1. A subject who has suffered from central serous chorio‐retinopathy in his left eye noticed differences in the colour of a given light as perceived by each eye alone. Standard screening tests (colour order and colour matching) indicated a tritan defect in the left eye; the right eye was normal on these tests.2. The subject was dichromatic in his left eye, trichromatic in his right. The left‐eye distimulus colour‐matching functions, spectral luminosity, and wave‐length discrimination functions were indistinguishable from corresponding data for congenital tritanopia. Comparable right‐eye data were normal.3. Spectral dichromatic colour matches were invariant under changes of intensity and under addition of a common light to both halves of the field. (Grassmann's laws of linearity are satisfied.)4. Increment threshold versus intensity (t.v.i.) curves for a blue (481·9 nm) test on a yellow background yielded the normal three branches (for Π4(μ), Π1(μ) and Π3(μ) respectively) in the trichromatic eye. In the dichromatic eye a single mechanism was found. It had the field sensitivity of Π4(μ) whether measured with the blue, or with a violet (429·5 nm) test. No trace of Π3(μ) or Π1(μ) was ever discovered in the tritanopic eye. Both are normal in the trichromatic eye.5. The field sensitivities of Π4, Π5 and Π3 of the normal eye are well fitted by linear combinations of the spectral colour‐matching functions of the trichromatic eye. Π4 and Π5 of the dichromatic eye are well fitted by linear combinations of the tritanopic matching functions.6. Colour matches made by the trichromatic eye do not match when viewed by the tritanopic eye, almost certainly because the ocular media of the two eyes have wave‐length‐dependent differences in absorption. For the largest difference (430 nm) the trichromatic eye transmits about 2·2 times more light than its fellow. When allowance is made for these differences, the field sensitivities of Π4 and Π5 of the two eyes do not differ. The field sensitivities of Π4 and Π5 of the normal eye, on the other hand, differ significantly from those of the average spectra obtained on four normal trichromats by Stiles, in a way that cannot be attributed to differences in transmittance of ocular media.7. It is concluded that classical (or acquired) tritanopia is not distinguishable in its manifestations from congenital tritanopia; furthermore, tritanopia can be regarded as a reduced form of normal trichromacy, once allowances are made for absorption of the ocular media and for variations among normal trichromats.8. Despite extensive search no evidence could be uncovered which might exclude the hypothesis that the colour vision in tritanopia depends exclusively upon absorption in only two foveal cone pigments, one long‐wave‐absorbing and one medium‐wave‐absorbing.