Do basic emotions have distinct autonomic signatures?
The oldest live debate in this cluster, running from 1884 to present. It is the empirically tractable core of the James–Lange controversy: not the (hard-to-test) temporal claim that bodily change precedes feeling, but the specificity corollary — that emotions like fear and anger have discriminable ANS patterns.
The two poles
- Specificity (James → Ax → Ekman/Levenson → Friedman): emotions have distinct, evolutionarily adaptive visceral signatures; apparent non-specificity is an artifact of univariate measures and weak inductions. Multivariate PCA studies are the strongest support.
- Non-specificity (Cannon → activation theory → Schachter–Singer): arousal is a single diffuse sympathetic state; emotional differentiation comes from central processing or cognitive labeling.
Where the papers land
- Friedman (2010) argues the evidence tilts toward specificity and endorses Cacioppo’s continuum reconciliation (patterned ↔ undifferentiated, somatovisceral afferents causal throughout).
- Dror (2017) complicates the non-specificity pole historically: the “undifferentiated arousal” premise entered via Schachter’s social-influence program, not from a physiological demonstration — so the debate may partly rest on a historically contingent framing (see origins-of-two-factor-theory).
- Seth (2013) arguably dissolves rather than settles it: under interoceptive-inference, “specific pattern” vs “labeled arousal” are recast as differing precision on interoceptive predictions vs prediction errors.
Scarantino’s third position: the prediction was badly motivated
Scarantino (2018) adds a stance that neither pole provides for, and it is the first source here to attack the hypothesis rather than the evidence.
The specificity claim is not a free-standing prediction. It descends from a specific auxiliary assumption — that affect programs are output-closed, producing a mandatory cascade across face, voice, ANS, and action once triggered (Ekman & Cordaro 2011). That is what made “there should be bodily signatures for each basic emotion” the central empirical hypothesis of traditional BET, and what has funded six decades of searching.
Scarantino grants the search has failed, and argues it should never have been launched. Output-closure is evolutionarily implausible: an emotion solving an abstractly-defined problem must issue flexible impulses, not fixed cascades, because the right action depends on circumstances the program cannot anticipate. Closure makes sense only where there is no time to adapt — looming objects for basic fear. So New BET’s programs are input-output open, and their outputs are Frijda’s action tendencies with control precedence.
The result is a genuine third cell:
| are the signatures there? | does their absence refute BET? | |
|---|---|---|
| Friedman / Levenson / Ekman | yes (multivariate) | — |
| Barrett / Cannon / Schachter | no | yes |
| Scarantino | no | no — BET never should have promised them |
What makes this more than a dodge is the pairing with HPC kinds: variability is not merely tolerated, it is predicted, because responses must adapt to circumstance. What makes it less than a victory is that the debt is only transferred — the theory now owes a demonstration that basic fear supports some scientifically interesting generalization, which Scarantino asserts and defers to prior work.
Note the awkward consequence for the specificity pole: Scarantino and Barrett agree on the data here. Friedman (2010) is left as the only source in the wiki still defending the signatures empirically, and his opponents now include a basic emotion theorist.
Status: open
Meta-analyses still conflict (Cacioppo et al. 2000: valence but not discrete-emotion differentiation; Stemmler 2004: robust fear-vs-anger specificity). The basic-emotions / natural-kinds dispute (Barrett) keeps the question alive at the conceptual level.
The Scarantino ingest shifts what the debate is about rather than settling it. If output-closure is dropped, “do basic emotions have distinct autonomic signatures?” stops being the test of basic emotion theory and becomes a narrower psychophysiological question — one whose answer (probably “not reliably”) no longer decides anything theoretical. The live question moves to affect-programs: what should an evolved emotion program be expected to produce? Recorded as widening the debate; no source here adjudicates it.
A near miss: the felt body is not the autonomic body
Volynets et al. (2020) arrives looking like the specificity pole’s long-awaited reinforcement — Jamesian framing, emotion-specific “fingerprints,” N = 3954, effect sizes that dwarf anything in the psychophysiological record — and it is not. Recording why, because the misreading is easy and would corrupt this page.
What they measure is where people report feeling activation and deactivation (bodily-sensation-maps, via embody), not what the viscera do. The paper’s own limitation section is unambiguous: interoceptive accuracy is “far from perfect” (Critchley et al. 2004), bodily changes “are not strongly coupled with subjective feelings” (Mauss et al. 2005), and therefore “the present data do not reflect a detailed description of emotion-specific physiological changes but the net consciously felt changes occurring in the body during emotions, and the BMs are thus subjective feeling maps.” They cite Siegel et al. (2018) — the meta-analysis whose title asks “Emotion fingerprints or emotion populations?” — in the introduction as leaving this debate open, and never claim to close it.
So the finding is orthogonal to the poles above rather than between them. Distinct felt maps are compatible with Friedman’s specificity and with Barrett’s non-specificity, and would be predicted by Scarantino’s input-output open affect programs too (flexible physiology, stable felt structure). It adds evidence to a layer this debate had not been arguing about.
What it does change is the burden on the non-specificity pole. If arousal were genuinely undifferentiated and identity came entirely from labeling, one might expect felt topography to be equally undifferentiated — yet it is sharply structured and stable across cultures. The constructionist has a ready answer (the structure comes from the concepts, and the cue was an emotion word in English), and that answer is now doing real work rather than sitting idle: it must explain a large, robust, replicated pattern rather than merely deny a weak one. Whether word-cued self-report can distinguish a felt body from a conceived one is the crux, and no source here settles it — see cultural-universality-of-emotion.
Nummenmaa’s actual position: the compound, not the channel
Filing the programme as a “near miss” was right about Volynets et al. and wrong about the programme. The founding paper (nummenmaa-2014-bodily-maps, ingested after Volynets) states a position on this debate directly, and it deserves recording as one:
physiological data have not revealed consistent emotion-specific patterns of bodily activation, with some recent reviews pointing to high unspecificity (Barrett 2006) and others to high specificity (Kreibig 2010). Our data reconcile these opposing views by revealing that even though changes in specific physiological systems would be difficult to access consciously, net sensations arising from multiple physiological systems during different emotions are topographically distinct.
The proposal: both poles are looking at single channels, where the answer really is “not reliably specific” and “not consciously accessible.” Specificity lives in the compound — the summed skeletomuscular, visceral and autonomic sensation that subjects “cannot separate” — and the compound has a spatial structure that no single-channel recording would ever show. That is a genuine fourth option on the table above, and it explains why the psychophysiological record could be as poor as Barrett says while felt emotion remains as differentiated as everyone’s experience suggests.
It is also unsupported by the data offered for it. Nothing in the paper measures the compound; it measures reports of the compound, and the paper says so (“cannot establish a direct link between the BSMs and an underlying physiological activation pattern”). Strip the physiological noun and the claim’s structure is identical to the constructionist’s — something is emotion-specific upstream of a specific report — with “net bodily sensation” where Barrett puts “conceptualization.” The proposed adjudicator is in the paper too, and remains unrun by anyone in this wiki: whole-body 15O-H2O PET perfusion during induced emotion, correlated against the BSMs. Until someone runs it, the reconciliation is a hypothesis with a diagram.
The brain-level parallel
Lindquist et al. (2012) runs the same specificity-vs-non-specificity logic on neuroimaging data rather than peripheral physiology, with Barrett again on the non-specificity side — and reaches the same conclusion (no consistent-and-specific mapping for any discrete category, this time onto brain regions rather than ANS patterns). See the dedicated locationist-vs-constructionist-brain-emotion debate, including Scherer’s objection that the locationist/non-locationist dichotomy itself may not exhaust the theoretical space.