is the only copy present in all arrangements, and it has been suggested that the Amy2 and Amy3 copies arose by duplication of Amy1 (Brown et al., 1990). In this study, we used only 5' and 3' flanking regions specific to the Amy1 gene.
A phylogenetic analysis of each flanking region generated the same branching topology, so we combined the two regions (Figure 2). In the deduced phylogeny, the TL arrangement splits off early from the SC–ST group, a finding concordant with the results of the previous RSP data analysis (Aquadro et al., 1991). An intuitive interpretation of this phylogeny is that the TL arrangement is ancestral to both SC and ST arrangements. and the SC–ST group calculated from our sequence data agrees well with that based on RSP data (2.9%). These findings seem to support the ancestral status of TL. What they do not provide is an exclusion of other gene arrangements as potential ancestors. Thus, we are still left with the following questions: Are our results consistent with another arrangement than TL being ancestral? And what about the HY arrangement, which has never been found in nature? Since no data are available for it, the phylogenies in Figure 2 cannot be used to rule out the possibility that HY was in fact ancestral.
To settle these questions, we developed a framework of competing hypotheses based on the assumption that each inversion type is monophyletic in origin, which accords with the RSP data (Aquadro et al., 1991), and the assumption that the relationships among the four central arrangements (Figure 1) are derived parsimoniously, with each inversion arising by two breakpoints from its parental arrangement. Under these assumptions, we successively chose each of the four possible arrangements to represent the ancestral type and asked in each case what the branching pattern of the resulting phylogenetic tree would be.