This work has several theoretical implications and strengthens existing literature in several ways. Concerning the processing of faces without awareness in cortical blindness, it is not yet established how the brain processes perception of unaware emotional stimuli (Khalid, Finkbeiner, König, & Ansorge, 2013). Moreover, few studies have investigated the implication of frontal regions in response to visual stimuli in blindsight patients. Whereas previous studies mainly relied on case studies, we included a group of control participants to test where and how a difference in oscillatory responses would appear. We highlighted activations to any kind of visual stimuli in blindsight and specifically to stimulus category in blindsight and in healthy controls.
Concerning the processing of fearful and neutral faces at different levels of awareness in epileptic patients, our results provide evidence that a rapid processing route conveys visual information to the orbitofrontal cortex, but the amygdala is not
necessary in this processing. Indeed, it is likely that the visual information takes an alternate route in the processing of visual stimuli, but the amygdala is not the only plausible candidate in this processing (Pessoa & Adolphs, 2011).
Finally, the investigation of the interaction between selective attention and consciousness has revealed interesting findings. It has to be highlighted that studies investigating the processing of subliminal emotional faces generally focus on the ability of the participant to detect the emotion, which was not the case in our experiment.
Moreover, few studies have investigated whether top-down processing can occur without consciousness. Our results suggest that when faces are task-relevant but the emotion is irrelevant, a shift of attention cannot occur if the cue is not consciously perceived. Moreover, when the emotion is task-irrelevant but the face has to be processed, it seems that emotion is bypassed, presumably because of the high working memory load due to the task.
The first theoretical implication that will be discussed includes the importance of frontal brain areas in the processing of relevant stimuli, particularly in detecting what is consciously versus unconsciously perceived by the individuals. The results of studies 1 and 2 will be contrasted to the existing literature. Despite the importance given to the amygdala in the processing of relevant stimuli without awareness, another network including frontal areas will be proposed. Moreover, the role of frontal regions in the detection of errors, as highlighted by specific brain oscillations occurring in frontal scalp sites, will be outlined.
The second theoretical implication that will be proposed is the contribution to the literature linking attention and consciousness. Indeed, while a debate on the requirement of attention for conscious perception has received much attention, whether consciousness is necessary for attention has been less studied in paradigms involving selective attention. Moreover, the modulation of this link by emotion when the cue is task-relevant has not been investigated, results therefore provide electrophysiological evidence for this processing. The implication of these selective attentional processes in the study of cognitive biases in mood disorders, such as anxiety, will be discussed.
2.1 The role of the frontal cortex in the processing of relevance without awareness
In the theoretical and experimental parts of this work, we mentioned that face processing generally activates a network of brain regions, including frontal areas.
Specific responses to face stimuli in the OFC (Rolls et al., 2005) and the prefrontal cortex (Tsao, Schweers, Moeller, & Freiwald, 2008) of the macaque brain have been
brain (Vuilleumier et al., 2001). In this work, we showed implication of frontal areas even if the faces are not consciously perceived, as it is the case in blindsight or when using the backward masked paradigm in healthy participants. Slow wave oscillations observed in frontal sites in blindsight and frontal responses to visual stimuli in a group of epileptic patients corroborate the theory that frontal brain areas encode consciousness. While it has for a long time been assumed that the activity of frontal areas only reflects the involvement of executive functions operated consciously, a more recent view suggests that these “high-level” processes may occur without awareness (see for example Badgaiyan, 2000; Dehaene & Naccache, 2001). Results of study 1 suggests implication of frontal areas during the processing of faces and scrambled faces, therefore implying a top-down cognitive control process even if the patient is not aware of visual stimulation. This top-down control would act as a warning during the processing of trials leading to incorrect responses, which is reflected in the increase in the theta band specific to that condition. Some authors have suggested that conscious and unconscious executive control processes are different in nature (Jiang, Zhang, &
van Gaal, 2015). This difference may be prominent as a function of control level and control maintenance duration. This assumption is based on previous findings showing that unconscious processing is characterised by its short effect (Greenwald, Draine, &
Abrams, 1996). Moreover, according to the global neuronal workspace model (Dehaene
& Naccache, 2001), there are modular cerebral networks that are constantly activated.
These networks would process a myriad of events that are not conscious. To become conscious, information has to rely on a top-down attentional amplification of the neural population sustaining its processing, including a joint activity of long-range modules distributed in the brain. When a specific network is activated for a certain duration, the information conveyed by this network becomes accessible to consciousness, and the dorsolateral prefrontal cortex (DLPFC) has an important role to play in this processing (Dehaene, Charles, King, & Marti, 2014). The phenomenon of blindsight is compatible with this model, showing that many processes can be done unconsciously, reflecting behavioural performance independently of awareness. Therefore, conscious and unconscious executive control are different; while conscious processing relies on the strong co-activation of several “higher-level” structures, unconscious processing is brief and reflects activation of a quite reduced brain network (van Gaal & Lamme, 2012). In their study, Jiang et al. (2015) showed that unconscious conflict detection is automatic and therefore unconscious, and that it is processed by the medial frontal cortex. On the other hand, the interaction between the medial frontal cortex and the lateral prefrontal cortex is unique to conscious detection. Both conscious and unconscious detections activated frontal theta-band oscillations, strengthening the interpretation of a top-down control process in blindsight in study 1.
The question of error monitoring has also been investigated in the field of consciousness. ERP studies have shown that the error-related negativity (ERN), a
specific EEG component related to error monitoring, can be observed even when errors are not conscious (Hughes & Yeung, 2011; Maier et al., 2008; O’Connell & al., 2007).
The anterior cingulate cortex (ACC) seems to be involved in this monitoring (Hester et al., 2005; Klein et al., 2007). Another study did not observe an ERN when using a masking procedure (Woodman, 2010), demonstrating that the debate on the presence of an unconscious error monitoring system is still an open question. The top-down control process observed in blindsight is therefore an evidence for the unconscious processing of errors, supporting the first hypothesis.
When comparing brain activations in response to sensory input in brain-damaged patients and control participants, it is important to keep in mind that neuronal plasticity may have implication on brain activations and behavioural outcomes in the patients. Therefore, interpretations relative to the results observed in the blindsight patient have to be taken with caution. Even if one study has shown that cortical blindness can be induced in healthy participants using TMS, leading to the same behavioural outcomes than the ones observed in blindsight (Jolij & Lamme, 2005), neural correlated of these outcomes may vary between control participants and blindsight. Indeed, there are several “short-cut” connections (Pessoa & Adolphs, 2011) originating from the retina and projecting to frontal areas. Using intracranial recordings in epileptic patients implanted in the amygdala and the orbitofrontal cortex, we showed that the hypothesis of an alternate route in the processing of relevant stimuli is plausible.
Indeed, the classical model of a fear module in the amygdala has received much attention, especially because of the effects observed in the amygdala during fear-conditioning. It has also been shown that conditioning leads to activation of an amygdala-hippocampus-frontal lobe network (Liu et al., 2011). Moreover, a network including the brainstem, the amygdala and fronto-temporal areas was suggested as reflecting unconscious perception of fear (Liddell et al., 2005). At it has been discussed in study 2, this route is not the only candidate for the processing of aversive (unconscious stimuli), and we therefore suggest the implication of another processing route projecting to the OFC, given the lack of effect of emotion in the amygdala in the implanted patients (Pessoa & Adolphs, 2011).
2.2 Attention and consciousness
As stated above, control processes in unconscious perception have been observed in some studies, suggesting that top-down control can occur without awareness. The aim of the third study was to test if the top-down control observed in blindsight may occur in healthy participants as well, when manipulating selective attention. We used a backward masked paradigm in a group of healthy volunteers to test the prediction that top-down control may occur even if the participant is not aware of
the cue that is displayed. In our experiment, top-down control was not observed: cueing manipulation was successful only in the supraliminal condition. Moreover, the lateralized potentials of the EEG usually observed in cueing paradigms were observed in posterior loads only when the cues were supraliminal, suggesting that subliminal selective attention is not possible in this case.
In visual search experiments, it has been shown that unattended stimuli do not access consciousness (Sperling, 1960; Treisman & Gelade, 1980). In a study, Mack and Rock (1998) showed that when participants were instructed to focus on a certain location in space, the detection of a detectable and unattended stimulus appearing in another location is impaired. Taken together, these findings suggest that attention is a necessary prerequisite of consciousness (Dehaene & Naccache, 2001), even if some authors have proposed that these two concepts act independently (Koivisto &
Revonsuo, 2007). As mentioned in the theoretical part of this thesis, attention shifting using faces has generally used face stimuli as targets for the ongoing task, or as primes preceding the apparition of a lateralized target. These studies have shown that emotional expressions influence behavioural outcomes in terms of accuracy and reaction times.
However, the effect of centrally presented faces predictive of the location of a subsequent target had never been investigated at a subliminal level. Therefore, the third study filled the gap that was present in the literature, by showing that consciousness is necessary for selective attention (at least in this case), and that emotion is not processed when the working memory is overloaded. This result is consistent with studies showing that when working memory load is high, the facilitated processing of threatening faces is abolished (Van Dillen & Derks, 2012). The functional interpretation of this effect might be that high working memory load leads to decreased attentional interference by negative task-irrelevant stimuli when they are presented in the attentional focus. Our results therefore support this idea, since no effect of emotion was observed during the attentional shifting.
In summary, the findings of our third study show that (i) consciousness is necessary for selective attention and (ii) emotion is not processed when the working memory load is high. These two effects may have clinical implications regarding cognitive biases in anxiety. Indeed, anxious participants show reduced top-down control in cognitive tasks (MacNamara & Proudfit, 2014), which is related to a reduced activity in the prefrontal cortex (Bishop, Duncan, Brett, & Lawrence, 2004), and this effect is modulated by unpleasant pictures in this subpopulation. Taken together, these results may lead to new treatment strategies in anxiety: targeting attentional control in order to increase top-down control and bypass effects of aversive stimuli (MacNamara &
Proudfit, 2014). Cognitive therapies have already proved their efficacy in the interventions focusing on generalised anxiety disorders and social phobia (McEvoy &
Perini, 2009). Only subjects with low scores in the state and trait anxiety scales participated in our study, but applying this procedure to anxious participants would be
of great interest. Indeed, it would allow determining if preconscious effects of threat-related images may have an effect on this subpopulation, which would have consequences on the treatment strategies as well.