“…The data given in Table 1 suggest that, except during the short-term memory (STM) stage of our model (up to 10 mm postlearmng, Gibbs & Ng, 1977), retention levels of chicks trained on a passive-avoidance task very rarely exceed 80-85% Furthermore, variability tends to be somewhat small among retention levels for TTIs beyond our STM stage. These findings are consistent with ours, although the variability observed in the studies reported in Table 1 generally yield chi-squares with p > 0.1 With respect to the latter, however, it may be noted that these studies either have large sample sizes (Cherkin, 1969;Lee-Teng, 1968;Lee-Teng et al, 1970) or a small number of TTIs (Bailey et al, 1969;Bell & Morgan, 1981, Stephenson & Andrew, 1981 or both A possible implication of the rather ubiquitous empirical retention ceiling of 80-85% is of interest to the issues raised by Roberts Taking a sample of 20, for convenience, the results from various studies suggest that 3-4 chicks in any group of 20, which apparently satisfy the criteria of training used, do not retain memory for the learned association beyond 10 mm postlearmng. If this is a relatively fixed natural retention ceiling, for whatever reason, then it may be argued that the effective sample size for any TTI longer than 10 mm postlearmng is not 20 but say 17.…”