39 Feeling the Future
Autonomic Prediction, Interoception, and Embodied Choice
39.1 A future can change the body before it arrives
Chapter 38 ended with an already embodied control state: selected information remained effective while the animal moved and its bodily needs continued to change. We now follow one part of that interaction more closely. How does an anticipated consequence prepare the body, and how does the resulting bodily information change ongoing control?
A thirsty animal reaches the edge of cover and sees water across open ground. The water is real, but the attack it risks is not. Nothing has yet moved in the grass. Even so, the exposed route can alter vigilance, posture, respiration, cardiovascular state, and readiness to retreat. These changes are not reactions to an attack. They are preparations for a possible future. They also become part of the choice: a route that reorganizes the body as dangerous does not compete on the same terms as one that leaves the animal prepared to continue.
Humans extend this prospective control far beyond the next few seconds. An imagined argument, accident, demanding performance, or long-awaited reunion can change the body before the event occurs. Experiments using emotional imagery have measured changes in cardiovascular and electrodermal activity while participants imagine emotionally significant situations [@schwartzetal1981imagery; @langetal1983fear]. The response varies with the person, the imagined content, and the task; it need not reproduce the full response to an actual event. What matters here is that a represented possibility can have a physiological consequence now.
This is not a separate form of valuation added after the fact. Chapter 37 defined value as the current weight of a predicted outcome for an organism in a present and anticipated state. It already included delay, effort, risk, and future bodily need. Chapter 38 then explained how selected information and priorities remain effective across time. The present chapter asks how a represented outcome enters autonomic, interoceptive, and affective state, and how those changes feed back into what the available actions are worth.
The problem is allostatic: regulation must take account of expected demands. A controller that waited for every consequence to become present would often regulate too late. It must prepare circulation, respiration, attention, posture, and action before the demand arrives. This predictive capacity did not originate with human deliberation. As the anticipatory hunger example in Figure 37.8 showed, older regulatory circuits already respond to conditions that predict a later need. Frontal participation extends the remembered and inferred circumstances that can influence such preparation (Section 36.1).
A represented future can change autonomic, interoceptive, and affective state before it occurs. Those changes feed back into valuation and help determine which action gains control.
The somatic-marker hypothesis gave this idea an influential neurological form. Damasio proposed that learned body-related signals, expressed peripherally or represented centrally, help constrain reasoning and choice [@damasio1996somatic]. One especially influential interpretation of the Iowa Gambling Task went further: nonconscious bodily signals guide advantageous choice before usable explicit knowledge is available. I take the embodied control problem seriously without treating that stronger temporal claim as established. Whether an anticipated event changes the body, whether bodily information influences choice, and whether it does so before reportable knowledge are different empirical questions.
39.2 Bodily response, interoception, affect, feeling, and value are not synonyms
The vocabulary surrounding embodied choice is easily compressed into the single word feeling. That compression hides the mechanism. Several variables must be kept apart.
An autonomic response is a change produced through sympathetic or parasympathetic control of organs and tissues. Heart rate, cardiac contractility, vascular resistance, pupil diameter, gastrointestinal activity, and sweating can all change as the organism prepares for what is expected. Endocrine secretion, immune activity, respiration, and skeletal-muscle readiness also contribute to the broader bodily state, although they are not all autonomic responses in the narrow sense.
Interoception is the nervous system’s sensing, integration, and representation of conditions within the body. It includes afferent evidence from organs, blood vessels, and tissues, together with central processing of what that evidence means in the current situation. Much of this processing supports regulation without reportable awareness. Conscious attention to a heartbeat is one interoceptive task; it is not a measure of the whole system.
Affect refers here to the organism’s broad orientation toward its current situation: attraction or aversion, activation or quiet, readiness to approach, withdraw, persist, or interrupt. A conscious feeling is the experienced and reportable aspect of that state. Affect can alter attention and action without becoming a clearly named feeling, and a conscious feeling incorporates context, memory, concepts, and interpretation in addition to peripheral physiology.
Subjective value is different again. It is the current influence of a predicted outcome on choice. Bodily and affective states help determine that influence, but value is not identical to heart rate, skin conductance, pleasure, or conscious preference. A threatening event can be highly arousing and strongly negative. A difficult performance can be highly arousing and positively valued. The same physiological direction can accompany different actions because physiology prepares the organism; it does not carry a complete semantic label for the situation.
The distinction is especially important for electrodermal activity. A skin conductance response reflects changes in eccrine sweat-gland activity under sympathetic control. It is one peripheral measure, not a meter of the organism’s complete autonomic state. Responses occur in many settings involving orienting, attention, effort, anticipation, or emotion, and their interpretation depends on recording conditions and timing [@boucseinetal2012electrodermal]. A larger response has no intrinsic positive or negative sign. Before calling it a warning, we must establish what it predicts about the option and the behavior that follows.
A somatic marker, in Damasio’s formulation, is a body-related state associated through learning with an outcome or class of outcomes. It can be enacted in the peripheral body, represented centrally, consciously felt, or expressed without a clear conscious feeling. That breadth made the hypothesis powerful, but it also means that several distinct causal claims can hide inside the same phrase.
- Does an anticipated outcome elicit a bodily response?
- Does that response covary with later choice?
- Does manipulating bodily input or its central representation change behavior?
- Is peripheral feedback necessary for performance in this particular task?
- Does a bodily influence on choice precede usable explicit knowledge?
These questions move from an observed response to progressively more specific causal and temporal claims. Necessity requires an intervention that tests the relevant feedback, not merely a correlation between physiology and choice. The last question also requires sensitive, repeated knowledge measurements. Universal dependence on fresh peripheral feedback is not a requirement of Damasio’s formulation: its as-if loop explicitly allows central body-state representations to contribute. Evidence about one question should not be treated as an answer to all five.
39.3 How preparation reaches the body and returns
The outgoing routes were introduced in Section 36.6. Selected orbital and medial frontal territories project to hypothalamus and to the periaqueductal gray, providing access to systems that organize autonomic and behavioral preparation. Rhesus tracing demonstrated frontal synapses in hypothalamus and mapped onward hypothalamic projections toward brainstem and spinal autonomic regions [@barbas2003autonomic]. This supports a route through intervening regulatory circuits; it does not establish every proposed link as a synapse on the same identified neuron. The final influences on heart, blood vessels, sweat glands, and other organs pass through the relevant autonomic effector pathways. Endocrine and skeletal-motor adjustments involve additional outputs.
The returning limb is sensory. Visceral afferents carried in the glossopharyngeal and vagus nerves reach the nucleus of the solitary tract (NTS) in the medulla [@becksteadnorgren1979afferents]. In monkeys, tracing from caudal NTS identified projections to medullary regions involved in autonomic control and to the parabrachial nuclei [@becksteadetal1980solitary]. These are parts of the returning and regulatory circuitry, not a single cable from an organ to a feeling. Other bodily afferents and circulating chemical signals provide additional information. The broader routes into hypothalamic, thalamic, insular, and frontal systems were introduced in Section 36.4.
At the cortical level, insular and somatosensory activity can contribute to the representation of bodily condition, while insular–orbital connections allow that information to interact with expected outcomes [@saleem2008complementary]. In a human fMRI experiment, attending to the timing of one’s heartbeats engaged insular, cingulate, and other regions; right anterior insular/opercular activity was related to heartbeat-detection accuracy [@critchleyetal2004interoceptive]. This is evidence about a particular interoceptive judgment, not proof that the insula is the sole source of feeling or that every bodily signal becomes conscious.
Bodily timing can also change how an external event is processed. Garfinkel and colleagues presented emotional faces at different phases of the heartbeat. Fearful faces were more readily detected, and in a separate task rated as more intense, at systole than at diastole; the imaging experiment also found a corresponding modulation of amygdala responses [@garfinkeletal2014heart]. Systole is associated with stronger arterial baroreceptor signaling. The experiment supports a contribution of ongoing cardiac state to threat processing, although it did not selectively manipulate the afferent pathway or test gambling decisions. It gives bodily feedback a specific experimental role without asking it to explain every choice.
Preparation therefore changes more than an output measure. It changes the organism from which the next sensory evidence will arrive. That is the returning limb of the control problem: what the nervous system does to the body can change what it subsequently makes of the world.
39.4 The somatic-marker hypothesis
The somatic-marker hypothesis grew from a clinical fact that older divisions between reason and emotion could not explain. Some patients with orbital and ventromedial frontal damage retained language, factual knowledge, and high scores on conventional tests yet repeatedly made choices that damaged their work, finances, relationships, and safety. Their deficit was not a general collapse of intelligence. It appeared when knowledge about consequences had to acquire enough current weight to govern a real course of action [@eslingerdamasio1985evr].
Damasio’s central proposal was that reasoning over possible futures is ordinarily constrained by learned body-related signals. Experience links situations and outcomes to patterns of autonomic, endocrine, motor, and affective response. When a similar possibility is considered later, part of that pattern is reinstated. The reinstated state changes attention, narrows or expands the set of options under serious consideration, and biases the competition among actions. A dangerous alternative can be rejected before every consequence is recomputed in detail; a promising alternative can gain priority before its reward is present [@damasio1996somatic].
This was an important reversal of the traditional story. Emotion was not an intrusion into an otherwise rational machine. Bodily and affective signals summarized what previous encounters had meant for the organism and made that history available to current control. Deliberation without such weighting would not become perfectly logical. It would become inefficient and poorly anchored to the consequences that matter.
Damasio distinguished a body loop from an as-if body loop. In the body loop, a represented outcome changes the peripheral body, and afferent signals return evidence about the resulting state. In the proposed as-if loop, central systems reinstate or modify a body-state representation without requiring the corresponding peripheral pattern to unfold first. This provides a possible shortcut through learned central representations, rather than a requirement to reproduce each bodily response afresh. The distinction belongs to the hypothesis; an anticipatory skin conductance response alone cannot identify which central mechanism is operating.
The same framework distinguished primary inducers, emotionally significant stimuli or events encountered directly, from secondary inducers, memories, thoughts, and represented possibilities that evoke a body-related state. The distinction concerns how a response is elicited, not whether its significance was innate or learned. In an early lesion study, both amygdala and ventromedial frontal groups performed poorly on the gambling task and failed to develop anticipatory skin conductance responses. The ventromedial group nevertheless produced responses to obtained rewards and punishments and acquired a conditioned response to a visual cue paired with a loud sound; the amygdala group was impaired on these measures [@becharaetal1999amygdala]. This pattern helped motivate different proposed contributions to eliciting and reinstating body-related signals. It does not divide the brain into a structure for the present and a structure for the imagined, nor does it show that the amygdala is required for every autonomic response.
The anatomy also resists a single-module account. As Section 36.3 established, human orbital and medial cortex contains multiple granular, dysgranular, and agranular territories [@onguretal2003architectonic]. The regional term vmPFC should not erase their differences or the extent of an individual lesion. Amygdala–frontal interaction also has several anatomical routes. Diffusion tractography in humans and macaques distinguished the courses of the uncinate fasciculus and ventral amygdalofugal pathway [@follonietal2019bundles]. These are complementary frontotemporal routes, not one dedicated emotion cable. Tractography estimates bundle organization; it does not by itself establish direction, synapses, or the information carried.
The hypothesis is strongest when expressed as a recurrent mechanism. A possible outcome changes activity in systems that represent bodily regulation, affective significance, context, and action. Those changes alter the value and priority of the possible outcome. No inner observer inspects a simulated scene and reads a bodily meter. The represented outcome itself reconfigures the controller. “Feeling the future” names this reconfiguration, not a homunculus entering an internal theater.
39.5 The Iowa Gambling Task made the hypothesis testable
A broad theory becomes scientifically useful when it predicts measurements. The Iowa Gambling Task was designed to place uncertain, temporally extended choice in a laboratory and to measure behavior, autonomic response, and explicit knowledge as learning unfolded [@becharaetal1994future; @bechara1996autonomic].
Participants choose repeatedly from four decks of cards, learning their payoffs through experience. Every selection produces a nominal gain; some also impose a penalty. The familiar fixed schedule can be summarized over a scheduled block of ten selections from each deck [@becharaetal1994future; @becharaetal1999amygdala]:
| Deck | Gain on every selection | Cards carrying a penalty per ten | Total penalties per ten | Net return per ten |
|---|---|---|---|---|
| A | $100 | 5 | $1,250 | −$250 |
| B | $100 | 1 | $1,250 | −$250 |
| C | $50 | 5 | $250 | +$250 |
| D | $50 | 1 | $250 | +$250 |
These are schedule totals, not a guarantee for every arbitrary run of ten trials across mixed decks. Versions of the task differ, so an experiment’s actual schedule matters.
A and B are disadvantageous by cumulative return, whereas C and D are advantageous. But the task also varies penalty frequency: A and C carry frequent penalties, B and D infrequent penalties. A penalty is a deduction from that card’s nominal gain; it need not produce a net loss on that selection. Some penalties cancel the gain or leave a smaller gain. Cumulative return, penalty frequency, and the frequency of net winning or losing trials must therefore be distinguished. The decks do not isolate a single dimension of goodness.
The exact contingencies are not announced. Participants sample the decks while the experimenter records their choices. In the physiological studies, electrodes measure skin conductance. An anticipatory response is operationally defined from a period before a recorded deck selection; an outcome response is measured after the gain or penalty. Periodic questions assess reportable knowledge. Skin conductance develops slowly, so the interval between trials and the analysis windows matter: activity caused by a previous outcome can overlap the period preceding the next choice [@boucseinetal2012electrodermal]. A response preceding the recorded selection is not necessarily earlier than the participant’s internal commitment to it.
The original findings suggested a compelling sequence. Healthy participants increasingly favored advantageous decks and developed larger anticipatory skin conductance responses before disadvantageous choices. In the original knowledge probes, this differentiation appeared before participants reported a correct account of the strategy. Bechara and colleagues described the result as “deciding advantageously before knowing the advantageous strategy” [@becharaetal1997advantageously]. The memorable interpretation was that the body knew first: a covert somatic signal warned against the bad decks before the participant could explain the choice. Whether the questions detected all usable knowledge became central to the subsequent debate.
Patients with bilateral orbital and ventromedial frontal damage showed a different pattern. As a group, they favored disadvantageous decks and showed weak or absent anticipatory responses. They could still respond electrodermally to obtained rewards and penalties, so the deficit was not a general inability to generate skin conductance responses [@bechara1996autonomic; @becharatraneldamasio2000characterization]. The combination implicated a contribution of the damaged region to anticipation and adaptive choice. It did not isolate the missing autonomic response as the cause of the behavioral impairment.
The experiment joined three observations that had previously been separated: real-world decision failure after focal frontal injury, a laboratory measure of learning about cumulative consequences, and a physiological signal measured before an action. It established that orbital and ventromedial injury can disrupt both anticipatory autonomic responding and adaptive choice. It also invited a stronger hypothesis: a bodily signal carries information that has not yet become explicitly available.
39.6 What the gambling task does—and does not—show
The Iowa Gambling Task linked frontal injury, autonomic anticipation, and impaired learning about the consequences of choice. The effects depend on the sample, task version, and measurement, and the task does not isolate one hidden mechanism. Its strongest original interpretation depended especially on how knowledge and skin conductance were measured.
Four streams of evidence must remain separate. Choice behavior shows which decks increasingly control action. Anticipatory skin conductance shows sympathetic activation before a selection. Outcome-evoked skin conductance shows activation after a gain or loss. Knowledge probes show what the participant can report about deck contingencies. A temporal ordering among these measures is meaningful only when each is sampled with enough sensitivity.
The original questions divided learning into broad periods: participants first had no idea, then a “hunch,” and eventually a conceptual understanding. These classifications could miss partial but usable knowledge. A participant need not explain the full payoff schedule to identify a preferred deck. Maia and McClelland used more specific questions about expected gains, penalties, and deck evaluations. Their participants reported knowledge sufficient to support advantageous behavior more often and earlier than the original probes suggested [@maiamcclelland2004]. This study measured knowledge and behavior, not concurrent skin conductance. It challenged the evidence for absence of knowledge; it did not establish a new physiological sequence or show that emotion was irrelevant.
Fernie and Tunney combined skin conductance recording with knowledge probes after trial twenty and then every ten trials. Most participants had usable reportable knowledge after approximately forty trials. The investigators found no reliable anticipatory differentiation between advantageous and disadvantageous decks before that knowledge emerged, although some outcome responses differentiated deck types [@fernietunney2013learning]. This failed to support the proposed covert anticipatory lead. A null result in one version does not exclude every bodily contribution, and intermittent questioning cannot pinpoint the instant knowledge became available. The claim that advantageous choice occurs because the body knows before the person does remains unsupported by these measures alone.
The structure of the decks adds a second complication. Deck B loses money over the schedule but yields a net gain on most individual selections. Lin and colleagues used simplified AACC and BBDD versions that compared decks within penalty-frequency classes. Participants could continue selecting B even after experiencing its large penalties [@linetal2007deckb]. This prominent Deck B phenomenon shows why grouping A and B as “bad decks” can conceal behavior sensitive to outcome frequency. It does not show that nobody learns cumulative return; it shows that the usual net score combines distinct influences.
The task also requires participants to overcome an early attraction to high-paying options as penalties accumulate. Fellows and Farah found impairment after both ventromedial and dorsolateral frontal lesions, with the ventromedial group’s performance particularly related to reversal-learning difficulty. Reordering the cards to reduce the initial misleading advantage of the high-paying decks improved ventromedial performance [@fellowsfarah2005underlying]. This implicates updating an initially favored option, without making reversal the sole explanation of every gambling deficit. More generally, different sampling histories, memory failures, or interpretations of the task can produce similar net scores.
Skin conductance presents a third interpretive limit. A larger anticipatory response before a disadvantageous deck is consistent with a warning signal, but it is not equivalent to one. It can reflect uncertainty, conflict, attention, anticipated magnitude, attraction to a large gain, or preparation for a consequential outcome. The response acquires meaning from its relation to deck history and later behavior. It does not directly encode “negative value” in the skin [@boucseinetal2012electrodermal].
Finally, correlation does not establish the direction of control. If a participant learns that a deck is dangerous, that knowledge can change both skin conductance and choice. The physiological response could contribute to the choice, express the same learned state, or both. Lesions that reduce both measures are also not decisive by themselves because orbital and ventromedial injury disrupts several interacting operations: outcome learning, reversal, state inference, future construction, value representation, autonomic control, and the use of context. A lesion does not remove one variable while leaving the rest of the network unchanged.
These qualifications do not make the Iowa Gambling Task uninformative. They clarify what its different measures can establish. Anticipatory autonomic responses can accompany learning about consequential options; orbital and ventromedial injury can disrupt both anticipation and adaptive choice. The task does not by itself establish that a peripheral signal is necessary for successful performance, that it encodes negative value, or that it guides choice before usable knowledge exists. Those are separable hypotheses, not interchangeable descriptions of one result.
The task’s lasting value is methodological as much as theoretical. It shows why a complex “frontal” test must be decomposed. Learning, exploration, gain frequency, loss magnitude, cumulative return, reversal, explicit knowledge, autonomic anticipation, and action all contribute to performance. The science advances when these variables are measured separately rather than when the total score is assigned to one faculty.
39.7 When knowledge is available but does not govern choice
The most durable insight from the lesion literature is not that reason survives intact while feeling disappears. It is that information can remain available for report without gaining enough control over behavior.
Patient EVR made this dissociation clinically unmistakable. After bilateral removal of orbital and lower medial frontal tissue, his measured intelligence remained superior and standard neurological and neuropsychological examinations showed many preserved abilities. Yet he could no longer organize his personal and professional life. Decisions that appeared simple in discussion led repeatedly to delay, poor judgment, lost employment, financial damage, and failed relationships [@eslingerdamasio1985evr]. A conventional test battery underestimated the disability because the central failure appeared when consequences had to be integrated across time and made effective in self-directed action.
The Iowa patients displayed a related pattern in a more constrained setting. Some could describe that particular decks were risky or disadvantageous and still continued to select from them. The proposition was verbally accessible. It did not reliably reorganize choice. That observation matters even after the strongest “body before knowledge” claim is set aside [@becharaetal1997advantageously].
Chapter 38 developed the same distinction through goal neglect. A person can repeat a rule and then fail to let it structure later behavior. The rule is psychologically available but behaviorally weak. The analogy does not imply that goal neglect and ventromedial decision failure are the same syndrome. It identifies a common control principle: representation is not enough; a representation must alter the state of the system that selects action.
Several operations stand between knowing and doing. The expected outcome can be represented too vaguely. Its value can fail to update when contingencies change. Immediate reward can dominate when later losses are poorly integrated. The current situation can fail to retrieve the relevant prior experience. Competing actions can receive inconsistent weights. The patient can also construct a thin or disorganized future rather than a vivid future that has simply lost its emotional tag. Bertossi and colleagues found impoverished descriptions of remembered and imagined personal events after vmPFC damage, implicating event construction as well as evaluation [@bertossi2016future]. This is a further experimental group, not another measurement from EVR or the original gambling patients.
This broader account preserves the force of the clinical evidence. Orbital and ventromedial cortex contributes where knowledge about outcomes must be integrated with current state, context, affective significance, and action. Damage can leave many facts and skills intact while weakening their influence on a sustained course of action. The earlier selective-devaluation evidence already separated changing an outcome’s value from using that change to guide choice (Figure 37.6). The clinical cases extend this distinction into less structured behavior. The result is not pure reasoning stripped of emotion. It is a controller whose components no longer cooperate effectively.
The body remains part of that cooperation. Autonomic and interoceptive changes alter attention, urgency, effort, and the perceived significance of an option. Conscious unease can interrupt a course of action; a subtler bodily state can bias sampling or persistence without being named. Yet bodily signals do not replace knowledge. Adaptive choice emerges when explicit propositions, learned affect, represented outcomes, current bodily state, and action competition constrain one another.
Knowing a consequence and being governed by that consequence are different achievements. Embodied valuation helps turn the first into the second.
39.8 Prospective valuation belongs to a distributed loop
A possible future acquires detail while the organism is already valuing, attending, and preparing. Remembering and imagining recruit overlapping hippocampal, medial temporal, posterior cortical, and medial prefrontal systems [@addisetal2007remembering]. Their interactions allow prior experience to contribute to a new possible event. We will examine constructive memory more fully in Unit 7. Here the relevant point is that a completed scene need not be delivered to a separate evaluator: current need and affect can influence which details are retrieved, while emerging details alter bodily preparation and the next act of sampling.
The anatomical framework is already in place in The Organ of Tomorrow. Its incoming routes connect bodily state, learned cues, and context to differentiated frontal territories; its recurrent circuits connect those territories with thalamic and basal-ganglia systems (Section 36.4; Section 36.5). A Common Currency showed how outcome identity, current value, and the inferred situation remain related. Holding the Line showed how selected information can remain effective while being revised. These are interacting contributions within prospective control, not completed stages that precede embodiment.
For this chapter, the crucial addition is the exchange with bodily regulation. Amygdalar and insular systems interact with orbital and medial frontal cortex as learned significance and bodily condition change. Hypothalamic and brainstem circuits help organize preparation. That preparation supplies new afferent evidence, which can alter attention and the competition among actions. The cortical and subcortical action routes in Section 36.6 allow these changing conditions to affect what the animal does. No region needs to finish its assigned psychological operation before the others become involved.
The body-loop/as-if distinction describes two proposed ways body-related information can participate in this exchange. Actual peripheral change provides new evidence about the organism; a centrally reinstated pattern can influence control without waiting for a matching new peripheral response. Both remain constrained by the body’s ongoing state. The hypothesis does not make a centrally represented body disembodied, and the presence of a peripheral response does not prove that its feedback determined the choice.
The returning bodily evidence does not arrive as a verdict. A racing heart can accompany fear, exertion, anger, excitement, or a mixture of them. Its influence depends on the situation represented by the rest of the network. Interoceptive signals become meaningful through their relation to context, memory, expectation, and available action. Embodiment therefore does not reduce choice to the viscera. It means that the state of the viscera is one of the variables from which the controller constructs significance.
Prospective thinking can measurably change choice. Peters and Büchel presented personally relevant future-event cues during choices between sooner and later monetary rewards. On average, participants discounted delayed rewards less in the episodic condition, with differences related to reported imagery; neural coupling between anterior cingulate and hippocampal regions was associated with the behavioral effect [@petersbuchel2010episodic]. This connects prospective content to present choice. It does not demonstrate mediation by autonomic feedback, and functional coupling does not establish a direction of neural influence. Enriching a future and giving that future bodily consequences remain related but experimentally distinguishable questions.
Return to the animal at the edge of cover. Thirst changes the significance of the visible water. An exposed stretch can evoke preparation for danger before an attack occurs; that preparation can increase vigilance and alter which environmental details the animal samples. A new sound, a change in breathing, or a newly noticed sheltered route can revise the expected consequences while the choice is still forming. Once the animal moves, visual and bodily feedback continue to change the same control state. There is no single pass through construction, valuation, feeling, and action. There is a continuing exchange in which each can change the conditions governing the others.
This is the control-theoretic core of feeling the future. Prediction prepares the organism for a state that does not yet exist. Feedback from the preparation changes the prediction’s practical meaning. Choice emerges from the recurrent settling of a distributed controller, not from a contest between a rational cortex and an emotional body.
39.9 Coda: another person’s future
The examples emphasized here have concerned the organism choosing: its water, danger, money, effort, delay, or loss. Human social life extends the same prospective architecture to outcomes experienced by other agents. Another person’s anticipated pain, relief, fear, disappointment, or gratitude can change the observer’s present affective and bodily state before anything happens. That change can alter what the observer is prepared to do.
Social control adds variables that the gambling task does not contain. The nervous system must represent whose outcome is at stake, what the other person knows, what they intended, whether the result was accidental or deliberate, what relationship exists, which norms apply, and what one’s own action will do to an autonomous agent. Language and culture make many of these variables explicit and extend their reach across institutions, reputations, promises, and obligations.
No single cortical region performs morality, and conscience is not a somatic marker pointed outward. Social conduct recruits the embodied valuation system described here, but it also depends on memory, perspective, causal inference, learned norms, language, and control of action. The next chapter follows prospective control into this interpersonal domain: the futures that matter because they belong to someone else.
Established findings. Anticipated and imagined outcomes alter autonomic and affective state before the outcomes occur. Skin conductance is a measure of sympathetic activation, not a direct measure of positive or negative value. Orbital and ventromedial frontal damage can impair flexible outcome-guided choice, anticipatory autonomic responding, future-event construction, and real-world organization while sparing many conventional intellectual and declarative abilities. Performance on the Iowa Gambling Task reflects several operations, including learning, exploration, gain–loss frequency, reversal, explicit knowledge, autonomic anticipation, and action selection. Information that is available for report does not always exert effective control over behavior.
Working synthesis. Prospective valuation is embodied and recurrent. Need, remembered possibilities, bodily preparation, sampling, and action can revise one another throughout a choice. The somatic-marker hypothesis made a consequential proposal about how body-related signals participate. Its body-loop and as-if-loop distinction helps separate actual peripheral feedback from central reinstatement, but neither label by itself identifies the causal mechanism in a particular experiment.
Open questions. The causal importance of particular peripheral feedback signals across different decisions remains to be specified. Experiments must determine when central body-state prediction is sufficient, when peripheral change adds indispensable evidence, what an anticipatory skin-conductance response represents in a particular task, and how explicit knowledge and autonomic learning interact over time. The original Iowa Gambling Task does not answer these questions by itself. They require tasks that manipulate the component processes separately.