Perceptual grouping and averaging studies.
- Palmer, S. E. (2002). Perceptual Grouping: It is Later Than You Think. Current Directions in Psychological Science, 11(3), article
- Argues "grouping" might occur at the perception level (rather than a retinal processing level) based on the findings observed in relation to lightness constantcy, illusory contours ve binocular depth.
- Whitney, D., & Yamanashi Leib, A. (2018). Ensemble Perception. Annual Review of Psychology, article
- Very good review article about ensemble encoding
- Visual system extract mean/average representation (and can do it <200 ms). Adapting to mean even leads to adaptation after-effects.
- Applies to high-level features like facial expressions, and biological motion too.
- Made me ask "what are the boundaries of the "similarity" (grouping)?"
- Ariely, D. (2001). Seeing Sets: Representation by Statistical Properties. Psychological Science, 12(2), article (A)
- Introduces concepts, like set property, argues that sets of items can be distinctly represented in of itself, separate than the sum of its parts, something like a unique entity.
- But when asked about the membership, people perform badly. They cannot reliably and accurately judge whether the size of the post-cued test spot (stimulus) was included in the previously seen set.
- Tested 500 ms, asked for yes/no membership questions and 2AFC, which one belonged to the set paradigms, cool techniques.
- Argues that this should also be observed in non-visual domains.
- Chetverikov, A., Campana, G., & Kristjansson, A. (2016). Building ensemble representations: How the shape of preceding distractor distributions affects visual search. Cognition, article
- Observers can build ensemble representations with surprising precision. They represent the distractor and target mean distributions (along with SD, and second-order statistics like kurotosis, skewness) in a visual search task.
- Yamanashi Leib, A., Chang, K., Xia, Y., Peng, A., & Whitney, D. (2020). Fleeting impressions of economic value via summary statistical representations. Journal of Experimental Psychology: General, article
- Are ensemble representations formed for other complex but commonly encountered visual stimuli that require abstraction?
- Authors used real retail prices of item images. Observers saw the items in displays of 4, 2, or 1 product, in equal proportion.
- Observers were able to extract an ensemble percept of product value!
- I liked this paper because this might suggest that ensemble perception operates beyond a basic visual feature analysis.
- Interesting! "Ensemble value would confirm that ensemble perception is not only an expedient visual mechanism for identifying redundant concrete features but may also provide a fast shortcut to abstract perceptual impressions".
- Leib, A. Y., Kosovicheva, A., & Whitney, D. (2016). Fast ensemble representations for abstract visual impressions. Nature Communications, article
- Participants rated "liveliness" of objects (life-likeness is a shared visual percept?)
- Big question: perception of abstract visual experiences also operate under ensemble perception mechanism?
- Observers can perceive the average lifelikeness of groups of objects in a fraction of a second.
- Corbett, J. E., & Oriet, C. (2011). The whole is indeed more than the sum of its parts: Perceptual averaging in the absence of individual item representation. Acta Psychologica, article
- Similar to Ariely's 2001 study, they showed that member identification and mean extraction is two different mechanism, observers can reliably represent the mean of a set while inaccurate in deciding if a test stimuli was a member of a previously seen set.
- Marchant, A. P., Simons, D. J., & de Fockert, J. W. (2013). Ensemble representations: Effects of set size and item heterogeneity on average size perception. Acta Psychologica, article
- Estimating the mean size of a set gets compromised as the set size and heterogeneity increase (perhaps ensemble processing is not a parallel process?)
- de Gardelle, V., & Summerfield, C. (2011). Robust averaging during perceptual judgment. Proceedings of the National Academy of Sciences, article
- Robust averaging, again. But compromised by 1) reducing the signal strength and 2) decreasing its reliability.
- And, observers disregard the extreme data points when averaging (when evidence is variable).
- Cavanagh, P. (2001). Seeing the forest but not the trees. Nature Neuroscience, article
- Again similar findings that shows mean representation is preserved (despite the inability to access the individual items).
- If the grating orientation differs by more than 45 degree, than linear averaging is disrupted.
- "In the case of 0 and 90 degrees, for instance, there is no average orientation, as it could just as well be 135 degrees as 45."
- Author also mention about the possible function of ensemble coding: rapid identification of deviant items with enormous savings in computational complexity.
- Summarizes the Parkes 2011 paper
- Jia, L., Cheng, M., Lu, J., Wu, Y., & Wang, J. (2022). Context consistency improves ensemble perception of facial expressions. Psychonomic Bulletin & Review. article
- Changes in the background (which is task-irrelevant) change facial expression judgment discrimination. Consistent background (between encoding and test phase) enhances discrimination.
- Norman, L. J., Heywood, C. A., & Kentridge, R. W. (2015). Direct encoding of orientation variance in the visual system. Journal of Vision, A
- Adaptation to high variance leads to seeing low variance at the subsequently presented medium variance texture -classical adaptation after-effects!
- This is interesting because it tells us that encoding of variance (of a display, like 10x10 grid of gabor gratings) is similar to encoding of size and contrast.
- Frame of reference? retinotopic, rather than spatiotopic, they showed (exp3), and possibly involves mid-level areas like V3 and V4 maybe even V2 (exp4).
- Hubert-Wallander, B., & Boynton, G. M. (2015). Not all summary statistics are made equal: Evidence from extracting summaries across time. Journal of Vision, A
- How does Summary representation work with different feature dimensions? Investigation of temporal position in the information stream.
- Instead of typically using the simultaneous presentation of displays, they tested the weight of presentation order in averaging.
- "Do all items or parts of the information stream contribute equally to the perceived mean? Or do early items or late items contribute more heavily?"
- they fit the response to the weighted average model (Juni et al., 2012).
- Average location -> primacy
- Average motion direction -> recency
- Average facial expression -> recency
- Florey, J., Clifford, C. W. G., Dakin, S., & Mareschal, I. (2016). Spatial limitations in averaging social cues. Scientific Reports, A
- When taking average of (of faces' head rotation & eye gaze) what is the N of subsampling? And, cartoon-like faces (like circle and a cone) also results in the same way as processing realistic faces?
- yes, no difference between social vs non-social averaging.
- They mention about 'equivalent noise which I didn't get it fully.
- Averaving is limited by two factors: external ve internal noise. External noise relates to subsampling and the increase in effective sample results in diminishes in perceptual performance. Internal noise: "any uncertainty in processing the stimuli in the brain".
- Need to read again sometime later.
- Alvarez, G. A., & Oliva, A. (2009). Spatial ensemble statistics are efficient codes that can be represented with reduced attention. Proceedings of the National Academy of Sciences, A
- Visual system loses clarity in representation for spaces are not attended. And, attention (and/or grouping?) reimburses that loss of clarity by pooling signals together which results in "efficient code", a somewhat compressed file of redundant signals.
- Participants engaged in dual task and results showed that structural changes in the background made them more sensitive (measured by d-prime) in detecting changes in local features.
- Efficient codes can be represented under reduced attention (reduced because participants were doing dual task, one involves in tracking foreground objects).
- Burr, D., & Ross, J. (2008). A Visual Sense of Number. Current Biology, article
- People see less of objects after being adapted to see many more objects. The percept looks dissimilar to the adaptor.
- Pretty neat finding.
- Chetverikov, A., Campana, G., & Kristjansson, A. (2017). Representing Color Ensembles. Psychological Science, article
- RTs following exposure to uniform distributions were constant when test targets were within the range of the learned distribution. In contrast, following exposure to Gaussian distributions, there was a monotonic decrease in RTs that paralleled the monotonic decrease in the probability density of that distribution.
- Observers were exposed to 1) uniform and 2) gaussian distribution visual displays, and results show that performance was better if items (colors) were distributed using a gaussian distribution (gaus > uniform, authors relate this to ecological validity, as nature provides more variance).
- shape of
distributionsare also encoded
not just mean and variance