Cats Were Not Linked to Child Anxiety — One Pattern Was

The headline travelled the way these headlines do. Cats linked to more anxiety and behavioural problems in young children, said the wire copy, repeated across a dozen outlets in near-identical words over a couple of days in September. Somewhere a parent read it at eleven at night and looked across the room at an animal that has been part of the household since before the child could walk, and the question that formed was not an academic one.

The paper underneath that wave is narrower than the coverage, and in one respect it says close to the opposite. It is also not new: it appeared in the World Journal of Pediatrics in October 2025, nearly a year before the syndication run that carried it into family inboxes. What it reports, read in full, is mostly a null result with one fragile exception attached — and the exception is the part that became the headline.

The data come from INMA, a Spanish population-based mother-child cohort begun in 2003 to study early-life environmental exposures. This analysis drew on four of its regional arms — Valencia, Sabadell, Asturias and Gipuzkoa — where mothers were recruited at their first prenatal visit before week 13 of gestation between November 2003 and February 2008, with baseline enrolments of 855, 657, 494 and 638 respectively. The analytic sample was 1,893 families.

Pets were recorded by an interviewer at two points, when the child was 1 and again at 4-5, as the presence in the previous year of five categories: any pet, a dog, a cat, a bird, or other animals, the last covering small mammals, fish and reptiles. Because each category was asked twice, every family fell into one of four patterns for each animal: never, always, only at age 1, or only at ages 4-5. That four-way split is the design's central move, and it is what the coverage lost.

Mental health was measured once, at ages 7-8, using the Strengths and Difficulties Questionnaire completed by a parent — twenty-five items across five subscales, each scored from 0 to 10, with higher scores meaning more difficulty. The authors combined them into two broadband scales running from 0 to 20: internalizing, from the emotional and peer subscales, and externalizing, from conduct and hyperactivity. The analysis used negative binomial regression adjusted for cohort, child age and sex, maternal education, family social class and weeks of breastfeeding, with further covariates specific to each outcome, followed by six sensitivity analyses.

Most of the results are absences. Owning any pet was not significantly associated with internalizing or externalizing problems. Neither was owning a dog. Neither was owning a bird. In the adjusted models, the animals that account for the overwhelming majority of pets in these households produced nothing.

One category did produce a signal, and it was narrow: cats present at ages 4-5 but not at age 1. For those families the relative rate ratio was 1.37 (95% CI 1.05-1.79) for internalizing problems and 1.27 (1.03-1.56) for externalizing problems, both measured at ages 7-8. The paper's abstract prints the externalizing figure as 1.26 (1.02-1.56); the results section gives 1.27. Families who had a cat throughout — the larger and more intuitive group — showed no such association.

A second signal ran the other way. Families who always had animals other than dogs, cats or birds had a lower rate of internalizing problems, at 0.80 (0.66-0.96); the externalizing estimate for the same group, 0.89 (0.78-1.03), crossed one. The authors speculate that small caged animals are often the child's own responsibility and may build something through that.

Scale matters for reading any of this. The median internalizing score in the whole sample was 3, with an interquartile range of 1 to 5, and the median externalizing score was 5, range 3 to 8, on scales that run to 20. In the bivariate comparison, the cat-only-at-4-5 group's median internalizing score was 4, range 2 to 6. Cats were also uncommon: 4.9% of families had one at both check-ins, against 11.6% for dogs and 24.0% for any pet at all.

A study that tests five animal categories against two outcomes across four ownership patterns is running many comparisons, and some will clear a significance threshold by chance. The authors knew this and applied Tukey's test and a false-discovery-rate correction. The result, in their words, was a borderline association between never having cats and having them at 4-5 years for internalizing problems, at P = 0.063, with the link to externalizing problems ruled out. The protective finding for other animals survived at P = 0.038.

This is worth stating plainly because the paper is not internally consistent about it. The abstract says the cat associations remained after multiple comparison testing and sensitivity analyses. The results section reports a borderline P value for one of them and the elimination of the other. A reader who stops at the abstract, as the coverage evidently did, gets a firmer finding than the body of the paper supports.

A further wrinkle sits in the sensitivity analyses. What was recorded was ownership, not whether the animal lived with the child; cohabitation data existed only at ages 4-5 and only for 1,528 participants. Comparing the two, cat ownership gave 1.16 (0.99-1.36) for internalizing against 1.19 (0.95-1.48) for cohabitation, and 1.19 (1.05-1.34) against 1.15 (0.97-1.37) for externalizing. The authors also report that only 32.4% of cohabiting cats fell into the only-at-4-5 group at all.

The most interesting sentence in the paper is one the coverage did not carry. Discussing why cats might show up this way, the authors note that cats are often chosen by families with children who have mental health conditions, because they need less maintenance than a dog — and then state that this hypothesis could not be tested here, as there was no information on mental health at earlier follow-ups. The direction of the arrow is, on the paper's own account, unresolved.

The study counts it as a strength that exposure was measured before the outcome, and in the narrow sense it was: pets were recorded at 1 and 4-5, behaviour at 7-8. But a household acquiring a cat when a child is four, in a year when that child is already difficult to settle, produces exactly the pattern reported, with the causality running backwards. Nothing in the design separates the two, because nothing measured the child's behaviour before the cat.

Two smaller cautions travel with the number. The questionnaire was completed by a parent, and the authors themselves cite work suggesting cat owners differ in ways that could colour how they rate a child's behaviour. And the other explanations they float — indoor animals offering limited engagement, or toxoplasmosis exposure — are hypotheses drawn from other literature, not things this study measured or tested.

The useful content of this paper is mostly the nulls. Across nearly two thousand families followed from pregnancy into primary school, the pets that most households actually have showed no measurable relationship with emotional or behavioural difficulties at ages 7-8. That is a reassuring finding, and it is the one the syndicated coverage omitted in favour of a subgroup that the paper's own correction left borderline.

The authors note that theirs is the first study to examine different types of pet ownership during the preschool years against mental health at primary-school entry. That is precisely the standing of the result: a first pass at an under-studied question, in which one of twenty or so comparisons came back positive and did not survive correction cleanly. Findings of that shape are how a field decides what to measure next. They are not how a household settles anything, and the distance between those two uses is the whole of what the week's headlines got wrong.

Read in full, this is a reassuring paper that was reported as an alarming one. The bulk of its content is null: in 1,893 families followed from pregnancy to primary school, having a pet, a dog or a bird bore no measurable relationship to emotional or behavioural difficulties at ages 7-8. The exception is one cell of a four-way split in the least common pet category, it did not survive the authors' own multiple-comparison test intact, and the authors name a reverse explanation they had no data to test — that families already coping with a child's difficulties tend toward the pet that asks less of them. That the abstract states the associations held after correction while the results section reports P = 0.063 and a ruled-out second outcome is a real problem with the paper, and it is the seam the coverage walked straight through. The honest summary is that this is a first look at a genuinely under-studied question, and that the most solid thing in it is the long list of things that turned out not to matter.