Abstract
Although the visual system detects statistical relationships between objects with remarkable efficiency, it is unclear how statistical learning occurs. Here we examine how object representations are activated during statistical learning. In the experiment, participants viewed a continuous sequence of objects at the center of the screen while performing a cover 1-back task during exposure. Unbeknownst to the participants, the sequence contained pairs of objects where one object always appeared before another (e.g., A always appeared before B). At the periphery of the screen, all unique objects in the sequence were presented in fixed locations at all times during exposure. This means that at any given trial, the object in the central sequence was also presented in the periphery, as well as its partner in the pair and all the other objects in other pairs. Participants' eye gaze was tracked throughout exposure. At test, participants chose pairs over foils as more familiar, indicating that they successfully learned the object pairs. Importantly during exposure, we found that when the first object in the pair was presented at the center of the screen, participants looked at its partner (the second object in the pair) in the periphery more than the other objects in other pairs. When the second object in the pair was presented at the center of the screen, participants looked at its partner (the first object in the pair) in the periphery more than the other objects in other pairs. This finding suggests that seeing the first member automatically activates the representation of the upcoming second member in a pair, and seeing the second member automatically activates the representation of the preceding first member. This study not only provides a novel paradigm to measure representation activation during statistical learning, but also elucidates the mechanism of how statistical learning occurs.
Meeting abstract presented at VSS 2018