Polarity – or what constitutes a primitive character state?
In their study of adult characters, McLaughlin & Lemaitre (1997) used Neoglyphea inopinata Forest & de Saint Laurent, 1975, as the out-group for testing the hypothesis that carcinization had occurred and that lithodids represented the advanced state. No information on development is available for N. inopinata, thus no parallel out-group analysis is possible.
Kluge (1985) and O’Grady (1985) both perceived ontogenetic data as providing three basic uses to phylogenetic systematics, i.e., to provide assessment of homology, to serve as an extra source of observations with which to judge historical relationships, and to serve to polarize character transformations. Although Christofferson (1987, 1988a, b, 1989) and Ng & Clark (2000) used both larval and adult characters in assessing phylogenetic relationships of the Eucarida, Caridea, and certain Brachyura, as pointed out by Hickman (1999), the exclusive use of invertebrate larval data in cladistic analyses is in its infancy. Only a few recent larval decapod studies have applied phylogenetic methods (Clark & Webber 1991; Marques & Pohle 1995, 1998). Polarity was determined by Clark & Webber (1991) and Marques & Pohle (1995) using the assumption that evolution had proceeded by oligomerization, i.e., the loss and reduction of segments and setation elements. However, Marques & Pohle (1998), using out-group analysis, tested the applicability of the oligomerization method of polarity determination, and found it unjustified in that it did not provide the most parsimonious explanation of the data set. In contrast, they believed their results demonstrated that only the use of out-group comparison to polarize character transformations would produce the most parsimonious hypotheses while allowing the researcher to recognize possible addition events. However, it should be noted that Koenemann & Schram (2002) now argue that maximum parsimony is a biased method to analysze development sequence data.
Marques & Pohle’s (1998) results not withstanding, and the out-group method’s common usage in assessing polarity in adult phylogenetic investigations (Meier, 1997), some researchers consider that the out-group method has disadvantages when applied to ontogenetic sequences (Nelson, 1978, 1985). For example, the out-group method of determining polarity assumes a knowledge of higher relationships (Nelson 1973a, b, 1978), and also is concerned, not so much with character polarity, but with connotations of ancestry (Williams et al. 1990). The selection of one or more taxa as out-groups theoretically establishes a hypothetical ancestor with all primitive characters, and character-state distributions are summarized at the out-group node (Maddison et al. 1984; Bryant 1997). Since our concern is for attributes pointing to possible carcinization events, our application of cladistic methods is once again somewhat unconventional (cf. McLaughlin & Lemaitre 1997: 96), thus determination of ancestry is not our specific aim.
An alternative to the out-group method is the ontogenetic polarity criterion, or Nelson’s Rule (Nelson, 1978). As elucidated by Bryant (1997), the distribution of character states through the ontogenies of members of the ingroup is used to infer the expected character states at the ingroup node. Specifically, Nelson’s criterion considers that the transformation of an ontogenetic character observed to be more general to one observed to be less general, represents a transformation from primitive to advanced. Nelson, however, did not interpret general and common as equivalents, as some of his subsequent critics have (Kluge & Strauss 1985; Kluge 1988; Kraus 1988). Patterson (1994, 1996) paraphrased Nelson’s (1978) criterion as “absence is more general than presence”. Unfortunately, generality is often in the eye of the beholder, and can be interpreted quite differently by different investigators (e.g., Mabee 1989, 1996; Patterson 1996). Similarly, absence does not universally precede presence (e.g., Fong et al. 1995).
In the opinion of de Queiroz (1985) “... characters do not transform in ontogeny; ontogenetic transformations are themselves the characters.” Importantly, Wheeler (1990) has made a justifiable distinction between character adjacency and character polarity. In an ontogenetic sequence, transformations in character states can be observed (character state adjacency), but such transformations do not provide information about polarity per se. Tests on the out-group criterion and “Nelson’s Rule” (Wheeler 1990; Meier 1997) have shown that both methods give approximately equally parsimonious results. Bryant (1992), in comparing the two methods, indicated that both require monophyly of the study group and comparison of equivalent sets of ontogenetic stages. Additionally, according to Bryant, the use of out-group analysis necessitates monophyly of a more inclusive group, or close relationship between the study group and particular out-groups, as well as an adequate survey of the distribution of character states among out-groups. Contrarily, Nelson’s Rule, as interpreted by Bryant, requires the retention of plesiomorphic states in the ontogenies of descendants. In the present study of megalopal and early juvenile development, available evidence suggests that all of the above criteria for the application of Nelson’s Rule have been met. However, as pointed out by Christofferson (1995), ontogenetic polarities of instantaneous characters, because of heterochrony, do not necessarily coincide with phylogenetic polarities of ontogeny. From our data, we have definitive evidence on character state adjacency. By applying Nelson’s Rule, we have, for some characters, been able to postulate polarities. However, in other character transformations, this method of polarity determination does not seem applicable. For example, there is no ontogenetic evidence to suggest that certain losses and/or transformations occur in ordered, stepwise manners among taxa. Specifically in coding pleopod reduction or loss and uropod loss or transformation, we have used the “intermediate method” of Wilkinson (1992, 1995), which hypothesizes that if a character state is intermediate in form, size or number between two other character states it is considered phylogenetically intermediate between those two other states. The rationales that have gone into our decision-making processes are presented for the characters used in our analyses.