Purchase this article with an account.
Michael Miuccio, Gregory Zelinsky, Joseph Schmidt; Evidence from contralateral delay activity that proto-objects are a good approximation of real-world set size. Journal of Vision 2021;21(9):1880. doi: https://doi.org/10.1167/jov.21.9.1880.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Decades of memory and visual search work demonstrate that increasing set size tends to decrease performance. Set size is easy to determine with simple stimuli, but how do you count the number of items in the real-world? Is a bowl of fruit one item, or many? Measures of clutter have been proposed as an alternative to counting items (Neider & Zelinsky, 2011). Clutter can be estimated by segmenting images into proto-objects (Yu et al., 2014), which predict human clutter rankings (Yu et al., 2013) and fixation density (Chen & Zelinsky, 2019). Contralateral delay activity (CDA; Vogel & Machizawa, 2004) measures the number of items maintained in visual working memory (Luria et al., 2016). Using data from Schmidt et al. (2014, Experiment 2), the current work assessed if the number of proto-objects in real-world stimuli, presented as target cues during a visual search task, affect CDA and later eye-movement metrics of visual search performance. Target cues were previewed for 400 ms, followed by a 1000 ms ISI in which CDA was assessed. The subsequent search array contained one target and three distractors from non-target basic level categories. Trials were separated into conditions based on whether targets had a low or high number of proto-objects. CDA in response to the target cue was greater (consistent with storing more items) when targets contained more proto-objects (p=.007). Additionally, search performance showed that targets with more proto-objects elicited poorer guidance (initial saccades to the target), longer RTs, and longer target dwell times (all p≤.001). We conclude that the number of target-related proto-objects predicts the resulting visual working memory load and later search performance in the context of real-world objects. Future work is needed to identify optimal methods of proto-object segmentation for set size prediction, and to test generalization to paradigms beyond search.
This PDF is available to Subscribers Only