Research & Evidence

Do Cats Read Other Cats' Emotions? What One Study Measured

By
By Clara LindqvistCertified Canine Behavior Consultant (CCBC) & Feline Environmental Specialist
Read time
4 min read
Published
Updated
Conceptual illustration of a cat observing a photo card and speaker in a familiar room; not a study photograph.

A 2026 PeerJ study asked a narrow question: can domestic cats respond differently to visual and auditory cues that represent other cats in different emotional states? In 24 cats, the visual task changed gaze and interaction patterns when the pictured cat showed negative valence and high activity. The isolated audio task did not produce the same pattern. That is evidence about a controlled experiment, not a way to label a cat's feelings from a home video.

Conceptual illustration of a cat observing a photo card and speaker in a familiar room; not a study photograph.
A conceptual scene for a study about cat responses to visual and auditory cues; it does not reproduce the study context or a study subject.

Short answer

The study found a visual response pattern that combined valence and activity. Cats looked longer and more often at the negative, high-activity image, and active images drew longer interaction regardless of valence. Valence and activity did not significantly change the measured responses to playback sounds. The result does not show that cats decode human-style emotions or that one response predicts a cat's temperament.

What the study tested

Spoelstra and colleagues presented 24 domestic cats with four visual and four auditory trials. The researchers used two dimensions for each stimulus: valence, meaning more positive or more negative, and activity, which they used as a proxy for arousal. The categories were designed to separate those dimensions rather than treat every posture or sound as one simple emotional label.

Route Positive, higher activity Positive, lower activity Negative, higher activity Negative, lower activity
Visual Approach posture, tail up Roll posture Back arch Crouch
Auditory Trill Purr Hiss Yowl

Independent experts rated the visual and auditory stimuli for valence. Their activity ratings were less consistent, which matters because activity was standing in for arousal. The paper therefore tests a carefully arranged stimulus set, not a universal dictionary in which one cat posture always carries one meaning.

Conceptual illustration comparing a cat observing posture photographs with a cat hearing a speaker; not a study photograph.
The study used separate visual and auditory trials. This illustration explains the two routes without reproducing study data.

Who took part, and how was it measured?

The sample included 24 healthy mixed-breed cats, 12 male and 12 female, from 13 households. Their mean age was 6.21 years, with a standard deviation of 3.73 years. All were at least one year old and neutered. Four cats lived in single-cat homes; the other homes had more than one cat. The researchers used snowball sampling, so this group cannot represent every domestic cat.

Testing took place in each cat's own home and a familiar room. Each cat completed eight two-minute trials: four visual and four auditory. The usual schedule placed two visual and two auditory trials in a session, with testing spread across at least two days. A cat started about 150 centimeters from the stimulus and could leave the area. The caregiver and experimenter did not interact with the cat during a trial.

Video coders recorded gaze duration and frequency, time spent interacting with the stimulus, proximity within 150 and 50 centimeters, approach latency, time out of sight, visits and other behaviors. The researchers coded the videos in BORIS with a pre-established ethogram and also used CatFACS action units when cats gazed at a stimulus. These are observable behavior measures. They are not direct emotion meters.

What changed in the visual trials?

The clearest result was an interaction between valence and activity. Cats gazed longer at the negative, high-activity visual stimulus than at the positive, high-activity stimulus. The reported model estimated a contrast of beta 1.72, with t = 2.22, p = .031 and a 95% confidence interval from 1.07 to 2.76. In the pairwise means, the back-arch image drew an average of 22.61 seconds of gaze, compared with 11.53 seconds for the approach image.

Gaze frequency followed the same broad pattern: the active negative image drew more looks. Stimulus interaction told a related but different story. Active images drew longer interaction regardless of valence. The reported means were 9.34 seconds for back arch, 11.26 seconds for approach and 2.86 seconds for roll.

Observed measure Reported pattern Reading boundary
Gaze duration Negative, high-activity visual cue drew longer gaze than the positive, high-activity cue Longer looking shows attention in this task; it does not identify a single feeling.
Gaze frequency Negative, high-activity visual cue drew more looks Repeated looks are still a task response, not a household label.
Stimulus interaction Active images drew longer interaction regardless of valence Activity and valence contribute different information.
Proximity and approach No significant valence or activity effects were reported Attention did not translate into a measured approach preference.

What happened with the sounds?

Valence and the valence by activity combination did not significantly affect gaze duration or gaze frequency in the auditory condition. The same was true for stimulus interaction, proximity, approach latency and time out of sight. In this experiment, isolated playback sounds did not produce the visual response pattern.

The authors interpret that difference cautiously. Cats may give only brief attention to an isolated sound, and a natural interaction usually contains movement, visual information and scent as well as sound. The auditory calls also had natural durations rather than one standardized length. Those features make the result useful for understanding this playback task, while limiting a direct comparison with an ordinary encounter between cats.

Conceptual illustration of a cat observation with gaze, interaction and distance as separate measures; not a study photograph.
Behavior studies can record gaze, interaction and distance as separate outcomes. The illustration contains no study statistics.

Evidence-led insights

1. The sensory route changes the question

Reported: the visual condition showed valence and activity effects, while the auditory condition did not. Inferred: a result about a picture and a result about a sound cannot be treated as interchangeable evidence about a cat's emotional understanding. The stimulus route is part of the question being measured.

2. Activity is part of the interpretation

Reported: active negative visual cues drew more gaze than active positive cues, and active images drew longer interaction overall. Expert ratings of activity were less consistent than ratings of valence. Inferred: a high-activity posture should not be treated as a direct meter of emotional intensity. The paper supports keeping activity and valence visible as separate parts of the design.

3. Attention and approach are different outcomes

Reported: gaze and interaction changed in parts of the visual task, while proximity, approach latency and time out of sight did not show the same effects. Inferred: looking longer does not by itself prove fear, preference, attraction or avoidance. When behavior measures disagree, the careful description keeps each measure separate.

4. Measurement agreement sets a boundary

The authors reported low inter-rater reliability for most CatFACS action units. Pupil dilation, nose licking and eye blinking were among the more reliable coded actions, while smaller facial movements were harder to score at the available distance. Inferred: broad, visible behavior measures may support a clearer summary than a fine-grained facial label when observers do not agree consistently.

What this study does not show

  • It does not create a home emotion detector or a way to label one cat from a short video.
  • It does not prove that cats experience the human words used to describe valence in exactly the same way people do.
  • It does not show that one cat's response predicts personality, wellbeing or future behavior.
  • It does not rank sounds, photos, toys, food or products.
  • It does not establish a health, treatment or safety decision.
  • It does not make the external study data VEP-owned data.

A careful takeaway for pet families

The practical lesson is to keep a behavior description close to what was observed. A cat may look at a sound, move toward a picture or leave a space for many reasons. A single response is easier to overread when gaze, interaction, distance and context are collapsed into one emotional label. This paper shows why researchers separate those measures and report the exact task that produced each result.

Source note and limits

This is a single-study Study Brief, not a systematic review and not a VEP-owned dataset. Spoelstra and colleagues published the paper in PeerJ on May 25, 2026, as volume 14, article e21292. Read the PubMed record, the DOI record and the full text in PubMed Central. The study used a small snowball sample of healthy mixed-breed cats, familiar-home testing, photographs and isolated sounds. It did not include natural scent context, and the expert activity ratings and several fine-grained facial codes had limitations. The underlying data are listed in DataverseNL.