Can we know whether other animals have conscious feelings?
Raised by LeDoux (2012), where the epistemic argument is not a side issue — it is the motivation for the entire survival circuit framework.
Why this drives the whole framework
LeDoux’s reasoning runs: emotion research uses introspective human feeling words as guideposts; those words are already questionable for studying humans; applying them to other animals compounds the problem “many fold.” If we cannot verify animal feelings, then a comparative science built on asking whether animals feel fear is built on an unanswerable question.
His fix is to change the direction of inquiry. Rather than asking whether human feelings are present in animals (backward-looking, anthropomorphic), ask to what extent circuits and functions present in animals are present in humans — a question animal methods can actually answer. Survival circuits are what you study once you accept the epistemic limit. See basic-emotions.
The asymmetry LeDoux concedes
He notices, and flags, that he argues in both directions depending on the target:
- For circuits: ask whether what is in animals is in humans.
- For feelings: ask whether what is in humans is in animals.
His defense is that the feelings question “seems only addressable in the other direction” — you must first know what underlies conscious feeling in humans (the only case with verbal report) before you can look for it elsewhere. Whether this asymmetry is principled or convenient is not settled in the paper.
The argument from other minds
The interesting move is why LeDoux permits attributing feelings to other humans but not to animals. His answer is architectural rather than behavioural: absent genetic mutation or brain damage, every human has “the same basic kind of brain, a brain with the same basic neural systems,” so we can assume comparable basic brain functions and mental capacities, and infer from behaviour with some confidence. Other species’ brains differ — and the key question is whether they differ in ways that matter, i.e. in the areas responsible for states of consciousness.
He answers yes, on two grounds:
- Cortical hardware. Conscious states depend on representation in a cognitive workspace involving prefrontal and parietal cortex (Crick & Koch; Dehaene & Changeux; Baars; Shallice). Dorsolateral prefrontal cortex is lacking in most mammals and less developed in nonhuman primates than humans; human granular prefrontal cortex has unique cellular features.
- Language. Humans alone have natural language, which permits semantic categorization of experience. There are >30 English words for gradations of fear. The brain “may be able to categorize emotional states in broad strokes without language, but it is unlikely that specific emotions (fear, anger, sadness, joy) could come about without words.” Syntax further allows simulating who will do what to whom without trial-and-error learning.
Both grounds are inherited commitments — a workspace theory of consciousness, and a strong linguistic-relativity thesis (Whorf, Lakoff, Lucy) — rather than results. That is the vulnerability.
Dependency on the anatomy
Ground (1) makes this debate hostage to where-are-feelings-constituted. If Craig is right that feelings are constituted in the anterior insula rather than a prefrontal/parietal workspace, LeDoux’s premise fails: other mammals have insular cortex. The skepticism is only as strong as the workspace anatomy it rests on, and that anatomy is contested by this wiki’s central sources.
The Scarantino (2018) ingest tightens this considerably, and from an unexpected direction — LeDoux’s own later work. Scarantino endorses LeDoux & Brown’s (2017) higher-order theory, in which the relevant “general networks of cognition” span prefrontal cortex, posterior parietal cortex, and insular cortex. If that list is LeDoux’s considered position, the DLPFC premise is no longer doing the work alone, and one of the three named substrates is one other mammals have. Whether the insula there is a workspace component (harmless to the argument) or a constituent of feeling (fatal to it) cannot be settled from a commentary. See higher-order-theory-of-consciousness; LeDoux & Brown (2017) is not in raw/ and is flagged as the most useful missing source for this debate.
Scarantino’s dissolution: attribute the motive state, not the feeling
Scarantino agrees with LeDoux where it counts against Panksepp — adaptive behaviour and feelings are implemented by different circuits — and then declines the epistemic problem rather than solving it.
The move follows from his terminological position. If “emotion” names a motive state that may or may not be felt, then attributing basic fear to a rat is not attributing an experience to it. Activating an affect program “does not necessarily involve the generation of any feeling,” which “allows us to ascribe basic emotions to infants and non-human animals even in the absence of agreed upon dependent measures for studying feelings in creatures without language.”
The contrast with LeDoux is instructive, because they start from the same premises and reach opposite practical conclusions:
- LeDoux: the feelings question is unanswerable, so change the subject — study survival circuits, and reserve “fear” for the feeling we cannot verify in animals.
- Scarantino: the feelings question is unanswerable, so make it unnecessary — “fear” names the motive state, which behavioural methods reach, and the feeling is a separate question one need not ask.
Both leave the hard question exactly where it was. Neither claims to know what a rat feels; Scarantino’s contribution is that his vocabulary lets comparative work proceed without the question being live at all, whereas LeDoux’s requires a neologism to achieve the same insulation. Whether that is an advantage or a relabelling of the same silence is unresolved — and note it does nothing for infants, where the same reasoning applies but where the wiki’s developmental material has an independent stake in what is felt.
What LeDoux would accept as progress
Not a resolution, but a constraint: “if we can find neural correlates of conscious feelings in humans (and distinguish them from correlates of unconscious emotional computations in survival circuits), and show that similar correlates exist in homologous brain regions in animals, then some basis for speculating about animal feelings and their nature would exist. While such speculations would be empirically based, they would nevertheless remain speculations.”
Item 10 of his future-directions list makes this the research program: study human conscious feelings vigorously — both full-blown (fear, joy, sadness, shame, embarrassment) and coarser (pleasant/unpleasant) — precisely so the animal question becomes askable.
Status: open
Open by the principal’s own framing — he does not claim to have shown animals lack feelings, only that the question is currently unanswerable and that the field should stop assuming an answer. The core-affect middle position (animals may have crude valenced states without human-style differentiated feelings) is the most likely site of convergence and is entertained but not developed in the source.