Contingencies and attentional capture: the importance of matching stimulus informativeness in the item-specific proportion congruent task

The proportion congruent effect is the observation that congruency effects are smaller when the proportion of incongruent stimuli is higher. The conflict adaptation account argues that this effect is due to a shift of attention away from the source of conflict. In contrast, the contingency account proposes that participants learn to predict the likely response on the basis of the distracter, and this produces a proportion congruent effect incidentally. However, some have argued that conflict adaptation can be observed in the restricted scenarios in which the mostly incongruent stimuli are not strongly predictive of the correct response. This opinion article argues that comparing predictive to non-predictive stimuli might be problematic. Some learning research would suggest that attention to the distracter should vary, but for an entirely different reason than that proposed by conflict adaptation theory: contingent stimuli attract attention. In the Stroop paradigm (Stroop, 1935), participants are tasked with the goal to ignore a distracting color word and to respond to the print color it is presented in. Only partially successful at doing so, participants respond slower and less accurately to incongruent stimuli (e.g., the word “blue” printed in red) than to congruent stimuli (e.g., “blue” in blue). Similar congruency effects are observed in the Simon (Simon and Rudell, 1967), flanker (Eriksen and Eriksen, 1974), picture-word (Rosinski et al., 1975), and various other comparable tasks. Such congruency effects are reduced if the relative number of incongruent trials is increased (Lowe and Mitterer, 1982). This proportion congruent (PC) effect is most commonly interpreted as evidence for the conflict adaptation account (e.g., Botvinick et al., 2001). This account argues that detection of conflict results in a decrease of attention to the source of conflict (e.g., the word in a Stroop task). Because conflict is more frequent in a mostly incongruent condition, attention to the word is particularly low. The word therefore has little impact on performance, and the congruency effect is resultantly small. Though seemingly intuitive, the conflict adaptation account does face some important challenges. For instance, consider the item-specific PC effect (Jacoby et al., 2003). In this variant of the PC task, some words are mostly congruent and others are mostly incongruent. These two item types are intermixed into one procedure, but there is nevertheless a smaller effect for mostly incongruent items. This might be described in terms of item-specific adaptations of attention (e.g., Blais et al., 2007), though this requires the unintuitive assumption that attention to the word is determined by the identity of the word (which must, of course, first be identified; but see Verguts and Notebaert, 2008). Alternatively, the contingency account proposes that the entire item-specific PC effect is explained by the learning of contingent relationships between distracting words and responses (Schmidt and Besner, 2008; Schmidt, 2013b; for a review, see Schmidt, 2013a). For mostly congruent stimuli, each word is presented most often with the congruent color. This means that, for instance, the word “blue” is predictive of a blue response. Congruent trials thus benefit from this prediction, and the congruency effect increases. For mostly incongruent stimuli, the reverse is true: each word is presented most often with a specific incongruent response. Thus, “red” might be predictive of a yellow response. Incongruent trials thus benefit, and the congruency effect decreases. Some evidence argues compellingly in favor of the contingency account (Schmidt and Besner, 2008; Atalay and Misirlisoy, 2012; Grandjean et al., 2013; Schmidt, 2013b). For instance, Schmidt (2013b) presents a dissociation procedure in which contingency learning and conflict adaptation could be separately assessed. Specifically, it was possible to compare sets of incongruent trials that were: (a) equivalent in PC (mostly incongruent) but that varied in contingency (high vs. low), or (b) equivalent in contingency (low contingency) but that varied in PC (mostly congruent vs. mostly incongruent). Thus, the former set allows an assessment of contingency learning in the absence of conflict adaptation, and the latter set allows an assessment of conflict adaptation in the absence of a contingency bias. These comparisons revealed a very strong contingency effect, with no evidence for conflict adaptation. The (item-specific) PC effect thus might have nothing to do with conflict adaptation at all. Some neuropsychological data even argues that the area claimed to be involved in (item-specific) conflict adaptation (viz., the anterior cingulate cortex; see Blais and