Autism is characterized by persistent difficulties in social communication and social interaction across multiple contexts, including social reciprocity, nonverbal communicative behaviors used for social interaction, and skills in developing, maintaining, and understanding relationships (American Psychiatric Association, DSM-5 Task Force, 2013). Motor coordination difficulties, although not highlighted within this description of autism, are a common feature and are increasingly becoming the subject of investigation (Gowen and Hamilton, 2013; Torres and Donnellan, 2015; Bhat, 2021). Motor coordination difficulties experienced by autistic individuals are apparent from childhood and persist into adulthood and may include altered fine motor control and eye-hand coordination, as well as postural instability and general clumsiness (Fournier et al., 2010; Gowen and Hamilton, 2013; Sacrey et al., 2014; Lim et al., 2017; Morrison et al., 2018; Lum et al., 2020; Gowen et al., 2023). Interest has grown in understanding how motor coordination difficulties might impact upon the social aspects of autism. It has been suggested that exclusion, lack of confidence and fatigue, due to increased effort, could all reduce participation in social activities such as sports, social hobbies or play, leading to less experience in engaging socially (Leonard et al., 2014; Hannant et al., 2016; Ohara et al., 2019; Gowen et al., 2023). Another link between motor and social difficulties is simulation. Simulation theory suggests that when an individual observes another person, they will internally simulate that person’s actions using their own motor, cognitive and emotional representations, in order to understand the other person’s actions (Jeannerod, 2001; Gallese, 2003). Motor simulation is thought to underlie a number of processes such as imitation, action prediction and understanding, as well as motor imagery (Wolpert et al., 2003; Rizzolatti and Sinigaglia, 2010; Gallese and Sinigaglia, 2011; Hurst and Boe, 2022). While some previous work has highlighted altered imitation (Wild et al., 2012; Sacrey et al., 2014; Vivanti and Hamilton, 2014; Gowen et al., 2020), action understanding (Vivanti et al., 2011; Todorova et al., 2019; Federici et al., 2020) and action prediction (Gowen et al., 2022) in autistic individuals, there has been less focus on motor imagery (MI). The aim of this systematic review was to assess the current knowledge around MI ability in autistic individuals.

MI is when an individual imagines performing an action without physically executing that action and is thought to involve similar, although not identical processes to those that program and prepare for the execution of physical movement (Jeannerod, 1994; Hardwick et al., 2018; Hurst and Boe, 2022; Frank et al., 2023). Jeannerod and Frak (1999) argue there are both implicit and explicit forms of MI where actions are imagined at the unconscious or conscious level, respectively. MI can also be performed in the visual or kinesthetic modality. Visual MI involves the visualization of an action and activates visual occipital brain areas, whereas kinesthetic MI consists of imagining the sensation of the action and activates motor-related brain areas such as the superior parietal and ventral premotor cortex (Jeannerod, 2001; Sirigu and Duhamel, 2001; Stinear et al., 2006; Guillot et al., 2009). MI can play a role in practicing, learning and improving execution of actions and is often used by athletes and sports players to enhance performance (Mulder et al., 2004; Simonsmeier et al., 2020; Toth et al., 2020). MI practice, particularly when combined with action observation (Vogt et al., 2013; Eaves et al., 2016, 2022) can also be used for rehabilitation of people with Parkinson’s disease (Tamir et al., 2007; Kikuchi et al., 2014; Caligiore et al., 2017; Bek et al., 2021), as well as stroke patients (De Vries and Mulder, 2007; Sun et al., 2016; Binks et al., 2023) leading to improvements in timed motor performance, limb functioning, daily living skills and reduction of freezing of gait. Therefore, MI training could provide a route to improving motor ability in autism, providing autistic individuals are able to perform MI.

MI can be investigated using several methods including behavioral experiments involving laterality judgments (e.g., right or left hand), mental chronometry and perspective taking, as well as self-report questionnaires (McAvinue and Robertson, 2008). Starting with implicit measures, the hand rotation task (Parsons, 1987) is a method regularly used to investigate implicit MI and many variations of the paradigm have been used. The task requires participants to judge the laterality of hands (right or left) that are presented either from the front (palm) or back view and at different orientations (0–360°). This task uses an implicit measure of MI as participants are not directly instructed to use MI, but engage in this technique to solve the orientation based problem. Outcome measures include accuracy and reaction times. Typically, reaction times increase as the angle increases away from neutral, termed the angle or slope effect and is considered evidence that participants solve the task by mentally stimulating the rotation of their own hand to match the orientation of the displayed hand (Parsons, 1994). However, as it is also possible that participants use a non-MI based strategy such as mental (object) rotation of the hand images (Gardner et al., 2013), a key indicator of whether participants use MI in this task is the effect of biomechanical constraints on reaction times. Reaction times tend to be faster for stimuli where the hands are in a medial rotation rather than lateral direction, due to it being physically easier to rotate your hand in a medial rather than lateral direction (Parsons, 1994; Butson et al., 2014). The stimuli presented in these tasks can vary, with studies either using photographs of real hands or line drawings and some studies use whole bodies, termed the Own Body Transformation task (Gardner et al., 2013) or body rotation tasks.

A variation of the hand rotation task consists of the presentation of two hands or body images that are rotated at different angles and participants are asked to indicate whether the hands are the same or different. This task is based on mental object rotation paradigms where participants are presented with 2D or 3D shapes or objects in different orientations and have to judge whether they match the target shape/object (Shepard and Metzler, 1971). It requires participants to use visual imagery to mentally rotate the object and spatial working memory to hold the image representation in their mind (Shepard and Metzler, 1971; Muth et al., 2014). The use of hands or bodies instead of objects opens the possibility that participants use MI instead of object rotation to complete the task. To differentiate between these two strategies, a steeper reaction time slope across the angles for objects compared to hand (palm) or body stimuli (facing toward) has been interpreted as an object rotation versus a more embodied approach, respectively (Parsons, 1987; Zacks and Tversky, 2005; Yu and Zacks, 2010; Steggemann et al., 2011).

For explicit measures of MI, two common questionnaires are the Kinesthetic and Visual Imagery questionnaire (KVIQ-10) (Malouin et al., 2007) and the Movement Imagery Questionnaire (MIQ) (Hall and Martin, 1997; Gregg et al., 2010). These questionnaires ask participants to perform physically, then explicitly imagine the same movement (e.g., elbow flexion) and rate certain aspects about the imagined movement using Likert-type scales. For the KVIQ-10, participants are asked to rate the clarity (visual subscale) and intensity (kinesthetic subscale) of the imagined movement, while for the MIQ they are asked to rate the ease/difficulty of seeing and feeling the imagined movement. While both measure visual and kinesthetic modalities of MI, unlike the KVIQ, the MIQ is self-administered and measures more functional movements.

Mental chronometry tasks compare the durations of imagined versus executed movements with smaller differences in times suggesting greater accuracy and ability in MI (Wong et al., 2013; Williams et al., 2015). For example, participants might be asked to execute a series of timed pointing movements, then asked to imagine performing the same movements, indicating when they have finished. Better MI ability would be represented as closer durations between the executed and imagined actions. Furthermore, the similarity between real and imagined movements has been taken as evidence of the close relationship between MI and action planning (Jeannerod, 2001; Munzert et al., 2009).

Other behavioral paradigms that potentially measure MI include perspective taking experiments, which require a participant to imagine themselves in a certain orientation and/or from the viewpoint of another person and to then judge the laterality or position of other objects. This type of task can be performed using an embodied strategy of imagining oneself in the new orientation or location of the other or it can be performed using a non-embodied approach such as a rule-based strategy, line of sight computation or by mentally rotating the objects in the scene (Michelon and Zacks, 2006; Kessler and Thomson, 2010; Kessler and Wang, 2012). However, even if individuals use the embodied strategy, these tasks do not necessarily require MI as they can simply be performed by visually imagining the viewpoint from the new perspective, without imagining one’s body parts (Ward et al., 2022). Therefore, whether perspective taking tasks investigate MI varies with different studies. For example, MI is more likely in perspective taking studies that involve referencing a body part or an action (e.g., Conson et al., 2015), as opposed to just reporting the location of objects (e.g., visual perspective tasks in Pearson et al., 2016). More generally, MI of whole body movements is less commonly investigated as highlighted by a meta-analysis by Hétu et al. (2013) who analyzed whole body with upper body tasks due to the small number of tasks in the former category. However, as there are no prior systematic reviews on the topic of MI in autism, we wished to provide a comprehensive review that included less “traditional” MI tasks particularly as perspective tasks are of interest to autism researchers. As the involvement of MI in perspective taking tasks is dependent on study design and instructions, the perspective taking studies included in this review will be individually considered, based on the details of the task to decide whether they measure MI and therefore should be included in the synthesis of results.

This systematic review aims to evaluate and bring together the findings of research investigating MI abilities in autistic individuals using the MI tasks described above. As MI relies on similar motor control processes involved in executing actions, it is possible that motor coordination difficulties frequently present in autistic individuals may lead to altered MI. This review is timely as although there are a growing number of studies in this area, it remains unclear if MI is affected in autistic individuals and if it is, what aspects or in what way is MI affected. A systematic review on the topic of autism and MI is yet to exist in the current literature. A more complete understanding of MI in autism can be helpful for advancing knowledge of simulation ability and whether MI could be used in therapies to support autistic motor coordination difficulties.

Please see Supplementary Information S1, S2 for PRISMA checklist.

This review aimed to provide an understanding of the current state of knowledge around MI in autism. Focusing on the implicit tasks, there appears to be good evidence of a biomechanical effect in children for hand laterality tasks (Conson et al., 2016; Chen et al., 2018; Johansson et al., 2022), suggesting the presence of MI. Where a biomechanical effect was absent in the autistic group, this seems to be explained by methodological differences such as reduced stimulus richness or limitations in power (Conson et al., 2013; Johansson et al., 2022). Interestingly, all of these hand rotation studies involve children and adolescents, with none being performed with autistic adults. For the body rotation studies that use a laterality judgment, there are mixed results with evidence of differences in processing between the two groups (Pearson et al., 2014; Conson et al., 2015) or no group differences (Ring et al., 2018). Although further studies are required, there does not appear to be a difference in findings according to age. However, the challenge with these tasks is that they do not consistently report the same measures (e.g., front vs. back and angle effect), and without these tasks being able to measure a biomechanical effect it is difficult to interpret how participants are performing the task. Furthermore, when an angle effect is measured, the presence of a steeper slope in autistic individuals (Pearson et al., 2014) could be interpreted as either difficulties in MI processing or “deeper” MI (Kessler and Wang, 2012). The latter fits with the observation of an increased posture effect in the autistic group (Conson et al., 2016).

The presence of implicit MI in the hand rotation task fits with findings of comparable automatic imitation ability in autistic and non-autistic groups (Cracco et al., 2019; Poliakoff and Gowen, in press), both of which are thought to involve motor simulation. However, group differences have been observed for other behaviors that involve motor simulation such as voluntary imitation (Wild et al., 2012; Edwards, 2014; Vivanti and Hamilton, 2014; Gowen et al., 2020), action understanding (Vivanti et al., 2011; Todorova et al., 2019; Federici et al., 2020) and action prediction (Gowen et al., 2022). It may be that in comparison to implicit MI and automatic imitation, these other behaviors require dynamic and more detailed sensory motor processing involving comparison with predictions to produce an accurate percept or action. This greater demand on sensorimotor integration could result in group differences (Gowen et al., 2023).

Three tasks used the same/different judgment variant of the rotation tasks (Soulières et al., 2011; Pearson et al., 2014, 2016). However, for these type of tasks, object rotation strategies appear to be the preferred strategy in typical populations (Zacks and Tversky, 2005; Yu and Zacks, 2010; Steggemann et al., 2011), limiting the conclusions that can be drawn from these studies. In addition, the biomechanical effect tends not to be examined in these tasks, although this is possible to include. Furthermore, none of the studies directly compared the angle effect between hand (palms) or body (facing toward) and object stimuli which has been the previous method in typical groups to differentiate MI from object rotation strategies (Parsons, 1987; Yu and Zacks, 2010; Steggemann et al., 2011). It is interesting to note that autistic and non-autistic groups perform more similarly on this task, suggesting that mental object rotation is similar in both these groups. Indeed, including an object rotation task with all hand and body rotation tasks is valuable for several reasons including distinguishing whether group differences could be due to general mental rotation and motor coordination difficulties, as well as difficulties in left/right judgments. Interestingly, those studies that included both a hand and object rotation task (Conson et al., 2013; Chen et al., 2018) showed that group differences were only present in the former task highlighting that that these group differences were specific to hand processing and unlikely to be caused by the above confounding factors.

A variety of tasks examined explicit MI in autistic children, adolescents and adults. As discussed in the introduction, visual perspective taking tasks do not necessarily involve MI (Ward et al., 2022), so we focused on those tasks that made reference to body parts and were more likely to elicit MI. There are too few studies to make firm conclusions, although the findings of Conson et al. (2015) showing that the responses of the autistic group did not vary when the observed actor was grasping an object suggests that MI was not used. Surprisingly only one study has used a questionnaire measure (Gowen et al., 2022) and found no group differences in their self-rated clarity and intensity of visual and kinesthetic imagery. Of note is that it has been suggested that questionnaires such as the KVIQ could be used to provide a “snapshot” of MI ability as they involve generation of a motor image, which may be most critical for MI ability (Kraeutner et al., 2020). The two studies that compared explicit MI with physical execution (Piedimonte et al., 2018; Xie et al., 2022), both revealed that the autistic group did not engage MI. These explicit tasks would benefit from qualitative input from the participants as in previous work with non-autistic groups (Zacks and Tversky, 2005; Yu and Zacks, 2010; Gardner et al., 2013) to understand what strategies they were using and add context to the findings. There were no clear differences in MI according to age group, although the same tasks would need to be conducted across ages to clarify this. A further consideration for both the reviewed and future explicit MI studies is that a full description of the instructions is provided. For example, Van Caenegem et al. (2022) highlighted substantial underreporting of the instructed modality and perspective which can lead to challenges with replication, understanding and synthesizing of studies, which we would add is particularly important when studying heterogeneous conditions such as autism. Future studies would benefit from following the recently published Guidelines for Reporting Action Simulation Studies (Moreno-Verdú et al., 2023).

A clear conclusion from this systematic review is that the field would benefit from a battery of implicit and explicit MI tasks being conducted across the same groups of autistic children, adolescents and adults. Recent work in typical populations has highlighted that MI involves many different dimensions. For example, Kraeutner et al. (2020) asked typical participants to perform a variety of MI tasks such as questionnaires, hand laterality and chronometry and conducted a PCA analysis on the data. They observed that tasks loaded onto three main components termed “generation,” “manipulation” and “maintenance” of motor images. Conducting a similar study in autistic and non-autistic individuals would allow greater understanding of whether specific MI processes are functionally equivalently or not, and whether social ability may be related to the different tasks. Similarly, it will also be important to directly compare implicit and explicit MI ability as they are likely to involve different processes: implicit MI mostly relies on motor representations within the parietal cortex and may not require motor preparation or control related to the activation of supplementary motor area usually associated with explicit MI (Hétu et al., 2013). A further aspect that future studies would benefit from is a comparison of MI with execution of the same action, or some measure of motor ability, to assess the potential relationships between motor coordination difficulties and MI. It is possible that autistic participants with motor coordination difficulties may still be able to accurately imagine their own (less accurate or slower) movements as has been found for people with Parkinson’s Disease (for discussion of this see, Poliakoff, 2013).

A better understanding of MI in autistic individuals has implications around support for motor coordination difficulties that are common in this population but infrequently assessed or treated (Zampella et al., 2021). Research is beginning to highlight the benefit of using MI to improve motor coordination in individuals with Developmental Coordination Disorder (Scott et al., 2021), suggesting that this may also be a useful approach for autistic people. Using qualitative focus group methods, Gowen et al. (2023) recently described the range of motor coordination difficulties and their impact from the viewpoint of autistic adults. They also explored what strategies participants used and some participants described the importance of pre-planning, visualization and imagining, although others found this too challenging. Therefore, employing MI techniques may well be beneficial for some autistic people, although an individual approach is likely to be most appropriate. Indeed, MI ability in typical populations is variable (Moran et al., 2012; Cumming and Eaves, 2018) and influenced by multiple factors such as sporting and dance experience, MI practice, motor and cognitive ability and tactile discrimination (Isaac and Marks, 1994; Arvinen-Barrow et al., 2007; Pelletier et al., 2018; Krüger et al., 2020; Dhouibi et al., 2021; Mao et al., 2022). Therefore, future research examining MI ability and impact of MI training on motor coordination in autistic populations will need to take into account individual differences.

One limitation of the literature on MI in autism is the presence of a sex bias with most of the studies mainly including males. Kessler and Wang (2012) showed that females are more likely to take an embodied approach (i.e., imagining oneself in the new orientation) to visual perspective tasks, whereas other studies suggest that males are better than females on mental rotation (Geiser et al., 2008). It will be important for future studies to include more balanced groups and examine the presence of possible sex differences. A further limitation is that the majority of studies involving the hand laterality task involve children, making it difficult to generalize the findings to the adult population. Explicit MI tasks were too varied and few to draw firm conclusions. A third limitation was the lack of studies investigating the neural processes underlying MI in autistic individuals. This would be a valuable future direction, particularly in relation to identifying possible differences in how autistic individuals might perform MI tasks. Finally, one study was included that was labeled as “exclude” according to quality assessment, but we felt it was important to present the study due to this being the first systematic review of the topic the overall small number of studies on the topic of MI in autism.

In summary, the presence of a biomechanical effect in autistic children performing hand rotation tasks suggests that they are able to implicitly use MI. However, further research on both hand and body rotation tasks is required in both children and adults to follow up suggestions that there may be group differences in the effort or depth of MI. Conclusions about explicit MI tasks are complicated by the small number of studies, the range of tasks conducted, with some the tasks potentially not tapping into MI. The field would benefit from research comparing a range of MI tasks in autistic and non-autistic children and adults to better understand what elements of MI may be affected and to assess the therapeutic potential of MI in these populations.

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

EG: Conceptualization, Data curation, Investigation, Methodology, Supervision, Writing – original draft, Writing – review & editing. EE: Data curation, Investigation, Methodology, Writing – original draft, Writing – review & editing. EP: Conceptualization, Investigation, Methodology, Writing – review & editing.