reinforce the experimenter's lack of knowledge of the storyline. The experimenter did not interrupt the participant once they began their narrative and all narratives were audio recorded for later transcription and coding. Scoring. Each narrative was transcribed by an independent transcriber. The narrative scripts were coded by one rater who was blind to participant diagnosis. A second rater then coded a randomly selected subset (n ϭ 14; 25%) of the narrative scripts. The narratives were scored on three key dimensions: length of the narrative, global structure of the narrative, and number of relevant semantic details given. Length (unbounded score). The overall length of the narrative in words was calculated after the deletion from the transcript of repetitions and disfluencies, such as "ums" or "errs." Global structure (from 0 to 6). The global structure measure was included as an index of the participant's understanding of the causal structure of the story (following the procedure adopted by Reilly, Bates, & Marchman, 1998; see also Semantic score (from 0 to 102). This score provides an indicator of the amount of relevant detail included in participants' narratives. The scoring procedure developed by Norbury and Bishop (2003) was adopted, and participants were scored on how accurately and fully they included a list of 51 story elements in their narrative. Participants were given a score of 2 for every story element that they accurately included in their narrative, with a score of 1 given for an element they included inaccurately or only partially elaborated (for detailed guidance, and a list of the 51 story elements; see Theory of Mind Task The animations task requires participants to describe interactions between a large red triangle and a small blue triangle, as portrayed in a series of silent video clips. Eight clips (taken directly from Each clip was presented to participants on a computer screen. To familiarize participants with the task, two practice animations were shown before the experimental stimuli (one physical and one mentalizing). Participants were asked to describe the behavior displayed in each of these clips, and experimenter feedback was given after each description. For the experimental animations, participants were asked to "watch the clip and give me a running commentary about how the triangles are interacting." For the experimental trials, a digital audio recording of participants' responses was made for later transcription. No feedback was given on the experimental trials. The order in which the experimental clips were presented was counterbalanced across participants. Scoring. Each description was transcribed by an independent transcriber. Participants' descriptions were scored on the basis of scoring criteria outlined in Abell et al. (2000; see their Appendix A for detailed scoring criteria). Participants' descriptions of each animation were given a score of 0, 1, or 2 according to their level of accuracy. Accuracy was defined as the extent to which the participant's description captured the intended meaning of the animation. Thus, the score achievable in each condition (mentalizing/physical) was between zero and eight. Each description was scored by an independent rater who was blind to group status. A second rater then coded each of the transcripts. Interrater reliability for scores across each of the eight animations was high, Cronbach's ␣ ϭ .99. Statistical Analyses Results were analyzed using the statistical software package SPSS (Version 19). A standard alpha level of .05 was used to determine statistical significance for all analyses. All reported significance values are for two-tailed tests. In the first instance, group differences on the two background tasks-the theory of mind animations and narrative control taskswere explored using univariate and multivariate ANOVAs, respectively. Next, the data from the main experimental (episodic memory, episodic future thinking, and scene construction) task were analyzed using a series of univariate ANOVAs and t tests. Where ANOVAs were used, we report the corresponding partial 2 values as a measure of effect size.