Eye movements dissociate between perceiving, sensing, and unconscious change detection in scenes

Abstract

Detecting visual changes can be based on perceiving, whereby one can identify a specific detail that has changed, on sensing, whereby one knows that there is a change but is unable to identify what changed, or on unconscious change detection, whereby one is unaware of any change even though the change influences one's behavior. Prior work has indicated that the processes underlying these different types of change detection are functionally and neurally distinct, but the attentional mechanisms that are related to these different types of change detection remain largely unknown. In the current experiment, we examined eye movements during a change detection task in globally manipulated scenes, and participants indicated their change detection confidence on a scale that allowed us to isolate perceiving, sensing, and unconscious change detection. For perceiving-based change detection, but not sensing-based or unconscious change detection, participants were more likely to preferentially revisit highly changed scene regions across the first and second presentation of the scene (i.e., resampling). This increase in resampling started within 250 ms of the test scene onset, suggesting that the effect began within the first two fixations. In addition, changed scenes were related to more clustered (i.e., less dispersed) eye movements than unchanged scenes, particularly when the subjects were highly confident that no change had occurred -providing evidence for change detection outside of conscious awareness. The results indicate that perceiving, sensing, and unconscious change detection responses are related to partially distinct patterns of eye movements.

The role of recollection and familiarity in visual working memory: A mixture of threshold and signal detection processes.

“…In addition, speeding responses in a binary response working memory task leads the linear ROCs to shift toward the chance diagonal (Yonelinas & Jacoby, 1995a; Figure 6c), suggesting that the proportion of items that are recollected depends critically on the processes that unfold during retrieval. In addition, in a recent working memory study for scenes, we found that perceiving-based working memory responses, but not sensing-based responses, depend critically on the subjects fixating on a region at study and then re-fixating on that changed region at time of retrieval (Ramey et al, 2022; also see Henderson & Hollingworth, 1999). These re-fixation results suggest that whether recollection contributes to WM performance depends critically on how attention is deployed during both encoding and retrieval.…”

“…The selective relationships between the intercepts and curvilinearity of ROCs on the one hand, and the mental states of perceiving and sensing on the other, indicates that the contents of working memory are available to conscious awareness; but in addition, they suggest that subjects can distinguish between two different types of conscious experience. Whether these two types of responses can be further differentiated from truly unconscious processes is not yet known (although see Ramey et al, 2022). In either case, the existing results indicate that subjects have conscious access to at least some of the processes underlying WM (Fechner, 1887; James, 1890), but points to the importance of distinguishing between cases in which the subjects have conscious access to the specific details that have changed and cases in which they can detect a change has occurred but are not able to recollect what has changed.…”

“…Thus, examining the shape of the ROC is useful in determining if there is a working memory threshold or not, but the ROC shape in itself does not tell us when in the encoding-storage-retrieval sequence that the thresholding of memory occurs. In principle, ROC studies could be designed to determine which period is most critical in altering the threshold, but as will be described below this has only rarely been directly examined (Ramey et al, 2022; Xie & Zhang, 2017b; Yonelinas & Jacoby, 1995a).…”