P. A. Marquet<strong>the</strong> Late Pleistocene caused <strong>the</strong> ext<strong>in</strong>ction <strong>of</strong> some lowland species, fur<strong>the</strong>r contribut<strong>in</strong>g to<strong>the</strong>se altitud<strong>in</strong>al and latitud<strong>in</strong>al patterns <strong>of</strong> diversity, and also acted as a strong selectiveenvironment favour<strong>in</strong>g <strong>the</strong> evolution <strong>of</strong> desert-adapted physiological traits <strong>in</strong> those speciescurrently found with<strong>in</strong> <strong>the</strong> conf<strong>in</strong>es <strong>of</strong> <strong>the</strong> coastal desert.Community StructureIn light <strong>of</strong> <strong>the</strong> complex <strong>in</strong>teractions among dispersal, diversification, adaptation andext<strong>in</strong>ction processes that shaped patterns <strong>of</strong> species diversity <strong>in</strong> <strong>the</strong> coastal desert andadjacent Andean area, one is tempted to predict a highly <strong>in</strong>dividualistic pattern <strong>of</strong> communitystructure at a regional scale (see Graham 1986; Brown and Kurzius 1987). I have juststarted to address this question, follow<strong>in</strong>g <strong>the</strong> lead <strong>of</strong> Brown and Kurzius' work on patterns<strong>in</strong> species distribution and coexistence at a regional scale. From <strong>the</strong> analysis <strong>of</strong> 120sites distributed with<strong>in</strong> <strong>the</strong> Atacama <strong>Desert</strong> and Puna areas <strong>of</strong> <strong>Peru</strong>, Argent<strong>in</strong>a and Chile(Marquet, unpublished data), <strong>the</strong> follow<strong>in</strong>g patterns arise. First, most sites (80%) are<strong>in</strong>habited by 1-3 species (Fig. 6); this is significantly different from that expected froma random association <strong>of</strong> species with sites (i.e, under a Poisson distribution) (G = 41.8,d.f. = 12, P < 0.001). Second, when <strong>the</strong> number <strong>of</strong> different species comb<strong>in</strong>ationsis analysed (Fig. 7), a distribution similar to that found <strong>in</strong> Fig. 6 emerges. However,because sites with few species are more likely to have redundant comb<strong>in</strong>ations <strong>of</strong> species,<strong>the</strong> frequency distribution <strong>of</strong> different comb<strong>in</strong>ations observed peaks at three, ra<strong>the</strong>r thantwo, species. A G-test shows that <strong>the</strong> observed distribution is significantly different froma Poisson distribution (G = 76.44, d.f. = 12, P < 0.001). F<strong>in</strong>ally, with regard to <strong>the</strong>Number <strong>of</strong> speciesFig. 6. Frequency distribution <strong>of</strong> <strong>the</strong> number <strong>of</strong> coexist<strong>in</strong>g species among local sites. Thecurved l<strong>in</strong>e represents expectations under a Poisson distribution.
<strong>Diversity</strong> <strong>of</strong> <strong>Small</strong> <strong>Mammals</strong> <strong>in</strong> a <strong>Coastal</strong> <strong>Desert</strong>Number <strong>of</strong> species per siteFig. 7. The number <strong>of</strong> different comb<strong>in</strong>ations <strong>of</strong> different numbers <strong>of</strong> locally coexist<strong>in</strong>g species.The curved l<strong>in</strong>e represents expectations under a Poisson distribution.Number <strong>of</strong> times comb<strong>in</strong>ation observedN = 63 different comb<strong>in</strong>ationsFig. 8. The number <strong>of</strong> times that different comb<strong>in</strong>ations <strong>of</strong> locally coexist<strong>in</strong>g species wereobserved.