This paper introduces the pilot project for the Syntactic Atlas of Welsh Dialects, setting out the procedures for data collection and sketching a case study for one variable within the dataset, namely the patterns of negative concord found with the negative modal cau ‘won’t’. This item is a relatively recent innovation, and it is currently undergoing increasing integration into the negative concord system. The atlas fieldwork establishes current patterns of dialect variation, showing significant age variation indicative of change in progress and the rise of negative concord in this context. On the basis of this, it is argued that the diffuse geographical patterns attested are best interpreted as evidence of multiple innovation across a wide area, with new speakers re-implementing the innovation (“multiple reactuation”). A formal analysis is sketched out, treating the change as moving along a pathway of feature change with semantic features shifting to interpretable syntactic features and then to uninterpretable syntactic ones. This analysis is consistent with the dialect patterns and interspeaker implicational hierarchies found in the data.