Differences in behaviour between ultimate and penultimate instar larvae of I. posita were examined using caudal lamella loss as an assay. Equal densities of each instar were separately assembled in 20 L aquaria using 39 vertical dowels as an artificial weed bed. After 24 h there were 5 deaths in the ultimate instar assemblages and 1 death in the penultimate assemblages. One hundred sixty one ultimate instar larvae (72%) suffered some loss of lamellae while only 54 penultimate larvae (24%) were so affected (PcO.OOl). Comparison of number of lamellae lost yielded similar results: 254 lost among ultimates, 65 among penultimates (PcO.OOl). The hypothesis that lamella loss follows a binomial distribution could not be rejected. This pattern suggests that lamellae are lost singly and that no individual has a higher probability of loss than another. Video recorded trials of paired, same instar larvae suggest that differences in lamellae loss result from differential use of agonistic behaviour patterns rather than differences in encounter rates due to locomotion patterns.