Sampling and ecology in three Early Miocene catarrhine assemblages from East Africa

Fossil catarrhines are abundant at the Early Miocene East African fossil sites of Songhor, Koru, and Napak. Despite their close temporal (19.5-20 Ma) and spatial proximity, the catarrhine species found at these sites vary; some species are found at all sites, while others are found only at one. Previous analyses of these catarrhines have assumed that the observed pattern of catarrhine distribution is unusual, and must be caused by habitat differences between these localities in the Early Miocene. But is it not equally likely that other factors, such as sampling and taphonomic biases, have also affected these assemblages?;I test the hypothesis that observed differences in catarrhine distribution at the Songhor, Koru, and Napak localities (49 localities) are due to habitat differences. Through analysis of presence/absence distribution patterns, relative abundances and rarefaction analysis, I demonstrate that at least some of the observed differences in catarrhine distribution amongst these localities are genuine. While there are large differences in sample size, the observed catarrhine differences cannot be explained solely by sampling error.;If catarrhine distributions are not due solely to sample size effects, then this supports the idea that habitat variation may be an important causal agent. Widespread habitat differences between localities should be reflected in the entire mammalian fauna, not just in catarrhine primates. First, I conduct a detailed analysis of overall taxonomic representation, body size representation, and skeletal part representation to determine whether localities are isotaphonomic. My results show that localities are generally similar in their depositional histories, with the exception of Napak II and Napak CCI.;Detailed comparison of the entire mammalian fauna using cluster analysis is unable to recover consistent patterns in faunal make-up across localities. This suggests that habitat variation is subtle, and does not cause major faunal differences between localities. To further test this idea, I apply the same multivariate techniques to a set of modern African forests from East Africa where habitat differences are known. Cluster analysis successfully groups these modern forests, suggesting that whatever habitat differences existed between Songhor, Koru, and Napak are weaker than those seen in modern East African forests today.