Basic emotions and natural kinds
James (1884) restricted his theory to the “standard”/“coarser” emotions (fear, anger, grief, surprise) — those with “distinct bodily expression” forming a “natural language.” This seeds the modern basic-emotions tradition (Ekman, Izard, Plutchik, Tomkins, Levenson): a limited set of discrete, universal, adaptive emotions each matching behavior/expression/ANS support to environmental demand. See william-james, friedman-2010-jamesian-perspective.
The “natural kinds” dispute
- Barrett (2005, 2006) argues against the natural-kinds assumption: emotions may not exist as entities independent of human perception/conceptualization; the assumption smuggles in discrete somatic patterns that may not exist. See lisa-feldman-barrett.
- The argument first-hand (Barrett 2017): the wiki carried this position through others’ papers for eight ingests. Barrett’s own statement grounds it in degeneracy — dissimilar neural populations produce instances of the same category, so “it is unlikely that all instances of an emotion category share a set of core features,” and the category’s summary representation is “an abstraction that need not exist in nature.” Her philosophical warrant is Darwin’s, via Mayr: a species is a population of variable individuals grouped by function, not a type with shared features; emotion categories are conceptual categories in exactly that sense. The irony she presses: this makes constructed emotion consistent with On the Origin of Species and inconsistent with The Expression of the Emotions in Man and Animals — the basic-emotions tradition claims Darwin, but from the book in which he abandoned his own innovation.
- Friedman/Levenson/Ekman counter that the weight of (multivariate) evidence for autonomic-specificity supports discriminable basic-emotion signatures, and that James himself — often cited by constructionists for acknowledging individual variation — actually held that within-emotion response similarities “trump” individual variability (“the symptoms of the angers… of different men still preserve enough functional resemblance… to lead us to call them by identical names”).
Discrete vs dimensional affective space
Orthogonal to natural-kinds: whether affect is best described by discrete categories or by dimensions (circumplex: valence × arousal; or orthogonal positive/negative axes, Watson & Tellegen). Friedman notes autonomic vs self-report measures may differentially support these structures, and endorses a hybrid where discrete emotions are unique points in dimensional space (Levenson). Relevant to autonomic-specificity-of-emotion.
The neuroimaging extension of the natural-kinds critique
Lindquist, Wager, Kober, Bliss-Moreau & Barrett (2012) generalize Barrett’s natural-kinds argument from psychophysiology/self-report into neuroimaging: a large meta-analysis finds no brain region consistently and specifically activated for any single discrete emotion category, proposing instead that categories emerge from domain-general core affect, conceptualization, language, and executive-attention networks. See locationist-vs-constructionist-brain-emotion for the resulting debate, including Scherer’s methodological objection that the locationist/constructionist dichotomy mischaracterizes non-constructionist alternatives like componential appraisal theory.
The eliminativist alternative: abandon the categories, keep the circuits
LeDoux (2012) adds a position that fits neither camp cleanly, and is easy to misfile given that its author is the amygdala-and-fear scientist. He concurs with Barrett that “the foundation of support for the idea that basic emotions, as conventionally conceived, have dedicated neural circuits is weak” — but denies this shows the mammalian brain lacks innate emotion-relevant circuits. It shows only that basic-emotion categories, derived from introspective English feeling words, are the wrong description of the circuits that exist. His replacement: survival circuits for defense, energy/nutrition, fluid balance, thermoregulation, and reproduction.
Three internal problems for the basic-emotions program that LeDoux presses:
- The lists don’t agree. Different theories posit different numbers of basic emotions and different names for similar ones; Panksepp’s animal-derived list “does not match up well” with Ekman’s human list — awkward for a program claiming evolutionary conservation.
- The methods are suspect. Identification often relies on forced choice rather than free labeling of facial expressions.
- The categories lack coherence. A single emotion label refers to a diverse range of states.
His constructive point is sharper than the critique: survival circuits deliberately fail to recover the Ekman six. There is no anger/aggression circuit (aggression segregates by context — defense, reproductive competition, predatory feeding) and no joy/pleasure circuit (such behaviours are outputs of the energy, fluid, procreation, and thermoregulation circuits). Converging on one feeling word “glosses over the underlying details of emotional processing.” Where Barrett dissolves the categories into construction, LeDoux dissolves them into evolved function — and unlike Barrett, keeps hard-wired circuitry.
Note the further consequence for autonomic-specificity-of-emotion: LeDoux is compatible with dimensional accounts (his circuits link to feelings only indirectly and nonobligatorily) but holds basic-emotions theory is incompatible with dimensionalism, since it derives feelings from hard-wired circuits. He also breaks with Barrett on method — see amygdala for the resolution dispute.
The reformist alternative: keep the circuits, change what a kind is
Scarantino (2018) responds to LeDoux directly, and occupies the position that the natural-kinds dispute above has been implicitly treating as unavailable: a basic emotion theorist who concedes the constructionist evidence. See andrea-scarantino.
What his New BET grants — more than Friedman grants, and essentially all of Barrett’s empirical case:
- Folk emotion categories are not natural kinds: “their extensions are too heterogenous for any scientifically interesting generalizations to be true of all of their members.”
- The bodily signatures are not there: “high correlations and specificities have not been demonstrated empirically.”
- Emotions have no essences and no fixed boundaries.
His diagnosis of why folk categories fail is the sharpest statement of the natural-kinds problem on this page: those categories “were not introduced into the language to carve nature at its joints, but rather to negotiate social transactions among language users. If folk emotion categories happened to carve nature at its joints without having been introduced for that purpose, this would be an unlikely and happy coincidence.” The coincidence did not occur.
What he keeps regardless: evolved, pan-cultural, causally potent affect programs with homologs in other species — a probabilistic latent variable model, against the constructionists’ emergent variable model. Emotions still cause expressions, autonomic changes, and actions; they do not emerge from them. In Lakatos’ terms, that is BET’s hard core; the essentialism and the output-closure were only the protective belt.
Two modifications make this consistent:
- HPC kinds (Boyd 1991). Kinds with no necessary properties, defined by clusters of relational and historical properties co-occurring because of imperfect mechanisms — as biological species are. Massive internal variability becomes a prediction (responses must adapt to circumstance), not a refutation.
- Explication (Carnap 1950). Replace the folk category with a narrower one — basic fear ⊂ fear — as memory science replaced folk “memory” with short-term/long-term memory. This changes what counts as counterevidence: “A critique of a theory of basic fear based on data about fear writ large is like a critique of a theory of short-term memory based on empirical data about memory writ large.”
Three ways to survive the natural-kinds critique
The wiki now holds three distinct escapes from the same pressure, and they should not be blurred together:
| keeps the categories? | keeps hard-wired circuits? | how | |
|---|---|---|---|
| Barrett / Lindquist | no | no | dissolve emotions into construction from core affect + conceptualization |
| LeDoux | no | yes | rename them — survival circuits carved by evolved function |
| Scarantino | yes (explicated) | yes | reconceive what a kind is — HPC, not essence |
LeDoux and Scarantino agree on nearly all the mechanism and disagree about vocabulary; that dispute has its own page, what-should-emotion-terms-refer-to. The useful contrast for this page is with Barrett: Scarantino concedes her data and denies her conclusion, which makes the natural-kinds argument’s decisiveness turn on an unstated premise — that natural kinds must be essentialist. Boyd’s account denies exactly that premise, and no source in the wiki defends it.
The cost is recorded on homeostatic-property-cluster-kinds: a kind with no necessary properties and expected variability is hard to falsify, and the only guard is the fruitfulness requirement (the category must support some scientifically interesting generalization), which Scarantino asserts rather than demonstrates.
The fourth response: change the measurement
Nummenmaa et al. (2014) and its scale-up Volynets et al. (2020) are not an escape from the natural-kinds critique — they are a refusal of it, and the only strategy here that defends the tradition by finding the discreteness somewhere new rather than by reinterpreting its absence. See lauri-nummenmaa.
The founding paper states the wager plainly and is worth quoting against the versions of it that came later: the distinct maps “are in line with the evidence from brain imaging and behavioral studies, highlighting categorical structure of emotion systems,” and suggest emotion information “is also represented in embodied somatotopic format.” Two hedges are in the same discussion and rarely travel with the citation — the authors do “not argue that the BSMs… would be the only components underlying emotional experience,” and concede the maps “may not relate directly to specific physiological changes.”
The programme’s wager is that the field has been looking in the wrong place. Signatures were sought in autonomic recording (weakly emotion-specific) and in brain regions (no consistency-plus-specificity, per Lindquist). Bodily sensation maps look instead at the felt body at high spatial resolution, and there the discreteness is unmistakable: 13 emotions plus neutral form separable clusters that hold across 101 countries (rs > 0.82) and both sexes (r > 0.80), with emotion category dominating every cultural and demographic variable in a ridge regression. The authors read this as “lending support to the basic emotion view (Ekman, 1992),” and add that the universalism extends past the canonical six to pride, shame, jealousy, contempt, love, anxiety and depression.
Two things keep this from settling the page’s dispute, and both are visible in the source.
It is a different claim than the one under attack. Barrett’s target is emotion-specific physiology; Volynets et al. measure emotion-specific phenomenology, and disclaim the physiological reading explicitly (“the BMs are thus subjective feeling maps”). Discrete felt maps sitting atop non-discrete bodily states is a coherent — indeed constructionist-friendly — possibility that the data cannot exclude. See autonomic-specificity.
The method’s cue is a word. Subjects were shown emotion terms, in English, and asked where they would feel them. That concordant answers might reflect shared concepts rather than shared bodies is exactly Barrett’s account of what emotion categories are, and the paper does not engage it (she is not cited). See cultural-universality-of-emotion and embody.
So the tradition gains its best contemporary evidence for discreteness and universality of felt emotion, and gains nothing on the natural-kinds question directly — while handing the constructionists a finding that their own theory predicts about as comfortably.
Where this leaves the page: the wiki now holds four responses to the same pressure.
| keeps the categories? | strategy | |
|---|---|---|
| Barrett / Lindquist | no | dissolve them into construction from core affect |
| LeDoux | no | re-carve by evolved function (survival-circuits) |
| Scarantino | yes (explicated) | reconceive kindhood (HPC, not essence) |
| Nummenmaa / Volynets | yes (unrevised) | change the measurement — find the fingerprints in felt topography |
Nummenmaa is the only one of the four who still expects Ekman’s categories to show up in data as they stand. He is also the only one who reports that they do.
Postscript: the two sophisticated positions made the same move
Visible only now that the wiki holds Barrett (2017) first-hand, and invisible from inside either source, since neither cites the other doing it.
Barrett and Scarantino — row 1 and row 3, the two positions on this page with a worked-out philosophy of categories — both defeat essentialism, both use biological species as the paradigm case, and both make within-category variability a prediction rather than a refutation. They just reach for different philosophers:
| device | via | conclusion | |
|---|---|---|---|
| Barrett | population thinking | Mayr, reading Darwin | categories are conceptual → emotions are constructed |
| Scarantino | HPC kinds | Boyd | kinds need no essences → basic emotions survive |
Same premise, opposite conclusions. Which tells you the real disagreement is not about essentialism — that fight is over, and essentialism lost twice. It is about whether a variable, functionally-grouped category can still be causally potent. Scarantino keeps the probabilistic latent variable: emotions cause expressions and autonomic changes. Barrett does not: categorizations are the events, and there is nothing underneath them doing the causing.
The cost is symmetrical and worth stating plainly, because the wiki has now recorded it twice from opposite directions. homeostatic-property-cluster-kinds flags that nothing in the imaging record can disconfirm New BET. degeneracy flags that nothing in the imaging record can disconfirm constructed emotion. The evidence this page was built to weigh can no longer decide between its two best theories, and it is degeneracy on one side and HPC kinds on the other that made it so.