How Small Can a Conscious Brain Be? Bees, Blindsight and the Minimum

Abstract

A bee has about a million neurons in a cubic millimetre. A human has around eighty-six billion in roughly 1,300 cubic centimetres. The obvious inference is that one of these can support conscious vision and the other cannot.

Marco Tamietto, Davide Orsenigo and Lars Chittka argue that the obvious inference is not available, because a neurological condition in humans severs the link between sophisticated visual behaviour and visual awareness in the opposite direction.

This is a proposal for research rather than a report of results, and we treat it as such. What makes it worth covering is that it puts a number on a question our other work in this department leaves abstract.

1.The Condition That Breaks the Intuition

Blindsight follows damage to the primary visual cortex, area V1. Patients report blindness in the corresponding part of the visual field — they say, sincerely, that they see nothing there.

Asked to guess, they perform far above chance. They localise stimuli, discriminate motion and wavelength, and avoid obstacles. A broad range of visual behaviour survives without conscious visual experience.

This matters because it removes the most natural test. If complex visual behaviour could occur only with awareness, we could infer awareness from behaviour. Blindsight shows it cannot, in the one species where we can ask.

The authors' framing of the puzzle: much of what we assume requires conscious experience — obstacle avoidance, detecting rewards and threats, recognising conspecifics — turns out not to.

2.What Bees Do That Blindsight Patients Cannot

Here the argument turns, and it is the most interesting move in the paper.

Non-conscious abilities preserved in blindsight have characteristic limits. They lack behavioural flexibility, context-dependent integration, cross-modal binding, perceptual constancy and generalisation across situations.

Bees are good at precisely those things. They learn and recall complex visual patterns, group elements flexibly depending on context, and discriminate stimuli by abstract relational properties independent of specific physical features. They also transfer learned information between vision and touch — cross-modal recognition, which is on the list of things blindsight does not deliver.

So the comparison does not run as expected. On the specific capacities that distinguish conscious from non-conscious vision in humans, bees resemble the conscious case and blindsight patients resemble the unconscious one.

The authors' conclusion is careful: this cautions against assuming that a miniature nervous system necessarily precludes visual awareness. Not that bees are conscious. That the size argument does not do the work it is asked to do.

3.What They Propose Doing About It

The substance of the paper is a set of experimental paradigms adapted from human and primate work.

Perceptual rivalry, where perception alternates between competing interpretations of an ambiguous image, already occurs in flies, accompanied by alternating activity in visual neuropils that precedes the behavioural switch. Spatial cueing produces opposite outcomes depending on whether a cue was consciously perceived, which makes it a candidate dissociation test.

And there are confidence-based measures used successfully in primates and rodents: opt-out tasks, where an animal can decline a difficult discrimination, and post-decision wagering, where it bets on its own accuracy. For free-flying bees these could be implemented through differential reward choices.

The target is specific. Demonstrating a dissociation between accuracy and confidence would be a comparative signature of conscious access to visual content — a bee that performs well but knows when it is guessing.

None of this has been done. The authors state plainly that these paradigms remain untested in bees and offer predictions rather than findings.

4.The Anatomy, and Why It Is Not Decisive Either

The neural comparison cuts both ways. In primates, V1 occupies about three per cent of the cortical surface in humans and up to fifteen per cent in macaques, and its loss abolishes conscious vision. So a small structure can be indispensable.

Bee brains have no laminated cortex. Neurons are packed into discrete neuropils, and bees lack the pyramidal cells that studies in mammals link to the sustained dynamics of conscious processing. What they have instead are Kenyon cells in the mushroom bodies, serving analogous integrative functions.

The authors' suggestion is that dense feedback loops between visual and central neuropils might permit global availability of visual representations — a functional parallel to cortical broadcasting, achieved with far fewer neurons. Which, if it held, would matter for global workspace theory as much as for entomology.

Insects and vertebrates diverged more than five hundred million years ago, so any similarity is convergence rather than shared inheritance. That makes functional analogy the only currency available, and functional analogy is exactly what is disputed.

5.Why This Belongs Beside the AI Question

In our series on indicators of AI consciousness, Ned Block argued that AI systems and simple animals are competitors for attribution of consciousness. Extrapolate from humans on computational properties and machines score well; extrapolate on the biological machinery underneath and simple animals do.

This paper takes the second branch seriously and asks what the minimum is. If a million neurons in a cubic millimetre can support visual awareness, the biological requirement is far less demanding than assumed, and the space of possible conscious systems is much larger.

If it cannot, the question becomes what specifically is missing — and that answer would constrain the machine case as well.

Either way the bee is more informative than the chatbot, for a reason worth stating: we know the bee evolved, we know roughly what it is made of, and we can run the experiment. None of those hold for a language model.

Editorial Comment

MindHeaven® makes no claim about insect consciousness, animal welfare or anything connected to either. We sell supplements to adults.

One disclosure the paper makes and we pass on: Lars Chittka's work is supported by a grant from Open Philanthropy for addressing evidence gaps in the science of insect sentience. That is not a criticism — the funding is disclosed, and someone has to fund the question. A reader weighing a paper that argues insect awareness deserves investigation is entitled to know that its senior author is funded to investigate insect sentience.

We record it because we would want the same standard applied to a trial of a compound we sell, and because we applied exactly that standard this week to a positive theanine study with manufacturer co-authors.

Nothing here shows that bees are conscious, and the authors do not claim it. They claim the question is answerable, and describe how.