The mental inference fallacy
Named in Barrett (2017):
This error, assuming that an action is equivalent to an emotion, which I call the mental inference fallacy, has wreaked havoc with the scientific accumulation of knowledge about emotion.
The canonical case: a rat freezes in response to shock; the experimenter calls it fear; the circuitry mapped is then reported as the circuitry for fear. The inference fails in both directions —
- Rats don’t freeze only when threatened (one behaviour → many categories).
- Rats do many things when threatened besides freeze (many behaviours → one category).
So the behaviour-to-state mapping is many-to-many, and the emotion word was supplied by the human watching. “Motor movements do not provide a direct indication of an internal state, be it in a rodent, a monkey or a human.”
The historical version
Barrett’s two best examples are a century old and both are reinterpretations rather than dismissals — which is what makes them useful.
Woodworth & Sherrington (1904) removed the cortex, thalamus and hypothalamus from cats and observed what they themselves called pseud-affective reflexes. Barrett takes their hedge seriously: the reflexive motor actions survived, but the allostatic driver was gone, so “these animals appeared to behave emotionally, but the actions were no longer in service of survival.” Pattern generators without the conceptual system.
Cannon & Britton (1925) — the decorticate cat that woke from anaesthesia spitting, clawing and arching its back. Cannon called it “sham rage.” Barrett’s point lands: in naming it rage, Cannon inferred a mental state from a set of actions. The word “sham” is doing the work the theory needs, and Cannon put it there himself. See walter-cannon.
The awkward part: LeDoux agrees
This is one of the few places where the wiki’s two principal antagonists make the same argument, and it is worth not smoothing over.
LeDoux (2012) built his entire programme on this complaint. Renaming “fear conditioning” to Pavlovian defense conditioning is a refusal of the mental inference fallacy — the circuit is defined by evolved function, “explicitly decoupled from feelings,” precisely because inferring the feeling from the freezing was the error. Barrett cites him for it (alongside her own 2012), which the polemic between them rarely acknowledges.
They diverge immediately afterward, and on what to do about it:
| diagnosis | remedy | |
|---|---|---|
| LeDoux | the emotion words misdescribe the circuits | re-carve by evolved function (survival-circuits) — keep the circuits, drop the words |
| Barrett | the emotion words are supplied by a perceiver, and there are no emotion-specific circuits to re-carve | dissolve the categories into construction; the circuits are pattern generators, not emotions |
Barrett presses the point further than LeDoux would like, and aims it at him: when scientists “create finer-grained typologies in an attempt to bring nature under control” (she cites Gross & Canteras, 2012, on “the many paths to fear”), “this does not avoid the problem.” Sub-dividing fear into fear-types is still inferring mental states from context-dependent behaviours; it just does it at higher resolution.
Whether survival-circuits escape this is genuinely open. LeDoux’s defence is that his circuits are not named for feelings — defense, energy/nutrition, fluid balance, thermoregulation, reproduction are functions, not emotions. That looks like a real escape. Barrett’s rejoinder is not made directly in this paper and the wiki should not manufacture one. See what-should-emotion-terms-refer-to.
The methodological corollary
The fallacy is why Barrett rejects the framing that human fMRI and animal optogenetics give different answers:
In reality, all are making physical measurements and mapping emotion concepts to them, and no set of findings is described appropriately by classical emotion concepts.
The measurements are fine. The concept mapped onto them is supplied by the experimenter in every case — human or rat, scanner or optrode. This is a sharper claim than the resolution dispute on amygdala, because it says the problem is not that the instrument is too blunt to see the emotion; it is that the emotion was never in the animal to be seen. It was in the person writing the paper.