The impetus for this special volume of the Anatomical Record grew out of an excited conversation with one of my good coworkers who has spent his entire professional life studying primate—especially New World monkey—morphology, enlightened by personal experience dissecting primates, collecting their fossils, chasing monkeys in the forest, and performing experiments in the lab. My rhetorical question was: “Is morphology dead?” His answer, paraphrasing, was: “It's never been more alive. Look at all the students going to the museum, coding characters.” To which I replied: “It's not about making a Grey's Anatomy of primate morphology. What more have we learned about the animals since all that began?”
The articles assembled here, which focus on platyrrhines, one of the major primate radiations, will hopefully demonstrate that primate morphology is anything but dead, nor is it in abeyance in the era of genomics. Primate morphology is more vibrant than ever, as it should be. We have refined the questions, deployed improved methods, are mastering advanced technologies that probe more deeply into the body and its workings, and continue to discover new fossils that reveal unexpected morphologies which expand the anatomical boundaries circumscribed by extant forms. However, as Simpson warned, it is the relationships among phenomena, the hardest thing to understand and the most important, that remains at risk when investigative agendas are very particularized and synthesis does not follow basic research. At that nodal point, where observation pivots toward meaning, the conversations we have behind the printed word indicates one whole version of morphological inquiry evinces controversy and uncertainty.
I refer to the morphology behind systematics, more specifically to the anatomical basis and methods of phylogeny reconstruction. One needs neither footnotes nor references to validate what is obvious in the pages of the leading journals that publish on primate evolution. Put bluntly: morphological articles dealing with phylogenetics are overwhelmingly biased toward studies that seem to use traits as if they are the parts of inanimate systems, without any reference to a feature's organic life and history. It is an ironic intellectual twist. The obsolete “descriptive anatomy” on which old-school systematics was built, and which deserved a much needed shot of fresh blood and thinking, has morphed into a nondescriptive anatomy beholden to “character states” that, by intention, are normally construed to be blind to hypotheses, suppositions, and assumptions. The result is a defleshed, binary anatomy. In adopting it, this new school of systematics atomizes and transmographies structural complexity in an effort to satisfy the twin goals of minimizing subjectivity and maximizing sample size, that is, the number of traits used and the number of taxa consulted, en route to producing machine-based cladograms. Is this a good thing? Some say, Yes; others say, No.
Evolutionary morphology, the theme of this volume, is neither a new construct nor a new label (e.g., Szalay, 2000). I use the term here simply as an expression meant to avoid the artificiality of separating the functional properties of characters from their phylogenetic properties. Both aspects are intrinsic to morphology although studying either facet often requires different tools and approaches. However, my contention is that there is more explanatory power in a synthetic morphology where form is brought together with function and phylogeny conceptually. I try to demonstrate this while introducing and contextualizing the articles delivered in the following pages, pointing out where an evolutionary or functional morphological perspective has been important in deciphering and explaining key points in platyrrhine phylogeny and history, as elucidated by anatomy. This approach is a break with the old schools of taxonomy, when nonfunctional paradigms dominated platyrrhine systematics. It is also the antipode of algorithmic, new-school systematics, an explicitly nonfunctionalist approach to phylogeny reconstruction. In my view, no other group of living primates illustrates better the benefits of a synthetic evolutionary morphology approach to systematics than the platyrrhines.
Following Szalay (e.g., 2000) and others, one can also make a distinction between evolutionary morphology and functional morphology as it is now practiced in that the former is inherently concerned with transformation. Functional morphology is surely steeped in evolution, as the explanations we have for “how things work” can be related to selective pressure and differential survival under ecological circumstances. Thus, folivory is seen as adaptive when one lives in a world where leaves are potential food and theoretically beneficial to survival. However, functional morphology does not address how adaptations like folivory arose. This represents a distinct set of questions and an approach that differs from the biomechanical. Tracing the origins of a trait or complex requires phylogenetic thinking, but it would be little more than a sterile exercise if it was not infused with possible explanations of “why” it happened. Evolutionary morphology, as the articles presented herein often show, gets us closer to establishing that crucial relationship between How and Why, which Simpson (1944) would surely have seen as part of the grand pattern and great process behind platyrrhine evolution.
Concerning the articulation of morphology and systematics, to be clear at the outset: I do not wholly reject computer-driven analyses of morphology designed to probe cladistics on philosophical or theoretical grounds. Phylogeny reconstruction is a difficult scientific challenge and all helpful approaches are welcome. However, now that we are about 20 years into the process of applying them to platyrrhines, there is ample empirical evidence to allow an objective assessment of these studies, most of which have used the parsimony method (i.e., PAUP; Swofford, 2002). This leads me to question how well they have been implemented as well as their capacity to perform as intended (see Matthews and Rosenberger, 2008; Rosenberger, 2010b). I also wonder how to evaluate the many inconsistencies they turn out (see below), and if they actually live up to the critical theoretical requirement that they produce explicitly (pure) cladistic results. Do the results of these studies justify the underlying proposition that we should assume as little as possible about the organization and evolution of anatomical characters so as not to contaminate the exercise by making supposedly unnecessary a priori postulates?
Such a viewpoint seems merely to shift the menu of complex scientific assumptions—all of science involves assumptions—from the input stage to the analytical. Because of it, hard earned information, hypotheses and causal explanations concerning functional morphology, adaptation and evolutionary history tend to be rejected as valuable evidence. They rarely inform the input routine, character selection and coding, and seem to hold little sway when they challenge output in the form of an alternative cladistic hypothesis. The atomization of morphology required by the algorithmic approach strips morphology of its effective and historical patterns, leading to an objectification which may have an effect opposite of what is intended—fabricating more noise. For in any orderly structural system, the information content of each unit is proportional to the size of the unit within the system and the level of its integration with other parts. Is there any doubt a primate's 10 toenails will say less about history than a whole primate foot functionally interpreted? Or, that the laptop's many keys say less about the manufacturer where it originated than the one motherboard to which each is connected? Ultimately, I am doubtful that some of the logical bases and expectations of the algorithmic agenda has been proven true, that more small-bore observations are commensurate with higher quality research, more accurate results, and the virtue of real objectivity.
In posing this contrast between approaches to phylogenetics, the platyrrhines serve as an object lesson in the value of integrating cladistics with an evolutionary, functional morphology approach to systematics. These themes speak to some of the relationships Simpson (1944) feared were falling between the cracks. More specifically, I take the view that the platyrrhines illustrate how character analysis can contribute to a holistic assessment of evolutionary history that accounts for phylogeny and adaptation within the same theoretical and methodological framework. One of the chief aims of this article is to highlight examples demonstrating this point based on the original studies presented in this volume.
Another aim of this article is to introduce a new classification of the platyrrhines that is designed to better reflect advances made recently, since the first morphology- and cladistics-based overhauls reshaped our thinking in the 1970s and 1980s. I draw particular attention to this point. The need for a revised classification is not based now on new phylogenetic hypotheses. Rather, it reflects major increases in our awareness of taxonomic diversity. There are more than three times as many fossil genera known today than in 1977, for example, and nearly three times as many fossil genera as there are living genera. Up until that point, classifications necessarily had to be anchored in information based on what animals exist today. However, new panoramas of platyrrhine biodiversity are being brought to light. An example. For one of the major modern NWM clades, the Sakis and Uakaris (Pithecia Chiropotes, and Cacajao), the addition of fossils has totally altered our picture of biodiversity and the ecological role this group has played in platyrrhine evolution (Rosenberger, 2002; Rosenberger et al., 2009). Once seen as a backwater ensemble of limited diversity and a nonfactor in terms of broader classification schemes, the pitheciins now appears to be a remnant of a central ecophylogenetic force in the history of the radiation. Like others who have recently advocated for the position (see below), I now regard this adaptive array as a separate family, Pitheciidae. However, this classification also comes with a caveat, for it can only be provisional. With many more taxa now in need of detailed study, the phylogenetic interrelationships of many fossil platyrrhines are still not well understood.
Structurally, the article is presented in three sections. The first provides a brief background to platyrrhine taxonomy and systematics. The second introduces the volume's papers in a discussion of their new contributions to platyrrhine evolution in the context of an evolutionary morphology perspective. The third section is a targeted summary and critique of the algorithmic approach as it has been applied in morphology-based studies of platyrrhine cladistics.
PART I. BACKGROUND TO PLATYRRHINE SYSTEMATICS
- Top of page
- PART I. BACKGROUND TO PLATYRRHINE SYSTEMATICS
- PART II. EVOLUTIONARY MORPHOLOGY AND PLATYRRHINE SYSTEMATICS
- PART III. EXPERIMENTS IN PARSIMONY
- LITERATURE CITED
Rosenberger (2002) argued that 1977 dated a paradigm shift in platyrrhinology. With the publication of Hershkovitz's (1977) massive volume on callitrichines (Figs. 1,2), which also included many companion studies of anatomical systems and their evolution among platyrrhines as a whole, the nonfunctional (descriptive) anatomical perspective which dominated taxonomic thought for decades came to a close, along with the Scala Naturae-like (gradistic) model of evolution. During the period prior, little progress was made either in classification or phylogenetics since about the 1920s, when scholars such as Pocock (1917, 1920, 1925) and Gregory (1922)—evolutionary morphologists in approach—made important contributions identifying, classifying and explaining “natural groups.” In his own approach, Hershkovitz used static keys to organize the classification system, rather than the monophyly principle ushered in via Hennig (1966) and cladistics.
Figure 1. An ecophylogenetic model of platyrrhine evolution based on the living forms (modified from Rosenberger, 2000). The major axes of differentiation, body size, food, and locomotion are emphasized. See Table 1 for a more specific taxonomic breakdown and a classification that includes fossils.
Download figure to PowerPoint
To achieve monophyletic groupings required dynamic analyses of characters, predicated on hypotheses of homology and polarity. In a philosophical shift, systematics embraced transformational thinking, which was consistent with functional morphology, though then, as now, applied rarely in studies of primate systematics. It became important to explain how (in what way) and why features might have changed. These new approaches and views were promoted, in different ways, in cladistic studies of platyrrhines by Ford (e.g., 1980, 1986) and Rosenberger (e.g., 1977, 1979b, 1980). The initial approach advocated by Ford was algorithmic (using Wagner tree parsimony). Rosenberger used character analysis (see Hecht and Edwards, 1977) infused with “adaptational analysis” (see Bock, 1977; Szalay, 1977, 1981; Delson et al., 1977; Szalay and Bock, 1991; Rosenberger, 1992). Homologies and polarities were imputed and tested by the comparative method (i.e., phenetics, anatomical and behavioral correlation, in-group and out-group commonality, temporal precedence, development, functional-transformational models of adaptation and behavior, etc.). I also weighed characters according to my level of confidence in character analysis decisions, emphasizing the uniqueness of the structures involved and their adaptive roles in the biological gestalt of the taxon. A central methodological concept shared by Ford and Rosenberger was the reconstruction of hypothetical ancestral morphotypes, along with behavioral and adaptive inferences that could be attributed to these hypotheses. Wood and Harrison (2011) have recently re-emphasized the importance of inferring morphotypes as part of the phylogeny reconstruction process.
Classifications illustrate the differences between these pre- and post-Hennigian approaches (Table 1). Hershkovitz (1977) used a large number of suprageneric categories in his scheme, including five family-level (family and subfamily) units devoted to eight genera of fossils alone. A similarly large number of higher taxa were used for the living genera. A classification as split as this lacks phylogenetic and adaptive coherence and cannot compliment diversity. The first example of the system I have followed is a composite of formal classifications (e.g., Rosenberger, 1981, 1992) and subsequent taxonomic additions based on new fossils and several topical studies. It is a marked contrast to the Hershkovitz model, organized around a monophyletic two-family system that tiers other monophyletic groups at lower ranks. The increased number of fossils added to the Platyrrhini during this period is important, for accommodating a large number of diverse new taxa would inevitably stress any system based initially on organizing the living forms, as this one was. However, what is most striking is the multiplication of genera related, or potentially related, to the pitheciins Pithecia, Chiropotes, and Cacajao. These three were often collected into a separate subfamily since the middle 1800s (Rosenberger, 1981), but with revised cladistic thinking and the addition of fossils, this group has now swollen to about 14 genera. As noted previously (Rosenberger, 2002), such dramatic changes in knowledge has important implications for considering the role of this clade in the ecological history of platyrrhines.
Table 1. A comparison of platyrrhine classifications
|Hershkovitz (1977)||Rosenberger (1981– 2002)||Rosenberger (present)|
|Family Callitrichidae||Family Cebidae||Family Cebidae|
| Callithrix|| Subfamily Cebinae|| Subfamily Cebinae|
| Cebuella|| Cebus|| Cebus|
| Saguinus|| Saimiri|| Saimiri|
| Leontopithecus|| †Neosaimiri|| †Neosaimiri|
|Family Callimiconidae|| †Laventiana|| †Laventiana|
| Callimico|| †Dolichocebus|| †Dolichocebus|
|Family Homunculidae|| Subfamily Callitrichinae|| †Killikaike|
| †Homuculus|| Tribe Callimiconini|| †Acrecebus|
| †Dolichocebus|| Callimico|| Subfamily Callitrichinae|
|Family Cebidae|| †Mohanamico|| Tribe Callimiconini|
| Subfamily Cebinae|| Tribe Sagunini|| Callimico|
| Cebus|| Saguinus|| †Mohanamico|
| Subfamily Saimiriinae|| Tribe Callitrichini|| Tribe Saguinini|
| Saimiri|| Callithrix|| Saguinus|
| †Neosaimiri|| Cebuella|| Tribe Callitrichini|
| Subfamily Aotinae|| Leontopithecus|| Callithrix|
| Aotus|| Subfamily Callitrichinae inc. sed.|| Cebuella|
| Subfamily Callicebinae|| †Micodon|| Leontopithecus|
| Callicebus|| †Patasola|| Subfamily Callitrichinae inc. sed.|
| Subfamily Atelinae||Family Cebidae inc. sed.|| †Micodon|
| Ateles|| †Branisella|| †Patasola|
| Brachyteles|| †Szalatavus||Family Cebidae inc. sed.|
| Lagothrix|| †Chilecebus|| †Branisella|
| Subfamily Alouattinae||Family Atelidae|| †Szalatavus|
| Alouatta|| Subfamily Atelinae|| †Chilecebus|
| Subfamily Pitheciinae|| Tribe Atelini||Family Atelidae|
| Pithecia|| Ateles|| Subfamily Atelinae|
| Chiropotes|| Brachyteles|| Tribe Atelini|
| Cacajao|| Lagothrix|| Ateles|
| Subfamily Tremacebinae|| †Caipora|| Brachyteles|
| †Tremacebus|| Tribe Alouattini|| Lagothrix|
| Subfamily Stirtoniinae|| Alouatta|| †Caipora|
| †Stirtonia|| †Paralouatta|| Tribe Alouattini|
| Subfamily Cebupitheciinae|| †Protopithecus|| Alouatta|
| †Cebupithecia|| Subamily Pitheciinae|| †Stirtonia|
|Family Xenotrichidae|| Tribe Pitheciini|| †Paralouatta|
| †Xenothrix|| Pithecia|| †Protopithecus|
| Chropotes|| †Solimoea|
|Suborder inc. sed.|| Cacajao||Family Pitheciidae|
|Family Branisellidae|| †Soriacebus|| Subamily Pitheciinae|
| †Branisella|| †Proteropithecia|| Tribe Pitheciini|
| †Nuciruptor|| Pithecia|
| †Cebupithecia|| Chropotes|
| Subfamily Pitheciinae inc. sed.|| Cacajao|
| †Lagonimico|| †Proteropithecia|
| †Carlocebus|| †Nuciruptor|
| †Antillothrix|| †Cebupithecia|
| Tribe Homunculini|| Tribe Soriacebinae|
| †Homunculus|| †Soriacebus|
| Aotus (incl. A. dindensis)|| †Mazzonicebus|
| †Tremacebus|| Subfamily Homunculinae|
| Callicebus|| †Homunculus|
| †Xenothrix|| Aotus (incl. A. dindensis)|
| || †Tremacebus|
| || Callicebus|
| || †Miocallicebus|
| || †Xenothrix|
| || †Antillothrix|
| || †Insulacebus|
| || Subfamily Homunculinae inc. sed.|
| || †Carlocebus|
| ||Family Pitheciidae inc. sed.|
| || †Lagonimico|
The second Rosenberger classification in Table 1 is a revision that accommodates the exceptional diversity of the “pitheciid” group by according it full family rank. Recognizing this third family is consistent with the suggestions of molecular systematists who have, since the 1990s, promoted a three-family classification using the same taxonomic concepts as I use here but with one significant difference in content (see Schneider and Rosenberger, 1996). Much of the rest, however, maintains the same structure, although newly discovered fossils have been added. The pitheciids continue to be a challenge systematically, and this new model will undoubtedly benefit from further adjustments as the internal affinities of larger groups (e.g., Tribe Pitheciini) become better known, or when intersubfamily relationships are better understood. For atelids, I have elected to use only one subfamily and two tribes as the terminology is well established and there appears to be no benefit to elevating the latter and adding additional tiers below at this time. As is evident, my philosophy about classification is that it should be consistent with phylogeny, first of all—and, hopefully, with an ecophylogenetic model—but that it also should be balanced as a taxonomic scheme relative to current concepts applied elsewhere. It should not in and of itself attempt to reflect overwhelming adaptive departures detached from a cladistic hypothesis by inflating ranks, as with the separation of the extinct Jamaican genus Xenothrix at the family level, which has been advocated (e.g., Hershkovitz, 1970, 1977).
It is pertinent in this regard to comment briefly on other schemes that have recently been published with regard to two taxonomic levels. Following several decades of taxonomic stability, we have entered a period of “taxonomic inflation” (Issac et al., 2004), wherein populations are being accorded full species status largely because new definitions or species concepts are being applied, less so because new knowledge has been gained. Issac et al. note that this phenomenon is more egregiously evident in classifications of New World monkeys than other primates. A similar effect is also evident at the genus and family levels. At least three additional genera of living NWM are currently being recognized beyond the 16 identified by Hershkovitz (1977) and Napier (1976), the cornerstones of modern platyrrhine classification. 1) Van Roosmalen and van Roosmalen (2003) differentially diagnosed Callibella (relative to Cebuella) based on its slightly larger size and coat color and pattern difference, a methodology that was universally rejected for most of the 20th century. As I interpret them, multivariate analyses of craniomandibular morphology also barely separates the three known specimens allocated to this form from Cebuella pygmaea (Aguiar and Lacher, 2003), nor do univariate and multivariate analyses of the postcranium (Ford and Davis, 2009). 2) Rylands et al. (2000) elevated Amazonian marmosets to the genus level (Mico) essentially to avoid purported paraphyly of the classical concept of Callithrix, but there is no morphological support for this view (see Hershkovitz, 1977). 3) Groves (2001) advanced the notion that Lagothrix flavicauda should be elevated to the rank of genus (named Oreonax), but this was shown to be based on a flawed parsimony analysis of poorly selected cranial traits (Matthews and Rosenberger, 2008). At a higher level, this inflationary trend has emboldened Rylands and Mittermeier (2009) to recognize five (!) platyrrhine families, with one designed exclusively for genus Aotus, and without considering fossils at all.
These taxonomic moves undermine the stability of platyrrhine classification that has been achieved over the last 30 years based on knowledge of morphology, adaptation, molecules, and a consistent interpretation of phylogenetics. Various recent articles review the cladistic models (e.g., Schneider et al., 2001; Schrago, 2007; Osterholtz et al., 2009; Wildman, 2009; Kay et al., 2008; Rosenberger et al., 2009). In my view, there remains one interesting and vexatious problem concerning the affinities of Aotus, for which morphology and molecules do not align (e.g., Rosenberger and Tejedor, in press). One school places Aotus among pitheciids and the other among cebids sensu Rosenberger (1981). A few other discrepancies exist in branching sequences when comparing morphological and molecular dendrograms, each important but less vital than the Aotus paradox. Thus, the revised, non-Hershkovitzian picture of platyrrhine systematics that began to emerge in the 1980s based on morphology, began to be corroborated by genetics in the 1990s. And the power of these adjustments has been demonstrated by the capacity of the new system to accommodate many new fossils without requiring still another major overhaul.
PART III. EXPERIMENTS IN PARSIMONY
- Top of page
- PART I. BACKGROUND TO PLATYRRHINE SYSTEMATICS
- PART II. EVOLUTIONARY MORPHOLOGY AND PLATYRRHINE SYSTEMATICS
- PART III. EXPERIMENTS IN PARSIMONY
- LITERATURE CITED
Kay and coworkers (e.g., Kay, 1990, 1994; Kay and Meldrum, 1997; Kay and Cozzuol, 2006; Kay et al., 2008) have been the most ardent users of parsimony algorithms (PAUP; e.g., Swofford, 2002) as a method to reconstruct the cladistics of NWM based on morphology. The effort makes their work an indispensable resource for evaluating the power of this approach in a controlled situation—one lab, using a consistent and expanding dataset based mostly on craniodental features, applied to a single adaptive radiation. It is noteworthy that this approach led Kay et al. (2008) to suggest that nearly all of the early middle Miocene platyrrhines from Patagonia, which comprises a very important collection of fossils key to understanding platyrrhine evolution, belong to a pre-crown radiation of NWM. This idea addresses central questions about the tempo and mode of platyrrhine evolution: How old are the extant lineages and how much have they changed over time? The model proposed by Kay et al. has been presented as an alternative to the Long Lineage Hypothesis (e.g., Rosenberger, 1979a; 2010b; Delson and Rosenberger, 1984), which holds that some of these same Patagonian forms known by interpretable material are actually early representatives of enduring extant clades, in a broad sense representing the ancestral groups from which platyrrhines evolved into the modern guilds as the Amazonian fauna assembled several millions of years later (Rosenberger et al., 2009). More accurately, it addresses a temporal question: When do these long-term lineages, documented by molecular systematists as well (e.g., Opazo et al., 2006; Schrago, 2006; Osterholtz, 2007; Schneider et al., 2001, Wildman, 2009), first appear in the fossil record?
Figure 7 presents summary cladograms of this body of work (e.g., Kay, 1990, 1994; Kay and Meldrum, 1997; Kay et al., 2008), chronologically. The dendrograms appear as published (a, b, and c) or as extracts taken from larger trees (d and e). They were selected to highlight the interpretations of Callicebus, one of the more interesting genera whose affinities have historically been problematic (see Rosenberger, 1981; Ford, 1986; Rosenberger and Tejedor, in press). The trees come from four different studies using an expanding dataset involving 15 to 199 craniodental characters, and as many as 30 primate genera. For one study, I reproduce two different dendrograms, one (e) based purely on morphology and another (d) that use a “molecular backbone” approach to reconstruct morphology, which essentially involves mapping morphological character states onto the topology of an existing molecular tree. The purpose of this particular application was to generate a list of traits useful for placing fossils within a presumably robust cladogram.
Figure 7. Five whole or partial maximum parsimony cladograms from several studies produced by Kay and coworkers (see text). (a) Kay (1990): 117 dental characters, 19 taxa [genera], Consistency index (CI) = 0.45. (b) Kay (1994) Lagonimico tree: 18 characters (15 dental), 9 taxa [genera]; CI = 0.65. (c) Kay and Meldrum (1997) Patasola “preferred” cladogram, 55 dental characters, 11 taxa [10 genera], CI = 0.54. Retention index (RI) = 0.55. (d) Kay et al. craniodental tree constrained by molecular backbone tree (Kay et al., 2008, Fig. 20), CI = 0.34. RI = 0.53, 199 characters craniodental. 30 taxa [genera]. (e) Kay et al. craniodental tree unconstrained by molecular backbone (Kay et al., 2008, Fig. 21); CI and RI not provided. Compare (a) and (e) for a fuller picture of branch-by-branch discordance of strictly morphology-based parsimony results.
Download figure to PowerPoint
The five trees reveal five different linkages for Callicebus. The first example (Kay, 1990) should perhaps be treated as an anomaly—a starter tree?—for the outstanding cladistic elements that give it shape have essentially been abandoned by Kay in all subsequent articles (e.g., compare with Fig. 7d,e), save for the recognition (but not the interrelationships) of three monophyletic groups consisting of callitrichines, pitheciins and atelines. For all intents and purposes, when stripped of the wayward interpretations of nonspecialists seeking to force fit classifications, these three clusters of genera have been understood to be coherent units for several generations (Rosenberger, 1981). However, in 1990, Kay interpreted Callicebus as a basal member outside the three clades, in a position below Cebus at the beginnings of the crown radiation. There is little question today, from morphology and molecules, that both Callicebus and Cebus are nested well within the Platyrrhini (e.g., Schneider and Rosenberger, 1996; Opazo et al., 2006; Schrago, 2006; Osterholtz, 2007; Schneider et al., 2001, Wildman, 2009).
However, in each of the four other morphology studies in the series Callicebus obtains different positions as well, sharing an immediate node with Pithecia (b), Cebus (c) and Aotus (e), or shown as the sister-group to pitheciins (d). There are many interesting aspects to these results. (1) Aotus is nearly always in the mix, never being more than one step removed from the Callicebus twig or node. Aotus and Callicebus have long been thought to be closely related (see Rosenberger, 1981). (2) In only one “unconstrained” case (b) is Callicebus tied directly to Pithecia, but not when the other pitheciins, Chiropotes and Cacajao, are included in the sample (d and e). Now, the close cladistic association of Callicebus and pitheciins is unanimously accepted by morphologists and molecular systematists (e.g., Rosenberger, 1992; Schneider et al., 2001; Opazo et al., 2006; Schrago, 2006; Osterholtz, 2007; Rosenberger et al., 2009; Wildman, 2009). (3) In two post-1990 iterations, Callicebus, Aotus, and Cebus form a exclusive monophyletic group (c and e). This contradicts the powerful cladistic linkage between Cebus and Saimiri, which is extended to encompass callitrichines as well by essentially all other studies (ibid.). (4) In none of the purely morphological iterations (a–c, e) is Aotus more closely related to cebines and callitrichines than to any atelids. The former hypothesis has received nearly universal corroboration from molecular systematists since the middle 1990s (e.g., Schneider et al., 2001; Opazo et al., 2006; Schrago, 2006; Osterholtz, 2007; Wildman, 2009), although I continue to be skeptical of this notion based on morphology, and also because the molecular results seem soft. In a survey of them, Aotus is typically linked to cebids via polytomies, and its node receives uncommonly low and variable statistical support (Rosenberger and Tejedor, in press).
These nonrepeated results concerning the interrelationships of Callicebus are major inconsistencies among the morphology-based parsimony tress. And, close inspection reveals that the discrepancies are not limited to one genus. They also involve Saimiri, Callimico, Leontopithecus, and so forth. While it is beyond the scope of this study to seek a specific explanation for errant results, it seems clear that no particular outcome can be said to be scientifically credible when it is not replicated, especially by the same methods, in one lab, using a consistently similar body of information. Confidence in any one of the trees, or in particular nodes, could have been gained because each of these iterations, coindependent by their very nature thus highly likely to repeat, was slightly or more progressively tweaked by design, for example, with enlarging matrices of characters and/or taxa but analyzed within the same experimental plan. However, this was not the case. The strongest common denominator among them in terms of broadly accepted and probably accurate results is found in the monophyly of callitrichines, morphologically and behaviorally one of the most derived platyrrhine clades—a finding not only expected but essentially required to avoid disqualification.
The final pair of cladograms was isolated from a larger study because it is instructive in another sense. One of the trees (e) is based exclusively on morphology and the other (d) is built from a molecular backbone. Again, prominent differences are seen with respect to Callicebus, Aotus, Cebus, Saimiri, and so forth. For the first time in the series, the morphology (e) suggests Saimiri is not related directly to cebids (d), a conclusion that is strongly contradicted by all other phylogenetic studies since the post-Hershkovitz (1977) paradigm shift. However, another aspect reveals the intrinsically limited cladistic signal of the analysis itself. By mapping the characters onto the molecular backbone (d), Kay et al. (2008) were able to obtain a statistical assessment of the “goodness of fit” between morphological characters and the tree. The Consistency index (CI), which is designed to measure the amount of homoplasy, was 0.34, and the retention index (RI), which is meant to measure the amount of synapomorphy, was 0.53. Since plesiomorphy and homoplasy contain no credible cladistic information by definition, the values obtained here indicate about one-third to one-half of the morphological characters mapped onto the molecular tree can be regarded as cladistically irrelevant noise. At most, only about half the tree can be reliably held, from this particular viewpoint, to be built on nodes supported by homologous derived craniodental characteristics.
While I do not assume the molecular tree has precedence over and is more likely to be more “right” than the morphological tree, the question is—What does this poor fit mean? If the molecules are correct, is this repudiation of morphology, or of the particular characters or methods used in the morphological study? It should also be recalled that two of the prior, purely morphological studies achieved much higher CI values, which is partly a function of having smaller matrices, but they also exhibited discordant branching sequences.
Kay et al.'s (2008) research strategy in this last study (d and e) has been critiqued at length (Rosenberger, 2010b; but see Kay and Fleagle, 2010). Various theoretical questions regarding specific sampling protocols and possible artifacts relating to them have been pointed out. However, the backbone methodology is problematic for other, deeper reason as well. It assumes, for example, a symmetrical correspondence between the patterns of molecular and morphological evolution (e.g., Assis and Rieppel, 2010). Moreover, Assis and Rieppel (2010:4) point out an epistemological dilemma associated with the procedure that is inherent in the presumed primacy of the molecular evidence: “Using mapped, rather that tested, characters as synapomorphies cannot lend support to the inference of monophyly, because such “synapomorphies” are empirically empty: they can never be shown to be wrong.” In other words, adhering to a backbone methodology makes the results and all contingent anatomical hypotheses oblivious to tests using morphology.
Irrespective of this caveat, Kearney and Rieppel (2006) also note that the procedures used in algorithmic analyses of morphology like the ones discussed here are similar to what Hull (1970) described as an “antitheoretical” method of “look, see, code, cluster” (pg. 31). This was Hulls' summary critique of early formulations of numerical taxonomy, a systematic approach that sought a high level of scientific objectivity but wound up being rejected as sound phylogenetic methodology, partly because nothing is theory-free in science. While acknowledging that practitioners vary in how they apply the method, its main contentions and philosophical positions are quite similar to those embraced by advocates of algorithmic cladistics. To paraphrase Hull (1970:31), their purported value lies in: 1) being based on unweighted or equally weighted characteristics, with no a priori weighting; 2) not being biased toward any scientific theory or prior research on the topic, or to ideas applied to character delineation, homologies, and taxonomic clustering.
Several other large points emerge from this assessment of the differing cladograms (Fig. 7). One has been demonstrated many times before in various studies of primates (see references in Rosenberger, 2010b) and other taxa (e.g., Hillis, 1998), including meta-analysis of dozens of morphological studies. The matrix configuration of a parsimony study has a profound influence on results. In other words, when it comes to real world applications, the method may be inherently unstable and unreliable, because no sampling of biodiversity can avoid omitting taxa and lineages that have gone extinct, and the consequences of this loss of information are imponderable but also clearly variable. Thus, we might assume every matrix is a skewed matrix. Wheeler (2005:71) makes this point in another way, which challenges the heart of parsimony as a phylogenetic method grounded in homology, saying: “Given a single cladogram, two features are homologous if their origin can be traced back to a specific transformation of a branch of that [Wheeler's emphasis] cladogram, but the same pair of features may not be homologous on alternative cladograms. Homology is entirely cladogram dependent…”
Wheeler's (2005) extreme formulation is meant to elucidate a concept of homology but it is also another way of saying parsimony analysis does not guarantee that the states attached to nodes are parts of a genetic continuum of traits or a discrete genetic pathway of descent. It only promises to find the shortest tree possible, that is, a tree with the fewest nodes, based on how select characters are distributed among select taxa. In other words, parsimony is an optimization method. It is not a tool that probes information in search of genetic qualities. It works by antiseptically redistributing character states among taxa in a manner deemed most efficient by predetermined rules, which may have nothing to do with morphological evolution or the evolvability of a particular radiation. Along the way, it brands traits as “synapomorphies,” not in the Hennigian sense of shared derived genetic homologies but as common denominators of taxonomic clusters. In reality, these common denominators, which unless homologized and polarized are nothing more than anatomical abstractions, may be primitive, derived, or analogous. As the CI and RI values indicate, the final cladogram is thus supported by an amalgam of cladistic and phenetic information. This is poignantly illustrated by the support values in the morphological cases presented above, and in the poor performance of the craniodental characters when mapped onto the molecular tree.
Does this disqualify parsimony as a tool for phylogeny reconstruction? It should for purists. However, not if we consider parsimony a heuristic device that may bring us a step closer to discovering interrelationships, and accept its many phylogenetic caveats at the same time. It does mean that a parsimony tree cannot be regarded as a true cladogram unless the input characters are stated to be (in theory) homologies, that is, unless it is preceded by character analysis. Why does parsimony work to the extent it does? Perhaps because a minimally adequate number of traits coded at the outset are indeed likely to be homologues. With this kind of “head start,” in many cases even purely phenetic resemblances can lead to a natural, monophyletic grouping of related taxa. However, if the parsimony tree is a phenetic-cladistic hybrid, it follows that only some of its nodes are trustworthy. More research, using alternative means, would be required to decide which ones deserve confidence.
Having seen some of the drawbacks of parsimony studies, in the present context the question now becomes: What role might evolutionary morphology have in improving the design of these studies? What roles can functional morphology play, more broadly, in cladistic studies? There are two operative levels in the process of phylogeny reconstruction, the dataset (taxon and character identification and coding) and the analysis (tree building). Functional morphology provides a testing platform for cladistic hypotheses, at the levels of character and clade. If one assumes adaptation is what drives much of morphological evolution, we can test how sensible cladistic hypotheses are by evaluating adaptive continuity and/or discord among potentially related taxa. For example, the association of Callicebus with pitheciins is one of the newer features of platyrrhine cladistics, but few researchers recognized any overlapping similarities between these groups until it was determined that they shared in common the unusual pattern of tall, narrow incisors, in addition to some subtle details of upper incisor morphology (Rosenberger, 1979b). This synapomorphic clue attained more cladistic power when it was recognized that the Callicebus condition was a good model for an early adaptational “stage” in the evolution of the hyperderived pitheciin incisor-canine battery (Kinzey, 1992; Rosenberger, 1992). At the same time, some of the trenchant differences between Callicebus and pitheciins were spelled out as adaptive departures related to the radically derived seed-eating habits evolved among the latter. This offered a rationale as to why close relatives looked less alike than propinquity of descent might predict, and it provided a sound hypothesis for the polarities of hypothesized changes.
In a much simpler example, the power of the hypothesis that Pithecia, Chiropotes, and Cacajao are monophyletic derives not only because one can generate a long list of shared anatomical novelties, but because we understand their anterior teeth as being functionally integrated with the cheek teeth—equally odd in shape—as an adaptive system geared to seed harvesting and processing. Thus, we attribute high “phyletic valence” to the characters. This set of features and analytical concepts has been used to place fossils with very incomplete anatomies, such as the early middle Miocene form Proteopithecia from Patagonia (Kay et al., 1998). In fact, the confidence placed in these character hypotheses is so high that Proteopithecia is the only Patagonian platyrrhine recognized by Kay et al. (2008) as belonging to the crown NWM clade. The other taxa, represented by characters without clear-cut functional explanations if any at all, he and his coworkers designate as stem platyrrhines. Interestingly, at the same time, they deny that another Patagonian fossil, Soriacebus, is also a pitheciine (see above) though it, too, exhibits a number of the same high-weight pitheciin traits.
Obviously, for cases involving taxa characterized by extreme morphologies, functionally interpretable, one might say homologies and polarities are simple enough that conventional character analysis can do the job and algorithms are simply overkill. However, what about the many other cases? One argument invoked to support the use of algorithms seems to be that parallelism and convergence is both so subtle and so common (see Kay and Fleagle, 2010) that the matter should not be left to the alleged subjectivity of conventional morphological methods. While the concern for parallelism is as old as Darwinian systematics itself, the entrenched contemporary view that parallelism is rampant can be attributed to meta-analyses of datasets which have attempted to empirically demonstrate the frequency of parallelism by examining the CI and RI statistics in parsimony projects (e.g., Williams, 2007), as well as to the CI and RI values in individual studies such as the series under discussion here. However, these comparative reports are fraught with difficulties, apart from the fact, recognized by Williams and others, that the measures are not independent of matrix size and cannot thus be compared easily across studies. More serious objections are that high levels of homoplasy appear to be 1) predicated on characters that are themselves nonindependent, and 2) involve traits that may not actually be suitable for higher phylogeny assessments. This situation arises directly from the decision to minimize assumptions about morphology when crafting data matrices. For example, the practice of coding serial homologues, for instance, hypoconulids on first, second and third molars, as three separate characters is one such example appearing in the platyrrhine literature (e.g., Rosenberger, 2010b) that violates several precepts: 1) hypoconulids vary at the population levels within some species and genera, making them difficult to code and untrustworthy at higher taxonomic levels and across clades, where they are prone to functional convergence; 2) the theory of homology dictates that such features strung together across adjacent morphological units constitute one feature, not two or three. According to Sereno (2009:626), and others, “character independence and the mutual exclusivity of character states” are the fundamental assumptions behind character state formulations.
The abundance of parallelisms detected by meta-analyses and individual parsimony studies is an issue, for all agree that it can become a significant nuisance factor in phylogenetic studies. As noted above, in the era of the supermatrix—when a single molar can contribute a dozen states or more—we are likely to find high degrees of parallelism because each one is an artificial subdivision of an integrated functional complex under selection. The essential point is that excising traits from their anatomical systems and treating them all separately strips them of their functional and genetic qualities [to Simpson (1944), relationships], making the effort to reconstruct phylogenetic interrelationships more challenging because it also detaches the analytical process from evolutionary theory. The more and more functional units are subdivided, the more information is lost about their genetic properties and their potential interaction with selective forces; the more noise is added to the system; the higher the measure of homoplasy.
In summary, for platyrrhines, morphology-based parsimony studies do not have a good track record. The “simple” clades can be easily retrieved, as they would likely be on purely phenetic grounds, but problematic genera wander about the trees in an inconsistent manner. Near-repeat studies do not replicate. Knowing these results tend to fail the test of repeatability, they must be viewed skeptically. Meaning, some nodes in a large tree are likely to be reasonable hypotheses of monophyly, but others are not. If there is any rule of thumb that can be applied here, it must be that nodes supported by features that make sense functionally in the lives of the animal are the ones deserving our attention. Those supported by volatile, atomized traits of limited independence and functional value, are to be suspect. As a heuristic device, parsimony analyses have a place, but without testing independently the homologies of the characteristics involved a parsimony tree cannot be accepted as a truly cladistic model. However, the under-performance of these studies in the platyrrhine domain should not be taken as evidence that morphology has failed. Or, that molecules are more reliable as phylogenetic evidence. Or, that parallelism is stultifying. These difficulties lie with the methods we use, not with the morphology itself.
- Top of page
- PART I. BACKGROUND TO PLATYRRHINE SYSTEMATICS
- PART II. EVOLUTIONARY MORPHOLOGY AND PLATYRRHINE SYSTEMATICS
- PART III. EXPERIMENTS IN PARSIMONY
- LITERATURE CITED
The diverse set of morphological studies presented here, and fieldwork bearing on morphological questions, continues to expand the body of knowledge concerning the adaptive radiation of NWM. Evolutionary morphological studies of platyrrhine primates now cover a wide range of questions concerning the current ecological and behavioral adaptations of New World monkeys. In both general and specific ways, they add confidence to the models of platyrrhine phylogeny and adaptation that began developing in the 1970s and early 1980s, when platyrrhines first became subject to cladistic studies informed by character analysis. They support the general model of the platyrrhines' current status as an ecophylogenetic array differentiated along the principle ecological axes of body size, diet, and locomotion, and serve to align other factors such as social organization, brain size, perception, cognition, and so forth. Moreover, they generate testable hypotheses about the adaptive shifts that must have occurred as the platyrrhine panorama unfolded.
One reason for their explanatory power is that the thinking behind these articles is closing the perceived gap between branches of morphology that are often held apart from one another, such as functional morphology and biomechanics, allometry, histology, systematics, and phylogeny reconstruction, while also addressing matters of ecology, behavior, and life history that articulate with anatomy. The result is a more expansive evolutionary morphology. Its perspective provides a robust platform for generating and testing hypotheses about platyrrhine evolution, how and why groups and characteristics changed over time, in addition to serving morphology's more conventional raison d'être, to investigate how things are structured, how they operate and how to categorize anatomical diversity. Evolutionary morphology also provides a primary way to rationalize false cladistic signals by exposing parallelisms as independently evolved paradaptations.
The notion of evolutionary morphology ties phylogeny and adaptation together as a matter of method. Many of the hypotheses about the cladistic interrelationships of platyrrhines are made plausible because there are strong adaptive hypotheses behind the characteristics put forth to support the phylogenetic models, with functional morphology proving a deeper understanding of the connections between form and biological role, and because the groups so identified occupy an understandable place in nature. Many of the homologies and polarities proposed as cladistic evidence in the first place were identified as being of potential value because of their evident adaptive significance.
Molecular systematics offers an independent empirical test of morphology-based phylogenetic results, a direct acid test of the hypothesized relationships and an indirect check of the evidence behind them. Overall, there is broad concordance between molecular and evolutionary morphological models, although some question marks—areas ripe for further research—still remain. What do I mean by broad concordance? Four major clades have been identified, and most workers see them as comprising two basic divisions of the radiation. Two of the four are unanimously agreed to be sister-taxa. There is consensus on how the major subclades within these groups sort out. Many of the links between pairs and triplets of genera are agreed on. That the molecules and morphology are also in sync in terms of the time scale inferred for the origins of these clades adds confidence and mutual corroboration of the phylogenetic outlines.
The main morphological method competing with the character analysis approach to phylogeny reconstruction over the past 20 years has been parsimony-based cladistics. These projects suffer from inconsistency, as shown by following the results obtained for the genus Callicebus in a series of related craniodental studies from the same lab. Cladistically, differences in the placement of Callicebus may be as small as one-node shifts or one-branch swaps, but they also get quite large: a cladistic position nested deeply inside versus one at the very edge of crown platyrrhines. In a taxonomic sense, the parsimony studies have aligned Callicebus with at least two different families. A number of other genera (e.g., Saimiri, Cebus, Aotus, Leontopithecus, and Saguinus) are also variably placed, nontrivially, in these assessments.
The NWM parsimony studies have also been shown empirically to have a low cladistic yield at the character level, that is, relatively few true synapomorphies, for the trees tend to exhibit low levels of homology. The reasons for this are not easy to determine, but various critiques of how the method has been used in primate studies, including platyrrhines, have pointed out a preponderance of issues. These could plague any phylogenetic study but they are more acute in rigid machine-based treatments. Among them: disparities in taxonomic samples used; a reliance on population-level traits for higher phylogeny studies; matrices biased by taxon and character choices; excessive use of nonindependent, correlated, and redundant characters; and, especially acute for studies involving fossils, matrices skewed by material limitations.
This relatively poor performance calls out for caution, especially when this approach is used to address, within a constrained taxonomic framework, the phylogenetics of individual, poorly known fossils. Without a complimentary assessment of the evidence by character analysis or evolutionary morphology, which has been shown to have a good track record, these projects are inevitably handicapped by compromised samples: an overabundance of missing biological information and an overatomized anatomy. Such is the case for several middle Miocene fossils of the circum-Amazonian basin. Examples are Lagonimico (said to be a giant tamarin, but more likely a pitheciid) and Solimoea (said to be a primitive atelin but more likely an aloauttin). Similarly, questionable results have been obtained in assessing a whole group of older fossils from Patagonia. Based on a very unevenly preserved information base, and using a molecular backbone approach in an effort to extract cladistically informative anatomical traits from the whole panoply of modern platyrrhine craniodental diversity, all the fossils were determined to be outside the crown clade. In this case, the resemblances between Tremacebus and Aotus, and Soriacebus and pitheciins, of features shown to be powerful ecophylogenetic indicators by evolutionary morphology, were rejected with little or no consideration, even though the trees' homoplasy indices were high.
These examples call into question the trustworthiness of morphology-based parsimony as a tool for cladistics, certainly as the ultimate arbiter. If platyrrhines are taken as an example, it has yet to be shown that parsimony-based methods, which tend to shun attaching evolutionary interpretations of homology, polarity, and adaptation as evidence inputs, have greater resolving power than a morphology which concentrates on fewer characters and is laden with interconnected evolutionary hypotheses. It has yet to be shown that anatomical information purged of its structural-functional context, that is, evidence of its genetic covariation and inheritance, outperforms methods that fuse morphology, phylogeny, and adaptation. Or, that sheer quantity of information, including characters and taxa devoid of evolutionary context, redundancies, or those immaterial to the question at hand because they are autapomorphic for other groups, yields more robust results than character analyses that fuse morphology and adaptation within a properly constructed phylogenetic framework.
As some have noted, the philosophical basis of algorithmic, morphology-based cladistics is akin to the “look, see, code, cluster” foundations of the now defunct school of numerical taxonomy. In other words, there is ample room for improvement. On the other hand, the rich evolutionary morphological tradition that is taking root in studies of the platyrrhines, like those presented herein, suggests progress could be made even by inserting only one additional term to the protocol: look, see, interpret, code, and cluster.