Geometric Morphometric Investigation of Molar Shape Diversity in Modern Lemurs and Lorises


  • Jess White

    Corresponding author
    1. Department of Sociology and Anthropology, Western Illinois University, Macomb, Illinois
    • Department of Sociology and Anthropology, Western Illinois University, 404 Morgan Hall, Macomb, IL 61455
    Search for more papers by this author


In the study of mammalian adaptation to the environment, teeth are of primary importance due to their role as one of the direct interaction points between an individual and its ecological surroundings. Here, molar shape and function are investigated through traditional multivariate statistics and Thin-Plate Splines deformations to compare the relative location of lower first molar occlusal structures (protoconid, metaconid, hypoconid, entoconid, cristid obliqua, and protolophid) in modern lemurs, lorises, tarsiers, and a non-primate outgroup taxon (Tupaia). Results suggest that shape is based both on tooth size and dietary patterns. Small teeth tend to be short (anteroposteriorally) with wide talonids, whereas larger teeth are generally characterized as being long and narrow. In considering non-size related shape trends, frugivorous and graminivorous taxa generally exhibit a relatively buccal intersection of the cristid obliqua with the base of the protolophid, and a relatively “perpendicular” position of the protolophid in relation to the anteroposterior axis of the tooth (defined as the axis connecting the protolophid and hypoconid). Morphological trends of folivores include a central (midline) position of the cristid obliqua-protolophid base intersection and an oblique angle of the protolophid. Insectivorous taxa (primate and non-primate) generally exhibit a central placement of the cristid obliqua-protolophid base intersection (as in folivores), along with a relatively perpendicular angle of the protolophid (as in frugivores). Omnivorous taxa exhibit shape patterns that are intermediate between these three former groups. This study provides a comparative baseline for the interpretation of morphological trends in fossil primate groups, particularly the Adapiformes. Anat Rec, 292:701–719, 2009. © 2009 Wiley-Liss, Inc.

Paleoecological and paleobiological interpretation of the primate fossil record continues to rely upon the quantification of morphological features and comparison with extant primate groups (Witmer,1995; Ungar,1998). Central to the present study is the exploration of biodiversity among adapiforms, fossil primates suggested to be most closely related to modern “tooth-combed” primates (lemurs and lorises). The diversity and evolutionary history of adapiform taxa is of interest in terms of their role in community ecology, as well as their evolutionary relationships to modern lemurs and lorises (Godinot,1988,2006; Masters et al.,2006). To explore these ecological roles, previous studies have utilized extant lemurs and lorises as baselines in the analysis of patterns observed in adapiform material (for illustrative examples, refer to Dagosto,1983; Gebo,1985; Lanèque,1993; Viguier and Tort,2000; Viguier,2002; White and Gebo,2004; Gilbert,2005; White,2005,2006). To provide further insight into the analysis of fossil adapiform taxa, this project aims to establish a comparative base-line for analyzing molar shape using occlusal structures in modern lemurs and lorises by asking:

  • 1What is the degree of allometric influence (as discussed by Gould,1966; Mosiman,1970; Klingenberg,1998) on the relative position of lower molar landmarks in a sample of recent primates and outgroup taxa, and what shape trends are predicted by molar size change?
  • 2Can components of molar size and shape be separated in the current sample so that morphological trends which are size-related, as well as those that are not, can be investigated to construct and infer form-function associations?
  • 3Can differences in the relative location of lower molar landmarks be related to various dietary patterns? How might these trends be applied in exploration of extinct primate adaptation and ecology?
  • 4How do the results obtained here compare to those commonly used methods where a separate measure of body size is required?

Living “tooth-combed” lemurs and lorises are primates distributed throughout sub-Saharan Africa and southern Asia. Lemurs (members of the families Lemuridae, Indriidae, Lepilemuridae, Cheirogaligidae, and Daubentoniidae) are restricted to Madagascar, whereas lorises and galagos (Lorisidae and Galagonidae) are distributed throughout the tropical regions of Africa and southern Asia. Lemurs and lorises are commonly referred to as “prosimians” or “strepsirrhines,” depending on the perspective authors take in reference to the relative placement of extinct and extant tarsiiform taxa within primates, as well as to the evaluation of soft-tissue and hard-tissue characteristics of the crania and postcrania (Pocock,1918; Rosenberger and Szalay,1980; Schwartz and Tattersall,1987; Yoder,1997; Fleagle,1999; Gebo et al.,2000). Here, the term “strepsirrhine” is used to distinguish the living lemur and loris taxa from tarsiiforms and “higher primates” (haplorhines) (Martin,2003). Modern strepsirrhines are incredibly diverse in their body size [between ∼30 g and 6840 g (0.3–6.84 kg)], ecological habitat, and dietary patterning (Fleagle,1999). Whereas lemurs and lorises are generally considered the most basal of the extant primates, particular systematic issues continue to be debated in the literature (Crovella et al.,1993; Purvis,1995; Yoder and Irwin,1999; Wyner et al.,2000; Yoder et al.,2001; Pastorini et al.,2003; Roos et al.,2004; Karanth et al.,2005; Masters et al.,2005). Discussion also continues regarding the relationships between primates and non-primate outgroup taxa. Several recent molecular analyses have scrutinized the relationships between primates, bats, modern colugos (Dermoptera), and tree shrews (Scandentia) (Adkins and Honeycutt,1991; Murphy et al.,2001; Springer et al.,2003; Nishihara et al.,2006). Accordingly, this study also considers primate and non-primate groups that “bracket” strepsirrhines in recent phylogenetic hypotheses to explore shape trends in dental morphology (Fig. 1).

Figure 1.

Summary of current phylogenetic hypotheses concerning the relationships between lemuriforms, lorisiforms, and outgroup non-primate taxa (Tupaia) (Crovella et al.,1993; Purvis,1995; Yoder and Irwin,1999; Wyner et al.,2000; Yoder et al.,2001; Pastorini et al.,2003; Roos et al.,2004; Karanth et al.,2005; Masters et al.,2005,2006). Branch lengths are not intended to indicate evolutionary time or distance.

It is anticipated that an examination of form-function relationships as designed here will provide a comparative framework for interpreting adapiform behavior. Teeth are particularly relevant in the study of environmental adaptation because they serve as one point of direct interaction between an organism and its ecology, specifically dietary adaptation. The benefit of examining dental morphology of extant groups is that form-function relationships between morphology and ecological adaptation can be hypothesized and explored using a variety of methods. Previously, research into molar morphology and adaptation at several different analytical levels has provided a host of methods to apply to extant groups and the fossil record (refer to Ungar,1998; Teaford,2000 for comprehensive reviews). Each of these methods address different aspects of function, be it an individual's dietary behavior during life as measured by microwear methods (for example refer to Ungar et al.,2003; Godfrey et al.,2004; Semprebon et al.,2004), shearing and/or crushing capabilities as investigated through comparative and quantitative studies of dental and mandibular morphology (for example refer to Kay and Hiiemae,1974; Kay,1975; Seligsohn and Szalay,1974,1978; Hiiemae,1978; Kay and Covert,1984; Ungar and Kay,1995; Ungar,1998; Yamashita,1998a; Strait,2001; Kirk and Simons,2001; Ungar et al.,2003,2004 and references therein), or the development of functional dental models (Spears and Crompton,1996; Evans and Sanson,2003; Polly,2004; Evans,2005 and references therein). Molar function has also been included in larger analyses of mastication dynamics and patterns (refer to Hiiemae,1984 and references therein). For example, one of the most popular methods of dietary interpretation utilized by paleoprimatologists is that developed by Kay (1975). In this classic study, an association was drawn between poorly developed shearing and crushing features and frugivory in primates. Folivorous and insectivorous primates, on the other hand, share increased shearing, crushing, and grinding capabilities facilitated by sharp shearing blades and molar cusps.

The present analysis adds to the quantification of shearing capabilities by examining the spatial pattern of shearing and crushing components of the lower molar and their relationship to dietary behavior in modern prosimians and mammalian outgroups. Specifically, I test the hypothesis that lower molar shape is associated with broad dietary categories among extant prosimians using geometric morphometrics. I predicted that if shape differences are associated with different broad dietary categories in closely related groups of mammals, convergent evolution of lower molar shape among prosimians with similar diets may be observed.


Assumptions and Caveats Regarding the Study of Dental Form and Dietary Behavior

The morphology of any structure of an organism is a result of a combination of genetic, evolutionary, and adaptive processes. Teeth are no exclusion to this pattern, and are commonly used to explore functional, evolutionary, or ecological hypotheses concerning living and fossil taxa. Primates generally exhibit between four and six cusps on mandibular molars (the paraconid, protoconid, metaconid, hypoconid, entoconid, and hypoconulid). In primates, the primary structures of the upper and lower molars are the paracone and protoconid, respectively. The acquisition of the protocone of the upper molar in early therian mammals was coincident with the appearance of a posterior talonid basin in the mandibular molars, resulting in a dental relationship that in addition to puncturing and shearing, also enhanced crushing and grinding capabilities (Patterson,1956; Crompton and Kielan-Jaworowska,1978; Butler,2000). This additional functional complex likely offered early mammals a system that could adapt to a variety of dietary patterns and adaptive niches (Clemens,1971). Ontogenetically, the pattern of cusp development closely mirrors the evolutionary pattern of development as the paracone and protoconid are the first to develop during embryonic growth (Hershkovitz,1971). Most recently, several key studies and literature summaries have established the embryonic patterns of growth and regulatory mechanisms responsible for molar morphogenesis (Vaahtokare et al.,1996; Jernvall and Jung,2000; Jernvall and Thesloff,2000; Jernvall et al.,2000; Cobourne and Sharpe,2003; Kangas et al.,2004; Salazar-Ciudad and Jernvall,2004; Kassai et al.,2005; Polly,2006). The implications of this for the functional analysis of tooth morphology are that the form and location of molar landmarks are determined early in embryonic development, and appear to be less susceptible to the effects of later development or of occlusion during fetal life. Therefore, the analysis of molar relative landmark position is appropriate for questions regarding initial functional adaptation and evolutionary change within mammalian lineages. These data, combined with other morphological information, can be used to refine the interpretation of adaptive change in the dentition of mammalian groups.

Sample and Categorization of Dietary Behavior

The taxa included for analysis here represent 12 primate genera (Galago, Otolemur, Loris, Perodicticus, Nycticebus, Lepilemur, Hapalemur, Lemur, Eulemur, Avahi, Propithecus, Tarsius), and one non-primate outgroup genus, Tupaia (Table 1). The sample was analyzed at the generic level to maximize sample size and provide an overall picture of shape variation to apply to the fossil record. Figure 1 illustrates the hypothesized relationships between the taxa included here. In total, 115 individuals were analyzed, housed in the following collections: Department of Mammalogy, Field Museum of Natural History (Chicago); the Department of Vertebrate Zoology, American Museum of Natural History (New York); and the Division of Mammals, Smithsonian National Museum of Natural History (Washington, DC).

Table 1. Summary of the genera and species included in the present study, including sample sizes (N) and dietary classification
  1. FOL, folivorous; FRG, frugivorous; INS, insectivorous; GRM, gramnivorous; OMN, omnivorous.

Lemurcatta7FOL/ FRG
Propithecusverreauxi5FOL/ FRG
diadema3FOL/ FRG

As field projects documenting the diversity of dietary behavior among lemurs and lorises continue to accumulate new data, it has become increasingly difficult to assign organisms to distinct dietary groups. Overlap in the use of food items, different field methods of observation, and seasonal use of resources by primate taxa are just a few of the notable explanations for the increasing complexity of dietary interpretation. However, as the use of shape analysis has great implications for paleoecology and interpreting dietary behavior in extinct organisms, the categorization of gross dietary patterns and the association of morphological patterns to dietary behavior remain useful. Given the incredible amount of information available, dietary groups used here were gleaned largely from composite summaries of regional primate behavior and ecology (Chapman et al.,1999; Gupta and Chivers,1999; Ganzhorn et al.,1999) (Table 1). Although individual field studies may differ slightly from the data presented in these studies, these dietary group assignments were chosen for the convenience of use in subsequent or comparative studies of dental morphology. In only one case, Hapalemur, was the published assignment of “folivorous” not used (Ganzhorn et al.,1999); instead, Hapalemur was assigned to a “graminivorous” category to acknowledge the overall differences between bamboo-feeding and the mastication of leaf material. Dietary classification of the non-primate genus, Tupaia, was based on Emmons (1991).

Geometric Morphometric Investigation of Shape

The methods used by this study fall under the rubric of “geometric morphometrics.” Geometric methods, based on the identification of relative landmark locations in either two- or three-dimensional space, are utilized because they provide both quantitative variables appropriate for multivariate statistical methods as well as representations of shape variation within a sample. The theory and practical application of geometric morphometrics have been thoroughly discussed in the biological and morphometric literature (refer to Rohlf and Marcus,1993; Roth and Mercer,2000; Adams et al.,2004; Zelditch et al.,2004; Slice,2007 for literature reviews and summary and Cardini and O'Higgens,2004; Monteiro et al.,2003; Monteiro and Dos Reis,2005; Maerz et al.,2006 for recent examples of two-dimensional studies). The geometric morphometric approaches used here to analyze shape variation within a sample of primate dental material are as detailed below:

  • 1Landmark Identification: In this study, five dental landmarks were identified on the first lower molar, each representing either the location in two dimensional space of a cusp tip (protoconid, metaconid, entoconid, hypoconid) or the intersection of major shearing structures (the intersection of the cristid obliqua with the base of the protolophid). For simplicity, the latter landmark will be referred to as the “CO-PRLPHD” point. The first lower molar of each specimen was aligned and photographed under standardized conditions so that the occlusal surface was parallel with the lens of a digital microscope (Digital Blue, QX3). Important considerations regarding photography of three-dimensional material has been raised in the literature (Mullin and Taylor,2002; Gharaibeh,2005). Although three-dimensional capture may not be the most practical in field situations and in situations requiring travel with 3-D scanners, care was taken to address the possibility of error introduction in using two-dimensional methods. Here, each individual specimen was centered under the lens of the digital microscope at the same magnification using molding clay. After each specimen was photographed, the position of the movable microscope platform, as well as the shape of the molding clay, was “reset” so that the next specimen would be treated in the same manner as previous material. With this standardized process, it is assumed here that any error that may have been introduced by the procedure is either negligible or consistent over the sample (Mullin and Taylor,2002). The landmarks were subsequently digitized into a coordinate system using the tpsDIG (v 2.0) (Rohlf,2004) software package. The lower first molar was chosen because it is the least variable along the molar tooth row in most primate and non-primate taxa and likely represents the single best tooth to study size distribution and variability in the fossil record (Gingerich,1974,1979).The relative location of molar cusps (protoconid, metaconid, hypoconid, and entoconid), along with the intersection of the anterior extension of the cristid obliqua with the protolophid base, were chosen as landmarks to capture either the gross dimensions of major crushing areas (the talonid basin of the lower molar) or the relative position of shearing structures (Fig. 2). The anterior landmarks (protoconid and metaconid) serve to represent the location of the protolophid, an anterior shearing crest and the anterior border of the talonid basin. The posterior landmarks (entoconid and hypoconid) function as the posterior corners of the talonid basin of the lower molar. Finally, the point of intersection between the anterior extension of the cristid obliqua (the primary talonid shearing crest) and the base of protolophid was also identified to assess the relative location of cristid obliqua (Landmarks 3 and 5) (Fig. 2). Since, in many cases, the anterior extension of the cristid obliqua is located at the base of the protoconid or metaconid (and does not continue to a cusp tip), the intersection of the crest with the protolophid base was identified as a uniform point within the sample. Among taxa, the breadth of the protolophid varies; therefore, the CO-PRLPHD point does not necessarily fall in line with the segment formed by the protoconid (Landmark 1) and the metaconid (Landmark 2). This point, which marks the buccal border of the talonid basin, is designed to provide not only an approximate measure of talonid basin size and dimensions, but also a marker of the direction and location of shearing capabilities.
  • 2Procrustes Superimposition and Geometric Morphometrics: Superimposition methods, such as those based on Procrustes distances, serve to register landmarks in a given sample to one common coordinate system through translation, scale, and rotation (Slice,2005). A generalized Procrustes analysis (GPA) was performed so that all individuals within the sample could be compared to an iteratively computed sample mean (Slice,2005). The transformed landmark data were then included in two-dimensional Thin-Plate Splines (TPS) analysis, designed to illustrate morphological trends in shape difference (Thompson,1917; Rohlf and Marcus,1993; Bookstein,1991,1996a,b; Birch,1997; Slice,2005). The geometric morphometric analyses were performed here using tpsRELW (version 1.42) software package (Rohlf,2005).
  • 3Statistical Analyses: The relationship between size and shape variables is necessary to investigate in any biometric study because size is one of the most important determinants of biological form in mammals (refer the discussions in Gingerich and Smith,1984). In this sample, log-transformed values of the centroid size (the square root of the summed squared differences between each landmark from the center of the form) was used as a proxy for tooth size and was derived only through the process of Procrustes analysis. Centroid size (and transformations thereof) is commonly used in geometric morphometric analyses as it is a size variable that is independent of random, isometric shape variation and can be used in regression analyses to test null hypotheses of isometry within a sample (as described below) and to explore how shape is correlated with size (allometric patterns) once scale is removed from a landmark dataset through superimposition methods (Zelditch et al.,2004). Thus, interpretation of size-shape relationships in the sample was not dependent on a separate proxy variable representing size (such as body mass, tooth length or width, or length of a cranial or postcranial element). Using a variable such as centroid size to examine the influence of allometry in a dataset is key, as shape patterns correlated with allometric scaling are important components in comparisons of biomechanic function, both at an ontological level (within one taxonomic group), as well as between species of difference sizes (Zelditch et al.,2004). Whereas a regression of centroid size on body mass within this sample would potentially provide further size-related information, an important goal of this study was to provide a framework in which to interpret fossil primate dietary behavior based on shape. Thus centroid size is a quantifiable metric in any landmark-based dataset that does not rely on other methods of estimating body mass, which can be difficult in the case of fossil taxa, particularly if other dental or postcranial remains are not available.
Figure 2.

Illustration of the landmarks included in the present study. Landmarks are labeled as: 1) protoconid, 2) metaconid, 3) intersection of the cristid obliqua with the protolophid base, 4) entoconid, and 5) hypoconid. Crushing areas are labeled as: TGB, trigonid basin; TLB, talonid basin; HF, hypoflexid region. Generalized lower molar based on Hershkovitz (1971).

In the present study, a multivariate regression analysis (using statistical definitions established by Slice et al.,1998; Monteiro,1999) performed by the tpsREGR (Rohlf,2007) software package was utilized to: 1) attain an r2 value (square of correlation coefficient), a statistic used to assess the predictive relationship between one independent variable (in this case, molar size as represented by log centroid size) and multiple dependent variables (the partial warps, which collectively describe shape, i.e., the relative distances between molar landmarks), and 2) examine whether the relationship between size and shape indicates a significant deviation from the null model of isometry as measured using a Wilks' Λ statistic. Instances in which r2 values approximating 1.0 suggest a strong predictive relationship between size and shape (i.e., a particular shape can be reliably predicted with a given measure of size). In the latter test, the Wilks' Λ value that is generated can be used to detect a significant difference against a model of isometry (P < 0.05 = allometry; P > 0.05 = isometry) (Rohlf,2007). The tpsREGR (Rohlf,2007) software package also produces a measure of what percentage [R2 of the shape variation in a given sample is not explained by a measure of size (centroid size); Goodall's F-test].

The partial warp scores, being multiple dependent variables that collectively describe shape, can additionally be analyzed using one-way multivariate analysis of variance (MANOVA). A MANOVA of the partial warps tests for collective, multivariate statistical differences in the partial warp scores (shape) among the genera, which yields traditional multivariate statistics (Wilks' Λ). If there is shown to be a significant multivariate difference in shape among the genera, subsequent post hoc one-way ANOVA comparisons tests (for each individual partial warp) are run following a traditional Bonferroni procedure with appropriately adjusted significance levels (α = 0.5/78 comparisons = 0.000064 for each comparison). Finally, a principal components analysis of shape (also often referred to “Relative Warps analysis”) is performed to consider the distribution of the sample in shape space. MANOVA analyses, principal components analysis, and relative warp analysis were performed using the tpsRELW (version 1.42) software package (Rohlf,2005) and SPSS 14.0 (using the exported data from tpsRELW).


Allometry and Statistical Analyses

To gain a measure of the overall effect of tooth size (log centroid size) on the pattern of shape variation within the sample, a multivariate regression analysis was performed. Results for the total sample (combined primate and non-primate taxa, N = 115) using uniform and non-uniform variables shows a moderate relationship between tooth size and molar shape (with an overall multiple r2 = 0.61) and significant shape allometry (the null hypothesis of isometry is rejected) (Table 2). In this sample, 73.98% (R2 = 26.02%) of the shape variation remains unexplained by the size-shape relationship, suggesting tooth size, while important, is not the sole factor determining molar shape variation.

Table 2. Summary of the results obtained through multivariate regression of the partial warp scores against log centroid size in the present sample
 r2Wilks's ΛdfP valueR2 (%)
Total uniform and non-uniform shape components0.61220.38776 (108)P < 0.0126.02%

Given that there was found to be a moderate relationship between tooth size (log centroid size) and molar shape, bivariate plots of the first principal component (generally representative of the greatest amount of variation in the sample) against log centroid size were produced to illustrate patterns of size-related shape change using both by generic and dietary classification (Figs. 3 and 4). Here, using PC1 values, shape changes are predicted to be associated with a transformation from short, wide molars at smaller tooth sizes to long, narrow molars at larger tooth sizes. At smaller tooth sizes, the distance between the entoconid and hypoconid is relatively larger, and the talonid basin is shortened with the decrease in the distance between the intersection of the cristid obliqua and the protolophid base (CO-PRLPHD) and the posterior landmarks (entoconid and hypoconid). At larger tooth sizes, the relative distance between the entoconid and the hypoconid decreases, whereas the talonid basin lengthens with an anterior shift of the CO-PRLPHD position (Figs. 3 and 4).

Figure 3.

Bivariate plot illustrating the relationship of log centroid size (X-axis) with the first principal component. Shape deformation grids along the X-axis represent predicted shape change in the total sample (smallest specimen to largest); shape deformation grids along the Y-axis represent the shape change along the first principal component. Taxonomic groups are as follows: ▾, Avahi; ○, Propithecus; •, Eulemur; +, Hapalemur; ▵, Lepilemur; ★, Nycticebus; ▪, Galago; ⋄, Loris; □, Otolemur; |, Perodicticus; ◃, Tarsius; ×, Tupaia; ◂, Lemur.

Figure 4.

Bivariate plot illustrating the relationship of log centroid size (X-axis) with the first principal component. Shape deformation grids along the X-axis represent predicted shape change in the total sample (smallest specimen to largest); shape deformation grids along the Y-axis represent the shape change along the first principal component. Dietary groups are as follows: ○, graminivorous; •, insectivorous; +, frugivorous; ▵, folivorous; □, omnivorous.

One-way MANOVA comparisons were performed to determine if differences in the mean configurations of the five molar landmarks for each genus were statistically significant in the sample. At the generic level, significant differences were detected in the sample, Wilks'Λ = 0.003, F(72, 533.541) = 13.599, P <0.001. Subsequent ANOVA of the partial warps generated post hoc Bonferroni comparisons that showed several significant differences at a 0.0000643 α-value (summarized in Table 3). Among lemuriform taxa, significant differences were noted between most other lemuriform and lorisiform genera, with the exception of between 1) Lepilemur and Avahi, Otolemur, Perodicticus, and Nycticebus, 2) Propithecus and Avahi and, 3) Eulemur and Lemur. Fewer significant differences were noted between lorisiform taxa; in fact, the only difference at the 0.000064 α-level was noted between Loris and Otolemur. In addition, no significant differences were observed between either Otolemur, Loris, Nycticebus with Tarsius, or between Loris and Tupaia. In comparing the sample using five dietary categories (frugivory, omnivory, folivory, insectivory, and graminivory), significant differences were detected in the sample, Wilks' Λ = 0.050, F(24, 367.511) = 20.892, P < 0.001. Subsequent ANOVA-derived post hoc Bonferroni comparisons for individual partial warps reveal statistical differences among all dietary groups.

Table 3. Summary of post hoc Bonferroni comparisons between taxonomic groups
  1. “SD” notates the presence of a significant difference at the 0.000064 α-level. “N/A” denotes no significant difference.

  2. TARS, Tarsius; GAL, Galago; OTO, Otolemur; PERO, Perodicticus; LOR, Loris; NYCT, Nycticebus; LEP, Lepilemur; LEM, Lemur; HAP, Hapalemur; EULM, Eulemur; PROP, Propithecus; AVH, Avahi; TUP, Tupaia.


Principal Components Analysis

A PCA of the partial warps for the lower first molar indicates that the first two principal components (PCs) account for 70.53% of the total variation within the sample; 48.59% and 20.94% are explained by the first and second axes, respectively (Table 4). The third PC accounted for an additional 14.62%. The dispersion of genera and dietary groups along the first three PCs is illustrated in Figs. 5 through 8. Along the first axis, primate and non-primate insectivores are found at the minimum values, whereas lemurid graminivores (Hapalemur) and frugivores are located at the maximum values. As illustrated in Figs. 5 and 6, shape variation along the first axis appears to correlate with the dimensions of the talonid basin, represented by the relative positions of the entoconid, hypoconid, and the CO-PRLPHD point. At the minimum range of PC1 values, the relative position of the five molar landmarks results in a short, wide tooth characterized by a relatively short, wide talonid basin. At maximum values, the overall dimensions of the tooth indicate a combination of a long, narrow molar with a long, narrow talonid basin. Additionally, the CO-PRLPHD occupies a relatively central location (toward the midline of the tooth) at minimum values, whereas it shifts in a relatively more buccal direction at maximum values (Figs. 5 and 6). These results suggest that the cristid obliqua, a major shearing structure of the lower molar, provides midline shearing capabilities at minimum values along the first axis. At maximum values, the cristid obliqua runs along the buccal border of the tooth, serving as an external shearing structure (Figs. 5 and 6).

Figure 5.

Principal components analysis of shape scatter plot (PC1 and PC2) and associated shape change (N = 115). Shape deformation along is illustrated at the extreme values for each axis. The “consensus” form is included in the lower left corner of the figure. Taxonomic symbols follow Fig. 3.

Figure 6.

Principal components analysis of shape scatter plot (PC1 and PC2) and associated shape change (N = 115). Shape deformation along is illustrated at the extreme values for each axis. The “consensus” form is included in the lower left corner of the figure. Dietary symbols follow Fig. 4.

Figure 7.

Principal components analysis of shape plot (N = 115) illustrating PC1 and PC3. Taxonomic groups are as follows: ▾, Avahi; ○, Propithecus; •, Eulemur; +, Hapalemur; ▵, Lepilemur; ★, Nycticebus; ▪, Galago; ◊, Loris; □, Otolemur; |, Perodicticus; ◃, Tarsius; ×, Tupaia; ◂, Lemur.

Figure 8.

Principal components analysis of shape plot (N = 115) illustrating PC1 and PC3. Dietary groups are as follows: ○, graminivorous; •, insectivorous; +, frugivorous; ▵, folivorous; □, omnivorous.

Table 4. Summary of the relative warps analysis (principal components analysis of shape) and the percentage of variation explained by individual components
ComponentPercentage of variance explainedCumulative percentage of variance explained

Along the second axis (representing 20.94% of the total variation), shape differences appear to correlate with the relative location of the CO-PRLPHD point, and the position of the protolophid relative to the long axis (anteroposterior axis defined by the line between the protoconid and hypoconid) of the tooth (Figs. 5 and 6). At maximum values, the CO-PRLPHD is positioned centrally, while the protolophid lies at a more acute angle relative to the anteroposterior (AP) axis of the tooth. At minimum values along the second principal axis, the protolophid lies at a relatively more perpendicular angle to the long axis of the tooth. In addition, along the minimum range of values, the CO-PRLPHD occupies a postero-buccal position. In comparing PC1 and PC3, further patterns emerge, although the third PC accounts for only 14.62% of the total shape variation (Figs. 7 and 8). Shape deformations along PC3 suggest variation in the length of the cristid obliqua and the position of the CO-PRLPHD intersection. Along PC3 (Y-axis), frugivorous prosimians (e.g., Eulemur and Hapalemur) overlap with some insectivorous taxa (e.g., Galago and Tarsius) exhibiting a shortened cristid obliqua and a buccal placement of the CO-PRLPHD point, whereas prosimians which incorporate a significant proportion of leafy matter in their diet (e.g., Propithecus, Avahi, and Lemur) are distinct. This latter group does overlap, however, with the spread of Tupaia, an insectivorous non-primate and exhibits a lengthened cristid obliqua and a midline position of the CO-PRLPHD point. This is a surprising result, given that Tupaia does not appear to overlap along PC3 with insectivorous primate taxa, suggesting a somewhat differing pattern of relative landmark position in insectivorous mammalian taxa. Additionally, in certain generic groups (such as Eulemur and Propithecus) in which several specific and sub-specific groups are recognized, a large spread is observed (Fig. 7); this may further suggest differentiation within generic groups and would warrant a future investigation of shape difference within these genera.


Geometric morphometric analyses facilitate linkages between raw data and biological patterns, and can be used to explore morphological shape. Here, it was hypothesized that tooth shape should vary significantly with broad dietary categories among the primate taxa examined here. My results suggest that my hypothesis is supported. Certainly, there is a level of phylogenetic influence in that lemuriform and lorisiform taxa separate along relative warp axes. Thus, phylogenetic relationships should not be discounted in any morphological interpretation (Yamashita,1998a). It would certainly be preferable to additionally investigate the correlation between taxonomic classification, size, and dietary behavior in a multi-factorial interaction model. Unfortunately, given that some of the generic groups are also the only representative of a dietary category (for example, Hapalemur is the only graminivore in the sample), such analyses are not possible statistically due to the limited sample size. Overall, however, the results of the analyses performed here document particular shape trends that are associated with dietary behavior.

Effect of Allometry on Molar Shape Trends

As indicated above, a moderate allometric relationship was demonstrated between tooth size and shape (multiple r2 = 0.61), with only ∼27% of the shape variation in the sample predicted by size. Thus, the following size-shape relationships can be summarized (Figs. 3 and 4).

  • 1Smaller teeth generally exhibit relative landmark locations indicating a short, wide tooth (dietary association with insectivory and omnivory).
  • 2Larger teeth generally exhibit a pattern of landmarks indicating a long, narrow tooth (dietary association with frugivory and graminivory).

Non-Allometric Shape Trends

Although size appears to be a factor in shape variation, ∼74% of the shape variation in the sample did not correlate with size. We can further summarize shape trends along the first three principal components that appear to distinguish between dietary behavioral patterns, rather than body size, as (Fig. 9):

  • 1Frugivorous Taxa: Larger talonid crushing areas, produced by the “perpendicular” position of the protolophid and the buccal position of the cristid obliqua. Talonid shearing structures are positioned buccally. It is important to note that Hapalemur (not a frugivorous prosimian) also exhibits a similar morphology, possibly reflecting several phylogenetic or functional factors (as discussed further below).
  • 2Folivorous Taxa: More midline shearing structures with an oblique position of the protolophid and the midline position of the cristid obliqua. Hypoflexid crushing areas are generally larger than in frugivores.
  • 3Insectivores: A relatively “perpendicular” position of the protolophid (as in frugivores), paired with a more midline position of the cristid obliqua (as in folivores), result in midline shearing structures, albeit broader talonid crushing areas than in folivores.
Figure 9.

Summary of shape patterns in dietary groups, (a) frugivory (Eulemur, FMNH 129378); (b) insectivory (Tarsius, FMNH 76860); (c) folivory (Lepilemur, FMNH 5658). Landmark designations follow Fig. 2.

Variation between taxonomic and dietary groups appears to be driven by the relative positions of the cristid obliqua, the protolophid, and the relative dimensions of the talonid basin. The relative location of shearing and crushing structures has direct implications for functional morphology. Occlusal wear patterns in primates have been discussed at length in the relevant literature (Kay and Hiiemae,1974; Hiiemae,1984) and suggest that during initial stages of Phase I occlusion, the cristid obliqua of the lower molar opposes the postparacrista on the upper molar (Fig. 10). As Phase I occlusion continues, the protolophid of the lower molar also contacts the preprotocrista of the upper molar, ending with the crushing action accomplished through the non-parallel position of the cristid obliqua with the preprotocrista. The function of the cristid obliqua is further constrained by the presence and size of the paraconule on the upper molar.

Figure 10.

Illustration of occlusal relationships between the protolophid, cristid obliqua, preprotocristia, and postparacrista in primates (as figured for adapiform Leptadapis magnus) (adapted from Gingerich,1972): (a) positions of the protolophid, cristid oblique, and postparacrista, (b) positions of the protolophid, cristid oblique, and the preprotocrista.

In individuals with a relatively buccal location of the CO-PRLPHD, the grinding surface of the anterior talonid basin is increased. In addition, as the talonid area increases, the basin area available for retention of food items also increases. On the other hand, a relatively midline location of the CO-PRLPHD may provide increased available crushing surface in the hypoflexid region (buccal to the cristid obliqua) (Fig. 2). Thus, the variation in the location of central shearing, crushing, and grinding complexes (represented by the relative positions of the landmarks examined here) may correlate with differing food properties.

The results provide evidence of additional morphological differences between extant folivorous, frugivorous, and omnivorous prosimians. Establishing morphological differences between insectivores and folivores has been difficult because the physical properties of leaf-structures and exoskeletal structures may demand similar tooth adaptations (tall cusps, expanded shearing surfaces). In many cases, the discernment of folivorous and insectivorous taxa using molar morphology often necessitates the inclusion of body size data (Kay,1975; Kay and Covert,1984). Interpretation of shape, as represented here by the relative location of occlusal landmarks, is not dependent on a known estimate of body size. Future analyses incorporating smaller-sized primate insectivores will further test the shape associations discussed here. In addition, it is necessary to model these patterns against physical properties of food (such as hardness or chemical properties) to explore correlations with shape variation.

Although not mutually exclusive (Lockwood,2007), these results highlight the simultaneous influence of both function and/or phylogeny in morphology. For example, Propithecus is generally classified as “folivorous” while including a large proportion of fruit items (and seeds) in its diet (Meyers and Wright,1993; Hemingway,1995; Dew and Wright,1998; Overdorff and Strait,1998; Simmen et al.,2003; Powzyk and Mowry,2003). Here, Propithecus did not group exclusively with other folivorous taxa (Avahi and Lepilemur) or frugivorous lemurs in the comparison of the first two Principal Components (Fig. 5). Rather, of the two indriids included in the sample, Propithecus did exhibit a tooth with a relatively more buccal CO-PRLPHD and a less oblique protolophid, as in frugivores. Propithecus may exhibit tooth morphology somewhat similar to Avahi as a result of phylogenetic relationships, but this may also reflect a functional convergence to similar shearing requirements (as measured, for example, by shearing quotients). Upon closer inspection, characteristics of an “indriid-type” tooth can vary in some details that may be related to dietary function (particularly the size of crushing areas).

These results for indriids (i.e., the presence of a phylogenetic shape signal overlaid on tooth morphology) would seem to bolster current understanding of evolutionary diversification in indriids, particularly in light of genetic data that suggests folivorous Avahi represents the sister-group of Propithecus and Indri (tooth data not presented here) (Rumpler et al.,1988; Razafindraibe et al.,1997; Roos et al.,2004). Moreover, it should be noted that the folivorous Lepilemur, a lepilemurid, is also characterized by morphology most similar to that of Avahi. This result potentially strengthens the inferred linkage between dietary adaptation and the landmark patterns reported here. Certainly, these inferences must be further explored using material from the Malagasy lemur fossil record (sub-fossil material, for example) before firmer statements regarding the dietary diversification of indriids may be proposed.

Likewise, the shape of the talonid basin observed in Hapalemur may also be a result of both phylogenetic and functional “co-opting” and compromise (Yamashita,1996,1998a,b) in which a feature originally efficient at processing hard seeds and fruit may have also been suitable for efficient mastication of bamboo stems and shoots. Similar functional demands related to “puncture-crushing” for processing both bamboo shoots and fruits/seeds, rather than shearing and grinding (as in folivorous primates) (Seligsohn and Szalay,1978), may be influencing the molar shape patterns reported here. Thus, these data are not incompatible with a phylogenetic hypothesis in which the ancestral lemurid was frugivorous (or graminivorous). Both Hapalemur and Propithecus serve as illustrative cases that suggest phylogeny and function play important, integrated roles in understanding the morphological adaptation, which should not necessarily be considered independently when inferring diet.

Lorisiforms (lorises and galagos), which tend to exhibit a more variable or omnivorous dietary pattern (in particular Perodicticus, Otolemur, and Nycticebus) (Charles-Dominique,1977; Fleagle,1999; Wiens,2002), lie in an intermediate position at the intersection between the ranges of primarily frugivorous, folivorous, and insectivorous taxa. With the exception of Loris, the lorisiform taxa included here overlap in shape trends with one another, which may reflect common phylogenetic patterns or a relative lack of morphological specialization. Although Loris is characterized here as “omnivorous,” it is notable that field observations have documented a high degree of insectivory (Nekaris and Rasmussen,2003). In this analysis, of the omnivorous lorisids, Loris groups most closely with other insectivorous taxa (particularly Tupaia). Otolemur and Nycticebus lie in positions along PC1 and PC2 intermediate between insectivorous taxa and folivorous Lepilemur and indriids (Fig. 5).

Thus, shape patterns in relative cusp and crest positions that overlap taxonomic groups serve to strengthen confidence in the proposed functional correlations suggested here. For example, the overall similarity in insectivore tooth shape (both primate and non-primate) suggests insectivory may be associated with a particular cusp and crest pattern. In addition, functional relationships can also be used to evaluate fossil taxa when utilized in an appropriate phylogenetic context. Morphological data should not be conceived of as static, but rather as a component of ecological data that provides a dynamic picture of adaptation and ecological change. Accordingly, examination of fossil material can result in the development of innovative views of change in community ecology and ecogeography through time (Albrecht et al.,1990; DiMichele,1994; Fleagle and Reed,1996; Lehman et al.,2005). For example, food resource quality, food availability, ecological variability (e.g., wet vs. dry forests), and diminishing soil quality are all continual pressures facing modern lemurs, particularly those dependent on fruits (Ganzhorn et al.,1999). Recent studies have suggested that certain folivorous Lemur species may be more capable of adapting to the change in forest ecology on Madagascar than other more “specialized” taxa (Lehman et al.,2006a,b).

Implications and Summary

The present analysis adds to extensive previous groundwork by investigating patterns of molar shape through exploration of cusp and shearing crest morphology, both components of molar morphology shown to develop early in embryonic development. The benefit of having several methods to pursue is obvious: with different strategies, we are provided the opportunity to test dietary hypotheses from several angles. Several different methods have been previously introduced to address the association between body size and dietary adaptation. For example, measurements of shearing, grinding, and crushing, combined with a measure of body size, appear to be indicative of dietary patterns. Although insectivores and folivores each exhibit parallel patterns of well-developed shearing capacities, only with a measure of body size they are distinguishable (Kay,1975). In general, insectivores also exhibit larger molars in relation to either frugivorous or folivorous primates of similar size. Crown area alone, however, may not sufficiently explain functional differences between primate taxa (Gingerich and Smith,1984).

As the goal of this work was to provide a comparative sample from which to infer dietary behavior in fossil primates, a method that did not require body size was preferable. By employing techniques analyzing the relative location of landmarks as described in the present study, insectivores, folivores, and frugivores were shown to exhibit distinctive shapes without the necessity of body size comparison. Thus, documenting how the relative location of occlusal features of the tooth is effected by size augments the results of Kay (1975) and others, adding a new dimension to the previous investigation of size and molar shape in primates ideally suited to interpreting diet from incomplete fossil remains.

Certainly, it would be short-sighted to suggest that the method presented here for inferring dietary specialization provides a complete picture. Since dietary specialization can be defined at the evolutionary, functional, and behavioral level (Ferry-Graham et al.,2002), various approaches will yield different but complimentary results. For example, information concerning an individual's dietary behavior gained through microwear analysis provides a valuable amount of information regarding the environment and food items available to that individual during a discreet slice of time. However, that individual may exhibit a “specialist” or “generalist” microwear pattern due to ecological constraints, competitive populations, or food availability at a specific point in time, and may not fully reflect its evolutionary background and the adaptive pattern of an ancestral population (Ferry-Graham et al.,2002). Thus, data concerning an individual's lifetime paired with data that reflect upon evolutionary ancestry and adaptation provide a broad picture of ecological change, population behavioral patterns, and morphological adaptation that has been canalized through genetic processes (Polly,2006). Furthermore, although a match in interpretation from all sides may be convenient, we must not overlook the value that these methods provide when they offer counter-interpretations, prodding us to search our functional questions more deeply.

To summarize, it was hypothesized and supported here that primate and non-primate taxa would differ in the relative shape and position of dental structures, and that these differences could be discussed in terms of functional demands placed on the organism as an adaptation to a particular dietary regime.

Using a geometric morphometric approach to investigating the relative position of major occlusal structures in the first lower molar of strepsirrhine primates, the shape trends observed are illustrated in Figs. 9 and 11, and can be summarized as:

  • 1Frugivorous Taxa: “Perpendicular” position of the protolophid; buccal position of the cristid obliqua; larger talonid crushing areas.
  • 2Folivorous Taxa: Oblique position of the protolophid; midline position of the cristid obliqua; smaller talonid crushing areas; expanded hypoflexid.
  • 3Insectivores: “Perpendicular” position of the protolophid; midline cristid obliqua; enlarged talonid and hypoflexid regions.
Figure 11.

Summary of shape trends observed in the current study. Symbols indicate: ▵, short, wide tooth, midline CO-PRLPHD, perpendicular protolophid, and large crushing areas (Insectivores); *, long, narrow tooth with oblique protolophid and small crushing areas (Folivores); ○, short, wide tooth, buccal CO-PRLPHD, perpendicular protolophid, and large crushing areas (Frugivores/Graminivores); ⊗, intermediate shape. Branch lengths are not intended to indicate evolutionary time or distance.

Considering the phylogenetic closeness of extant lemur and lorisiforms to fossil adapiforms, these data provide a broad base from which to compare molar shape and draw conclusions regarding dietary trends. This is not to suggest, however, that adapiforms will be expected to replicate this pattern, or that they are directly ancestral to either living lemurs or lorises. What these data do provide is a comparative baseline from which to propose hypotheses of dietary adaptation and allometric molar-shape scaling trends in adapiforms. The body size of adapiforms has been suggested to be fairly broad [100–6900 g (Fleagle,1999)], and future comparisons will bear out whether adapiforms exhibit similar relationships between size and molar shape. This analysis also suggests that certain features of the molar, such as the position of the protolophid and the cristid obliqua, and the breadth of the anterior talonid basin, may be related to functional demands for shearing and/or crushing. Thus, if differentiation in these features is also documented among adapiforms, further functional hypotheses may be posited regarding this diverse fossil group. These results and the approach used here strongly encourage the collaborative use of several lines of morphological data to reveal the most complete picture of dietary adaptation among fossil primates. Moreover, this approach, in combination with other previously successful methods, suggests that understanding behavioral and ecological diversity is always enhanced by the combination and comparison of different methods.


The author thanks the members of the University of Iowa dissertation committee for their valuable assistance: Dr. Russell Ciochon (UI–chair), Dr. Robert Franciscus (UI), Dr. James Enloe (UI), Dr. Christopher Brochu (UI), and Dr. Gregg Gunnell (University of Michigan). She is further indebted Dr. Matthew Bonnan (WIU) and Dr. Gregg Gunnell (UM) for their comments and lengthy discussions in regard to this project since the completion of her dissertation. Her gratitude is extended, as well, to the following for their valuable assistance in accessing mammal collections at their institutions: Dr. Gregg Gunnell (UM Museum of Paleontology), William Stanley (Field Museum, Chicago), Minh-Tho Schulenberg (Field Museum, Chicago), Eileen Westwig (American Museum Natural History, New York), Linda Gordon (National Museum of Natural History, Smithsonian Institution, Washington, DC). She also wants to recognize the conveners and participants of the 2006 Vienna MORPHOFEST for their invaluable advice and assistance. Finally, she owes thanks to the following for their continuing support: Steven Miller, Lindsay Eaves-Johnson, Sara Filseth, Susan Meiers, and Matthew Bonnan.