Understanding why different marine food-webs have different structures will give insights into how changing environmental conditions and other stressors might affect ecosystems. But how can we tell if one food-web is different from another? And what do we mean by food-webs being “different” anyway? Researchers often have idiosyncratic ways of dealing with missing or uncertain information when developing marine ecosystem models, and this leads to models having “model personality”: models reflect the implicit assumptions and semi-arbitrary judgements made during their development as well as genuine differences in food-web structure. In this presentation I present a method previously used to remove model personality and suggest why it is of limited wider utility. Two new preliminary approaches are presented that together move towards a more general test of statistical significance in differences between food-web models. The first approach uses stable isotopes to help guide how initial parameter estimates are adjusted to obtain a balanced models. The second perturbs the balancing process to develop a large number of balanced food-web models. Combining the two approaches allows us to compare within-model and between-model variation as a statistical measure of differences in food-web structure.