Listening to radio or television one notices the tendency for standard Dutch (ABN) to become more and more differentiated, i.e. regionally colored. Hoppenbrouwers (1990) showed the opposite tendency for dialects. Being influenced by standard Dutch and by each other they have become less differentiated and fused to larger wholes: regiolects (see also Hinskens (1993), Auer & Hinskens (1996), and Hinskens, Auer & Kerswill (2005)). While earlier scholars usually describe this change in terms of single linguistic phenomena, we plan to investigate this change using modern web-based and computational techniques to obtain an overall picture of this change. Our goal is to examine how the change from dialects to regiolects is reflected in the production and perception of the dialect speakers. The results of our research will give insight into the nature of language change and dialect levelling. The research is important for historical linguists since it gives information about the direction and rates of sound change. The research will be based on representative Dutch dialects of approximately 80 locations in the Netherlands and North Belgium. Perceptive distances are obtained on the basis of a web survey in which speakers listen to recordings. Computational distances are found on the basis of the transcriptions of the recordings. In the experiments two groups are distinguished: conservative dialect speakers (old males) and innovative dialect speakers (young females). We will test three hypotheses. First, perceptive distance measurements which are based on the recordings of innovative speakers will suggest larger areas than those which are based on the recordings of conservative speakers. Second, the change from dialect to regiolect affects the lexical level (‘kopstubber’ becomes ‘roagebol’) more strongly than the phonological (‘hoes’ becomes ‘huus’) and phonetic levels. Third, this change also affects the perception of the speakers, but perception lags behind production.