Small, adult (19.50 ± 2.01g wet weight) Lytechinus variegatus (Lamarck, 1816) were fed 8 formulated diets with different protein (1236% dry weight as fed) and carbohydrate (2139% dry weight) levels. Each sea urchin (n = 8 per treatment) was fed a daily ration of 1.5% of the average body weight of all individuals for 9 wk. Akaike information criterion (AIC) scores were used to compare 6 different dietary composition hypotheses for 8 growth measurements. For each physical growth response, different mathematical models representing a priori hypotheses were compared using the AIC score. The AIC is one of many informationtheoretical approaches that allows for direct comparison of nonnested models with varying numbers of parameters. Dietary protein level and protein:energy ratio were the best models for prediction of test diameter increase. Dietary protein level was the best model of test with spines wet weight gain and test with spines dry matter production. When the Aristotle's lantern was corrected for size of the test, there was an inverse relationship with dietary protein level. Log-transformed lantern to test with spines index was also best associated with the dietary protein model. Dietary carbohydrate level was a poor predictor for growth parameters. However, the protein × carbohydrate interaction model was the best model of organic content (percent dry weight) of the test without spines. These data suggest that there is a differential allocation of resources when dietary protein is limiting and the test, but not the Aristotle's lantern, is affected by availability of dietary nutrients.