The behaviour of the sand martin fleaCeratophyllus styx jordaniSmit was studied in relation to its ecology.The cocooned resting imago in the host's old nest is the main over-wintering stage. Mechanical disturbance of the cocoons by the exploratory habits of the newly returned sand martins elicits emergence of the imagines in spring. The seasonal rise in temperature is not, by itself, important in causing emergence, as it does not become effective until many weeks after the martins have returned, by which time undisturbed fleas are likely to have died inside their cocoons.When the imagines break out of their cocoons they are negatively photo-tactic, but become positively phototactic within 24 h. This response takes them outwards along the martin's disused burrow until the increasing intensity of light reduces their activity so that they aggregate on the lower lip of the entrance. Positive phototaxis prevents them dispersing downwards from the entrance.Periodically the fleas bury themselves in sand; this response may function for water conservation. The proportion buried is greatest at night.The sand martin's habit of hovering close to a succession of entrances renders it accessible to the fleas, whose main host-finding response is an outward jump from the burrow entrance. This jump is released by a sudden decrease in light intensity and is directed towards dark objects. Vibration and air currents do not release jumping.Dispersal from aggregations in disused burrows may occur by transport on the host, or by spontaneous horizontal emigration along the cliff face, or by falling after an unsuccessful host-finding jump. Fleas which have fallen become negatively geotactic and positively anemotactic. Fleas wandering on the cliff face visually detect burrow entrances up to 30 cm away, and turn towards them. Preference for moister sand and, possibly, a negative phototactic response, may induce the fleas to remain in newly found burrows.Small aggregations of fleas also occur at the entrances of burrows currently in use by martins. Fleas circulate between the entrance and nest chamber of these burrows, until the martin begins to incubate its eggs. The entrance aggregation then disappears and fleas accumulate in the nest chamber. There is some interchange of fleas between infested burrows in the martin's pre-incubation period.The general pattern of behaviour resembles that ofC. gallinae. Behavioural differences between the two species are related to the ecology of their hosts.C. styxis adapted to dispersal and host finding in its host's breeding site, whereasC. gallinaeis adapted to reach foraging birds. This difference partly accounts for the narrow host specificity ofC. styxand the wide host range ofC. gallinae.My grateful thanks are due to Chris. Mead and Giles Pepler for information on sand martins' behaviour and migratory arrival. I would also like to thank Brian Little, who introduced me to several colonies in the Tyne Valley. The valuable advice of the Hon. Miriam Rothschild and Dr E. T. Burtt is especially acknowledged.