Behavioural responses to variables associated with habitat and level on the shore are often given as explanatory models for vertical patterns of distribution or vertical gradients of size of intertidal animals. These models are frequently examined using experimental transplants in which animals are moved to levels (or habitats) where they are not usually found and their movements subsequently monitored. There are many problems with the design, execution and interpretation of most of these studies because of lack of appropriate experimental design and inadequate analyses to identify interactions among variables. This study examined the role of behaviour in determining patterns of distribution of two small periwinkles, Littorina unifasciata Gray and Nodilittorina pyramidalis (Quoy and Gaimard) on rocky shores in New South Wales, Australia, using experimental transplantations. Also examined were the experimental and analytical procedures necessary for such experiments and spatial and temporal variability in results obtained. N. pyramidalis and L. unifasciata responded to being transplanted downshore or, for L. unifasciata only, upshore by dispersing further and moving directionally towards their level of origin. These responses were persistent and relatively widespread across many sites. There was considerable variability in the results obtained from place to place and time to time, although transplanted snails were less variable in behaviour than were snails in other treatments or controls. Translocation to a different site at the same level on the shore caused snails to disperse longer mean distances, although not as far as moved by transplanted snails, but did not have a consistent effect on directionality of movement. Experiments such as these must include controls for disturbance and translocation if responses of transplanted animals are to be appropriately interpreted. They must also include proper design and replication so that variability in responses from site to site, day to day and among similar experiments can be quantified. Only then can processes influencing such behaviour be investigated and the role of behaviour in determining patterns be understood.