Visual mental imagery abilities in autism

Sensory atypicalities such as over-reactivity to auditory input or unusual interest in perception-based information are a common feature of autism (Ben-Sasson et al., 2019). However, these atypicalities as not limited to the “sensory” dimension as higher levels of perceptive functioning are also atypical in autism (Bertone et al., 2005; Mottron, 2019; Samson et al., 2011). The Enhanced Perceptual Functioning (EPF) theory suggests a higher role, autonomy and performance of perceptive abilities in autism (Mottron et al., 2006; Mottron & Gagnon, 2023; Samson et al., 2011). Perception in autism is argued to be more precise and less likely to be altered by prior knowledge. As visual mental imagery and perception activate the same neural networks and rely on the same content-dependent representations in visual areas (Kosslyn et al., 2006), the enhanced visual abilities described in autism could induce enhanced visual mental imagery abilities. Thus, mental imagery might be a key interface between a particular cognitive functioning and the sensory particularities in autism. However, this cognitive mechanism has seldom been studied so far. The objective of this study was therefore to evaluate in depth mental imagery abilities in autism with the hypothesis of superior aptitudes in autism.

Mental imagery is defined as the evocation of a representation and its associated sensory information in the absence of a direct external stimulus (Pearson et al., 2015). Mental imagery plays a central role in cognition as it allows one to remember past events, plan the future, represent oneself in space or even make decisions. These various uses of mental imagery are made possible by its four different stages: generation, maintenance, inspection and manipulation of mental images (Kosslyn et al., 2006; Pearson et al., 2013). A mental image is indeed “generated” as it is constructed “step by step” (Koenig et al., 1991). Once generated, a mental image is subject to rapid decay with an average duration of only 250 ms (Kosslyn, 1994). An active maintenance of the image in our attention window is then required. Subsequently, this image can then be inspected or manipulated in space. The “scanning” of mental images is one possible form of inspection activity (i.e., moving from one point to another in a mental image) (Finke & Pinker, 1983). Mental images then can also be modified/transformed in space (Pearson et al., 2013).

Visual mental imagery has anecdotally been indicated as an autistic strength (Grandin, 2009). According to the self-reports of some autistic adults, autism entails a particular “way of thinking” described as “thinking in pictures”. Autistic individuals report using mental visual representation, and hence mental imagery, more frequently than typical participants (Bled et al., 2021). Regarding an objective evaluation of mental imagery, among the four stages described, only the manipulation of mental images has been studied in autism, through mental rotation tasks. Several studies have shown superior performances of autistic compared to typically developed individuals on mental rotation tasks (Falter et al., 2008; Hamilton et al., 2009; Pearson et al., 2016; Soulières et al., 2011). However, other studies have failed to find a difference between autistic and typically developed participants in mental rotation performance (Beacher et al., 2012; Conson et al., 2013; Rhode et al., 2018; Silk et al., 2006). Moreover, a meta-analysis showed only a nonsignificant advantage for autistic individuals over typical individuals for other aspects of the mental rotation task (working memory, information processing or decision making), and none regarding the aspect of rotation in space per se (Muth et al., 2014).

Even if, from a behavioral perspective, the results concerning mental rotation in autism are not consistent, functional imaging studies have confirmed particularities in brain activation in autistic individuals during this type of task. Studies in functional magnetic resonance imaging (fMRI) showed decreased neuronal activity in regions associated with working memory and executive functions, but greater posterior brain activity (in visual areas), in autistic individuals compared to typical individuals during a mental rotation task (Hooven, 2004; Silk et al., 2006). Thus, autistic individuals engage visual perceptual areas more but involve their working memory less during this mental imagery task. These studies suggest differences in the strategies used for mental rotation in autistic versus typical individuals.

As for the generation of mental images, it has never been studied in autism. The Image generation task is a paradigm first developed by Podgorny and Shepard (1978) and modified by Kosslyn et al. (Kosslyn, 1988; Podgorny & Shepard, 1978). This task evidences that mental images are generated in parts and not as a whole (the first parts of the images are created faster than the last part). Variations of this task have been used in neuroimaging studies of mental imagery in the typical population (Kosslyn et al., 1993) or in studies of group differences with depressed or anxious patients (Morrison et al., 2011; Zarrinpar et al., 2006) but never in people with neurodevelopmental conditions. A possible manifestation of superior mental imagery abilities in autism could be a faster and/or more detailed generation of mental images with results approaching those obtained in the simple perception condition.

Tasks involving the maintenance of visual information in short-term memory (i.e., visual working memory tests) have previously been studied in autism and showed that visuospatial working memory is compromised in autism (Desaunay et al., 2020; Kercood et al., 2014; Wang et al., 2017). However, the maintenance of mental images, per se, has never been evaluated in this population. The Visual Pattern Test consists of a matrix, with some of its cells filled in black, which is presented for a short duration. Participants then have to reproduce the pattern previously seen. This test is used to measure the visual component of working memory (without spatial properties) (Della Sala et al., 1999), and is an indirect measure of the maintenance of mental images (Pearson et al., 2013). Image maintenance tasks have been used, for example, to study imagery processes in depression (Cocude et al., 1997).

The inspection of mental images has never been studied in autism either. The Image scanning test is based on a scanning paradigm first introduced by Finke and Pinker (1982) and later refined by Borst et al. (2006) (Borst et al., 2006; Finke & Pinker, 1982). In this test, a pattern of dots is presented on the screen; the pattern is then removed and is replaced by an arrow. Participants have to decide whether the arrow points to a location previously occupied by one of the dots. As the distance between the arrow and the target dot increases, the time to make the decision increases, suggesting that participants scan their mental image (Kosslyn, 1988). Again, a possible manifestation of superior mental imagery abilities in autism could be a more detailed, and hence more accurate, inspection of mental images. The exploration of mental images may also be atypical and more locally driven than in the typical population.

Lastly, a visual thinking style implying a high use of visual mental representation in daily life activities is reported in autism (Bled et al., 2021). Furthermore, significant mental imagery abilities can be observed in autistic individuals. However, only their ability to manipulate mental images has been investigated to date. In this paper, we systematically evaluated the four stages of mental imagery (generation, maintenance, inspection and manipulation) in autistic versus typical adults, to offer the first detailed characterization of this cognitive process in autism. This paper also aims to try and answer our hypothesis of significant mental imagery abilities in autism.

Main results for the Autistic group and the Typical group are presented in Table 1.

We evaluated autistic individuals' performances in the four stages of mental imagery with four experimental tasks, in order to have a more detailed characterization of this mental process in autism. Globally, we observed intact, sometimes superior, mental imagery performances in the autistic group. For the Image generation task (generation of mental images), we found the typical effect of the probe location in both groups, as well as similar performance for autistic and typical participants. For the Visual pattern test (maintenance of mental images), we found a greater visual span in autistic than in typical individuals. Concerning the Image scanning test (scanning of mental images), we found an absence of effect of the difficulty. Contrary to expectations, we did not find any group difference for the Mental rotation test (manipulation of mental images). Lastly, correlations between the four stages of mental imagery indicated that all the stages are linked together except for maintenance which correlates only with the manipulation of mental images. This may be due to the fact that the test used (the Visual pattern test) is an indirect measure of the maintenance of mental images and is also involved in other cognitive processes such as visual working memory. The correlation patterns were similar between typically developed and autistic participants. In the following, we will discuss the result of each stage separately.

In the Image generation task, participants responded faster in the Early location condition compared to the Late location condition. This confirms the findings of previous research, suggesting that the mental image of the letter is indeed generated segment by segment and not as a whole (Kosslyn, 1988). Participants with autism were equivalent to controls for both reaction times and correct responses in image generation. The patterns of generation of mental images appeared to be similar between autistic and typical individuals.

Results from the Visual pattern test evidenced a greater ability to maintain mental images in autistic individuals. The active maintenance of mental images requires the central executive component of visual working memory (Pearson et al., 2013). Working memory in autism could be impaired by challenged executive attentional resources known in autism but it can also be enhanced by domain-specific representations reinforced by more efficient visual perceptual encoding (Hamilton et al., 2018). Indeed, visuospatial working memory is compromised in autism (Kercood et al., 2014; Wang et al., 2017). However, in the majority of studies, spatial working memory is the primary focus of evaluation (Hamilton et al., 2018). Spatial working memory could be challenged in autistic individuals, while visual working memory could be more efficient with more detailed visual representation. In line with this hypothesis, one study evidenced a significant positive relationship between autistic traits and visual working memory performance in children (Hamilton et al., 2018). As the Visual Pattern Test is limited in spatio-sequential properties, it represents a relatively pure measure of the visual component of working memory (without spatial properties) (Della Sala et al., 1999), which may explain the better performance of autistic participants in this task.

Concerning the third mental imagery stage, with the Image scanning test, many studies using this test have reported a linear increase in response times with increasing scanning distances (Borst et al., 2006; Borst & Kosslyn, 2008; Pinker, 1984). Our results are in line with these previous findings, suggesting that participants indeed created a mental image of the pattern of dots and then scanned the distance between the arrows and the dots. Furthermore, the way mental representations are processed reflects the spatial structure of the representations used in the tasks. Regarding the inspection of mental images, autistic individuals were just as accurate and fast as typical participants. We did not find any effect of distance on their accuracy and response times. Their response times did not increase with the level of difficulty, unlike what was observed in typical participants. This can be explained by a more detailed mental imagery, allowing them to accomplish tasks of increasing difficulty with the same ease. Furthermore, Enhanced Perceptual Functioning (EPF) in autism entails an enhanced role, performance and autonomy of perception (Mottron et al., 2006; Mottron & Gagnon, 2023; Samson et al., 2011). EPF might consequently lead to a more “accurate” and detailed visual mental representation, but also one that is more independent from top down influences in autism.

Last, considering the manipulation of mental images, with the Mental rotation test, autistic individuals performed as well as typical participants. Although several studies have shown superior performances of autistic compared to typically developed individuals on mental rotation tasks (Falter et al., 2008; Hamilton et al., 2009; Pearson et al., 2016; Soulières et al., 2011), our study failed to replicate this result. Autistic individuals encounter difficulties in timed tasks, due to difficulties with initiation and psychomotor speed with a slow and accurate response style (Hill & Bird, 2006; Johnston et al., 2011). Unlike the previously cited studies, the version of the Mental rotation test used in our study involved a time limit (e.g., complete as many items as possible within 3 min). This may explain why we did not replicate those previous results. It is possible that autistic individuals are more accurate than typical individuals, but answer fewer items within the time limit.

In summary, our findings support the hypothesis of typical or superior mental imagery abilities among autistic individuals. In a broader context, these results are in line with previous studies that have shown common mechanisms between mental imagery and perception (Pearson et al., 2015). Indeed, a more accurate and less top-down influenced (i.e., context-dependent) perception has been demonstrated in autistic individuals with experimental paradigms using visual illusions, for instance (Mitchell et al., 2010). Visual illusions have long been used to explore higher-level perceptual functioning in autism. Autistic individuals are found to be less sensitive to the “McGurk effect,” as they are less influenced by incongruence between visual and auditory stimuli (Stevenson et al., 2014). They also exhibit peculiarities during the “flash-beep” illusion (Bao et al., 2017). Autistic individuals may integrate the inducing context of information to a lesser extent (Happé, 1996). Thus, mental representations (i.e., mental images) in autism, like perception, may be more precise and context-independent. A reduced top-down feedback may be one potential explanatory mechanism for more accurate perception and mental imagery in autism (Park et al., 2022).

The neurophysiological integrated model of the visual system discussed by Bullier (2001) states that the visual system integrates local analysis and global information by exchanging information between neurons in higher-order areas responsible for different attributes. This can be achieved by retroinjecting computations from neurons in higher-order areas through feedback connections to neurons in lower-order areas such as V1 and V2 (Bullier, 2001; Layton et al., 2014). V1 and V2, which serve as “detailed general-purpose representations,” can then function as active “blackboards” (Bullier, 2001). According to this integrated model, as autistic individuals exhibit reduced cortical feedback input from higher visual areas (Isler et al., 2010; Kessler et al., 2016), this reduced top-down feedback connectivity in autism and fewer connections between these visual areas may result in a greater dominance of local processes in lower visual areas (V1 and V2). As these areas are “topographic and retinotopic,” the visual projection on the “blackboard” may then be locally processed throughout the representation. Hence, neuronal atypicalities and reduced feedback from higher areas reported in autistic individuals, in the framework of this integrated model, would explain the reduced contextual modulation reported in autism (Park et al., 2022) and the enhanced perception and mental imagery in this population. This is supported by the alteration of large-scale connectome asymmetry in the sensory and default-mode regions in brain imaging studies (Wan et al., 2023; Yoo et al., 2024).

One limitation of our study is the small number of participants. There was also a gender bias with more female participants in the Autistic group than expected according to the sex ratio in autism (Rødgaard et al., 2022; Zeidan et al., 2022). In addition, due to the complexity of the protocol, all participants had an average, or above average, level of intellectual functioning. Therefore, our sample is not entirely representative of the autism spectrum. Besides, one possible explanation for the fact that we did not find group differences for some mental imagery tasks is the significant heterogeneity of cognitive profiles that is well known in autism (Nader et al., 2015). It is indeed possible that only a “subprofile” of autistic individuals develops increased mental imagery abilities. For example, autistic individuals who experienced delayed language acquisition may develop significant visual capacities and use mental imagery as an alternative mechanism to verbal thinking. It would therefore be interesting to study mental imagery abilities specifically in autistic individuals with delayed language acquisition to verify whether there is a link between language and visual capacities.

The results of this study, which is the first to explore the different stages of mental imagery in autism, indicate preserved to greater visual mental imagery abilities in this population. Maintenance of mental images is greater in autistic individuals. Particularities in the inspection of mental images in autism can be related to the atypical perceptual functioning present in this population, including the bias towards a more local processing of information and a lesser top-down effect. Future studies might focus on how these mental imagery abilities are related to the cognitive functioning of autistic individuals.

This work was funded by the French National Research Agency (ANR-19-CE28-0012).

The authors have no conflict of interest to declare.

The study was approved by the Research Ethics Committee of Toulouse University (file number 2019–139), by the Ethics Committee of Rivière-des-Prairies Hospital and is in accordance with the European General Data Protection Regulation (GDPR).

All participants read an information sheet with explanations concerning the study and indicated their consent before beginning.