Paranasal sinuses are bony cavities, lined with mucosal epithelium that develop as expansions of various recesses of the cartilaginous nasal capsule. The sinuses “pneumatize” the face and basicranium while maintaining communication with the nasal cavity via small openings or “ostia.” Most nonaquatic placental mammals have at least one, the maxillary sinus, but many have more, including frontal, anterior ethmoid, posterior ethmoid, and sphenoid sinuses. The conservative nature of nasal and paranasal anatomy in mammals makes it phylogenetically informative at higher taxonomic levels. As was noted by Moore (1981, p 240):
The fundamental configuration of the (nasal) fossae is remarkably constant throughout the great majority of mammalian groups, the structure of this part of the skull being largely uninfluenced, apparently, by factors such as total body size, type of diet and mode of life which have had such profound modifying effects upon the morphology of other cranial regions.
Indeed, nasal and paranasal anatomy has been considered to be diagnostic of several primate higher taxa, such as the Haplorhini (Cave, 1967, 1973), Cercopithecoidea (Fleagle and Kay, 1987; Benefit and McCrossin, 1993), Hominoidea (Ward, 1997b), Ponginae (Pongo and Sivapithecus) (Ward and Brown, 1986), and Homininae (African apes and humans) (Cave and Haines, 1940), and often features prominently in phylogenetic arguments (e.g., Ward and Kimbel, 1983; Ward and Pilbeam, 1983; Ward and Brown, 1986; Andrews and Martin, 1987; Begun, 1987; Harrison, 1987; Brown and Ward, 1988; Begun, 1992, 1994, 1995; Benefit and McCrossin, 1995, 1997; Moyà-Solà and Köhler, 1995; Alpagut et al., 1996; Pickford et al., 1997; Rae, 1997; Walker, 1997; Rae et al., 2002). However, until recently, the actual polarity of these characters within Anthropoidea was left ambiguous by our ignorance of the conditions in living platyrrhines and stem catarrhines.
New information reported in recent years has more than doubled the number of anthropoid genera, for which confident assessments of sinus identity can be made, and precipitated new hypotheses of sinus homology (Koppe and Ohkawa, 1999; Rossie, 2005, 2006). As a result, it is likely that the phylogenetic meaning of commonly cited characters such as the ethmoid and frontal sinuses have changed. In this article, the phylogenetic significance of anthropoid paranasal sinus anatomy will be explored in light of this new information.
The methods employed in the present endeavor will be those of “character optimization” or “character mapping.” In other words, paranasal sinus anatomy will not be used to generate phylogenies. Rather, the distribution of sinus characters will be mapped onto several current competing hypotheses of phylogeny, permitting an evaluation of the “fit” of paranasal sinus data to each tree (cf., Asher, 2001). Because there are no published phylogenies of sufficient taxonomic breadth and precision to be used at the level of Anthropoidea, character optimization will first be performed for platyrrhines and catarrhines separately. Then the best fitting of the two sets will be combined for an analysis of all anthropoids.
Reluctance to generate phylogenies based on sinus data alone stems not from a lack of confidence in the utility of these characters, but merely from the belief that such phylogenies serve no purpose. To quote Cartmill (1978, p 109):
No matter what procedure we decide to adopt in phyletic reconstruction, it is not legitimate to choose one system or one part of the body arbitrarily and proceed as if that aspect of the animals under study uniquely holds the key to their relationships.
Like any other character, the phylogenetic significance of sinuses cannot be determined by fiat (MacPhee and Cartmill, 1986); it must be evaluated by assessment of congruence with other characters (Patterson, 1982). Indeed, the purpose of this and previous work by the author is to provide a taxonomically comprehensive and biologically sound argumentation for the identification of putative homologs in paranasal sinus anatomy so that these characters can be incorporated in future phylogenetic analyses of primates in a meaningful fashion. The character optimization pursued here is merely intended as a starting point—a preliminary investigation of the behavior of sinus characters in various phylogenetic scenarios, and a guide to how best to define and delineate sinus characters and states.
MATERIALS AND METHODS
The anatomical observations published elsewhere (Rossie, 2005, 2006) are distilled here into seven sinus characters. Many of the more interesting anatomical discoveries made in these works are not represented here because they constitute autapomorphies of a single genus (e.g., the sphenoidal fossae of Leontopithecus). The product may seem an almost insignificant number of characters, but it is my intention to produce character data of high quality, not quantity, having been subjected to thorough “pre-congruence” homology testing (Rieppel and Kearney, 2002).
To preserve the distinction between my primary hypotheses of homology and my final postcongruence assessment, I will present the former now, before proceeding to the congruence test. It is at this juncture that one must translate their primary hypotheses of homology into a character matrix amenable to the congruence test. To characterize the decisions made in character delineation and coding in a cladistic analysis as nontrivial would be a gross understatement (Pleijel, 1995; Wilkinson, 1995; Hawkins et al., 1997; Strong and Lipscomb, 1999; Wiens, 2001; Rieppel and Kearney, 2002). Decisions made during this process essentially determine the outcome of the analysis (Cartmill, 1982). As Stevens (1984) stressed, one of the most important properties of a homology is the taxonomic level at which it constitutes a synapomorphy. This level is a property of taxonomic distribution that can emerge only once the characters are mapped onto a phylogeny, but it is constrained by the manner in which character states are delineated. If anatomies are excessively atomized, they have no chance of emerging as high-level synapomorphies. For this reason, I see no justification for proliferating states for a given character on the basis of variations in shape or size that have no apparent relationship with the underlying developmental processes responsible for their genesis. Such a practice would not only produce the mistaken impression of synapomorphies at lower taxonomic levels (or worse, only autapomorphies), it would also prevent the detection of the real higher-level synapomorphy.
How can this be avoided? Ideally, the criterion for delimiting character states would be Wagner's (1989a, b) criterion for identifying “individualized” structures or potential homologs, which are, after all, character states. On this basis, most sinuses would be coded as simply present or absent, although the various forms of the cupular sinus would warrant some subdivision, not simply because they differ in their final adult form, but rather because they appear to be produced by a second set of diverticula that, although in direct continuity with the cupular sinus proper, possesses a morphogenetic potential of their own. A similar argument would be made for the anterior ethmoid sinuses, which can be expressed as lamellar cells or as frontal sinuses or both. According to the developmental data described elsewhere (Rossie, 2003, 2006), I would recognize the following fully individualized potential homologs or character states. Brief definitions of the structures are provided here (see Rossie, 2006 for more detailed discussion):
1lateral recess (not recessus lateralis);The recessus lateralis is an internal concavity of the lateral wall of the fetal nasal capsule. It is a product of the pars intermedia being overlapped posteriorly by the pars posterior and anteriorly by the pars anterior. When this recess (or the cartilaginous elements that circumscribe it) ossifies without the occurrence of any secondary pneumatization, the resulting concavity in the middle meatus is called a lateral recess (Maier, 2000; Rossie, 2006). The misuse of this term in the literature lead Rae and Koppe (2003) to advocate its abandonment, but as defined here and elsewhere (Rossie, 2003, 2006), it remains a useful term if applied consistently.
2maxillary sinus;A maxillary sinus is an excavation of the bones surrounding the recessus lateralis produced by secondary pneumatization. It therefore opens into the nasal cavity via an ostium in the middle meatus (Cave, 1967; Maier, 2000; Rossie, 2006).
3cupular recess (not recessus cupularis);Like the lateral recess, the cupular recess is produced by ossification of the cartilaginous recessus cupularis (Rossie, 2006). The latter structure is the posterior portion of the nasal capsule, consisting of the pars posterior or lamina antorbitalis, and is floored by the lamina transversalisposterior (Maier, 2000).
4cupular sinus (a.k.a. sphenoid sinus);The cupular sinus is an excavation of the bones surrounding the recessus cupularis produced by secondary pneumatization. In humans and some apes it pneumatizes mainly the sphenoid bone, and opens into the nasal cavity via ostia in the anterior portion of the sphenoid bone (Cave & Haines, 1940). For this reason it has been termed the sphenoid sinus in most literature. However, in other primates its ostia may be found in the ethmoid or frontal bone simply because these are the membranous bones that happen to overgrow the underlying cartilaginous recess (Rossie, 2006). This illustrates the danger in using the site of the ostium as the ultimate guide to sinus homology, since it is only a proxy for the identification of the cartilaginous recess from which pneumatization emanates. Accordingly, I find it more informative to name the sinus after the true arbiter of homology—the recess.
5cupulofrontal sinus;These are bilateral diverticula that develop from the roof of the cupular sinus and invade the supraorbital portion of the frontal bone in some primates (Hershkovitz, 1977; Rossie, 2006). Their development begins late, after pneumatization of the sphenoid body (Rossie, 2006).
6anterior ethmoid sinuses;The superior portion of the recessus lateralis is termed the recessus frontalis (Maier, 2000). In some primates (e.g, African apes, Aotus, Alouatta, Cebus), sinuses develop from this space and invade mainly the ethmoid bone (Cave & Haines, 1940; Rossie, 2006). These sinuses are separated from the posterior ethmoid sinuses by the base of the first ethmoturbinal, which is the ossification of the anterior edge of the pars posterior (Moss-Salentijn, 1991; Rossie, 2006). In other words, the anterior ethmoid sinuses develop from within the recessus lateralis, whereas the posterior ethmoid sinuses do not.
7frontal sinus;In some primates (e.g., Saguinus, Callicebus, African apes and humans), one or more of the anterior ethmoid sinuses expands superiorly to invade the interorbital and supraorbital portions of the frontal bone (Cave & Haines, 1940; Rossie, 2006). This is called a frontal sinus (Paulli, 1900). This expansion occurs late in postnatal development, after the ethmoid sinuses are nearly full-sized (Cave & Haines, 1940).
8posterior ethmoid sinuses;These small sinuses develop from the spaces between the 1st, 2nd, and 3rd ethmoturbinals on the lateral wall of the pars posterior (Moss-Salentijn, 1991; Rossie, 2006). The homology and individualization of ethmoid sinuses is a complex problem that is discussed in greater detail elsewhere (Rossie, 2006).
These character states are conditions that, if found in two taxa, would be assumed homologous unless refuted by the congruence test. To summarize the “precongruence” hypotheses of homology between platyrrhines and catarrhines that were articulated in the work of Rossie (2006), I assume that these eight features are homologous among all anthropoids in which they are found because no aspect of their ontogeny has been found that contradicts this. Note that this list of homologs will not correspond in number with the character states that are ultimately constructed to portray them (see below), which will include several states that represent the absence of these structures.
Conspicuously absent from this list of characters is the ethmoid labyrinth. The proliferation and hypertrophy of anterior and posterior ethmoid sinuses that produces the catarrhine ethmoid labyrinth is difficult to portray in terms of an individualized character, being a modification of two separate characters (anterior and posterior ethmoid sinuses), and may best be considered an emergent property. However, its distribution within catarrhines is still of interest and will be investigated, particularly because of its occurrence in Proconsul, and possibly Aegyptopithecus (Rossie et al., 2002; Rossie, 2005).
Before one can proceed to the congruence test, these primary hypotheses of homology must be translated into a character matrix, and here one encounters a problem. Some of the characters above are mutually exclusive (e.g., maxillary sinus, lateral recess) and are easily portrayed as alternative states of a character, but others are logically dependant upon one another (e.g., cupular sinus, cupulofrontal sinus). The methods for coding logically dependent characters are a source of considerable debate at present (see Maddison, 1993; Pleijel, 1995; Wilkinson, 1995; Hawkins et al., 1997; Lee and Bryant, 1999; Strong and Lipscomb, 1999), and there appears to be little consensus beyond the conclusion that none actually solve the problem, and different methods are justifiable for different purposes. There are basically two justifiable methods, which will be summarized here.
In the first, dubbed “composite coding” (Wilkinson, 1995), all dependent conditions are coded as states of one compound character (Maddison, 1993). In Maddison's (1993) example of tail presence and tail color, tail colors are coded as part of states of the character “tails” in which “tail absent” is also a state. In the case of the cupular sinus, there would be one character with the states: 0, absent; 1, present; 2, present with cupulofrontal extension. The advantage of this method is that every taxon can be scored for one of these three states, so there will be no missing or inapplicable data. This is not the case in the second method, “reductive coding” (Wilkinson, 1995), in which the character would be divided into two characters:
Cupular sinus: 0, absent: 1, present.
Cupulofrontal sinus: 0, absent; 1, present.
The advantage of such a division is to preserve the transformational independence of the cupulofrontal extensions and the cupular sinus so that each can be a synapomorphy at a different taxonomic level (e.g., Wilkinson, 1995; Lee and Bryant, 1999). In other words, composite coding robs the character data of one of its potential synapomorphies, and reductive coding restores it. In essence, the homology of the cupular sinus itself becomes ambiguous in the composite character (Lee and Bryant, 1999). In Maddison's (1993) example, the presence of a tail is no longer a synapomorphy of a clade if that clade includes taxa with different tail colors because the two will be coded with different states (e.g., tail present and blue, tail present and red). Maddison (1993) has proposed that the full phylogenetic signal could be recovered by either differential weighting or ordering of states in the composite character, but these procedures are not always justifiable (Lee and Bryant, 1999), and in the present case they are not even desirable. For these reasons, I will follow Hawkins et al. (1997) in scoring the independent part (e.g., cupular sinus) as present or absent and the logically dependent conditions (e.g., cupulofrontal extensions) as states of a separate character.
The only drawback to this reductive coding method is that taxa that lack a cupular sinus must also be coded for the presence or absence of a cupulofrontal extension—the “inapplicable data” problem alluded to earlier. There are three options here: they can be coded as “absent,” “?,” or “9.” To code them as “absent” produces a potential synapomorphy that has no evidentiary basis (Maddison, 1993). That is, the absence of cupulofrontal extensions is not observable in the taxa that lack cupular sinuses. Coding them as “?” is problematic because current parsimony algorithms read this as uncertainty, not inapplicability, and will infer a state for them based on the states of other taxa (Maddison, 1993). Maddison (1993) described two problems that can result from such coding. The first problem, assignment of impossible states to ancestral nodes, only occurs when the entity (tail in his case) evolves convergently in two separate clades. This happens because the tailless stem taxa of each clade are coded as “?” for tail color and are reconstructed by PAUP or MacClade according to the distribution of the tail colors in the tailed clades. This results in stem taxa being coded as lacking a tail, but as having a tail color.
However, if the tailless taxa are instead coded as “9” for tail color (e.g., Novacek, 1992) the ancestral nodes would be reconstructed as “9” for tail color, which the investigator knows to mean “inapplicable data.” The only remaining problem is that “9” could become a potential synapomorphy of false clades, but as Maddison notes this will not be the case if the “9”s are confined to paraphyletic or monophyletic groups, in which case they will only falsely bolster true clades. Because the essential problem is intractable with currently available algorithms (Hawkins et al., 1997), this solution seems the lesser of several evils.
According to the principles outlined above, the following are the characters and states of paranasal sinus anatomy that best reflect the primary hypotheses of homology in recent years (Rae, 1999; Rossie, 2003, 2006). In some cases, comments are made here about the coding of certain taxa, particularly fossils, in the analysis.
1Maxillary sinus: present (0), absent (1).This is essentially a convenient abbreviation of “developmental fate of ‘recessus lateralis’: maxillary sinus (0), lateral recess (1),” which would amount to coding the ontogenetic series as advocated by De Queiroz (1985). This holds true for all the sinus characters described herein, but the simpler wording will be used.
2Cupular sinus: present (0), absent (1).As described in Rossie (2006), the commonly used term “sphenoidal sinus” does not adequately express the homology between the structure found in hominoids and in other primates. In some platyrrhines the ostium of this sinus is found within the ethmoid and even parts of the frontal bone instead of the sphenoid. That it is the same sinus is made clear by its ontogeny—it is the product of secondary pneumatization from the “recessus cupularis.” A consistent nomenclature would refer to this as the cupular sinus, just as we refer to the product of secondary pneumatization from the “recessus maxillaris” as the maxillary sinus (Rossie, 2006).
3Cupulofrontal extension: present (0), absent (1).In some platyrrhines, the cupular sinuses give rise to a pair of diverticula that pneumatize the supraorbital portion of the frontal bone (Rossie, 2006). Extensions of the cupular sinus that may be homologs of the cupulofrontal sinuses are reportedly found in gibbons (Cave and Haines, 1940), but the ontogeny of the sinuses in Hylobates remains largely undocumented. The single adult scanned by the author shows some development of a frontal extension, though it is unclear whether it reaches the frontal bone. As a conservative approach, gibbons will not be considered to have these sinuses until it can be convincingly demonstrated.
4Posterior ethmoid sinuses: present (0), absent (1).Aegyptopithecus and Proconsul are coded as having both anterior and posterior ethmoid sinuses because of their position relative to surrounding structures, including the first ethmoturbinal, and their expression as an ethmoid labyrinth (Rossie, 2005). The ethmoid cells in Morotopithecus described by Pilbeam (1969) appear to also include both anterior and posterior ethmoid sinuses based on their position relative to the frontomaxillary suture and lacrimal bone (cf., Cave and Haines, 1940).
5Anterior ethmoid sinuses: present (0), absent (1).These include both the lamellar cells seen in such platyrrhines as Callimico, as well as the frontal sinus because these are simply two expressions of the anterior ethmoid sinuses. They cannot be coded as a compound character because the two are not mutually exclusive. That is, a taxon can have both lamellar cells and a frontal sinus. In accordance with the discussion of dependant character coding above, the possession of any anterior ethmoid sinus is coded as character 5, while the presence of a logically dependant, but transformationally independent, frontal sinus is coded in the subsequent character.
6Frontal sinus: present (0), absent (1).Afropithecus is coded as having a frontal sinus, rather than equivocating over the possibility that it is a cupulofrontal sinus because examination of the fossil reveals no evidence of posterolateral canals such as those found in platyrrhines. The same applies to Dryopithecus (Begun, 1987), Ankarapithecus (Alpagut et al., 1996), and Oreopithecus (Harrison and Rook, 1997). Morotopithecus is coded as having a frontal sinus because some of its ethmoid cells are at the level of the frontomaxillary suture (Pilbeam, 1969).The frontal sinus of platyrrhines will be assumed to be homologous with that found in catarrhines, again because it develops from the frontal recess in both infraorders (Rossie, 2006). Aegyptopithecus was initially interpreted as having frontal extensions of the ethmoid sinuses (Rossie et al., 2002), but more recent high-resolution CT scanning of these specimens by Drs. Alan Walker and Tim Ryan (courtesy of Elwyn Simons) demonstrate that this was an artifact of the low resolution of the original scans (personal communication; see Simons et al., 2007). Accordingly, this taxon is regarded as having only ethmoid sinuses, and not frontal sinuses.
7Ethmoid labyrinth or inflated lateral mass: present (0), absent (1).This is evident in Proconsul heseloni (KNM-RU 7290), Morotopithecus (Pilbeam, 1969), and Aegyptopithecus zeuxis (DPC 2803), but cannot be discerned in any other fossil catarrhines (Rossie, 2005).
Character optimization was performed by comparing the length of competing trees using MacClade version 4.03 (Maddison and Maddison, 2001). The platyrrhine data were traced on trees composed of only the 10 genera for which data was collected. While this precludes analysis of a few lower-level relationships, it is an honest portrayal of the data at hand. In the absence of first-hand data on the six missing genera, nothing can be said about relationships among Alouatta, Lagothrix, Ateles, and Brachyteles, which are viewed differently by various researchers (e.g., Kay, 1990; Rosenberger, 1992). Similarly, relationships among Callimico, Saguinus, and Leontopithecus can be assessed, but the position of Callithrix and Cebuella relative to them cannot. Not all previously proposed platyrrhine phylogenies were evaluated here. Instead, a subset was chosen to represent the alternative placements of taxa currently entertained by systematists (Figs. 1 and 2). The trees of Ford (1986), Horovitz (1999), Schneider et al. (1996), von Dornum and Ruvolo (1999), Rosenberger (1992, 2002), and MacPhee and Horovitz (2002) were pruned of the six genera not included here, and the cladogram of Kay (1990) was modified in an attempt to reflect his more recent view of the position of Callicebus (Kay, 1994; Meldrum and Kay, 1997). Some of these trees include polytomies involving the genus Aotus. In these cases, the polytomies were treated as “hard” so that tree length would not be underestimated (Maddison and Maddison, 1992).
The only fossil catarrhines that were included in the character optimization were those that preserve more than maxillary sinus morphology. This excluded all of the East African small-bodied “apes,” as well as several other phylogenetically significant genera such as Equatorius, Kenyapithecus, Nacholapithecus, Rangwapithecus, and Nyanzapithecus. The conclusions drawn here may require revision when better material is known for these taxa. There are no published phylogenetic hypotheses available that encompass all of the catarrhine taxa included here. As a result, character optimization was performed on a set of three cladograms that are based largely on that of Begun et al. (1997), later modified by Begun (2002) to include Ankarapithecus (Fig. 3a). For the purpose of this baseline phylogeny, cercopithecoids were placed in a conservative position between Aegyptopithecus and Proconsul. Proconsul, Afropithecus, and Morotopithecus are placed as stem hominoids because they lack many of the postcranial synapomorphies of crown hominoids, particularly in the shoulder joint and thorax (Harrison, 1982, 1987, 2002; Ward et al., 1993; Pilbeam, 1996; Gebo et al., 1997; Walker, 1997; Ward, 1997a). Morotopithecus is placed above Afropithecus because of orthograde features in its lumbar vertebra and scapula (Gebo et al., 1997). Aegyptopithecus is placed as the sister taxon to crown catarrhines because of its retention of an entepicondylar foramen and an annular ectotympanic (Fleagle, 1984; Harrison, 1987).
This baseline tree was then modified to reflect the views of Harrison (e.g., 1987, 2002) and Andrews (1992, 1996; Andrews and Bernor, 1999) so that they could also be evaluated (Fig. 3b,c). Unfortunately, this requires certain inferences that may not reflect the precise views of those authors. For example, Ankarapithecus was not discussed by Andrews (1992, 1996) in his most inclusive systematic reviews, but was considered a sister taxon or junior synonym of Sivapithecus elsewhere (Andrews and Tekkaya, 1980; Andrews and Bernor, 1999). The “Andrews” tree also requires inferential placement of Oreopithecus, from which Andrews sensibly abstains from in his recent work. Because he recently referred it to “Hominidae indet” (Andrews, 1992), it seems safe to place it as a stem hominid above his Afropithecini (or Afropithecinae in Andrews, 1996) on the basis of its postcranial anatomy. It is not positioned as a sister to Dryopithecus because he specifically rejects such a scenario (Andrews, 1992). Construction of the “Harrison” tree requires only inference in the placement of Ankarapithecus, because all other taxa are included in either Harrison (2002) or Harrison and Rook (1997). Here again, it is placed as a stem pongine following Begun (2002).
Rather than figure all 28 possible character state trees (seven characters on four trees), the figures in the following discussion will use the tree of Horovitz (1999) as an exemplar unless otherwise specified. This is not intended as an endorsement of this particular tree, though it is one of the only two trees to employ both morphological and molecular data, which may have some philosophical appeal (Eernisse and Kluge, 1993). Of the trees evaluated, four were found to fit the sinus data equally well. These were the trees of von Dornum and Ruvolo (1999), Horovitz (1999), MacPhee and Horovitz (2002), and Schneider et al. (1996). Interestingly, all four of these analyses included at least some molecular data (Fig. 1), while the purely morphological analyses (Fig. 2) were a poorer fit to the sinus data. All four of the most parsimonious trees will all be included in the combined anthropoid analysis below in order to assess the effect of global parsimony (Maddison et al., 1984) on their tree length.
The four trees all feature a cebid clade including callitrichines, Cebus, Saimiri, and Aotus. In three of these, the other taxa comprise a monophyletic atelid clade, but in the tree of von Dornum and Ruvolo (1999), the atelines and pitheciines are successive sister groups to the cebid clade. This cebid/atelid division is supported most strongly by the presence of posterior ethmoid sinuses in the former, and the absence of these sinuses in the latter (Fig. 4). The polarity of this transformation is not clear when only platyrrhines are considered, but it will be clarified by the combined anthropoid analysis (see below). Apart from this, the four trees differ in: (1) their placement of Saimiri, as either the sister of Cebus alone, or as sister to the callitrichines (i.e., a monophyletic or paraphyletic Cebinae); (2) in the relationships among the three callitrichids, with each of the three possibilities represented; (3) the position of Aotus as either a basal cebid or in a polytomy with a cebine clade and a callitrichine clade.
It is not surprising that the placement of Saimiri is equivocal, because it has lost all sinuses, and requires the attendant steps unless paired with Cacajao. The frontal and cupular sinuses are the only features that vary within the callitrichines studied, and both favor a basal position for Callimico because it retains a cupulofrontal sinus but lacks the frontal sinus shared by Saguinus and Leontopithecus (Fig. 5). On the present evidence it is not certain that the absence of the cupulofrontal sinus in Saguinus is shared by Callithrix or Cebuella, but according to Hershkovitz (1977) it is indeed the case.
The ambiguous placement of Aotus is a result of its combination of plesiomorphic conditions (i.e., cupulofrontal, posterior ethmoid, and maxillary sinuses) and derived conditions that represent independent losses in every tree (i.e., loss of anterior ethmoid sinuses). Its possession of posterior ethmoid sinuses does, however, support its inclusion in the Cebidae (e.g., Horovitz, 1999; MacPhee and Horovitz, 2002) rather than in the Pitheciinae (e.g., Kay, 1994; Rosenberger, 2002).
Sinus anatomy generally favors an atelid clade that includes Callicebus but excludes Aotus. This is probably the main reason for the good fit of the four most parsimonious cladograms, and is strongly favored by the molecular data (e.g., Schneider et al., 1996; von Dornum and Ruvolo, 1999). Other close approximations to this atelid or ateloid clade (e.g., Rosenberger, 1992, 2002) suffer from including Aotus in the Pitheciinae. In this respect the anatomy of Tremacebus would be most interesting because its affinities appear to be with either Aotus (Fleagle and Tejedor, 2002) or possibly Callicebus (see comparisons in Rose and Fleagle, 1981). On the basis of its external anatomy, my guess is that it would resemble Callicebus internally, but it might reveal intermediate character combinations that would demonstrate a closer link between the two living genera than is supported by the present evidence from paranasal anatomy.
The distribution of the maxillary and cupular sinuses is such that their loss is unequivocally independent in each case (Fig. 6). Saguinus, Saimiri, and Cacajao are not closely related in any of the phylogenetic hypotheses considered here, so the loss of the cupular sinus in all three and the loss of the maxillary sinus in the latter two are all independent events. Further evidence for the independent loss of the maxillary sinus in living pitheciins derives from the presence of a large maxillary sinus in the stem pitheciin Cebupithecia (Meldrum and Kay, 1997). The small maxillary sinus that fails to invade the supradental region in Aotus has been reported in the type specimen of Aotus dindensis, which Kay (1990) refers to the hypodigm of Mohanamico (Meldrum and Kay, 1997). If one accepts the latter generic attribution, it is noteworthy that its maxillary sinus morphology also matches that of Saguinus as well as Callimico, to which Mohanamico may be closely related (Rosenberger et al., 1990). Finally, Lagothrix is the only genus that possesses a cupular sinus but lacks a cupulofrontal sinus, although this loss might ultimately prove to be a synapomorphy of an ateline clade excluding Alouatta.
The paranasal characters mapped here do not favor any of the particular phylogenetic hypotheses evaluated for catarrhines. All three trees were of equal length, so the character evolution implied by each must be considered in the combined analysis below.
As in the discussion of the platyrrhine results, the tree of Horovitz (1999) will again be employed for the purpose of figures; this time in combination with the “Harrison” tree unless otherwise specified. In contrast, however, this may be taken as a tacit endorsement of Harrison's views (e.g., Harrison, 1982, 1987, 1988, 2002; Harrison and Gu, 1999; Harrison and Sanders, 1999) regarding the relationships among fossil and living catarrhines. Regardless, the “Begun” and “Andrews” trees will also be evaluated.
Not surprisingly, the tree of von Dornum and Ruvolo (1999) is now less parsimonious than the other three platyrrhine trees because of the extra steps in the evolution of the posterior ethmoid sinuses implied by atelid paraphyly (Fig. 7). The addition of the catarrhines polarizes these characters so that the loss of the posterior ethmoids is a synapomorphy of the atelid clade. Accordingly, only the three other platyrrhine trees will be considered further. Beyond the effect just described, the reconstruction of ancestral states within the infraorders is relatively unaffected by combining the two. Accordingly, the following discussion will mostly concern nodes within catarrhines and at the infraorder and suborder levels.
The history of the maxillary sinus is invariant in all trees considered, having been lost independently in Saimiri, Cacajao, and the stem cercopithecoid lineage only to be regained in the genus Macaca (Fig. 8). Similarly, all trees indicate that the cupular sinus was lost independently in Saimiri, Cacajao, Saguinus, and cercopithecoids. The frontal extension of the cupular sinus follows the same pattern of independent loss in Saimiri, Cacajao, and Saguinus (Fig. 9). It appears to have been lost in Lagothrix, despite the presence of the cupular sinus, but this may be an artifact of my sample size because Hershkovitz (1977, fig. IV, p 34) illustrates a cupulofrontal sinus in one specimen of Lagothrix.
The cupulofrontal sinus is unequivocally absent in primitive catarrhines (Fig. 9), and is only potentially present in Hylobates according to some authors (Cave and Haines, 1940), although none was found in the single specimen examined here. The primitive condition for anthropoids remains ambiguous, but, because its presence is primitive for platyrrhines, it does represent a useful differentia at the infraordinal level, albeit one of uncertain polarity at present.
Anterior ethmoid sinuses, either lamellar or frontal, are primitive for catarrhines, platyrrhines, and anthropoids (Fig. 10). In all trees they are independently lost three times: in cercopithecoids, hylobatids, and pongines. In platyrrhines, their loss is independent in Saimiri, Cacajao, and Aotus. Posterior ethmoid sinuses are also primitive for anthropoids, platyrrhines, and catarrhines (Fig. 11). Within catarrhines, their pattern of loss parallels that of the anterior ethmoid sinuses, while in platyrrhines their loss constitutes a synapomorphy of an atelid clade that includes Callicebus but not Aotus (cf., Horovitz, 1999; MacPhee and Horovitz, 2002).
When mapped on the “Horovitz” and “Begun” or “Andrews” trees, the frontal sinus appears as a synapomorphy of the hominoid clade (Fig. 12a), being present in the stem hominoid Proconsul, as well as Afropithecus and Morotopithecus, which are either stem hominoids (in the “Begun” tree) or stem great apes (in the “Andrews” tree). In these scenarios, the sinus is lost independently in gibbons, cercopithecoids, and the Sivapithecus/Pongo clade. When the “Harrison” tree is employed, several catarrhine branches appear to become equivocal (Fig. 12b). However, the equivocation on the stem hominoid and stem catarrhine branch above Aegyptopithecus is not between absence and presence, but rather between presence and inapplicable (“9”). In other words, MacClade is indicating that these branches are characterized by either a frontal sinus, or the absence of anterior ethmoid sinuses. Because the possession of anterior ethmoid sinuses is unequivocal on all these branches, the presence of a frontal sinus is indicated as well. Far from being a problematic behavior of the program, this is precisely what MacClade was instructed to do when the decision was made to employ reductive coding in order to preserve transformational independence between the anterior ethmoid sinuses and the logically dependent frontal sinus (see above). In contrast, the equivocation on the stem catarrhine branch below Aegyptopithecus is truly equivocal.
The ancestral anthropoid, platyrrhine, atelid, and cebid conditions are significantly affected by which platyrrhine tree is used. With the “MacPhee and Horovitz” tree, the frontal sinus is always independently evolved in atelids and cebids (Fig. 13a). When the “Horovitz” tree is employed, the primitive platyrrhine condition is either equivocal or inapplicable, depending on the catarrhine tree with which it is combined. The “Schneider” tree reconstructs the frontal sinus as primitive for platyrrhines in all combinations. Interestingly, the “Schneider” tree also affects polarity within catarrhines, and makes the frontal sinus a primitive property of stem catarrhines and hominoids (Fig. 13b). In sum, the evolutionary significance of the frontal sinus is highly dependent upon which phylogenies are employed. Consequently, the presence or absence of the frontal sinus at these nodes cannot be unambiguously determined on the present evidence, but the plausibility of homology between the platyrrhine and strepsirrhine frontal sinus (Rossie, 2006) suggests that it was present in the ancestral anthropoid morphotype.
The ethmoid labyrinth is not present in any platyrrhines, although Cebus approaches the condition. Its presence in Aegyptopithecus, Proconsul, Morotopithecus, and the African ape clade would seem to make it a clear symplesiomorphy of catarrhines (Fig. 14a). However, with the set of taxa included here, it is equally parsimonious to infer that it evolved in stem catarrhines, was lost before the divergence of cercopithecoids, and was then regained independently in Morotopithecus and in the African ape clade (Fig. 14b). This scenario would be eliminated if Afropithecus turns out to bear evidence of an ethmoid labyrinth. Considering its facial similarities to Morotopithecus, Proconsul, and Aegyptopithecus (Leakey et al., 1991), this would not be surprising.
Summary of Ancestral Morphotypes
The preceding character analysis has demonstrated that paranasal morphology, while variable in detail at low taxonomic levels, provides useful synapomorphies at higher levels. Beyond providing strong support for the cercopithecoid, pongine, and atelid clades, paranasal characters define unique ancestral morphotypes for catarrhines, platyrrhines, and potentially anthropoids.
The ancestral catarrhine morphotype is characterized by the presence of cupular, maxillary, anterior ethmoid, posterior ethmoid, and possibly frontal sinuses as well as the absence of a cupulofrontal sinus. Of these features, the absence of a cupulofrontal sinus may be derived in catarrhines if its presence in platyrrhines is primitive.
The platyrrhine ancestral morphotype includes anterior and posterior ethmoid, maxillary, cupular, and cupulofrontal sinuses. As described previously, it is also possible that the frontal sinus was present as well. Of these, only the cupulofrontal sinus is diagnostic of platyrrhines.
The ancestral anthropoid morphotype is somewhat ambiguous at present because of the need for data from strepsirrhines and tarsiers, but several significant possibilities exist. The cupular, posterior ethmoid, and maxillary sinuses were certainly present (contra Cave, 1967), but neither the cupulofrontal sinus nor the ethmoid labyrinth were present (unless they exist in strepsirrhines). The anterior ethmoid sinuses were also present, but whether these were manifested as lamellar cells or frontal sinuses remains equivocal (see above). Pending proper documentation of tarsiers, strepsirrhines, and euarchontan mammals, the best prospects among these for anthropoid synapomorphies are the cupular and posterior ethmoid sinuses.
Character mapping fails to clarify the homology of some of the sinuses investigated here. In particular, the frontal sinus may or may not be homologous across anthropoids according to this “taxic” approach to homology. Without descending into a lengthy discussion of homology, it is worth mentioning some ways in which a strictly taxic approach to sinus homology is problematic. The maxillary sinus of anthropoids is now fairly well studied, and it provides a useful illustrative example of these problems.
The macaque maxillary sinus represents, on the preponderance of evidence (Rae, 1999; Rae et al., 2002), a reacquisition. It is therefore not homologous with that of other anthropoids in a taxic sense. That is, because it is not a synapomorphy of macaques and other anthropoids, it cannot be considered homologous. This equivalence of synapomorphy and homology is the cornerstone of the taxic homology concept, and it seems logically valid.Valid as it may be, it is not clear that the synapomorphy = homology argument is sound. One of the premises of the argument is that all homologies are synapomorphies at some hierarchical level, and this may be false in some cases. In certain instances of parallelism, it may be that there is continuity of information (sensu Van Valen, 1982) without continuity of phenotypic expression, which leads to incongruence and the impression of homoplasy. Brundin (1976) and Saether (1979) have argued that such cases of parallelism were phylogenetically informative. Briefly, the cases of interest are those in which a character state (primary homology) appears to evolve independently in more than one closely related clade. Although the taxonomic distribution of the feature does not indicate its presence in the common ancestor of the two (or more) clades, the feature may be sufficiently unique or ontogenetically similar in all its instances to suggest that the capacity for its development must have been present in the common ancestor (Brundin, 1976). This capacity itself can then be thought of as the synapomorphy. Because such cases involve primary homologies that fail the test of congruence, parsimony algorithms cannot treat them as support for the larger clade in which they are found. Their phylogenetic meaning is therefore relegated to an a posteriori consideration, although notably the same primary homology might pass the congruence test with only slight changes to the same character matrix, as was the case in the present mapping of the frontal sinus on competing trees.
Saether (1979) recognized that a distinction should be made between parallelisms that are caused by parallel selection and those that are caused by the same inherited genetic and developmental mechanisms. In the former case, the similarity of the convergent features is a result only of the limitations imposed by the body-plan of a clade (e.g., similarities in postcranial morphology between suspensory platyrrhines and catarrhines). These may be taken as supporting evidence for, in this example, an anthropoid clade, but they are not homologous features by any standard. In contrast, parallelism resulting from shared inherited genetic and developmental factors might be considered homologous in the biological sense. Saether (1979) termed these “underlying synapomorphies” and provided several examples of such features in clades of Chironomidae.
Our current understanding of the catarrhine maxillary sinus indicates that it may be an example of an underlying synapomorphy. The absence of the maxillary sinus in the stem cercopithecoid Victoriapithecus, all colobines, and all but Macaca among cercopithecines suggests that the structure was lost early in cercopithecoid evolution only to be regained in the genus Macaca (Rae et al., 2002). However, nothing in our knowledge of the biology and development of the macaque sinus would lead us to doubt its homology with that of other anthropoids (Maier, 2000; Rossie, 2003). In other words, it fails only the congruence test. Given the similarities in the ontogeny of the macaque and hominoid maxillary sinus, it appears that the same set of morphostatic constraints are at work. If so, then the discontinuity of information that results in the failure to develop a maxillary sinus in most cercopithecoids is not a loss of information that pertains to how to build a maxillary sinus, but rather that which dictates whether or not to build one. If the maxillary sinus in macaques and other catarrhines are the product of a shared set of morphostatic constraints, then they are biologically homologous (sensu Wagner, 1989a). As stated by Rieppel (1993, p 21), “[w]hat is inherited, …, is not characters, traits, or structures, but rather the information and therewith the potential to create characters or structures through development.”
Moreover, there may have been no discontinuity of information whatsoever if the suppression of the sinus in cercopithecoids is caused not by a loss of morphogenetic information but simply by structural constraints that impinge upon its development. As Roth (1994, p 318) noted, “[i]f the development of a particular character is blocked by environmental factors in one generation, for example, additional copies may nonetheless reappear in later generations”. The ontogenetic data from recent studies (Rossie, 2003, 2006; Smith et al., 2005) shed some light on the potential identity of these factors in the case of the maxillary sinus, where the dental germs and periorbital tissues appear to have morphogenetic primacy over the sinus epithelium. Suffice it to say that what is lost in the case of underlying synapomorphies is not continuity of information, but continuity of phenotypic expression. If there is a distinction between these two phenomena (as suggested by Rieppel, 1993; Roth, 1994; Hall, 1995), then there is truly a difference between homology and synapomorphy. Because this distinction exists, and because the discovery of parallelisms of this sort has important implications for the mechanisms of development that are responsible for maintaining the structural similarities that allow us to identify homology (cf., Wagner and Misof, 1993), I find it counterproductive to advocate the complete replacement of the term homology with synapomorphy. I therefore continue to consider the maxillary sinus of macaques, apes, and platyrrhines a viable homology, albeit not a synapomorphy. A similar argument pertains to the frontal sinus, which is either a symplesiomorphy of anthropoids, or an independent acquisition in platyrrhines and catarrhines according to the character mapping exercise above. Given the ambiguity of the character mapping results, and the similarity of the developmental pattern of the sinus in all anthropoids (Rossie, 2006), I consider the hypothesis of primary homology unrefuted in this case as well.
The preceding discussion can be distilled into a brief evaluation of several long-held hypotheses about sinus morphology in primate evolution.
1The ethmofrontal sinus is not a synapomorphy of hominines and is at least primitive for hominoids (contra Cave and Haines, 1940; Cave, 1967). The frontal sinus of platyrrhines, and possibly strepsirrhines, probably represents a homologue of the ethmofrontal sinus that differs only in the morphology of the ethmoid sinuses that give rise to it during ontogeny.
2The maxillary sinus is a primitive feature of the Primate order, and instances in which it is lost in anthropoids constitute derived conditions. The notion that an enlarged sinus is a derived feature of hominoids (Andrews and Martin, 1987) has recently been refuted on the basis of sinus volumes measured in platyrrhines and catarrhines (Koppe et al., 1999; Rae and Koppe, 2000), and this present study confirms that a large sinus that excavates the alveolar process is present in many platyrrhines and stem catarrhines. Differences in the topology of the sinus in certain clades may have phylogenetic meaning at a lower level, but based on the distribution of such differences among platyrrhines and the intraspecific variation observed in fossil catarrhines, the distinctive morphology of Sivapithecus and Pongo (Ward and Brown, 1986) may represent the only meaningful case.
3The ethmoid sinuses are not hominine synapomorphies (contra Cave and Haines, 1940), but rather are primitive for anthropoids, and the posterior group may constitute a much needed synapomorphy of the Anthropoidea.
4The cupular sinus is clearly primitive for anthropoids, but its distribution beyond this remains uncertain. It may be an anthropoid synapomorphy, or that of a more inclusive clade. The cupulofrontal extension may be a platyrrhine synapomorphy, or its absence may be a catarrhine synapomorphy. Both issues depend on the condition in non-anthropoid primates.
The author thanks Andrew Hill, Eric Sargis, Steve Ward, Richard Sherwood, Erik Seiffert, Richard Lawler, and Todd Rae for helpful discussions on sinus homology and related issues and also Lawrence Witmer and Gunther Wagner, whose works made the present endeavor less theoretically problematic. Finally, the author thanks Sam Márquez for his invitation to contribute to this special issue.