Participants
Twenty-two autistic individuals without intellectual disability (ID) (two women and 20 men) were included in the present study. The participants were recruited at the Autism Resource Center of Colmar (France) and at the Psychiatric Department of Strasbourg’s University Hospital (France). All the individuals had obtained an established diagnosis of ASD in these centers, according to the DSM-5 criteria and the French recommendations for the ASD diagnosis in adults. During the diagnosis process, all individuals were assessed with ADI-R46 and ADOS module 447. Individuals with intellectual quotient (IQ) lower than 70 were not included. A comparison group of 22 nonautistic participants was matched to the autistic individuals with respect to age, gender, and level of schooling (see Table 1). None of the nonautistic participants reported a history of psychiatric diagnosis or medication use. No participants from either group had a history of neurological disorders or substance abuse. Specific data on race or ethnicity are not allowed to be collected in France and are thus not presented in the present study. The Institutional Review Board of Lille III approved this study, and all experiments were conducted in accordance with approved guidelines and regulations. All participants provided written informed consent.
Material
Clinical measures. The Autism Spectrum Quotient (AQ)48, validated in French49, is a typical and rapid screening instrument for ASD. All participants completed this self-administered questionnaire of 50 items divided into five ten-item subscales assessing social skills, attention shifting, attention to detail, communication, and imagination. High scores reflect high levels of autism characteristics. The Empathy Quotient (EQ)50, validated and translated in French49, is a short 40-item self-administered questionnaire used to measure empathy in all participants. High scores represent high levels of empathy.
Cognitive measures. Two cognitive tests were administered to assess executive functions involved in the elaboration of life narratives in all participants. Verbal fluency was assessed using a semantic (animals) and a phonological (French words starting with P) fluency task of 120 seconds each51. Mental flexibility was assessed using the Trail Making Test (Parts A and B) by subtracting the connecting time in Part A from that in Part B52. An estimation of the participant’s IQ was assessed using the French version of the national adult reading test53.
Biographical practices. Given that the frequency of one’s biographical practices can influence the construction of his or her life narrative, we assessed the frequency of various biographical activities (e.g., keeping a diary, looking at old pictures, and talking about problems with friends) in all participants21.
Sense of Coherence. Participants self-rated their subjective sense of coherence, a salutogenic index reflecting one’s resilience abilities28. The scale measures the three components of the sense of coherence: the comprehensibility of external events, their manageability, and their meaning. Higher scores reflect a higher sense of coherence.
Life narratives.
Procedure. Participants’ life stories were collected using the protocol developed by Habermas and de Silveira21. First, participants were asked to recall the seven most important events they had ever experienced and to write them down on cards. Second, participants were asked to narrate their life story in approximately 20 minutes. Instructions specifically invited participants to orally narrate a story of their whole life, from birth to the present time, integrating the seven most important events previously recalled into the narrative and explaining how the individuals had become the person they are today. The experimenter did not interrupt participants but encouraged them to pursue the narrative if they had time remaining. All life narratives were audio-recorded and transcribed verbatim. Afterward, the participants dated the seven most important events and rated their vividness, emotional features, and consequences on 7-point Likert scales 54. Participants also completed the short version of the Centrality of Event Scale (CES55), assessing the extent to which each event had become central to their identity.
Coding. All narratives were coded according to the coding schemes developed by Habermas and colleagues27,30.
Segmentation. First, narratives were divided into propositions corresponding to minimal meaningful sentences. Two independent coders segmented 25% (n = 11) of the life stories into propositions with an agreement of 92.6%. Each disagreement was resolved by discussion, and the remaining narratives were segmented by one of the two coders.
Beginnings and endings. The degree of elaboration of beginnings and endings was assessed using 3-category (for beginnings) and 4-category (for endings) coding schemes. These coding schemes assessed the time of narrative beginnings (i.e., at birth, after birth or unclear) and the time and elaboration of narrative endings (i.e., at the present time, at the present time with a retrospective, at the present time with future projections, at the present time with both retrospective and prospective thoughts, or an arbitrary ending). A life narrative starting at birth and ending in the present is considered more elaborate than a narrative starting implicitly in the middle of adolescence and ending with a random event29.
Local indicators of coherence. The presence of local indicators of coherence was quantified for each type of coherence, temporal, causal-motivational and thematic. Indicators of temporal coherence include landmarks (such as date, life period, age, distance from the present, and historical events) that allow the listener (or the reader) to locate events within the course of the story. Indicators of causal-motivational coherence correspond to self-event connections reflecting identity changes (e.g., “That journey changed many things for me; and since then I am a little more self-confident”) and other biographical arguments placing local experiences in the context of someone's entire life (e.g., “Then, for the first time in my life, I had to take on responsibility”). They indicate that the narrator intends to maintain self-continuity despite life disruptions or changes. Local indicators of thematic coherence include self-event connections reflecting the stability of identity (e.g., “I’ve always been a shy person and had problems making friends”).
The proportion of indicators was calculated by dividing the number of local indicators coded by the total number of propositions separately for each type of coherence and considered in statistical analyses27. Based on Habermas and colleagues’ manuals, all types of local indicators of coherence were coded by two independent coders for 23% of the life narratives, with good interrater reliability (temporal coherence, Κ =.83; biographical arguments, Κ =.73; self-event connections, Κ =.75).
Global coherence. At the global level, temporal, causal-motivational and thematic coherences were rated using three 7-point Likert scales based on the overall impression gained while reading the entire narrative. Higher scores reflect higher levels of global coherence. The temporal orientation scale assesses how well the life narrative enables the interviewer to determine when and in what order something occurred in the narrator’s life. The causal-motivational rating scale captures how clear the development of the personality becomes in terms of turning points and motives throughout the narratives. The thematic scale reflects the extent to which the individual elements of the life story are thematically connected.
Based on Habermas and colleagues’ manuals, all types of global coherence were rated by two independent coders for 23% of the life narratives, with good interrater reliability (temporal coherence, ICC = .91; causal coherence, Κ =.81; thematic coherence, Κ =.74).
Statistics
Statistical analyses were performed under a Bayesian framework56. In addition to avoiding reasoning based on the criticized p values, Bayesian statistics provide a distribution of the probability that the performance of participants in one group (ASD) differs from the other group (C). This should not be confounded with the null hypothesis testing of classical statistics, although p values are often misleadingly considered to reflect this probability.
For all analyses, the probability of the score of each measure being higher in the ASD group than in the comparison group [indicated as Pr(ASD > C)] was calculated. A probability higher than 97.5% (i.e., Pr(ASD > C) > 0.975) that scores of autistic individuals would be higher than those of nonautistic participants was considered meaningful. It is worth noting that Pr (ASD > C) > 0.975 is equivalent to a probability lower than 2.5% that scores of autistic individuals are lower than those of nonautistic participants (i.e., Pr (ASD < C) < 0.025). Therefore, Pr (ASD < C) < 0.025 was also considered meaningful. Interactions between factors were written as Pr (OR > 1), OR corresponding to the exponential of the interaction coefficient. All analyses were performed using noninformative priors.
A burn-in of 5,000 iterations followed by 100,000 iterations was used for each of the three chains, yielding a final 300,000-iteration sample for retrieving posterior distribution characteristics. The convergence of the MCMC sample chains was checked graphically and observed in each case. All computations were performed using R software (R, Development Core Team, 2018) with all the required additional packages (in particular, Rjags; Plummer et al., 2018).
No community members were involved in the present research.