B.2.3. A biased-competition account of working memory

Both endogenous orienting and resisting interference depend on WM, especially on executive WM. Claiming that they both require attentional or WM resources is insufficient, however, as the underlying mechanistic principles are not explored. Recent theoretical developments allow to address these principles. Sala and Courtney (2007; Courtney et al., 2007) presented a biased-competition model of WM, congruent with the models of Miller and Cohen (2001) and Duncan (2001). They postulated that information held in WM is subject to a competition, similar to the one claimed to occur in the perceptual systems (Desimone & Duncan, 1995), that could be biased endogenously or exogenously. Indeed, Duncan (1996) argued that competition was a very widespread cognitive phenomenon, common to many cognitive systems and based notably to the mutual lateral inhibitions in the neuronal connectivity (see also Miller & Cohen, 2001). The information held in WM could be of various sorts: it could concern objects, locations, but also task rules, and it could be contextual or motivational, and so forth.

Duncan and colleagues (Duncan, Emslie, Williams, Johnson, & Freer, 1996; Duncan, Parr, Woolgar, Thompson, Bright, Cox, Bishop, Nimmo-Smith, 2008) addressed the issue of dual-tasking in a similar perspective. They postulated that, when achieving a new or unfamiliar task, one has to construct, and then keep active, a representation of "relevant facts, rules, and requirements [... as] an effective control structure or mental program", which they call a task model (Duncan et al., 2008, p.133). In this task model, the goals of the task are represented and maintained active. Consistently with the model of Sala and Courtney (2007), these goals compete against each other. This competition could lead to the de-activation or neglect of some of these goals (Duncan et al., 1996; 2008), since the representing capacity in WM is limited (Courtney et al., 2007). Thus, increasing the number of the goals concurrently maintained active (in WM) would naturally lead to decrease their respective activations. Indeed, Duncan et al. (2008) evidenced that a goal becomes more prone to be neglected as the task model becomes more complex. Obviously, increasing the complexity of the task models increased the number of goals to keep active, and thus made the competition fiercer. Conversely, Bourke and Duncan (2005) showed that the complexity of the template describing the target to be sought increased interference with a concurrent dissimilar task (story memory). Koechlin, Basso, Pietrini, Panzer, & Graffman (1999) showed that the fronto-polar pre-frontal cortex could be specifically involved in maintaining a goal in WM while simultaneously achieving secondary goals. This biased-competition perspective, applied to goals and task rules, helps understanding the very reasons why WM or attentional resource are limited. This competitive account might also explain why loading executive WM could hamper attentional selection, only when endogenously controlled. Indeed, only endogenous, not exogenous, control should depend on a goal activation. This might also explain why an executive WM load would be more prone to disrupt a visual search task than a mnesic WM load (e.g. Han & Kim, 2004). Indeed, the complexity of the task model is usually greater for an executive WM task than for mnesic WM one. On these bases, one could expect a high executive WM load to disrupt both endogenous orienting and resistance to interference.