Research indexPreprint
Visual Statistical Learning
Pellegrini, Ansorge
30 Jun 2026Open access
NobleIDark:/48914/w1/20261105/9741A6E0DOI10.17605/osf.io/ht9s8
Visual Statistical Learning
Pellegrini, Ansorge · 2026
NobleIDark:/48914/w1/20261105/9741A6E0DOI10.17605/osf.io/ht9s8
Abstract
Visual statistical learning (VSL), i.e., the extraction of visual regularities from the environment, plays an important role in guiding attention. Recently it has been shown that VSL also comes to play in visual search tasks, a context characterized by competition for visual selection. Statistical learning of likely distractor locations can bias attentional selection and probably enhances inhibition of capture by salient distractors. Open questions concern the role of value-driven learning, the true nature of the interference by the salient distractors as due to capture versus dwelling, and the exact role of attention on the acquisition or learning of the regularities. Regarding the possibility of value-driven learning, many studies used visual search tasks with relevant targets and irrelevant but salient distractors and found that interference by the distractors was stronger for distractors at less likely than for distractors at more likely locations. In this situation, learning could be driven by value because, for example, successful suppression of the distractors improves task performance, and improved task performance is a value and, thus, rewarding in itself (e.g., Anderson, 2016). Duncan and Theeuwes (2020) claim to have observed statistical learning in the absence of such task-relevance. They used their additional singleton protocol (Theeuwes, 1991, 1992) with one salient distractor to show that participants can learn (to suppress) the position in which a salient color singleton distractor appeared with a high probability, even though the singleton distractor was not relevant for the task during the learning phase as no searching was required during learning, just the discrimination of the overall stimulus layout as diamond-shaped versus circular. However, if the singleton distractor attracted attention in a bottom-up way also during learning, then the shape of the singleton distractor was either congruent or incongruent to the overall layout. This was due to the fact that the singleton distractor could also be a circle or a diamond, just as the shape of the overall arrangement, and varied orthogonally to the shape of the overall arrangement. As a consequence, the distractor was response-congruent in half of all trials (i.e., a diamond-shaped singleton distractor was presented in a diamond-shaped arrangement or a circular singleton distractor was presented in a circular stimulus arrangement) and it was response-incongruent in the other half of all trials (i.e., a diamond-shaped singleton distractor was presented in a circular arrangement or a circular singleton distractor was presented in a diamond-shaped arrangement). In the current study, we, therefore, aim at replicating and extending the findings of Duncan and Theeuwes (2020), and will address the questions above left open by the original study. The first question that we want to address is that of the task-irrelevance of the singleton distractors in the learning phase. We will test if the singleton distractors lead to a congruence effect – with faster and/or more accurate responses under congruent than incongruent conditions – because a congruence effect would indicate some type of task relevance, namely response-relevance, of the seemingly task-irrelevant distractors. To note, this does not necessarily call into question the entire reasoning of Duncan and Theuwese as a congruence effect may also be expected based on the fact that distractors first captured attention in a bottom-up way as a necessary precondition for the task-dependent response-congruence effect to show up. However, a congruence effect would definitely suggest to run further control experiments to find out if it is not the other way round and the response-relevance of the singletons is a necessary pre-condition for statistical learning, which would mean that the jury would still be out for the final verdict on the role of value in the acquisition of statistical regularities. The second question left open by prior studies is that of the role of capture by the singleton distractor versus dwelling on the singleton distractor (for the distinction, see, e.g., Becker et al., 2023; Wolfe & Horowitz, 2017). While capture denotes the attraction of attention by a stimulus (e.g., a singleton distractor), dwelling denotes the time that attention lingers at a stimulus before then next stimulus can be selected. In research on statistical learning, the contributions of the two processes to more interference for distractors at rare locations than for distractors at more likely locations has not been investigated, and, yet, it could be of theoretical relevance. Past studies have not only demonstrated capture by but also more dwelling on less expected or more surprising stimuli (e.g., Fuchs et al., 2026; Horstmann & Herwig, 2015). Thus, unless this is investigated, it is an open question if less capture and, thus, suppression of the likely singleton distractors and/or more dwelling on the less likely singleton distractors account for the difference in interference between these two conditions in VSL of likely distractor locations. To investigate this open question, we will collect eye movement data to quantify attentional capture and dwelling during the learning and the transfer phase. Specifically, fixation probability and fixation duration are used as estimates of capture versus dwell time on the additional singleton. Fixation probability and fixation duration will serve as separate indices of attentional capture and dwelling, respectively. We will examine the extent to which each measure predicts behavioral interference by regressing the difference in interference between rare and frequent distractors on the corresponding differences in fixation probability and fixation duration. Finally, it has been argued that attention on the learned stimuli, here, the singleton distractors, is a necessary prerequisite for learning (Theeuwes et al., 2022), and we will put this hypothesis to test, too. On the one hand, we will test whether the magnitude of the (possible) congruency effect correlates with the magnitude of learned suppression in the test phase. Such a relationship would suggest that VSL is stronger in observers for whom the additional singleton reliably captured attention during learning, be this now due to bottom-up capture or task-dependent capture (see above). On the other hand, we can also use the probabilities of fixations on distractors during learning as a proxy for the capture of attention by these distractors. Again, participants that show a lot of such distractor fixations during learning should show stronger probability effects on interference by the singleton distractors during transfer. References: Anderson, B. A. (2016). The attention habit: How reward learning shapes attentional selection. Annals of the New York Academy of Sciences, 1369(1), 24–39. https://doi.org/10.1111/nyas.12957 Becker, S. I., Grubert, A., Horstmann, G., & Ansorge, U. (2023). Which processes dominate visual search: Bottom-up feature contrast, top-down tuning or trial history?. Cognition, 236, Article 105420. Duncan, D., & Theeuwes, J. (2020). Statistical learning in the absence of explicit top-down attention. Cortex, 131, 54–65. https://doi.org/10.1016/j.cortex.2020.07.006 Fuchs-Leitner, I., Horstmann, G., & Ansorge, U. (2026). Wouldn’t it be nice to not fall for it twice? Prior experience does not abolish the impact of expectancy violations on attention capture. Vision, 10(2), Article 32. Horstmann, G., & Herwig, A. (2015). Surprise attracts the eyes and binds the gaze. Psychonomic Bulletin & Review, 22(3), 743-749. Theeuwes, J. (1991). Cross-dimensional perceptual selectivity. Perception & Psychophysics, 50(2), 184–193. https://doi.org/10.3758/BF03212219 Theeuwes, J. (1992). Perceptual selectivity for color and form. Perception & Psychophysics, 51(6), 599–606. https://doi.org/10.3758/BF03211656 Theeuwes, J., Bogaerts, L., & van Moorselaar, D. (2022). What to expect where and when: How statistical learning drives visual selection. Trends in Cognitive Sciences, 26(10), 860-872. Wolfe, J. M., & Horowitz, T. S. (2017). Five factors that guide attention in visual search. Nature Human Behaviour, 1(3), Article 0058.