Imitation learning

Learning a new movement requires the sensorimotor system to form a novel representation of its kinematic structure. This often occurs through imitation learning, where lower‑level sensorimotor processes encode the biological‑motion features of an observed action. However, top‑down processes, such as attention, task goals, and social context, can also influence how these movements are represented.

To examine how these bottom‑up and top‑down processes interact, we used several variations of an imitation protocol that essentially involve observation of novel, typical or atypical biological‑motion kinematics, alongside a constant‑velocity control model.

Across several experiments, we showed that adult participants reproduced movement sequences containing typical or atypical biological‑motion cues. Kinematic analyses showed that participants reliably encoded and imitated the velocity‑based features of biological motion, and that these representations differed from the constant‑velocity control condition.

Although the presence of end‑state targets did not influence the imitation of biological‑motion kinematics, they did affect movement‑time accuracy, suggesting that attentional demands modulate certain aspects of performance.

Additional dual‑task and selective‑attention manipulations further demonstrated that while biological‑motion coding is primarily driven by bottom‑up sensorimotor processes, it can be strengthened or weakened through top‑down attentional control.

Together, these findings indicate that imitation learning reflects a complementary interaction between lower‑level sensorimotor mechanisms and higher‑level attentional processes, each shaped by the environmental and task context.

Imitating cyclical arm movements can be involuntarily disrupted when a person simultaneously observes another movement that conflicts with their own. This phenomenon is known as motor contagion, and is an important mechanism that supports rapport, cooperation, and smooth social interaction.

In our research, we examined how social context and specific movement features influence this interference using tasks that resemble everyday social interactions. Participants performed horizontal arm movements while observing either congruent horizontal motion or incongruent curvilinear motion, and were primed with pro‑social or anti‑social words beforehand. As expected, incongruent biological motion produced greater movement deviation.

Notably, interference was stronger following anti‑social primes, suggesting that the effect of social cues depends on how they interact with an individual’s self‑concept and top‑down social processing.

In further experiments, we explored which aspects of observed movement drive motor contagion by manipulating the congruence of trajectory and end‑points. Participants observed horizontal (congruent), vertical (incongruent trajectory and end‑points), or curvilinear movements (incongruent trajectory but congruent end‑points).

Motor contagion was greater for vertical than horizontal stimuli, and even stronger for curvilinear stimuli, indicating an additive effect of mismatched trajectory and shared end‑points. This pattern persisted even when participants faced perpendicular to the display, confirming that end‑points were coded as part of the movement rather than as external spatial cues.

Together, these findings show that motor contagion arises from shared features between observed and executed actions and is shaped by both bottom‑up processing of biological motion and top‑down social modulation.

They support the theory of event coding, which proposes that interference increases when observed and executed actions overlap in their representational features.

People often learn new skills by copying the actions of a teacher, parent, or carer in everyday settings such as the home or classroom. For some autistic individuals, this can be challenging, and these difficulties have been linked to the processes involved when observing a new action.

To examine this more closely, the study tasked autistic and non‑autistic participants with learning a new visuo-motor skill solely by watching a model, without subsequent voluntary imitation. The results showed that autistic participants matched the model’s movement characteristics (such as movement time and style) just as closely as non‑autistic participants.

This suggests that previously reported imitation difficulties do not stem from how autistic people watch or process another person’s movements, but instead relate to sensory‑motor challenges involved in planning and preparing to imitate. These findings highlight the importance of considering autistic motor difficulties when teaching new motor or everyday skills through modelling.

Read more about Observational learning

Autistic individuals are able to learn new motor skills, but the movements they produce are often less accurate and more variable than those of non-autistic adults.

These differences are thought to reflect how the brain integrates sensory information and plans actions. In this study, we examined the mechanisms involved in sensorimotor learning by measuring accuracy, variability, movement timing, and the balance between feedforward (planning‑based) and feedback (online correction) control.

Both autistic and non‑autistic participants improved their performance with practice, particularly when given feedback about their results. However, autistic participants remained less accurate overall. Detailed movement analysis showed that autistic individuals displayed greater spatial variability early in the movement at the point where force and direction are being planned, but similar variability later in the movement when feedback helps guide the action.

This pattern suggests that autistic movement differences are linked to how actions are planned and initiated, rather than how feedback is used during execution. These findings highlight that feedforward and feedback control processes may operate differently in autism, and understanding these mechanisms can help explain how autistic people acquire everyday motor and social actions.

Read more about Specificity in planning and feedforward control

Humans are highly skilled at learning new sensorimotor behaviours, but research shows that the way motor commands, sensory feedback, and visual information are integrated during learning differs in autism. This can lead to atypical internal action models.

In this study, we examined how autistic and non-autistic adult participants learned a novel movement sequence timing task, and then if this transferred to a related visual‑perception task.

Although autistic adults showed less accurate motor timing overall, both autistic and non‑autistic participants improved their motor timing during practice with the availability of feedback.

This improvement was still evident in retention without the availability of feedback. In the subsequent perception task, both groups performed similarly in judging timing differences. However, only the non‑autistic group showed a significant relationship between their motor‑timing accuracy and their perceptual‑timing judgments.

These findings suggest that while autistic adults adapt through practice, the sensorimotor processes that support internal action model formation operate differently in autism.

Read more about Sensorimotor learning