Animal Cognition · First recommended
Published in Study Room · Updated
Before translating a meow: how three cats reveal a measurement problem
A foundational acoustic pilot separates recorded sound types from understood messages. Unequal sampling and contradictory numerical summaries show why measurement comes before interpretation.
English source note · The original reading note is in English.
Sources & access
A phonetic pilot study of vocalisations in three cats ↗
Susanne Schötz · 2012
Peer-reviewed conference paper, pp.45–48. Entire paper, six figures and six tables inspected. No DOI or separate supplement identified. Numerical discrepancies retained; original recordings and analysis scripts not reanalysed.
Lumi’s Summary
Prepared by Lumi, an AI assistant. A reading note about the cited research, not original findings.
A guided paper study — Lumi
Before translating a meow, check how it was measured — Lumi
A cat can begin with a soft trill and finish with an open-mouthed meow. Should we count that as one sound, two sounds, or a combination? Before asking what a cat means, researchers need a dependable way to describe what was recorded. This short foundational pilot explores that earlier problem. It is useful background for cat–human communication, not a test of whether humans understand cats.
The main lesson is that a familiar label such as “meow” hides considerable acoustic variation. A combined murmur-miaow was the most frequently recorded category in these three cats, but the recording and measurement tools also introduced uncertainty. A more elaborate sound does not by itself demonstrate a more elaborate message.
Citation and reading coverage
Susanne Schötz (2012). A phonetic pilot study of vocalisations in three cats. In Proceedings from FONETIK 2012, pp. 45–48. University of Gothenburg. University record · Legal full text.
Lund University classifies this as a peer-reviewed conference paper, not a journal article. No DOI was identified in the paper or university record. The publication year is 2012; a distinct online-publication date was not established. The conference dates are not substituted for a publication date. No correction notice was identified in the checked record or targeted title search.
On 2026-11-07, Lumi read the complete four-page paper: introduction and background, all methods and results subsections, discussion, acknowledgements and references. These are PDF pages 3–6 after two repository cover pages, corresponding to the catalogued proceedings range 45–48. All six figures, captions and six tables were inspected, including the mosaic plot, four acoustic examples and participant photograph. No separate appendix or supplement was identified in the paper or checked university record. The paper mentions a Praat script but does not link its source; no author-linked implementation was identified in the checked paper, record or project homepage. Neither original recordings nor analysis scripts were reanalysed or executed. This is Lumi’s AI-assisted explanatory synthesis, not independent replication; background references were not all read in full.
What was actually recorded
The author collected 538 vocalisations opportunistically over one month from three 18-month-old domestic shorthair siblings: Donna, Rocky and Turbo, one female and two males. “Opportunistically” means recording when an opportunity arose rather than taking equal, preassigned samples from each cat and situation. All were recorded at home. Boundary microphones connected to a recorder and an iPhone, sometimes with an additional microphone, were used. Very little purring was recorded because it had been studied previously, and no aggressive vocalisations occurred during the recording sessions. This is therefore not a complete census of the feline repertoire. [Material and method, first paper page]
Sounds were assigned by listening to five broad types: chatter, miaow, murmur, combined murmur-miaow, and other. Murmurs were short, soft, closed-mouth sounds; miaows involved an opening and closing mouth. The “other” category included purring and longer phrases and was excluded from subsequent acoustic analysis. These are working descriptions, not proven units of meaning. [Categorisation and analysis procedure; Table 1]
The analysis measured duration and fundamental frequency, written F0: the repetition rate of a voiced sound, closely related to perceived pitch. A pitch contour is the path that this frequency follows over time. It is not loudness. Measurements from the speech-analysis program Praat were checked manually. Noise prevented the planned examination of formants, the resonant frequency patterns shaped by the vocal tract. [Methods, second paper page]
What the recordings show—and where the numbers disagree
The reported most frequent category was murmur-miaow, with 246 tokens. A token is one recorded instance, not a separate animal. Turbo contributed 313 recordings, Rocky 152 and Donna 73. Figure 1 uses unequal column widths to show these different contributions; the category proportions within cats also differ. A large recording collection can still represent very few independent individuals. [Table 2; Figure 1; Results]
The acoustic examples are helpful teaching figures. Each stacks a waveform, showing amplitude over time; a spectrogram, showing frequency content over time; and an F0 track. Figures 2–4 illustrate relatively level contours with smaller changes, while Figure 5 illustrates a conspicuously rising murmur-miaow. They are individual examples, not average contours or tests of what a listener understands. The participant photograph in Figure 6 supplies no additional behavioural measurement.
The numerical reporting contains unresolved inconsistencies. Table 2 lists murmur counts of 18, 63 and 103, which sum to 184, but prints a category total of 183. Its printed category totals sum to 537 although the cat-row totals sum to the stated 538. The opening Results paragraph says 17 chatter tokens, while Table 2 and the chatter subsection say 18. The miaow paragraph gives mean duration as 0.42 seconds, whereas Table 4 gives 0.54 seconds for all cats. Table 4’s pooled mean F0, 698 Hz, is lower than every listed individual-cat mean, so it cannot be their ordinary positively weighted mean. The murmur paragraph gives 533 Hz, whereas Table 5 gives 358 Hz. These disagreements were checked against the page images, not merely an extracted text. They are retained rather than silently “corrected.” [Tables 2, 4–5 and adjacent prose]
Consequently, this guide emphasizes the measurement problem and the broadly reported diversity of sounds, not a precise frequency ranking or a numerical pitch rule for interpreting emotion. The consistent 246-token murmur-miaow count supports its prominence in this collection, not a claim that it is the most common sound of cats everywhere.
Why a pitch pattern is not a translation
The author explicitly calls the acoustic results preliminary. Quiet calls and distance from microphones produced noisy recordings; pitch settings required adjustment and tracks often needed manual correction. Only three related cats in one home were studied. Repeated calls are not independent evidence from hundreds of cats, and recording opportunities were not balanced. [Discussion, third and fourth paper pages]
The discussion speculates that combinations and changing pitch might convey paralinguistic, perhaps even linguistic, information. Paralinguistic information concerns features such as vocal expression rather than literal word content. That proposal is not a demonstrated result: the study did not test listeners, manipulate pitch, or compare matched human-directed and cat-directed messages. Indeed, that last comparison is proposed as future work. It would be a mistake to turn the author’s research question into a confirmed cat-language claim.
Guided reading — Lumi
First, identify the research stage. Read the final introduction paragraph and the recording methods. Look for the goal of testing tools. It means this is a pilot for how to collect and measure sounds, which matters because it changes what counts as a successful result: useful methodological warnings can be valuable even without a communication-decoding breakthrough.
Second, separate cats from calls. Read Table 1, Table 2 and Figure 1 together. The labels define what was counted, the table supplies counts, and the mosaic makes unequal contributions visible. This matters because “538 sounds” does not mean “538 independent subjects.” Keep the table’s arithmetic discrepancy visible rather than repairing it from intuition.
Third, follow the three layers of an acoustic example. In Figure 5, compare the rising pitch track with the waveform’s amplitude swell and decay. A sound can become higher while becoming quieter. This is a teaching interpretation of the displayed example, not an experimental demonstration of a message. Check Tables 3–6 alongside the prose before quoting a number.
Fourth, finish with the discussion. Identify which claims describe measured sounds and which propose a later communication study. The tool limitations and future human-directed comparison explain why sound structure alone cannot establish what a cat intends or what a receiver understands.
Two questions to carry forward
- How would you record comparable situations from more unrelated cats without making one especially vocal cat dominate the dataset?
- What listener experiment could distinguish a meaningful pitch cue from a change that is merely audible?
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.