Studies investigating the interaction between attention and emotion have first been based on the well-known effect that attention to one location in space leads to enhanced accuracy and shortened reaction times when participants have to discriminate a stimulus at that location (Posner, Snyder, & Davidson, 1980). Attention has been divided into two different types: endogenous and exogenous attention. Exogenous attention refers to a bottom-up process, where an external cue captures attention, leading to enhanced sensory input at that location and therefore to behavioural outcomes (see Fig. 4A – left side of the figure). Endogenous attention (also known as selective attention) refers to a top-down processing, where, for instance, a central stimulus (an arrow) orients attention to a certain location in space, also leading to better and faster discrimination at that location (see Fig. 4A – right side of the figure).
When investigating endogenous attention, valid trials lead to faster reaction times than invalid trials, independently of the duration between the cue and the target (SOA; Posner, 1980) (see Fig. 4C). In the case of exogenous attention the effect of
Figure 4: (A) The Posner paradigm with on the left a target preceded by a peripheral cue and on the right a target preceded by a central cue. (B) Reaction times as a function of SOA for valid and invalid trials in non-predictive peripheral cues. (C) Reaction times as a function of SOA for valid and invalid trials in central predictive cues. Reproduced from Chica, Bartolomeo and Lupianez, 2013.
validity is SOA-dependent: a short SOA leads to faster reaction times for valid than invalid trials, whereas a longer SOA leads to slower reaction times for valid than invalid trials (see Fig. 4B). The effect of validity with a short SOA has been interpreted as an effect of facilitation of the processing of the target by attentional capture (Cameron, Tai
& Carrasco, 2002). On the other hand, the effect observed when using a longer SOA has been labelled “Inhibition of return” (IOR; Posner, Rafal, Choate, & Vaughan, 1985). If attention is oriented out of the location that has just been explored by the individual, he will be less efficient when exploring again this region. After an attentional shift, the IOR prevents attention to be redirected to previous locations. This phenomenon therefore acts like a visual field exploration facilitator.
4.1 Electrophysiological correlates of spatial attention
Attention has been shown to modulate components of the EEG as early as 100 ms (P1) after target onset and spreading to the N1 component, at occipital and parieto-occipital scalp sites. The enhancement in the amplitude observed on these components would reflect selective amplification of sensory input (Hillyard & Anllo-Vento, 1996;
Mangun, Hillyard, & Luck, 1993). However, these components reflect amplification of sensory input contralateral to the attended position rather than deployment of attention.
Therefore, electrophysiological responses relative to cue onset might inform us about the process underlying the allocation of attention to a specific location. For this purpose, several components have been highlighted using the difference in amplitude between the two hemispheres, reflecting the deployment of attention to the right or to the left visual field respectively reported. These are lateralized components. The most commonly observed lateralized component in the field of visual search is the N2pc (N states for negativity; 2 states for the timing since it appears at ~200 ms post-stimulus; pc states for the localisation: posterior-contralateral; Luck & Hillyard, 1994), which is generally amplified over the visual cortex contralateral to the attended location.
However, due to the lateralized characteristics of the N2pc, we do not compute analysis on this component with centred stimuli, but only when stimuli are lateralized, allowing the computation of the difference between ipsilateral and contralateral presentations (see Luck, 2012).
Several studies have also discussed specific components relative to cue onset referring to selective (endogenous) attention, or when the cue is presented at the centre of the screen (Eimer & Van Velzen, 2002; Hopf & Mangun, 2000; Jongen, Smulders, &
Van Breukelen, 2007; Nobre, Sebestyen, & Miniussi, 2000). Three components have especially been reported: the early directing attention negativity (EDAN), the anterior directing attention negativity (ADAN) and the late directing attention positivity (LDAP) (Harter, Miller, Price, Lalonde, & Keyes, 1989).
About 200 ms after cue onset, the EDAN has been observed over posterior areas and proposed to reflect the encoding of the significance of a cue and the induction of a consequent attentional shifting, even if it has been argued that this early component only reflects differences in the physical properties of the stimuli used (left versus right arrows; Harter et al., 1989; Yamaguchi, Tsuchiya, & Kobayashi, 1994; Nobre, Sebestyen, & Miniussi, 2000; Velzen & Eimer, 2003; Jongen, Smulders, & Van der Heiden, 2007). It has been concluded that the EDAN acts like an N2pc, and reflects the selection of available information relevant to the ongoing task.
The ADAN appears in frontal and central scalp sites about 300 ms after cue onset and would reflect attentional shifting following the processing of the cue (Jongen et al., 2007). This interpretation is partly based on the frontal distribution of this component, linking it to the more general process of attentional control proposed earlier in the literature (LaBerge, Carlson, Williams, & Bunney, 1997; Posner & Petersen, 1990).
Finally, the LDAP appears at about 500 ms after cue onset at posterior scalp sites and has first been associated with cortical excitability, anticipating the apparition of a stimulus at attended locations (Harter et al., 1989). This view has been challenged by several studies observing a decrease in the LDAP before the onset of the target (Hopf
& Mangun, 2000) and which were not able to show an increase of this component in covert attention shifting (Mangun, 1994; Nobre et al., 2000; Talsma, Slagter, Nieuwenhuis, Hage, & Kok, 2005; Yamaguchi, Tsuchiya, & Kobayashi, 1994).
Following these observations, several studies have investigated whether emotion interacts with this attentional capture and how emotion may influence neural correlates of attentional processing.
4.2 The role of relevance and awareness
Studies on attentional processes using relevant stimuli have shown that these two concepts interact such that relevance enhances attention. As it was stated earlier, faces represent a stimulus category that is preferentially processed by individuals (Itier &
Taylor, 2004). In the field of attentional processes, this has been shown using the
“flicker paradigm”. This paradigm has been introduced by Rensink, O’Regan, and Clark (1997). Two images separated by a mask are presented sequentially, and the task of the participant is to detect as quickly as possible the changes occurring in the second image as compared to the first image. Using this paradigm, participants are faster and more precise when detecting changes appearing in face stimuli than other objects (Ro, Russell, & Lavie, 2001). This effect has been interpreted as attentional capture by faces as compared to other objects in a visual scene.
Following this observation, the question remained whether this attentional capture is automatic. The Posner paradigm has been used to test automaticity in the facilitation of processing of faces (Fox, Russo, Bowles, & Dutton, 2001). In this study, emotional words and faces (positive, negative, neutral) were used as exogenous cues to orient attention to the right or to the left side of the screen, where a valid or invalid target appears. Reaction times were faster for valid than invalid trials, especially for targets cued by a negative emotion. This effect would act as exogenous attention, but can be modulated by certain personality traits like anxiety. The first interpretation of this result has been that faces capture attention automatically, and that emotion plays a crucial role in this capture (Mathews et al., 1997; Vuilleumier, 2002; Yiend & Mathews, 2001). However, this view has been challenged by Fox et al. (2001) who were interested in invalid trials cued by threat, which led to slower reaction times in high-anxious individuals. According to these authors, this is due to a difficulty to disengage from negative stimuli in these patients. In another study, Georgiou et al. (2005) wanted to generalise this result to other negative emotions. They presented faces expressing fear and sadness at the centre of the screen, followed by a target letter appearing on the top, the bottom, the left or the right. They showed that anxious participants disengage less easily from fearful faces but not from sad faces than non-anxious participants. The role of anxiety in attentional processes will be discussed in section 4.3.
In order to test automaticity of attentional capture by faces, Theeuwes and Van der Stigchel (2006) used saccadic eye movements in their paradigm. They proposed to use bilateral presentation of faces and objects instead of unilateral presentations. After the bilateral display representing a face on one side and an object on the opposite side, a pointer presented in the centre of the screen indicated in which direction an eye movement had to be made. Saccadic latencies were measured in order to determine if the phenomenon of IOR would be observed at the location previously filled with a face.
Results showed that saccadic latency was more important for faces than objects, suggesting that IOR was involved, and therefore that attentional capture by faces is automatic since that location has already been investigated by the individuals, leading to slower eye movements to that location.
If attentional capture is automatic, we may wonder what happens if faces are presented under the threshold of awareness. This question has been investigated by Mogg & Bradley (1999). In their study, a pair of faces (angry versus neutral or angry versus happy) was displayed under the threshold of awareness (14 ms) and was immediately followed by a mask (a scrambled face). The detection of a target following this display was faster for valid trials where the target replaces an angry face than a neutral or a happy face. Moreover, this effect was more important when the target appeared on the left visual field, suggesting the implication of the right hemisphere in this processing. This effect has also been shown to be more important in a group of
that attentional capture is mediated by the amygdala for fearful masked faces. In their paradigm, a pair of faces (fearful and neural) was immediately followed by a mask composed of a pair of neutral faces. Then, participants had to detect the location of a target which replaced either the fearful or the neutral face. An EEG study showed activation of the N170 component contralateral to fearful masked faces (Carlson &
Reinke, 2010). The processing of negative emotions can even be processed when faces are outside the attentional focus (Pegna et al., 2011), with modulation of the N170 specific to masked fearful faces. These findings suggest that emotional faces capture attention more importantly than other stimuli and that this capture can even be observed pre-consciously.
4.3 The role of anxiety
Trait anxiety is characterised by attentional biases towards threatening stimuli, especially facial expressions (Mogg & Bradley, 1999; Mogg, Millar, & Bradley, 2000;
Fox, 2002) and even when they are presented under the threshold of awareness (Mogg
& Bradley, 2002; Li, Zinbarg, & Paller, 2007). These attentional biases are characterised by an increase in the viewing time on threatening images (Quigley et al., 2012) and faster eye movements for bodies representing angry or fearful postures directed towards a peripheral target (versus away from the peripheral target) (Azarian et al., 2016). In attentional cueing tasks, the interaction between facial expression and gaze direction has been observed using fearful faces in anxious participants (Mathews et al., 2003; Tipples, J., 2006; Fox et al., 2007). Another’s gaze can orient attention in the viewing direction and lead to selective amplification of the sensory input at this location, influencing reaction times and accuracy in congruent as compared to incongruent trials. In general, averted fearful gaze amplifies this effect in anxious (versus non-anxious) participants. This attentional shifting by eye gaze is observed even if the face is not consciously perceived (Sato et al., 2007; 2016; Xu et al., 2011).
Subliminal gaze cues can orient attention and are predictive of reaction times when the cue is non-predictive of target location. However, Al-Janabi et al. (2012) observed an effect of masked gaze cues only in the case when the masked cues are in the context of unmasked predictive cues, therefore suggesting the involvement of volitional control (top-down process), in contrast with reflexive mechanisms during the processing of unconscious gaze cues.
At the neuronal level, a study (Ewbank et al., 2009) has shown more important activation of the right amygdala in response to attended angry (versus neutral or fearful) faces and more important activation of the left amygdala in response to unattended fearful (versus neutral or angry) faces in high anxious individuals. Therefore, the processing of these two types of emotions can be distinguished at the neuronal level as a
function of the participant type, raising the question of vulnerability. Another study (Ewbank et al., 2010) showed increased right dorsal amygdala in response to angry faces with an averted gaze and to fearful faces independently of gaze direction in high state anxious participants.
All these observations were made with stimuli expressing negative (fearful, angry) emotions, subsequently raising the question of whether positive or neutral stimuli could affect behavioural performances as well. Morel et al. (2014) showed that the P1 component of the EEG is enhanced when processing happy (versus neutral) faces in participants showing higher levels of trait anxiety. In a recent study, Gmaj et al.
(2016) used a Continuous Attention Task with neutral stimuli in patients with anxiety disorders and showed cognitive impairment influencing, among others, the engagement of attention. An fMRI study showed a reduction in the connectivity of regions implied in attentional processes during resting state in anxiety (Modi et al., 2015). A behavioural study using subliminal and supraliminal primes (arrows) showed that trait anxiety is linked to increased distractor processing in the supraliminal condition only.
This suggests that consciousness is an absolute precondition for the observation of attentional bias in anxiety. An fMRI study using neutral faces has shown that supraliminal as well as subliminal averted gazes activate a network involved in attentional processes (Sato et al., 2016). These recent evidences suggest that markers of vulnerability could be present in anxiety and could in part explain the maintenance of anxiety (Thibodeau, Jorgensen, & Kim, 2006).