Badminton expertise shapes cue extraction and evidence accumulation during perceptual prediction

This study aims to investigate the impact of badminton-specific experience on the mechanisms of cue extraction and processing during perceptual prediction from a computational modeling perspective.

Thirty-six participants (18 experts and 18 novices) were recruited for the experiment. Utilizing a temporal occlusion paradigm, three cue stages were established: 40 milliseconds before racket-shuttlecock contact (T1), at contact (T2), and 40 milliseconds after contact (T3). Participants were tasked with predicting the landing location of the shuttlecock. The study not only recorded behavioral metrics (accuracy and reaction time) but also employed the Drift Diffusion Model (DDM) to jointly model the behavioral data, extracting three latent processing parameters: drift rate, decision threshold, and non-decision time.

Accuracy analysis revealed significant main effects for cue stage and group, alongside a significant interaction. Simple main effect analysis between groups indicated that experts' accuracy at T1 and T2 was significantly higher than that of novices, whereas the difference was non-significant at T3. Reaction time analysis also showed significant main effects for cue stage and group, alongside a significant interaction; experts exhibited significantly shorter reaction times than novices across all three time points. DDM results demonstrated that experts had a significantly higher drift rate than novices at T1 and T2, while this inter-group difference disappeared at T3. As information increased, experts showed a more pronounced upregulation in the decision threshold at T3, whereas novices exhibited greater threshold fluctuations. Non-decision time was consistently shorter for experts across all stages.

The optimization of perceptual prediction mechanisms driven by badminton-specific experience exhibits stage specificity. The expert advantage is primarily concentrated in the early processing of kinematic cues when information is incomplete, manifesting as a higher rate of evidence accumulation. In stages with more abundant information, experts maintain output stability through an optimized dynamic speed-accuracy tradeoff (threshold upregulation) and retain shorter non-decision times across all stages. This suggests that long-term training facilitates the automated synergy of stimulus encoding and motor preparation.