Placement of target in a trial introduces bias in statistical learning studies using the RSVP task

What is it about?

Among other things, we prove that even in the absence of any regularities in a set of shapes (Experiment 4), it is possible to observe a trend in the response times of target detection, that resembles the effect of "statistical learning", as present in many previous studies using the Rapid Serial Visual Presentation (RSVP) task. This misleading trend is present due to biased responses to shapes such that shapes presented later in a trial's sequence are responded to, quicker than earlier shapes. This trend, though misleading, matches what one would expect from responses based on the regularities present among the shapes (that each shape is part of a unique and fixed triplet, where its position within the triplet renders it a certain amount of predictability).

Featured Image

Why is it important?

Results of our experiments point to a confound present in the RSVP task - the stream location artifact - that biases RTs, irrespective of the predictability of the shapes, in a visual statistical learning paradigm. These results put certain previous reports into question, which attributed the trend in the RTs to the learning of the sequential relations among the shapes. This discrepancy demands a second look at previous reports that interpret this trend as proof of statistical learning.