These effects of emotion on face perception have led researchers to question whether unconscious perception of emotional stimuli might affect behavioural performances and neural responses as well. Fear-conditioning has especially gained interest in this field, since it relies more on autonomous circuits detecting danger than the explicit subjective experience of fear (Ledoux, 2000). Faces, as being relevant stimuli for the humans, have been used to test the unconscious perception of emotion.
Stimuli that are biologically relevant (for example the ones that signal danger) have to be processed quickly and the activation of the amygdala during the processing of emotional faces has been interpreted as a quick and dirty route to rapidly detect signals of potential danger (Vuilleumier & Pourtois, 2007). Based on the phenomenon of affective blindsight and perception of subliminal stimuli in healthy subject, the neural correlates of this processing will now be discussed.
3.1 Affective blindsight and the study of emotional faces at a non-conscious level
One of the first motivations that initiated the study of the non-conscious processing of affective signals in the human brain has been the observation that affective blindsight patients are able to perceive emotional signals of fear despite the destruction of the visual cortex (De Gelder, Vroomen, Pourtois, & Weiskrantz, 1999).
Blindsight (or cortical blindness) refers to the neurological state where the striate cortex (V1) is damaged, resulting in blindness in the contralateral visual field. Despite this blindness, some patients have been able to discriminate features of visual events, as motion direction or orientation (Weiskrantz, Barbur, & Sahraie, 1995). These patients thus perform a cognitive task in the absence of any explicit knowledge, illustrating a case of implicit processing where awareness is dissociated from perception (Weiskrantz, 1991). One possible explanation of this phenomenon is the existence of a parallel visual pathway which would pass through the superior colliculus and the pulvinar. This pathway is believed to be able to process visual information even when the visual cortex is damaged (Weiskrantz, 1986).
These observations involving blindsight patients were first made on basic representations, like detecting a light spot in the blind field, relying on low-level
processes. This raises the question of whether higher level representations can take this non-conscious route as well. Interestingly, stimuli conveying an emotional message seem to represent a specific category that is processed non-consciously in blindsight patients. In a pioneer study by De Gelder et al. (1999), patients were able to guess above chance level the expression of emotional faces, even without conscious awareness.
Moreover, another study on a patient with a bilateral destruction of the visual cortices (Pegna, Khateb, Lazeyras, & Seghier, 2005) showed that emotional expressive faces can elicit brain responses, particularly in the right amygdala. As was stated by Ledoux (2000), a subcortical pathway allows the processing of threatening stimuli very rapidly, without getting to the visual cortex and passing through the thalamus before reaching the amygdala. This route is a plausible candidate for the processing of emotional stimuli presented in the absence of awareness, especially faces expressing a negative emotional valence (Morris, Öhman, & Dolan, 1999). In support of this theory, direct connexions between the thalamus and the amygdala have been found in rodents (Cowey & Stoerig, 1991; Linke, De Lima, Schwegler, & Pape, 1999). This theory has been confirmed using a fear conditioning paradigm with an angry face associated with a 95dB white noise in healthy participants. fMRI results showed that for stimuli that were consciously seen, visual areas were activated. On the other hand, when the conditioned non-seen faces were processed, the superior colliculus and the pulvinar were activated, and this activity was associated with the one of the right amygdala.
Concerning the processing of faces without awareness, it is not yet established how the brain processes perception of unaware emotional stimuli (Khalid, Finkbeiner, König, & Ansorge, 2013). Structures involving two subcortical processing routes have been proposed in the literature to account for this unconscious processing (see Tamietto
& De Gelder, 2010). As described above, visual stimuli are known to project to the primary visual cortex through retinal fibres to the lateral geniculate nucleus of the thalamus. From there, visual information conveys information to the striate and then the extrastriate cortex along the ventral and dorsal streams. However, a small portion of fibres project to the superior colliculus, the pulvinar, the amygdala, the substantia innominata and the nucleus accumbens (see Fig. 3 a). The latter pathway is a non-visual pathway processing emotional stimuli which includes the following brain structures: the locus coeruleus, the periaqueductal grey, the nucleus basalis of Meynert, the basal ganglia, the hypothalamus and the hippocampus (see Fig. 3 b).
Each of these structures plays a different role in human cognition and some structures from these two pathways are proposed to interact during the processing of unconscious emotional signals. As mentioned above, the superior colliculus, the pulvinar and the amygdala seem to act in synergy in this processing.
3.2 Processing of subliminal emotional faces in healthy subjects
In healthy subjects, the effects of non-conscious representations have been investigated by some experimental designs, including backward masking or binocular rivalry. In the first paradigm, a neutral stimulus is presented to the subject immediately after the presentation of an emotional target stimulus which appears at a non-conscious level (Stigler, 1910). Whalen et al. (1998) used this design to study the perception of masked faces (i.e. without awareness). In their paradigm, a face was flashed during 33ms, immediately followed by the presentation of a visible face presented during 167ms. The flashed faces were not consciously perceived and were representing either happy or frightened expressions. The visible stimuli were always representing happy faces. Using fMRI, they observed that the amygdalae were significantly more activated by the flashed faces representing an expression of fear, as compared to the happy faces.
It seems therefore that even if the emotional representations are not consciously perceived, activation of the amygdala may occur. Following this discovery, Morris, Figure 3: The two pathways involved in unconscious emotional processing. (a) The visual pathway represents direct projections from the retina to the visual cortex through the LGN in the thalamus (Th; thick arrows). The alternative visual pathway represents the minority of fibres projecting to the superior colliculus (SC) and the pulvinar (Pulv), before reaching the extrastriate cortex (thin arrows). (b) The non-visual pathway involved in emotion processing consists of cortical and subcortical structures such as the amygdala (AMG), the substantia innominata (SI), the nucleus accumbens (NA), brainstem nuclei (the periaqueductal grey (PAG) and the locus coeruleus (LC), the orbitofrontal cortex (OFC) and the anterior cingulate cortex (ACC). Taken from Tamietto & De Gelder, 2010.
Öhman, and Dolan (1998) used PET to explore the route taken by the non-conscious stimuli before reaching the amygdala and found that the subcortical route activating the superior colliculus was engaged. In summary, these studies have highlighted brain structures involved in the perception of non-consciously perceived emotional stimuli (Morris et al., 1998; Pasley, Mayes, & Schultz, 2004; Whalen et al., 1998; Williams, Morris, McGlone, Abbott, & Mattingley, 2004) including more or less the same regions involved in the non-conscious perception of emotional stimuli in blindsight patients, namely the amygdala, the superior colliculus, the basal ganglia and the pulvinar.
Interestingly, the activity of these structures was enhanced in the condition of an emotional non-consciously perceived stimulus, as compared to the activity observed in the condition of a consciously perceived stimulus (Anderson, Christoff, Panitz, Rosa, &
Gabrieli, 2003).
Binocular rivalry refers to an effect of visual perception where two different images are presented alternatively to each eye, which leads to perceptual dominance of one image at a time while the other is suppressed (Blake & Logothetis, 2002). Using this paradigm with fearful faces versus images of chairs, Pasley, Mayes and Schultz (2004) showed increased left amygdala activation following fearful faces, but no increase in the inferior temporal cortex (IT). This region is a part of the visual system responsible for object recognition (Gallant, 2000). Fearful faces therefore bypass this high-level visual area, directly accessing the amygdala through a phylogenetically older visual system. In another study, Williams, Morris, McGlone, Abbott and Mattingley (2004) used fearful, happy and neutral faces which were systematically presented with the image of a house with binocular rivalry. They showed that extrastriate visual areas were only activated when faces and houses were consciously perceived, whereas fearful (versus neutral) faces activated the amygdala bilaterally in both conscious and suppressed conditions. Concerning happy faces, they elicited greater amygdala activation only when suppressed. Therefore, the amygdala would refer to a rapid rough processing route, conveying basic non-specific information about visual input.
3.3 Electrophysiological correlates of subliminal emotion face processing
When interested in the timing of processing of emotional stimuli presented under the threshold of awareness, we may ask if non-conscious presentation of fearful facial expressions would have an effect on the N170 component specific to faces as well. Fear enhanced N200 was found to occur in fronto-central sites in subliminal presentations (Liddell et al., 2005) while N400 and P300 were associated with supraliminal presentations. Kiss and Eimer (2008) found earlier activation for the same condition (140-180ms) in anterior sites. An ERP (event-related potential) study investigated the
modulation of the N170 using the backward masked paradigm, where the target stimuli composed of neutral, happy or fearful faces appeared at 5 different durations (16ms, 33ms, 66ms, 133ms and 266ms) followed by a neutral stimulus (Pegna et al., 2008).
The task of the participants was to respond if they saw a fearful face or not. Behavioural results showed that the longer the duration of the target stimulus, the shorter the reaction times. As for the electrophysiological results, the subliminally presented fearful faces produced a stronger N170 than non-fearful faces, which was also the case for the supraliminal conditions. This activation was located in right temporal and temporo-occipital sites, again suggesting the involvement of the right amygdala in this type of processing.
When one is interested in the perception of faces under different levels of awareness, one may also ask how awareness modifies the electrical activity in the brain.
This has indeed been the central issue in many studies on brain processes. The synchronised gamma-band activity occurring in the visual cortex has been associated with visual awareness (Crick & Koch, 1990; Sewards & Sewards, 1999): when neurons fire in synchrony at this frequency (gamma), awareness arises. Moreover, when a stimulus reflects goal conducive features, gamma band oscillations were observed in a study (Grandjean & Scherer, 2008). Phase synchrony between beta and alpha frequencies has been linked to consciousness (Meador et al. 2002; Gross et al. 2004;
Palva et al. 2005; Srinivasan et al. 1999; Doesburg et al. 2005) and observed when participants report a subjective feeling towards a stimulus (Dan Glauser & Scherer, 2008). Theta band activity has also been reported in this processing, with induced activation during recollection of personal events (Guderian & Düzel, 2005) and awareness of perceived words in frontal regions (Melloni et al. 2007). Prestimulus alpha synchronisation has also been highlighted when predicting seen (versus unseen) stimuli (Mathewson, Gratton, Fabiani, Beck, & Ro, 2009).
Gamma band synchronisation has also been observed in the amygdala when participants process emotional stimuli consciously (Oya et al. 2002; Luo, Holroyd, et al.
2007), as well as in a distributed network including amygdala and visual, prefrontal, parietal and posterior cingulate cortices when processing supraliminal and subliminal emotional stimuli, suggesting that the gamma band does not only reflect consciousness but may also sustain emotional processing in some circumstances (Luo et al., 2009).