Establishment of trial duration
To determine an appropriate duration for behavioural trials, in which dyads would consistently form stable dominant/subordinate relationships, a preliminary test, using the same protocol as the observer trials, had established five hours to be a reasonable period in which to assess the development of dominant/subordinate relationships: of 27 preliminary size-matched dyads, 26 formed a clear dominant/subordinate relationship after 5 h (χ2 = 23.15, df = 1,p  < 0.001, showing a goodness-of-fit departure from a 50:50 ratio of clear to unclear hierarchies), with a mean time ± s.d. to detection by the observing researcher of 3 h 21 min ± 54 min. The dyad was left in the test tank for a further 24 h after the end of the trial. The stability of the dyad relationship was then re-tested by 5 min observation of behaviour followed by confirmatory bloodworm feeding. The re-test showed that most (24 of 26) dominant/subordinate relationships remained stable (χ2 = 18.62, df = 1,p  < 0.001, showing a goodness-of-fit departure from a 50:50 ratio of hierarchies that remained the same after 24 h to those that became unclear or changed). Specimens used in the preliminary trials were not used again in the main behavioural trials.