While the post hoc test clearly showed significant differences among the experimental groups, the statistical analysis indicated replicate interactions for SGR and condition factor (Table 3). The nested ANOVA showed a significant relationship between carbon dioxide concentration and weight on day 23 (pb 0.05), although no significant differences between mean values were found. On day 38, the mean weight of the high and very high group was significantly reduced when compared to the control group (p b 0.0001). Both weight and condition factor decreased linearly with increasing carbon dioxide concentration on day 38 (pb 0.01). No significant effects from CO2 exposure were found on mean length on any days (Table 3). Mean SGR was significantly reduced for the three highest carbon dioxide groups during period I compared to the control group (Table 3; p b 0.05). During period II, only the SGRs of the two highest carbon dioxide groups were significantly reduced. For the medium group, the mean SGR for the whole period (Period I and period II) was significantly reduced compared to the control group (p b 0.05), and the lowest mean SGR was found in the very high carbon dioxide group. It should be noted that the SGRs for the low, the medium and the high carbon dioxide group were about the same during period I and period II. In the very high carbon dioxide group, however, the mean SGR was much lower during period II (as a result of increased carbon dioxide level). For period II, a second-order polynomial model described the relationship between carbon dioxide concentration and individual specific growth rate (Fig. 3a; p b 0.01). Some individual fish grew well even at high CO2 concentration (38 mg L−1). Nephrocalcinosis was not observed in any of the groups and significant histomorphological lesions were not found in the gills. Mortality was not observed during the experiment.