In Moggi’s computational calculus, reduction is the contextual closure of the rules obtained by orienting three monadic laws. In the literature, evaluation is usually defi ned as the closure under weak contexts (no reduction under binders): E = [] | let x:=E in M. We show that, when considering all the monadic rules, weak reduction is nondeterministic, non-confluent, and normal forms are not unique. However, when interested in returning a value (convergence), the only necessary monadic rule is beta , whose evaluation is deterministic. The proof relies on tools coming from a calculus inspired by linear logic.