In the talk, I explore the empirical landscape of Czech comparatives and related forms. Drawing on Caha, de Clercq and Vanden Wyngaerd (2018), I give reasons to think that the comparative decomposes into two projections. Against this background, I explore the implications of a curious class of comparatives which violate a candidate universal, discussed recently by Grano and Davis (2018), to the effect that comparatives are always based on the positive degree.