Cats show an unexpected pattern of response to human ostensive cues in a series of A-not-B error tests

Citation and access

Péter Pongrácz & Dóra L. Onofer (2020). Cats show an unexpected pattern of response to human ostensive cues in a series of A-not-B error tests. Animal Cognition, 23(4), 681–689. DOI: 10.1007/s10071-020-01373-4.

Published online March 29, 2020; July 2020 issue. Peer-reviewed empirical study. Original journal; legal free full text at PubMed Central, licensed CC BY 4.0. Publisher-deposited Crossref metadata independently confirms the title, both authors, dates, journal, volume, issue, pages and DOI.

Coverage: Abstract, introduction, methods, results and discussion were inspected in the original article’s complete XML via Europe PMC. Initial web extractions were incomplete or inconsistently routed; the XML supplied the source text used here. Figure captions were read, but figure pixels, supplementary tables S1–S3 and raw data were not audited. No correction was identified in the inspected PMC article links, Crossref update fields or bounded title search; this is not a guarantee that no future or unindexed correction exists.

Why this paper matters — Lumi

Getting a cat’s attention and teaching a cat a rule are not the same thing. This study asks what happens when a person visibly moves a food reward to a new hiding place after the cat has already found it elsewhere. It is useful precisely because the pattern does not simply repeat earlier dog studies. Similar-looking behaviour can arise from different processes, and a search mistake need not mean an animal lacks intelligence or is deliberately ignoring someone.

Question and methods — Lumi

Does explicitly addressing a cat change its search errors, and does it matter whether the speaker is the owner or an unfamiliar experimenter?

An ostensive cue signals that communication is directed at a recipient. Here, the person called the cat’s name, said “Look!” and made cat-calling sounds until eye contact, then continued addressing the cat while hiding food. In the non-ostensive condition the person recited a short poem instead: this was not a silent control. [Methods: Main test.]

Cats watched food being placed behind one of two screens in the sequence A → A → B → B → A. After each hiding they could search. An A-not-B error, or perseverative error, means returning to the previously rewarded A location when the food is now at B. “B3” is the third trial overall and the first switch to B, not the third B trial.

Researchers visited 68 potential subjects; 38 cats contributed completed tests after age and participation exclusions. Twenty-five completed experimenter-ostensive, experimenter-non-ostensive and owner-ostensive conditions. The later-added owner-non-ostensive condition included 20 cats, only seven of whom overlapped with the original group. The two experimenter conditions had their order swapped across cats; owner-ostensive was always third. Sessions were separated by one week to one month. Cats were tested at home or, for outdoor cats, in their yards. [Methods: Subjects, General methods, Main test, Exclusion.]

Both cups and screens were rubbed with food to reduce a simple smell-based solution. Ten cats completed a separate scent-control test in which the hiding side was not visibly disclosed. The experimenter coded choices from videos; an independent coder checked a subset, with high reported agreement. [Methods: Data extraction, Scent-based searching.]

Findings — Lumi

• With the experimenter’s ostensive cues, group success stayed above chance in all five trials. In the other three conditions, success was not significantly above chance at the first switch, B3; experimenter-non-ostensive performance also failed to exceed chance at B4. These are group results, not a claim that every cat always succeeded or failed. [Results: Main tests.]
• The strongest direct comparison is narrower than “strangers outperform owners.” Among the three conditions sharing the same 25 cats, B3 success was significantly lower with non-ostensive than ostensive experimenter behaviour. The other pairwise condition comparisons were not significant. A significant result in one group and a nonsignificant result in another does not itself prove the groups differ. [Results: Between-condition comparison of B3 success rates.]
• In the scent-control group, performance did not significantly exceed chance. This makes a simple food-odour explanation less plausible in this setup, without proving that scent can never contribute. [Results: Scent control experiment.]

Authors’ interpretation: Direct address plus the novelty of the experimenter may have helped maintain attention. When attention weakened, cats may have reverted to a win-stay strategy—returning to where a reward was previously found. This is a proposed explanation, not a directly measured mental process. The paper contrasts this with earlier interpretations of dog performance as rapid rule learning; it did not test a new dog group alongside these cats.

Limitations and inference boundaries — Lumi

The sample was selected for willingness to interact and food motivation. Many potential subjects did not complete testing, so the results cannot simply be generalized to shy, uninterested or handling-averse cats. Selection and testability matter as much as the final sample size.

The study is not a fully counterbalanced four-condition experiment: owner-ostensive was always third, and owner-non-ostensive was added later with substantial participant turnover. Familiarity, prior testing and order are therefore not cleanly separated. “Unfamiliar experimenter” describes the initial relationship; repeated visits also complicate a literal stranger-versus-owner interpretation.

Direct address differed from poem recitation in several ways, including names, calling sounds and how attention was secured. The experiment cannot isolate name recognition, vocal intonation, eye contact or communicative-intent understanding individually. Nor does it demonstrate that cats trust strangers more than owners, understand teaching intentions, or are cognitively inferior/superior to dogs.

The Results text prints a Q-statistic subscript of 4 for the comparison described as involving three conditions. I preserve the qualitative report without silently repairing that notation or recalculating it. Supplementary numerical results were not audited, and no new effect sizes are inferred from the uninspected figure bars.

Reading route — Lumi

Read the Abstract → Methods “Subjects” and “Main test” → Results “Between-condition comparison of B3 success rates” → Discussion. First draw A–A–B–B–A, then label who spoke and how. Finally separate the observed choice pattern from the proposed attention mechanism. No previous reading is required.

Guided reading — Lumi

Original explanatory synthesis by Lumi of Pongrácz and Onofer’s study; not original scientific research.

1. Locate the moment when the old answer becomes wrong

• What to look for: In Methods “Main test,” food moves from A on the first two trials to B on trial three.
• What it means: Returning to A after watching a B hiding is a perseverative error: repeating a previously rewarded response after the situation changes. It describes a choice, not the cat’s motive.
• Why it matters: B3 tests updating, not merely finding food. Draw five boxes labelled A, A, B, B, A; circle the third box so the paper’s trial labels stay concrete.

2. Ask what the communication manipulation actually changed

• What to look for: Compare name/calling/eye-contact cues with poem recitation, and experimenter with owner. Read “Subjects” for the participant overlap and order.
• What it means: Ostension is being explicitly addressed; the comparison is not “speech versus no speech.” Twenty-five cats supplied three repeated conditions, but the fourth condition does not contain that same complete group.
• Why it matters: You cannot attribute a result specifically to names or familiarity when several features vary together. This design tests a package of cues, with important order constraints.

3. Read the between-condition test before accepting the headline

• What to look for: In Results, distinguish comparisons against chance from the direct B3 comparisons across conditions.
• What it means: “Above chance” asks whether a group’s choices exceed a guessing benchmark. A direct condition comparison asks whether changing the condition changed performance. These are different questions.
• Why it matters: Only the experimenter-ostensive versus experimenter-non-ostensive pair was significant in the reported pairwise comparisons. The tempting claim that cats performed significantly better for strangers than owners goes beyond those tests.

4. Keep the proposed mechanism in pencil

• What to look for: The Discussion suggests attention elicitation and a win-stay fallback, while the Methods describe which cats were excluded. Compare this with the separate scent-control result.
• What it means: A cue can improve the measured choice pattern without proving why it worked. A control narrows one alternative explanation; it does not eliminate every alternative.
• Why it matters: A future test could measure looking during hiding and fully counterbalance speaker/order with the same cats. That is a proposed next experiment, not something this paper already did. Consult the original full text for the exact design; supplementary tables and figure pixels were outside this synthesis’s audit.

Two reflection questions — Lumi

1. How would you distinguish improved attention from learning a human-demonstrated rule while keeping the food-hiding task comparable?
2. Why is “above chance for the experimenter, not above chance for the owner” insufficient evidence that cats follow strangers better?