Method

The parsimony analyses were generated using the computer program PAUP* version 4.0 beta (Swofford, 1998). The phenotypic data provided in Appendix Table 1 were analyzed in two separate ways in order to determine the impact that different data sets may have on resulting topologies. First, all 92 characters were analyzed, and second, only those characters preserved in S. tchadensis or K. platyops were analyzed. Each data set was analyzed using the following procedures: (1) a strict consensus tree was generated; (2) a number of 50% majority-rule consensus trees was used, with the tree length continually increased up to three additional steps; and (3) a bootstrap analysis with 1,000 replications was generated.

All characters analyzed, except for characters 1, 3, 29, 74, and 77, were treated as ordered. Ordered characters are weighted in the sense that all intermediate stages are considered to have occurred; e.g., a change from

0 to 3 is weighted to reflect the three steps. Following Strait et al. (1997), it is also maintained that a change between adjacent states (e.g., between 2 and 3) is treated as a single step in a tree; i.e., equally weighted state changes are used throughout this study.

The strict consensus tree was obtained using the "heuristic" search option. The consensus tree is shown, along with its length, consistency, retention, and rescaled consistency indices. The consistency index (CI) is calculated from the number of homoplasies that must be assumed in the most parsimonious solution. The consistency index for a cladogram, as a whole, is equal to the total number of derived character states scored in the matrix, divided by the number of steps required to produce a tree. As such, the CI decreases as the level of homoplasy increases. Also, the amount of homoplasy will generally increase as the number of genera included also increases (Smith, 1994). The retention index (RI) measures the proportion of terminal genera that retain the character identified as a synapomorphy for that group. For example, if a character identified as a synapomorphy for a clade is present in all the terminal genera, it is given an RI of 1.0. If this same character, however, through later transformation or reversal, is present in only 50% of the terminal genera, its RI will now be just 0.5 (Smith, 1994). The rescaled consistency index (RCI) is calculated by multiplying the CI by the RI (Farris, 1989). It has been argued that the RI and the RCI are more robust in terms of being less sensitive to variations in maximum and minimum tree-length (Farris, 1989; Strait et al., 1997).

In each analysis a number of 50% majority-rule consensus trees was generated. These trees were continually generated by increasing the tree length by one step each time, until they had reached three steps. A 50% majority-rule consensus tree chooses the topologies that appear most often among the alternative cladograms. Only groups that appear in more than a specified percentage of all rival cladograms are used to construct a majority-rule consensus tree. Thus, adapting a 50% cut-off means that any group that appears in the majority-rule consensus tree is found in more than half of the competing cladograms (Schoch, 1986; Smith, 1994).

Finally, a bootstrap analysis was generated. This method of analysis selects characters at random from the data matrix, with replacement, thus constructing a new data matrix of the same dimensions as the original matrix table. The new data matrix may include the same character more than once, while others may be deleted altogether, because characters are chosen at random with a replacement option. This new matrix table is used to calculate a topology, and the process is repeated many times. In the analyses to follow, 1,000 replications were requested. Thus, a branch appearing in only 250 replicates represents just 25% (Noreen, 1989; Smith, 1994; Swofford, 1998).

Results

1. Analyses with all 92 Characters Included

The strict consensus tree of all 92 characters is shown in Figure 5.1, generated from 8 trees (tree length = 387; CI = 0.485; RI = 0.629; and RC = 0.302). From this analysis, after the divergence of the outgroup, Pan emerges, followed by Ardipithecus, then the anamensis group, followed by Praeanthropus. At this point we see a polytomy containing four clades. Sahelanthropus shares a common ancestor with the garhi group, with an expanded hominin clade containing Australopithecus, Kenyanthropus, and

Kenyapithecus Dryopithecus Graecopithecus Pongo Gorilla Pan

Sahelanthropus K. platyops K. rudolfensis H. habilis H. ergaster H. sapiens Australopithecus Garhi P. walkeri P. boisei P. robustus Praeanthropus Anamensis Ardipithecus

Figure 5.1 ► Strict Consensus Tree of 92 characters (see text for details).

Figure 5.1 ► Strict Consensus Tree of 92 characters (see text for details).

Homo, and another consisting of Paranthropus species. Within the expanded hominin clade Australopithecus splits off, followed by two sub-clades, one containing species of Kenyanthropus and the other the three species of Homo.

Following this analysis, a 50% majority rule consensus tree was requested, increasing the tree length by an additional step; i.e., to 388. This resulted in the generation of 86 trees, the consensus of which is shown in Figure 5.2. This tree is the same as the consensus tree just discussed. The replication values for this scheme are robust, even the lowest value is still relatively well supported; i.e., the placement of the Paranthropus clade with the Homo group in 62% of cases. The next analysis included the tree length increased by 2 extra steps (i.e., 389) resulting in the topology shown in Figure 5.3, from 502 likely trees. This tree again reproduces the consensus tree, and replication values are relatively high. The final analyses increased the tree length by

Kenyapithecus

Dryopithecus

Graecopithecus

Pongo

Gorilla

Sahelanthropus

0 0

Post a comment