In Berk and Green (2004), diseconomies of scale in asset management combined with performance-driven flows drive expected excess returns to zero at the optimal fund size. However this model incorrectly accounts for variations away from the optimal size due to investment returns. In this paper I explain the errors in Berk and Green (2004) and present an alternative model in which managers profess to have skill at identifying expected returns based on public information. Updating about whether managers possess this skill leads to contracts and performance/flow relationships which conform with those we observe. The model allows competition among managers and so, in general, updating depends not only on performance relative to a benchmark but relative to a peer group also.