If birds reared in isolation learn visual features of their own bodies, they could generalize from such learned perceptions and choose to associate with members of their own species rather than with other species when subsequently placed with other animals. This self-perception hypothesis could account for some published studies showing preferential choice of own species by birds reared in isolation. A series of experiments was conducted to test this hypothesis. Typically, newly-hatched chicks were dyed red, or green, or were left their natural yellow colour. After being reared for 8 days in visual isolation or in small social groups, the chicks were tested in a three-choice test apparatus for their approach to red, green, and yellow chicks. A chick was scored for a choice if it reached a goal-box and tried to push through the separating wire-mesh. In Experiment I, red and green chicks, reared socially or in isolation, were tested for their choice of red, green, and empty goals. The socially-reared chicks clearly chose goal chicks of their own colour. The isolate-reared chicks responded in an indiscriminate manner. In Experiment 2, chicks of the natural yellow colour were used along with red and green ones. Again, social-chicks chose goals of their own body colour, while isolate chicks responded indiscriminately. Some generalization appeared to occur in the social-chick responses, particularly between green and yellow, with green and yellow chicks clearly avoiding the red goal. To a lesser extent, red was generalized to yellow, so that red chicks avoided the green rather than the yellow goals. In Experiment 3, newly-hatched yellow chicks responded to red, green, and yellow goals indiscriminately, showing that they have no innate companion-colour preference. However, 8-day isolate-reared yellow chicks behaved like 8-day social-chicks and avoided red goals in favour of green and yellow ones. Experiment 4 tested an hypothesis that isolate-chicks learn their own colour from their reflection in their drinking water. Green and yellow chicks, reared in isolation for 8 days with no drinking water (which was administered by pipette directly into the crop, either in light or in darkness), failed to show any tendency to choose green and yellow goals preferentially to red and empty goals. In Experiment 5, chicks coloured red, yellow or green were reared in mixed groups; one chick of one colour with two companions of another colour. In this way chicks experienced companions either of one, or of two, colours. The tests at 8-days showed that when two colours had been experienced either colour may be chosen, but that green was more often chosen than might have been expected. It appeared that green was most, and red least, readily learned and responded to. The 'natural' yellow colour tended to be confused with green. Experiment 6 tested for the possibility of individual recognition in 8-day social-chicks by using cage-mates in one of the goals. Responding was indiscriminate in yellow chicks tested in this way with yellow goal chicks. Taken together, these experiments show that newly-hatched chicks have no preference for red, yellow, or green chicks, but that they learn the colour of their companions and selectively approach them by 8-days of age. These results are compared with the published work on colour and approach and following responses in chicks. In general there seems to be more evidence against colour preferences in such responses in newly-hatched chicks. The apparent stronger preference shown by chicks that have experienced colours at the extremes of the visible spectrum could be due to the greater possibilities for generalization in birds trained to colours in the middle of the spectrum. In any event, there seems no doubt that the colour of the companion is learned. The self-perception hypothesis was not supported by the results of Experiments I and 2. However, if the colour groups are combined in Experiment 2, the isolate-chicks appear to have chosen their own body-colour significantly more than would be expected by chance. Furthermore, in Experiment 3 yellow isolate-chicks clearly avoided red goals. It appears, then, that isolates can develop a weak tendency to choose colours not too different from their own body-colour. Experiment 4 showed that such a tendency did not develop if the chicks were not able to see their own reflection. It must be concluded, therefore, that the self-perception hypothesis is tenable at least if perception through reflection in water is included. It is not known whether, in the absence of reflected self-images, such a tendency would develop over a longer period of time through the bird seeing its own body directly. This conclusion is discussed in relation with published studies of species and familial recognition in birds reared in isolation or in the company of other breeds or species.