Nonconceptual content is a contrastive term that can be explicated in relation to conceptual content whose properties it lacks in some or several respects. Accordingly, the notion of nonconceptual content has been invoked in explaining many types of mental states whose content cannot be circumscribed by the subject's conceptual capacities. These states include representational states at the subpersonal or subdoxastic level of information processing; sense perceptions by virtue of their phenomenological or logical properties or both; and representational states of non-human animals and human infants who do not seem to possess concepts in the first place (Bermudez, 2008). However, the third category is theoretically redundant because the representational states of non-human animals and human infants are nonconceptual in some other sense and do not constitute a kind of nonconceptual content of their own. Therefore, I focus on three types of nonconceptual content - subdoxastic, phenomenological, and inferential - and their role in emotions. I first introduce each version of nonconceptual content. Then I examine how each of these versions has been applied in the discussion on emotional content and whether any of these versions offers a valid account of the content of all or most human emotions at the algorithmic level of analysis.
A. Subpersonal and subdoxastic nonconceptual content
Characterizations of nonconceptual content as subpersonal, subdoxastic, paralinguistic and non-propositional are negative as they focus on properties that nonconceptual content lacks in comparison with conceptual content. Dual-process models of information processing place subpersonal and subdoxastic representations within the implicit and associative type 1, but as I have argued above, this categorization alone is not very informative because implicit processes can operate with either conceptual or nonconceptual content. Subpersonal and subdoxastic representations are evolutionarily old and shared by humans and animals alike, and their computational system is highly independent from rule-based explicit processing. A classic analysis of such a computational system is Jerry Fodor's (1983) distinction between central and modular processing that some advocates of affect programme theories have applied in their accounts of emotional processing.
Affect programme theorists argue that emotions are special kinds of information-processing devices that emerge rapidly in situations that are relevant to the organism's fundamental life tasks. The rapid onset of emotions suggests that they are independent from conceptual processes. Yet the fact that mutually similar emotions emerge in otherwise dissimilar situations indicates that there is an internal system that is capable of appraising the emotional relevance of situations very quickly on the basis of salient perceptual cues. Accordingly, Paul Ekman (1999; 2003) has posited, following the work of Charles Darwin (1965 ) and Silvan Tomkins (1962), the existence of basic emotions with distinct affect programmes. On the one hand, Ekman describes affect programmes as the operating mechanisms of basic emotions whose characteristics are distinctive universal signals, distinctive physiology, automatic appraisal mechanism, distinctive universals in antecedent events, distinctive developmental appearance, presence in other primates, quick onset, brief duration, unbidden occurrence, distinctive thoughts, memories and images, and distinctive subjective experience. In this sense, "the affect programme of an emotion is situated in the central part of the somatic component and is put forward as the cause of several other components (motivational, peripheral somatic, motor) in the emotion", as Moors (2012, 259) points out. On the other hand, Ekman suggests that the different elements of response "enter into" affect programmes that come out as "complex, coordinated, and automated" emotional responses, as Griffiths (1997, p. 77) has summarized. However we interpret the concept of affect programme, these programmes involve automatic appraisal mechanisms that compare perceptually submitted information to stored information about classes of stimuli that have previously, in either the ancestral or the individual past, elicited similar emotional responses. Biologically determined stimuli can be amplified or dampened by social learning. The automatic appraisal mechanisms are implemented in developmentally ancient, subcortical brain structures such as the amygdale, basal ganglia, and lateral temporal cortex that we share with many other vertebrates, especially with other mammals, and they operate without the control of conscious thinking and semantic memory. Instead, their functioning is modular in the sense suggested by Fodor (1983).
Affect programmes exhibit several characteristics of Fodor's modular systems, as Griffiths (1997) has pointed out. Their operation is opaque, informationally encapsulated, mandatory, domain specific, and neutrally hard-wired in identifiable circuits and locations of the brain. Affect programmes are opaque as we can be aware of their inputs and outputs but not of the automatic appraisal processes themselves. Due to their informational encapsulation, automatic appraisals cannot access information that is stored in specific cognitive systems, and they can store information that contradicts consciously available or retrievable information. Informational encapsulation explains the mandatory and recalcitrant nature of many emotions: they cannot be chosen, prevented, or terminated at will, and they can persist even if we know that they are off the mark. This is possible because the appraisal system that evaluates the stimulus and determines a fixed response to it operates on infradoxastic representational items that are semantically less complex than propositional attitudes and conscious thoughts (see Charland, 1996). Finally, each affect programme operates in its own domain, defined by its characteristic "core relational theme", and there is evidence that basic emotions, such as fear, anger, and disgust, possess specific command circuits with distinct neurochemical coding systems in the mammalian brain (Panksepp, 1998).
Jaak Panksepp advocates the most sophisticated and ambitious version of contemporary affect programme theories in his view that there are seven fundamental emotional-motivational systems in subcortical regions of the mammalian brain: seeking, rage, fear, lust, care, grief/panic, and play. Each emotional system consists of neural networks that coordinate specific patterns of arousal, distinctive instinctual behaviours such as flight or freeze in fear and aggression in rage, as well as certain "raw" affective feelings that are closely enmeshed with the neural circuits that engender instinctual emotional behaviours. Phenomenally, emotional feelings are characterized by a certain valence, arousal, and power, but they do not relate to anything beyond the organism even if they are elicited by certain evolutionarily hard-wired environmental stimuli; hence their "raw" quality. Emotional affects are also "raw" in an evolutionary sense as their underlying neural networks reside in the subcortical areas of the brain that are among the most ancient in the brain. Coherent emotional behaviours and the adjacent raw feelings can be provoked by applying unpatterned electrical stimulation to the relevant subcortical area of the brain. The same responses remain intact also in animals that are decorticated at a young age. These observations provide Panksepp evidence for the independence of emotion from cognition, as no higher cognitive activity is required for the elicitation of emotions at the primary-process level that we share with non-human mammals. Panksepp's affect programme theory is based on his research of primary-process brain networks in the mammalian brain. However, he believes that the results of animal studies are relevant to human emotions as well by virtue of homologous subcortical networks in the mammalian brain.
Nevertheless, the main limitation of affect programme theories - including that of Panksepp - is their applicability to post-infantile human emotions. The problem is not that we do not have pure primary-process level emotions - hard-wired responses to unconditional stimuli - but that they constitute only a small minority of our emotional episodes, as Russell (2012a, 282) points out. Panksepp admits this point in his notion of tertiary-level emotions that emerge when our unconditional emotional-affective responses are blended with learning, conceptual thinking, and cultural norms. Indeed, Panksepp argues that the emotional system in humans extends beyond primary-process brain networks, involving interrelations between emotion and cognition in both directions: emotion controlling cognition and cognition instigating emotion. He even writes that "in the intact adult human brain, the relations between affect and cognition are enormous and almost impossible to disentangle" (Panksepp, 2012, 59). Yet he sees the primary-process level emotional circuits as underlying mechanisms that also control emotions at higher levels. However, psychological constructionists reject this assumption about the noncognitive core of all human emotions, maintaining that empirical evidence does not support the existence of emotion-specific subcortical neural networks any more than it supports the existence of emotion-specific facial expressions, autonomic nervous system responses, action tendencies, or subjective experiences (see Russell, 2012a, 281 for a summary, and Lindquist et al., 2012 for a meta-analysis of brain studies). Accordingly, constructionists argue that the human brain typically processes emotions holistically rather than in a bottom-up fashion as affect programme theorists suggest.
Considerations of modularity provide further arguments against the wide applicability of affect programme theories. Even if some human emotions are strongly modular in the Fodorian sense, the global validity of affect programme theories is undermined by the fact that many human emotions are weakly rather than strongly modular. The pervasive regulation of emotional responses before, during, and after their generation shows that their malleability far exceeds that of sense perceptions, Fodor's paradigmatic modular states. The apparent similarity between sense perception and human emotion emerges from the fact that all human emotions are capable of persisting in the face of rational counterevidence when we are in their grip, so to speak, before contrary information is taken in, as I pointed out above. Yet this weak modularity of ordinary human emotions does not render them similar to strongly modular sense perceptions and evolutionarily primitive emotions. "Emotional lag is less like visual illusion than it is like the ordinary cognitive phenomenon of beliefs outlasting acceptance of the evidential base on which they were first formed" as Karen Jones (2007, p. 23) remarks. It appears to me that this claim applies, with reservations, even to those pathological emotions whose dysfunction lies in the context of emotion regulation rather than in the underlying organismic structures and pathways. Contrary to persistent perceptual illusions, such as the Muller-Lyer illusion, there is a real possibility of bringing pathological emotions of the former kind in line with our considered judgments in therapies that may involve several types of treatment; cognitive, behavioural, and pharmaceutical (e.g. Leahy, 2004; Young, Klosko, & Weishaar, 2003; Dobson, 2001). This is the main reason for regarding those emotions as weakly or moderately rather than strongly modular.
Griffiths' (2004) own example of an affect programme emotion, anger aroused by being unexpectedly poked in the small of the back in a nightclub queue, is a case in point. Griffiths suggests that the event launches an affect programme of anger that prepares the agent for an aggressive behavioural response. However, the interpretation of the situation is defused and the response aborted as soon as one turns around and finds out that the "hit" was actually caused by the stumbling of a feeble old man. If this is an example of a "pure" affect programme, as Griffiths maintains, then it testifies against strong modularity of subpersonal emotional processing instead of supporting this view, for "one of the basic principles of automaticity contends that genuinely automatic processes cannot be disrupted by controlled processes", as Lieberman (2003, p. 55) points out.
An affect programme theorist may object that instrumental behaviour is not part of emotion, unlike the "square-mouthed" facial expression of anger that the subject expresses while turning around with clenched fists. However, we can grant this concession to Griffiths without defusing the argument because it is important to notice that in addition to the abortion of instrumental behaviour, the entire response would not have arisen in the first place if full perceptual information about the situation had been available to the agent during emotion generation. Visual information about the feeble old man would have supplemented tactile information, overriding the unfitting automatic appraisal of an intentional offence as soon as such an appraisal had been made. Indeed, if the automatic affect programme appraisals can be overridden by slower and more careful semantic appraisals before these programmes run their course from elicitation to full-fledged emotional response, these emotions are weakly or moderately rather than strongly modular, and therefore significantly dissimilar from sense perceptions. Accordingly, subpersonal and subdoxastic content may accommodate the emotions of subjects whose personal-level semantic processing is hindered for developmental, psycho-pathological, or physiological reasons but it is hardly capable of accommodating all human emotions.
-  Ekman emphasizes that an "appraisal is not always automatic. Sometimes the evaluation of what is happening is slow, deliberate, and conscious" (Ekman 1977, 59; quoted from Ekman 1999, 51). Ekman associates his view of emotional appraisal with multi-level appraisal theories and Lazarus (1991) in particular. However, extended appraisals are not part of basic emotions and their role in Ekman's theory remains open because he explicitly states that there are no "non-basic" emotions.
-  The argument that modularity in emotion is a matter of degree was first presented by de Sousa (1987). For more recent discussions on the topic, see Faucher & Tappolet (Eds.) (2007).