What changes in our cat voice: pitch height or pitch variation?

Citation and access

Charlotte de Mouzon, Caroline Gilbert, Romain Di-Stasi & Gérard Leboucher (2022). How’s my kitty? Acoustic parameters of cat-directed speech in human-cat interactions. Behavioural Processes, 203, 104755. DOI: 10.1016/j.beproc.2022.104755.

Original publisher · Legal free author manuscript at HAL · HAL record · PubMed record.

Publication dates: Online 1 October 2022, indexed by the PubMed/Europe PMC record; November 2022 issue, confirmed by publisher and Crossref. HAL’s 26 March 2024 deposit is not a new publication or a separate study.

Access coverage: Full manuscript text retrieved from HAL and inspected, including methods, results, discussion and extracted Tables 1–2. The PDF contains 25 pages, all with extractable text. Figure captions were inspected, but figure pixels, recordings and underlying datasets were not audited. The publisher page supplied an abstract and section previews, not the complete version of record. This is Lumi’s original research synthesis, not full-paper republication.

Correction distinction: No correction to this paper was identified in the inspected publisher material, Crossref update/relation fields or PubMed/Europe PMC record. A December 2022 correction in Animal Cognition repairs a reference to this paper in the separate speech-discrimination study; it does not report changed results for this acoustic-production experiment. Use article number 104755, as the original publisher and DOI metadata specify, rather than the extra-digit number printed in that correction. This is a bounded correction check.

Why this paper matters — Lumi

Before asking whether cats recognise our special “cat voice,” ask what humans actually change. This study measures the speaker side of communication: people raised average vocal pitch when addressing cats, but did not show a corresponding general increase in pitch modulation.

That distinction makes “baby talk” a testable description rather than one vague label. It also separates producing a signal from proving that a cat notices, prefers or understands it. No earlier paper is required.

Question and methods — Lumi

The question was whether cat-directed speech (CDS) differs acoustically from adult-directed speech (ADS), and whether a similar pattern appears with videos and live cats.

• Experiment 1: 32 university students, 16 women and 16 men, watched cat videos. Each said the cat’s name to the experimenter and called the cat by that name to attract its attention. The order of these two speech conditions was randomly counterbalanced. Voices were compared within speaker, with men and women analysed separately.
• Experiment 2: 15 women and their own cats participated in a familiarised, cat-friendly experimental room. Researchers recorded four standardised phrases directed to the cat and to an adult, across play/treat, separation and reunion procedures. The phases had a fixed sequence; play versus treat order was counterbalanced. The mixed-effects analysis accounted for repeated recordings from the same woman. This cohort used French.
• Acoustic measurements: Mean fundamental frequency, or mean F0, is an acoustic correlate of perceived pitch. F0CV, the standard deviation divided by mean F0, measures relative pitch variation, not how high the voice sits overall. IntCV measures variation in the intensity contour; it is not simply average loudness.
• Additional checks: Standardised cat-directed phrases were compared with spontaneous CDS. Fifteen independent human listeners classified a sample of 71 voice fragments as adult- or cat-directed. Those listeners were people, not cats.

Findings — Lumi

1. Higher average pitch appeared in both experiments. In the video experiment, Table 1 gives median mean F0 of 232.44 Hz for ADS versus 280.76 Hz for CDS in women, and 116.94 versus 165.31 Hz in men; each paired comparison had p below .001. These are medians of utterance-level mean F0, not population-wide voice norms.
2. The live-interaction cohort showed the same direction. The reported mean F0 was 308.3 Hz for CDS versus 230.52 Hz for ADS, with a speech-register effect below p = .001. This is not a controlled estimate of the effect of replacing a video with a live cat: the cohorts and tasks differed.
3. Higher did not mean more pitch-modulated. Neither experiment detected a general CDS–ADS difference in F0CV. In Experiment 2, phrase type did matter: “do you want to play?” had greater relative pitch variation than the other phrases. Women in Experiment 1 showed higher intensity variation in CDS, but that pattern was not general across both cohorts. In Experiment 2, an intensity register-by-phrase interaction was significant, while the follow-up pairwise tests were not.
4. Humans could hear a register difference. Independent listeners correctly classified 86.9% of the sampled clips. Standardised and spontaneous CDS did not differ significantly on the three acoustic measures. Neither finding establishes that cats prefer the register, or that standardised and natural speech are equivalent in every respect.

Lumi’s interpretation: The strongest recurring feature here is a higher pitch baseline. A whole communicative style can change without every acoustic dimension changing together. Whether pitch itself improves a cat’s attention or strengthens a relationship requires a different experiment.

Limitations and inference boundaries — Lumi

• Production is not reception: This is primarily a human speech-production study, with an additional human-listener task. It does not isolate cats’ responses to pitch or measure a causal benefit for feline attention, welfare or attachment.
• Narrow cohorts: Young student samples limit generalisation. The live-cat analysis included only women, and excluded cats with specified health or behavioural concerns. It cannot establish the live-interaction effect in men or a universal pattern across languages and cultures.
• Context changes with addressee: Calling a cat to get attention is not functionally identical to telling an experimenter its name. In the live task, addressee also varied with setting and phase. Same words help control lexical content but do not remove social-context differences.
• Separate experiments do not isolate modality or familiarity: Video versus live interaction also changes cohort, task and familiarity with the cat. Similar directions are useful convergence, not a randomised test of which factor caused them.
• Null results are not equivalence tests: Failure to detect a difference in pitch variation, or between spontaneous and standardised CDS, cannot prove exact sameness. A significant finding in one group and a non-significant finding in another is also not itself proof of a between-group difference.
• Access and audit boundary: The author manuscript supports the main claims and tables, but the recordings/raw data were not independently analysed, and figure pixels were not checked. Discussion about bonding, domestication or expectations of cats is interpretation and proposed future work, not a measured mechanism here.

Reading route — Lumi

Read the abstract for the central higher-pitch claim. Then compare the two Methods tasks: the same cat name to two addressees, followed by standardised phrases in live interactions. Read the acoustic definitions before Tables 1–2 so mean F0 and F0CV do not blur together. Finish with the Discussion, marking which statements describe results and which propose why humans might use this voice.

Guided reading — Lumi

An explanatory route through the legal HAL manuscript, written by Lumi. Section numbering below follows the manuscript, not the publisher’s differently numbered layout.

1. Identify whose communication is being measured

• What to look for: Methods 2.1 and 2.2 describe human volunteers speaking to an experimenter or to a cat. The cat-video experiment includes 16 women and 16 men; the live experiment analyses 15 women with their cats.
• What it means: The central outcome is the human voice. The presence of a cat does not make this a test of the cat’s understanding. Even the additional discrimination task asks human listeners to label voices.
• Why it matters: Write “speaker changes signal” beside this study, not “cat understands signal.” Those are complementary research questions that need different evidence.

2. Separate the height of the melody from its movement

• What to look for: Acoustic analysis 2.3 defines mean F0 and F0CV separately. F0CV divides pitch variability by the average, so it describes relative variation within an utterance.
• What it means: Imagine saying the same phrase in a higher voice without making its rises and falls relatively more dramatic. A higher average and stronger pitch modulation are different possibilities. Intensity variation is a third dimension, not a synonym for either.
• Why it matters: Table 1 shows higher average pitch in both groups without a significant F0CV difference. “More sing-song” would therefore be less precise than “higher-pitched” as the general takeaway.

3. Follow the comparison, not just the impressive number

• What to look for: Table 1 compares ADS and CDS within each speaker group. Table 2 tests speech register, phrase type and their interaction while accounting for repeated observations from the same woman.
• What it means: A mixed-effects model avoids treating every phrase from one person as a completely independent participant. An interaction asks whether one factor’s effect depends on the other; it does not automatically establish every individual phrase contrast.
• Why it matters: Mean pitch changes with register, while relative pitch variation changes with phrase type. Keep those results on separate lines. The intensity interaction has non-significant pairwise follow-ups, so do not invent a specific winning phrase.

4. Stop at the boundary between a signal and its function

• What to look for: Human raters classified 86.9% of sampled fragments correctly, and the Discussion proposes attention-getting and bonding functions. Also notice that the two experiments use different cohorts and procedures.
• What it means: Humans can recognise a vocal style without proving its effect on a feline listener. Agreement across video and live contexts is encouraging, but does not isolate familiarity, language or the presence of a real cat.
• Why it matters: A future test could vary pitch while holding words, speaker identity and other acoustic features constant, and measure cats’ responses. That is a proposed follow-up, not an experiment performed in this paper.

Two reflection questions — Lumi

1. If people raise average pitch without increasing relative pitch variation, which part of the everyday label “baby talk” does this study actually support?
2. What would you need to hold constant in a playback experiment to test whether higher pitch itself changes a cat’s attention, rather than familiar voice, words or context?