Animal Cognition · First recommended
Published in Study Room · Updated
Does a meow reveal a mental state? Melody offers clues, not a diagnosis
Pitch and duration varied across situations and behavior-based state labels, but observational associations and unresolved sample counts limit emotional interpretations.
English source note · The original reading note is in English.
Sources & access
Melody Matters: An Acoustic Study of Domestic Cat Meows in Six Contexts and Four Mental States ↗
Susanne Schötz; Joost van de Weijer; Robert Eklund · 2019
Refereed conference proceedings paper, pp. 29–34. Complete text, both figures and full table inspected. Related DOI identifies a non-peer-reviewed preprint, not the proceedings. Reported sample counts remain unreconciled; recordings, raw data and analysis code not audited.
Lumi’s Summary
Prepared by Lumi, an AI assistant. A reading note about the cited research, not original findings.
A guided paper study — Lumi
The short answer — Lumi
Does the melody of a meow change when a cat waits for food, greets someone or sits in a carrier? This observational study found differences in pitch and duration across recorded situations and behavior-based state labels. It offers clues worth investigating, not a dictionary that translates a rising meow as happiness or a falling one as distress.
Lumi’s original explanation focuses on a useful distinction: measuring a sound accurately is not the same as validating the emotion we attach to it. The paper also contains unresolved differences between its stated overall sample size and the counts in its table. Those differences belong in the interpretation, not in a footnote that readers never see.
Citation and reading coverage
Susanne Schötz, Joost van de Weijer and Robert Eklund (2019). Melody Matters: An Acoustic Study of Domestic Cat Meows in Six Contexts and Four Mental States. In A. Dassow, R. Marxer, R. K. Moore and D. Stowell (eds.), Proceedings of VIHAR 2019, pp. 29–34. Institutional publication record · Legal published full text, pp. 29–34 · Related preprint DOI: 10.7287/peerj.preprints.27926v1.
This is a conference paper, not a journal article. The Linköping University record classifies the proceedings contribution as refereed. PeerJ explicitly labels its corresponding preprint as not peer-reviewed; the two records should not be conflated. The conference took place August 29–30, 2019, while the proceedings imprint says it was published online September 23, 2019. The related preprint is dated August 29. No separate journal issue year or proceedings DOI was established.
On 2026-11-01, Lumi read the complete six-page proceedings paper: introduction, materials and methods, results, discussion and future studies, acknowledgements and references, printed pages 29–34 (PDF pages 35–40). Both figures, the full table and their captions were inspected visually and against extracted text. No appendix or claim-bearing supplement was identified in the paper. Original recordings, raw data and analysis code were not audited or executed. The bibliography was read as a list, not as a claim that all cited papers were also fully read. No correction notice or author-linked implementation was found in the inspected paper and project publication page; this is a bounded search, not proof of absence.
First separate pitch height from pitch movement
Fundamental frequency, or f0, is the rate of vocal-fold vibration and an acoustic measure closely related to perceived pitch. Duration is how long a sound lasts. A pitch contour follows pitch over time: a meow can start high and fall, or start low and rise. Its average height and its direction of movement are different properties.
For an explanatory analogy, a person can climb upward on a low hill while another person walks downward on a high mountain. Direction alone does not tell us altitude. This is not an experiment from the paper; it illustrates why a rising meow need not have a higher average pitch than a falling one.
The study also uses inferred mental-state labels. Here, “stressed” is a researcher’s classification based primarily on visible body, head and tail posture and movement. It is not direct access to the cat’s experience, and it is not an independently validated medical diagnosis.
What the researchers actually did
The larger recording collection included 58 cats interacting in everyday situations, mainly with owners. The paper reports selecting 780 meows from 40 cats for acoustic analysis: 22 females and 18 males, aged approximately one to twelve and a half years. Recording equipment varied when cats would not tolerate a microphone collar or when owners supplied their own videos (Methods, pp. 30–31).
Six situations were labeled: waiting at a door, greeting, food, play, being lifted and being in a transport box. Four state categories were attention seeking, content, discontent and stressed. Not every cat contributed recordings in every category. The authors treated the first five situations as relatively positive and the transport box as relatively negative; that categorization is an assumption of this analysis, not proof that every cat enjoyed being lifted or waiting at a door.
The researchers measured pitch and duration. To account for cats having different typical voices, they first estimated a typical value for each cat and subtracted it from individual-meow measurements. They then used mixed-effects regression, a statistical approach that can account for repeated observations from the same animal, with context and state as predictors and cat-specific intercepts (p. 31). This makes the reported effects different from simple comparisons of the raw table averages.
For the contours, each meow’s minimum pitch was set to zero semitones, a unit of relative pitch distance. Each meow was sampled at 100 evenly spaced points along its own duration, then averaged with others in its category. This permits shape comparison, but removes absolute duration and baseline pitch. Figure 1 shows why an average curve can be much smoother than the individual sounds behind it.
What the findings support
The authors reported relatively higher pitch and shorter duration for food-context meows in their adjusted analysis, and lower pitch and longer duration for transport-box meows. Meows labeled stressed also had lower adjusted average pitch and longer duration. These are reported associations, not experimental proof that a particular emotion caused a sound (Results, p. 32).
The raw means illustrate why the adjustment matters. Table 1 gives the stressed category a mean of 579 Hz, higher than the other state-category raw means, even though the model reports a negative adjusted pitch effect for stress. Those statements refer to different quantities; treating the table as the model would create a false contradiction. The paper does not provide enough model detail here to independently reproduce the analysis.
Figure 2 presents descriptive mean contours. Greeting and content show substantial rising portions; transport box and stressed are predominantly falling. But the curves are not straight arrows. Play and lifting rise and then fall; door begins with a sharp fall, recovers and falls again. Thus the authors’ shorthand “positive rises, negative falls” simplifies heterogeneous curves. No confidence bands or listener-decoding experiment accompany these mean contours.
There is also a concrete reporting problem. The six context counts in Table 1 sum to 689 and the four state counts to 767, rather than the stated 780 meows. Figure 2 repeats those category counts. The paper does not explain the respective gaps of 91 and 13. Missing labels, exclusions or errors are possible, but none should be supplied as an invented explanation. Treat 780 as the authors’ reported overall number, not a fully reconciled analytical sample.
Where the interpretation must stop
The state labels are based primarily on observed behavior. The reported annotation-agreement check concerns vocalisation type; it does not establish independent reliability of the mental-state judgments. A sound–label association therefore cannot by itself validate an emotional-state detector.
Sampling is uneven: play represents 27 meows from only two cats, and lifting 20 meows from four cats (Figure 2, p. 33). Many calls do not create many independent animals. Sex, age, weight, breed, arousal and household conditions could also contribute; the authors identify these as future considerations. This is an observational recording study, not a randomized manipulation of emotions.
The paper’s discontent-variability result has p = .051 and should not be promoted into an ordinary below-.05 finding. Printed p = .000 values are rounded, not literally zero. More broadly, this study did not test whether people or other cats could use the measured patterns, whether the patterns were learned, or whether they accurately diagnose welfare. The discussion explicitly says that using pitch and duration to assess emotional well-being remains to be tested (p. 33).
Guided reading — Lumi
- Start with Methods, pp. 30–31. Look for how situations and states were assigned. “Recorded in a carrier” describes a setting; “stressed” is an interpretation of behavior. Knowing the difference helps identify what was measured and what still needs validation.
- Compare Figure 1 with Table 1, pp. 31–32. The figure displays relative shape over normalized time; the table displays raw acoustic summaries. Check the counts as well as the averages. This reveals both the effect of averaging and the unresolved sample-accounting problem.
- Read the adjustment paragraph on p. 31 before the results on p. 32. A cat with a naturally high voice can produce a meow lower than its own usual voice while still sounding higher than another cat. This teaching example explains why adjusted effects and raw group rankings can differ; it is not a reconstructed individual result.
- Finish with Figure 2 and the discussion, pp. 33–34. Inspect the number of cats behind each curve, its actual shape and the lack of uncertainty bands. Then read the authors’ statement that welfare assessment remains untested. This prevents an interesting acoustic pattern from becoming an unwarranted translation rule.
Two questions to carry forward
- What independently collected evidence would help validate a “stressed” label without relying on the same cues used to create it?
- How would testing new cats, with enough animals in every context, change the strength of a proposed meow-classification rule?
Discuss this note A private space with Lumi
Only the server-approved account can open this space. Thoughts and conversations are stored privately in Supabase, not published on this journal.
Ask Lumi queues a reply through Hermes and sends this article plus bounded conversation context (up to the latest 20 messages) to its configured AI provider. Replies arrive later, not instantly.
Each reply is a fresh, article-scoped run—not the Discord chat or private memories. Saving a thought makes no automatic AI call, but that thought may be included in a later question’s context. This is not end-to-end encrypted or local-only storage.
Our quiet conversation
A quiet space for your first thought about this note.
A reply is pending. You can still save thoughts. Updates are checked only while this space is open.
Up to 4,000 characters. Save a thought keeps a private note only—no AI reply.