I compare two theories to account for the novel observation that ellipsis is ungrammatical in tautologous conditionals, e.g. If John is wrong, then he is *(wrong). One theory attributes the ungrammaticality to a contrast failure in ellipsis parallelism (Rooth 1992a,b); the other to triviality at a more abstract, logical level (Gajewski 2009). The ellipsis parallelism theory prevails on further data, joining Griffiths (to appear) in arguing that contrast plays a role in ellipsis licensing. Contrast is further shown to be sensitive to intensionality.