Neural representations of visual statistical learning based on temporal duration

Time perception is an essential aspect of daily life, and transitional probabilities can be learned based on temporal durations that are independent of individual objects. Previous studies on temporal and spatial visual statistical learning (VSL) have shown that the hippocampus and lateral occipital cortex are engaged in learning visual regularities. However, it remains unclear whether VSL on temporal duration unlinked to object identity is represented in brain regions involved in VSL and object recognition or in those involved in time perception without sensory cortex involvement. We examined this question by adapting a VSL paradigm to time perception using functional magnetic resonance imaging. Thirty-four students participated in the VSL experiment, comprising a familiarization scan and a subsequent familiarity-decision test. The region-of-interest (ROI)-based classification showed chance-level performance across all ROIs, but only the left medial frontal gyrus, which is involved in subsecond time perception, showed a moderate effect size with 95% confidence intervals not crossing the chance level of 50%. Moreover, searchlight analysis showed that the right orbitofrontal cortex successfully decoded brain responses related to the processing of structured timing sequences. Meanwhile, representational similarity analysis suggested that the neural signal patterns could not be divided between the structured timing and pseudo-random sequences in the lateral occipital cortex. Our findings serve as a pilot study suggesting that the medial frontal and orbitofrontal regions are involved in VSL based on temporal duration, independent of visual object processing, which is a key and common timing mechanism for predicting sequential events.