Dialectometric intensity estimation as introduced in Rumpf etal. (2009) and Pickl and Rumpf (2011, 2012) is a method for the unsupervised generation of maps visualizing geolinguistic data on the level of linguistic variables. It also extracts spatial information for subsequent statistical analysis. However, as intensity estimation involves geographically conditioned smoothing, this method can lead to undesirable results. Geolinguistically relevant structures such as rivers, political borders or enclaves, for instance, are not taken into account and thus their manifestations in the distributions of linguistic variants are blurred. A possible solution to this problem, as suggested and put to the test in this paper, is to use linguistic distances rather than geographical (Euclidean) distances in the estimation. This methodological adjustment leads to maps which render geolinguistic distributions more faithfully, especially in areas that are deemed critical for the interpretation of the resulting maps and for subsequent statistical analyses of the results.