Newly hatched autumn-spawned herring larvae Clupeu lzurengus were released in two 2500-m' outdoor mesocosms and reared over a 2-month period. Hydrographic conditions were similar in the two mesocosms, but the average plankton density was initially more than 10 times higher in mesocosm B compared to mesocosm A (>1 1Half-way through the experiment the feeding conditions reversed with three times higher average densities in mesocosm A than in mesocosm B ( > 3 1 ~ ' v. -1 1 ~ '). Herring larvae were sampled with a 0.3-m2 two-chambered net twice weekly, and survivors were harvested by draining the mesocosms at the end of the experiment. Otolith growth trajectories of individual larvae were determined by relating radial otolith size with number of increments from the outer edge of the otolith (days before capture). The increment widths during the first 3 weeks after hatching, including the first-check size, were generally wider among larvae from mesocosm B (relatively good initial feeding conditions) than among those from mesocosm A (poor initial feeding conditions). The otolith growth pattern also confirmed that the surviving herring in mesocosm A belonged to the upper size range of larvae in the mesocosm after only 2-3 weeks from hatching; no such trend was found in mesocosm B. In both mesocosrns the otolith size-at-age indicated that with the present sampling gear, herring larvae larger than 20--25 mm were underrepresented in the net samples. The information obtained from otolith-size-at-age is compared with other morphometric and biochemical measures of size and condition of larvae v. 10.1 1 ~ I ) .obtained throughout the experiment.(c) 1997 The Fisheries Society of the British lsles