One-hundred and fifty-one participants from the United Kingdom took part. We included only participants from the United Kingdom since diagnostic processes and barriers may differ by country (Huang et al., 2020). One hundred and seventeen participants identified as cisgender female (77.6%), with 30 cisgender male (19.7%) and four non-binary or transgender participants (2.7%). Participants’ age ranged from 18 to 65 years old, with a mean of 31.26 (SD = 10.23). Further participant characteristics are available in Table 1, indicating that most participants were White British, educated to degree level, and in employment.

Participants were required to have a formal autism diagnosis since the study considered the experience of receiving a diagnosis. Eighty-six participants (57.0%) self-reported a diagnosis of “Autism Spectrum Condition,” with the remaining 65 participants (43.0%) reporting “Asperger’s Syndrome.” As this information was self-reported, the Ritvo Autism and Asperger Diagnostic Scale (RAADS-14; Eriksson et al., 2013) was administered to validate the presence of diagnosable autistic characteristics. All participants scored above the RAADS-14 cut-off score of 14. Participants predominantly received their diagnosis through NHS services (96.0%), with few receiving their diagnosis privately (4.0%). Participants received their diagnoses between 2000 and 2020, reporting ages at diagnosis ranging from 6 to 62 years old (mean = 26.42, SD = 11.18, 78.8% diagnosed over the age of 18). On average, participants had received their diagnosis 4.95 years previously (SD = 4.09, range 0–20). Over half indicated they had additional neurodevelopmental or mental health diagnoses (58.3%).

Participants were recruited online using snowballing methods, with adverts posted on public and private social media (e.g., Facebook and Twitter) and sent directly to relevant autism groups, organisations, and charities between December 2019 and March 2020. Although we were particularly interested in the experiences of those diagnosed late, the survey was open to all. This study was reviewed and approved ethically by the Research Ethics Committee at Royal Holloway, University of London. All participants provided informed consent prior to participation.

We used “Qualtrics” as the online survey platform. Participants first completed questions concerning demographics and diagnoses. Due to variation in language preferences within the autistic community (Kenny et al., 2016), participants were given the opportunity to customise the survey to reflect their preferred terminology (“autistic person,” “person with autism,” or no preference). Accordingly, we presented participants with subsequent questions using either identity-first (64% preferred) or person-first (2%) language, or a combination if no preference (34%). Participants then completed the four measures in the order outlined and an optional qualitative question (see below).

This study used a cross-sectional mixed methods survey design. We selected mixed methods approaches to maintain a high level of empiricism while also acknowledging the exploratory nature of the study. Including qualitative elements in autism research has been highlighted as particularly important in understanding issues from the perspective of autistic people themselves (e.g., Bölte, 2014).

We analysed quantitative data using SPSS version 27. There were no missing data for any of the quantitative measures. We first calculated descriptive statistics and variables were tested for compliance with standard parametric assumptions. All z scores calculated for skewness and kurtosis were lower than 2.58 (p > 0.01), and therefore we considered the data to be normally distributed. We consider p values between 0.05 and 0.005 as suggestively significant and p < 0.005 as significant (Ioannidis, 2018).

For the first hypothesis (personal autistic identity, age of diagnosis, and recency of diagnosis), we conducted two separate multiple regressions with personal autistic identity (QDIO subscales) as the dependent variables, and age of diagnosis and recency of diagnosis as predictor variables (controlling for gender). These analyses aimed to show how diagnostic timing related to personal autistic identity, over and above any gender differences in diagnosis.

For the second hypothesis (autistic personal identity, self-esteem, and wellbeing), we used two hierarchical regression analyses. The first hierarchical regression had self-esteem (RSE scores) as the outcome variable. We entered control variables into the first step, specifically autistic characteristics (RAADS-14), wellbeing (WEMWBS), gender (female vs. male only, due to small n of non-binary/other genders), and recency and age of diagnosis. We analysed self-esteem and wellbeing as two separate outcome variables: although there is shared variance between these variables, these concepts can be viewed as distinct, with discriminant validity between the two (Robins et al., 2001; Lyubomirsky et al., 2005), and we aimed to understand how identity contributed to each one uniquely. We thus controlled for wellbeing/self-esteem in the regression analysis predicting the other to ensure that variance explained by these inter-related concepts was taken into account in the model. In step 2, we entered the personal autistic identity variables (the two QDIO subscales: autism pride and exclusion/dissatisfaction) as predictors.

Similarly, we carried out a second hierarchical regression with wellbeing (WEMWBS) as the dependent variable, with the same variables as above controlled for in step 1 (replacing wellbeing with self-esteem), and personal autistic identity (QDIO subscales) in step 2. Together, these analyses aimed to show how personal identity contributed to wellbeing/self-esteem, over and above gender, diagnostic timing, autistic characteristics, and shared variance between wellbeing and self-esteem. All assumptions were met for the regressions, including homoscedasticity and multi-collinearity.

Table 2 shows the means and SDs for the quantitative variables included in the study, and Table 3 includes correlations between the variables.

Results from the two multiple regressions, with the personal autistic identity subscales as outcomes and age at and recency of diagnosis as predictors (controlling for gender), are summarised in Table 4. With autism pride as the outcome, the model explained 2.2% of the variance and was not significant [F(3,143) = 1.07, p = 0.36], with no significant predictors. With exclusion/dissatisfaction as the outcome, the model explained 21.5% of the variance and was significant [F(3,143) = 12.07, p < 0.001]. Here, as number of years since diagnosis increased, exclusion/dissatisfaction decreased (Figure 1).

The first step, including wellbeing, autistic characteristics, gender, and recency and age of diagnosis explained 77.1% of the variance in self-esteem (Table 5), and the model was significant [F(5, 141) = 95.21, p < 0.001]. Adding the QDIO subscales into step 2 explained an additional 5.7% of the variance in self-esteem, which was a significant increase [F(2, 139) = 22.84, p < 0.001]. The final model was significant [F(7,139) = 95.61, p < 0.001] and explained 82.8% of the variance in self-esteem. Specifically, lower exclusion/dissatisfaction significantly predicted higher self-esteem, and greater autism pride predicted higher self-esteem (Figure 2). Wellbeing was also a significant predictor in the final model, such that higher wellbeing predicted higher self-esteem. No other variables were significant.

The first step, including self-esteem, gender, autistic characteristics, and recency and age of diagnosis, explained 75.9% of the variance in wellbeing scores (Table 6), and the model was significant [F(5, 141) = 89.01, p < 0.001]. Adding the QDIO subscales as predictors explained an additional 0.9% of the variance in wellbeing, which was not a significant increase [F(2, 139) = 2.84, p = 0.062]. The final model significantly explained 76.9% of the variance in wellbeing scores [F(7,139) = 66.05, p < 0.001], with greater exclusion/dissatisfaction predicting lower wellbeing at a suggestively significant threshold (Figure 3). As before, higher self-esteem also predicted higher wellbeing, and no other variables were significant in the final model.

Fifty-four participants (35.7% of the sample) responded to the optional qualitative question, describing how receiving an autism diagnosis impacted on their thoughts and feelings about themselves. Table 7 illustrates the identified categories and sub-categories, number of coded responses for each category, and example quotes.

The most reported category focused on an “adjustment process,” with responses generally reflecting changes experienced by the respondents following diagnosis. Within this category, we identified three sub-categories. First, participants discussed “Making sense of it all over time” whereby they referred to a process of re-evaluating their lives following their diagnosis. Further, participants talked about an “Emotional reaction to diagnosis,” describing various reactive emotional responses including both positive (such as relief and happiness) and negative feelings (such as confusion and being overwhelmed). Additionally, some participants discussed “Permanence of difficulties,” conveying a sense of grief or frustration over having to accept that their difficulties would not go away.

The next most reported category concerned “Self-exploration,” referring to the experience or process of self-discovery following diagnosis. There were three sub-categories: most often, participants discussed “Knowing and understanding who I am,” highlighting how they had an increased understanding of themselves after diagnosis, in addition to allowing themselves more self-compassion. Some participants discussed “Being myself,” describing feeling increasingly able to be more authentic in themselves. However, some participants discussed “Feeling the same about myself,” reflecting the sense that the diagnosis had not led to any change in their self-perceptions.

Within the category of “Learning and support needs,” participants reflected on a lack of information or knowledge about autism following diagnosis: Within the sub-category “I’ve learned a lot,” participants mentioned their own preconceptions about autism before and after receiving the diagnosis and discussed finding out more about autism by doing research following diagnosis. The sub-category “More support needed post-diagnosis” referred to a perceived lack of post-diagnostic support and services available. A few participants mentioned “Late identification,” questioning how it had taken until adulthood for others to recognise them as autistic. Relatedly, some participants specifically mentioned “Autism in women,” describing a specific lack of information available about autistic females.

The next category focused on “Responses from others,” referring to the reactions (both positive and negative) from other people to the respondent’s diagnosis. Here, within the sub-category “Lack of understanding and acceptance from others,” participants discussed difficult interactions they had with others on receiving their diagnosis, particularly in managing others’ poor autism knowledge and being disbelieved or not accepted by others. Some discussed the sub-category of “Issues with disclosure,” reporting concerns about disclosing their diagnosis to others. A few participants described “Positive reactions,” giving examples of others responding positively to the diagnosis, often in terms of increased understanding.

In the category “Autism as a positive difference,” responses included descriptions of positive aspects of receiving a diagnosis. Here, within the sub-category “Autistic pride and appreciation,” participants discussed the strengths associated with being autistic, in addition to experiencing a positive autistic identity. Some participants mentioned “Thinking differently/being different” where they discussed positive aspects of difference and described the diagnosis as replacing a previous sense of something being “wrong” with themselves. A few participants discussed the sub-category of “Self-advocacy,” reflecting on an increased ability to advocate for themselves and their needs following diagnosis.

Less often, we coded responses into the category “Challenges of the diagnosis,” where participants reflected on negative aspects of receiving a diagnosis. Within the sub-category “Struggling to come to terms and find my place,” participants reflected on feelings of low self-confidence and not relating to their pre-existing perceptions of autism. We coded a few responses as “Autism as a negative,” where participants described a sense of negative difference or focused on difficulties associated with being autistic.

The final category, “I now feel part of something with others like me,” reflected a sense of connection with the autistic community, how receiving an autism diagnosis helped participants to feel less lonely, and provided a sense of hope that there were other people who have similar experiences.

This study particularly highlights the experiences of autistic people who were diagnosed late, and demonstrates relationships between autistic personal identity, self-esteem, wellbeing, and diagnostic timing. Self-esteem and wellbeing can both be understood as closely related to broader psychological health and functioning (Lyubomirsky et al., 2005). Although the correlational nature of the current study design inhibits the ability to determine causation, our results suggest that exploring potential identity-based support for psychological health should be a priority for future research. In the disability literature, it has been recommended that disabled people should be encouraged to engage with their personal disability narrative to aid the development of coping strategies and positive identity development (Dunn and Burcaw, 2013). For autistic people, this could involve a programme supporting newly diagnosed autistic people to think about strengths, challenges, and fostering connections with the wider autistic community. A recent study has highlighted the specific benefits of such group programmes being autistic-led in terms of developing a positive outlook on being autistic (Crane et al., 2020), and research highlights the benefits of autistic-autistic peer communication (Crompton et al., 2020a,b).

Further, Forber-Pratt et al. (2019) suggest that professionals are often the most significant or accessible resource for many following diagnosis, and therefore have a role to play in influencing people’s relationship with their autistic identity. However, some participants in the present study noted that lack of post-diagnostic professional support was an issue, which has been reported previously – with 42% of respondents in one survey not offered any type of post-diagnostic support (Jones et al., 2014). Although it is reassuring that many of the participants in this study connected with other autistic people online for information and support, the current findings suggest there is a need for services and clinicians to provide higher quality post-diagnostic support.

There were several limitations with our sample: only four participants identified as either non-binary or transgender, which was too few to include in statistical analyses. This limitation restricted the quantitative analyses to cisgender participants and is particularly relevant given increased gender identity variance in autistic individuals (De Vries et al., 2010; Pasterski et al., 2014). Further, participants were predominantly female (77.6%), suggesting a potential issue in terms of overall representativeness of the female experience. Since there has been a systematic under-identification of autism in females, with females at higher risk of being misdiagnosed or diagnosed late (Mandy et al., 2012; Kreiser and White, 2014; Trubanova et al., 2014), partly due to the male-biased development of assessment measures and diagnostic criteria (Kreiser and White, 2014; Loomes et al., 2017), the study may have been of particular interest to females. Given the challenges faced by autistic females in terms of unmet support needs, social exclusion, and isolation (Baldwin and Costley, 2016), our study indicates that understanding and supporting autistic females to explore their autistic identity may be beneficial. Nonetheless, future research should aim to establish whether the current findings are replicable in a larger sample of males.

The sample was also predominantly white, and university-educated. However, the demographic figures reported in the present study are broadly similar to comparable survey research (e.g., Cooper et al., 2017; Cage and Troxell-Whitman, 2019). Lack of diversity is a frequent criticism of autism literature (Pellicano et al., 2014). However, the sampling of autistic adults and particularly females builds upon a previous lack of representation in autism research (Kirkovski et al., 2013; Pellicano et al., 2014). Further, this study utilised convenience sampling through social media and groups, which may have led to biases in the sample. For example, in a study of autistic adults’ participation in research, factors such as altruism, a sense of community, and a keenness to be listened to and understood were found to be particularly motivating (Haas et al., 2016). Additionally, our sample is biased in that it consisted of mostly well-educated individuals recruited via the internet, and this sample will not be representative. Our findings represent only a subset of autistic people, and although there was variation in autistic identification, autistic pride was generally high. Future research should attempt to recruit participants from a wider variety of online and offline sources and find ways to capture the views of the autistic community more broadly. Finally, the measure of personal autistic identity was adapted from the disability literature (Darling and Heckert, 2010), and may not have fully captured a personal autistic identity. The development of measures specifically to capture the incorporation of “autism” into personal identity would thus be beneficial. The constructs of wellbeing and self-esteem were also based on measures that have been created by and validated with non-autistic people.

The present study explored the relationships between autistic personal identity, diagnostic timing, and psychological health, with a focus on late-diagnosed autistic people. With more time, since diagnosis there was less dissatisfaction with being autistic and autism pride and exclusion/dissatisfaction significantly predicted self-esteem, and exclusion/dissatisfaction suggestively predicted wellbeing. Qualitative descriptions of diagnosis experiences described a post-diagnostic process that included emotional reactions and self-exploration, which developed into self-acceptance and belonging. Our results add to the literature concerning the experiences of late-diagnosed autistic adults, with implications regarding the need for more frequent and comprehensive provision of information and post-diagnostic support and finding ways to enable all autistic people to explore their autistic personal identity.