This study examined the cranial and dental morphometric patterns among South American spider monkey taxa (Ateles), using various multivariate data reduction techniques to corroborate or falsify the predictions of alternative vicariance or parapatric models for explaining neotropical biodiversity. Data from 284 specimens were standardized to eliminate sex differences and then transformed by factor analysis to obviate the effects of body size on shape relationships. We found the independent discriminant analyses of dental and cranial data to be not at all dissimilar, nor did body size appear to distort geographical relationships. Thus a pooled data set of 50 measurements was further reduced by cluster analysis to produce a dendrogram of taxonomic relationships that is discordant with current classification. Three independent taxa are indicated, with the Guianan samples being most distinct, followed by the animals of northwest Venezuela and adjacent Colombia. Elsewhere, the samples describe a cline of relationships around Amazonia that resembles a ring species. Of particular interest is the apparent clinal intergradation across major tributaries of the Amazon, which is inconsistent with the concept of these rivers as barriers in their present form. Instead these observations, and the congruence between the multivariate clusters and previously identified centers of endemism, support both tierra firme and forest refugia models of historical biogeography. However, the latter model is somewhat falsified within Amazonia by character stability for alternative pelage patterns on opposite sides of the same major rivers that showed morphometric intergradation, but not as predicted within the refugia. This parapatric enhancement of pelage differences, which originated presumably in allopatry, suggests a mechanism of incipient speciation by frequency-dependent assortative mating with no apparent hybrid disadvantage or disruption of widespread gene flow.