In traditional dialectology, maps are divided into dialect areas on
the basis of isoglosses or with the use of the arrow method
\cite{daan:1969}. However, the choice of isoglosses is subjective. A
second shortcoming for both methods is that there is no way to
characterize relations between entire varieties.  Comparison is
inevitable atomistic. Third, existing methods records differences in
varieties without distinguishing degrees of differences. The view that
some differences vary along a continuum has no analytical
foundation. This paper focuses on the Levenshtein method. Using this
method, pronunciation differences can be measured for corresponding
pronunciations in two varieties. We can apply the method to a large
sample, providing an objective foundation for further analysis. The
differences can be added, which allows one to relate entire varieties,
aggregating the atomic differences. After comparing dialects on the
basis of their distances the dialects can be classified by clustering
or multidimensional scaling. Using clustering we get a sharp
classification in the form of a tree, where the dialects are the
leaves. Using multidimensional scaling we get a plot on which like
dialects are plotted nearby and unlike dialects are distant. When
scaling to three dimensions, a map can be colored, where the
dimensions represent respectively the intensity of red, green and
blue, while the areas between the dialects are colored by
interpolating from the dimensions of the dialects. In that way,
dialect variation is visualized as a continuum.