Explicit Access to Detailed Feature Distribution Representations
Poster Presentation: Sunday, May 18, 2025, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Perceptual Organization: Ensembles
Schedule of Events | Search Abstracts | Symposia | Talk Sessions | Poster Sessions
Vladislav Khvostov1,2, Julie Golomb1, Árni Gunnar Ásgeirsson3, Árni Kristjánsson4; 1Department of Psychology, The Ohio State University, USA, 2HSE University, Russia, 3University of Akureyri, Iceland, 4University of Iceland, Iceland
The human visual system can quickly process groups of objects (ensembles) and build compressed representations of their features. The current thinking is that conscious access to this representation is very limited: observers cannot report any distributional aspects beyond simple summary statistics, such as the mean or variance. However, people subjectively feel they can perceive much more than that. Here, we introduced a new paradigm (Feature Frequency Report) to thoroughly probe this explicit representation. We showed that in this case, the subjective impression is closer to the truth than the scientific consensus. Observers viewed 36 disks of various colors for 800 ms and then reported the frequency of a randomly-chosen color using a slider (0-8 disks). In Experiment 1 (N=10), the colors of the disks had a Gaussian, uniform, or bimodal distribution with a random mean color. The distributions of averaged responses (both aggregated and at the individual-observer level) followed the shape of the presented distribution, revealing that people indeed represent more than summary statistics. Model simulations demonstrated that performance reflected integrated information from the whole set rather than the sub-sampling of a few items. In Experiment 2 (N=17), observers were presented with symmetrical and skewed Gaussian distributions and demonstrated that they could also explicitly represent distributional skewness. These results show that after only a brief exposure to a color set, the visual system builds detailed representations of feature distributions that observers can explicitly access. This necessitates a fundamental rethinking of how ensembles are processed. We suggest that distribution representation, not summary statistics, is the default way to consciously represent object groups. The summary statistics can be simple derivatives of these distribution representations rather than the core units of ensemble perception. Our finding helps to explain people‘s impression of having a rich perceptual experience despite severe attentional and working memory limitations.
Acknowledgements: Supported by grant #228366-051 from the Icelandic Research Fund (AK, AGA), NIH R01-EY025648 (JG), and NSF 1848939 (JG)