Numeric.Probability.Either -> Control.Monad.Trans.Except Object -> Experiment? Object avoid functional dependencies or type class, at all Examples: Election and prognoses generalize fixed Dist to (Distribution prob) whereever possible Use pretty printer in PrintList? Which one? current pretty function is nice for single Dist values, but not for (Dist a, Dist b) et.al. Collection.draw using StateT QuickCheck properties use a non-empty list structure for the distribution generalize ToFloat class to Num Need for multi-parameter type classes? create export list new data type for continuous distributions: C (Float -> Float) prove correctness of >>=? (What did Martin mean with this comment?)