Potential characters will be looked for at all scales and at all levels of resolution, therefore light microscopy, computer-assisted morphometrics, and SEM will be learned and employed by all trainees. The recognition and definition of characters and character states is the empirical basis for all systematic studies. In particular, phylogenetic analysis is composed of two distinct phases. The first, character analysis, can be defined as all the steps leading up to the production of a data matrix, i.e., OTU circumscription (what are the rows?), character definition (what are the columns?) and state recognition (what are the correct entries?). The second, cladistic analysis, can be defined as the formal transformation of the data matrix into a phylogenetic tree. The second phase has attracted the most theoretical attention, while the conceptual bases of the first phase have remained relatively less explored (but see Mishler and De Luna, 1991).

Within the first phase of phylogenetic analysis, the questions of OTU circumscription and character definiton have received attention, the former under the guise of the "species problem" (Mishler and Brandon, 1987), the latter in discussions of the concept of homology (Roth, 1988). The question of character-state division is the most problematic at present (e.g., see extensive discussion of empirical problems by Stevens, 1991). In the approach to characters and character-state division we are advocating here, semaphoronts (Hennig 1979: individual, character-bearing "snap-shots" of the life cycle; essentially individual collections in mosses) should be examined for features that appear to unite them into homogeneous groups. Not all of these features will turn out to be legitimate taxonomic characters, of course, since they may fail one or another test in character analysis (Mishler and De Luna 1991). Characters should be heritable, independent, and have discrete states (Wiley, 1981).

Operational Taxonomic Units (OTU's) are hypotheses of minimal phylogenetic groupings given the evidence at hand. They are formed in an iterative process: initial groupings may either be joined (if the features originally thought to distinguish them are found to be non-heritable, etc.) or further split (if a new distinctive character is found). In this iterative process, character-state decisions are necessary at each step. In practice, highly distinctive differences are noted first, and other, more subtle, differences recognized later. This basic empirical procedure for defining characters and states (Mishler and De Luna, 1991) can be summarized as follows: (1) relatively discrete taxonomic characters are used to recognize tentative OTU's; (2) less obvious distinctions (i.e., potential taxonomic characters showing greater overlap among OTU's) are tested for their association with (and thus support of) tentative OTU's using analysis of variance (ANOVA), followed by multiple range tests. Potential characters with discrete states still need to be examined for their heritability and independence from other characters. Heritability will be examined through comparative growth experiments, using standard techniques (e.g., Mishler, 1985); independence will be examined through multivariate statistical techniques to explore the correlation structure among the measured variables.

Each character in a data matrix can be viewed as an individual hypothesis of taxic homology. There is not space to fully justify a parsimony approach (see Mishler 1994), but the most parsimonious tree can be viewed as the joint hypothesis that allows the largest number of the individual hypotheses of homology to be maintained. We will use MacClade (Maddison and Maddison, 1992) for handling the data matrix and for examining trees, and PAUP (Swofford, 1991) for reconstructing phylogeny. A preliminary set of potential characters for Mitthyridium is presented in Table 2; these will be considerably refined in the course of the proposed research.

Table 2. Preliminary list of morphological characters potentially useful in defining the genus Mitthyridium and at the species level within it. Quantitative characters will be divided into states following the ANOVA technique of Mishler &;De Luna (1991); see text for short explanation.

1. Leaf arrangement around the stem (loosely arranged vs. tightly crispate around stem). 2. Plant habit (repent with ascending-erect branches vs. stems erect, not repent). 3. Length of leaf. 4. Width of leaf. 5. Position of widest point on leaf (length of widest point from base divided by total leaf length). 6. The length of leaf in relation to cancellinae. 7. Angle between costa and middle of cancellinae (measured from the point cancellinae meets the chlorophyllose cells on the costa and the middle point at which cancellinae meets the chlorophyllose cells). 8. Angle between costa and edge of leaf (the furthest end at which cancellinae meets the chlorophyllose region). 9. Length of border from base (total length of border from bse divided by total leaf length). 10. Width of border at the widest point. 11. Point at which dentation begins. 12. Point at which dentation ends. 13. Length of teeth at margin of leaf. 14. Thickness of leaf margin (unistratose vs. bi- or multistratose). 15. Thickness of cell walls. 16. Angle from leaf tip to a point 1.5 cm away on the edge of leaf. 17. Angle from leaf tip to a point 4 cm away on the edge of leaf. 18. Peristome (absent vs. present). 19. Nature of calyptra (enclosing capsule and open by slits vs. cucullate. 20. Size of plants. 21. Apex (plane, involute, or funnel-shaped).

To Next Section: Molecular Studies

Return to Mishler PEET Home Page

Return to Mishler Lab Home Page