Affect programs

Ekman’s mechanism for a basic emotion: a causally powerful latent variable that mediates between stimuli and responses, selected to solve a recurrent evolutionary problem. The challenges Tooby & Cosmides list are the paradigm cases — “[f]ighting, falling in love, escaping predators, confronting sexual infidelity, experiencing a failure-driven loss in status, responding to the death of a family member.”

Latent variable is the operative phrase. An affect program causes facial expressions, autonomic changes, and actions; it is not constituted by them. This is precisely what constructionists deny — see core-affect and the emergent-variable model below.

Input-open, output-closed — and why that mattered

The traditional (Ekman) specification, drawing on Mayr’s (1974) notion of closed behavior programs:

• Input-open: learning can shape which stimuli activate the program, via appraisal. Elicitors differ from person to person — which is why Ekman insists basic emotions are not reflexes ([18]: reflexes are easy to elicit and reliably shown).
• Output-closed: once running, the program produces a mandatory cascade. Ekman & Cordaro: a “cascade of changes (without our choice or immediate awareness) occurs in split seconds in: the emotional signals in the face and voice; preset actions; learned actions; the autonomic nervous system activity that regulates our body.”

The mode of operation is therefore reflex-like even though the program is not a reflex — and this is what generated basic emotion theory’s central empirical hypothesis: there should be bodily signatures for each basic emotion, consisting of highly correlated, emotion-specific changes across facial expression, autonomic activity, and preset/learned actions.

That hypothesis is what the wiki’s autonomic-specificity-of-emotion debate has been about for decades. It descends from output-closure.

The three fates of the construct

The wiki now holds three distinct responses to the failure of the signature evidence, and they are worth keeping separate because they are often conflated:

1. Defend the signatures — Friedman (2010), Levenson, Ekman. The multivariate evidence for autonomic-specificity is better than the constructionist reading allows; James himself held that within-emotion similarities trump individual variation.

2. Abandon the latent variable — Barrett, Russell, Lindquist et al. (2012). Replace it with an emergent variable model: emotions do not cause facial expressions, autonomic changes, and actions — they emerge from them, via categorization of an underlying state of core affect.

3. Keep the latent variable, drop output-closure — Scarantino (2018). A probabilistic latent variable model: emotions still cause responses, but not the same responses every time.

Position 3 is the one this page exists to record, because it is new to the wiki and it reprices the other two.

Input-output open: the New BET revision

Scarantino’s affect programs are input-output open: learning shapes both what activates the program and what responses it brings about. Their output is not a response cascade but prioritized action tendencies — Ekman’s evolutionary adaptations merged with Frijda’s (1986) action tendencies with control precedence.

The argument for this is evolutionary, not evidential, and that is what makes it interesting. Output-closure “never made much evolutionary sense to begin with”:

To solve abstractly defined evolutionary problems like dealing with dangers or slights, basic emotions cannot afford to produce anything more than impulses to behavior that are flexibly adaptable to the specific circumstances of elicitation.

The exception proves the rule: closure is expected only where speed is everything and there is no time to adapt — suddenly looming objects for basic fear.

So the decades of searching for bodily signatures were chasing a prediction that a better-specified BET would never have made. Scarantino accepts the negative empirical record (“high correlations and specificities have not been demonstrated empirically with respect to anger, fear, happiness, sadness, surprise, and disgust”) and denies it refutes BET. Combined with HPC kinds, variability stops being an embarrassment and becomes a prediction.

This is a third position on autonomic-specificity-of-emotion that the wiki did not have: not “the signatures are there” and not “their absence refutes discrete emotions,” but their absence was predictable, and the theory should never have promised them.

Relation to survival circuits

LeDoux’s survival circuits are the same kind of thing structurally — special-purpose mechanisms, evolved for specific challenges, generating a global motive state, reliant on a general-purpose cognitive system for adaptation to circumstances. Scarantino says so explicitly, and reads both as instances of Gallistel’s two-level control structure.

The differences are two, and neither is about closure:

• Carving. Affect programs are carved by explicated folk category (basic fear, basic anger); survival circuits by evolved function (defense, energy/nutrition, fluid balance, thermoregulation, reproduction). LeDoux’s carving deliberately fails to recover the Ekman six — no aggression circuit, no joy circuit.
• Naming. See what-should-emotion-terms-refer-to.

Both authors have converged on flexible output. LeDoux gets there via instrumental behaviour and cortical regulation; Scarantino via action tendencies with control precedence. The wiki should note that output-closure has no defender among its current sources — Friedman defends the signatures, which is a weaker claim, and Ekman’s own late position (via Ekman & Cordaro 2011) is the only statement of closure on record here, quoted by its critic.