Zoological Journal of the Linnean Society, 2008, 153, 187–211. With 12 figures Shape variation in the mole dentary (Talpidae: Mammalia) EUGENIE BARROW1* and NORMAN MACLEOD2 FLS 1 2 Department of Earth Sciences, University of Oxford, Parks Road, Oxford OX1 3PR, UK Department of Palaeontology, Natural History Museum, Cromwell Road, London SW7 5BD, UK Received 11 April 2006; accepted for publication 6 July 2007 The clade Talpidae consists of specialized fossorial forms, shrew-like moles and semi-aquatic desmans. As with all higher jawed vertebrates, different functional, phylogenetic and developmental constraints act on different parts of dentary influencing its shape. In order to determine whether morphological variation in the dentary was unified or dispersed into an integrated complex of structural–functional components, a morphometric analysis of the mole dentary was undertaken. The dentary was subdivided into component parts – horizonal ramus; coronoid, condylar, angular processes of the ascending ramus – and outline-based geometric morphometric methods used to quantify, compare and contrast modes of shape variation within the clade. These were successful in revealing subtle differences and aspects of shape important in distinguishing between mole genera. Closer examination of shape variation within the two fully fossorial mole clades (Talpini and Scalopini) revealed several similarities in ascending ramus shapes between genera from each clade. For example, the broad, truncated appearance of the coronoid process in the talpine genera Talpa and Parascalops was shared with the scalopine genus Scapanus. Also, the more slender, hook-shaped coronoid process of Euroscaptor and Parascaptor (Talpini) closely resembles that of Scalopus (Scalopini). Interestingly, subspecies (one from each clade) more closely resembled genera other than their own in coronoid process shape. Important distinctions in horizontal ramus shape were found to exist between the two clades, such as the extent of curvature of the ventral margin and relative depth of the horizontal ramus. Results show shape variation in this region is correlated with dental formulae and the relative sizes of the teeth. The taxonomically important dentition differences characteristic of mammals are also reflected in the horizontal ramus results. Moreover, these results suggest size may be affecting shape and the extent of variation in, for example, the coronoid and condylar processes between the semi-aquatic moles Desmana and Galemys. It is likely that the effects of morphological integration seen at this level of analysis – covariation between shapes of dentary components – may exist because interacting traits are evolving together. Horizontal ramus and coronoid process shape, for example, are similar across Scapanus and Parascalops, but both these shapes have diverged in Scalopus. © 2008 Trustees of the Natural History Museum (London). Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211. ADDITIONAL KEY WORDS: ascending ramus – desmans – eigenshape analysis – fossoriality – horizontal ramus – morphology – morphometrics – relative warp analysis – shrew moles – Talpinae. INTRODUCTION The analysis of shape is important for understanding patterns of morphological evolution. Variation in shape across taxa, clades, groups and/or samples may result from many different factors such as response to *Corresponding author. E-mail: [email protected] selective pressures, different functional roles, changes in developmental processes and even disease or injury (e.g. Zelditch et al., 2004). Shape has long been used to describe and classify taxa and often provides useful characters for phylogenetic studies. Improved understanding of shape variation may help to resolve taxonomic problems and may provide a method for finding new phylogenetic characters (see MacLeod, 2002). Studies of shape variation may also reveal the effect © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 187 188 E. BARROW and N. MACLEOD of ecological factors or developmental processes that override phylogenetic signal and how such demands limit or direct evolutionary change (e.g. Björklund & Merilä, 1993; Schluter, 1996; Hallgrímsson et al., 2005; Klingenberg, 2005). Unfortunately, finding appropriate shape-based features that support good group characterization and interspecies discrimination, and finding useful ways of describing those features is not always easy (Costa & Cesar, 2000). Morphological examination may sometimes be more instructive if it concentrates on substructures rather than the structure as a whole as it is often small aspects of the object that are important phylogenetically or taxonomically. Quantitative analyses can be applied to parts of a structure, in separate analyses, thereby retaining geometric information that would otherwise be lost if the whole structure were examined in a single analysis (MacLeod, 2002). Considering the complete structure in a single analysis will produce more generalized results and may obscure important patterns of regional shape differentiation. This is most likely to be a problem if the taxa under study are closely related or only subtle differences exist between individuals. Moreover, morphometric analysis of small sets of landmarks scattered over a shape, while sufficient for examining questions related to the relative position of substructures across a sample, often fail to capture shape differences existing among corresponding substructures themselves. THE MAMMALIAN DENTARY The mammalian mandible is a complex morphological structure that consists of two symmetrical dentary bones. Several studies of mandibular morphological variation have been performed (e.g. Atchley et al., 1992; Atchley, 1993; Cheverud, 1996; Humphrey, Dean & Stringer, 1999; Duarte et al., 2000; Badyaev & Foresman, 2004; Hylander, 2005; Monteiro, Bonato & dos Rees, 2005; Rees, 2005). Examination of rodents has revealed parts of the dentary that are more or less variable than others, but these findings are based on relative landmark positions for the whole dentary (Klingenberg, Mebus & Auffray, 2003; Monteiro & dos Reis, 2005). The dentary can be divided into different regions (Fig. 1). The horizontal ramus supports the teeth and the ascending ramus provides attachment sites for several masticatory muscles (Hildebrand, 1982). The ascending ramus consists of three processes. The temporalis muscle inserts onto the coronoid process and the masseter and medial pterygoid muscles onto the angular process. The condylar process provides an attachment site for the lateral pterygoid muscle as well as articulating with the cranium. These regions also correspond to the morphogenetic components described by Atchley & Hall (1991). Figure 1. Regions of the mammalian (talpid) dentary. Variation in dentary form arises from changes in the development of its components, and variability in the patterns of integration between those components into a functioning complex structure (Atchley, 1993). Development of the ascending ramus is governed by the density of mesenchymal condensations (from the neural crest cells in embryonic stages of development, see Atchley et al., 1992) followed by development of the associated muscles, whereas the horizontal ramus is mostly dependent on tooth development (Cheverud, 1996). The adult form of the mandible therefore results from interactions between these functional and developmental processes. However, the extent to which these factors interact may vary. For example, the effect of a particular muscle on the ascending ramus only determines the form of the individual process to which the muscle attaches (Hall, 2003). Nevertheless, the structure as a whole must be able to perform effectively and at some level these processes must be integrated. Allometric effects, as well as external factors (e.g. diet and food acquisition), will also contribute towards functional needs. The evolutionary history of a group represents a combination of these functional–developmental factors and factors imposed by the group’s ancestry. Moles are a diverse group with complex phylogenetic history. Here we consider the question of whether shape variation occurs to different extents in individual parts of the mole dentary, in the hope of distinguishing between those regions primarily influenced by taxonomic/phylogenetic factors and those in which other factors (e.g. functional and size differences) may be predominant. As the mole clade contains two fully fossorial subgroups we will also examine questions relating to the existence and extent of intraclade functional convergence and/or phylogenetic unity at a variety of organizational levels within the overall dentary structure. THE TALPIDAE Diverse lifestyles have evolved among the Talpidae. They are distributed throughout Eurasia and North America and include species that are ambulatory (the shrew-like moles), semi-aquatic (desmans), semi- © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY 189 Table 1. Groups within the Talpidae, their functional category and distribution Subfamily Tribe Genus Common name Functional category Distribution Asiatic shrew moles Russian desman Pyrenean desman Star-nosed mole Ambulatory Semi-aquatic Condylurini Uropsilus (4)* Desmana (1)* Galemys (1)* Condylura (1)* Eurasia Eurasia Eurasia N. America Scaptonyx (1)* Urotrichus (1)* Dymecodon (1) Neurotrichus (1)* Scalopus (1)* Scapanus (3)* Parascalops (1)* Scapanulus (1) Talpa (9)* Euroscaptor (6)* Mogera (5)* Parascaptor (1)* Scaptochirus (1) Long-tailed mole Japanese shrew moles Japanese shrew moles American shrew moles Eastern American mole Western American mole Hairy-tailed mole Gansu mole Old World moles Southeast Asian moles East Asian moles Indian mole Short-faced mole Uropsilinae Desmaninae Talpinae Scaptonychini Urotrichini Neurotrichini Scalopini Talpini Semi-aquatic/ semi-fossorial Semi-fossorial Semi-fossorial Fully fossorial Fully fossorial Eurasia Eurasia Eurasia N. America N. America N. America N. America Eurasia Eurasia Eurasia Eurasia Eurasia Eurasia Taxonomy at the genus level follows Hutterer (2005) and includes Nesoscaptor within Mogera (Motokawa et al., 2001). *Genera from which one species is represented in this study. Numbers in parentheses indicate the number of species in each genus. fossorial (shrew moles) and fully fossorial (Hutchison, 1976; Yates & Moore, 1990). Talpids are largely faunivorous, but to varying degrees depending on their habitat. Aquatic forms consume small crustaceans and fish whereas terrestrial forms concentrate, to different degrees, on worms, insect larvae and other small invertebrates (Raw, 1966; Nowak, 1999; Silcox & Teaford, 2002). There are 17 recognized extant talpid genera and approximately 42 species (Nowak, 1999), separated into several clades (Table 1). The Talpinae form the largest group within the Talpidae and include both fossorial and semi-fossorial moles. Shrews and hedgehogs are thought to be the most likely talpid sister groups, and together these three groups form the clade Eulipotyphla (Waddell, Okada & Hasegawa, 1999; Asher, Novacek & Geisler, 2003). Recent phylogenetic studies have been based on both morphological and molecular data and concur that Uropsilus is the most basal talpid genus (Hutchison, 1976; Whidden, 2000; Motokawa, 2004; Cabria et al., 2006; Sánchez-Villagra, Horovitz & Motokawa, 2006). However, the position of other groups within the clade is less certain. For example, there is inconsistency in the relationship between the shrew moles Urotrichus and Neurotrichus (from Japan and North America, respectively; see Moore, 1986; Whidden, 2000; Shinohara, Campbell & Suzuki, 2003) and the relationships between and within the two fully fosso- rial clades, Talpini and Scalopini (Rohlf, Loy & Corti, 1996; Okamoto, 1997; Shinohara et al., 2004). The most recent comprehensive talpid cladogram is that of Sánchez-Villagra et al. (2006; Fig. 2). This is based on a maximum-parsimony analysis of 157 morphological characters and is an extension of Motokawa’s (2004) research on talpid phylogenetics. The fully fossorial lifestyle most likely evolved from semi-fossorial moles, but uncertainties in their phylogenetic relationships have triggered some debate as to whether it evolved in a single evolutionary step (Whidden, 2000) or as a result of parallel adaptations (Motokawa, 2004). The fully fossorial clade Talpini ranges throughout Eurasia while the Scalopini is known mostly from North America (one genus, Scapanulus, appears endemic to China but only a few specimens have been found to date). Sánchez-Villagra et al. (2006) argued that the monospecific genus Scaptonyx was sister group to the Talpini clade. Nowak (1999) points out that little is known about this mole but it is thought to be semi-fossorial, indicating that the fully fossorial lifestyle evolved at least twice (Motokawa, 2004; Sánchez-Villagra et al., 2006). Morphological studies have identified characters of the dentary used in talpid phylogenetic studies, for example by Motokawa (2004: characters 51–55). These have also been used in species descriptions. True (1896) described the form of the coronoid and angular processes and features of the horizontal © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 190 E. BARROW and N. MACLEOD ling 14 out of the 17 extant genera recognized in the current taxonomy (see Table 1). Sample sizes for each genus varied from three to 15 individuals. Specimens used came from the mammalogy collections at The Natural History Museum, London (BMNH), and the Museum für Naturkunde, Humboldt University (MNHU), Berlin. A full list of all specimens used in this study is available from the senior author on written request. METHODS Figure 2. Phylogenetic relationships among the Talpidae proposed by Sánchez-Villagra et al. (2006) based on a maximum-parsimony analysis of 157 morphological characters. Clades indicated (in grey) are discussed in the text (see also Table 1). ramus in his review of the American talpid species and subspecies. A comparative study of the mole Scalopus aquaticus and the shrew Blarina brevicauda highlighted important differences in the dentary (Gaughran, 1954). These include the scoop-like angular process of Scalopus (monospecific) compared with the narrow, spicular angular process of Blarina brevicauda and the larger, heavier condylar process but thinner coronoid process in the Scalopus as compared with the shrew Blarina brevicauda. Dental characters have also been used as a source of information in talpid phylogenetic and taxonomic studies. Absence of the lower canine in the East Asian mole genus Mogera distinguishes them from the Old World moles Talpa, with which they were once included (Nowak, 1999). Ziegler (1971) examined how tooth formula differentiations, including fossil talpid taxa, related to their functional groups/lifestyles. The effect of variation in dental formula (and tooth size) is likely to contribute to aspects of the dentary form, particularly the shape of the horizontal ramus. MATERIAL AND METHODS MATERIAL A total of 124 mole specimens were used in this study. One species was used to represent each genus, total- Advances in techniques used to describe and compare shapes, including imaging and data-acquisition methods, provide powerful approaches to the study of morphological variation (Rohlf, 1990). Geometric morphometrics offers an important tool in the analysis of shape and have been used in biological systematic and developmental studies (Zelditch et al., 2004). The most appropriate method depends heavily on the objects under study. Landmark-based techniques require selecting discrete points on a form that are locatable across all specimens comprising a data set (Bookstein, 1991). Often the number of appropriate landmark points is limited, particularly if the relationship among the study taxa extends beyond closely related or intraspecific levels (MacLeod, 1999) and may result in a highly biased summary of the true shape variation. When there is a lack of landmarks, or when the shape of an object’s outline is of interest rather than its relationship to various landmarks, outline methods are usefully employed (Rohlf, 1990). One outline-based method is eigenshape analysis (MacLeod, 1999). This technique can be used to study an open curve between two landmarks and, by adding internal landmarks, can constrain points along a curve in order to maintain biological correspondence of landmarks across the shape (extended eigenshape). The following approach uses outline-based methods to capture the complexity of the shape of each part of the talpid dentary (otherwise limited to the small number of possible landmarks). Standard eigenshape analysis (MacLeod, 1999) is used to analyse the outline curves defining each process of the ascending ramus. In order to include information regarding the position of the teeth with the horizontal ramus shape, a combination of outline and landmark data are used in a relative warp analysis (Rohlf, 2003). The left dentary of each specimen was photographed in lateral view using a digital camera. If the left dentary was damaged or not available, the right was used instead and the image transformed so that it could be included in the analysis. Left–right asymmetry of the mandible, which undoubtedly exists, is assumed to be unlikely to affect the results significantly. To standardize the orientation of each object – © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY Figure 3. Landmark locations on a talpid dentary. Anatomical descriptions of landmarks are as follows (types correspond to classification of Bookstein, 1991). (1) Anterior-most point of the dentary where the bone meets the anterior edge of the first incisor (Type 1). (2) Maximum curvature on the ventral border between the angular process and the most ventral point on the horizontal ramus (Type 2). (3) Maximum curvature of the posterior boundary between the angular process and the condylar process (Type 2). (4) Maximum curvature of the dorso-posterior boundary between the coronoid process and condylar process (Type 2). (5) Point where posterior edge of the 3rd molar meets the dentary bone, at the base of the coronoid process (Type 1). (6) Point where anterior edge of the 1st molar meets the dentary bone (Type 1). with the lateral surface of the bone facing upwards, parallel to the camera lens – each was positioned such that when viewed from a ventral position a horizontal line could be drawn from the most anterior point of the dentary to the most posterior point of the condylar process. Similarly, when viewed from a posterior position the ascending ramus, from coronoid to angular process, was horizontal. The dentary bone in most talpids is relatively flat, thus facilitating orientation. However, in certain individuals the angular process was slightly medially inverted, making orientation more difficult. Dentary (outline) shapes Outline data were collected from each image using the Media Cybernetics Image Pro Plus (v. 5.1) and tpsDig (v. 2.0, Rohlf, 2004) image analysis software packages. Between 150 and 200 semi-landmark coordinate points were judged to capture the complexity of each curve and were recorded between selected landmarks representing different parts of the dentary. Landmarks were chosen to represent skeletal areas and associated sites of muscle insertion and tooth development and are described in Figure 3. One Scalopus specimen was damaged at the anterior region of the horizontal ramus and one Scaptonyx specimen was missing its angular process. These specimens were not used in analyses involving those particular regions of the dentary. Ascending ramus The ascending ramus was divided into its three processes. The angular process was characterized by the 191 outline curve between landmarks 2 and 3, the condylar process between landmarks 3 and 4, and the coronoid process between landmarks 4 and 5. Once the x,y coordinate points had been collected they were converted into the f shape functions (Zahn & Roskies, 1972) using the xy-phi program of MacLeod (see MacLeod, 1999). One hundred evenly spaced coordinate points were specified for each outline segment. The (unstandardized) f shape functions expressed as patterns of deviation from the sample means shape for the three processes were each submitted to standard eigenshape analysis. The between-shapes covariance matrix was used as the basis for the analysis (rather than the correlation matrix) so the full range of shape variation observed in the sample was expressed (see MacLeod & Rose, 1993). The eigenshape analysis is essentially equivalent to a relative warps analysis of the outputted shape variables (eigenshape scores) explaining different proportions of the f shape functions and results include a set of singular values (expressing the amount of shape variation represented by each eigenshape vector) and a set of eigenshape vectors (= eigenshapes) expressing a set of variance-maximized, composite shapevariation trends that are uncorrelated with each other. The original mean-deviate shape functions can then be projected onto these new eigenshape axes to achieve a low-dimensionality representation of shape similarity and difference patterns inherent in the sample. Ordinations for each process of the ascending ramus were plotted using the resulting eigenshape scores to ordinate the distribution of taxa across the first three shape axes. These ordinations were used to identify clusters relating to particular taxic groups and to spot outliers. The mean shape and eigenshape results for each region of the dentary were then used to model shape change along each of the resulting shape variables. Shape modelling procedures are useful in the interpretation and subsequent analysis of the eigenshape results and follow the discussion provided in MacLeod (1999). Horizontal ramus The outline curve between landmarks 1 and 2 along the ventral edge of the horizontal ramus was also digitized. To include the relative depth of the horizontal ramus in the analysis, landmarks 5 and 6 were employed. First, the xy-phi program was used to interpolate 14 evenly spaced semi-landmark points along the outline between landmarks 1 and 2. The coordinate points from landmarks 5 and 6 were then mean-centred and added to the 14 previously meancentred points outputted by the xy-phi program. This provided a new set of 16 landmark points (12 of which not strictly landmarks but semi-landmarks) that were © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 192 E. BARROW and N. MACLEOD used in a relative warp analysis using the software tpsRelw (Rohlf, 2003). Similar to eigenshape analysis, the relative warp analysis outputs new shape variables (relative warps) that summarize the variation among specimens. Thin-plate spine deformations were then used to visualize changes in shape variation along each relative warp. Talpini and Scalopini There is a large representative of fully fossorial moles in this study. These include four of the five genera from the Eurasian clade Talpini, and the three North American genera from the clade Scalopini. Shape analysis of each part of the dentary was repeated using just these two fully fossorial clades in order to examine differences, if any, between genera within each clade. New shape variables were obtained that differed from the original set because they only explained variation within this smaller sample and were therefore not affected by the greater variation across the larger group. Measurement error Measurement error was assessed following methods discussed by Arnqvist & Martensson (1998). These authors stress the importance of calculating measurement error in geometric morphometrics and its possible sources. Potential sources of error relevant (or measurable) in this study are largely associated with personal error, namely orientation of the objects and selecting landmark positions. In order to assess measurement error a subset (N = 6) of representative individuals was selected from the whole sample. This subset included individuals that had both joined and separate dentary bones (it was easier to orientate dentaries that were separate) and from within the same and separate genera. Individual specimens were photographed three times and outline data were collected from each image for each region of the dentary. Furthermore, outline data were recorded three times on the same photograph. This totalled seven repeated measures for each specimen (one original, three from the same image and three from separate images). Photographs and outline data were collected over a 3-week period so that sequential familiarity factors did not further complicate the analysis. Shape variables were obtained for each repeated measure in either eigenshape or relative warp analysis (with the rest of the sample) depending on dentary region. A one-way ANOVA was performed on each shape variable explaining up to 90% of the variance and the repeatability was calculated using Arnqvist & Martensson’s (1998) equation: R = S2 A ( S2 W + S2 A ) (1.1) 2 where S A is the among-individual variance component (= (MSamong - MSwithin)/n; n is the number of repeated measures per individual; and S2W is the within-individual variance component (= MSwithin). To compare the effect of error resulting from outline and landmark selection with that caused by variation in orientation of each specimen under the camera lens, two repeatability scores were calculated: R1, the repeatability of repeated measures for each individual on the same image; and R2, the repeatability of measures taken from different images of the same individual (and same image). RESULTS MEASUREMENT ERROR Repeatability scores for shape variables explaining up to 90% of the variation for each dentary region are shown in Appendix 1. In each region of the dentary repeatability values decrease as the percentage of variation explained by each variable decreases. Generally, repeatability remains high in the first several shape variables. This is an expected result as the sample signal-to-noise ratio should be higher for axes that explain greater proportions of variance. However, repeatability analysis provides an indication of how far into the eigenshape (and relative warps) decomposition it is ‘safe’ to go when making interpretations. On the whole those variables accounting for less than 2% contribute to more than 20% of the measurement error (i.e. r < 0.80). Repeatability scores become increasingly erratic, and below 90% were sometimes negative. The differences between R1 and R2 are minimal, demonstrating that orientation and landmark/outline collection were equally likely to affect measurement error. One might expect R1 scores to be higher, as they were collected from the same image. However, the outline was sometimes obscured by associated connective tissue (particular on the condylar process) or by the presence of the other dentary bone if the dentaries were attached. Based on these results we conclude that measurement error should have a negligible impact on the results presented below. DENTARY SHAPES The 100 interpolated points along each outline proved sufficient to achieve at least 90% shape characterization accuracy for the individual processes (coronoid, condylar and angular) over the entire sample. It is difficult to picture patterns of shape change along all these axes, but scatterplots of the first three eigen- © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY shapes (ES-1 versus ES-2 and ES-2 versus ES-3), and relative warps for the horizontal ramus shape (RW-1 versus RW-2 and RW-2 versus RW-3), enable visualization of the distribution of taxa within a threedimensional shape space. It should be remembered, however, that a proportion of shape variation remains undepicted (as much as 38% in the case of the coronoid process) on such plots. Nevertheless, eigenshapes and relative warps subsequent to the first three were each responsible for a relatively small proportion of the variation (l ⱕ 10%). In Figures 4–10 and 12 symbol shading corresponds to locomotory/ functional categories (see Table 1): black represents ambulatory, semi-fossorial and semi aquatic moles, white represents fully fossorial moles from the clade 193 Talpini and grey represents fully fossorial moles from the clade Scalopini. Legends for each figure identify individual genera. Models are used to represent shape variation along the eigenshape axes and so provide a generalized baseline for interpreting the geometries depicted by the shape space. Coronoid process shape The first two eigenshapes account for 50.2% of the variation about the mean shape (ES-1, l = 32.9%; ES-2, l = 17.3%). Figure 4A shows that, when projected into the two-dimensional space represented by these axes, the majority of taxa are clustered about the mean shape (the origin of the coordinate system), but extend from approximately -0.2 to 0.3 along ES-1. Figure 4. Distribution of talpid coronoid process outline shape in the ES-1 versus ES-2 (A) and ES-2 versus ES-3 (B) shape planes and corresponding outline shape models (C and D). l values = percentage variance accounted for by each axis based on singular values. For symbol shading conventions see text. © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 194 E. BARROW and N. MACLEOD This cluster contains the fully fossorial moles (Talpini and Scalopini clades) plus Urotrichus and Galemys. Distinct from this group, with ES-1 scores of -0.3 and lower, are Condylura and Neurotrichus. Uropsilus and two of the three Scaptonyx specimens also have relatively low ES-1 scores (the third Scaptonyx specimen is positioned close to the mean shape). Low ES-1 scores are indicative of a coronoid process outline with a short posterior edge and a top that tapers into an upward- and backward-pointing ‘tip’ (see models along ES-1 in Fig. 4C). At ES-1 scores above -0.3 the posterior edge increases in length and the tip rounds off and becomes more upright, characteristic of taxa in the larger cluster. Condylura and Neurotrichus are also distinct along ES-2 (Fig. 4A). Condylura has a relatively high ES-2 score (> 0.1), the outcome of a coronoid process that is more triangular than those represented by lower scores. The triangular shape results from the proximal ends of posterior and anterior margins extending away from one another particularly along the anterior margin to where it meets the tooth row. Like Neurotrichus, Uropsilus and Desmana, both have low ES-2 scores. This is attributable to a narrower, uncinate (True, 1896) coronoid process (ES-2 < -0.1, see models in Fig. 4C). Within-genus variation along ES-2 is quite large in both Condylura (0.10–0.30) and Neurotrichus (-0.34 to -0.12). In each case the specimen with the highest ES-2 score has a wider, less hooked coronoid process outline. However, the distribution along this axis is no greater than in those genera in the larger cluster (which have larger sample sizes). The third eigenshape (ES-3) summarizes a further 11.6% of observed shape variance information and is plotted against ES-2 in Figure 4B. Variation along ES-3 is largely associated with differences in Condylura compared with the rest of the sample. A low score (< -0.2) indicates a coronoid process with a short posterior margin and an irregularly shaped, backward-directed vergence (as in Condylura). As the posterior margin lengthens the outline over the coronoid process vergence smoothes and flattens (see models along ES-3, Fig. 4D). In order to model shape change within the ES-2/ ES-3 shape plane, variation along ES-1 has been constrained at the mean shape (ES-1 = 0.0). As Figure 4C shows, Condylura score particularly low along ES-1. For this reason models for the region occupied by Condylura in Figure 4D (positioned at the mean ES-1 value) exhibit a generalized, but not a close, representation of the true Condylura outline shape (see models at top left of Fig. 4D). The apparent model-reality discrepancy stems from the fact that the Condylura specimens occupy a mean ES-1 position -0.45, not 0.0, and so occupy a region considerably more remote from the ES-2/ES-3 plane that served as the basis for models. Moreover, some details of the geometric distinction between Condylura and the rest of the taxa used in this analysis are best expressed on subsequent eigenshapes. It is apparent in Figure 4B that, within the fully fossorial clade Scalopini, the genus Scalopus is quite separate from Parascalops and Scapanus along ES-3. Similarly, in the Talpini clade Talpa is shown to be distinct from Euroscaptor and Parascaptor, but overlapping with Mogera, along the same axis. Shape variation among fully fossorial moles was examined more closely in a separate eigenshape analysis. Within the Talpini the distinction between Talpa and Euroscaptor and Parascaptor remains in the ES-1 versus ES-2 shape space (shown by the least convex hulls on Fig. 5A). Mogera specimens in this study are represented by the species Mogera wogura, four of which are of the subspecies M. wogura kanai. As indicated by the dotted line in Figure 5A, these four specimens have low scores for ES-1 and are close to the mean shape along ES-2, resembling Talpa, whereas the rest of the genus Mogera is closer to Euroscaptor and Parascaptor in this shape space. A further two Mogera specimens (from MNHU collection) also resemble Talpa along ES-1 (and are included within the dotted line), but their subspecies is unknown. Low ES-1 scores indicate broad, truncated coronoid process outlines. Those with higher scores are more slender and uncinate and are represented by models in Figure 5E. It is important to note that these are new eigenshapes and therefore do not correspond with the models shown in Figure 4C. Among the scalopine genera (Fig. 5C) Scapanus and Parascalops are positioned in a region of the shape space that also contains Talpa. These differ from one another along ES-2 where Scapanus has a more concave posterior margin than the straight posterior outline of Parascalops. Scalopus has a high ES-1 score, closer (in this view of the overall highdimensional shape space) to that of the Talpini genera Euroscaptor and Parascaptor. The genus Scalopus is monospecific. One Scalopus aquaticus specimen included in this study was of the subspecies S. aquaticus machrinus (BMNH: 29.11.7.11). It is more similar to Parascalops and Scapanus than to the other Scalopus specimens along ES-1. A further two Scalopus specimens (MNHU A2832.1 and MNHU 13645, unknown subspecies) share a low ES-1 score with S. aquaticus machrinus. The position of these three specimens is indicated by the dotted line on Figure 5C. The arrow identifies the S. aquaticus machrinus specimen (BMNH: 29.11.7.11). There is considerable overlap along ES-3 in both Talpini and Scalopini clades (Fig. 5B, D). However, the three Scalopus specimens mentioned above have relatively high scores along this axis (indicated by © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY 195 Figure 5. Distribution of coronoid process outline shape among fully fossorial moles in the ES-1 versus ES-2 and ES-2 versus ES-3 shape planes for Talpini (A and B, respectively), Scalopini (C and D) and corresponding outline shape models (E and F). Least convex hulls show position within the shape space that each genus occupies. Dotted lines/arrows refer to position of specimens mentioned in the text. © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 196 E. BARROW and N. MACLEOD Figure 6. Distribution of talpid condylar process outline shape in the ES-1 versus ES-2 (A) and ES-2 versus ES-3 (B) shape planes and corresponding outline shape models (C and D). l values = percentage variance accounted for by each axis based on singular values. For symbol shading conventions see text. dotted line on Fig. 5D), and are characterized by a slight protuberance along the posterior margin of the coronoid process (see models along ES-3 in Fig. 5F). This resembles a ‘mammiform tubercle’ seen in the subspecies S. aquaticus typicus but not usually found in S. aquaticus machrinus according to True (1896). Although size was not taken into account, Scalopus specimens with higher scores along ES-3 (> 0.075) were noticeably larger than those with lower scores. Condylar process shape The outline shape of this process shows no distinct clustering of taxa in the most information-rich, threedimensional shape space. Instead there is a continuous pattern of shape change exhibited along each axis (Fig. 6A). The ES-1 axis represents 60.9% of the shape variation about the sample mean with ES-2 representing 9.1% of the shape variation. Along ES-1, the condylar process outline changes from having long ventral and short dorsal edges at its lowest value (ES-1 = -0.6, e.g. Uropsilus) to the reverse at its highest value (ES-1 = 0.45, e.g. Euroscaptor, Fig. 6A). At both ends of the scale the distal end is narrow, but between these two extremes – close to the mean shape – its depth increases and the two edges are more equal in length (see models in Fig. 6C). Generally the fully fossorial clades and Condylura and Galemys are skewed towards high ES-1 values whereas Uropsilus, Neurotrichus, Urotrichus, Scaptonyx and Desmana are distributed along the lower end of the scale. © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY Those familiar with multivariate ordination analyses will, no doubt, recognize the presence of a strong ‘horseshoe effect’ in the Figure 6A plot (see Greenacre, 1984; Reyment, 1991; Reyment & Jöreskog, 1993). The pattern reflects the existence of a particularly well-structured gradient of shape variation within the condylar process data set. The Uropsilus species exhibit condylar morphologies that are very similar to one another, but very different from the Euroscaptor species (which are, in turn, similar among themselves). All other species within this data set exhibit condylar morphologies intermediate between these extremes such that essentially all linear intermediate alternatives have been realized. As a result, the eigenshape analysis is forced to represent differences between pairs of Uropsilus and Euroscaptor species as being subequidistant on the plot and pairs of intermediate taxa (e.g. Uropsilis– Urotrichus, Uropsilis–Scalopus. Euroscaptor–Talpa, Euroscaptor–Desmana) as having well-structured, intermediate, relative-distance values. This causes the scores of the extreme morphologies to be ‘pulled’ in along ES-1 like a strung bow (see Greenacre, 1984, for additional examples and discussion). This particular data set is noteworthy in that extremal morphotypes are represented by small numbers of species relative to genera with intermediate morphologies. It may well be the case that this horseshoe primarily results from this taxonomic discrepancy. Methods exist to ‘detrend’ such data and so remove the effect, but these are ad hoc adjustments that may serve to introduce other, much more difficult to interpret, artefacts to the result. For the most part we agree with Reyment & Jöreskog’s (1993) observation that horseshoe plots tend to ‘frighten’ data analysts more than they should. The reason for the effect is well known and taking the time to interpret the pattern correctly usually leads – as shown above – to a better understanding of one’s data (see also MacLeod, 2006). A low ES-2 score (< -0.2) represents a condylar process outline that is broad and almost rectangular at its distal end. This characteristic becomes narrow with a narrow articular point and concave dorsal edge at higher values (Fig. 6C). Uropsilus and Euroscaptor have the lowest ES-2 values, but are very different along ES-1. ES-3 accounts for 6.4% of the shape variation about the mean shape and largely contributes to variation in the ventral edge of this outline shape (Fig. 6B, D). Low ES-3 scores indicate a concave ventral edge (for example in Galemys) while high scores indicate a convex ventral edge. The dorsal margin becomes increasingly concave at higher scores resulting in an ‘upturned’ articular tip. This is most distinct in (some) Condylura specimens and, to a lesser extent, in Scaptonyx. 197 Examination of fully fossorial clades in a separate eigenshape analysis revealed little distinction between genera (Fig. 7A–D), particularly among Talpini (shown by overlapping least convex hulls in Fig. 7A, B). The Scalopini were found to have generally lower scores along ES-1, indicting that the condylar process outline has a longer ventral margin (Fig. 7C, E) than some Talpini specimens. Although Parascalops and Scapanus can be distinguished from one another in the ES-1 versus ES-2 shape space, Scalopus overlaps both these clusters. On the whole Scalopus has the highest ES-2 score, representing a narrow distal end of the condylar process. As ES-2 scores fall it becomes relatively wider in Scapanus followed by Parascalops. The Scalopus aquaticus machrinus specimen (BMNH: 29.11.7.11) has a relatively low score along ES-1 and ES-2 positioning it within the shape space defined by Scapanus (as labelled on Fig. 7C). Those specimens resembling the Scalopus aquaticus machrinus specimen in coronoid process outline shape are not distinguished among Scalopus in condylar process outline shape. Although similar geometric interpretation can be made along these axes (Fig. 7E, F) they differ from those in Figure 6 because they result from a separate eigenshape analysis. Angular process shape Eigenshapes 1 and 2 explain 41.3 and 13.5% of the observed shape variation about the sample mean, respectively. The scatterplot of scores along these axes (Fig. 8A) shows that Condylura and Uropsilus have high ES-1 scores making them distinct from the rest of the group. In this case both sides of the process are approximately equal in length, the dorsal edge is somewhat concave and the distal end is relatively narrow (Fig. 8C). Desmana and Galemys cluster closely to one another and have a positive, relatively high score along ES-1 and ES-2. This is indicative of their broad but equal-sided, rectangular-shaped angular process outline. At lower ES-1 scores the outline becomes broader and the dorsal edge shorter than the ventral edge. The distal end is broad (see models in Fig. 8C). The rest of the taxa are widely distributed along ES-1 (from -0.4 to 0.35) and ES-2 (from -0.21 to 0.27). Along ES-2 the angular process outline changes from being narrow and uncinate (ES-2 = -0.2) – for example in Neurotrichus – to a broad and rectangular shape (ES-2 = 0.3) characteristic of Scalopus, Desmana and Galemys (see Fig. 8A, C). ES-3 is responsible for 10.2% of the shape variation about the sample mean. Positive scores indicate the angular process outline has a proximal end narrower or equal to the width of the distal end. The dorsal margin is concave and the distal end rounded. At negative ES-3 scores the dorsal margin is straight, © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 198 E. BARROW and N. MACLEOD Figure 7. Distribution of condylar process outline shape among fully fossorial moles in the ES-1 versus ES-2 and ES-2 versus ES-3 shape planes for Talpini (A and B, respectively), Scalopini (C and D) and corresponding outline shape models (E and F). Least convex hulls show position within the shape space that each genus occupies. Arrow refers to position of specimen mentioned in the text. © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY 199 Figure 8. Distribution of talpid angular process outline shape in the ES-1 versus ES-2 (A) and ES-2 versus ES-3 (B) shape planes and corresponding outline shape models (C and D). l values = percentage variance accounted for by each axis based on singular values. For symbol shading conventions see text. the distal end is flattened and the proximal end is wide (see models in Fig. 8D). No distinct clusters of taxa are visible in the plot of ES-2 versus ES-3 in Figure 8B. Condylura exhibits the highest scores (ES-3 > 0.2) and Galemys the lowest (plus one Mogera outlier – see below) (ES-3 < -1). Differentiation among fully fossorial moles can be seen in Figure 8A and B and is identified more clearly following an additional eigenshape analysis on these groups. Among the talpines, Talpa and Parascaptor form adjacent clusters within the ES-1/ES-2 shape space (Figure 9A), but Mogera and Euroscaptor overlap these groups. The Mogera subspecies M. wogura kanai are not distinguished from the rest of the genus along these shape axes, but one other Mogera specimen (BMNH: 1922.8.24.4) has particularly low scores along each axis (as indicated on Fig. 9A). This results from a wide, relatively evensided angular process outline shape (see models in Fig. 9E), resembling that of Galemys (see Fig. 8). The three scalopine genera form distinct clusters in this shape space (Fig. 9C). Scalopus is separated from Parascalops and Scapanus along ES-1. The high scores in Scalopus are characteristic of a short dorsal margin and broad distal end to the angular process outline. Parascalops and Scapanus are distinct along ES-2. Parascalops (low ES-2) has a broad angular process with relatively straight sides whereas Scapanus (high ES-2) maintains a broad ‘tip’ but has a concave dorsal margin. Parascalops fills a shape © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 200 E. BARROW and N. MACLEOD Figure 9. Distribution of angular process outline shape among fully fossorial moles in the ES-1 versus ES-2 and ES-2 versus ES-3 for Talpini (A and B, respectively), Scalopini (C and D) and corresponding outline shape models (E and F). Least convex hulls show position within the shape space that each genus occupies. Arrow refers to position of specimen mentioned in the text. © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY 201 Figure 10. Distribution of talpid horizontal ramus shape in the RW-1 versus RW-2 (A) and RW-2 versus RW-3 (B) shape planes and corresponding shape models (C and D). l values = percentage variance accounted for by each axis based on singular values. For symbol shading conventions see text. space similar to that of the Talpini genera in the plot of ES-1/ES-2 whereas Scalopus and Scapanus are positioned outside the Talpini shape space (see Fig. 9A, C, E). The Scalopus aquaticus machrinus specimen is not separated from the rest of the Scalopus sample along any of the three eigenshapes (Fig. 9C, D). Although there is considerable overlap between the two clades along ES-3, the Talpini genera Euroscaptor and Mogera and Scalopini genus Parascalops have relatively low scores and an angular process outline that has a longer ventral margin and greater proximal width than those genera from each clade with higher ES-3 scores (Fig. 9B, D, F). Horizontal ramus shape The first three relative warps of these data explain up to 91.1% of the total outline shape variation about the sample mean. Plots of relative warps 1 and 2 (RW-1, l = 58.3%; RW-2, l = 23.9%) separate shapes into tight clusters (Fig. 10A). Galemys and Desmana have RW-1 scores below -0.07, characterized by a horizontal ramus with a ventral outline that curves upwards anteriorly. There is little posterior curvature where it joins the ascending ramus (at landmark 2; see thin plate spline deformation models, Fig. 10C and the landmark positions, Fig. 11). Landmarks 5 and 6 (the molar row) are positioned more posteriorly than in the mean shape and the distance between them – the © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 202 E. BARROW and N. MACLEOD Figure 11. Horizontal ramus shape showing position of landmarks (numbered coordinates, see Fig. 3) and semilandmarks (un-numbered coordinates) relating to models in Figures 10C–D and 12E–F. length of the molar teeth – is short. Consequently, the distance along the antemolar region is relatively long. RW-1 for Condylura is also relatively low (but not to such a great extent). Fully fossorial and semi-fossorial moles have similar positions along RW-1. A high RW-1 value (0.07) in Uropsilus is reflected in the relatively flat anterior ventral margin, short antemolar region and relatively long molar row length. Condylura is distinguished from all other genera by having a RW-2 score well below 0.0 (Fig. 10A). This is depicted by a long, shallow horizontal ramus along the anterior ventral margin and an extended antemolar region of the tooth row (see models in Fig. 10C). The other extreme is seen in Urotrichus and Scalopus. Conversely, these have a deep horizontal ramus curved upwards anteriorly and a short antemolar length. This feature separates Scalopus from the rest of the Scalopini clade and distinguishes between the semi-fossorial Urotrichus and Neurotrichus genera. Scaptonyx is closer along the second relative warp to Neurotrichus than Urotrichus. Within the RW-1/RW-2 shape space Scaptonyx closely resembles the fully fossorial mole Talpa from the Talpini clade. Relative warp 3 accounts for 8.9% of the shape variation about the sample mean in the horizontal ramus. A high score in Neurotrichus, Scaptonyx and Parascalops separates them from the Talpini cluster. Similarly, higher scores among Urotrichus distinguish them from Scalopus (Fig. 10B). Variation along the third relative warp is concentrated in the anterior region of the dentary. At high scores the point where the dentary bone meets the first incisor (landmark 1) is positioned below the level of where the dentary meets the anterior edge of the first molar (landmark 6) resulting in a concave curve along the corresponding ventral edge of the dentary. At low values landmark 1 is positioned above the level of landmark 6 and the ventral edge is less concave (see models in Fig. 10D). The two fully fossorial clades are almost completely separated in the RW-2/RW-3 shape space (Fig. 10B). This is also apparent in the results of a separate relative warp analysis performed solely on the fully fossorial moles. Relatively high scores for RW-1 and RW-2 among talpine genera distinguish them from the Scalopini (Fig. 12A, C). Differences result from a deeper horizontal ramus in Scalopini compared with a shallow horizontal ramus in Talpini. The Talpini are tightly clustered but on the whole Euroscaptor and Parascaptor have slightly higher scores along RW-2 than Talpa and Mogera (Fig. 12A). Higher RW-2 values represent a flatter anterior portion of the ventral margin compared with the more concave nature of this margin at lower scores (Fig. 12E). Within Scalopini, Scalopus is clearly distinct from Parascalopus and Scapanus in the RW-1/RW-2 shape space (Fig. 12C). A low score along RW-1 and relatively high score along RW-2 for Scalopus results from a deep anterior region and straight ventral edge. In Parascalops and Scapanus the horizontal ramus has a shallower anterior region corresponding to a considerable concave curve along the ventral edge (Fig. 12E). Variation along the third relative warp is minimal in both fully fossorial mole clades (Fig. 12B, D). However, Scapanus can be distinguished from Parascalops. The latter has a relatively greater distance between landmarks 2 and 5 and a longer molar region, depicted by a relatively low RW-3 (see Figs 11, 12F). Scalopus is less distinct in this regard. Summary of dentary shapes Examination of dentary shapes within threedimensional shape spaces defined by various eigenshape/relative warps axes reveals that shape variation among genera is least distinct in the condylar process. There is a general pattern across the shape space defined by the first two eigenshapes of genera from Uropsilus, the most basal talpid genus, through the semi-fossorial moles to a culmination of fully fossorial moles. Genera are more distinct in their coronoid and angular process outline shapes; for example, Condylura and Uropsilus are separated from other taxa along axes explaining the majority of the variation in these two processes. The coronoid process shape in the semi-fossorial mole Neurotrichus is quite different from that of the semi-fossorial genus Urotrichus, but they do share a closer resemblance in outline shape of the condylar and angular processes. Scaptonyx clusters amongst Urotrichus and the fully fossorial moles in coronoid and angular process outline shape. The semi-aquatic desmans, Galemys and Desmana, only cluster with one another in the shape of their angular process and horizontal ramus. The Desmana coronoid and condylar processes are closer in shape to those of Uropsilus and Neurotrichus whereas Galemys bears a closer resemblance to the fully fossorial moles. The fully fossorial mole clades, Talpini and Scalopini, share a similar © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY 203 Figure 12. Distribution of horizontal ramus shape among fully fossorial moles in the RW-1 versus RW-2 and RW-2 versus RW-3 shape planes for Talpini (A and B, respectively), Scalopini (C and D) and corresponding shape models (E and F). Least convex hulls show position within the shape space that each genus occupies. © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 204 E. BARROW and N. MACLEOD three-dimensional shape subspace across each process of the ascending ramus, though the Scalopini shape range is often smaller (this could be due to smaller sample sizes for this clade). Shape distinctions within the horizontal ramus shape space are particularly striking. The desmans are tightly clustered and well separated from the rest of the taxa, as is Condylura. Talpini forms a tight cluster that is almost completely separated from Scalopini. Within Scalopini Scapanus and Parascalops closely resemble one another and are well separated from the Scalopus horizontal ramus shape. Of the semi-fossorial moles, Urotrichus clusters closely with Scalopus whereas Neurotrichus clusters with Parascalops and Scapanus. Scaptonyx more closely resembles Neurotrichus than Urotrichus in its horizontal ramus shape. DISCUSSION In this study shape variation was explored across talpid genera by dividing the dentary into separate morphological units so localized or regional aspects of shape variation could be studied in isolation. Outlinebased methods proved particularly appropriate as the shape of each unit could be measured at high resolution. The concave or convex nature of the side margins of each process, for example, is clearly important in shape characterization between genera, but these would have gone unnoticed using landmarkbased methods. Problems do arise when using outline analyses, for example non-correspondence within the variable sequence. There are strategies that overcome this to a large extent (e.g. extended eigenshape analysis; see MacLeod, 1999), but in many cases there is no good information regarding interobject correspondences at this detailed level. Nevertheless, qualitative taxonomists use outline-based clues to interobject similarity and difference on a routine basis. Any method that does not take similar information into consideration can only be compared with qualitative assessments/interpretations with difficulty (MacLeod, 2002). In the absence of such information, and especially in cases where detailed information about the geometric character of outlines is needed, the analysis of semi-landmarks placed at equally spaced locations along the outline represents a geometrically rigorous way of approaching quantitative morphological analysis. The relative warp analysis in this study used outline information combined with landmark data (ventral margin of the horizontal ramus with position of the molar teeth) and provides another powerful method for identifying and characterizing groups. Landmarks focus on geometric changes in a small number of discrete (or quasi-discrete) locations on the structure whereas outline analyses summarize more information about the shape of the structure itself. Thus, in many landmark analyses structures may appear more (or less) distinct than they actually are because only a very small part of the overall geometry is participating in the analysis. This effect might be partly responsible for the very tight clustering of taxa in the relative warp plots for the horizontal ramus results. The eigenshape and relative warp methods were also useful because the shape characteristics of each individual could be projected into the space defined by the first three covariance-based shape vectors and the distribution of individual specimens within this ordination space examined. These ordinations clearly show that differences exist between genera across different parts of the talpid dentary. Some regions show greater variation within genera and considerable overlap between genera along the outline shape variables than others. Despite this, distinctions between closely related genera can be made. These results also demonstrate that common constraints (e.g. geometric, developmental, functional) and/or selective pressures appear to be determining the shape of some parts of the dentary while others have evolved more independently. SHAPE VARIATION Condylura cristata, the star-nosed mole, differs from other moles in that it has 22 fleshy appendages on its muzzle used for navigation and food location (Catania, 2002). A unique shape also characterizes its dentary, which is more elongated and slender than those of other talpids. The foraging apparatus is thought to have been maximized for exploiting large quantities of small prey at high speed (Catania & Remple, 2005). It appears that development of a unique nasomaxillary articulation and nasolabial musculature associated with the starry-nose relates to the evolution of the long proboscis. This has shifted the plane of the anterior teeth, lengthened the mandibular ramus and weakened the masticatory mechanism (including a reduced size of the temporalis and masseter muscles) compared with other talpids (Grand, Gould & Montali, 1998). Correspondingly, our results show that the Condylura dentary also displays unique outline shapes for each part of the dentary. Among other moles in this study distinctions could be made across one or more parts of the dentary and factors contributing to the similarities or differences are discussed below. Variation within the horizontal ramus In this study the horizontal ramus provided the clearest distinction among genera compared with the parts of the ascending ramus. It appears therefore that © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY variation in horizontal ramus outline shape is more tightly constrained within each genus than parts of the ascending ramus. Furthermore, variation is not evident between the (small number of) subspecies within Mogera and Scalopus, unlike in the outline shape of parts of the ascending ramus. Shape of the horizontal ramus is mostly affected by tooth development (Cheverud, 1996). Inclusion of dental information has probably strengthened the discriminant power of horizontal ramus shape in the distinction between genera in this study. Position and length of the molar row, and its distance from the ventral margin (i.e. the depth of the horizontal ramus), explained a large proportion of the horizontal ramus variation about the mean shape. There are also important distinctions in the antemolar formulae among talpids (Ziegler, 1971) and differences, such as the number of antemolar teeth or their relative sizes, appear to determine the depth and curvature of the horizontal ramus. Dentition differences coincide with horizontal ramus shapes found among genera in this study. In Desmana and Galemys the first two incisor teeth are large (particularly the second incisor) relative to the rest of the antemolar dentition and the anterior region of the horizontal ramus is deep. Similarly, Urotrichus has an enlarged second incisor (the first is missing in this genus) and a corresponding deep anterior region of the horizontal ramus. In the latter case, however, our results indicate the horizontal ramus is shorter as it supports a reduced number of antemolar teeth, compared with the full complement held by Desmana and Galemys (Ziegler, 1971). In Uropsilus the number of antemolar teeth is also reduced and the molars occupy the majority of the space along the horizontal ramus. The anterior region of the horizontal ramus is similarly short but shallower, a result of the enlarged, procumbent (shrew-like) incisor tooth. Talpine genera are all characterized by an enlarged first premolar tooth and small incisor and canine teeth, unlike Scalopini. Scapanus and Parascalops, within the Scalopini, have largely unspecialized and uniform antemolar dentition. These differences appear to be correlated with the outline shape of the anterior region of the horizontal ramus. The horizontal ramus shape in Neurotrichus was found to be similar to those of Scapanus and Parascalops. Neurotrichus also has relatively unspecialized incisors and canine teeth, but lacks two premolars (made up for by the remaining two being relatively larger). The reduced number of antemolar teeth and enlarged second incisor in Scalopus distinguish it from the other two Scalopini genera and results in its horizontal ramus shape resembling that of Urotrichus instead. In Condylura it is not so much the number or size of the teeth as the substantial elongation of the anterior region (i.e. antemolar 205 length) of the horizontal ramus that distinguishes this genus from the rest, presumably relating to unique features of the upper jaw and cranium. Variation within the ascending ramus Greater variation between and within some genera was found in the outline shape of the ascending ramus processes than in the horizontal ramus. Changes in the shape of these parts are initially determined by the density of mesenchymal condensations during development (Cheverud, 1996) followed by variation in the development of masticatory muscles (Atchley et al., 1992). Outline variation is substantial in the coronoid process of Mogera, Parascaptor, Scalopus and Urotrichus, both in the broadness of the process as a whole and in the relative straightness or otherwise of the posterior margin. Similarly, variation in the angular process is quite substantial in some talpine genera and Urotrichus in the width of the process, and the length and degree of curvature of the dorsal margin. [Note: potential source of error resulting in variation in the angular process could be due to the fact that in some individuals this process is slightly medially inverted.] Despite variation within these two processes and their subsequent overlap within the threedimensional shape space, some genera were still clearly distinguishable from the rest of the sample, occupying their own place within the shape space. Uropsilus, the most basal member of the Talpidae, and Condylura both have a distinct angular process shape. It is narrow and spicular (Gaughran, 1954), resembling that of outgroup (shrew) taxa. This feature has been used as a phylogenetic character (Motokawa, 2004). Uropsilus, Neurotrichus and Desmana have distinct coronoid process shapes. On closer inspection of the fully fossorial moles, important distinctions became apparent in the coronoid and angular process shapes among the Talpini and Scalopini genera. It is interesting to note the variation in the coronoid process between subspecies of the same species, and the extent of variation between subspecies could be examined further in a larger study. Some specimens in the present study resembled the identified subspecies along measurements characterizing coronoid process shape. One could tentatively assign these individuals to the relevant subspecies, but to determine this confidently would require additional information not available (e.g. locality data). The relationship between bone morphology and muscle attachment has been known for some time (Slijper, 1946; Grüneberg, 1952; Atchley & Hall, 1991). The muscles that attach onto the ascending ramus processes (see Introduction) are developmentally linked and each underlying morphogenetic component of the dentary is governed by relatively independent © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 206 E. BARROW and N. MACLEOD direct genetic or epigenetic control. Hall (2003) showed how congenital absence of the lateral pterygoid muscle resulted in the condylar process failing to form but had no effect on the other parts of the dentary. This might explain the greater variation among genera in parts of the ascending ramus than the horizontal ramus, if during development additional factors ‘allow’ for greater shape change in the processes. Continuous distribution of genera along variables representing the outline shape of the condylar process shape (i.e. lack of any distinct groups) was found. Even Condylura could not be clearly distinguished along the first three eigenshapes for this outline. Unlike other morphogenetic regions of the mammalian dentary the processes are capable of forming specialized cartilage that is later replaced by endochondral bone (Hall, 2003). Therefore, in the condylar process, for example, the basal portion forms by intramembraneous ossification whereas the distal end forms by development of the cartilage (known as condylar cartilage). Similar, but more transitory cartilage is found on the angular and coronoid processes (Hall, 2005). Development of the cartilage results from supplementary growth in response to local factors (i.e. mechanical factors such as strains and stresses and muscle development) (Tomo, Ogita & Tomo, 1997). The condylar process also articulates with the cranium forming the temporomandibular joint, and complex adaptive and compensatory growth takes place in order to maintain this joint. It is possible that this additional function and prolonged development of condylar cartilage may restrict variation among talpid condylar process outline shape compared with the coronoid and angular processes. SHAPE AND PHYLOGENETIC RELATIONSHIPS The semi-aquatic desmans, Desmana and Galemys, undoubtedly form a monophylum as found in phylogenetic studies (Hutchison, 1976; Whidden, 2000; Shinohara et al., 2003; Sánchez-Villagra et al., 2006). Nevertheless, differences among their dentaries were found and may result from a number of factors during a separate evolutionary history of several million years. This investigation has documented greatest similarities in aspects of shape characterizing the angular process and horizontal ramus. The outline shapes of the condylar and coronoid processes were, however, quite distinct; Galemys more closely resembles the shape seen in the fully fossorial moles. Similarity in the shape of the horizontal ramus is not surprising given their dentition (see above). It is possible that some other underlying factor has preserved the shape of the angular process and independent evolution of the coronoid and condylar process has been occurring since the time the two lineages split. Although size variation was excluded from our analysis, Desmana is clearly larger than Galemys. Size could be constraining some aspects of the desman dentary. The coronoid and condylar processes of Galemys more closely resemble the outline shape of more similarly sized, fully fossorial moles, suggesting that size has acted on the evolution of the shape of these two processes. Determining whether this is the case would require examination of fossil Desmaninae species (e.g. Desmagale & Mygalea; Hutchison, 1968, 1974). Older considerations of talpid phylogeny hypothesized that the two semi-fossorial shrew moles (Neurotrichus and Urotrichus) were sister groups (Van Valen, 1967; Yates & Moore, 1990; Whidden, 2000). This analysis found that coronoid process and horizontal ramus shape between these two genera are quite distinct, supporting more recent phylogenetic hypotheses that these two genera are not closely related (e.g. Shinohara et al., 2003, 2004; SánchezVillagra et al., 2006; NB: given the variability in the coronoid process shape the horizontal ramus provides better ‘support’ for this). Sánchez-Villagra et al. (2006) placed the Japanese shrew mole, Urotrichus, most basal after Uropsilus and before the desmans while Neurotrichus formed a sister group relationship with the rest of the taxa (Scalopini, Talpini, Scaptonyx and Condylura). Judging by the appearance of the shapes of the ascending ramus in this study Urotrichus more closely resembles the Talpini and Scalopini clades than does Neurotrichus. However, in the shape of the horizontal ramus Neurotrichus more closely resembles the fully fossorial moles (excluding Scalopus – see below), particularly to Scapanus and Parascalops, and relates to greater similarities in their dentition compared with Urotrichus. The presumed semi-fossorial mole Scaptonyx has been described as ‘[looking] like a mole with the feet of Urotrichus, or like a Urotrichus with the head of a mole’ (Allen, 1938: 66; ‘mole’ refers to a talpid adapted to fossorial life in this instance). Hutchison (1968) described a talpine appearance of the Scaptonyx lower antemolars. Furthermore, the third incisor is sometimes missing (Ziegler, 1971), as in Mogera. Given these facts it is not surprising that the Scaptonyx horizontal ramus shape resembles that of the Talpini. However, Scaptonyx also shows a close phenetic affinity with the American shrew mole, Neurotrichus, and the American (fully fossorial) moles Scapanus and Parascalops, in horizontal ramus shape. Scalopus has changed considerably in relation to Scapanus and Parascalops and has the most reduced antemolar dentition of all known scalopine moles (Ziegler, 2003). Sánchez-Villagra et al. (2006) positioned Scaptonyx as sister group to the Talpini (suggesting that the fully fossorial lifestyle evolved twice), forming a clade that is in turn sister group to the Scalopini (see Fig. 2). © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY However, they recognized that this grouping was weakly supported and required further testing. The two fully fossorial clades could only be fully distinguished from one another based on horizontal ramus shape. It is important to note that in different parts of the ascending ramus there are shape ‘equivalents’ in each clade. For example, the Parascalops and Scapanus coronoid process and angular process outline shape is similar to that of Talpa whereas the coronoid process shape of Scalopus resembles the other talpine genera (although its angular process does not). Note, this difference could be interpreted as arising due to chance (there are smaller sample sizes of scalopine genera) or it could be a consequence of other factors causing similar patterns of morphological divergence among closely related talpids, such as character displacement (e.g. Dayan & Simberloff, 1994) resulting in similar radiation patterns. Additional investigation will be needed to resolve this issue. A clade consisting of Scalopus and Scapanus has been strongly supported by numerous analyses (e.g. Hutchison, 1976; Whidden, 2000; Shinohara et al., 2003, 2004; Motokawa, 2004) and Parascalops as sister group to them (Yates & Moore, 1990; Shinohara et al., 2003, 2004). In outline morphology parts of the Scapanus dentary are more similar to Parascalops than to Scalopus along axes explaining the majority of the variation (except in characteristics of the angular process shape where Scapanus and Scalopus are both distinguishable from Parascalops). The Scalopus subspecies S. aquaticus machrinus more closely resembles Scapanus than the rest of the Scalopus specimens in aspects of coronoid and condylar process shape. Obviously, this observation is based on only one specimen of this subspecies and so should be treated with caution. Scalopus specimens previously mentioned that share similar coronoid process shape characteristics with S. aquaticus machrinus do not do so in condylar process shape. They are, however, all of a similarly larger size than the rest of the Scalopus sample. S. aquaticus machrinus is the largest of the Scalopus subspecies (True, 1896). Scalopus has the largest range of any of the scalopine moles (Yates & Schmidly, 1978) and, like Parascalops, is from eastern North America whereas Scapanus is restricted to the west coast. The Scapanus species used in this study (S. townsendii) is the largest of the three Scapanus species and is noticeably larger than the more similarly sized Parascalops and Scalopus. Sánchez-Villagra et al. (2006) hypothesized Talpa as the sister group to a Mogera–Scaptochirus (not used in this study) clade, followed by Euroscaptor and Parascaptor. Measurements characterizing variation among structures examined resulted in different degrees of overlap in the pattern of variation between these genera. For example, variables accounting for 207 angular process shape distinguished between Talpa and Mogera, but variation in Euroscaptor and Parascaptor overlaps these two genera. Talpa can be distinguished from the other three talpine genera (except for the Mogera subspecies M. wogura kanai) in the coronoid process shape. As mentioned above, differences within subspecies suggest important variation may also be evident at the species level. A greater degree of overlap of shape measurements among Talpini genera than Scalopini genera may be partly due to smaller sample sizes in the latter. In their study of morphological evolution in the mandible of spiny rats, Trinomys (Rodentia: Echimyidae), Monteiro & dos Reis (2005) found that coronoid process shape exhibited the greatest conservatism during genetic divergence – the coronoid process was responsible for shape differentiation between two clades. Although the most basal genus Uropsilus is fairly distinct in the coronoid process shape, Urotrichus and Galemys share a similar shape space with the two fully fossorial clades to which they are not closely related. Despite this there is subtle variation among closely related genera (e.g. within Scalopini) in addition to possible differences between subspecies. The coronoid process plays a more dominant functional role in the mole dentary than in the rodent dentary, which might explain the overlap of coronoid process outline shapes between mole genera that are not closely related. Among moles, movements are mostly governed by the temporalis muscle pulling on the relatively large coronoid process. But in rodents the coronoid process is small and dentary movements are dominated by action of the medial pterygoid and masseter muscle on the angular process. The less significant functional demands on the rodent coronoid process might explain why this part of the dentary in rodents is most closely associated with common descent (Monteiro & dos Reis, 2005). Convergent morphologies are likely to confuse phylogenetic signal and, as has been shown in carnivoran tarsal morphology, younger clades that have not had time to diverge may not be affected by convergence (Polly, 2008). Scalopini and Talpini clades probably diverged before the Miocene (M. R. Sánchez-Villagra & E. C. Barrow, unpubl. data) allowing substantial time for the morphospace to be explored. However, some genera within each of these two clades and their subspecies may have diverged more recently. It is tempting to speculate that this might explain why shape differences are apparent at this level, but the entire question needs further investigation. Another factor that might be important in determining shape variation in different parts of the dentary could relate to morphological integration and modular organization within this structure. Traits interacting during development or function tend to © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 208 E. BARROW and N. MACLEOD be inherited together and therefore evolve together (Olson & Miller, 1958; Cheverud, 1996). Functional interactions between parts may affect performance of the dentary, or developmental interactions may occur during dentary formation. When integration is stronger between some subsets of traits than others they form ‘quasi-independent’ modules (Schlosser & Wagner, 2004). An example might be the horizontal ramus and coronoid process outline shape differences in Scalopus compared with those of Scapanus and Parascalops. The latter two genera closely resemble one another in these parts of the dentary, but Scalopus differs in both parts and it is possible that the shape of one is, to a certain extent, dependent on the other in order to function sufficiently. Modules exist at different organizational levels within an organism and their independence depends on the level at which they are delimited (Bolker, 2000; Duarte et al., 2000; Klingenberg et al., 2003; Klingenberg, Leamy & Cheverud, 2004; Hallgrímsson et al., 2005; Monteiro et al., 2005). Klingenberg et al. (2003) identified separate integrated units in the anterior and posterior regions of the mouse dentary. Different developmental processes act on these parts (Cheverud, 1996). Therefore, it is likely that traits within the ascending ramus will be more strongly integrated to one another than they will be to the horizontal ramus. The modular structure of pleiotropic quantitative trait loci (QTL) effects have also been studied in mouse mandible morphology and is it is likely that the modular structure also exists at the level of the processes of the ascending ramus (Klingenberg et al., 2003, Klingenberg et al., 2004). Closely related species are expected to share similar patterns of covariation. But in order for a functioning structure to be maintained greater integration between certain traits may exist, regardless of phylogenetic relatedness. Functionally integrated traits within the dentary have been identified across intergeneric, interspecific and within populations in the rodent family Echimyidae (Monteiro et al., 2005). Similarly, traits associated with muscle attachment sites in the mole’s sister taxa, shrews, are more strongly integrated than those that are not functionally related and the relationship persists at different taxonomic levels (Badyaev & Foresman, 2004; Badyaev, Foresman & Young, 2005; Young & Badyaev, 2006). Covariation between different parts of the mole dentary requires further study to identify patterns of morphological integration in this functionally diverse group. CONCLUSIONS Detailed analysis of individual regions of the dentary has revealed important aspects of variation across mole genera. This would not have been possible had the dentary been considered as a whole. Our results showed that morphological variation occurs to different extents in individual parts of the dentary. The condylar process shape showed least variation between genera, the coronoid process and angular process shapes showed greatest convergence among genera, and variation in horizontal ramus shape provided the clearest distinctions between genera. Horizontal ramus shape was more strongly influenced by taxonomic–phylogenetic patterns of variation than other parts of the dentary. Although variation was closely correlated with taxonomically important dentition differences, phylogenetic relatedness could not be used to account for variation between all genera. For example, differences in horizontal ramus shape between Urotrichus and Neurotrichus (not closely related) are very similar to those that separate Scalopus from the closely related Scapanus and Parascalops. Variation in the ascending ramus shapes are more strongly influenced by functional controls than taxonomic or phylogenetic relatedness, particularly in coronoid process shape where allometric effects may also be responsible for differences among closely related genera (e.g. Galemys and Desmana). We found the influence of these various factors even more apparent when variation was examined among the two fully fossorial clades. The horizontal ramus shape provided the clearest distinction between the two clades, reflecting their phylogenetic separation. We documented shared patterns of variation in ascending ramus shapes between the two clades, corresponding to their similar functional roles. Within each clade the results revealed distinctions between genera in coronoid process (and to a lesser degree angular process) shapes, including between subspecies. One explanation for discrimination at this level concerns the time since these taxa diverged (it is possible they have not had time to converge) and, as implied by variation in Scalopus, may also be associated with size differences. Dentaries are among the most common elements in the talpid fossil record. Therefore, results of this study could be useful in providing interpretations of fossil as well as modern talpid biology. The present study could be expanded to include fossil taxa and the patterns of morphological variation might help identify evolutionary guilds (for an example, see MacLeod & Rose, 1993). Additionally, measuring discrete sections of bone structure might be well suited to studying fossil records where structures are all too often incomplete. Closely related genera within this study share similar lifestyles and diets. It is therefore difficult to determine the effects of function (feeding habits and the effects on masticatory muscles) over phylogeny on dentary morphology. This will require additional study. © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY Insight into intrageneric and intraspecific shape variation among Talpidae, for example in diverse genera such as Mogera (Abe, 1999), may also provide a potential avenue for future study. This study may also aid in the discovery of new phylogenetic characters as well as help define character states and features of potential phylogenetic value (as recommended by MacLeod, 2002 and Sánchez-Villagra et al., 2006). Larger sample sizes are required to clarify boundaries between genera and variation at the species level, in the hope of discovering characters that would be useful in phylogenetic studies. Examining the pattern of morphological integration could be particularly interesting in understanding the evolution of this structure and could be studied at developmental as well as evolutionary levels. Integrated traits are selected for together and therefore their use as cladistic characters is questioned (Strait, 2001; Ackermann, 2005). Characters that are not independent might bias the results of such analyses. ACKNOWLEDGEMENTS We thank Daphne Hills, Louise Tomsett and Paula Jenkins (Natural History Museum, London) and Robert Asher and Willi Willborn (Museum für Naturkunde, Berlin) for providing access to the material examined in this study. A research visit to Berlin was funded by SYNTHESYS whose support is gratefully acknowledged. All eigenshape analyses were carried out using programs written by N.M. (available from http://www.nhm.ac.uk/hosted_sites/paleonet/ftp/ ftp.html). We also thank two anonymous reviewers for their useful comments and especially thank Marcelo Sánchez-Villagra (Natural History Museum, London) for his help and advice throughout this study and his comments on previous versions of the manuscript. REFERENCES Abe H. 1999. Nation-wide investigation of moles in Japan – geographic and interspecific variations of an insectivore. In: Yokohata Y, Nakamura S, eds. Recent Advances in the Biology of Japanese Insectivora. Proceedings of the symposium on the biology of insectivores in Japan and on the wildlife conservation. Hiba: Hiba Society of Natural History, Shobara, 15–19. Ackermann RR. 2005. Ontogenetic integration of the hominoid face. Journal of Human Evolution 48: 175–197. Allen GM. 1938. The mammals of China and Mongolia. Natural history of Central Asia, vol. XI, part 1. New York: American Museum of Natural History. Arnqvist G, Martensson T. 1998. Measurement error in geometric morphometrics: empirical strategies to assess and reduce its impact on measures of shape. Acta Zoologica Academiae Scentiarum Hungarica 44: 73–96. 209 Asher RJ, Novacek MJ, Geisler JH. 2003. Relationships of endemic African mammals and their fossil relatives based on morphological and molecular evidence. Journal of Mammalian Evolution 10: 131–194. Atchley WR. 1993. Genetic and developmental aspects of variability in the Mammalian Mandible. In: Hanken J, Hall BK, eds. The vertebrate skull. Chicago: University of Chicago Press, 207–247. Atchley WR, Cowley DE, Vogl C, McLellan T. 1992. Evolutionary divergence, shape change, and genetic correlation structure in the rodent mandible. Systematic Biology 41: 196–221. Atchley WR, Hall BK. 1991. A model for development and evolution of complex morphological structures. Biological Review 66: 101–157. Badyaev AV, Foresman KR. 2004. Evolution of morphological integration. I. Functional units channel stress-induced variation in shrew mandibles. American Naturalist 163: 868–879. Badyaev AV, Foresman KR, Young RL. 2005. Evolution of morphological integration: developmental accommodation of stress-induced variation. American Naturalist 166: 382–395. Björklund M, Merilä J. 1993. Morphological differentiation in Carduelis finches: adaptive vs. constraint models. Journal of Evolutionary Biology 6: 359–373. Bolker JA. 2000. Modularity in development and why it matters to evo-devo. American Zoologist 40: 770–776. Bookstein FL. 1991. Morphometric tools for landmarks data: geometry and biology. Cambridge: Cambridge University Press. Cabria MT, Rubines R, Gómez-Moliner B, Zardoya R. 2006. On the phylogenetic position of a rare Iberian endemic mammal, the Pyrenean desman (Galemys pyrenaicus). Gene 375: 1–13. Catania KC. 2002. The nose takes a starring role. Scientific American 2002: 54–59. Catania KC, Remple FE. 2005. Asymptotic prey profitability drives star-nosed moles to the foraging speed limit. Nature 433: 519–522. Cheverud JM. 1996. Developmental integration and the evolution of pleiotropy. American Zoologist 36: 44–50. Costa LF, Cesar RM. 2000. Shape analysis and classification: theory and practice. New York: CRC Press. Dayan T, Simberloff D. 1994. Character displacement, sexual dimorphism, and morphological variation among British and Irish mustelids. Ecology 75: 1063–1073. Duarte LC, Monteiro LR, von Zuben FJ, dos Reis SF. 2000. Variation in mandible shape in Trichomys apereoides (Mammalia: Rodentia): geometric analysis of a complex morphological structure. Systematic Biology 49: 563–578. Gaughran GRL. 1954. A comparative study of the osteology and myology of the cranial and cervical regions of the shrew, Blarina brevicauda, and the mole, Scalopus aquaticus. Miscellaneous Publications Museum of Zoology University of Michigan 80: 1–82. Grand T, Gould E, Montali R. 1998. Structure of the proboscis and rays of the star-nosed mole, Condylura cristata. Journal of Mammology 79: 492–501. © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 210 E. BARROW and N. MACLEOD Greenacre MJ. 1984. Theory and applications of correspondence analysis. London: Academic Press. Grüneberg H. 1952. The genetics of the mouse. Hague: Martinus-Nijhoff. Hall BK. 2003. Unlocking the black box between genotype and phenotype. Biology and Philosophy 18: 219–247. Hall BK. 2005. Bones and cartilage: developmental and evolutionary skeletal biology. Amsterdam: Elsevier Academic Press. Hallgrímsson B, Yardley Brown JJ, Hall KH. 2005. The study of phenotypic variability: an emerging research agenda for understanding the developmental-genetic architecture underlying phenotypic variation. In: Hallgrímsson B, Hall BK, eds. Variation: a central concept in biology. Amsterdam: Elsevier Academic Press, 525–551. Hildebrand M. 1982. Analysis of vertebrate structure, 2nd edn. New York: John Wiley and Sons, Inc. Humphrey LT, Dean MC, Stringer CB. 1999. Morphological variation in great ape and modern human mandibles. Journal of Anatomy 195: 491–513. Hutchison JH. 1968. Fossil Talpidae (Insectivora, Mammalia) from the later Tertiary of Oregon. Bulletin of the Museum of Natural History of the University of Oregon 11: 1–117. Hutchison JH. 1974. Notes on type specimens of European Miocene Talpidae and a tentative classification of Old World Tertiary Talpidae (Insectivora: Mammalia). Geobios 7: 211– 256. Hutchison JH. 1976. The Talpidae (Insectivora, Mammalia): evolution, phylogeny, and classification. PhD Thesis, University of California, Berkeley. Hutterer R. 2005. Soricomorpha. In: Wilson DE, Reeder DM, eds. Species of the world: a taxonomic and geographic reference. Baltimore: Johns Hopkins University Press, 220–311. Hylander WL. 2005. The functional significance of primate mandibular form. Journal of Morphology 160: 223–239. Klingenberg CP. 2005. Developmental constraints, modules and evolvability. In: Hallgrímsson B, Hall BK, eds. Variation: a central concept in biology. Amsterdam: Elsevier Academic Press, 219–247. Klingenberg CP, Leamy LJ, Cheverud JM. 2004. Integration and modularity of quantitative trait locus effects on geometric shape in the mouse mandible. Genetics 166: 1909–1921. Klingenberg CP, Mebus K, Auffray JC. 2003. Developmental integration in a complex morphological structure: how distinct are the modules in the mouse mandible? Evolution and Development 5: 522–531. MacLeod N. 1999. Generalizing and extending the eigenshape method of shape space visualization and analysis. Paleobiology 25: 107–138. MacLeod N. 2002. Phylogenetic signals in morphometric data. In: MacLeod N, Forey PL, eds. Morphology, shape and phylogeny. Systematics Association Special Volume. London: Taylor & Francis, 100–138. MacLeod N. 2006. Rs and Qs II: correspondence analysis. Palaeontological Association Newsletter 62: 60–74. MacLeod N, Rose KD. 1993. Inferring locomotor behaviour in Paleogene mammals via eigenshape analysis. American Journal of Science 293: 300–355. Monteiro LR, Bonato V, dos Reis SF. 2005. Evolutionary integration and morphological diversification in complex morphological structures: mandible shape divergence in spiny rats (Rodentia, Echimyidae). Evolution and Development 7: 429–439. Monteiro LR, dos Reis SF. 2005. Morphological evolution in the mandible of spiny rats, genus Trinomys (Rodentia: Echimyidae). Journal of Zoological Systematics and Evolutionary Research 43: 332–338. Moore DW. 1986. Systematic and biogeographic relationships among the Talpinae (Insectivora: Talpidae). PhD Dissertation, University of New Mexico. Motokawa M. 2004. Phylogenetic relationships within the family Talpidae (Mammalia: Insectivora). Journal of Zoology (London) 263: 147–157. Motokawa M, Lin LK, Cheng HC, Harada M. 2001. Taxonomic status of the Senkaku mole, Nesoscaptor uchidai, with special reference to variation in Mogera insularis from Taiwan (Mammalia: Insectivora). Zoological Science 18: 733–740. Nowak RN. 1999. Walker’s mammals of the world, Vol. 1, 6th edn. Baltimore: The John Hopkins University Press. Okamoto M. 1997. Phylogeny of Japanese moles inferred from mitochondrial CO1 gene sequences. In: Recent advances in the biology of Japanese insectivora. Proceedings of the symposium on the biology of insectivores in Japan and on the wildlife conservation. Hiba: Hiba Society of Natural History. Olson EC, Miller RL. 1958. Morphological integration. Chicago: University of Chicago Press. Polly PD. 2008. Adaptive zones and the pinniped ankle: a 3D quantitative analysis of carnivoran tarsal evolution. In: Sargis E, Dagosto M, eds. Mammalian evolutionary morphology: a tribute to Frederick S. Szalay. Dordrecht: Springer, 165–194. Raw F. 1966. The soil fauna as a food source for moles. Journal of Zoology (London) 149: 50–54. Rees RW. 2005. Morphologic variation in the mandible of the white-tailed deer (Odocoileus virginianus): a study of populational skeletal variation by principal component and canonical analyses. Journal of Morphology 128: 113–130. Reyment RA. 1991. Multidimensional paleobiology. Oxford: Pergamon Press. Reyment RA, Jöreskog KG. 1993. Applied factor analysis in the natural sciences. Cambridge: Cambridge University Press. Rohlf FJ. 1990. Fitting curves to outlines. In: Rohlf FJ, Bookstein FL, eds. Proceedings of the Michigan Morphometrics Workshop. The University of Michigan Museum of Zoology, Special Publication No. 2, 167–177. Rohlf FJ. 2003. tpsDig, relative warps analysis, Version 1.36. Stony Brook: Department of Ecology and Evolution, State University of New York at Stony Brook. Rohlf FJ. 2004. tpsDig, digitize landmarks and outlines, Version 2.0. Stony Brook: Department of Ecology and Evolution, State University of New York at Stony Brook. Rohlf FJ, Loy A, Corti M. 1996. Morphometric analysis of © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211 SHAPE VARIATION IN THE MOLE DENTARY Old World Talpidae (Mammalia, Insectivora) using partialwarp scores. Systematic Biology 45: 344–362. Sánchez-Villagra MR, Horovitz I, Motokawa M. 2006. A comprehensive morpholgical analysis of talpid moles (Mammalia) phylogenetic relationships. Cladistics 22: 59–88. Schlosser G, Wagner GP. 2004. Modularity in development and evolution. Chicago: University of Chicago Press. Schluter D. 1996. Adaptive radiation along genetic lines of least resistance. Evolution 50: 1766–1774. Shinohara A, Campbell KL, Suzuki H. 2003. Molecular phylogenetic relationships of moles, shrew moles, and desmans from the new and old worlds. Molecular Phylogenetics and Evolution 27: 247–258. Shinohara A, Suzuki H, Tsuchiya K, Zhang Y-P, Luo J, Jiang X-L, Wang Y-X, Campbell KL. 2004. Evolution and biogeography of talpid moles and continental East Asia and the Japanese islands inferred from mitochondrial and nuclear gene sequences. Zoological Science 21: 1177–1185. Silcox MT, Teaford MF. 2002. The diet of worms: an analysis of mole dental microwear. Journal of Mammology 83: 804–814. Slijper EJ. 1946. Comparative biologic–anatomical investigations on the vertebral column and spinal musculature of mammals. Amsterdam: North-Holland Publishing Co. Strait DS. 2001. Integration, phylogeny, and the hominid cranial base. American Journal of Physical Anthropology 114: 273–297. Tomo S, Ogita M, Tomo I. 1997. Development of mandibular cartilages in the rat. Anatomical Record 249: 233–239. True FW. 1896. A revision of the American moles. Proceedings of the US National Museum 29: 1–115. Van Valen L. 1967. New Paleocene insectivores and insectivore classification. Bulletin of the American Museum of Natural History 135: 217–284. Waddell PJ, Okada N, Hasegawa M. 1999. Towards resolving the interordinal relationships of placental mammals. Systematic Biology 48: 1–5. Whidden HP. 2000. Comparative myology of moles and the phylogeny of the Talpidae (Mammalia, Lipotyphla). American Museum Noviates 3294: 1–53. Yates TL, Moore DW. 1990. Speciation and evolution in the family Talpidae (Mammalia: Insectivora). In: Nevo E, Reig OA, eds. Evolution of subterranean mammals at the organismal and molecular levels. New York: Wiley-Liss, 1–22. Yates TL, Schmidly DJ. 1978. Mammalian species. Scalopus aquaticus. American Society of Mammalogists 105: 1–4. Young R, Badyaev AV. 2006. Evolutionary persistence of phenotypic integration: influence of developmental and functional relationships on complex trait evolution. Evolution 60: 1291–1299. Zahn CT, Roskies RZ. 1972. Fourier shape descriptors for closed plane curves: IEEE Transactions on Computers v. C-21: 269–281. Zelditch ML, Swiderski DL, Sheets HD, Fink WL. 2004. Geometric morphometrics for biologists: a primer. Amsterdam: Elsevier Academic Press. Ziegler AC. 1971. Dental homologies and possible relationships of recent Talpidae. Journal of Mammology 52: 50–68. 211 Ziegler R. 2003. Moles (Talpidae) from the late Middle Miocene of South Germany. Acta Palaeontologica Polonica 48: 617–648. APPENDIX 1 Estimates of repeatability of shape variables explaining up to 90% of the variation in each region of the dentary. R1 is calculated from repeated measures taken from the same images of each individual and R2 included repeated measures taken from separate images of the same individual. Variable Coronoid process ES-1 ES-2 ES-3 ES-4 ES-5 ES-6 ES-7 ES-8 ES-9 ES-10 ES-11 Condylar process ES-1 ES-2 ES-3 ES-4 ES-5 ES-6 ES-7 ES-8 ES-9 ES-10 Angular process ES-1 ES-2 ES-3 ES-4 ES-5 ES-6 ES-7 ES-8 ES-9 ES-10 ES-11 ES-12 Horizontal ramus RW-1 RW-2 RW-3 RW-4 % Cumulative % R1 R2 37.5 19.0 11.2 7.9 4.1 3.5 2.2 1.4 1.2 1.2 1.1 37.5 56.5 67.7 75.6 79.8 83.3 85.5 86.9 88.2 89.4 90.4 0.98 0.90 0.99 0.79 0.98 0.98 0.82 0.85 0.36 0.33 0.60 0.98 0.92 0.99 0.83 0.96 0.96 0.73 0.86 0.38 0.22 0.60 65.4 10.3 5.6 2.2 2.0 1.6 1.1 1.0 0.7 0.7 65.4 75.7 81.3 83.5 85.5 87.1 88.1 89.2 89.9 90.6 0.95 0.83 0.96 0.77 0.64 0.65 0.45 0.49 0.20 0.29 0.97 0.87 0.95 0.83 0.72 0.75 0.49 0.47 0.12 0.16 41.6 14.6 11.8 7.4 5.6 2.0 1.6 1.6 1.4 1.2 0.8 0.7 41.6 56.2 68.0 75.4 81.0 83.0 84.7 86.2 87.6 88.8 89.6 90.3 0.99 0.99 0.95 0.99 0.96 0.66 0.84 0.83 0.82 0.89 0.69 0.65 0.98 0.98 0.96 0.98 0.96 0.46 0.71 0.82 0.82 0.87 0.73 0.65 45.6 28.4 15.0 5.4 45.6 73.9 88.9 94.3 0.95 0.99 0.99 0.94 0.95 0.98 0.98 0.93 © 2008 Trustees of the Natural History Museum (London) Journal compilation © 2008 The Linnean Society of London, Zoological Journal of the Linnean Society, 2008, 153, 187–211
© Copyright 2026 Paperzz