The importance of geographic isolation in speciation has been debated since the 19th century. Since the beginning of the 20th century, the consensus has been that most speciation involves divergence in allopatry. This consensus was based largely on decades of observations by naturalists and verbal arguments against speciation without isolation. Recent attempts to quantify the importance of allopatric versus sympatric speciation using comparative methods called "age-range correlation" (ARC) suggest that allopatric speciation is more common than sympatric speciation. However, very few taxa have been studied and there are concerns about the adequacy of the methods. We propose methodological improvements including changes in the way overlap between clades is quantified and Monte Carlo methods to test the null hypothesis of no relationship between phylogenetic relatedness and geographic range overlap. We analyze 14 clades of mammals, chosen because of the availability of data and the consensus among mammalogists that speciation is routinely allopatric. Although data from a few clades clearly indicate allopatric speciation, divergence with gene flow is plausible in others and many results are inconclusive. The relative rarity of significant correlations between phylogenetic distance and range overlap may have three distinct causes: (1) post-speciation range changes, (2) relative rarity of range overlap, and (3) a mixture of geographic modes of speciation. Our results support skepticism about ARC's power for inferring the biogeography of speciation. Yet, even if few clades provide clear signals, meta-analytic approaches such as ARC may set bounds on the prevalence of alternative modes of speciation.