Comparison of our results with those of Kay et al. was complicated because the number of characters in the data then posted (281) was greater than the number in the article (256). Taxa included in the outgroup, as well as those in the ingroup, differed from those described in the article. Kay et al. did not provide the number of resulting trees, their lengths, or describe what type of consensus formed the basis for their conclusions. The similarity of our results to those of Kay et al., when we ran the posted data for only 30 replications, is not unexpected because we did not override PAUP's (4) random number seed of 1. Our more parsimonious results after 1000 replications probably differed as a result of the greater number of replications with better sampling of possible cladograms.

Five conclusions offered by Kay et al. form the basis of their scenario of anthropoid origins. Our analysis of the data posted at the time of publication (Fig. 1) found little support for these points: (i) Haplorhine-Strepsirhine dichotomy. The monophyly of Haplorhini was not supported, questioning the usefulness of recognizing this dichotomy. (ii) Adapiformes-Lemuriformes relationships. We found adapiformes to be the sister group of extant lemuriformes, but this clade includes Macrotarsius and Rooneyia, both traditionally considered omomyids. (iii) Omomyid relationships. No omomyids formed a sister group relationship with anthropoids, and interrelationships among omomyids were essentially unresolved. (iv) Eosimiidae-Anthropoidea relationships. Our strict consensus tree (Fig. 1A) indicated that Eosimias and the tentatively assigned "Eosimias petrosal" do not share a sister group relationship with anthropoids, but a 50% majority rule consensus of the same results (Fig. 1B) provided limited support for the dentition and jaw (but not the petrosal being anthropoid). (v) Tarsius relationships. Tarsius is nested within omomyids and is the sister taxon to washakiin omomyids. It did not appear to be the sister taxon of anthropoids in either our strict consensus or our 50% majority rule consensus.

Kay et al. changed the Web site data (as is noted on the Web site) since our first analysis. While the ingroup now contains only the taxa cited in their article, the taxa in the outgroup are still different. The number of characters used in the analysis is still 281 (rather than the 256 stated in the article by Kay et al.), and at least one coding of morphology has been changed. We now find 21 equally most parsimonious cladograms of length 94,906 steps, and the five conclusions stated by Kay et al. appear to have more support. However, a strict consensus of our most parsimonious cladograms is still inconsistent with the phylogeny they described, and neither the data posted at the time of publication nor the data currently posted seem to be identical to those they describe in their article.

Discovery of more parsimonious interpretations of the data presented originally by Kay et al. reopens the issue of primate relationships that Kay et al. describe as resolved. Maddison (5) and Templeton (6), analyzing the "Eve Hypothesis," found phylogenetic problems on this scale to require intensive application of computing resources, and a heuristic analysis with 30 replications is not enough. Archiving original data and reporting results in an unambiguous fashion should be the norm for publication in phylogenetics as for other disciplines, especially when they address issues of widely recognized importance.

Jonathan I. Bloch
Daniel C. Fisher
Philip D. Gingerich
Gregg F. Gunnell
Museum of Paleontology and
Department of Geological Sciences,
University of Michigan,
Ann Arbor, MI 48109-1079, USA
Elwyn L. Simons
Department of Biological Anthropology and Anatomy,
Duke University,
Durham, NC 27705, USA
Mark D. Uhen
Cranbrook Institute of Science,
Bloomfield Hills, MI 48303-0801, USA

REFERENCES AND NOTES

1. R. F. Kay, C. Ross, B. A. Williams, Science 275, 797 (1997) [Abstract/Free Full Text] .
2. Kay et al. state [in figure 3 of (1) on page 799] that their data are "available at" http://www.sciencemag.org/feature/data/kay.shl. When we checked this site, it had no data, but instead referred the reader to the personal home page of co-author Callum Ross, at http://www.informatics.sunysb.edu/anatomy/cross.html. Our first analysis was based on data downloaded from this site on 24 March 1997. These data have been changed subsequently, as stated in a brief note on the C. Ross Web site, at http://www.informatics.sunysb.edu/anatomicalsci/aboutmx.
3. D. C. Fisher, in MacClade, Analysis of Phylogeny and Character Evolution, W. P. Maddison and D. R. Maddison, Eds. (Sinauer, Sunderland, MA, 1992), pp. 124-129; D. C. Fisher, in Interpreting the Hierarchy of Nature, L. Grande and O. Rieppel, Eds. (Academic Press, San Diego, CA, 1994), pp. 133-171.
4. D. L. Swofford, PAUP: Phylogenetic Analysis Using Parsimony, Version 3.1 (Illinois Natural History Survey, Champaign, IL, 1993). Our tree lengths treat multistate codings as uncertainty.
5. D. R. Maddison, Syst. Biol. 40, 355 (1991) .
6. A. R. Templeton, Science 255, 737 (1992) [Free Full Text] .
7. We thank F. Ankel-Simons, W. C. Clyde, D. L. Fox, E. A. Kowalski, and J. A. Trapani for comments and discussion. B. Miljour aided in production of the figure.

Response: Our group's cladistic analysis of dental and osteological characters supported the fundamental Haplorhine/Strepsirrhine dichotomy in primate evolution extending back to the early Eocene [figure 1 in (1)]. Bloch et al. state that our data do not support these results and that we ran an insufficient number of replicates of our analysis. We have performed supplemental phylogenetic analyses, and these support our original conclusions. We are unable to reproduce the results given by Bloch et al. either with the data that we originally presented on the World Wide Web or with any other data set.

In our supplemental phylogenetic analyses, we used three data sets (Table 1):

Table 1. Summary of our supplemental phylogenetic analyses. Clade relationships are described in the text and in Figs. 1 and 2. An X indicates that majority-rule consensus supports the proposed clade. Ad, Adapidae; An, Anthropoidea; E, Eosimias; H, Haplorhini; O, Omomyidae; S, Strepsirrhini; T, Tarsius. Data set Number of trees found Clade relationships (i) H/S (ii) Ad/S (iii) O/H (iv) E/An (v) T/E/An 1. Table 3 of (1) 21 X X* X X X 2. Web site 1 X X X --§ X 3. Combined 17 X X* X X X * Donrussellia sister to all Haplorhines. Rooneyia is near base of strepsirrhines. The taxon "Eosimias petrosal" falls within Omomyidae. § The position of Eosimias is unresolved with respect to the T/An clade: the taxon "Eosimias petrosal" is sister to (T/An), whereas the taxon "Eosimias" is sister to Tarsius.

1) Data from table 3 of our article (1). Our original analysis [summarized in figure 3 and table 3 of (1)] was based on 50 ingroup taxa and three outgroup taxa. We selected the ingroup taxa because they sampled the full phenetic range of the fossil taxa, are relatively complete, and have historic importance in the debates about anthropoid origins. Eosimias was represented as two taxa in all analyses; the data on a petrosal bone of uncertain assignment were run separately.

2) Web site data. Bloch et al. initially analyzed a data set that we placed on the Web (http://www.informatics.sunysb.edu/anatomy/cross.html) that inadvertently excluded two of the taxa listed in our report (Strigorhysis and Uintanius) and included two added taxa not included in the report (a second species of Macrotarsius and the early anthropoid Qatrania). An equivalent dataset is now archived at http://www.sciencemag.org/feature/data/46622.shl.

3) Combined data. Data from table 3 in (1) with two taxa added (Macrotarsius sp. and Qatrania).

Bloch et al. state that there is a discrepancy in the number of characters in our posted data sets (281), as compared with the number of characters we mentioned in the article (256). This discrepancy is apparent, not real: The posted character list included (and still includes) a number of characters that are invariant in the taxa examined here, but crop up in other taxa analyzed by us in a more exhaustive analysis now in preparation (2). Of the 281 characters posted, 256 are "informative" ones [see, for example (3), p. 199]. Further, G. F. Gunnell (one of the co-authors of Bloch et al.) kindly pointed out to us that the matrix entry for the number of incisors in Eosimias was incorrect; this has been rectified in the current analyses.

As in our article (1), all analyses used the phylogenetic analysis program PAUP (4) with a "proportionate" weighting scheme for multistate characters. This allows us to discern a variety of states for some characters without "penalizing" dual state characters. So as to ensure primate monophyly, 50 two-state "dummy" characters were added, each of which scores the outgroups as primitive and the primate taxa as derived. All analyses were undertaken with the use of the "heuristic" search mode, selecting "add-sequence" and "subtree pruning, regrafting" options of PAUP, with 1000 repetitions. Exhaustive searches were not feasible, given the large numbers of characters and taxa. Comparisons between our most parsimonious "Web" tree and the majority consensus found by Bloch et al. were undertaken with the use of the MacClade program (3).

All of the supplemental analyses (with salient features summarized in Table 1 and Figs. 1 and 2) produce results in agreement with the five main conclusions stated in our article (1) and contrary to the trees found by Bloch et al. (their figure 1B). All of our trees show that (i) a primary dichotomy exists among living taxa dividing living Haplorhini (Tarsius+Anthropoidea) from Strepsirrhini (Lemuriformes); (ii) Adapidae are assignable to the strepsirrhine side of the dichotomy; (iii) Omomyidae are assignable to the haplorhine side of the dichotomy; (iv) Eosimiidae are sister to late Eocene-to-Recent Anthropoidea; and (v) Tarsius is either the sister group to (Eosimias, Anthropoidea) or nested within omomyids, depending on allocation of a petrosal bone to Eosimias.

Bloch et al. describe finding 21 trees with the use of data from table 3 in (1), presumably the same 21 that we found (1), although they have declined our offer to exchange original trees to confirm this point. They state that a "strict consensus" of these 21 trees is "inconsistent" with our conclusions. However, the strict consensus of these 21 trees does support our conclusions (Fig. 2).

We find a single most parsimonious tree with the use of the original Web site data with 1000 repetitions, and the tree supports our original finding. Our tree is about 2% shorter than the majority consensus of 98 trees illustrated by Bloch et al., although without examining the individual trees one cannot calculate precisely how much shorter.

The results of some (not all) of the reanalyses depart from those in figure 3 in our article in that Rooneyia, placed by us (with a query) as a haplorhine falling outside of Omomyidae now falls as a basal member of Strepsirrhines. In the original analysis, we depicted the clade [(Eosimias, Anthropoidea) Tarsius], as always arising out of the Omomyidae and specifically allied with washakiine omomyids. The revised analysis always depicts this clade as arising out of Omomyidae, and sometimes, but not always, supports a sister-group relationship with the Washakiinae.

Bloch et al. state that their reanalysis of our data challenges the usefulness of recognizing a haplorhine-strepsirrhine dichotomy of living primates. In this regard, we note that our data set does not include any characters of physiology, development, or soft-tissue anatomy because we were trying to place the Eocene-Oligocene fossil primates, for which such data are unavailable, into phylogenetic context. Many other characters not preservable in the fossil record, as summarized by Martin (5), also support the halplorhine-strepsirrhine dichotomy, including loss of the ability to synthesize vitamin C, chromosome morphology, the absence of a rhinarium, the reduced development of the nasal turbinals, several aspects of the organization of the visual cortex, several features of the anatomy of the eye, distinctive features of placentation, and early development. Our findings strengthen this fundamental dichotomy and allow it to be extended to fossils.

Richard F. Kay
Department of Biological
Anthropology and Anatomy,
Duke University,
Durham, NC 27710, USA
Callum Ross
Department of Anatomical Sciences,
Health Sciences Center,
State University of New York,
Stony Brook, NY 11794, USA
Blythe A. Williams
Derek Johnson
Department of Biological
Anthropology and Anatomy,
Duke University

REFERENCES

1. R. F. Kay, C. F. Ross, B. A. Williams, Science 275, 797 (1997) .
2. C. Ross, B. A. Williams, R. F. Kay, in preparation.
3. W. P. Maddison and D. R. Maddison, MacClade Version 3.4 for Macintosh (Sinauer, Sunderland, MA, 1992).
4. D. L. Swofford, PAUP: Phylogenetic Analysis Using Parsimony, Version 3.1.1 and accompanying manual (Illinois Natural History Survey, Champagne, ed. 3.0, 1993).
5. R. D. Martin, Primate Origins and Evolution (Chapman & Hall, London, 1990).