Body loop and as-if body loop
The most consequential idea in the somatic-marker-hypothesis for a wiki about interoception, because it is where the hypothesis specifies how much body a somatic marker requires. From Bechara & Damasio (2005), Fig. 2 — and, nine years earlier and with a different answer, Damasio (1996).
Read the 1996 section below before the 2005 material. The wiki built this page from the later source and framed the as-if loop as a structural weak point that the 2005 indexing claim rescues. With both timepoints in hand, that framing is incomplete: in 1996 the as-if loop is not a weak point being exploited, it is the mechanism being advertised.
The two loops
When the VM cortex activates somatic effectors in hypothalamus and brainstem — attempting to reconstitute a somatic state that belonged to an earlier exteroceptive–interoceptive conjunction — two chains of events are possible.
The body loop. The somatic state is actually re-enacted in the body proper. Signals from its activation are relayed back to subcortical and cortical structures, “especially in the insular and SII and SI cortices.” Many channels carry body information centrally — spinal cord, vagus nerve, humoral signals — and the paper claims the vagal route is especially critical (Bechara 2002), corroborating older evidence for the vagus in emotional modulation of memory (Roozendaal et al. 1996). This is the straightforwardly Jamesian arm.
The as-if body loop. The body is bypassed altogether. Because expressed somatic states leave patterns in brainstem nuclei and insula/SII/SI, those patterns can be activated directly, inducing changes in neurotransmitter release “without engaging the body.” The somatic signal is real; its peripheral origin is not. Somatic states are “simulated” intra-cerebrally.
Damasio’s own gloss elsewhere in the paper: among the responses an emotion directs at the brain is “an active modification of the state of somatosensory maps such as those of the insular cortex (as-if-body-states).“
1996: the as-if loop is the default, and the periphery is written out
The earlier primary source, Damasio (1996), contains both loops in recognisable form and orders them differently. The body loop is the original mechanism from both evolutionary and ontogenetic perspectives — and has been superseded by the as-if loop, which Damasio judges the more frequently used of the two.
That is not a claim about availability. It is a claim about replacement: the peripheral route is the ancestral implementation, and the simulation is the upgrade that retired it.
Then Everitt asks the question this page exists to ask. In the transcribed discussion following the paper, B. J. Everitt puts it directly: to what extent do peripheral somatic changes contribute to the risk-taking in the gambling test — if you manipulated those peripheral changes experimentally, would you expect behaviour to change?
Damasio’s answer is that in the more frequently operational mode, changes in the periphery ought to have relatively little impact on central functions related to emotion/feeling and reasoning/decision making. In most circumstances the as-if loop is engaged, the periphery is bypassed entirely, and only the neural structures representing body states are modified.
Three qualifications are offered and all three are worth keeping:
- The role of peripheral signals varies between developmental and adult phases. During development, and in non-average situations, body states are actively engaged and very much part of the emotion/feeling loop.
- Evidence from cord damage and peripheral neuropathy suggests there may be some effect of the periphery on central processes — despite, Damasio notes, the obligate incompleteness of such lesions.
- The final answer requires further study.
Why this is the most important thing on this page. The founder of the field’s most-cited bodily theory of decision-making was asked, on the record, whether the body matters to the mechanism, and answered: mostly not, any more. No secondhand source in this wiki reports it, and the material exists only in the Royal Society discussion format.
Note what it does and does not undercut. A theory of unfelt bodily signals is still a bodily theory — that is the version does-somatic-feedback-guide-decisions argues with, and SCR is the right measure for it. A theory in which the periphery is bypassed in most circumstances is a theory about somatosensory cortex, in which the body is a developmental scaffold subsequently dispensed with. Those are different retreats and the wiki had them merged.
What this does to the 2005 indexing claim
Read forward from 1996, the certainty→as-if / ambiguity→body mapping below looks less like a standing view finally specified and more like a reversal: the body reacquires a job its author had written out of the adult brain. Neither paper mentions that anything changed, so whether Damasio regarded it as a revision is not knowable from these two sources.
The wiki should therefore stop describing the indexing as the constraint the framework always had and stated late. It is a change of position — and, on the evidence of both texts, a change in the direction of more body rather than less.
The developmental qualification is the useful part
Buried in a concessive clause: body states are actively engaged during development, then the periphery is bypassed.
That is the structure of primary → secondary induction as stated in 2005 (you must have felt pain to imagine it; once acquired, secondary induction becomes independent of primary) — meaning the developmental dependency was in hand nine years before the architecture built on it. It is also the structure of Farb et al.’s simulation-map: a body model built from bodily experience and then run offline. Damasio in 1996 and the simulation-map literature are describing the same developmental arc from different vocabularies, and the 2005 paper — which has the inducer architecture but not this framing — is the one place it could have been said and is not.
The 2005 indexing claim — and why it matters
The wiki’s earlier one-line treatment presented the as-if loop mainly as the device that reconciles Valins’s (1966) false-feedback effect with a physiological reading (per Friedman 2010). That is true but incomplete. In the primary source the loops are indexed to the epistemic character of the decision:
| decision type | outcome | loop engaged |
|---|---|---|
| ambiguity (Iowa gambling task — probabilities never learnable) | unknown, unpredictable, unestimable | body loop |
| risk (Rogers’ “betting” task — odds stated, 90%→60%) | intermediate | intermediate |
| certainty | explicit and predictable | as-if body loop |
The stated evidence: normal subjects generate minimal anticipatory SCRs on the betting task, especially for the most certain choices, and the overall average is lower than on the gambling task. The body loop “becomes increasingly prominent as decisions move from certainty to risk, and to ambiguity.”
This is the hypothesis’s most important constraint, and it is thin. It is one preliminary comparison across two tasks that differ in far more than epistemic character (payoff structure, instructions, provenance — Rogers’ task was designed for a different purpose entirely), reported without statistics, in a paper that flags the section as preliminary.
Why the wiki should care about that thinness specifically. Without the indexing, the as-if loop is an escape hatch: any decision showing no peripheral somatic change can be attributed to intra-cerebral simulation, and “somatic marker” becomes satisfiable without a body. With the indexing, the framework makes a falsifiable claim — you may not appeal to the as-if loop under ambiguity. Whether that constraint holds is, as far as this wiki’s sources go, unknown. It is the single most useful open empirical question the somatic marker hypothesis poses to interoception research.
And it is now sharper than “unknown.” The 1996 material above means the constraint is not merely untested — it is nine years younger than the position it constrains, and its author’s stated view before it existed was that the periphery could be manipulated with relatively little behavioural consequence. Anyone running the ambiguity/certainty test would be testing a claim Damasio arrived at late and never flagged as new. That raises the value of running it: the two texts make opposite predictions about Everitt’s experiment, and a single well-matched study would tell you which Damasio was right.
As-if is not weaker
Bechara & Damasio explicitly block the natural inference that as-if states are fainter than enacted ones. They need to: §3.2.4 has posterior VM triggering stronger somatic states than anterior VM, while §2.3 assigns posterior/certain decisions to the as-if loop. If as-if meant weak, the two gradients would contradict each other.
The resolution offered is that the loops differ in “the complexity of the mechanisms for triggering somatic states as opposed to the strength or magnitude of somatic states” — different qualities, not different amounts. Their illustration: finding a million dollars on a table inside a bank (certain, limited response options) suppresses the impulse to take it “quickly and robustly” with no body loop needed; finding a million dollars in a dark alley (ambiguous, numerous conflicting options) may well engage the body. Neither somatic state is stronger; “the quality is different.”
This is asserted rather than demonstrated, and the bank/alley pair is an intuition pump. But note the structural point stands regardless of the illustration: somatic states detected in the body are not an index of somatic state strength. Any study inferring marker magnitude from peripheral measurement — which is to say, most of them — is measuring loop mode as much as intensity.
Relation to the wiki’s predictive-coding material
The as-if loop is a body model run offline. So is Farb et al.’s simulation-map, and so, in a different vocabulary, is the descending generative model of interoceptive-inference: in all three, activity in somatosensing cortex stands in for the body rather than reporting it.
The convergence is real and none of the three sources acknowledges it. But do not collapse them — they differ on what the offline model is for and on which way the signals run:
- As-if loop: the simulated state’s function is to bias a decision. Its direction is ambiguous in the 2005 source — the pattern is activated “directly,” but whether that constitutes a prediction, a memory read-out, or an efference copy is never specified. The 1996 source resolves this, and not in prediction’s favour: the as-if loop is an attempt to reconstitute the somatic state that belonged to an earlier exteroceptive–interoceptive conjunction, via convergence zones that hold no content and only the potential to reactivate. That is retrodiction — a memory read-out with no comparator and no error signal. See convergence-zones for why this makes the predictive-coding parallel more precise and less flattering.
- interoceptive-inference: the model’s function is perception itself; predictions descend and prediction errors ascend, and there is no interoceptive percept that is not model-generated.
- simulation-map: the model’s function is regulation, and its layers are explicitly hierarchical.
Seth’s framework would say Damasio’s as-if loop was reaching for descending prediction without the apparatus to state it. That is a reading, not a claim either author makes, and the honest position is that the somatic marker hypothesis predates the machinery that would settle where it sits on feedforward-vs-predictive-interoception. Bechara & Damasio’s insula is an undifferentiated store-and-feel block with no posterior→anterior gradient and no comparator — neither Craig’s ascending hierarchy nor Seth’s generative model. See insular-cortex.
The 1996 paper adds one anatomical commitment the 2005 one drops without trace: the somatosensory cortices carrying the loop (insula, SII, SI) are specified as mattering especially in the non-dominant hemisphere, stated without a supporting citation. By 2005 the laterality claim has migrated to a different structure entirely (left VM ↔ positive, right VM ↔ negative) on unrelated lesion evidence. Two laterality claims, two structures, one programme, never reconciled — and neither engaged with Craig’s right-AIC asymmetry, which is the wiki’s third.