Assessing Statistical Learning through Alternating Serial Reaction Time Task

The Alternating Serial Reaction Time (ASRT) task was used to assess statistical learning and consolidation (49 (link), 50 (link)). In this task, four horizontally arranged empty circles are presented on the screen and a stimulus (a dog's head) appeared in one of the circles (51 (link)). Participants were instructed to press a corresponding key (Z, C, B, or M on a QWERTY keyboard) as quickly and accurately as they could when the stimulus occurred using their index and middle fingers. After the correct response of the participant, the next stimulus appeared 120 ms later. Unbeknownst to the participants, the presentation of stimuli followed an eight-element sequence, within which pattern (P) and random (r) trials alternated with each other (e.g. 2-r-4-r-3-r-1-r; where numbers indicate the four locations on the screen from left to right, and r denote a randomly chosen location out of the four possible ones; see Fig. 1).
Due to this alternating sequence, some runs of three consecutive trials (triplets) were more probable than others. In the example sequence 2-r-4-r-3-r-1-r, triplets 2-X-4, 4-X-3, 3-X-1, and 1-X-2 (where X indicates the middle element of the triplet) occurred with a higher probability because they were presented in every sequence repetition (P-r-P) and could also be formed by chance (r-P-r, see Fig. 1B). Note that here, we use X to indicate the middle element of the triplet because, for example, 4-X-3 (e.g. 4-2-3 in Fig. 1B) can appear both as a P-r-P structure (where the first and last element of the triplet belong to the predetermined pattern) and as a r-P-r structure (where the first and last elements are random, and the middle element is part of the predetermined pattern). In contrast, triplets 2-X-1 and 3-X-2 occurred with a lower probability since they could only be formed by chance (that is, their structure could only be r-P-r). The former triplet types are referred to as high-probability triplets and the latter ones as low-probability triplets. Overall in the task, high-probability triplets were five times more probable than the low-probability ones (27 (link), 41 (link)). Note that triplets were identified using a moving window throughout the stimulus stream. Thus, each trial was categorized as the third element of a high- or a low-probability triplet, and this categorization was used in our analyses; the same trial then served as the middle and the first element for the categorization of the following triplets.
The ASRT task enables us to separate statistical learning from general skill improvements. Statistical learning is defined as faster and more accurate responses to high-probability elements than to low-probability ones (50 (link)). In contrast, general skill improvements refer to average speed-up and changes in accuracy which are independent of the probabilities of events. These improvements reflect more efficient visuomotor and motor–motor coordination due to practice (9 (link), 18 (link)).