On the interaction between implicit statistical learning and the alternation advantage: Evidence from manual and oculomotor serial reaction time tasks.

In this study, we examine how implicit statistical learning (ISL) interacts with the cognitive bias of the alternation advantage in serial reaction time (SRT) tasks. Our aim was to disentangle perceptual from motor aspects of learning, as well as to shed light on the cognitive sources of this alternation effect. We developed a manual (Study 1) and an oculomotor (Study 2) two-choice SRT task, with visual stimuli following the regularities of two binary artificial grammars (Fibonacci and its modification Skip). While these grammars share some deterministic transitional regularities, they differ in their probabilistic transitional regularities and distributional properties. The pattern of manual RTs in Study 1 provide evidence for ISL, showing that subjects learned the deterministic and probabilistic transitions in the two grammars. We also found a bias toward alternation (vs. repetition) in correspondence to non-deterministic points, regardless of their statistical properties in the grammars. Study 2 provides further evidence for both ISL and the alternation advantage, in terms of shorter manual RTs and higher accuracy rates of anticipatory eye movements. Saccadic responses preceding stimulus onset allow us to argue for the perceptual nature of ISL: participants detected regularities in the string by forming S-S associations based on the sequence of the perceived stimuli. Moreover, we propose that shifts in visuospatial attention preceding oculomotor programming play a role in the occurrence of the alternation advantage, and that such an effect is driven by the spatial location of the stimulus. These findings are also discussed with respect to the presence of two (possibly interacting) parsing strategies: statistical generalizations on the string vs. local hierarchical reconstruction.