Clingfishes are widely distributed on rocky coasts (Briggs, 1986, 1990), but knowledge of their social behaviour is extremely scarce. This is probably because most species are very small and cryptic and occupy microhabitats, such as among algae (Diplecogaster spp., Eckloiaichthys spp.: Breder & Rosen, 1966), under boulders (Lepadogaster spp., Chorisochismus spp.: Breder & Rosen, 1966), inside cavities (Lepadogaster spp.: personal observations), and even among the spines of sea urchins (Diplecogaster spp.: Patzner et al., 1992) and associated with crinoids (Lepadichthys spp.: Fishelson, 1966). Available data refer mainly to the spawning habitat, descriptions of egg masses, parental care and sexual dimorphism (Breder & Rosen, 1966; Gibson, 1969, 1982; Thresher, 1984). As far as we know, no descriptions of agonistic behaviour are available for the family. Lepadogaster lepadogaster purpurea (Bonnaterre) is a common clingfish in the inter- tidal zone of the Eastern Atlantic and the Mediterranean (Briggs, 1986, 1990). In this paper, the agonistic behaviour of L. l. purpurea in captivity is described. Fish were collected in December 1992 (group 1), March 1993 (group 2), August 1994 (group 3) and October 1994 (group 4) under boulders during low tide with a hand net at Arrábida, on the Atlantic coast of Portugal (38)28*N, 8)59*W), and kept in 63#32#31 cm and 80#40#38 cm aquaria with gravel and boulders to provide shelter. Natural photoperiod was maintained by a fluorescent light (15 W), and temperature varied from 14-17·5)C (group 1), 18-24)C (group 2) and 21-24)C (groups 3 and 4). They were fed with shrimp (Palaemon spp.), common cockles (Cerastoderma spp.), and other live marine invertebrates. A total of 24 fishes (total length: males: mean 6·95 cm, s.d. 0·48, range 6·3-7·8 cm, n=12; females: mean 5·45 cm, s.d. 0·69, range 4·5-6·8 cm, n=12) was observed for 190 h, including 110 h of focal and scanning sampling (sensu Martin & Bateson, 1993). Each group was composed of three males and three females. Males were larger and more colourful than females with reddish unpaired fins and more conspicuous blue eye spots. Details of behaviour patterns were based on frame-by-frame and slow motion analysis of videotape recordings. Statistical analysis was performed using the PC computer program Statgraphics (version 4.0) and the simulation stat- istical program ACTUS (Estabrook & Estabrook, 1989) designed for the analysis of contingency tables. This program uses random numbers to simulate 1000 tables, each with the same row and column total of the original table. Each simulated table is compared with the original data table and values smaller than 50 are considered to be