Development of caravaning behaviour in the house musk shrew (Suncus murinus) was described in detail and discussed in relation to sensory and motor development. With age of pups, caravaning changed in its formation pattern (I-V) as well as its quantitative aspects such as frequency of occurrence, duration and distance. The shift in formation pattern from II to III was simultaneous with the onset of vision (Experiment I). In darkness, frequency of caravaning significantly decreased at ages later than 11 days. This was due to the selective disappearance of Patterns III and V (Experiment II). These findings suggest the role of vision in caravan formation at the later stage of the period of occurrence. After the onset of vision, caravaning is released mostly by visual motion of the object, whereas caravaning at the earlier stage is released nonvisually. Caravaning and the pup's locomotor activity were suppressed and delayed until the age of weaning in the enclosure set up to simulate the natural habitat. The period of occurrence was reduced by disappearance of some formation patterns observable in the laboratory open-field and the onset of caravaning exactly synchronized with that of spontaneous and independent locomotion. Development of caravaning in the outdoor enclosure closely resembled the development of Pattern V in the open-field (Experiment III). The adaptive significance of caravaning can be fully understood only by comparing behaviours which occur in different situations. Such an approach could bridge the gap between observations in the laboratory and those in the natural habitat. In Experiment I, caravaning was compared between domesticated laboratory animals and animals at the first two generations of nonsib-mating from wild ones. The age-dependent tendency was more distinct in the latter group. This suggests that mother-offspring relations changed in the process of domestication.