Aim: The relationship between geographic distance and similarity in species composition is regularly used as a measure of species turnover and beta diversity. Distance decay analyses are applied, cited and compared despite variable extent, and different grain sizes of records (e.g. plots, islands, states) are regularly used within such analyses. Currently, differences among distance decay relationships that cover different grain sizes and extents are attributed to ecological processes that are suspected to operate differently over varying extent and grain size. We assess whether the implicit assumption that the distance decay relation is independent from grain size and study extent is valid, or whether sampling design could be the underlying cause for observed differences. Location: An artificial one-dimensional "landscape". Methods: The distance decay relationship was quantified in simulated communities. Grain and study extent were varied systematically. In each sampled data set the linear relation of Simpson and Sørensen similarity to geographic distance (on both a log-transformed and the original scale) between 100 even sized equidistant plots was assessed using linear regression and generalized linear regression with log-link function. Regressions were applied either including or removing zero similarities from the data. Results: Both, slope (measuring turnover) and goodness of fit measure r2 (quantifying the influence of space on species composition) of the distance decay relationship were strongly influenced by grain and study extent. Approaches that are able to cope with zero similarity values of large distance comparisons were less dependent on grain and extent. Main conclusions: Reported differences between landscapes detected by current distance decay measures cannot be explicitly traced back to ecological scale-specific processes. Instead, they can largely be attributed to sampling design and are highly sensitive to grain size and study extent. More appropriate approaches for the study of distance decay and the understanding of scale specific processes are required.