Neural mechanisms of high-level cognitive processes in task preparation and implementation

Resumen

To adjust our behavior based in goals is a core human ability that allows us to rapidly adapt to new demands or new environments. Cognitive control, that is, the mechanism that regulates our thought and actions upon internal representations of our goals (Norman & Shallice, 1980), is thought to underlie this ability. Due to its central role in our cognitive activity, control is involved in a plethora of phenomena, and therefore it can be analyzed from different perspectives. Specifically, this thesis takes advantage of a temporal classification: reactive versus prospective control. While reactive control refers to the immediate deployment of control upon conflict detection, prospective processes allow the anticipation and corresponding adjustment to forthcoming demands (Braver, 2012). The main aim of this thesis is to advance our knowledge about proactive control. The Experimental Series I of the present thesis aimed at assessing the relationship between preparatory processes and consciousness. While some studies point to an unconscious bias in the selection of different action schemas (e.g. Reuss et al., 2011), other studies suggest the need for a conscious experience to effectively exert proactive control (Kiefer et al., 2012; Kiefer & Martens, 2010). To answer this question, we employed a cuing paradigm together with visibility manipulations to alter the conscious experience of cues. Results show a clear effect of subliminal cues on task setting, even when visibility of cues was extremely restricted. Crucially, this unconscious effect depended on conscious intentions, insofar a correct executive setting had to be configured consciously for these unconscious influences to take place. Last, this experimental series revealed a distinct pattern of results when using an objective threshold of consciousness compared to a subjective one, suggesting that these two measures might be indexing different phenomena (Snodgrass & Shevrin, 2006). Another aim of this thesis was to evaluate the advantages and disadvantages of using fMRI in the quest for the neural correlates of cognitive processes. This was crucial given the predominance of fMRI evidence in this thesis. Thus, we reviewed the most relevant considerations regarding fMRI use prior to carrying out our fMRI studies. In Experiment I of the thesis we studied the neural correlated of preparatory processes. Preparation, defined as the adjust to task rules prior to the onset of a target (Rogers & Monsell, 1995), can be elicited by showing cues that signal to-be-presented demands. At the neural level, these cues engage both control and selective processing brain regions. However, previous studies did not control for interference between tasks, which is known to confound preparatory effects. Thus, it remained unknown whether preparatory mechanisms are domain-specific or domain-general. In this experiment, we found that even in a context with low interference between tasks, preparatory cues can elicit anticipatory activations both in control and selective processing regions. More specifically, some of these selective processing regions were later on engaged by targets, revealing the proactive nature of preparatory activations. Last, in the Experiment II of the present thesis we assessed the neural indexes of preparation in a different context. Most of preparation studies employ paradigms that capitalize on the alternation between two task sets. However, control demands are thought to be larger in novel situations (Norman & Shallice, 1980). In these contexts, a new task set has to be created from scratch rather than merely being reactivated (Cole et al., 2013). Despite a profound involvement of control in this type of situation, the correlates of preparatory mechanisms in complex, novel task sets remain unclear. In this experiment, we thus assessed the neural indexes underlying the implementation of verbal instructions, insofar they require the formation and execution of novel task sets (Hartstra et al., 2011). Our results reveal, first, an initial effect of instructions that engaged frontal regions, and, second, an involvement of frontal together with parietal areas during active preparation for implementation. Moreover, our results are the first to show how different brain regions carry information in their patterns about the category of the instruction seconds before its implementation. These regions include both control and selective category regions during the preparatory period. Crucially, the quality of these representations had a direct influence on subsequent behavior, illustrating the important role of proactive control in novel contexts. All of our experiments highlight the predominance of top-down influences in preparatory mechanisms. However, bottom-up elements also play a role in control demanding context. The biggest piece of evidence in this line comes from Experimental Series 1, which revealed an unconscious bias in task set selection. This result is coherent with previous studies showing unconscious influences in high-level processes (van Gaal et al., 2012). It therefore shows that at least some proactive control mechanisms can be altered by bottom-up information. Subsequently, in conjunction with previous similar studies (e.g. De Pisapia et al., 2011), they suggest the dissociation between proactive control and consciousness (Hommel, 2017). Experiments I and II also show how some stimuli, such as instructions, can elicit control processes automatically (Liefooghe, Wenke, & De Houwer, 2012). The evolutive relevance of fast learning (Cole et al., 2013) can underlie this automatic effect of instructions. Despite the relative automaticity described before, our data show an overwhelming predominance of top-down effects in proactive control. First, in Experimental Series I, the unconscious effect is only found when the executive setting is configured properly according to conscious expectations (Kiefer, 2012). Therefore, subliminal perception can affect but not initiate control processes (van Gaal et al., 2012). Moreover, in Experiment I, conscious expectations regarding future demands were shown to modulate brain activity during preparation. In the proactive control framework (Braver, 2012), the reported category specific activations can be understood as the outcome of a top-down influence, originating in control regions, on incoming information. Last, Experiment II reveals how actively preparing to implement novel task sets involves a large set of control areas. Moreover, category specific information could be decoded from selective processing regions seconds before target onset, which shows again a bias in the processing of incoming information based on internal goals. Altogether, our results suggest that proactive cognitive control sets up our information-processing system in a top-down manner to allow some extent of automaticity (Dehaene & Naccache, 2001; Kiefer, 2012; Kiefer & Martens, 2010). This, in turn, would make our control systems more efficient by reducing costs associated to maintained monitoring, and therefore, optimizing the consecution of our goals (Kiefer, 2012). In sum, the present thesis reveals a dynamic relationship between bottom-up and top-down processes. We interpret this relationship within the predictive coding framework (Friston, 2005), which suggests that our psychological experience is the result of an iterative interaction between bottom-up information and top-down predictions that bias this information to guide perception. Such mechanism would allow the evolutive development of a proactive control system (Buschman & Miller, 2014).