Skip to content
NatureHQ

Claim check

Can a fish recognise a human face?

Mostly supported

The core is right; the usual phrasing overstates part of it.

A fish can be trained to pick one human face out of dozens, which is a genuinely important result — it shows the task does not require the specialised brain structures usually invoked to explain it. But that is discriminating a photograph, not knowing who you are. No study has shown a fish recognising its owner.

The claim as it circulates

“Fish can recognise their owner’s face, proving they are far more intelligent than anyone thought.”

Where you may have met it: Coverage of the archerfish face-discrimination experiments; Aquarium and pet-keeping communities; Posts arguing that fish are underestimated

Taking it apart

What was claimed
That fish recognise individual humans, and specifically that a pet fish knows its owner’s face.
What was actually observed
Archerfish, which knock prey off vegetation with a jet of water, were trained to spit at one human face displayed on a screen above the tank. They then chose that face from among dozens of unfamiliar ones well above chance, and continued to do so when the images were converted to greyscale and standardised for head shape and brightness.
What the evidence supports
That an animal with no neocortex and no evolutionary history of needing to tell people apart can perform fine-grained discrimination between human faces, using the internal features rather than colour or outline.
What it does not support
That fish spontaneously recognise familiar people. The fish were trained on one face and rewarded; they had no prior relationship with anyone in the images. It also does not show that fish process faces *as faces* rather than as complex patterns that happen to be faces — a distinction that matters, and that this design cannot settle.

This study earns its place for a reason that has nothing to do with fish being underestimated. Face discrimination in animals is usually explained by pointing at specialised regions in mammal and bird brains. An archerfish has neither, and does the task anyway. That is a problem for the explanation, not a compliment to the fish.

It also illustrates the distinction this whole page rests on. Telling two faces apart, recognising a face you have seen before, and knowing whose face it is are three different abilities requiring different evidence. Popular coverage routinely collapses them, and almost every overstated claim about animal face recognition comes from doing so.

Where the evidence lives

The claims underneath

Each one carries its own evidence, scope and caveats. Expand any of them to reach the studies.

Telling one human face from another does not require a large or mammal-like brain

Well supported

Good evidence backs this, though some details remain open.

Discrimination between human faces has been demonstrated in animals without a neocortex or a large brain, including archerfish trained on photographs and honey bees trained on face-like patterns, as well as in wild birds and domestic horses.

Who this applies to
the species in which human-face discrimination has been directly tested
Studied in
Toxotes chatareus, Apis mellifera, Mimus polyglottos, Equus caballus
Why we rate it this way, and what the caveats are
Well supportedHigh confidence

Independent demonstrations in a fish, an insect, a wild bird and a domestic mammal, using different methods. The convergence across such distant lineages is what carries the claim.

How far it can be extended

Assembled from primary studies in each species. The claim concerns which kinds of nervous system can perform the task, so its strength comes from the taxonomic spread rather than from any single result.

Caveats

  • Discriminating a face from other faces is a smaller claim than knowing who someone is.
  • Several results are trained discriminations of photographs rather than spontaneous recognition of familiar people.
  • Which cues each animal uses — face, gait, size, movement — is usually not isolated.

Still unanswered

  • Do any of these animals process faces as faces, or as complex patterns that happen to be faces?
  • How many of these abilities are used on humans in the wild rather than only under test?

Last reviewed 2026-08-09

The evidence (5 studies)

Go deeper

Recognising individual humans

A crow, an elephant and a bee all "recognise" people. Not one of them means the same thing by it.

Last reviewed 2026-08-09