Distinct neural networks for target feature versus dimension changes in visual search, as revealed by EEG and fMRI Stefanie I. Becker a, ⁎, Anna Grubert b , Paul E. Dux a a School of Psychology, The University of Queensland, Australia b Department of Psychological Sciences, Birkbeck College, University of London, UK abstract article info Article history: Accepted 31 August 2014 Available online 6 September 2014 Keywords: Attention Priming of pop-out Dimension weighting EEG fMRI In visual search, responses are slowed, from one trial to the next, both when the target dimension changes (e.g., from a color target to a size target) and when the target feature changes (e.g., from a red target to a green target) relative to being repeated across trials. The present study examined whether such feature and dimension switch costs can be attributed to the same underlying mechanism(s). Contrary to this contention, an EEG study showed that feature changes influenced visual selection of the target (i.e., delayed N2pc onset), whereas dimension changes influenced the later process of response selection (i.e., delayed s-LRP onset). An fMRI study provided convergent evidence for the two-system view: Compared with repetitions, feature changes led to increased activation in the occipital cortex, and superior and inferior parietal lobules, which have been implicated in spatial attention. By contrast, dimension changes led to activation of a fronto-posterior network that is primarily linked with response selection (i.e., pre-motor cortex, supplementary motor area and frontal areas). Taken together, the results suggest that feature and dimension switch costs are based on different processes. Specifically, whereas target feature changes delay attention shifts to the target, target dimension changes interfere with later response selection operations. Crown Copyright © 2014 Published by Elsevier Inc. All rights reserved. Introduction In visual search, changing the target feature across trials (e.g., from a red to a green target) or the target dimension (e.g., from a size target to a color target) typically leads to performance impairments (e.g., increased reaction time) relative to pairs of trials where target defining properties are repeated. For instance, in search for an “odd-one-out”, changing the target feature from a smaller item to a larger item or from a red item to a green item slows responses to the target (e.g., Maljkovic and Nakayama, 1994; Becker, 2008a,b,c, 2010a,b). Similar switch costs occur when the stimulus dimension of the target is changed, for example, from a target differing in size to a target differing in color (e.g., a large to a red target; Müller et al., 1995). Originally, both types of intertrial effects were attributed to an attentional weighting mechanism. According to this attentional biasing account, selection of the target feature or the target dimension primes or biases attention to select the same feature or dimension on the next trial, by increasing the weights (or gains) of cor- responding feature-specific and dimension-specific maps, respectively (e.g., Maljkovic and Nakayama, 1994; Müller et al., 1995). In contrast to this early attention view of target switch costs, it has been proposed that they may reflect later processes involved in response selection (e.g., Cohen and Magen, 1999). According to a response biasing account, repeating the target could bias response selection mechanisms to repeat the response from the last trial as well, even when repetitions of the target feature and the response are statistically independent. For example, in a target present/absent search task, changing the target dimension (e.g., from a color to a size target) would automatically create a bias to change the response as well, which would lead to switch costs when the target changes but the instructed response is the same as in the previous trial (e.g., a target present response; Becker, 2008a, 2010a; Cohen and Magen, 1999; Mortier et al., 2005; Pollmann et al., 2006; Yashar and Lamy, 2011). In line with the response biasing account, reaction times are often faster when both the target and response repeat than when only either the target or the response repeats (e.g., Hillstrom, 2000; Huang and Pashler, 2005; Meeter and Olivers, 2006; Müller and Krummancher, 2006; Töllner et al., 2008; Yashar and Lamy, 2011). Yet, at least with respect to feature changes, changing the response requirements does not completely eliminate target switch costs, indicating that they cannot be fully accounted for by response biasing (e.g., Hillstrom, 2000; Yashar and Lamy, 2011). As a consequence, most researchers to date advocate a dual stage account, which holds that target changes can incur costs both at an early attentional level and at a later, response-selection level (e.g., Meeter and Olivers, 2006; Mortier et al., 2005; Rangelov, Müller and Zehetleitner, 2011; Yashar and Lamy, 2011; Zeheitleitner et al., 2012). NeuroImage 102 (2014) 798–808 ⁎ Corresponding author at: School of Psychology, The University of Queensland, McElwain Building, St Lucia QLD 4072, Queensland, Australia. E-mail address: s.becker@psy.uq.edu.au (S.I. Becker). http://dx.doi.org/10.1016/j.neuroimage.2014.08.058 1053-8119/Crown Copyright © 2014 Published by Elsevier Inc. All rights reserved. Contents lists available at ScienceDirect NeuroImage journal homepage: www.elsevier.com/locate/ynimg