Csibra and Gergely argue that, as part of an adaptation for ‘naturalpedagogy’, infants are prewired to recognise ‘ostensive signals’ asindicating that a speaker is acting with communicative intent (Grice’sproperty of ‘non-natural meaning’). I raise a series of objections totheir account. First, I argue that the signals Csibra and Gergely identifyare too loosely correlated with communicative intent to function as theyclaim. Without the introduction of a more careful specification of how themechanisms supporting the NPH work, this would make natural pedagogymaladaptive. This leads to a dilemma for Csibra and Gergely’sadaptationist approach. It has three horns, none of which are attractive.(1) One could posit an adaptation that tracks a property that we cannotspecify, and that may turn out not to exist (leaving the proposedmechanism inadequately specified). (2) One could propose a stepwise seriesof adaptations, the first developmental stages of which seem maladaptive(making its initially undergoing selection implausible). (3) One couldposit a single adaptation that that is in itself maladaptive (and soimplausible) but the functioning of which is refined by lots of individuallearning (thereby making the explanatory value of positing an initialadaptation mysterious). I conclude that given such a dilemma, we shouldnot posit any adaptive specialisation, and I finish with a discussion ofwhen adaptationism should be an acceptable default strategy.