Purchase this article with an account.
Simona Buetti, Yujie Shao, Zoe Jing Xu, Alejandro Lleras; Re-examining the linear separability effect in visual search for oriented targets. Journal of Vision 2020;20(11):1244. doi: https://doi.org/10.1167/jov.20.11.1244.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The linear separability effect refers to a benefit in search performance observed in a feature-search task, where target and distractor features vary along a continuous feature dimension: search performance is best when there is a boundary in feature space that separates the features present in the distractor stimuli from the feature that defines the target. Search is qualitatively more difficult when there exists no such boundary separating the target from the distractor features. Here, we re-examined this effect in the context of a new procedure from Lleras et al. (2019) that quantifies the impact of distractor heterogeneity on search performance. First, one measures how well observers can find the target in homogeneous displays and then one uses the observed log-search efficiencies to predict how long it ought to take observers to find the target in heterogeneous displays in a separate experiment. Following this strategy, in Experiment 1, we evaluated how long it takes to find a tilted target oriented 20 degrees to the right of vertical, when surrounded by either distractors oriented 20 degrees to the left or 60 degrees to the right, using only distractor homogeneous displays. The log slopes of -20 and 60 degree distractors (312 and 498 ms/ln(ss)) were then used to predict performance in Experiment 2, where the target was presented in displays containing both types of distractors simultaneously. The number of distractors of each kind varied independently (2, 4, 6 or 8). Using Lleras et al.’s formula, total variance accounted for in Experiment 2 was 98.4%. Results suggest there is no “linear separability effect” in search for oriented targets. We concluded that search becomes harder due to stimulus heterogeneity, which reduces parallel processing efficiency by a factor of 1.62 (in log scale), just as is observed when intermixing images of real-life objects.
This PDF is available to Subscribers Only