B.2.7.3.1. Master Activation Map model

To begin with, the Master Activation Map model (MAM, Michael et al., 2006) proposed a hybrid cognitive model of visual selective attention, comprising an independent inhibitory process. Michael and collaborators (2006; Michael et al., 2001b) presented the case of a patient with a lesion of the ventral pre-motor cortex (right inferior frontal gyrus, rIFG). This patient performed more or less normally in simple visual search tasks that did not demand to resist interference, for instance in the no-distractor conditions of several visual search experiments similar to the present ones. This suggested the authors that orienting attention processes were effective in this patient. On the contrary, this latter was much more perturbed than control participants by the presence of a salient distractor, even when knowing that this distractor was completely irrelevant. He also had difficulties in several tasks usually associated with "inhibitory processes" (e.g. Go/No-Go, Stroop test, Trail-Making Test B...).

Michael and colleagues (2006) proposed the MAM model on the basis of these results. The most important features of this model are a salience map and a relevance map, integrated in a master activation map, and an independent inhibitory process. One the one hand, the visual processing is considered to be split in two, according to the spatial or non-spatial nature of the visual information (Mishkin, Ungerleider, & Macko, 1983). As in older classical models (Koch & Ullman, 1985; Theeuwes, 1993; Treisman & Souther, 1985), all the basic non-spatial dimensions are processed in specific "feature maps", whose activities are thereafter combined in a salience map. On the other hand, the model postulated a "relevance map", representing the knowledge potentially available about the target (e.g., location, features, etc.). The activities from these two maps are then supposed to integrate in a "master activation map". Through interactions with a spatial map, this MAM controls the attentional orienting. Inhibition of return is considered a part of orienting processes. In addition, the model postulated an independent process of inhibition, in charge of avoiding orienting attention toward salient but known-to-be-irrelevant items. Indeed, such items can capture attention, which would hamper performance (Theeuwes, 1992; Yantis & Jonides, 1984; Yantis & Egeth, 1999). In these circumstances, the model predicted that the irrelevant activations in the master activation map, induced via salience map, would undergo inhibition. This process of inhibition is explicitly assimilated to the one proposed by Watson and Humphreys (1997) to account for visual marking phenomenon. The MAM model posited that inhibitory processes were independent from orienting ones, essentially on the basis of the patient R.J.'s performances (Michael et al., 2001b, 2006), described above. In the experimental psychology domain, Watson and Humphreys (1997) showed that the hypothesized inhibitory process was affected by the concurrent achievement of an attention demanding load task. However, the authors argued that this process was not itself resource-limited, because the preview benefit did not vary as a function of the number of items to inhibit. They proposed that the limitations were linked to the goal states of the participants. Indeed, the inhibitory process depends on the maintaining of the appropriate task goals. Yet the number of goals that could be maintained simultaneously active is limited (Kimberg & Farah, 2003). Watson and Humphreys (1997) suggested that these goals could be mutually inhibitory. These ideas were strongly reminiscent of the biased competition hypothesis (Desimone & Duncan, 1995), and very similar to the propositions of Sala and Courtney (2008) or Duncan et al. (1996; 2008). In the MAM, both inhibitory and relevance processes are dependent on the activations of participant's goals, and thus on executive functions and WM.

The present results could be interpreted within the MAM model framework (Michael et al., 2006). Indeed, both endogenous orienting and inhibition were thought to depend on WM, through the activation of the task goals. The model thus predicted that both processes could be disrupted by loading working memory, as was done in these experiments, except if they called for a different quantity of resources. Since both processes could depend on the same WM resources, their impairment could probably be correlated amongst people. Loading executive WM should hamper both processes in a similar way in participants, according to their working memory capacities. In this perspective, the WM-induced increase of AC, observed in the present experiments, should be attributed to a sub-optimal inhibition of the irrelevant distractor. This would be caused by a deactivation of the goal to inhibit this distractor, following the reasoning of Watson and Humphreys (1997; Sala and Courtney, 2008; Duncan et al., 1996; 2008).

Some neuroimaging studies supported the hypothesis of an inhibitory process, postulated by the MAM. For instance, the rIFG was activated in suppressing irrelevant targets in Go-No Go task or Wisconsin Card Sorting Test (Konishi, Nakajima, Uchida, Kikyo, Kameyama, & Miyashita, 1999) and antisaccade tasks (Chikazoe, Konishi, Asari, Jimura, & Miyashita, 2007). Aron, Russel & Poldrack (2004) proposed a review of the literature. They claimed that neuroimaging data are compatible with an implication of the right inferior frontal cortex in inhibitory processes, since it was often activated in these processes, among other prefrontal areas. They further claimed that the human neuropsychology supported more directly this implication, since damage to this structure induced deficits in various kinds of inhibition, particularly in stop-signal inhibition and task-switching.