1. The spectral sensitivity of 7 broiler fowl was determined in a behavioural test. 2. Initially the birds were trained to receive a food reward by pecking at a clear, Perspex panel behind which was a light stimulus (circular, diameter = 9 mm), originating from a tungsten-halogen lamp. Subsequently, they were trained to choose between 2 panels only 1 of which was lit; the assignation of light and dark on each panel was randomly ascribed between trials. The colour of the lit panel was determined by the wavelength of the light transmitted through one of 13 closely defined narrow bandwidth filters between 326<lambda<694 nm. The flux of photons and hence the intensity of the stimulus received by the bird could be controlled by changing the voltage across the lamp. The photon flux was gradually reduced until the birds failed to detect accurately the lit panel. Success was defined as a choice of the lit panel 9 or more times in a sequence of 10 trials, providing that the sequence contained at least 4 changes in the position of the stimulus. 3. Generally, the birds showed a peak sensitivity between 540<lambda<577 nm. The results agree with electrophysiological data between 507<lambda<694 nm and psychophysical data between 500<lambda<700 nm but our data showed higher sensitivities between 380<lambda<507 nm compared with electro-physiological findings. 4. Our findings confirm that broilers can 'see' into the UV(A) range and that their spectral sensitivity is different to the human. The implication of this is that the measurement of light intensity in poultry housing using the lux unit does not accurately describe the intensity perceived by fowl. 5. Experimenters using colour, for example differently coloured lighting or food, need to account for this different sensitivity when interpreting their results.