Does your cat actually care, and why two labs got opposite answers. · published 29 September 2026
The famous 2019 cat-attachment result is real, but a more tightly controlled study four years earlier, using nine counterbalanced episodes instead of three fixed phases, found no secure attachment at all — and a 2025 paper argues a test built for primate infants was never the right tool for a solitary hunter, so the number that travelled online was one of two answers.
These are the working notes, unedited. Everything the video
states as fact comes from here, and nothing else may be stated as fact.
The sections on limits and on what the finding does not show are
part of the brief, not an afterthought — see
the method.
The claim that went round. In 2019, Kristyn Vitale, Alexandra Behnke and Monique Udell published “Attachment bonds between domestic cats and humans” in Current Biology, and the headline everywhere was that cats bond with their people much as babies do. The numbers are why: of 70 classifiable kittens, 64.3 per cent came out securely attached; of 38 adult cats, 65.8 per cent. In human infants the figure is about 65 per cent. A six-week socialisation course afterwards changed nothing, which suggests the style is stable rather than trained.
How that was measured. The Secure Base Test, borrowed from infant research and already used on dogs. Three phases, two minutes each: the cat and its person in an unfamiliar room together, then the person leaves, then the person comes back. What is scored is the reunion - whether the animal settles and then gets on with exploring, which is the secure pattern, or stays clingy or avoidant, which is not.
The study that found the opposite. Four years earlier, Alice Potter and Daniel Mills published “Domestic Cats Do Not Show Signs of Secure Attachment to Their Owners” in PLOS ONE. They used a fuller version of the same paradigm - the Ainsworth Strange Situation, nine episodes across two conditions, with the order counterbalanced so that the sequence itself could not produce the result. Their conclusion was that adult cats are largely autonomous even in their social relationships, and do not use a person as a source of safety. Only vocalisation behaved as the attachment account predicts. They suggested cats do prefer their owner to a stranger, but that this is built from ordinary interaction rather than being a security-giving bond.
Why both can be defended. Vitale has the far bigger sample - 108 animals against 18 - and Potter and Mills have the more demanding protocol. Nine counterbalanced episodes is a harder test than three fixed phases over six minutes, and the shorter test is the one that found attachment. That is a real difference in design and not a detail: Potter and Mills built their version specifically because earlier work lacked those controls.
And a third position. A 2025 paper titled “The biological irrelevance of ’Cattachment’ - It’s time to view cats from a different perspective” argues the whole frame is wrong - that importing a paradigm built for an intensely social primate infant, then applied to a cooperative-hunting dog, onto a solitary hunter with a short domestication history, asks a question the species was never shaped to answer.
Limits
Neither study is large by the standards of human attachment work. Both classify a relationship from a few minutes of behaviour in one strange room. And “secure attachment” is a technical label from a specific coding scheme, not a measure of affection - a cat can matter to you, and you to it, with the label going either way.
What this does not show
Not that cats do not care about people. Not that the 2019 result is fake, or that the 2015 one is outdated because it is older - it is the more controlled of the two. The honest position is that a finding which travelled as settled is contested at the level of method, and the version you were shown online had one of the two studies in it.