Understanding abstract knowledge structures in implicit perceptual sequence learning

Traditional approaches to studying perceptual sequence learning (SL) often employ adaptations of the classical serial reaction time task, albeit these tasks suffer from confounding factors such as (oculo)motor learning effects. Unlike motor SL, the extent to which pure perceptual SL can occur implicitly without (oculo)motor learning remains uncertain. We adapted a previously formulated task (Garvert et al., eLife, 6, 1-20, 2017) to isolate perceptual sequence learning, without the interference of (oculo)motor confounds, and to determine whether perceptual sequence learning can occur implicitly. Fifty participants judged whether each object appeared in its original or mirrored form, gradually improving performance based on feedback. Unbeknownst to participants, the succession of these objects followed a probabilistic sequence. A training phase consisting of 8 regular blocks was followed by a testing phase, where 5 random and 5 regular blocks were presented alternatingly. A force-choice recognition test probing knowledge about specific transitions in the task was also used to assess explicit knowledge. Our findings indicate robust perceptual SL effects, as indicated by slower reaction times (RTs) in random blocks than regular blocks. Notably, transitions between objects with higher communicability (i.e., a metric of objects' connectedness within the underlying grid) showed lower RTs in regular, but not random blocks. This indicates that perceptual SL in our task may rely on strategic cognitive processes in response to violations of expectation. Importantly, our results also demonstrate that explicit knowledge of the underlying structure did not influence perceptual SL in any way, suggesting that learning was driven by implicit knowledge.