Psychopathy and interoceptive awareness
Nentjes, Meijer, Bernstein, Arntz & Medendorp (2013), Journal of Personality Disorders 27(5):617-624 (Guilford Press). A four-page brief communication, and the wiki’s first source on psychopathy and its first forensic/offender sample. No DOI is printed on the paper; Guilford’s DOI scheme (10.1521/pedi_YYYY_VV_NNN) is not mechanically recoverable from anything on the page (no PII, and the article number is not derivable), so the doi field is left blank rather than guessed — the same convention used for Damasio (1996).
The design is simple: give 75 male Cluster-B-disordered offenders a heartbeat discrimination task and the PCL-R, and correlate. The interest is entirely in which part of psychopathy the interoceptive deficit attaches to — and the answer is the opposite of the one the paper set out to find.
The result, and the twist inside it
The hypothesis was Jamesian by way of Damasio: psychopathy is a disorder of emotional deficiency, interoception underlies emotional experience, so psychopaths should perceive their bodies poorly. The data half-cooperated. Interoceptive accuracy was inversely related to psychopathy — but not to the emotional part.
| PCL-R component | what it captures | r with d’ |
|---|---|---|
| Factor 1 | interpersonal / affective | −.01 (ns) |
| Facet 1 | affective (shallow affect, no remorse) | −.01 (ns) |
| Facet 2 | interpersonal (lying, manipulation) | −.04 (ns) |
| Factor 2 | antisocial / deviant lifestyle | −.29 (p<.05) |
| Facet 3 | lifestyle (impulsivity, irresponsibility) | −.22 (p<.10) |
| Facet 4 | antisocial (poor controls, criminal versatility) | −.24 (p<.05) |
| PCL-R total | — | −.20 (p<.10) |
The affective core — the “marked emotional deficiencies” the introduction leads with — is at r = −.01. The whole relationship lives in the antisocial-behaviour dimension. Regression with covariates confirms it: Factor 2 (β=−.32, p=.006) and, at facet level, Facet 4 (β=−.30, p=.01) survive; heart rate, impulsivity and trait anxiety drop out.
So the paper’s own framing — reduced interoception as the bodily face of psychopathic coldness — is not what it demonstrated. It demonstrated reduced interoception as a correlate of psychopathic antisociality. The authors pivot cleanly to a somatic-marker reading (attenuated bodily signals → weaker behavioural guidance → more norm-violation), which fits Factor 2 better than the affective story ever fit Factor 1. But the wiki should record that the confirmed hypothesis is a reconstruction: the affective-deficit prediction failed, and the antisocial correlation is what remained.
The chance-level problem the paper walks past
The single most important number for reading this study is not in the abstract. Mean d’ was 0.00 (SD 1.18). On the Brener–Kluvitse simultaneity method, this forensic sample did not discriminate its own heartbeats above chance, on average — exactly what the wiki’s method page records this task family “typically” does (chance-level performance, via Van der Does et al.).
That reframes “reduced interoceptive awareness.” The Factor-2 correlation is not “good detectors are prosocial, poor detectors are antisocial”; it is variation around a mean of zero, where higher psychopathy pushes d’ from ~chance toward the negative (systematic misjudgement) tail. A signal-detection d’ below zero is not a smaller amount of a real percept — it is confident wrongness about beat timing, which on a strict reading is closer to the schema-driven percepts Van der Does describes than to blunted perception. The paper does not address what a chance-anchored, sometimes-negative accuracy score means; it treats d’ as a graded quantity of “interoceptive awareness” throughout. Held here as the study’s central interpretive soft spot.
Where it sits in the wiki’s arguments
On somatic markers. This is the framework’s extension to a genuinely new population, and it comes with the framework’s own supporting citations: Werner et al. (2009) — poor heartbeat detection tracks disadvantageous Iowa Gambling Task play (now read first-hand: a real effect, but extreme-groups, marker-free, and significant on deck choice not net gain — so the anchor Nentjes leans on is softer than the citation implies); Gao, Raine & Schug (2012) — “somatic aphasia,” a mismatch between reported and actual autonomic change in psychopathy; Katkin, Wiens & Öhman (2001) — good detectors show superior fear conditioning. Psychopathy is independently known to impair IGT performance (Mitchell et al. 2002) and fear conditioning (Birbaumer et al. 2005), so the interoceptive deficit slots into an existing behavioural-regulation gap. This is suggestive convergence, not a test: Nentjes et al. measured no decisions and no markers — only heartbeat detection and a trait score — so the somatic-marker chain is inferred, not observed. See does-somatic-feedback-guide-decisions.
On is more awareness better? This is a rare row where less interoceptive accuracy tracks a worse outcome (antisocial behaviour), i.e. a “more is better” datum — and it points opposite to the panic literature on the same page, where better cardiac perception predicts worse outcome (Ehlers 1995). Interoception’s clinical valence flips by disorder: too little may disinhibit the antisocial offender, too much may sensitise the panic patient. But the row is weak — cross-sectional, mean-at-chance, effect confined to the behavioural factor, R²≈.11 — and it should sit in the collision table as another reason the naive dose-response reading fails, not as a clean vote.
On interoceptive-sensitivity and the Dunn gain-term view. Nentjes et al. report a direct zero-order correlation between accuracy and an outcome (the psychopathy factor), which is the kind of main effect Dunn et al. argue interoceptive accuracy does not have. The tension is soft — the outcome here is a personality trait, not felt arousal or decision quality, and the effect is modest — but it belongs on the record alongside Pollatos’s arousal main effect as another accuracy↔outcome correlation that the pure gain-term reading does not predict.
Honest weaknesses, recorded
- Hypothesis/result mismatch (above): the affective-deficit rationale failed; the antisocial correlation is the survivor.
- Chance-level mean d’ (above): the interpretive anchor is zero, and the score can be negative.
- Forensic confounds: institutionalisation (~7 years), unreported medication, a task the offenders found tedious — all plausibly depress d’ and covary with Factor 2. Stroke volume unmeasured, so the cardiodynamic confound is untouched.
- The unexplained IQ effect: higher IQ → lower detection (β=−.23). Carried as a covariate, never interpreted; worth flagging because it runs against the grain of the somatic-marker story (one would not expect the better decision-makers to be the worse perceivers).
- Small, atypical sample: N=75, single session, no replication, and the Factor 1/Factor 2 independence (r=.08) that makes the dissociation clean is itself unusual.
Bottom line
A modest but genuinely novel data point: in a forensic sample, reduced heartbeat-discrimination accuracy tracks the antisocial, not the affective, dimension of psychopathy, best read through the somatic marker hypothesis as attenuated bodily guidance of behaviour. Its value to the wiki is as the first psychopathy and first forensic source, and as a “less is worse” counterweight in is-more-interoceptive-awareness-better — held loosely, because the confirmed hypothesis is a post-hoc reconstruction, the sample discriminated at chance, and nothing but a correlation was measured.