An automated reasoner that thrashes about, energetically producing irrelevant inferences without getting any nearer its goal, is reminiscent of the behaviour of a human agent with a particular form of brain damage. Patients who have suffered damage, perhaps as a result of tumour, to the ventromedial part of the frontal cortex, just above the nose and behind the eyes, are capable of reasoning just as well as before, but their decision-making is poor. This apparent paradox should be illustrated by an example. Damasio reports discussing with a ventromedial patient the scheduling of his next appointment: “I suggested two alternative dates, both in the coming month and just a few days apart from each other. The patient pulled out his appointment book and began consulting the calendar […] For the better part of a half-hour, the patient enumerated reasons for and against each of the two dates: previous engagements, proximity to other engagements, possible meteorological conditions, virtually anything that one could reasonably think about concerning a simple date” (Damasio 1994 p193).
The ventromedial patients can perform deduction perfectly, and even score as well as before the damage occurred on standard IQ tests. Nevertheless they are unable to hold down jobs, stay happily married, or even make simple decisions about when to see the doctor. To explain what has gone wrong in such cases, Damasio formulated the somatic-marker hypothesis. As background to the hypothesis, it should be noted that ventromedial patients have three features in common: damage to the ventromedial prefrontal cortex, a flattening of the emotions (particularly the social emotions), and an apparent inability to learn from experience. Let us look in turn at the role played by the ventromedial prefrontal cortex in emotions, at the evidence that ventromedial patients fail to learn from experience, and finally at the mechanism responsible simultaneously for relevance, analogy, and learning from experience.
Emotional responses are of various sorts. Primary emotions are those which are wired in at birth. For example: “A baby chick in a nest does not know what eagles are, but promptly responds with alarm and by hiding its head when wide-winged objects fly overhead at a certain speed” (Damasio 1994 p132). The amygdala plays a key role in primary emotions. Secondary emotions involve not only the amygdala but also the prefrontal cortices, and are learned (or at least refined by experience). When iconic representations in the early sensory cortices are processed, signals arise that activate dispositional representations in the prefrontal cortex which embody knowledge of how certain types of situations have been paired with certain emotional responses. The damage suffered by ventromedial patients prevents them from activating the dispositional representations that pair certain categories of situation with emotions. Specifically, disturbing images that produce feelings of disgust or horror in normal individuals leave ventromedial patients unmoved, although they can remember that previously they would have experienced a strong reaction.
Nor can new dispositional representations of this kind be acquired. When ventromedial patients and normal participants were asked to play a card game that involved learning from rewards and penalties, the outcomes were very different. Each player starts with some money and gets four decks of cards. Two of the decks are good, in the sense that they confer more rewards than penalties, while the other two are bad and contain penalties that outweigh the rewards. Players do not know this but must learn by sampling. The sampling is rigged (unbeknownst to the participants) so that the initial cards from the bad decks offer large rewards. Both normal participants and ventromedial patients therefore gain an initial preference for the bad decks. Subsequently, cards from the bad decks impose even larger losses, while cards from the good decks maintain a pattern of smallish gains and even smaller losses. Normal participants learn to avoid the bad decks; ventromedial patients do not. (Damasio 1994 pp212-217) While the ventromedial deficit may not involve a complete inability to learn from experience, it amounts at least to an inability to revise beliefs in the face of new evidence.
The somatic-marker hypothesis is that emotions are connected, by learning, to the predicted future outcomes of certain scenarios and that this serves to increase the accuracy and efficiency of decision-making. In this view, decision-making is not disembodied; instead biological information steers the decision-making process towards outcomes that are advantageous, based on past experiences with similar situations.
If a dispositional representation is to connect an emotion to the iconic representation of some event, then emotions must be representations too. But what representations are these? The answer lies in the notion of a body map.
The survival of the organism requires, for homeostasis, the co-ordination of the myriad body functions on which life depends. Energy sources must be found, energy incorporated, an interior chemical balance maintained, wear and tear repaired, and external agents of disease or injury fended off. In order for the brain to co-ordinate these activities, it needs to have information about what is going on in different body sectors so that various functions can be slowed down, or halted, or called into action. The brain needs, in short, to have maps in which the state of each body sector is represented moment by moment. The brain’s somatosensory complex represents the body by reference to a body schema having midline parts (trunk, head), appendicular parts (limbs), and a body boundary (skin). The somatosensory complex is distributed over several areas (the insular cortices, S1, S2) and concentrated in the right hemisphere. The body maps include a ‘primordial representation’ against which the current state of the body can be compared. Such comparisons allow a register of changes to the body to be kept up to date. This includes a representation of the outside world in terms of the modifications it causes in the body (Damasio 2003 pp96-118, Damasio 1994 p230/1).
Let us term any change to the body map(s) an affect. Some changes set off alarm bells (metaphorically speaking); these are the negatively valenced affects. Some changes restore the body map to something more closely resembling the primordial representation; these are the positively valenced affects, those we may experience as well-being. Every significant change to the body map gives rise to signals that go to the body (including the brain) in order to orchestrate a response to the change in body map. Signals via peripheral nerves place the viscera (e.g. heart) in the state most commonly associated with the type of triggering situation; signals to the motor system prompt skeletal muscles to adjust body posture or facial expression; endocrine and peptide systems are activated and their chemical actions result in further changes to body and brain states; neurotransmitter nuclei are activated (Damasio 1994 p138). One consequence of the neurotransmitter responses is a change in the speed with which mental images are formed, discarded, attended, evoked, as well as a change in the style of thinking that operates on those images (Damasio 1994 p163). The effect is to associate with the triggering affect a momentary thought-action repertoire. To illustrate, consider affects that correspond to specific emotions: fear is linked with the urge to escape, anger with the urge to attack, disgust with the urge to expel. These urges are not invariably translated into action; rather, people’s ideas about possible courses of action narrow in on a specific set of behavioural options. The resultant thought-action repertoire aids decision-making by, as it were, shining the light on a subset of options and eliminating the remaining options by simply removing them from consideration. Reasoning has been equipped with a notion of relevance based on affect, which in turn is based on embodiment.
The presence of somatic markers, deriving from changes to the body maps, supports a kind of cognition quite distinct from deduction. Instead of a slow, step-by-step process which is intentional, controllable, consciously accessible and viewable, sequential, and demanding of attentional resources, the system based on somatic markers typically operates swiftly and effortlessly, unintentionally and automatically, without demanding attentional resources. We may call the system based on somatic markers intuition, because only the results of processing enter awareness, not the intermediate steps. Intuition is basically a one-step process as far as conscious awareness is concerned. Not only is a notion of relevance built into intuition, this notion is more general than ‘relevance as attention-paid-to’, because the operation of somatic markers can be unconscious.
Finally, the operation of intuition does not exclude the possibility of symbol manipulation, but clearly does not require it, and instead offers a natural form of analogical ‘reasoning’ by virtue of the links between a categorical representation and its subsidiary iconic representations. The human brain is a semantic engine, rather than the syntactic engine envisaged by Dennett.