Lee and Hobson (1998) assessed semantic autobiographical memory in a group of children with ASD with low verbal ability. Although they were found to produce qualitatively similar descriptions of their physical characteristics when compared to controls, differences appeared in their descriptions of psychological and social self‐
descriptive statements. Indeed, children with ASD produced significantly fewer psy- chological or social self‐statements. This study clearly highlights the question of the relation between psychological self and memory, and how one might influence the other. In fact, these findings fit with the SMS (Conway, 2005; Conway & Pleydell‐
Pearce, 2000) according to which autobiographical memory is structured hierarchi- cally at different levels of specificity. In this hierarchy, the first type of memory representation is the more general semantic autobiographical memory. The second type of autobiographical memory representation is specific episodic autobiographical memories. These are sensory perceptual memories for personally experienced events (Conway, 2005; Tulving, 1972). In the SMS, episodic and semantic autobiographical memories are thought to provide the content for various self‐images. In line with this idea, it has been found that autobiographical memories, which are cued by self‐images, tend to cluster temporally around the time of emergence for that particular self‐image,
a phenomenon known as the reminiscence bump (Rathbone, Moulin, & Conway, 2008). For example, people have a tendency to recall more personnel events from their adolescence or adulthood. In the context of our research, we decided to explore further the relation between self‐images, autobiographical memory, and future think- ing in autism. Indeed, the SMS holds that normal autobiographical memory function is crucial for a well‐functioning self‐concept (Conway, 2005). As such, impairments in autobiographical memory like those observed in ASD samples, might have detrimen- tal effects on self‐concept development, and indeed, research on the self and memory has provided indirect evidence for this idea.
For example, previous studies have revealed reduced self‐reference effect (Millward et al., 2000; Toichi et al., 2002) and a lack of the enactment effect in autism (i.e., a facilitation effect in retrieving self‐performed actions; Zalla et al., 2010). This sug- gests that memory in individuals with ASD does not improve when information is processed in relation to the self. To assess more directly the links between autobio- graphical memory and the self, we tested individuals with ASD on the Twenty Statements Test (TST) (Rathbone et al., 2008). Participants were asked to generate I AM statements (e.g., I AM a mother) and then to recall episodic autobiographical memories to some of those statements. We also decided to test adolescents with ASD (unpublished data) as self‐identity formation has been suggested as the main develop- mental task of adolescence (Erikson, 1968).
A total of 14 adolescents with ASD (male = 13, female = 1; mean age = 15.1, SD = 2.41; age range = 12–19; full scale IQ = 113.00, SD = 13.76) and 12 control par- ticipants (male = 10, female = 2; mean age = 14.9, SD = 2.97, age range = 11–19; full scale IQ = 115.08, SD = 10.11) took part in the study. There were no significant dif- ferences: gender [t(24) = 0.74, p = .469], age [t(24) = 0.21, p = .832], or full scale IQ [t(24) = 0.43, p = .669] indicating close matching. Inclusion criteria for ASD partici- pation was a clinician confirmed diagnosis of either Asperger syndrome (AS) or high‐
functioning autism (HFA) with no language impairment and a full scale IQ of >70 as assessed by Weschler abbreviated intelligence scale (WASI; Weschler, 1999). Ethical approval was obtained from Leeds University Institute of Psychological Sciences Ethics Committee (ethics reference code 12‐0067). Prior to their participation, all participants gave their informed consent. In addition, parental consent was obtained if the participant was below 18 years of age.
To measure the current self‐concept a modified version of the TST (Kuhn &
McPartland, 1954) was used. Participants were instructed to generate 20 self‐
statements (under no time pressure) until 20 statements had been produced or until they knew with certainty that no more self‐statements could be generated. The task employed to measure participant’s future self can be thought of as an extension of the TST, which allows the assessment of a person’s future self‐concept. The procedure and format of the future self‐task were identical to those of the current self‐task with the modification that this task required participants to provide statements describing their future self‐beginning with “I will be.” As was the case in the current self‐task, participants were provided with possible categories (social roles, personality traits, physical traits, or interests) of responses and examples of these categories (e.g., “I will be a mother for a future social role”). All participants were informed that the future self‐statements provided were to be plausible and realistic.
Following the TST task, participants were asked to select the three self‐statements they felt best described their current self‐concept from the ones they had generated in
the current self‐task. Each of these three statements was then used to cue three epi- sodic autobiographical memories of related events. Participants were instructed that the memory should be of a personally experienced event that occurred over one year ago and reflected an event where the particular self‐image was present (e.g., if the I am statement was musician, then the memories needed to be of an event relating to being a musician). Participants described their memories verbally and were asked to provide as much details as they could possibly recall. Following the I Will Be task, participants were asked to select from their future self‐statements the three future self‐images they felt would best come to describe their future self. Participants were then asked to imag- ine three specific future events linked to each of these three future self‐statements and to provide as much details as possible about the imagined future event.
Self‐strength was first measured as the mean total number of current and future self‐statements generated per group (max = 20). A 2 (group; ASD/control) × 2 (con- dition; current/future) mixed ANOVA generated a significant main effect of condi- tion, F (1, 23) = 4.83 p = .03, such that both groups generated more current (M = 17.8, SD = 4.35) than future self‐statements (M = 16.48, SD = 4.79). The main effect of group was not significant, F (1, 23) = 1.31, p = .26. The Group × Condition interac- tion was also non‐significant, F (1, 23) = 0.27, p = .61.
Moreover, a measure of self‐quality was created using both current and future self‐
concepts statements. These statements were coded using an extensive self‐statement coding scheme developed for use with the TST by Rhee, Uleman, Lee, and Roman (1995). Following this method, each statement was categorized as either autonomous (denoting self‐statements that were “stable over time and context and that repre- sented internalized feelings, beliefs, or abilities”) or social (denoting self‐statements that include reference to others, specific locations, or social context). Independently of this, each statement was also coded as either specific (denoting self‐statements, which were qualified by specific details) or abstract (denoting self‐statements lacking in specific details). This permitted a measure of self‐quality.
To assess group differences in social versus autonomous current self‐quality, an independent samples t‐test was carried out on social responses. This revealed a signifi- cant effect showing that the controls generated a higher proportion of social state- ments (M = .36, SD = .13) than the ASD group (M = .26, SD = .12), t (23) = 2.02, p = .05. An independent samples t‐test was also carried out in order to assess group differences in the proportion of abstract versus specific statements. This showed that controls generated significantly more abstract responses (M = .33, SD = .13) than did the ASD‐group (M = .23, SD = .12), t (23) = 2.06, p = .05. For future self‐statements, an independent samples t‐test revealed no significant group differences in the propor- tion of social responses provided (Control group M = .37, SD = .09, ASD‐group M = .38, SD = .15), t (23) = 0.164, p = .871. The independent samples t‐test on the proportion of abstract responses did however reveal that controls generated a signifi- cantly higher proportion of abstract self‐statements (M = .17, SD = .07) than the ASD group (M = .08, SD = .09), t (23) = 2.73, p = .012.
The findings in both current and future self‐quality motivated a more in depth analysis on self‐quality. As such, a set of independent samples t‐tests were carried out on the proportion of statements that fell into four of the categories (Rhee et al., 1995). The analysis of current self‐statements revealed a number of trends showing that the ASD group produced more physical descriptive statements, t (23) = 1.84, p = .07, and more specific attributes, t (23) = 1.87, p = .07, but fewer trait‐based
statements, t (23) = –1.82, p = .08. For future self‐statements, the independent sam- ples t‐tests revealed that controls generated significantly more trait based state- ments, t (23) = –3.08, p = .005, and significantly less specific attribute statements, t (23) = 2.18, p = .04.
Data from the autobiographical memory and future experience tasks were coded according to the methodology employed by Lind and Bowler (2010). In accordance with this coding scheme, a past or future event was classified as episodic if it met three criteria. First, it needed to be a specific personal event that lasted, or would last, less than one day and had occurred or would occur on only one occasion. Second, the event needed to be situated in time to within a two‐year range (such as 8 or 9 years old). Third, the event needed to be situated in space. This criterion was considered met if participants provided a specific location such as, “my classroom in school.”
Simply stating a town or a country was regarded insufficient. A future experience or autobiographical memory was coded as non‐episodic if the participant was unable to remember or imagine any event associated with the particular self‐statement used as a cue, if the event lasted or would last longer than a day, had occurred or would occur on repeated occasions, or was not located in time and space.
A second rater, blind to participant group, independently coded 50% of the auto- biographical memories and future experiences. Cohen’s Kappa indicated that inter‐
rater reliability was substantial at 0.96 for autobiographical memory and 0.93 for future experiences. Episodic scores for both autobiographical memories and future events were generated by assigning a score of 1 to any autobiographical memory or event classified as episodic and a 0 to events and autobiographical memories classified as non‐episodic. Thus, a proportion of episodic autobiographical memories and epi- sodic future events could be calculated by dividing the number of truly episodic auto- biographical memories and future events by the total autobiographical memories and future events provided. As such, the maximum episodic score for both was 1.00.
The mean episodic scores for autobiographical memories and future events were generated by calculating the mean proportion of strictly episodic autobiographical memories and future events per group (max = 1.00). A 2 (group; ASD/controls) × 2 (condition; past/future) mixed ANOVA revealed a significant main effect of group, Table 5.1 Mean proportion of current trait, social identity, specific attribute, and physical descriptive statements for each group (SDs in parentheses).
Trait Social identity Specific attributes Physical description
ASD 0.23(.15) 0.21(.15) 0.19(.18) 0.15(.07)
Control 0.34(.15) 0.26(.18) 0.08(.07) 0.10(.07)
Table 5.2 Mean proportion of future trait, social identity, specific attribute, and physical descriptive statements for each group (SD in parentheses).
Trait Social identity Specific attributes Physical description
ASD 0.07(.08) 0.28(.15) 0.15(.15) 0.21(.19)
Control 0.16(.06) 0.28(.11) 0.07(.06) 0.16(.10)
F (1, 23) = 10.04, p = 0.004) resulting from the greater proportion of episodic auto- biographical memories and future events generated by controls than the ASD‐group.
There was also a significant main effect of condition, F (1, 23) = 38.30, p < .001) dem- onstrating that both groups obtained significantly higher episodic scores in the past (M = .61, SD = .27) than future condition (M = .27, SD = .26). The Group × Condition interaction was however nonsignificant, F (1, 23) = 0.27, p = .612. See table 5.3 for Means and SDs.
The results of this study are in line with the growing body of research that has already demonstrated that episodic autobiographical memory impairments are a com- mon feature of the disorder (Bruck et al., 2007; Crane et al., 2009; Crane et al., 2011;
Goddard et al., 2007; Lind & Bowler, 2010; Tanweer et al., 2010;). The impaired ability to use current self‐images to cue episodic autobiographical memory was mir- rored in the future condition by a reduced ability to use future self‐images to cue episodic future events. This appears to confirm previous findings of impaired future thinking in ASD (Lind & Bowler, 2010). This result converges with previous research on ASD, which has shown that impairments in episodic autobiographical memory are associated with impairments in episodic future thinking (Lind & Bowler, 2010).
Finally, these results are consistent with the wider literature, which demonstrates that the two abilities are closely interlinked (D’Argembeau, Raffard, & Van der Linden, 2008; Spreng & Levine, 2006).
The finding of impaired episodic autobiographical memory in the present study does also extend the literature in an interesting way. In the bidirectional self‐memory relation, the self is thought of as a structure that controls the encoding and retrieval of autobiographical memories (Conway, 2005). Because the present study used cur- rent self‐images to cue episodic autobiographical memories, the finding of impaired episodic autobiographical memory might suggest that individuals with ASD have a specific deficit in using the self to organize retrieval from autobiographical memory.
That is, individuals with ASD might fail to use their self as an effective organizing structure in memory. This idea supports Crane and Goddard (2009), who suggested that the reduced ability of individuals with ASD to use self‐relevant cues to access episodic autobiographical memories indicated a deficit in using the self to organize memory retrieval.
However, the hypothesis of a reduced ability to use self‐relevant cues clearly con- trasts with our findings showing that adolescents (and adults, Tanweer et al., 2010) are able to produce as many self‐images as controls, and therefore, do not show a diminished self. In other words, despite the impairments in episodic autobiographical memory, adolescents with ASD appear to develop a strong and accessible self. As pre- vious studies have shown that individuals with ASD might have intact semantic auto- biographical memory (Crane & Godard, 2008; Tanwer et al., 2010), one potential explanation of such findings is that patients with impaired episodic autobiographical Table 5.3 Mean proportion of autobiographical memories and future images classified as episodic in nature (max = 1.0).
Episodic memories Future experiences
ASD group 0.48(.28) 0.17(.22)
Control group 0.75(.17) 0.38(.25)
memory simply rely on their intact conceptual semantic autobiographical memory to maintain or develop a strong sense of self (Conway et al., 2004; Rathbone, Moulin,
& Conway, 2009). One possibility is that in the same way that the current self might be supported by the intact semantic autobiographical memory in ASD, the future self‐concept can be supported by semantic future thoughts. In this way, the simple knowledge of events that may occur in the future could support retrieval of episodic future events (Atance & O’Neill, 2005). For example, it is conceivable that cultural life scripts (Berntsen & Rubin, 2004), which can be thought of as a schema of norma- tive life events and their usual timing (e.g., leaving school or having children) might support the future self‐concept in ASD.