Evaluation of Long Bone Surface Textures as Ontogenetic Indicators in Centrosaurine Ceratopsids


  • Allison R. Tumarkin-Deratzian

    Corresponding author
    1. Department of Earth and Environmental Science, Temple University, Beury Hall, Philadelphia, Pennsylvania
    • Department of Earth and Environmental Science, Beury Hall, 1901 North 13th Street, Philadelphia, PA 19122
    Search for more papers by this author
    • Fax: 215-204-6081


The search for criteria for aging non-mammalian fossil vertebrates has preoccupied paleobiologists in recent years. Previous studies of the long bones of pterosaurs and modern and subfossil birds as well as of cranial material of centrosaurine ceratopsid dinosaurs have documented variations in surface textures that seem to be ontogenetically related. In this study, long bones from the centrosaurine ceratopsid genera Centrosaurus, Einiosaurus, and Pachyrhinosaurus are examined to test the hypothesis that changes in bone surface textures and reduction of surface porosity could be correlated with size (and presumably age) classes, as has been previously documented in pterosaurs and birds. The data set includes 141 bones representing all six long bone elements, collected from monodominant centrosaurine bone beds. Bone surface patterns are documented by macroscopic visual examination, and a sequence of five texture classes ordered by decreasing surface porosity is described based on the common distributions of these patterns. Calculations of Spearman's rank correlation coefficients reveal significant correlations between texture class and size. The smallest bones are invariably associated with porous midshaft textures that grade to fibrous and long-grained patterns proximally and distally [Texture Class (TC) 1]. Post-hoc analysis after Kruskal–Wallis ANOVA on ranks confirms that the mean size of TC1 bones is, in most cases, significantly different than the mean size of bones in other texture classes. Results of this study suggest the presence of an ontogenetic surface textural signal in centrosaurine long bones; however, comparison of texture classes with size-independent maturity criteria is needed to clarify further the potential ontogenetic significance of higher texture classes. Anat Rec, 292:1485–1500, 2009. © 2009 Wiley-Liss, Inc.


Absence of soft tissue anatomy and absolute age data makes reliable determination of ontogenetic status for fossil remains considerably more difficult than for extant organisms. Differentiation of juvenile, subadult, and adult individuals, however, is as critical an issue in paleontological as in neontological research, particularly for phylogenetic and paleoecological studies where it is necessary to distinguish between juveniles of larger taxa and adults of smaller, related taxa. Within a single taxon, size-based distinctions may be useful, because, on average, subadults can be expected to be larger than juveniles, and fully grown adults to be larger than subadults (Ryan et al., 2001). Over-reliance on size as the sole age proxy may become problematic, however, if there is a high degree of intraspecific size variation (Horner and Padian, 2004; Sander and Klein, 2005; Bybee et al., 2006), or in comparisons of multiple taxa.

A variety of nonsize-based criteria have been used to evaluate skeletal ontogenetic stages in modern and fossil taxa. These include ossification of limb bone ends (e.g., Johnson, 1977; Brinkman, 1988; Bennett, 1993); fusion of long bone epiphyses and compound elements of the skull and axial skeleton (e.g., Carey, 1982; Sadler, 1991; Brochu, 1996; Gotfredsen, 1997; Sampson et al., 1997; Ryan et al., 2001; Carrano et al., 2005; Tumarkin-Deratzian et al., 2006; Irmis, 2007); relative development of suspected sexual display structures such as cranial ornamentation (e.g., Dodson, 1975, 1976; Sampson et al., 1997; Ryan et al., 2001; Ryan and Russell, 2005; Goodwin et al., 2006; Horner and Goodwin, 2006, 2008), and bone microstructure (e.g., Bennett, 1993; Chinsamy, 1995; Curry, 1999; Erickson and Tumanova, 2000; Horner et al., 2000; Sander, 2000; Steyer et al., 2003; Erickson et al., 2004; Horner and Padian, 2004; Padian et al., 2004; Ray and Chinsamy, 2004; Botha and Chinsamy, 2004, 2005; Ray et al., 2005; Bybee et al., 2006; Reizner and Horner, 2006; Lee, 2006, 2007). An additional method relies on the fact that bones from individuals of different ages often possess distinctive surface textures that macroscopically reflect changes in microscopic ossification patterns in successive relative maturity stages (Bennett, 1993; Tumarkin-Deratzian et al., 2006; Tumarkin-Deratzian, 2007, in press).

It is important to stress that in this context “maturity” refers to development of the skeleton toward somatic (specifically osteological) maturity, and does not presume to convey information about an individual's sexual or reproductive status. This distinction is crucial, because the relative timing of somatic and reproductive maturation varies among extant amniotes, with some groups such as squamates, crocodilians and large-bodied mammals attaining sexual maturity before somatic maturity (Chabreck and Joanen, 1979; Andrews, 1982; Shine and Charnov, 1992; Wilkinson and Rhodes, 1997; Lee and Werning, 2008). Others such as birds and small-bodied mammals follow the reverse pattern (Craighead and Stockstad, 1964; Ricklefs, 1968; Owen, 1980; Tumarkin-Deratzian et al., 2006; Lee and Werning, 2008).

Although specific bone surface textures and patterns of textural change are not universal across anatomical elements and taxonomic groups, the general trend is one of decreasing overall surface porosity and increased “finishing” of bone surfaces as growth rates slow with increasing size and osteological maturity (Johnson, 1977; Callison and Quimby, 1984; Bennett, 1993; Sampson et al., 1997; Ryan et al., 2001; Tumarkin-Deratzian et al., 2006). Histologically, this trend has been shown to be associated with changes in the extent and organization of periosteal vasculature as bones mature and bone growth (both longitudinal and appositional) decreases (Bennett, 1993; Horner and Currie, 1994; Tumarkin-Deratzian et al., 2006; Tumarkin-Deratzian, 2007, in press).

Not all groups examined to date show consistent patterns of ontogenetic bone textural change. Diagnostic ontogenetic trends have been well documented in extant birds (Callison and Quimby, 1984; Benecke, 1993; Cohen and Serjeantson, 1996; Gotfredsen, 1997; Sanz et al., 1997; Serjeantson, 1998, 2002; Mannermaa, 2002; Tumarkin-Deratzian et al., 2006) but seem to be absent in crocodilians. Tumarkin-Deratzian et al. (2007) hypothesized that an absence of predictable textural changes in a growth series of Alligator mississippiensis long bones may reflect a high degree of individual growth variation stemming from sensitivity to local and regional environmental conditions coupled with long periods of growth before attainment of somatic maturity. Among fossil amniotes, surface textural differences that permit reliable differentiation of immature and mature bone have been noted in basal synapsids (Brinkman, 1988), ichthyosaurs (Johnson, 1977), pterosaurs (Bennett, 1993), non-avian dinosaurs (Callison and Quimby, 1984; Horner and Currie, 1994; Jacobs et al., 1994; Sampson et al., 1997; Carr, 1999; Brill and Carpenter, 2001; Ryan et al., 2001; Ryan and Russell, 2005; Brown, 2006; Tumarkin-Deratzian, 2003, 2007, in press; Brown et al., 2007, 2009) and birds (Sanz et al., 1997).

The short-frilled centrosaurine ceratopsid dinosaurs are excellent candidates for analysis of ontogenetic trends, because several genera (Centrosaurus, Einiosaurus, Pachyrhinosaurus, Styracosaurus, and possibly Albertaceratops) are known from monodominant bone bed assemblages comprising remains of multiple individuals from a range of sizes and presumably ages (Langston, 1975; Currie and Dodson, 1984; Rogers, 1990; Sampson, 1995; Ryan et al., 2001; Dodson et al., 2004; Eberth and Getty, 2005; Ryan and Evans, 2005; Ryan and Russell, 2005; Ryan, 2007; Currie et al., 2008). Moreover, current taxonomy of centrosaurines is heavily based on cranial ornamentation features, which develop comparatively late in ontogeny (Sampson et al., 1997; Ryan et al., 2001; Dodson et al., 2004). Reliable recognition of skeletal markers of different ontogenetic stages is therefore also critical for centrosaurine phylogenic studies, because unornamented individuals could conceivably represent somatically immature representatives of ornamented taxa, or distinct basal and/or paedomorphic forms lacking strong ornamentation. Such is the case with the problematic centrosaurines Monoclonius, Brachyceratops, and Avaceratops, which have been variously considered as distinct taxa (Dodson, 1990; Dodson and Currie, 1990; Tumarkin and Dodson, 1998; Penkalski and Dodson, 1999; Chinnery, 2001 [Avaceratops]; Tumarkin-Deratzian and Dodson, 2005), nomina dubia (Sampson et al., 1997; Ryan et al., 2001; Chinnery, 2001 [Monoclonius, Brachyceratops]), and “possible Centrosaurinae” (Table 23.1 in Dodson et al., 2004).

Several authors have examined diagnostic ontogenetic textural changes in the centrosaurine skull (Sampson et al., 1997; Ryan et al., 2001; Brown, 2006; Tumarkin-Deratzian, 2003, 2007, in press; Brown et al., 2007, 2009), and have defined three relative age classes based on a consistent association of distinctive surface textures with particular size ranges, degrees of element fusion, and extent of cranial ornamentation (Sampson et al., 1997; Ryan et al., 2001). Juvenile-sized bones are characterized by long-grained texture, comprising fine striations generally following the direction of longitudinal growth. Subadult-sized bones are typically marked by co-occurrence of long-grained texture with a mottled texture characterized by finely pitted surfaces with an irregular unfinished appearance. Adult-sized bones are typically characterized by finished surfaces lacking both long-grained and mottled textures, although the surface topography may vary from smooth to rugose.

This study begins documentation of surface textures in the centrosaurine postcranial skeleton, which to date has received little attention, through examination of long bones of the genera Centrosaurus, Einiosaurus, and Pachyrhinosaurus. Representative textures are described in detail, and the distributions of textures within each genus and in the total sample are analyzed to assess whether a general pattern of textural change may be reasonably linked with increasing size and/or age classes.


Examined Material

The study sample comprises 141 long bones collected by the Royal Tyrrell Museum of Palaeontology (TMP, Drumheller, Alberta) and Museum of the Rockies (MOR, Bozeman, Montana) from monodominant bone beds of Centrosaurus (both C. apertus and C. brinkmani) (Ryan et al., 2001; Eberth and Getty, 2005; Ryan and Russell, 2005), Pachyrhinosaurus lakustai (Currie et al., 2008), and Einiosaurusprocurvicornis (Rogers, 1990; Sampson, 1995). The sample includes only elements that met the following conditions: adequate unaltered surface was preserved for textural observations, the element was sufficiently complete and/or undistorted to obtain useful measurements for size analysis, and sample size of a given element in a given genus was larger than N = 2. A breakdown of the total sample by taxon and skeletal element is summarized in Table 1. Because of the very small number of examined C. brinkmani long bones, both species of Centrosaurus were analyzed together as a single group.

Table 1. Summary of study sample by taxon and anatomical element
C. apertus1096111046
C. brinkmani1124
E. procurvicornis8816
P. lakustai171472010775

Surface Texture Coding

Elements were examined grossly with the unaided eye and with a 10× hand lens. Textural features of well-preserved surfaces were documented qualitatively through a combination of detailed specimen notes, sketches, and photography. Bones were grouped based on similarity of overall surface type without reference to size measurements; incomplete elements were grouped with more complete bones showing similar surface patterns in preserved regions. Textural classes were then defined based on characteristic surface patterns of these groupings, and ordered based on decreasing surface porosity. Relationships between texture and ontogeny were examined only after establishment of the texture class sequence. (As the purpose of this study was primarily a documentation of general textural trends, relative proportions of different surface patterns within texture classes were not quantified. Development of quantitative definitions for texture classes will be addressed in a separate study.)

Determining Size and Skeletal Maturity

For the purposes of this study, element size was used as a proxy for skeletal maturity. Ideally, size should not be the only criterion used to distinguish relative age classes of skeletal material, as individual growth variations can cause body size variation within age classes (e.g., Horner and Padian, 2004; Sander and Klein, 2005; Bybee et al., 2006). On average, however, body size within a given taxon may be expected to increase from juvenile through subadult to adult stages (Ryan et al., 2001), and therefore size can be useful as an initial indicator of relative ontogenetic status, provided its limitations are considered.

For analysis of subsamples comprising single anatomical elements, the size measurement used was element length in millimeters. Lengths reported for the fibula, humerus, ulna, and radius are maximum lengths. Because of poor preservation of proximal and distal ends of some femora and tibiae, however, maximum length from the proximal-most to distal-most point was not always available for those elements. Reported femoral lengths are measured from the proximal-most point on the femoral head to the distal-most point on the medial condyle. Reported tibial lengths are midline lengths measured from the proximal articular surface to the proximal-most point of the astragalar facet. In most cases, reported lengths were measured directly from the specimens with digital calipers or tape measure, and the mean of three replicates was used as the reported length value.

Lengths could not be directly measured from 15 incomplete elements (five femora, seven tibiae, and three humeri). Correlation and regression analyses using PAST 1.81 (Hammer et al., 2001) were used to predict lengths for these incomplete bones based on comparison with other dimensions that could be measured from both the fragmentary elements and more complete and uncrushed bones of the same genus. For Centrosaurus, femoral length was compared with length of the fourth trochanter; tibial midline length was compared with distal width; and humeral maximum length was compared with proximal and distal widths. (Incomplete Centrosaurus humeri did not consistently preserve either the proximal or the distal ends, necessitating two separate comparisons.) For Einiosaurus, femoral length was compared with shaft width at the distal end of the fourth trochanter; tibial midline length was compared with midshaft width. For Pachyrhinosaurus, tibial midline length was compared with maximum length. Simple linear (Model I) regression was used rather than reduced major axis (Model II) regression, following the recommendation of Sokal and Rohlf (1995) that Model I regression techniques are appropriate when a regression line is being fitted primarily for predictive purposes, even in cases where both the independent and dependent variables are measured with error.

To compare data across anatomical elements, it was necessary to convert bone lengths into a size proxy for an individual animal. For Centrosaurus, percent “adult” size values were calculated for each bone. Following Ryan et al. (2001), Yale Peabody Museum (YPM, New Haven, Connecticut) 2015 (Centrosaurus “flexus”) as measured and figured by Lull (1933) was used as the adult size standard. YPM 2015 is assumed to be a fully or near-fully grown adult individual based on its well-developed frill ornamentation and skeletal measurements falling within the upper size limits known for this genus. Lull's reported femoral length is measured from the proximal end of the greater trochanter to the distal end of the lateral condyle, and tibial length is the maximum length including the astragalus. In these two cases, approximate values for the femoral head-to-medial condyle length and tibial midline length used in this study were calculated by ratios from Lull's measurements and illustrations of YPM 2015.

Without reasonably complete associated adult-size individuals of Einiosaurus and Pachyrhinosaurus, it was necessary to devise another proxy for individual body size for these genera and for combined analysis of all taxa. For this purpose, femur lengths were used as measured or calculated from the regression analyses, and length measurements for other elements were converted to estimated femur lengths for an animal with bones of that size, using ratios of reported element lengths from YPM 2015. This method allows the sizes of all bones from all genera to be expressed as a scaled estimated femur length, which functions as an estimate of the relative size of the animal. It relies on the assumption, however, that long bone length proportions in Einiosaurus and Pachyrhinosaurus are similar to those in Centrosaurus.

Statistical Relationship Between Texture Class and Ontogeny

Relationships between texture class and bone length (for separate elements) or individual size (for genera and total sample) were evaluated using values obtained for Spearman's rank correlation coefficient (rs). This nonparametric statistic tests whether two categories are independent, using rank indices, and therefore does not require assumptions of normality. Relationships among mean sizes in different texture classes were analyzed using Kruskal–Wallis ANOVA on ranks, with post-hoc analysis of differences among groups by pair-wise Mann–Whitney tests. All statistical analyses were performed using PAST 1.81 (Hammer et al., 2001). Post-hoc tests were performed both with and without Bonferroni correction. Performing multiple pair-wise comparisons on the same sample via Mann–Whitney or similar methods raises the risk of Type I errors, in which significant results are obtained simply by chance. Bonferroni-corrected p-values compensate for this by multiplying the normal uncorrected p-value by the number of comparisons being made. It has been argued, however, that the Bonferroni correction compensates too far, raising the risk of Type II error (interpretation of a result as nonsignificant, when in reality it is) to the point where some significant differences may be overlooked entirely (Bland and Altman, 1995; Perneger, 1998).


Types of Surface Patterns

Six distinctive types of surfaces were identified on centrosaurine long bones:

Long-grained texture is characterized by fine sharp subparallel ridges and deep narrow furrows, typically oriented longitudinally or in a radiating pattern that follows the direction of bone growth [Fig. 1(A)]. This texture is equivalent to the long-grained texture previously described on cranial bones of juvenile and subadult-sized centrosaurines (Sampson et al., 1997; Ryan et al., 2001).

Figure 1.

Centrosaurine long bone surface patterns. A: Long-grained, TMP 89.55.1468, humerus, caudal surface of deltopectoral crest. B: Fibrous, TMP 2002.29.02, tibia, metaphyseal region of cranial surface. C: Porous, TMP 89.55.1468, humerus, midshaft cranial surface. D: Co-occurrence of porous and dimpled (arrows), porous dominant, TMP 86.55.63, femur, midshaft cranial surface. E: Co-occurrence of porous and dimpled (arrows), dimpled dominant, TMP 89.55.1507, femur, midshaft caudal surface. F: Grooved, TMP 86.55.200, tibia, distal region of cranial surface. G: Smooth, TMP 89.55.1549, humerus, midshaft caudal surface. Scale bars = 1 cm. All images Pachyrhinosaurus lakustai.

Fibrous texture is similar to long-grained texture, but striations are shorter and less regular [Fig. 1(B)].

Long-grained and fibrous textures are both associated with pores that penetrate the bone surface. They are most common in proximal and distal metaphyseal regions, as well as on the deltopectoral crest of the humerus and olecranon process of the ulna.

Porous texture is characterized by a finely pitted surface [Fig. 1(C)]. Pores are small (≤0.5 mm) and closely spaced, and may be seen to penetrate the bone surface in well-preserved areas. This surface pattern is most common in midshaft regions and often grades into fibrous and/or long-grained textures proximally and distally.

Dimpled texture is characterized by coarse indentations that are larger (millimeter-scale), deeper, and more widely spaced than those of porous texture, and lack association with pores that penetrate the bone surface. It is most common in midshaft regions, where it may occur alone or co-occur with and overprint porous surfaces [Fig. 1(D,E)]. In the latter case, it is usually possible to identify either the porous [Fig. 1(D)] or the dimpled [Fig. 1(E)] pattern as dominant.

Grooved texture is characterized by longitudinally directed ridges and furrows, which are coarser, more rounded, and less regular than those of long-grained texture [Fig. 1(F)]. The grooved pattern may occur midshaft or in proximal and distal regions. In midshaft regions, it may, like dimpled texture, co-occur with and overprint porous texture. Proximally and distally, grooved texture often imparts a corrugated appearance to the bone surface [Fig. 1(F)].

Smooth texture is characterized by a relatively featureless surface lacking any of the other textures [Fig. 1(G)]. It is most common midshaft, but may occur in any region.

Numerical Texture Classes

A sequence of five textural classes ordered based on overall decrease in surface porosity (loss of long-grained, fibrous, and porous surface types associated with penetrating pores) may be defined based on common occurrence and distribution patterns of these six surface types (Table 2). Classes are primarily defined based on midshaft features, although proximal and distal features are included in Table 2 for the sake of completeness. Proximal and distal ends were more likely to be poorly preserved, and surface types in those regions were more variable and difficult to separate into well-marked classes. The five texture classes are largely consistent across elements and genera, although not all classes were observed on all elements of all taxa.

Table 2. Texture classes for centrosaurine long bones, ordered based on overall decrease in midshaft surface porosity
Texture ClassCharacteristic surface features
1Porous presentLong-grained/fibrous present
 Dimpled/grooved absent 
 ±Isolated smooth areas 
2Porous presentLong-grained/fibrous present
 Dimpled/grooved present 
 ±Isolated smooth areas 
3Porous absentLong-grained usually absent
 Dimpled/grooved dominantFibrous present
4Porous absentLong-grained usually absent
 Isolated dimpled/grooved areas±Fibrous
 Smooth dominantDimpled/grooved/smooth present
5Porous absentLong-grained absent
 Dimpled/grooved absent±Fibrous
 Smooth presentDimpled/grooved/smooth present

Texture class 1 (TC1) is characterized by porous texture midshaft, which gradually transitions to fibrous and/or long-grained texture in proximal and distal regions. Dimpled and grooved textures are absent midshaft, but isolated areas of smooth texture may be present. Samples of all examined elements in all taxa contained at least one TC1 bone.

Texture class 2 (TC2) is similar to TC1, but midshaft areas contain a mix of porous and dimpled and/or grooved textures. The latter may occur in patches distinct from porous areas, or may co-occur on the same surface in an overprinted pattern. As in TC1, isolated smooth areas may or may not be present midshaft, and proximal and distal regions bear fibrous and/or long-grained textures. TC2 bones occurred in samples of all examined elements of Centrosaurus and Pachyrhinosaurus, and in the femur of Einiosaurus.

Texture class 3 (TC3) is characterized by an absence of porous texture. Dimpled and/or grooved texture is dominant (occupies greater than 50% area) midshaft, although areas of smooth texture may also be present. Textures in proximal and distal metaphyseal regions, the deltopectoral crest of the humerus, and the olecranon process of the ulna are variable. Fibrous, dimpled, grooved, or smooth textures may be present in those areas, but long-grained texture is usually absent. TC3 was documented in the femur, fibula, humerus, and ulna of Centrosaurus; femur and tibia of Einiosaurus; and fibula and radius of Pachyrhinosaurus.

Texture class 4 (TC4) is dominated by smooth midshaft surfaces (occupying greater than 50% area), although dimpled and/or grooved textures occur in isolated patches. Porous texture is absent. Proximal and distal surfaces vary as in TC3, but again long-grained texture is usually absent. TC4 was documented in the fibula and humerus of Centrosaurus; femur and tibia of Einiosaurus; and femur, tibia, and humerus of Pachyrhinosaurus.

Texture class 5 (TC5) is characterized by completely smooth midshaft surfaces. Porous, dimpled, and grooved textures are absent from midshaft regions. Proximal and distal textures are variable, but long-grained texture is absent. TC5 was documented in the ulna of Pachyrhinosaurus, and femur and tibia of Einiosaurus.

For purposes of assigning elements to texture classes based on midshaft surface features, the following general key, which emphasizes the most important features of each texture class, may be used: TC1, porous texture; TC2, co-occurrence of porous and dimpled/grooved textures; TC3, dimpled/grooved textures with <50% smooth areas; TC4, smooth texture with <50% dimpled/grooved areas; and TC5, smooth texture.

Correlation and Regression Analysis for Prediction of Immeasurable Bone Lengths

Regression line equations, Pearson correlation coefficients, p-values, and standard errors for each comparison are listed in Table 3. Measurement values were log transformed to obtain linear relationships in all cases. Bone lengths derived from the regression equations are indicated with alphabet “a” in Table 4.

Table 3. Regression equations used in calculations of approximate element length values for incomplete bones
 Regression Line Equation (y = ax + b)Pearson's rPNError aError b
  1. DW, distal width; HMCL, head to medial condyle length; MDL, midline length; MDW, midshaft width; ML, maximum length; PW, proximal width; TRL, 4th trochanter length; TRW, shaft width at distal end of 4th trochanter.

 FemurLog HMCL = 0.85139 (log TRL) + 0.892640.921530.02650.207130.42116
 TibiaLog MDL = 0.94432 (log DW) + 0.545310.97503<0.00170.0961840.20379
 HumerusLog ML = 0.85476 (log PW) + 0.777170.971510.00160.104250.21255
 Log ML = 0.94954 (log DW) + 0.601780.98422<0.001100.0603480.12367
 FemurLog HMCL = 0.5881 (log TRW) + 1.53290.98691<0.00160.0480540.094243
 TibiaLog MDL = 0.69816 (log MDW) + 1.31390.998390.00240.0280180.046776
 TibiaLog MDL = 0.96686 (log ML) + 0.0661050.99965<0.001100.00904430.022941
Table 4. Raw data for centrosaurine long bones examined in this study
 NumberTaxonElementTCLength (mm)Other measurements (mm)
  • TC, texture class. Refer to text for discussion of length measurements used for different elements.

  • a

    Indicates lengths estimated using regression equations in Table 3. Other measurements listed as relevant for these specimens and those used in derivation of the regression equations. Measurement abbreviations as defined in Table 3.

TMP79.11.52C. apertusFemur3690 
TMP82.18.68C. apertusFemur1430TRL, 128
TMP89.18.120C. apertusFemur3559aTRL, 151
TMP95.12.146C. apertusFemur3656aTRL, 182
TMP95.176.4C. apertusFemur3787 
TMP95.400.134C. apertusFemur3576 
TMP95.400.144C. apertusFemur1331TRL, 77
TMP95.401.8C. apertusFemur2362TRL, 99
TMP95.400.48C. apertusFemur3625TRL, 152
TMP96.173.3C. apertusFemur2470aTRL, 123
TMP79.11.161C. apertusFibula4516 
TMP82.18.138C. apertusFibula188 
TMP87.18.19C. apertusFibula3433 
TMP95.401.11C. apertusFibula4494 
TMP95.401.3C. apertusFibula4517 
TMP96.176.19C. apertusFibula3508 
TMP80.18.302C. apertusHumerus1340aPW, 113
TMP95.175.98C. apertusHumerus2436DW, 125
TMP95.400.107C. apertusHumerus1172DW, 57; PW, 57
TMP95.400.12C. apertusHumerus1209DW, 67; PW, 64
TMP95.400.56C. apertusHumerus2348DW, 112
TMP95.401.21C. apertusHumerus2439DW, 122; PW, 115
TMP95.401.19C. apertusHumerus4547DW, 198
TMP96.173.2C. apertusHumerus3551DW, 190; PW, 218
TMPP79.11.61C. apertusHumerus2267DW, 80
TMPP80.24.4aC. apertusHumerus4359aDW, 114
TMPP86.18.75C. apertusHumerus1236DW, 74; PW, 75
TMP79.11.33C. apertusTibia1338aDW, 126
TMP79.11.56C. apertusTibia1362DW, 127
TMP80.18.92C. apertusTibia2272DW, 104
TMP81.18.199C. apertusTibia1237DW, 83
TMP95.400.234C. apertusTibia1320DW, 136
TMP95.400.25C. apertusTibia2378DW, 140
TMP95.400.41C. apertusTibia1391DW, 144
TMP98.12.66C. apertusTibia1154aDW, 55
TMP96.173.1C. apertusTibia2573DW, 214
TMP79.11.99C. apertusUlna3255 
TMP80.18.111C. apertusUlna3264 
TMP81.18.179C. apertusUlna2273 
TMP89.18.38C. apertusUlna1313 
TMP91.18.23C. apertusUlna1183 
TMP95.400.16C. apertusUlna1284 
TMP95.400.28C. apertusUlna2289 
TMP95.400.36C. apertusUlna2283 
TMP95.401.42C. apertusUlna2320 
TMP95.401.66C. apertusUlna1207 
TMP2002.68.78C. brinkmaniFemur2398TRL, 96
TMP2002.68.76C. brinkmaniFibula2490 
TMP2002.68.81C. brinkmaniHumerus2429aPW, 148
TMP2002.68.83C. brinkmaniHumerus2547DW, 176; PW, 207
MOR456E. procurvicornisFemur4533aTRW, 107
MOR373-7-27-6-1E. procurvicornisFemur1446TRW, 71
MOR373-8-6-6-3E. procurvicornisFemur3571TRW, 133
MOR456-8-10-7-14E. procurvicornisFemur2452TRW, 77
MOR456-8-10-7-32E. procurvicornisFemur1289TRW, 41
MOR456-8-18-6-1E. procurvicornisFemur3644TRW, 145
MOR456-8-19-87-3E. procurvicornisFemur4567TRW, 119
MOR456-B2-108E. procurvicornisFemur5506aTRW, 98
MOR456E. procurvicornisTibia4468aMDW, 88
MOR373-8-19-85E. procurvicornisTibia1287MDW, 43
MOR456-8-10-6-1E. procurvicornisTibia1231MDW, 31
MOR456-8-27-87E. procurvicornisTibia1235MDW, 34
MOR456-8-8-87-15E. procurvicornisTibia3413aMDW, 73
MOR456-A2-55E. procurvicornisTibia4418aMDW, 75
MOR456-A2-90E. procurvicornisTibia5435aMDW, 79
MOR456-B3-124E. procurvicornisTibia5495MDW, 95
TMP86.55.246P. lakustaiFemur2412 
TMP86.55.63P. lakustaiFemur2388 
TMP87.55.114P. lakustaiFemur2397 
TMP87.55.225P. lakustaiFemur1254 
TMP87.55.226P. lakustaiFemur2370 
TMP87.55.307P. lakustaiFemur2393 
TMP87.55.5P. lakustaiFemur1273 
TMP88.55.118P. lakustaiFemur2280 
TMP88.55.17P. lakustaiFemur1351 
TMP88.55.210P. lakustaiFemur2373 
TMP88.55.71P. lakustaiFemur2441 
TMP89.55.1425P. lakustaiFemur2389 
TMP89.55.1445P. lakustaiFemur4674 
TMP89.55.1507P. lakustaiFemur2675 
TMP89.55.362P. lakustaiFemur1231 
TMP89.55.384P. lakustaiFemur4617 
TMP89.55.39P. lakustaiFemur2382 
TMP86.55.216P. lakustaiFibula1254 
TMP86.55.88P. lakustaiFibula2500 
TMP87.55.182P. lakustaiFibula2496 
TMP87.55.192P. lakustaiFibula3477 
TMP87.55.325P. lakustaiFibula1306 
TMP89.55.1422P. lakustaiFibula2429 
TMP89.55.555P. lakustaiFibula2434 
TMP86.55.119P. lakustaiHumerus2563 
TMP86.55.161P. lakustaiHumerus1239 
TMP86.55.186P. lakustaiHumerus1297 
TMP86.55.199P. lakustaiHumerus1249 
TMP86.55.43P. lakustaiHumerus1303 
TMP86.55.58P. lakustaiHumerus2335 
TMP86.55.73P. lakustaiHumerus4535 
TMP87.55.265P. lakustaiHumerus2310 
TMP88.55.61P. lakustaiHumerus2538 
TMP89.55.1301P. lakustaiHumerus2371 
TMP89.55.1349P. lakustaiHumerus2312 
TMP89.55.1468P. lakustaiHumerus1194 
TMP89.55.1549P. lakustaiHumerus2570 
TMP89.55.250P. lakustaiHumerus1205 
TMP89.55.400P. lakustaiHumerus2290 
TMP89.55.407P. lakustaiHumerus2556 
TMP89.55.470P. lakustaiHumerus2222 
TMP89.55.510P. lakustaiHumerus1305 
TMP89.55.695P. lakustaiHumerus1287 
TMP89.55.940P. lakustaiHumerus2498 
TMP86.55.264P. lakustaiRadius1229 
TMP86.55.96P. lakustaiRadius1196 
TMP87.55.21P. lakustaiRadius1203 
TMP89.55.1247P. lakustaiRadius3362 
TMP89.55.1313P. lakustaiRadius1205 
TMP89.55.1339P. lakustaiRadius2341 
TMP89.55.1462P. lakustaiRadius1190 
TMP2002.29.02P. lakustaiTibia1191mL, 195
TMP86.55.120P. lakustaiTibia1293 
TMP86.55.169P. lakustaiTibia1294mL, 303
TMP86.55.200P. lakustaiTibia4452mL, 482
TMP87.55.236P. lakustaiTibia1278amL, 288
TMP87.55.256P. lakustaiTibia1297 
TMP87.55.71P. lakustaiTibia1332mL, 344
TMP89.55.1093P. lakustaiTibia2516mL, 555
TMP89.55.1118P. lakustaiTibia2503 
TMP89.55.1383P. lakustaiTibia1308mL, 320
TMP89.55.271P. lakustaiTibia1186mL, 190
TMP89.55.643P. lakustaiTibia2529mL, 547
TMP89.55.666P. lakustaiTibia1204mL, 211
TMP89.55.884P. lakustaiTibia2497mL, 527
TMP86.55.109P. lakustaiUlna1235 
TMP86.55.49P. lakustaiUlna2247 
TMP87.55.176P. lakustaiUlna2210 
TMP87.55.200P. lakustaiUlna2392 
TMP87.55.218P. lakustaiUlna1244 
TMP87.55.263P. lakustaiUlna1225 
TMP87.55.89P. lakustaiUlna5392 
TMP88.55.110P. lakustaiUlna1209 
TMP89.55.1203P. lakustaiUlna2414 
TMP89.55.227P. lakustaiUlna2397 

Relationship Between Texture Class and Size

Plots of the relationship between surface texture class and element length for each of the six long bones (Fig. 2) reveal that for all elements in all genera TC1 is confined to bones less than 500 mm in length, and less than 400 mm for all but the femur. There is a defined break between the maximum size of TC1 and the minimum size of TC3, TC4, and TC5, where recorded, for the femur, tibia, fibula, and radius. There is also a defined break between the maximum size of TC1 and the minimum size of TC2 in the fibula and radius, and in the tibia of Pachyrhinosaurus. When all three genera are considered together, TC2 occupies the widest size range in the femur, tibia, humerus, and ulna. When the genera are considered separately, however, this does not always remain the case: In Centrosaurus, TC1 and TC2 occupy nearly identical size ranges in the femur, and TC1 occupies the widest size range in the ulna. In Einiosaurus, TC1 and TC5 occupy the widest size ranges in the femur and tibia, respectively. In Pachyrhinosaurus, TC1 occupies the widest size range in the tibia and radius, although given the very small sample size of radii (N = 7) and the fact that TC2 and TC3 are each represented by only one bone, this latter result may not be consequential.

Figure 2.

Scatter-plots of the relationship between texture class and bone length in millimeters for each of the six long bones.

Plots of the relationship between texture class of all elements and the size of the individual (based on percent “adult” size in Centrosaurus and scaled estimated femur length for all genera) (Fig. 3) show an overall trend of increase in the minimum body size marked by succeeding texture classes, although the total size ranges occupied by each texture class overlap to a considerable degree. This trend holds in Centrosaurus for all recorded texture classes (TC1 through TC4), and in Einiosaurus and Pachyrhinosaurus for TC1, TC2, and TC3. (As TC2 is represented in Einiosaurus by only one bone, however, the validity of this result for Einiosaurus is questionable.) When the entire data set from all genera is considered together (Fig. 4), there is an overall trend of increase in minimum size from TC1 through TC4 and TC5 (the minimum size of TC4 and TC5 is the same). TC2 occupies the broadest size range, from a scaled estimated femur length of 280 mm up to the largest value in the data set of 971 mm. TC3, TC4, and TC5 occupy similar size ranges.

Figure 3.

Scatter-plots of the relationship between texture class and size of the individual for each genus examined. The relationship in Centrosaurus is the same whether percent “adult” size or scaled estimated femur length is used as the individual size proxy.

Figure 4.

Relationship between texture class and size of the individual for the complete centrosaurine long bone data set, compared with the same relationship in long bones of the extant Canada goose (Branta canadensis). Texture classes are not directly equivalent in the two taxonomic groups, but increasing texture class corresponds to decreasing surface porosity in both. The Branta sample includes femora, tibiotarsi, and humeri (data from Tumarkin-Deratzian et al., 2006).

When each element from each genus is considered separately there is a statistically significant relationship between texture class and bone length for the following elements: femur, fibula, and humerus of Centrosaurus; tibia of Einiosaurus; femur, tibia, humerus and radius of Pachyrhinosaurus (Table 5). When elements are considered across genera, there is a significant relationship for all elements. Analysis by taxon identifies a significant relationship between texture class and the size of the individual in all three genera. A significant relationship is also obtained when the entire data set is analyzed as a single sample, without separation by element or taxon.

Table 5. Spearman's rank correlation coefficients (rs) for analyses of the relationship between texture class and size
Centrosaurus (percent “adult” size)0.62174<0.00150
Centrosaurus (scaled estimated femur length)0.61265<0.00150
Total sample0.60716<0.001141

Results of the Kruskal–Wallis tests and post-hoc analyses (Table 6) reveal significant differences between mean lengths of bones of different texture classes for the femur and humerus of Centrosaurus; and the femur, tibia, and humerus of Pachyrhinosaurus. Post-hoc analysis with Bonferroni correction indicates that in these elements of Pachyrhinosaurus, the mean length of bones in TC1 is distinct from that in TC2. Post-hoc analysis reveals significant differences in mean length of texture classes in Centrosaurus only when Bonferroni correction is not applied; in that case, the mean length of femora in TC2 is distinct from that in TC3, and the mean size of humeri in TC1 is distinct from that in TC2.

Table 6. Statistically significant results of post-hoc analyses after Kruskal–Wallis ANOVA on ranks
 Texture Classes with Significant Mean Size DifferencePN
  • a

    Indicates significant differences in post-hoc analyses with Bonferroni correction; listed probabilities for these differences are Bonferroni-corrected.

 FemurTC2 and TC30.02811
 HumerusTC1 and TC20.02513
 FemuraTC1 and TC20.02217
 TibiaaTC1 and TC20.02114
 HumerusaTC1 and TC20.01320
FemurTC1 and TC20.03136
 aTC1 and TC30.009 
 TC1 and TC40.008 
 aTC2 and TC30.006 
 TC2 and TC40.011 
TibiaaTC1 and TC20.02231
 TC1 and TC40.008 
 TC1 and TC50.027 
FibulaTC1 and TC20.03714
HumerusaTC1 and TC20.00233
 TC1 and TC40.012 
UlnaTC1 and TC20.02420
Centrosaurus (percent “adult” size)TC1 and TC20.01350
 aTC1 and TC30.004 
 aTC1 and TC40.029 
 TC2 and TC40.028 
Centrosaurus (scaled est. femur length)TC1 and TC20.01350
 aTC1 and TC30.005 
 aTC1 and TC40.029 
 TC2 and TC40.031 
EiniosaurusTC1 and TC30.03716
 TC1 and TC40.020 
 TC1 and TC50.037 
PachyrhinosaurusaTC1 and TC20.01275
 TC1 and TC30.021 
 aTC1 and TC40.014 
Total sampleaTC1 and TC2<0.001141
 aTC1 and TC3<0.001 
 aTC1 and TC4<0.001 
 aTC1 and TC50.025 
 TC2 and TC40.025 

When elements are considered across genera, significant differences between mean bone lengths in different texture classes are obtained for all elements except the radius (Table 6). Post-hoc analysis with Bonferroni correction indicates that the following relationships are significant: the mean lengths of TC1 and TC2 are distinct from that of TC3 in the femur, and the mean length of TC1 is distinct from that of TC2 in the tibia and humerus. Without Bonferroni correction, the following additional relationships are also recognized as significant: the mean length of TC1 is distinct from that of TC2 and TC4, and the mean lengths of TC2 and TC4 are distinct from each other in the femur; the mean length of TC1 is distinct from that of TC4 and TC5 in the tibia; the mean length of TC1 is distinct from that of TC2 in the fibula and ulna; and the mean length of TC1 is distinct from that of TC4 in the humerus.

Analysis by taxon indicates significant differences between mean sizes of individuals with bones of different texture classes in all three genera (Table 6). Post-hoc analysis with Bonferroni correction indicates that the following are significant: the mean size of individuals with TC1 is distinct from that of TC3 and TC4 in Centrosaurus, and the mean size of individuals with TC1 is distinct from that of TC2 and TC4 in Pachyrhinosaurus. Without Bonferroni correction, the following additional relationships are also recognized as significant: the mean size of individuals with TC1 is distinct from that of TC2, and that of TC2 is distinct from that of TC4 in Centrosaurus; the mean size of individuals with TC1 is distinct from that of TC3, TC4, and TC5 in Einiosaurus; and the mean size of individuals with TC1 is distinct from that of TC3 in Pachyrhinosaurus. Relationships recognized as significant in Centrosaurus remain the same whether percent “adult” size or scaled estimated femur length is used as the proxy for individual size.

For the entire data set, Kruskal–Wallis analysis recognizes significant differences between mean sizes of individuals with different texture classes (Table 6). Post-hoc analysis with Bonferroni correction indicates the mean size of individuals with TC1 is significantly different from that of all other texture classes. Without Bonferroni correction, the mean size of individuals with TC2 is also recognized as significantly different from that of TC4.


A common suite of surface patterns and texture classes may be recognized across all elements in all three genera, although not all texture classes were recorded on all elements. This may represent a true absence of certain classes from specific elements in specific taxa, but it may also be a function of sample sizes and the incomplete nature of some specimens. When all elements from each genus are considered together (Fig. 4), all texture classes occur in Einiosaurus and Pachyrhinosaurus, and all except TC5 in Centrosaurus. When the data set is broken down by element, all texture classes occur in the femur and tibia, four of five classes occur in the fibula, humerus, and ulna (the fibula and humerus lack TC5 and the ulna lacks TC4), and three of five classes occur in the radius. In general, elements represented by more specimens (Table 1) record the presence of more texture classes. (The humerus, with the second largest sample size of N = 33, is an exception in that it lacks TC5.) Also, the texture classes which are occasionally absent (TC4 and TC5) are those which are associated with larger size bones, for which the sample size is smaller. It is therefore reasonable to infer that missing texture classes from certain elements more likely represents a sampling effect, rather than a true biological absence. The complete absence of TC5 Centrosaurus bones is noteworthy, but may also be a sampling issue. TC5 is comparatively rare in the sample as a whole, occurring for example in only 1 of 75 examined bones in Pachyrhinosaurus.

The statistical analyses identify a significant relationship between size and texture class for all genera, and for all elements except the radius (Table 5). The lack of a significant relationship in the radius may be a function of the very small sample size (N = 7) for this element.

Results of the Kruskal–Wallis tests and post-hoc analyses reveal that in general significant differences in mean size exist between TC1 and other classes (Table 6). This result is consistent whether the data set is analyzed by anatomical element (with the exception of the radius) or by genus, and suggests that TC1 is biologically linked to smaller body size and (presumably) earlier ontogenetic stage. In a few cases (femur across genera, Centrosaurus, total data set) significant differences are also identified between mean sizes of TC2 and higher classes.

There is a question, however, as to whether significant differences recognized only without Bonferroni correction should be considered valid. The correction reduces the risk of Type I errors (identification of a significant difference in the mean sizes of different texture classes, when in fact no such significant difference exists) but increases the risk of Type II errors (identification of no significant difference in the mean sizes of different texture classes, when in fact a significant difference does exist). Comparison of the statistical results (Table 6) with visual inspection of the texture class versus size plots (Figs. 2 and 3) suggests that in some cases the correction may in fact be masking real differences, such as those between TC1 and TC4 in the femur, tibia, and humerus, TC1 and TC5 in the tibia, TC1 and TC2 in the fibula, TC1 and TC4 or TC5 in Einiosaurus, and TC1 and TC3 in Pachyrhinosaurus.

The results of this study demonstrate that some overall decrease in surface porosity (progression from lower to higher texture classes) does occur throughout centrosaurine ontogeny, and that certain textural patterns are linked with various size and (presumably) age classes. The data suggest that TC1 is associated with the smallest individuals, TC3, TC4, and TC5 with the larger-bodied individuals, and TC2 with an intermediate stage that spans a range of sizes. Assuming that size can function as a general overall indicator of relative age, TC1 may be hypothesized to occur first in younger individuals, and TC3, TC4, and TC5 in presumably older animals. Interpretation of potential ontogenetic importance of TC2 is more uncertain. Because its defining midshaft characteristics combine the porous surfaces of TC1 and the dimpled and grooved surfaces of TC3 and TC4, it is tempting to consider TC2 as a marker of an intermediate “subadult” ontogenetic stage. The wide size range occupied by TC2 elements, however, offers little support either for or against intermediate status.

In Centrosaurus and Pachyrhinosaurus the minimum size of individuals with TC2 bones is greater than the minimum size of individuals with TC1 bones; although this difference is relatively slight (∼5 cm) in Pachyrhinosaurus. In both genera, moreover, TC2 bones occupy the entire rest of the size range, including the largest bones examined. In Pachyrhinosaurus, there is some suggestion that the large size spread of TC2 bones may be related to different rates of growth and maturation in different skeletal elements; TC2 femora tend to cluster in the lower half of the size range, tibiae and fibulae in the upper half. This might suggest that in this taxon, femora reach greater osteological maturity earlier in animals of smaller sizes. In Centrosaurus, separation of TC2 elements is less distinct, but TC2 femora do cluster in the lower half of the size range as in Pachyrhinosaurus. There is also a suggestion that femora may advance more quickly to more advanced texture classes at smaller body sizes in the Einiosaurus data set; however, the small sample size for Einiosaurus makes this pattern less robust. Differences in the timing of osteological maturation in different long bones have been well documented in modern taxa, particularly in mammals where epiphyseal fusion occurs at different times in different elements (e.g., Gilbert, 1990 and references therein) and precocial bird species in which the hind limb elements attain mature lengths and surface textures before those of the wing (e.g., Owen, 1980; Sedinger, 1986; Tumarkin-Deratzian et al., 2006).

It is reasonable to assume that decreasing textural porosity is more tightly linked to increasing osteological maturity than it is to simple increases in body size. In a previous study on extant birds, size-based age proxies were able to separate out a probable juvenile texture class, but size-independent proxies were necessary to reveal a characteristic “subadult” texture signal linked with intermediate levels of skeletal maturity (Tumarkin-Deratzian et al., 2006). In contrast, a companion study on extant crocodilians (Tumarkin-Deratzian et al., 2007) revealed little correlation between long bone texture and either size-based or size-independent age proxies in animals older than hatchlings. Therefore, the fact that a size-based age proxy is able to identify a distinct probably juvenile postcranial texture (TC1) in centrosaurines is promising, and suggests that a more detailed signal might be revealed by future comparisons of texture classes with size-independent age proxies. The size-independent maturity indices used by Tumarkin-Deratzian et al. (2006, 2007) relied on parsimony and cluster analyses of presence–absence data for muscle scars and other prominent bony landmarks to determine which of these features appeared in a predictable order during ontogeny. In theory, a similar method could be used to create a size-independent maturity index for each of the centrosaurine taxa examined in this study. A possible difficulty, however, that was not encountered with modern taxa is the incomplete nature of many of the bones. From the current data set, the best possibility for success in developing a size-independent maturity index lies with Pachyrhinosaurus, for which 93% of examined elements were complete, when compared with 75% for Centrosaurus and only 38% for Einiosaurus.

Previous studies examining the link between long bone surface textures and bone microstructure in archosaurs (Bennett, 1993; Tumarkin-Deratzian et al., 2006, 2007) suggest that bone textural changes may be reliable indicators of relative skeletal maturity in taxa with uninterrupted determinate growth (pterodactyloid pterosaurs, extant birds), but do not seem to be reliable for taxa with interrupted and/or indeterminate growth regimes (extant crocodilians). In some respects, the relationship between long bone surface texture and size in centrosaurines resembles the relationship seen in extant birds (Fig. 4), particularly in the overall lack of higher (less porous) texture classes in smaller individuals. The pattern in centrosaurines is more complicated, but this likely reflects the different growth regimes of the taxa under consideration.

Nearly all extant birds reach adult body size and full osteological maturity in less than 1 year, and at least two lines of evidence suggest that immature centrosaurines required several years to reach full adult size and skeletal maturity. The first evidence is the existence of a continuous size range from juvenile to adult within monodominant Centrosaurus apertus bone beds interpreted as mass-death assemblages (Ryan et al., 2001). Had young matured in 1 year, a die-off of an entire single population would be expected to yield remains from individuals of all adult size, or a bimodal distribution of adults and one non-adult size class, depending on whether the die-off occurred before or after that year's young had reached adult size. The second evidence is long bone microstructure. Recent examination of growth series of Einiosaurus (Reizner and Horner, 2006) and Centrosaurusapertus (Lee, 2006, 2007) suggests that individuals could have taken at least 3 years to approach adult sizes. This longer period of growth to adult size in centrosaurines may account for at least some of the variability observed in the relationship between size and long bone surface texture. A multiyear period of active immature growth could allow for greater growth variation among individuals and size variation within age classes than that observed in faster-maturing taxa such as the extant birds studied by Tumarkin-Deratzian et al. (2006). Greater intraspecific growth variability could also obscure a distinction between “juvenile,” “subadult,” and “adult” animals when body size is used as an age proxy.

An attempt to link size differences between centrosaurine long bone texture classes to discrete ontogenetic stages (juveniles versus subadults versus adults) is at this point premature. This is especially true for a “subadult” age class. Even if TC2 bones are the best candidates for an intermediate age class, the definition of “subadult” needs to be considered carefully, if it is to be anything more than a wastebasket class between juvenile and adult. If bones of TC2 are in fact the best candidates for such an intermediate stage, results reported here for Centrosaurus are not consistent with the previous “subadult-size” standard of 50–67% adult size (Ryan et al., 2001), because long bones of C. apertus bearing TC2 range from 44 to 100% adult size (Fig. 3).

This raises a question concerning the potential relationship, or lack thereof, between the long bone textures documented in this study and the cranial texture types previously examined (Sampson et al., 1997; Ryan et al., 2001; Brown, 2006; Tumarkin-Deratzian, 2003, 2007, in press; Brown et al., 2007, 2009). Given that surface texture changes throughout the skeleton are linked to bone growth processes and vascularization patterns on the microstructural level (Bennett, 1993; Tumarkin-Deratzian et al., 2006, 2007; Tumarkin-Deratzian, 2007, in press), both cranial and postcranial elements in centrosaurines would be expected to show similar patterns of an overall decrease in porosity and increase in surface finishing throughout ontogeny. This is, in general, what the data show. There are, however, certain marked differences between cranial and postcranial textures, most notably the presence of mottled texture on subadult-sized cranial bones, and that of fine porous and dimpled textures on the long bones. The metaphyseal regions of the long bones seem to follow a pattern of textural change most similar to that of the skull, with both demonstrating long-grained texture in smaller size classes, and smooth/rugose (skull) or smooth/grooved texture (long bones) in larger individuals. (The intermediate mottled texture that occurs on the skull seems to be a unique feature that has to date only been documented in the cranial bones of ceratopsid dinosaurs.) This general correspondence may be related to shape changes during ontogeny. Growth-related modeling and remodeling is more extensive and complex in both cranial elements and long bone metaphyses when compared with the neutral midshaft regions of long bones which undergo primarily appositional growth.

Even if specific long bone texture classes can ultimately be linked with specific ontogenetic stages, these stages may not directly correspond with those defined based on cranial elements, as patterns of growth and rates of osteological maturation are not necessarily the same throughout the skeleton. Ontogenetic stages for the individual as a whole, therefore, may need to include compound definitions, such as “texture class x” on cranial elements and “texture class y” on long bones. It is also possible that the current concept of subadult may need to be refined further, and perhaps subdivided into early and late stages, to accommodate different rates of osteological maturation in the cranial versus postcranial skeleton.

In summary, five distinct texture classes may be recognized on long bones of the centrosaurine genera Centrosaurus, Einiosaurus, and Pachyrhinosaurus. These range from highly textured bones with widespread fine porosity midshaft grading to fibrous and long-grained texture proximally and distally (TC1) to bones with primarily smooth midshaft regions (TC5). Textural patterns are largely consistent across the anatomical elements and taxa examined. Analysis of the relationship between texture class and size reveals a significant correlation in most cases; this seems to be driven largely by the fact that TC1 is invariably associated with the smallest bones, and the mean size of TC1 bones is in most cases significantly different than the mean size of bones in other texture classes. The relationship of the other textural classes to ontogenetic trends is less certain. The data do suggest that TC3, TC4, and TC5 are characteristic of larger individuals, and TC2 may be associated with intermediate sizes, although this is complicated by the fact that TC2 occurs over a wide size range, overlapping with that of the other texture types. Results of the current study do not rule out the possibility that distinct long bone texture types may be linked to juvenile, subadult, and adult stages in attainment of skeletal maturity; however, quantitative definitions of texture classes and comparison with size-independent maturity criteria such as progressive acquisition of bony landmarks and histological data will be necessary for better understanding.


The following individuals and institutions granted access to specimens under their care: Don Brinkman, David Eberth, James Gardner, Brandon Strilisky (TMP); John Horner, Pat Leiggi (MOR). Peter Dodson, Andrew Farke, and Michael Ryan reviewed an earlier draft of this manuscript and provided much helpful criticism. David Eberth, Barbara Grandstaff, Andrew Lee, Scott Sampson, Darren Tanke, and David Vann provided much stimulating and informative discussion. David Deratzian provided technical assistance with photography and digital imaging.