Call knowledge where so-and-so, knowledge who so-and-so, etc., knowledge-wh. The reductive view says that knowledge-wh reduces to the two-place knowledge relation Ksp. Schaffer (2007) argues that this view has no viable response to the problem of convergent knowledge: how can a knowing-wh ascription be reduced to a Ksp ascription if a second knowing-wh ascription intuitively inequivalent to the first can be reduced to the same Ksp ascription? Instead he suggests that knowledge-wh be understood as a three-place knowledge relation Kspq, where q is a contextually salient contrast proposition. I argue firstly that once we realise that wh-questions can have more than one true answer, the reductivist has an obvious response to this problem. Secondly, I pose a revenge problem for Schaffer's contrastivist alternative: how can a knowing-wh ascription be reduced to a Kspq ascription if a second knowing-wh ascription intuitively equivalent to the first can be reduced to a distinct Kspq ascription?