Barnaby D. Dunn

The wiki’s principal critic of the somatic-marker-hypothesis, and — unusually for a critic — the person who then went and collected the evidence he said was missing.

Created with the dunn-2006-somatic-marker-evaluation ingest. does-somatic-feedback-guide-decisions had carried him as a field position marked ‘Position not yet sourced first-hand’ since the Bechara & Damasio ingest, on the strength of Friedman’s (2010) passing mention and his co-authorship of Farb et al. (2015). That placeholder is now filled.

The critique

Dunn, Dalgleish & Lawrence (2006) is not a refutation and does not claim to be: ‘none of these reservations falsify the SMH; they just suggest that other sources of evidence need to be gathered.’ What it does is disaggregate. The hypothesis had been argued about as one thing; Dunn et al. show it is at least three, with three different verdicts.

• The neural substrate — VMPFC, amygdala, insula, somatosensory cortex — is ‘reasonably well supported,’ with revisions (right-lateralized; DLPFC involved; the region badly named).
• The psychological mechanism requires revision: the reward schedule is cognitively penetrable, the anticipatory SCR admits several readings, and ‘myopia for the future’ is one of at least five candidate explanations of the deficit.
• The evidential base — the iowa-gambling-task — ‘is no longer sufficient to be a major source of evidence for the SMH.’

The move that gives the paper its staying power is the last one. Attacking a theory is cheap; showing that the single paradigm carrying it cannot carry it, by cataloguing every published use of that paradigm (Table 1, ~35 clinical studies to July 2005), is not.

The two things he saw that this wiki independently re-derived

Worth recording, because it calibrates how much credit the wiki’s own readings deserve.

1. The unfalsifiability structure. does-somatic-feedback-guide-decisions built a table of ‘the framework’s own escape hatches’ from reading Bechara & Damasio and Damasio (1996) directly. Dunn et al. state the same conclusion in 2006, from the empirical side: nearly every negative causal finding is absorbed by appealing to residual peripheral routes or to the as-if loop, and ‘it seems virtually impossible to test the theory in a scenario where all peripheral feedback routes are disturbed and the as-if loop cannot be utilised.’

2. Interoception is the missing variable. Four wiki pages independently observe that the Iowa programme measures somatic signalling and never asks whether anyone perceives the signal. Dunn et al.’s future-directions section names it: ‘an individual differences approach could be adopted, for example looking at how variance in interoceptive ability relates to IGT performance’ (citing Katkin et al. 2001).

Neither observation is original to this wiki. Both were in print in 2006.

Then he ran the study

The reason Dunn matters here rather than only in the decision-making literature. Having identified interoceptive ability as the untested moderator, he went and measured it with the heartbeat-detection-task — and built a task to measure it against.

Now ingested: Dunn et al. (2010). This page previously carried the finding at second hand, on Friedman’s account, as “the relationship is not monotonic,” with an instruction not to state his 2010 position until the ingest happened. The instruction is discharged, and the second-hand summary was directionally right but structurally wrong. The finding is not that the relationship is non-monotonic. It is that there is no relationship at all.

Interoceptive accuracy predicted nothing in either study on its own: not felt arousal, not felt valence, not cardiac response, not decision quality (r = .08, p = .46), not even how informative the participant’s own body was (r = .07, p = .56). Every effect the paper reports lives in a product term. Better perceivers’ bodily responses coupled more tightly to their felt arousal (ΔR² = .12) — and not at all to their felt valence (ΔR² = .00, tested against each other, Z = 2.45). Better perceivers decided better when their anticipatory bodily signals favoured the profitable decks and worse when those signals favoured the unprofitable ones (ΔR² = .08).

So Dunn’s position on is-more-interoceptive-awareness-better is sharper than “not monotonic”: interoception is not on the axis at all. It is a gain term on a signal whose sign it does not determine. Ask whether more is better and the answer is a question — better at what, and is your body currently right?

The other half is the task. Study 2 does not use the iowa-gambling-task; it uses the intuitive-reasoning-task, which Dunn built to repair the IGT flaws his 2006 review catalogued — reversal confound removed, deck position counterbalanced, magnitude crossed with profitability, pre-decision physiology separated from pre-feedback. This is the most direct case in the wiki of a critique converted into an instrument. It is also the reason to read him sceptically here rather than only admiringly: the person who designed the test had a published position on what the test should find, and the paper’s central claims rest on two unreplicated interactions at n = 58 and n = 92.

Clinical trajectory

Dunn’s later work moves to depression and anhedonia at Exeter, and his presence on Farb et al. (2015) is the visible hinge: the person who established that the Iowa programme never measured interoception joins a working group whose entire subject is interoceptive training. See norman-farb, mindfulness-interoceptive-training.