The effect of dynamic motion and simultaneous presentation on the statistical learning of nonadjacent dependencies in manual gesture sequences

Statistical learning enables people to uncover regularities in sensory input. However, the effectiveness of this process hinges on how data is organized within each sense. A preregistered study explored how motion cues and temporal/spatial arrangement affect the understanding of non-adjacent dependencies within sequences of sign-like manual gestures.

Presentation format (serial or simultaneous) and the level of dynamism (dynamic or static) were altered to observe their impact. Participants engaged in an online task where reaction times and accuracy were monitored to gauge learning. They also completed an offline grammaticality-judgment task. Across all conditions, 123 participants demonstrated an ability to comprehend the underlying structure: reaction times shortened during training, spiked when input was randomized, and offline accuracy surpassed chance levels.

Simultaneous presentation led to greater online improvements during exposure, yet offline grammaticality judgments did not vary between conditions. Motion cues did not enhance model fit. These results imply that visual statistical learning for structured manual gesture input gains more from concurrent spatial information than from sequential presentation.

This indicates that the most effective input structure hinges on modality-specific perceptual constraints. This research was funded by a Lendület Research Grant from the Hungarian Academy of Sciences.