Autumn-spawned North Sea herring eggs were fertilized artificially, incubated in the laboratory, and released into two mesocosms (A and B) in September 1991 as 1-d-old larvae (hatched within 24 h) . The duration of the experiments was 60 d . The released larvae experienced a high initial temperature (.-18°C) in both mesocosms ; the temperature decreased to approximately 10°C at the end of the experiments. The temperature in mesocosm A was on average 0 .7°C higher compared to mesocosm B . The prey density was higher initially in mesocosm B (> 1 1 -1) compared to mesocosm A (<0.1 1-1) . The situation reversed around day 23, resulting in higher prey densities in mesocosm A (> 3 1 -t) compared to mesocosm B (1 1 -1). Differences in the somatic and otolith growth rates of the herring larvae were observed, reflecting the pattern of prey densities in the two mesocosms. However, changes in the otolith growth rate showed a delayed response to changes in the body growth rate . The results show that starving and slow-growing herring larvae have a relatively higher otolith growth rate than somatic growth rate. Both the somatic and otolith growth rates increased with increasing prey density ; however, they were affected inversely by temperature . The thermally stable marine environment indicates that decoupling of otolith and body growth rates caused by temperature alone will be of only limited importance . When back-calculating to a previous body size, decoupling might not be a serious problem above a minimum body growth rate (>0 .19 mm d-1) in herring larvae, but the problem may arise when starved and slow-growing herring larvae are included .