vol. 154, no. 2 the american naturalist august 1999 Wednesday Jul 21 12:14 PM/980258/ll/alw Using Phylogenies to Test Macroevolutionary Hypotheses of Trait Evolution in Cranes (Gruinae) Arne Ø. Mooers, 1,2,* Steven M. Vamosi, 1 and Dolph Schluter 1 1. Department of Zoology and Centre for Biodiversity Research, University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada; 2. Department of Biology and Zoological Museum, University of Amsterdam, P.O. Box 94766, 1090 GT Amsterdam, The Netherlands Submitted July 28, 1998; Accepted March 18, 1999 Keywords: speciation, phylogeny, macroevolution, Gruinae. We present a novel phylogenetic approach to the study of trait evolution. We compare explicit models describing the evolution of phenotypic characters over the phylogeny of a group, employing a method that avoids point estimates of ancestor reconstructions (see Maddison 1990, 1995; Lo- sos and Miles 1994; Schluter et al. 1997; Zhang and Nei 1997; see also Pagel 1997). We contrast well-characterized models of change in species-specific phenotype: gradual change, in which differences in trait values among species are expected to be proportional to the time available for change; genetic change, in which the differences between species in the tested trait are correlated with the number of inferred substitutions that have accumulated between them (Omland 1997; Pagel 1997); and speciational change (Rohlf et al. 1990; Martins and Garland 1991; see Diaz- Uriarte and Garland 1996; Pagel 1997), in which differ- ences are proportional to the number of speciation events separating species. We also consider a new nonhistorical model in which changes are completely independent of history, such that differences among species are expected to be the same, regardless of relationship. These four mod- els are in turn contrasted to a fifth, null hypothesis (the free model), whereby individual lineages accrue pheno- typic change at their own idiosyncratic rates. We illustrate this method by examining a number of traits presumed to be involved in sexual signaling in the * E-mail: mooers@bio.uva.nl. Am. Nat. 1999. Vol. 154, pp. 000–000.  1999 by The University of Chicago. 0003-0147/1999/15402-0011$03.00. All rights reserved. bird family Gruidae (cranes): plumage characters, unison call characters, and the shape of the bones that house the convoluted trachea presumed responsible for the species- specific calls. If sexual selection is intimately associated with the speciation process, then changes in sexually se- lected traits will be closely associated with speciation events, rather than changing gradually or idiosyncratically. Theory (Lande 1981, 1982; Schluter and Price 1993; Turner and Burrows 1995), comparative evidence (West-Eberhard 1983; Turner 1994; Barraclough et al. 1995; Mitra et al. 1996), and field observations and experiments (Baker and Baker 1990; Houde and Endler 1990; Endler and Houde 1995; studies reviewed in Carson 1997; Saetre et al. 1997) all point to a correlation between sexual selection and speciation. If the traits we model are involved in mate choice, we predict that the speciational change model will offer a better fit to the data than the alternatives. If trait changes are caused by continuous selection pressures that change direction often (Felsenstein 1981, 1988) or drift in large populations (Lande 1976; Lynch 1990), the gradual model should outperform the others. If trait evolution is governed by the same processes causing molecular sub- stitutions, then the genetic model should be a good de- scriptor. If the evolution of the characters is not con- strained by phylogenetic history, then the nonhistorical model will offer the best description. Finally, if change has a strong historical component but is highly idiosyncratic, the free hypothesis should do best. This last free model also offers a detailed description of how the traits have actually evolved over the history of the group, informing us about other possible processes besides those suggested here. Methods Hypotheses Figure 1 depicts the five models, which are all based on a Brownian motion process. In each, the branch lengths represent the a priori expectation for the amount of change in the modeled traits. The gradual model (fig. 1A) is the one most commonly employed for change in continuous