This paper reports on experimental research undertaken to analyse the information processing mechanism by which the fry of substrate spawning cichlid fish visually recognise their guarding parent(s), already from the earliest time they are able to swim. The study is inspired by LORENZ' concept of 'innate releasing mechanism' (IRM) and aims at penetrating into the problem of how species-characteristic predispostions to repond selectively to particular environmental stimuli could be brought about. Stimulus selection is studied by means of dummies. A discussion on the methodology of dummy experiments, especially on the pros and cons of successive and simultaneous presentation of the different dummies, precedes the presentation of the results (Chapter II). In Chapter III of this paper, the effectiveness of relative brightness, melanistic patterning and overall colour of the parent in eliciting the approach response was comparatively examined in fry of 15 different cichlid species, by means of choice tests between two dummies. Although selection for the same stimuli was often found to be shared by different species, the set of different stimuli inducing approach varied in a for each species characteristic way. This set only roughly matches the colouration shown by the parents in the period they are guarding the fry, a roughness which cannot be attributed to limitations of the visual capacities of the young fish. It is argued that this suggests an information processing procedure in which species-characteristic behaviour elements (fixed patterns and clusters of them) are activated when signals from a particular set of detectors have reached a 'selector' unit specifically controlling the response. Such a model (see Figs. 1 and 2) was earlier developed by BAERENDS (1982, 1985) for the recognition of eggs as an object to incubate in gulls. This information processing construction would provide a naive animal with an 'identification key' for recognizing environmental situations that are vital to it. The capacity of an animal to build the species-specific connections between each selector and a set of detectors is thought to have developed through selection during the evolution of the species. Cichlasoma nigrofasciatum provided an opportunity to investigate this because, beside the grey and darkly barred common morph of this species, also an amelanistic white morph exists. Chapter IV of this paper reports on experiments testing the relative effectiveness for eliciting approach of each of 10 dummies in which features of the dark morph (grey groundcolour, black bars) had stepwise been modified towards the overall white of the mutant. By means of successive presentation, rank orders of decreasing effectiveness of these dummies were established for fry of the dark morph, the white morph, of mixed pairs and of F1-hybrids between both morphs, either kept with the parents or raised with peers only. These rank orders did not differ significantly. An entirely black dummy and a dummy with black vertical bars on a white background (a contrast exceeding that on the fish) came out at the top. The conclusion was drawn that a predisposition for recognizing the colour pattern of the parent is present which roughly matches the appearance of the dark morph and that this same predisposition is present in the white mutation. However, a comparison of the variation in the rank orders found for dark morph and for white morph fry, both raised with parents, indicated that during the parental period acquisition of knowledge about the actual appearance of the parents takes place in the young. In the dark morph this learnt knowledge reinforces the predisposed information and causes the statistical significance of the rank order to increase. In the white morph both sources of information act largely contradictory. However, the effect of learning, which was also demonstrated in choice tests, did not entirely override the predisposed preferences. The suggestion that the predisposed information about features of the adequate situation for eliciting an action pattern under natural conditions guides a relevant extension of knowledge about that situation through learning, is further investigated in chapter V in experiments on Cichlasoma meeki. The conclusion drawn by BAERENDS & BAERENDS-VAN ROON (1950), from results of successive dummy presentation, that the fry of this species have to learn that the parental dress shows a pattern of red and grey, had to be rejected after the more sensitive choice experiments carried out in the present study had demonstrated a predisposed preference for this combination in naive fry. Experienced fry, however, were found to have strengthened this preference and clearly selected for dummies with the red colour on the ventral side. Finally in Chapter VI some questions concerning the nature of predispositions are considered and the literature relevant to the functions and consequences of learning about features of the parents for partner selection in later life is discussed.