This article provides a qualitative study of variation in negation in 17 th century letters from people of different regions and social classes of the Dutch speaking area. This intermediate language stage between negative concord in Middle Dutch and single negation in Modern Dutch is affected by both bottom-up change (the negative clitic started to erode due to functional redundancy) and top-down change (some elite writers, like P. C. Hooft and Joost van den Vondel, started to omit the negative clitic completely). The letters display different surface varieties: optional deletion of the negative clitic, conservative negative concord and progressive single negation. However, in the underlying syntactic structure, the syntactic features of the negative clitic polarity features projecting PolP – discriminate four different derivational systems: (i) a high PolP as part of an extended CP; (ii) a low PolP that attracts the finite verb to T; (iii) both high PolP and low PolP; and (iv) no PolP. We will argue that the letters with (incidental) negative concord represent a stage in which one or two PolPs are present. In particular, the letters with optional deletion of the negative clitic show that the clitic extends its function from true negation to emphasis on negation of a presupposition (as found in West Flemish) to general contrast on the polarity of the clause. This less specific function is indicative of semantic bleaching and a motive for eventual loss of the clitic. The letters without negative concord show completion of this process and hence loss of the PolP. However, we also show that at least it is feasible that a subset of letters with single negation still has a PolP underlyingly, which means that although the negative clitic is absent from the surface, its polarity features are present at the underlying syntactic structure. The negative features are present during the syntactic derivation in a PolP, however, we infer, they are not spelled-out at the surface, possi-bly as an effect of upper-class ...