Electrophysiological evidence of attentional control from working memory and long-term memory: how conjunction features are represented during target repetition

In three experimental chapters, we asked whether features of conjunctive stimuli are represented as separated or integrated fashion in vWM. We measured several electrophysiological indices while participants were cued to search a constant target that was defined by color and shape conjunction in six consecutive trials. Based on the previous observation that attentional template would be off-loaded from vWM to an alternative mechanism during the same target learning, the underlying assumption of the present study is, if conjunctive features are represented in a separated fashion, their impact on task performance should be largely independent when attentional template was off-loaded from vWM. We then manipulated the similarity between search targets and distractors in the last two trials. Specifically, all search distractors could match either target’s shape or color, thereby blocking the role of shape and color during the target selection respectively. We also included a baseline condition to make comparison, in which all search distractors have no target features overlap. Experiment 1 & 2 first revealed that mean amplitude of SPCN and LPC time-locked to the memory display systematically decreased as a function of target repetition, suggesting the demands on vWM to maintain the attentional template was lessened. This phenomenon is likely due to an off-loading of the template from vWM to the vLTM. Results of the last two repetition trials provided fruitful information to evaluate the impact of color features and shape features on search performance. We found that, when all distractors matched the target shape, search efficient was the same as the baseline condition (i.e., all distractors are heterogeneous) in the behavioral level, but the ERP results showed an attentional guidance by search targets along with an attentional suppression by shape-matched distractors. Moreover, the target selection and distractor suppression appeared to be working in parallel when we further divided the data based on the vertical elevation in Experiment 2. Contrarily, search slope significantly dropped down relative to the baseline when all distractors matched the target color, but we did not observe the distractors suppression in the ERP level. Instead, targets elicited SPCN, presumably due to the guidance of attentional switched from feature-based to object-based manner. Further, we found the SPCN and FN400 time-locked to the cue increased in the memory phase when encountered color-matched distractors in the previous trial, suggesting a strategical resampling to enhance the search performance in the next trial. Experiment 3 was designed to further examine whether objects are encoded in their entirety in the memory display. Participants implicitly learned which features of the to-be-remember object would direct to search target, targets could match either the color or shape of the memory cue in six consecutive trials. Again, in the last two repetitions, we then instructed participants to identify a full memory matched target (conjunction) instead of the previous single feature matched target. The successful identification revealed that participants did not discard the task-irrelevant feature regardless of search intentions required them to configure a color template or shape template. Besides, search efficiency was better when encountering the conjunction target in the remember shape series than remember color series, suggesting the color feature acquire better learning even when search task emphasized the role of shape features. These findings indicated that the format and the structure of remembered information in vWM are better to be considered including both object-based and feature-based levels. That is, the initial object encoding follows an object-based manner, whereas conjunctive features are bound indirectly in a hierarchical structure.