In order to probe the production of kinematic cues to signal boundaries in action sequences, adults performed sequences of three actions on an object, with or without an action boundary following the second action. Movement of the hand was recorded via motion tracking, and it was found that the boundary was marked by a lengthening of the pre-boundary action and by a pause. These cues are also found in prosody to signal phrase boundaries in speech, suggesting a close coupling of the mechanisms underlying boundary production in both domains.

From about six months of age onward, infants observing an action, such as a grasp, start to shift their gaze from the moving agent to the goal before the action is complete. A variety of factors that influence such goal-predictive gaze have been identified. However, the underlying cognitive processes are heavily debated. We propose that our minds structure sensorimotor dynamics into probabilistic, generative event-predictive models, and, choose actions with the objective to minimize predicted uncertainty. We implement this proposition by means of event-predictive learning and active inference. Trained on manual object-manipulations, the generation of goal-predictive gaze emerges: The model starts fixating the anticipated goal at the start of an observed event when a familiar agent (i.e., a hand) is involved. Meanwhile, the model keeps tracking unfamiliar agents (e.g., a claw) performing the same movement. We conclude that event-predictive learning combined with active inference may be critical for eliciting action-goal predictions.

Infants develop the ability to anticipate action goals during their first year, as shown by anticipatory gaze behavior. As they grow older, this is first evident for most familiar actions and agents, e.g., human hands performing a reaching action; later also for unusual agents (e.g., mechanical claws). We argue that this ability emerges as infants attempt to segment the world they observe into events—to infer the currently unfolding events and to predict their consequences for minimizing anticipated uncertainty. We propose a computational model that explains this development from a functional, algorithmic perspective, CAPRI² (Cognitive Action PRediction And Inference in Infants). Our model integrates proposals about the development of object files, event files, and physical reasoning abilities into a learning and probabilistic planning-as-inference framework. While observing goal-directed, or arbitrary, interactions between two objects (i.e., potential agent and patient), CAPRI²'s active inference processes infer both maximally consistent event interpretations and motor actions (here, eye fixations), where the latter are executed in the service of further minimizing current and anticipated uncertainties. As a result, CAPRI² models typical developmental patterns of infants' anticipatory gaze behavior in an emergent manner. In particular, to successfully model the emergent developmental pattern, our model suggests that infants activate object event files, implicitly reason about object interactions in an event-oriented manner, infer consistent interpretations of their observations, and control their gaze shifts to minimize anticipated uncertainty. We propose that these mechanisms, as reflected in our model, may constitute fundamental building blocks for developing goal-predictive capacities in infants.