Abstract
This experimental study compared the precision and accuracy of the biological intercept ( BI ), modified fry ( MF ) and time‐varying growth ( TVG ) backcalculation models in estimating the early growth of the tropical freshwater purple‐spotted gudgeon Mogurnda adspersa . Larvae were reared up to 41 days post hatching under two temperatures and four different feeding regimes. Food and temperature treatments induced complex growth profiles among fish, and although total length ( L T ) and otolith radius were related under all conditions, some uncoupling was evident in the otolith–somatic‐growth ( OSG ) relationship of fish subjected to periods of changing food availability. Furthermore, otolith growth was found to be significantly influenced by temperature, but not by food availability. Analysis of backcalculation residuals by linear mixed effects modelling revealed that BI and TVG were equally precise in predicting somatic growth, with the highest accuracy provided by TVG . The performance of all the three models declined as the OSG relationship weakened under low‐food conditions, with maximum errors estimated to be 39, 60 and 36% of observed L T for the BI , MF and TVG models, respectively. The need for careful validation of backcalculation models is emphasized when examining fishes subjected to variable environmental conditions, and when exploring the differential influence of temperature and food on fish L T and otolith growth.