The behavioural mechanisms by which Delia brassicae (Hoffm.) locates its host plants are reviewed. In a large wind tunnel, the principal mechanism appears to be upwind orientation and flight stimulated by host plant odour. Gravid females landed frequently while in an odour plume, the mean flight length being 0.5 m for cultured flies under a variety of conditions. On leaving the odour plume while overshooting the source or moving across the air flow, the majority of females turned downwind, circled and re-entered the plume.These conclusions are seen as an hypothesis of host plant finding and adequate means of testing its validity in field situations are required. The approach developed here was to identify features which were likely to be peculiar to this type of host plant finding. A computer model was built to simulate the events downwind of an odour source and this predicted that the "average" cultured female moves upwind at the relatively slow rate of 65m/day. In two mark and recapture experiments in the field, cultured females moved upwind at estimated mean rates of 80 and 95m/day when released downwind of a brassica plot.Analysis of the behavioural mechanisms underlying the location of host plants by insects has been carried out largely under laboratory conditions. The progression of research on the cabbage root fly, Delia brassicae (Hoffm.) appears to be typical, having passed through similar stages to that on other species. Some field experiments have been carried out (de Wilde, 1947;Traynier, 1965;Hawkes, 1974Hawkes, , 1975 but when detailed observations of behaviour or an analysis of the causation of behaviour are required, the difficulties of field experimentation have forced a retreat to laboratory situations (Traynier, 1967 a, b;Coaker & Smith, 1970; Nair & McEwen, 1976;Hawkes & Coaker 1976, 1979 Hawkes, Patton & Coaker, unpubl.). The present knowledge on host plant finding in D. brassicae has then been outlined by field experiments but elaborated in the laboratory. This paper reviews the existing knowledge and suggests that the conclusions of laboratory experiments should be seen as an hypothesis and not as an endpoint. Such hypotheses need to be tested in the field to determine whether they provide a valid basis for the understanding of the ecological process of host plant finding. This paper illustrates an approach to the problem of validation of hypotheses. *) Present address: --Fasanv/igen 4, 752 52 Uppsala, Sweden.