Can one define and test the hypothesis of (un)bounded rationality in stochastic choice tasks without endorsing Bayesianism? Similar to the state specificity of assets, we rely on state-specific goal formation. In a given choice task, the list of state-specific goal levels is optimal if one cannot increase the goal level for one state without having to decrease that for other states. We show that this allows to relate optimality more easily to bounded rationality where we interpret goal levels as aspirations. If for the latter there exist choices satisfying all state-specific aspirations and if one such choice is used, we speak of satisficing which may or may not be optimal.