Let us turn to the processing of emotional stimuli in the brain. When the brain receives information, it generates hypotheses based on the input and what it knows from past experiences to guide recognition and action, and these hypotheses are tested against the incoming sensory information (Bar 2009: Barrett 2017; Clark 2013; Friston 2010; Raftopoulos 2019). In addition to what it knows, the brain uses affective representations, that is, prior experiences of how the input had influenced internal bodily sensations. In determining the meaning of the incoming stimulus, the brain employs representations of the affective impact of the stimulus to form affective predictions. These predictions are made within ms. and do not occur as a separate step after the object is identified; rather they assist in object identification (Bar 2009).

There is substantial evidence that the OFC, which is one of the centerpieces of the neuronal workspace that realizes affective responses, plays an important role in forming the predictions that support object recognition. The earliest activation of the OFC owing to bottom-up signals is observed between 80–130 ms (Bar 2009) . This activity is driven by fast LSF information through magnocellular pathways. Since it takes at least 80 ms. for signals from OFC to reenter the occipital cortex, OFC affects in a top-down manner perceptual processing after about 160 ms. that is, after early vision has ended. A second wave of activity in the OFC is registered at 200 to 450 ms, probably reflecting the refinement and elaboration of the initial hypothesis. There is evidence that the brain uses LSF information to make an initial prediction about the gist of a visual scene or object, that is, to form a hypothesis regarding the class to which the scene/object belongs. This hypothesis is tested and details are filled using HSF information in the visual brain and information from visual working memory (Kihara and Takeda 2010).

Barrett and Bar (2009) argue that the medial OFC directs the body to prepare a physical response to the input, while the lateral parts of OFC are integrating the sensory feedback from the bodily states with sensory cues. The medial OFC has reciprocal connections to the lateral parietal areas in the dorsal system from where it receives LSF information transmitted through magnocellular pathways. Using LSF information, the medial OFC extracts the affective context in which the object has been experienced in the past and this information is relayed to the dorsal system where it contributes to the determination of the sketchy gist of the scene or object. The lateral OFC, in its turn, has reciprocal connections with inferior temporal areas of the ventral stream, whence it receives HSF information through parvocellular pathways. Its role is to integrate sensory with affective information to create a specific representation of the scene or object, which eventually leads to conscious experience.

The role of OFC in the modulation of perceptual processing suggests that even the early activation of OFC does not affect early vision. It plays a significant role in the formation of hypotheses concerning the identity of the stimulus and their testing but this takes place during late vision. This entails, in turn, that the role of any cognitive states in relating the stimulus to past experiences and determining their emotive significance or task relevance (as indicated by the N1 ERP component elicited at 170 ms. poststimulus) for the viewer, does not affect early vision.

Another significant part of the brain that processes emotive stimuli is the amygdala (Domınguez-Borras and Vuilleumier 2013). The profile of amygdala reactivity is generally compatible with biases observed in behavioral performance or in sensory regions. The amygdala seems to be activated primarily in response to the arousal or relevance value of sensory events, rather than to negative valence only, although arousal-valence interactions and stronger responses to threat are frequently observed. The amygdala is well poised to modulate cortical pathways involved in perception and attention, because it has bidirectional connections with all sensory systems, and is also connected with fronto-parietal areas subserving attention. Studies in the macaque show that projections to visual cortices are highly organized, so that rostral regions of the amygdala project to rostral (i.e., higher level) visual areas, whereas caudal regions of the amygdala project to caudal (i.e., lower level) visual areas (Freese & Amaral 2006). At the microscopic level, there is evidence that projections from the amygdala reach pyramidal neurons in early visual areas with synaptic patterns suggestive of excitatory feedback (Freese & Amaral 2006). MRI studies in humans using diffusion tensor imaging (DTI) have identified topographically organized fibbers in the inferior longitudinal fasciculus that directly connect the amygdala with early visual areas and might contain such back-projections (Gschwind, et al 2012).

Functional studies of the amygdala in humans suggest that amygdala activation reflects the integration of perceptual information with emotional associations of the stimuli (Oya et al. 2002). The amygdala in humans processes the emotional content of facial expressions at 140–170 ms. after stimulus onset (Conty et al. 2012), or at 200 ms. (Pessoa & Adolphs 2010), or at two distinct latencies, a transient early and a later sustained period (Krolak-Salmon et al. 2004). Krolak-Salmon et al. (2004) argue that the time course of amygdala involvement and its dependence on the attended facial features confirm the critical implication of cortical frontal areas for visual emotional stimuli. There seems to be a functional link between the amygdala, the visual occipito-temporal stream, and OFC. These three regions may belong to a temporally linked triangular network implicated in facial expression processing, especially when specific attention is engaged. Intracranial recordings show that the fear effect on amygdala is recorded between 200–300 ms. after stimulus onset. Finally, Kawasaki et al. (2001) report an earlier activation onset (120–160 ms.) in ventral sites of the right prefrontal cortex for aversive visual stimuli.

The previous studies focused on the interactions of amygdala through cortical pathways. However, the amygdala and OFC receive inputs and are activated through subcortical circuits from subcortical regions in the basal ganglia, thalamus, and brainstem (Kawasaki et al. 2001; Schmid et al. 2010; Tamietto & de Gelder 2010; Vuilleumier, et al. 2003) that bypass the occipital cortex. This means that the amygdala might exert direct influences on early visual areas very early, independent of any subsequent attentional effects modulated by the affective stimulus. Tamietto and de Gelder (2010) argue that their study, which combined MEG and MRI methods, revealed early, event-related synchronization in the posterior thalamus (probably in the pulvinar), as fast as 10–20 ms. after onset of the presentation of fearful facial expressions, followed by event related synchronization in the amygdala at 20–30 ms. after onset. By comparison, synchronization in the striate cortex occurred only 40–50 ms. after stimulus onset. Thus, amygdala process fearful expressions earlier than the visual processing of the stimulus in the visual striate cortex and could affect through reentrant connections visual processing. What may explain this early onset is the fact that responses in the superior colliculus and pulvinar are tuned to coarse information in low spatial frequencies. Consistently with these findings, the subcortical pathway to the amygdala is sensitive to the presentation of fearful faces in low spatial frequencies.

Let us turn now to the empirical evidence concerning the affective modulation of visual perceptual processing by signals emanating from the amygdala or the OFC. Studies by Domınguez-Borras and Vuilleumier (2013), Pourtois & Vuilleumier (2006), Pourtois et al. (2004), and Pourtois et al. (2005), in which emotional or neutral stimuli were presented as cues before the presentation of a target stimulus, usually a bar, either at the location of the cue (valid trials), or at some other location (invalid trials), suggest that the C1 ERP component that is generated very early (in less than 80 ms.) in the early striate visual cortex had a significantly higher amplitude for fearful faces cues than for happy or neutral faces. Since the EEG recordings were time-locked to the emotive cue (a fearful face, for example) and not to the subsequent bar-target, any brain response observed is likely the result of the effect of the affective cue on visual processing rather than a response to the subsequently shown target-bar. Moreover, since the sites of C1 are mainly in the cuneus and lingual gyrus in the occipital visual system, they clearly belong to early vision. The enhancement of C1 by fearful faces is clearly a direct emotive effect on early visual processing that is independent of any attentional effects, which is consistent with the findings that C1 is not mediated by any sort of attention, as it is known that C1 is not affected by spatial attention (Di Russo et al. 2003).

This early effect could not have been produced by low-level features of faces rather than the emotional trait of the face, since experiments with inverted faces did not produce similar effects. This is supported by EEG experiments that contrasted the involuntary effects triggered by exogenous and emotional cues (Brosch et al. 2011) and revealed that these two factors operated during two distinct time windows: ERPs time-locked to the exogenous cue showed a specific enhancement of the N2pc component, consistent with a rapid shift in attention to the cued side, whereas the emotional cue enhanced the P1 time-locked to the target, consistent with enhanced visual perception. Thus, ERPs clearly differentiated between processes mediating attentional biases induced by emotional meaning and biased caused by the physical properties of the stimuli. In Saito’s et al. (2022) study, finally, participants were engaged in an associative learning task wherein neutral faces were associated with either monetary rewards, monetary punishments, or zero outcome in order for the neutral faces to acquire positive, negative, and no emotional value, respectively. Then, during the visual search task, the participants detected a target-neutral face associated with high reward or punishment from among newly presented neutral faces. Their findings showed that there were no prominent differences in terms of visual saliency between neutral faces with and without value associations or between high- and low probability faces. This indicates that an efficient search for emotional faces can emerge without any influence of visual saliency.

Attar et al. (2010) did not examine the time course of emotional processing of stimuli per se but its effects on attentional resource allocation in a primary task with respect to which the emotional stimuli functioned as distractors. Their findings suggest that highly arousing emotional pictures consume much more processing resources than neutral pictures over a prolonged period of time, which means that emotional distractors receive prioritized processing despite severe resource limitations. This effect, however, is of relatively small size when compared to the effects of general picture processing during task-related activity, where irrelevant whole pictures without any emotional value that act as distractors have a detrimental effect on task related activity. More importantly for this paper, Attar et al (2010) found, at the behavioral level, significant decreases in target detection rates when emotional compared to neutral pictures were concurrently presented in the background. At the neuronal level, the effect was accompanied by a stronger decrease of SSVEP amplitudes directed to a primary task for emotional relative to neutral pictures. The earliest onset for the affective deflective amplitude was at 270 ms. According to our knowledge about the neural sites at which SSVEP signals are generated, the deflection observed stems from sources in early visual areas (Andersen et al. 2012).

Pourtois et al. (2005) also found that even though the C1 responses might reflect the enhanced visual processing of fearful faces, this early effect of emotional face expression in V1 was no longer present in the EEG at the time of target onset. No valence effect was observed for the P1 and N170 ERP wave forms succeeding C1 time-locked to the fearful face cue. This means that there is no early direct emotive effect on perceptual processing other than that observed for C1. Thus, the subsequent effects observed in the scalp topography 40– 80 ms. after target onset in the fear valid trials probably correspond to a distinct modulation of the mechanisms of spatial attention towards visual targets appearing at the same location, perhaps triggered by the initial emotional response to facial cues. As we shall see, this means that the functional coupling between temporal and posterior parietal regions that direct the focus of spatial attention and occipito-temporal cortex through top-down signals from the former to the latter might be enhanced following threat-related cues (such as fearful faces). The positive correlation between the early temporo-parietal activity and the subsequent extrastriate response was significantly higher in valid fear trials than in all other conditions. This correlation may provide a neural mechanism for the prioritized orienting of spatial attention towards the location of emotional stimuli (Pourtois et al. 2004; Vuilleumier 2002).

The early C1 emotive effect may result from a rapid modulation of visual processing in V1 by reentrant feedback from the amygdala (Amaral et al. 2003; Tamietto & de Gelder 2010; Vuilleumier et al. 2004), since, as we have seen, the amygdala is connected to the early visual cortex and it is likely that is activated subcortically by emotional stimuli through signals bypassing the occipital cortex, at earlier latencies than the early visual striate cortex. Pourtois and Vuilleumier (2006, 76) speculate that this early differential response to emotional faces, as evidenced by the enhancement of C1 ERP component in V1, may serve a rapid decoding of socially relevant stimuli in distant regions such as the amygdala that begin responding to faces at 120 ms. poststimulus as a result of receiving signals from V1. Alternatively, it may reflect a rapid modulation of V1 by reentrant feedback from the amygdala that process emotional stimuli at earlier latencies through signals that they receive from subcortical pathways bypassing the occipital cortex. These two functions are not mutually exclusive; the early amygdala responses to emotive stimuli subserved by subcortical pathways rely on coarse low-spatial frequency information, whereas the later response of the amygdala to signals from V1 may reflect the more refined processing of the emotive stimuli based on high-spatial frequency, more detailed information. This is consistent with findings showing that the face-responsive area in fusiform gyrus responds to emotional faces after the registration of N170 ERP waveform that signals the recognition of faces.

This is a case in which the CI of early vision seems to be threatened. If fearful faces or other threatening stimuli affect visual processing at these early latencies, and if the assessment of threat relies on cognitive states that are formed as the result of past experiences, this clearly means that early vision is CP. Since the onset of C1 is too early to be attributable to any top-down cognitive influences, CP could occur only if the cognitive information guiding the attribution of threatening character of the stimulus was embedded within the early visual system, the way the formation principles (Burge 2010), or operational constraints (Raftopoulos 2009) were thought to be embedded in the visual system allegedly rendering visual perception CP and the state transformations in it similar in nature to discursive inferences. The problem with this line of thought is that it is likely wrong, for the same reasons that view that the operational constraints at work in vision render early vision CP is wrong (Raftopoulos 2019).

Pourtois & Vuilleumier (2006), based on work by Lang et al. (1998) and Thiel et al. (2004), suggest that the early activity of the amygdala in response only to threatening and not to other affective stimuli possibly reflect some general alerting or arousal effect triggered by fearful faces. Along these lines, Tamietto & de Gelder (2010) offer an intriguing explanation of the early registration of threatening stimuli in the amygdala. They argue that data from human and animal studies suggest the continuity of subcortical emotion processing across species and its evolutionary role in shaping adaptive behavior. They discuss a proposal by Isbell (2006), who suggests that fear detection has played a major part in shaping the visual system of primates, and in its integration with an emotion circuit centered on the amygdala. Snakes, for example, presumably represented a major threat for our ancestors and the need to detect them on the ground accelerated the development of greater orbital convergence, allowing better shortrange stereopsis, particularly in the lower visual field. The koniocellular and magnocellular pathways from the retina were developed further to connect the superior colliculus with the pulvinar promoting fast and automatic detection of snakes. It is likely that this mechanism generalized for detection of other fear-relevant stimuli. The fast reactivity of this subcortical pathway to visual stimuli with low resolution has also promoted the parallel development of a cortical visual pathway to the amygdala with complementary features to those of the subcortical pathway. Consistent with this claim is evidence that the subcortical pathway for processing emotional stimuli emerged early in phylogenesis, as evidence in human and non-human primates indicates that the formation of these structures is more developed at birth compared to the relatively immature development of the cortical areas involved in visual and emotional processing.

Hodgson (2008, 345) reaches the same conclusion. ‘Animals... were integral to the evolution of the human brain to the extent that the encoding of animal forms seems to have become a dedicated domain of the visual cortex.’ Considering that in the Palaeolithic our hunters-gatherers ancestors were constantly living in close proximity with animals some of which were dangerous (cave-bears, cave-lions, wolves) and that animals were an integral part of human life, it is only natural to assume that the conditions in which they lived shaped the way they perceived, and thought of, animals and, thus, their brains (which are the same as our brains) were accordingly shaped. From these considerations, it is it very plausible that our brains have developed mechanisms embedded in our perceptual systems that allow tracking of threating animal-related stimuli very fast, enabling fast responses to avoid danger. This requires that our perceptual systems be linked directly through fast interconnections with the brain systems that encode the emotive significance of the stimulus and is not conditioned on interactions with, and modulation by, the cognitive areas through attention. As Kim et al. (2017, 7) remark ‘Although attention and arousal are often considered linked processes, the origins of their modulatory signals are quite distinct. Arousal signals have primarily been attributed to the locus coeruleus–norepinephrine system, whereas the attentional-control signals stem from a cortical constellation encompassing both dorsal and ventral frontoparietal networks. Thus, while they are potentially complementary modulatory signals, it remains unclear as to whether these two processes influence response properties in the brain interactively or they act as two independent processes.’

This idea is supported by findings that the perceptual and emotive-significance-assessment brain areas have phylogenetically developed prior to the formation of the areas that are closely related with the cognitive control of the neuronal activity elsewhere in the brain. These are the association areas that assess and integrate information across different brain areas subserving different modalities (Preuss 2011; Schoenemann 2006; Neubauer et al. 2018). The cognitive areas are mostly associated with a network involving frontal and parietal areas (Jung & Haier 2007; Barbey et al. 2012). The brain has formed a system of threat appraisal and avoidance behavior, most likely hardwired in the brain, well before the development of the cognitive areas. In view of the fact that evolution never undoes previous constructions and functionalities that successfully addressed evolutionary pressures in order to reshape the brain a nouveau, even when the newer cognitive areas developed and new white matter tracks—the superior longitudinal fasciculus, the arcuate fasciculus, the uncinate fasciculus and the cingulum (Lebel and Leoni 2018) — were formed connecting these new areas with the older areas of the brain (allowing, thus, cognitive control of the neural computations), the previously developed mechanisms of threat-appraisal and reaction remained intact and fully functional enabling fast reaction to possible threat. It goes without saying that the new connectivity enables further elaboration of the threat assessment, it allows integration with other mental abilities, etc., but it does not abolish the established instinctive and hardwired fast threat assessment and reaction, which is advantageous to the agent because it allows fast responses to perceived threat.

If these considerations are on the right track, the detection of the threatening character of the stimulus by the early visual perceptual system does not rely on the workings of some cognitive states that are formed on the basis of past experiences and are embedded within the early visual system. Most likely, the relevant mechanisms are hardwired in the early visual system, exactly the way the operational constraints are hardwired, as a result of the phylogenetic development of our species in its environment. For this reason, it is very probable that no contentful states are involved in the assessment of threat, and, thus, no CP could occur. One might retort that even if it is correct that the attribution of threat in a stimulus does not require any cognitive involvement, it is still the case that organisms with different past experiences in differing environments will make different assessments concerning the threat that a stimulus might pose for the organism. This is correct, of course, but it is either irrelevant to the problem of the CP of early vision, or does not pose any threat to the claim that early vision is CI, for the same reasons that perceptual learning does not (Raftopoulos 2009).

Let us move to the emotive influences on P1. The evidence we have examined shows that there are no valence effects on P1 time-locked to the fearful cue, which means that this cue does not affect directly P1, even though there is a positive correlation between the C1 modulated by the cue and P1 enhancement. Studies in humans (Domınguez-Borras and Vuilleumier 2013; Olofsson et al. 2008; Pourtois & Vuilleumier 2006; Pourtois et al. 2004; Pourtois et al. 2005; Vuilleumier et al. 2004; Vuilleumier and Driver 2007) show that fearful or threatening vs. neutral or happy faces presented as cues induce a higher amplitude of VEP (visual evoked potentials) and an enhancement of the P1 ERP component time-locked to the stimulus presentation at the cued location at about 120 ms. Thus, the enhanced evoked potential concerns the perception of the target-bar presented at the location of the emotional cue (or at another location in invalid trials) and not the perception of the cue per se. P1 originates in extrastriate areas and is considered to be the hallmark of the effects of exogenous spatial attention on visual processing, that is, the effects of the automatic orienting response to a location where a sudden stimulation occurs. This entails that the emotion-related modulation of the visual cortex arises prior to the processing stages associated with fine-grained face perception indexed by the N170 component for face recognition. Emotional affects are prior to, and help in determining, the categorization of the stimuli and can collaborate with attentional effects by enhancing the processing of spatially relevant and emotionally significant stimuli. Notice that.

The Pourtois et al. (2004; 2005; 2006) experiments were designed with short cue-target intervals (CTI) so that allocation of attention be automatic and free from any cognitive manipulations. The short CTI are also important to avoid a phenomenon observed in a variation of the cueing task that we have discussed thus far, namely the dot-probe task in which two cues, one emotional and the other neutral, are presented simultaneously and then a target appears at the location of the emotional cue (valid trails) or the location of the neutral cue (invalid trials). While attention is allocated automatically to one of the two cues during the initial presentation, participants often try to move their attention back strategically to the central position if the presentation of the cues is extended for a longer duration (CTI > 300 ms.). Should this happen, the re-allocation of attention to the initial position is inhibited and this may result in a reversed validity effect; for example, Cooper and Langton (2006) found a threat bias for 100 CTI, but a reversed effect at 500 ms.

EEG recordings show that targets preceded by emotional cues elicit a larger visual P1 component, relative to those preceded by neutral cues. This is consistent with enhanced visual processing of targets when pre-cued by emotional stimuli. When the target is preceded by an emotional valid cue, activity to the target is first increased in parietal areas (50–100 ms.), probably reflecting top-down attentional signals induced by the emotional cue and responsible for faster spatial orienting to the target location and enhanced processing in extrastriate visual cortex at the P1 latency (100–150 ms.).

In addition to enhancing responses in perceptual regions, emotion signals may also increase activity in brain regions associated with attention control, including the posterior parietal cortex (Vuilleumier 2005). This might in turn influence top-down attentional signals on perceptual pathways. In particular, such effects have been observed in dot-probe tasks with fear-conditioned images (Armony & Dolan 2002), threatening faces (Pourtois, et al 2005), or even positive affective stimuli. Presentation of a peripheral threat-related cue in these tasks typically produces an increased activation in frontoparietal networks, reflecting a shift of attention to the location of the emotional cue (Armony & Dolan 2002). Moreover, when a neutral target (dot) is preceded by a pair of face cues (one neutral and one emotional), those targets preceded by neutral cues, as compared to by emotional ones, elicited reduced BOLD responses in intraparietal sulcus (IPS) ipsilateral to the targets, consistent with a capture of attention by the emotional cue on the contralateral side and a reduced ability to reorient to the target on the ipsilateral side (Pourtois, et al. 2006, 2013). Targets appearing after an emotional cue produce stronger BOLD responses in the lateral occipital cortex (Pourtois et al. 2006) and a larger P1 component in EEG recordings (Pourtois et al. 2004), consistent with improved visual processing and better target detection. A detailed analysis of the time course of these effects with EEG (Pourtois et al. 2005) suggests that the modulation of parietal areas may be triggered by an initial response to the emotional cue, and subsequently induce top-down spatial attentional signals responsible for enhanced processing of the target, but only in valid trials.

Notice that the effects of fearful faces on ERPs time-locked to the target bars affected the lateral occipital P1 and N170 in the fusiform gyrus but did not affect C1 generated in the primary visual cortex or N1 generated in higher visual areas within occipito-parietal cortex.

The suggestion that emotional cues play a causal role in the enhancement of P1 is reinforced by the finding of a positive correlation between the enhancements of C1 and P1 ERP components even though the former is time-locked to the cue and the latter is time-locked to the target (Pourtois et al. 2004, 2005). Correlation analysis revealed a significant positive correlation between the amplitude of C1 and P1 in the fear condition, which, however, was restricted to the left hemisphere. There was, also, no significant correlation between the amplitude of the C1 and the P1 validity effect with happy faces in ei