Statistics Learning of Target Regularities in a Pop-out Search:
Behavioral Performance and Neural Mechanisms
Abstract
The study examined human performance and related neural mechanisms in a
pop-out search with different probabilities of target location and the
relation between the target location and feature. In a search array, we
introduced the binding relation between two target features and two
kinds of location probability. Moreover, in the second half of the
experiment, such a probability pattern for location/feature binding was
reversed. Behavioral results revealed successful statistical learning of
probability for both absolute target location and target’s
location-feature binding indicated by faster RTs in the high-probability
conditions for both location and location-feature binding. Moreover, the
learning benefit for the probability of location-feature binding
acquired during 1st (training) phase was still expressed in the 2nd
(reversal) phase despite the actual binding probability was reversed.
ERP results suggested that both the attentional selection and response
selection process were affected by such learning revealed in the
difference in N2pc and LRP amplitudes between the two conditions with
different binding probability in the reversal phase. An expectation to
the high probability for location-feature binding was also suggested
from time-frequency analysis and Multi-Variate Pattern Classification
(MVPC) indicated by larger alpha ERD magnitude and lower decoding
accuracy, respectively, when the target appeared at the high-binding
location in the training phase instead of the reversal phase. Overall,
we have demonstrated behavioral evidence and 4 EEG markers for the
associative learning of the probability of relation between location and
feature of target in a pop-out search.