Does a cat prefer red or green coloured food? This experiment tested whether a British Shorthair cat (Mi) shows a statistically significant preference for red- or green-dyed cat food. Regular dry cat food was dyed with food colouring and presented in two side-by-side bowls over 30 days. The bowl Mi approached first was recorded as his preference for that day. A chi-squared test was used to determine whether the observed preference differed significantly from chance.
| Category | Details |
|---|---|
| Subject | British Shorthair cat (Mi) |
| Food | Regular dry cat food (Kirkland) |
| Dye | Red and green food colouring |
| Serving | 10 pieces per bowl per trial |
| Duration | 30 days (August-September 2024) |
Coloured cat food was prepared by mixing regular dry food with red and green food colouring in separate jars. Each day, two bowls were placed side by side — one with 10 red pieces and one with 10 green pieces. Bowl positions were alternated daily to control for side bias. The colour Mi approached and began eating first was recorded as his preference. Remaining pieces, serving time, and food dye concentration were also tracked. Days where Mi did not eat from either bowl were excluded. After 30 days, a chi-squared goodness-of-fit test was applied to the observed preferences.
Raw data was recorded on handwritten data sheets, photographed as JPEG images. Variables tracked per trial include: date, serving number, pieces served and remaining, bowl position (left/right), serving time, and food dye drops used.
Over 30 days, Mi chose red on 13 days and green on 17 days. A chi-squared goodness-of-fit test with one degree of freedom yielded a test statistic of 0.533, well below the critical value of 3.841 at 95% confidence. The null hypothesis (no colour preference) was not rejected. Mi shows no statistically significant preference for red or green cat food.
Note — correction to written report: Claude identified an arithmetic error in the original PDF report. The report calculates χ² = 0.266 by dividing each (O − E)² term by 30 (the total number of observations). The correct chi-squared formula divides each term by the expected count for that category (15), not the total N. The corrected value is χ² = (13 − 15)² / 15 + (17 − 15)² / 15 = 0.267 + 0.267 = 0.533, as computed in the reproducible notebook. The conclusion is unchanged — both values fall well below the critical value of 3.841 — but 0.533 is the correct test statistic.
See the written report, the static notebook or run the reproducible analysis yourself.