Can we rate the similarity among invisible objects?

Conscious visual experience can be studied at least with two distinct methodologies. One method is a detection paradigm, studying what allows a stimulus to be consciously perceived, and what renders it otherwise. The other method is a similarity paradigm, which tries to characterise a quality of an experience based on its similarity relations to other experiences. However, the relationship between results obtained by the two methods has not been systematically understood. While the positive ability for similarity rating naturally implies above chance detection, the reverse is unclear due to possible contribution of non-conscious processing to similarity judgment. To address this, we devised a novel paradigm, which asked participants to simultaneously detect the locations and judge the similarity of masked colour patches. To manipulate patch visibility, we presented the patches in 16ms, 66ms and 133ms SOAs (Stimulus Onset Asynchrony). Our results showed that participants could not rate similarity with chance detection. Additionally, even under high detection accuracy (> 0.82), participants’ colour similarity structure under 16ms was more clustered than those longer SOAs, revealing nuanced differences in the quality of colour experiences across different SOAs. We concluded that similarity judgements reliably reflect the quality of conscious experience.