Although great effort has been devoted to the descrip-tion of richness patterns (see, as most recent examples,Svenning and Skov 2007, Montoya et al. 2007, Reyjolet al. 2007, Whittaker et al. 2007), much less literatureon large scale beta diversity patterns is available (but seeQian et al. 2005). In a recent contribution, Soininenet al. (2007) performed an interesting meta-analysissearching for relationships between different predictorsand the decay of similarity with geographic distanceacross a wide range of organisms, ecosystems andgeographical gradients. Though the aim, objectivesand conceptual background of this paper are veryinteresting, ecologists should be aware of strong draw-back in their analytical framework that could poten-tially invalidate their major conclusions. One of theirmost remarkable results is the finding of a fasterdistance decay of similarity at higher latitudes, contra-dicting all current data and disagreeing Rapoport’s rule,which predicts a positive correlation between speciesrange size and latitude. I show here how the resultmay be produced by the confusing effect of richnessgradients combined with the selection of a simila-rity measure inappropriate for the hypothesis tested.Soininen et al. (2007) sought to describe beta diversitypatterns based on Sorensen similarity measure, which isknown to be strongly biased by richness patterns (Koleffet al. 2003). Soininen et al. (2007: p. 8) acknowledgedthat this characteristic of the Sorensen metric could be aproblem for the interpretation of their results whendiscussing the relation between similarity decay andlatitude. In fact, there is a problem in all their testsinvolving variables covariating with richness (as trophicposition or body weight, for example), because richnessgradients distort turnover patterns if the similaritymeasure (Sorensen in this case) incorporates differencesin richness as differences in composition. The solutionwould be to select a measure of beta diversity notinfluenced by richness differences.