Exploring the Development of Past and Future Episodic Memory in Adolescents with Autism Spectrum Disorder: A Preliminary Longitudinal Study

Adolescence is a critical period for individuals to establish their identity and solidify their interactions with others. However, for adolescents with Autism Spectrum Disorder (ASD), the development of identity and social boundaries is precarious. These challenges in the development of identity and social ties may be related to difficulties in autobiographical memory (AM), whether recalling past events (past episodic memory; past EM) or imagining future scenarios (future episodic thinking; EFT). To date, little is known about the developmental patterns of AM in individuals with ASD over time. Eleven adolescents with ASD or typical development (TD) completed annual assessments of past EM and EFT over three years. Preliminary results indicate that past EM has become more detailed over the years for adolescents with ASD, while TD adolescents showed no changes. Interestingly, only the content elements of the narrative events increased, not the contextual elements. Moreover, EFT continuously developed in the TD group but remained stable in the ASD group. This first multi-case longitudinal study on AM requires replication with more subjects, but it seems to indicate a heterogeneous evolution of AM in ASD. These results will guide future research to explore the hypotheses of developmental delay and the factors influencing AM development in ASD. Ultimately, understanding these developmental pathways highlights the importance of personalized therapeutic approaches to support the social integration, identity construction, and future planning of adolescents with ASD.

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by deficits in social interaction and restricted repetitive behaviors. Additionally, executive function deficits and memory challenges closely associated with these core features are found in ASD. Importantly, complex memory processes, including autobiographical memory (AM), have been described as altered in ASD.

AM refers to the memory of an individual’s life. It combines specific events of personal experience (accompanied by perceptual and sensory experiences), corresponding to episodic memory, and personal knowledge, referring to semantic components. In our daily lives, AM manifests as the consistency of our identity over time (identity function), our ability to use past experiences to guide our choices and solve problems, enabling us to adapt to future events (directive function), and our capacity to share and communicate with others (social function). AM is closely related to self-awareness, as it allows for the storage and retrieval of memories consistent with personal goals and purposes, enabling us to establish and maintain a sense of identity and continuity over time. Furthermore, AM allows us to maintain social connections with others by recalling and sharing experiences, memories, and life stories.

While “memory” is intuitively a concept related to the encoding and retrieval of past events (i.e., past episodic memory; past EM), AM also refers to the ability to imagine and plan for future events, a cognitive ability known as future episodic thinking (EFT). EFT is the ability to project oneself into the future to pre-experience events. According to the theory of “constructive episodic simulation,” imagining future events is based on recombining several details belonging to memory. These are then linked together through semantic knowledge, which serves as the basis and support for constructing these future simulations according to the “semantic scaffold hypothesis.” Following this line of thought, AM can be divided into two components of “past” and “future.” However, this distinction is not clear-cut. In fact, similar cognitive abilities have been found between tasks involving imagining possible future events and recalling past events. Moreover, similar subsets of brain regions activate when subjects engage in tasks that require recalling the past or imagining the future.

Investigating AM in ASD highlights difficulties in both past and future components. Previous studies on children and adults with ASD reported impairments in past EM, including less specificity, detail, and coherence in retrieval memories. Regarding EFT, studies have shown that ASD individuals perform worse compared to typically developing (TD) subjects, both children and adults.

Overall, individuals with ASD demonstrate poorer past EM and EFT compared to TD subjects across different age groups. However, there is little research on how these abilities develop over time in ASD. In TD, AM appears to develop gradually from childhood through adolescence to adulthood. According to a review by Reese and colleagues, AM emerges around the age of 3-4, at which point children can provide comprehensible memories. Although 8-year-olds may not provide temporal links between different past events, their memories reach a comprehensible level. By age 12, the evoked memories continue to develop, with more precise spatiotemporal contexts and temporal connections. Additionally, experiences start to interconnect in memory, forming a more cohesive life narrative. The personal significance of each memory, as well as the connection between AM and self-construction, is not fully established until around age 16 and continues to develop into late adolescence. As for EFT, although this function appears around the age of 3-4, its developmental extent seems far behind the episodic memory inherited from childhood. Only from age 14 do adolescents develop the ability to make detailed and complete predictions about the future.

A developmental study involving two different age groups of ASD indicated that scores for incidental past EM in late adolescence (ages 12-18) were better than those in early adolescence (ages 8-12). To date, the only longitudinal exploration of AM in ASD has been a single-case study conducted on children, which pointed out limited development of past EM over time, with personal knowledge gradually forgotten. To our knowledge, no studies have investigated the development of EFT in ASD.

Cross-sectional studies provide snapshots of AM abilities across different age groups but fail to capture the dynamic changes that occur as individuals age and develop. On the other hand, longitudinal studies provide valuable insights into how past EM and EFT develop and interact during critical developmental periods. This distinction is crucial in ASD, as the developmental patterns in ASD may differ from those observed in the TD population. Therefore, it seems necessary to deepen this work by considering the crucial role of AM in constructing personal identity, social life, and problem-solving.

To fill this gap, we wanted to know how the AM of adolescents with ASD evolves compared to their typically developing peers. In fact, our study investigated the longitudinal development of past EM and EFT in adolescents with autism compared to a non-autistic control group. By conducting a three-year follow-up study with a group of adolescents, assessing their AM abilities annually, we aimed to identify specific developmental patterns. Based on the literature, we hypothesized that AM in TD would evolve during adolescence. Regarding ASD, the first hypothesis is that AM difficulties remain stable during adolescence. The second hypothesis is that the development of AM skills during adolescence is different or slower compared to TD.

2.1 Participants

Twenty-six children began this study (13 ASD and 13 TD), but some only completed one or two years of the study; others did not want to continue, moved away, or could no longer be contacted. We chose to retain only those adolescents who participated in all three studies. Thus, we included six boys with autism but without intellectual disabilities and five boys with intellectual disabilities, all aged between 12 and 18. The control group was matched with the ASD group in age and intelligence (Table 1). Intelligence quotient was assessed using the Wechsler scales. Participants had already been diagnosed with ASD at the time of inclusion in the study. The diagnosis was based on DSM-5 criteria, using the Autism Diagnostic Interview-Revised and/or the Autism Diagnostic Observation Schedule-Generic (Table 2). For all participants, exclusion criteria included a history of attention-deficit/hyperactivity disorder with or without hyperactivity, schizophrenia, head trauma leading to loss of consciousness, recent or frequent alcohol or drug abuse, chronic neurological or endocrine diseases, and medications that might interfere with memory assessments, as well as intellectual disabilities assessed through WISC IV or WAIS-IV. All participants and their parents signed consent forms. The study followed international and local ethical procedures.

2.2 Procedure

Adolescents completed three assessments (T0, T1, and T2), each separated by one year. We measured AM using the “From Past to Future Task” developed by Anger. Participants were asked to freely recall two events that happened to them in the past (the day before, i.e., recent, and the last summer vacation, i.e., remote) and imagine two events that would happen to them in the future (tomorrow, i.e., recent, and the next summer vacation, i.e., remote). The past events had to be real memories that occurred to them (e.g., “Do you remember what happened to you yesterday? I want you to recall it with a lot of details as if you were reliving this event, and your description should allow me to imagine this event”), while the future events had to be imagined (e.g., “Can you imagine what you might do tomorrow? It can be something planned or something completely new, but I want you to imagine it with a lot of details as if you were experiencing it, and your description must allow me to visualize this event”). There was no time or word limit; participants had to provide as many details as possible. After one minute of no response, the interviewer asked the children an open-ended question (e.g., “What else do you remember?”). After another minute of no response, the interviewer would ask more specific questions about the events (these questions were not scored in this study).

Each event was scored out of 5 points, with the following elements each worth 1 point: “When,” “Where,” “What,” “Who,” and “How.” We summed the scores of the two past events to obtain the past EM score (maximum score of 10), and similarly summed the scores of the two future events to obtain the EFT score. The higher the score, the more detailed the event. We then separately explored the contextual elements (“Where” and “When” details) and content elements (“What,” “Who,” and “How”). In case of disagreements, they were resolved through discussion. The Cohen’s Kappa coefficient between the two raters was 0.79, with a confidence interval of 0.66-0.93, representing a moderate to high level of agreement.

2.3 Statistical Analysis

Non-parametric statistical analyses were performed using R software. We conducted separate group comparisons (TD vs. ASD) at each time point using the Mann-Whitney test. We used Friedman’s variance analysis for within-group comparisons to evaluate the evolution at the three assessment time points: T0, T1, and T2. We then repeated the same analysis by separating the content items (“What,” “Who,” and “How”) and contextual items (“When” and “Where”). Post-hoc comparisons used the Wilcoxon paired sample test (differences in past AM and EFT scores between T0-T1, T1-T2, and T0-T2 for the ASD group and TD group) and Bonferroni correction. Effect sizes were calculated using Rosenthal’s r for Wilcoxon Mann Whitney tests; for Friedman variance analysis, Kendall’s W was used. Effect size interpretations are as follows: 0-0.3: small effect, 0.3-0.5: medium effect, 0.5 and above: large effect.

Regarding group comparisons, there were no significant differences between past EM and EFT at T0 (past EM: U=21, p= 0.115, r= 0.20; EFT: U= 39, p=1, r= 0.16), T1 (past EM: U= 33, p= 0.645, r= 0; EFT: U= 36.5, p=0.890, r=0.27), and T2 (past EM: U= 10.5, p= 0.460, r= 0.21; EFT: U= 5.5, p= 0.093, r= 0.49).

Within-group comparisons for T0, T1, and T2 showed that the past EM scores for the ASD group significantly increased (χ2 = 6.86; p< 0.05; W = 0.57) (Figure 1); the TD group showed no significant changes (χ2 = 4.13; p> 0.05; W = 0.41). Post-hoc analyses did not reveal any significant differences. The evolution of the past component in the ASD group was characterized by a significant increase in content (χ2 = 7.6, p<0.05; W= 0.63) sub-scores, while the contextual sub-scores showed no significant increase (χ2(2, n=6)= 0.67, p=0.716; W= 0.23); the TD group showed no significant changes (Table 3).

Regarding the group comparisons of EFT scores at T0, T1, and T2, no significant changes were observed in the ASD group over the three assessment times (χ2 = 1.2; p>0.05; W= 0.10; see Figure 2), while the TD group showed a significant increase in EFT scores (χ2 = 6.50; p<0.05; W= 0.65; see Figure 2). Post-hoc analyses did not reveal any significant differences. Regarding EFT sub-scores, there were no significant changes in the ASD group; in the TD group, both content (χ2 = 6.78; p<0.05; W= 0.68) and contextual (χ2 = 6.13; p<0.05; W= 0.61) EFT sub-scores significantly increased over time (Table 3).

This first longitudinal group study showed that adolescents with autism had more detailed past EM over the past few years, while EFT showed no changes. In the TD group, the results highlighted an increase in EFT performance, while past EM showed no changes. An exploration of event details revealed that the evolution of past EM in the ASD group was accompanied only by an increase in content elements, with no evolution related to context. In contrast, the evolution of EFT in TD adolescents was associated with increases in both content and context elements.

In this three-year longitudinal study, both groups exhibited different AM developmental patterns during adolescence. In the ASD group, we found that past EM performance improved over these three years. The evolution of past memory development in ASD aligns with the findings of Goddard’s cross-sectional study and Bon’s case study, both of which described an improvement in the ability of ASD individuals to recall past memories during adolescence. According to Atance and O’Neill, AM emerges around the ages of 3-4 in TD individuals and continues to develop into early adolescence. Thus, past memories in ASD individuals appear to begin developing during adolescence and continue into adulthood. In fact, adults with ASD show lower past EM skills compared to adults with TD. Given that the average age in these studies was under 40, this raises the question of whether past EM will continue to develop to match TD at some age or whether it will slow down or stop at some point, remaining below TD performance. More research is needed to confirm this hypothesis.

Regarding the future component of AM, while the EFT of TD subjects significantly improved over three years, the ASD group showed no significant changes. This change in TD is consistent with the results of Gott and Lah, which emphasize that TD adolescents have more developed EFT than children. The static nature of EFT throughout adolescence in ASD adolescents seems to indicate that their developmental characteristics are atypical compared to TD peers. However, since there was no significant difference between the two groups, we cannot further explain this result. Nevertheless, this opens a field for future research to explore the hypothesis of delayed AM development.

In this regard, previous studies have suggested that the development of EFT may occur later. Lind and Bowler proposed that more cognitive mobilization is needed to create future predictions than to recall past memories, regardless of the population (ASD or TD adults). Many studies have shown that TD individuals recall past events more accurately, specifically, and in detail than future events. Behavioral data accompanying these tasks indicate increased recruitment of the hippocampus during EFT tasks compared to past EM tasks. This increase may be associated with the recruitment of complex cognitive mechanisms involved in imagining new events. In fact, according to the theory of “constructive episodic simulation,” EFT requires a highly flexible system that can not only select and extract elements from past events but also manipulate and recombine these elements to form a coherent scene in the future.

The ASD group exhibits combined issues related to central coherence, which may lead to difficulties in linking past details to constructing future predictions. Additionally, the executive function deficits exhibited by ASD may affect the mobilization of complex memory processes, such as EFT. Goddard reported that AM difficulties in ASD are related to executive function deficits. Recently, a meta-analysis also highlighted the barriers faced by ASD patients of different ages (adults and children) in “mental time travel,” including projecting oneself into the past and/or future. Other studies have also reported poorer EFT skills in ASD children, noting that this is related to narrative abilities in ASD children, spatial navigation difficulties in ASD adults, and difficulties in scene construction, all of which may complicate the development of EFT in adolescents with autism. This is a promising avenue for future research.

Another interesting finding of this study is that improvements in past EM in the ASD group were primarily concentrated in content items (“What,” “Who,” and “How”), with no improvements in contextual elements (“Where” and “When”). In the control group, improvements in EFT were achieved through content and contextual details. Previous research has described difficulties faced by ASD patients when recalling past and future narratives. The observed lack of spatial elements in the development of AM may relate to spatial navigation difficulties in ASD adults, which are associated with impaired scene construction and self-projection in ASD, especially for future projections. Scene construction refers to the psychological process of generating and maintaining multimodal spatial representations. This process involves the hippocampus using relational memory to combine these details, which is impaired in ASD. In a recent review, Agron noted that, in addition to the widely explored social deficits in ASD, difficulties in scene construction and spatial navigation are part of the reasons for AM impairment and may represent a new pathway for understanding ASD. Time perception has been studied less in ASD but seems to be affected as well. Boucher proposed the existence of atypical “temporal thinking” (the ability to understand a fact or a group of facts over time), which may be related to the lack of coherence in autism.

The heterogeneous evolution of AM in adolescents with and without ASD leads us to hypothesize developmental delay. Our preliminary findings align with previous research, emphasizing that past EM continues to develop during adolescence in ASD, while this function is typically developed in TD adolescents and continues to evolve into more complex mechanisms, such as EFT. This viewpoint has been corroborated by some longitudinal studies on executive function. Ozonoff and McEvoy reported that the ASD group showed slight improvements in executive function during adolescence, but their performance did not catch up with TD participants over the three years of the study. Similarly, Skogli reported that adolescents with ASD face more difficulties in daily executive function compared to TD. After two years of longitudinal follow-up, these difficulties remained relatively stable, with no changes occurring in the ASD group. Furthermore, the hypothesis of developmental delay also appeared in a longitudinal study involving a group of ASD adolescents over a decade. This study indicated that over time, ASD adolescents improved in certain aspects of executive function, such as working memory abilities. However, they still exhibited difficulties relative to the TD group and did not catch up.

Considering between-group comparisons at each time point for the ASD and TD groups, we found no significant differences between the two groups in past AM and EFT. Although this is the first longitudinal group study, our sample size remains small, and insufficient study power may account for this lack of difference. Nevertheless, Anger conducted the same AM task study using a cross-sectional approach with a larger sample, showing that adolescents with ASD performed lower than the control group in both past EM and EFT. These results align with other findings indicating that ASD children generally perform worse than their TD peers in both past EM and EFT. However, another explanation may be that we combined recent and remote events for scoring in each past and future direction. Goddard found that children with ASD performed poorly on remote past EM but retained recent memories. Future research needs to explore the evolution of remote temporal perspectives.

Understanding the developmental trajectory of AM in ASD is crucial for designing targeted interventions that support the social integration, identity formation, and future planning capabilities of this population. Reflecting on past actions and decisions to adapt to future challenges can influence the self-regulation abilities of individuals with ASD. As Westby noted, this ability is critical for self-regulation, action, and decision-making in ASD. Furthermore, as previously mentioned, AM plays a key role in an individual’s social life. The connection between social difficulties in ASD and AM has been widely discussed in the literature. Insufficient or delayed learning of AM may also lead to social difficulties for this population.

This must encourage clinicians to better consider AM in interventions. In fact, in previous work, we established a psychological education program around AM for adolescents with ASD. We emphasized the beneficial impact of this training on social aspects. These results are encouraging and warrant continued efforts. Current research indicates that there are different patterns of AM development, highlighting the importance of cultivating this cognitive skill. The key period of adolescence/early adulthood is particularly significant, during which critical decisions arise, such as social-vocational orientation and life project selections, which pose challenges for individuals with ASD and their support individuals.

Due to the small sample size, the results presented in this paper need to be replicated in larger populations. In fact, this limits the power of our study to demonstrate eventual significant differences between our study groups (ASD vs. TD), thus constraining our interpretations. On the other hand, small samples may further strengthen within-group variability. To account for this variability, we chose graphical representations for each individual separately. Additionally, the limited number of participants also restricts the generalizability of our results. For instance, the comparisons between the two groups revealed a moderate effect size for age, indicating a moderate difference in age between the two groups, so interpretations of the results should be made cautiously. Nevertheless, based on the delayed development hypothesis, we found strong effect sizes in all significant analyses and weak or moderate effects in all non-significant analyses, which align with our results. Despite these limitations, the current longitudinal approach is novel and provides new dimensions for understanding the development of AM in ASD. Although this study is preliminary and cannot provide accurate conclusions about the developmental profile of AM in ASD, it does open interesting questions regarding the possibility of developmental delay. Given the role of AM in adolescent development, this also encourages more research on this topic.

This study aims to pave the way for a better understanding of the development of AM in adolescents with ASD. Our research provides the first longitudinal suggestions regarding the heterogeneous developmental trajectories of AM in ASD, emphasizing the need for further research with larger sample sizes to verify and expand these findings. We consider this a crucial scientific endeavor, given the close connection between AM in adolescents with ASD and their daily life challenges. Future research is needed to confirm this developmental pattern and explore the cognitive factors involved, including social impairments, executive function, central coherence, or mental time travel. Future directions should investigate the potential benefits of specific interventions to enhance past and future episodic memory. By addressing these areas, we can better support the social development of adolescents with ASD, ultimately helping them integrate into society and construct their identities.