In the literature on the spatial spread of linguistic innovations, traditional models of diffusion have been adopted largely uncritically from (economic) geography. The validity of geographical critiques of these traditional diffusion models is demonstrated through an examination of the diffusion of eight linguistic innovations from the southeast of England to the Fens, a rural area approximately 150 km north of London. Whilst diffusion models in sociolinguistic dialectology have tended to expect vigorous innovations to eradicate conservative dialect forms in their path, and some of the innovations examined here have indeed conquered local variants, it is found that some of the spreading dialect forms are unsuccessful and others only partially so. This article reconceptualizes the diffusion of innovations as a form of dialect contact and demonstrates that in examining changes in this way, we are much better able to account for the outcomes of innovation diffusion than has traditionally been the case. Rather than innovations wiping out traditional dialect forms, they engage in contact with them in local communities: the outcomes of diffusion in the Fens bear all the hallmarks of contact-induced koineization, such as interdialect formation, levelling, and reallocation (Trudgill 1986). The success of innovations appears to be contingent on the social, interactive, and structural compatibility of the new form with the old one. If that compatibility is not present, diffusion may be slowed down, succeed only partially, or in some cases, be rejected entirely.