Influence of Sequence Heterochrony on Hadrosaurid Dinosaur Postcranial Development


  • Merrilee F. Guenther

    Corresponding author
    1. Department of Biology, Saint Joseph's University, Philadelphia, Pennsylvania
    • Department of Biology, Saint Joseph's University, 5600 City Avenue, Philadelphia, PA 19131
    Search for more papers by this author
    • Fax: +610-660-1832


A goal of modern dinosaur paleobiology is to synthesize the understanding of dinosaurian development and phylogeny. This study explores the Iguanodontia, one clade that includes several taxa for which growth series are preserved. It is hypothesized that analysis of growth series of iguanodontian taxa will reveal important developmental differences at play during the evolution of the clade. Such differences can reflect the impact of sequence heterochrony on iguanodontian evolution. Data were collected on the growth stages of the postcranial skeleton of a basal iguanodontian, Tenontosaurus; a lambeosaurine hadrosaurid, Hypacrosaurus; and two hadrosaurine hadrosaurids, Brachylophosaurus and Maiasaura to test the hypothesis. The event-pairing method provides a framework for the comparison of ontogenetic sequences among related taxa to detect significant evolutionary changes in developmental sequence. Significant developmental events are identified for the group and then the relative timings of those events in the developmental sequence are compared among the different species. Synapomorphic characters of Hadrosauridae, including the development of a prominent biceps tubercle on the coracoid, appear to be the result of changes in the developmental sequences of hadrosaurid taxa compared to basal iguanodontian taxa. This analysis also recognizes the developmental sequence differences that result in the development of the robust appendicular elements that characterize Lambeosaurinae. This comparative methodology shows that the development of characters such as the well developed supraacetabular process of the ilium results from the early onset of the development of those characters in the lambeosaurine hadrosaurid Hypacrosaurus compared to the hadrosaurines Brachylophosaurus and Maiasaura. Anat Rec, 292:1427–1441, 2009. © 2009 Wiley-Liss, Inc.

Evolutionary change can be difficult to study in taxa for which only fossils remain. In particular, understanding changes in developmental timing, or heterochrony, as potential mechanisms for evolutionary change (Gould, 1977, 1982; McKinney and McNamara, 1991) has been difficult because we lack the opportunity to observe fossil organisms through life. However, methods have been recently developed that better permit the utilization of fossil data. This began initially with the development of “event-pairing” (Mabee and Trendler, 1996; Smith, 1996; Velhagen, 1997; Mabee et al., 2000), which allows for the analysis of ontogenetic sequences by pairing two discreet developmental events in multiple taxa and subsequently comparing the relative timing of those events. Event-paired data can then be mapped onto a previously determined phylogeny to determine how the relative timing of two developmental events differs between taxa and changes over the course of the evolution of the clade.

The event-pairing method was developed for the purpose of analyzing the influence of sequence heterochrony on the evolution of a particular group of taxa, rather than solely determining the influence of differing growth rates. The method was first applied to the craniofacial development of eutherian and metatherian mammals (Smith, 1996, 1997). The event-pairing methodology allows for the analysis of a number of developmental events for multiple taxa, each of which may have different developmental rates. Later methods include event-pair cracking, which analyzes all event-pairs at a node together, and identifies those events that move in relation to the greatest number of other developmental events and move consistently in the same direction (Jeffery et al., 2002b). The event-pair cracking methodology has been modified, in the form of the Parsimov method, to better incorporate the principle of parsimony into the process by identifying the minimum number of events required to explain event-pair changes (Jeffery et al., 2005). This study does not utilize these later methods, but rather focuses on what information can be gained from the application of the original event-pairing methodology.

When attempting to study the impact of sequence heterochrony on the evolution of a clade, using developmental data, the method used should satisfy two criteria. First, the method should remove any issues related to the lack of chronological age or absolute timing data. This places the focus on the sequence of developmental events, rather than growth rates, and ensures that fossil data sets can be used (Webster and Zelditch, 2005). Second, the method should allow the analysis to take place within a phylogenetic framework so that observed symplesiomorphic and synapomorphic aspects of the sequences can be differentiated from one another (Jeffery et al., 2002a). The event-pairing method satisfies both requirements.

Ontogenetic studies are limited by the specimens and growth stages that have been preserved. Although discreet growth stages can be preserved histologically, the general morphological changes of a single individual cannot be followed throughout the course of its ontogeny. Therefore, assumptions have to be made about the degree to which an available set of specimens represents one particular growth stage of an average individual and how that relates to the averages observed for subsequent growth stages. Resolution of absolute timing, absolute developmental rates, and the variation in ontogeny within a taxon are impossible with a fossil group. These problems can be overcome by analyzing the ontogeny of a taxon in terms of relative timing and relative rates (Smith, 1997).

When comparing developmental series, even in extant taxa, there are difficulties in standardizing and quantifying developmental data so that they may be compared across species. Establishing growth stages can be one way of comparing developmental data. With both extant and extinct taxa, several definitions of a growth stage are possible, including those based on a percentage of adult (full) body size, chronological timing, or the appearance of particular morphological characters. Each of these can be accomplished with fossil taxa, but there are difficulties in defining stages based on a percentage of the size of a mature individual when growth rates are not known and maximum size may be indeterminate. This leaves the appearance of select morphological characteristics as the easiest change to recognize in fossil taxa. However, if heterochrony has occurred, then the sequence of development of crucial morphological structures may differ among taxa. This problem is shared by both extinct and extant taxa.

The hypothesis to be tested in this study is that event-pairing method can be informative when applied to a fossil group, and that this method can be used to recognize differences in developmental sequences within such a group. This study tested whether the resolution of data from fossil growth series is sufficient to detect differences, if they exist, in the sequence of developmental events among fossil taxa. Members of Hadrosauridae, along with a more basal iguanodontian, were chosen as the experimental group because they are well studied, well sampled, and include some of the most detailed growth series in the Dinosauria (Norman, 2004). The event-pairing methodology has been previously shown to be informative in groups with different growth strategies, including mammals (Smith, 1997) and reptiles (Velhagen, 1997). One hypothesis is that sequence heterochrony is an evolutionary mechanism behind the differing adult hadrosaurid postcranial morphologies, specifically how lambeosaurines evolved more robust postcranial skeletons than the more gracile hadrosaurines (Richardson, 1999; Horner et al., 2004).

During the development of any organism, the ultimate phenotype achieved is a product of genotype and environment (Dobzhansky, 1950; Lamb, 2000). A challenge in analyzing heterochrony, especially in a fossil group, is attempting to separate genetically induced versus environmentally induced differences in development. There is variation in the environments in which iguanodontian taxa lived (Forster, 1990; Maxwell and Ostrom, 1995; Varricchio, 1995; Horner, 1999; Fricke et al., 2008) and developed and this must be taken into account if genotypic developmental variation among taxa is to be determined. Work with extant taxa suggests that varied environmental conditions can alter the growth rate and the duration of developmental events, but does not appear to alter the sequence of developmental events (Gozlan et al., 1999; Kováč, 2002; Rozzi et al., 2005). It appears that the sequence of developmental events is selected for and is adaptive (Kováč, 2002).

Iguanodontia, and Hadrosauridae especially, present one of the greatest opportunities to study the ontogenetic trajectories of a dinosaur group. Growth series are known for several taxa, the most complete being the lambeosaurine Hypacrosaurus stebingeri (Horner and Currie, 1994) and the hadrosaurine Maiasaura peeblesorum (Horner, 1999). This study examines the similarities and differences among iguanodontian ontogenies from a different perspective. The growth series of several hadrosaurid taxa, including Maiasaura, Hypacrosaurus, and Brachylophosaurus were compared to the more basal iguanodontian Tenontosaurus. Carr (2003) considered application of the event-pair cracking method for tyrannosaurids, but this is the first event-pair study utilizing an ornithischian dinosaur group.


In analyzing the ontogenetic changes within a taxon or a group of taxa, it is advantageous to have as much developmental data as possible, beginning as early in ontogeny as possible. Embryological data are useful for understanding the full picture of the development of an organism, but other than from late stage embryos, such data are not available for fossil groups such as iguanodontians. This is due to a lack of preservation of the cartilaginous precursors of bones and of soft tissue morphology. As a result, analyses of the ontogenies of extinct taxa are restricted to studying the development of osteological characters of the ossified skeleton.

Despite such limitations, there are still a number of well-preserved embryonic and juvenile specimens with which to work. Taxa for which the event-pairing methodology could be applied must include specimens from at least two, but preferably four or more, growth stages. This greatly reduces the number of qualifying taxa within the Dinosauria; iguanodontians have a larger proportion of taxa represented by multiple growth stages than practically any other dinosaurian group. Ideally, the taxa in the study include hatchling or embryonic material. However, only three taxa have definitive hatchling material: Maiasaura peeblesorum (Horner, 1999; Horner and Weishampel, 1994), Hypacrosaurus stebingeri (Horner and Currie, 1994), and Tenontosaurus tilletti (Forster, 1990). The other taxon included in the study, Brachylophosaurus canadensis, is known only from posthatchling juvenile stages and the adult stage (Prieto-Marquez, 2005). The issue of preservation quality, specifically completeness and taphonomic alteration of morphology, becomes a factor in the analysis because it limits specimens that can be included in the study. The Hypacrosaurus stebingeri specimens are the best preserved specimens for each growth stage; the specimens of Tenontosaurus tilletti are the least well preserved in the hatchling stages, but have well-preserved juvenile and adult specimens. Characters chosen for analysis reflect the constraints of the material: the appearances of distinctive morphological features, rather than changes in proportion or anatomical location, made prime event characters because the latter could be altered by diagenetic compression, which seems to have affected some of the early growth stage elements.

In assembling growth series for these taxa, establishing the taxonomic identity of the specimens is critical to the integrity of the results of the analysis. This is particularly true when working with postcranial elements, which can be less informative than cranial material, especially for early growth stage individuals. Some well-preserved specimens represent the early developmental stages of an indeterminate ornithopod, and could not be included in this analysis because they lacked taxonomic certainty. The specimens utilized for this study (Table 1) are from nesting horizons and bonebeds in the cases of Hypacrosaurus and Maiasaura, bonebeds in the case of Brachylophosaurus, and are associated with adult material in the case of Tenontosaurus, thus providing a high degree of certainty as to the identities of the specimens.

Table 1. Specimens used in the study and the elements available for each taxon
TaxonSpecimen numbersGrowth stages representedElements available for the taxon
  1. Museum abbreviations are as follows: CMN, Canadian Museum of Nature; MOR, Museum of the Rockies; RTMP, Royal Tyrrell Museum of Palaeontology; OMNH, Sam Noble Oklahoma Museum of Natural History; YPM, Yale Peabody Museum of Natural History.

MaiasauraMOR 0055Scapula, coracoid, humerus, radius, ulna, pubis, ilium, ischium, femur, tibia, fibula
MOR 547
YPM 22400
YPM 22405
YPM 22432
YPM 22472
HypacrosaurusMOR 495Scapula, coracoid, humerus, radius, ulna, pubis, ilium, ischium, femur, tibia, fibula
MOR 355
MOR 548
MOR 553
MOR 559
MOR 600
RTMP 94.385.1
BrachylophosaurusCMN 88932Scapula, coracoid, humerus, radius, ulna, pubis, ilium, ischium, femur, tibia
MOR 1071
TenontosaurusOMNH 101443Scapula, pubis, ilium, ischium, femur, tibia, fibula
OMNH 34191
OMNH 34785
OMNH 58340

Before the relative sequences of each taxon can be determined, the concept of “growth stage” must be defined so it can be applied consistently across taxa. Each taxon is made up of several “growth stages,” each of which is defined as a collection of individuals within the bonebed or nesting material that cluster by size. In most cases, there were natural gaps between the size clusters. Because none of the taxa is represented by a continuous growth series, growth stages are easier to define using the assumption that different elements of a certain size range all pertain to individuals of the same growth stage. As there are few articulated elements available, same-sized elements are assumed to belong to the same growth stage, but not necessarily the same individual. Each growth stage is composed of individual elements that are no more than fifteen percent smaller than the largest element in that size cluster. The presence or absence of key morphological features of each element, including all of the vertebral, pectoral girdle, forelimb, pelvic girdle, and hind limb elements, were compared to each of the other elements within each growth stage. A relative timeline of the appearances of morphologic features during different growth stages can then be constructed for each taxon (Table 2).

Table 2. Hypothetical developmental sequences for four different taxa
TaxonGrowth stage 1Growth stage 2Growth stage 3Growth stage 4
  1. In each growth stage, several different morphological features appear (developmental events A–P; see Table 3 for key to developmental event letter assignments).

Taxon AA, C, F, H, MB, J, K, LE, N, O, PD, G, I
Taxon BB, C, F, IA, J, K, MD, E, H, NG, L, O, P
Taxon CA, C, K, JB, D, I, K, ME, H, L, NF, G, O, P
Taxon DC, F, M, HA, J, K, M, PB, D, N, OE, G, I, L, P

For the event-pairing method, as described by Smith (1996, 1997), the relative timing of pairs of various developmental events are used to construct a matrix for each taxon that shall hereafter be known as the STEP matrix (single taxon event pairs) (Fig. 1A). Each developmental event—for example, the ossification of a certain element—is compared to every other developmental event across all stages, and the relative timing of the two events within each pair is established. The developmental events are scored in the matrix in a manner similar to the coding of morphological characters in a phylogenetic analysis: the event is scored (0) if it occurs before the event it is being compared with, (1) if it occurs simultaneously with the second event, or (2) if it occurs after the second event (Fig. 2). Missing data are coded as (?). It is unlikely that there are many truly simultaneous developmental events. However, because these are fossil data and the growth stages are relatively spread apart, missing gaps in ontogeny, certain developmental pairings appear simultaneous when, in fact, they are not, and would not appear so if missing growth stages were preserved. This is illustrated by Brachylophosaurus, which has the fewest represented growth stages among the studied taxa, resulting in a larger number of developmental event-pairs being classified as “simultaneous.” This problem is analogous to a polytomy in a cladogram resulting from limited data. Increasing the resolution of the sample would help alleviate this problem, but this was not possible with currently available specimens.

Figure 1.

Two-step method for creating a data matrix that can be used to analyze event-paired data. A: In the first step, referred to as the STEP matrix (single taxon event pairs), each developmental event is compared to each other event in the sequence of a single taxon to determine the relative timing. Relative events are coded as having occurred before (0), simultaneously with (1), or after (2) focus events labeled at the top of each column. B: In the second step, referred to as the ATEP matrix (all taxa event pairs), event-pairs determined in step one are combined into a data matrix that includes event-pair data for all taxa included in the study.

Figure 2.

For each pair of developmental events, the events can have three possible relationships within the sequence: developmental event A can occur before (0), simultaneously with (1), or after (2) developmental event B.

Because developmental events, rather than morphological features, are used as “characters,” a clarification about the method's terminology, initially formalized by Jeffrey et al. (2002a), must be made. The initial data matrices that are compiled for each individual taxon are composed of developmental events—these events are based on the appearance of actual morphological characters. As shown in Fig. 2, the relative timing of each developmental event is compared to that of every other event. Because the STEP matrices are not designed to be analyzed, this set of “characters” will not be referred to in the results. In the matrix that is actually analyzed, hereafter referred to as the ATEP matrix (all taxa event pairs) (Fig. 1B), each of the developmental events from the STEP matrix is paired with every other developmental event to form a series of event-pairs. An event-pair [referred to as a “sequence unit” by Velhagen (1997)] is the summary of the relative timing between the developmental events being compared in the STEP matrix. The event-pairs for all taxa are included in the ATEP matrix, for which event-pairs are functionally the equivalent of characters in a phylogenetic analysis. Further terms are defined as follows: a focus event is a developmental event labeled at the top of each column of a STEP matrix; a relative event is an event labeled at the beginning of each row of a STEP matrix. Relative events are compared to focus events, and event-pairs are taken from the perspective of the focus events, which are held constant [a modification of Jeffrey et al. (2002a)]. For example, relative event “A” occurs before focus event “B.” In the STEP matrix, the event state would be coded (0). In the ATEP matrix, the state for event-pair “AB” would therefore also be coded (0). Finally, an event-pair synapomorphy is an event-pair state transformation reconstructed along a branch of a phylogenetic tree. It is a change in relative timing between two events occurring during the evolution of the group.

Event-pair states for each taxon were coded and entered into a STEP matrix that expresses the relative timing of events for that taxon. Once the STEP matrices for each individual taxon were established, they were combined into an ATEP matrix composed of the event-pairs for all of the taxa. The ATEP matrix was entered into the computer program PAUP (Swofford, 2001) and a parsimony analysis was completed. Event-pair state transformations were mapped onto previously established phylogenies for Iguanodontia in order to provide a phylogenetic context for the developmental data (Smith, 1997). Although there is no universal consensus regarding the phylogenies of Iguanodontia and Hadrosauridae as a whole, the relationships between the taxa included in the study are not in doubt; the phylogenies of Norman (2004) and Horner et al. (2004), respectively, were chosen for use in this study.

Event-pair state transformations were determined based on the event-pair maps and developmental events that commonly differed in their position in the sequence among the different taxa were identified. The following is an outline of the morphological characters chosen as developmental events to be paired in the STEP and ATEP matrices.

Morphological Characters

In choosing developmental events to include in the data matrix, care must be taken to choose events that are defined by the appearance of morphological characters shared by all taxa included in the study. A single exception was the appearance of the extensor tunnel of the femur, a postcranial synapomorphy for hadrosaurids (Weishampel and Horner, 1990). This morphological character was included despite its absence in Tenontosaurus because its appearance in the developmental sequences of hadrosaurid taxa differs between lambeosaurines and hadrosaurines.

Different appendicular skeletal characters appear at different stages of development and grow at different rates. The appearances of these morphological characters were chosen as the focus of this study because postcranial elements are more abundant for specimens that represent early growth stages, particularly embryonic material (see Table 3 for a detailed description of characters). There are 16 developmental events for each taxon; the pairing of those events results in 120 event-pairs (Table 4; Fig. 3). Tenontosaurus has only 104 valid event-pairings due to its lack of an extensor canal of the femur.

Figure 3.

Morphological characters coded as developmental events, see Table 3 for detailed description of characters. Elements are not to scale.

Table 3. Description of morphological characters of the postcranial skeleton that are used as developmental events for the event-pair method
VertebraeA) Complete ossification of vertebrae
CoracoidB) The closure of the coracoid foramen
C) Development of biceps tubercle
ScapulaD) Development of acromion process
HumerusE) Development of humeral head
F) Development of muscle scar of latissimus dorsi
G) Development of muscle scars on lateral border of humerus
IliumH) Development of supraacetabular process
I) Development of pubic peduncle
IschiumJ) Development of obturator process
PubisK) Development of ischiadic peduncle
FemurL) Development of muscle scar for caudofemoralis brevis on the fourth trochanter
M) Development of muscle scar for femorotibialis externus on the lateral side of the fourth trochanter
N) Development of deep intercondylar groove
O) Development of extensor tunnel
TibiaP) Development of medial tibial condyle of the tibia
Table 4. The ATEP matrix for event-paired data for hadrosaurid and iguanodontian taxa (letter assignments as per Table 3)


The analysis of the ATEP matrix provided detailed information about the similarities and differences in the developmental sequences of the analyzed taxa. There were different proportions of the three different event-pair states (Table 5). (The high proportion of unknown or missing data for Tenontosaurus results from the event-pairs that include the appearance of the extensor canal on the femur, which Tenontosaurus lacks.) For each taxon, the number of (0) and (1) states was approximately equal. States coded (1) are simultaneous events and are the least informative of the states. Event-pair state transformations are summarized in Fig. 4.

Figure 4.

Distribution of some event-pair transformations on cladogram of relationships within iguanodontia.

Table 5. The event-pair states for each taxon used in the analysis (total of 120)
TaxonEvent-pair state 0 (event occurs before)Event-pair state 1 (event occurs simultaneously)Event-pair state 2 (event occurs after)

Mapping event-pairs onto a phylogeny of Iguanodontia (Norman, 2004) elucidates which constitute event-pair synapomorphies at various taxonomic levels (Table 4). Many of the event-pairs have states that are shared by all of the taxa (Fig. 4). Uniform event-pair states across taxa (here, within Hadrosauridae) indicate that the relative timing of two events is unchanged during the evolution of the taxon.

The event-pairs are analyzed individually to determine the significance of each one in the evolution of the group. Because event-pairs in this context are used in an attempt to understand a phylogeny rather than to provide a diagnosis for any particular clade, these event-pairs can be analyzed on a pair by pair basis. Traditional parsimony analyses aimed at recovering phylogenies can include developmental data, but are typically dominated by morphological characters. Nevertheless, for elucidating possible phylogenetic patterns, developmental data can be as informative as morphological data (Grandcolas et al., 2001).

The distribution of event-pairs among taxa can be used to support the monophyly of a taxon. The distribution can also identify event-pair autopomorphies and synapomorphies, and can illuminate previously unrecognized convergences in developmental sequences. In the present analysis, this was observed in Tenontosaurus and Hypacrosaurus (e.g., Fig. 5C,D).

Figure 5.

Examples of event-pair states and their distribution among taxa. Phylogenetic trees are simplified; many taxa and clade names are excluded. Double lines indicate event-pair state (0), dotted lines indicate event-pair state (1), and solid lines indicate event-pair state (2). A: Focus event—closure of coracoid foramen, relative event—development of lateral humeral muscle scars; B: Focus event—development of muscle scar for latissimus dorsi on the humerus, relative event—development of supraacetabular process; C: Focus event—appearance of obturator process, relative event—development of extensor tunnel of femur; D: Focus event—development of obturator process, relative event—development of medial tibial condyle of the tibia.

In the ATEP matrix, the distribution of event-pair states among the taxa is relatively consistent (Table 5). For all taxa, the event-pair state (0) comprises ∼30% of the event-pairs. State (1) also accounts for ∼30% of the event-pairs, and state (2) comprises the remaining ∼40% of the event-pairs. The similar proportions of state (1), which represents unresolved parts of the sequence, indicate that the amount of ambiguous sequence data is similar for each taxon. The distribution of the Tenontosaurus matrix differs slightly because of the missing data ((?) codings in the matrix).

The iguanodontian postcranial skeleton is somewhat conservative, resulting in many event-pairs for which the coded event-pair state is the same across all taxa, meaning that no event-pair transformations have occurred. Of the 120 event-pairs, more than half (63%) have a single state shared by all taxa. For example, the state is the same for all taxa for event-pairs KL and KM (Table 4), meaning that in all sequences the ishiadic peduncle appears before the development of the femoral muscle scars.

A number of event-pairs support a monophyletic Hadrosauridae and have shared event-pair states across the hadrosaurid taxa to the exclusion of the more basal iguanodontian, Tenontosaurus. Those event-pairs that include developmental events of the coracoid (B and C, Table 3) show that the appearance of the biceps tubercle and the closure of the coracoid foramen occur earlier in the developmental sequences of hadrosaurids than they do in Tenontosaurus. Further differences in the pectoral girdle are exhibited by the emergence of the acromion process (D, Table 3) early in the hadrosaurid sequences compared to the sequence for Tenontosaurus.

Also uniting the hadrosaurid taxa are event-pairs that include the development of the supraacetabular process of the ilium (H, Table 3). The process is well developed in hadrosaurids, especially lambeosaurines, and the hadrosaurids share a developmental sequence in which the supraacetabular process forms early compared to other morphological structures. Some event-pairs reflect an earlier appearance of this process in Hypacrosaurus than in either the hadrosaurines or Tenontosaurus.

The early appearance of the supraacetabular crest is involved in all the event-pair autopomorphies for Hypacrosaurus. Because of the limited taxon set in the study, it is unclear if this is an event-pair autopomorphy for the genus or an event-pair synapomorphy for Lambeosaurinae.

Unexpectedly, some event-pairs coupled Tenontosaurus with Hypacrosaurus. The event-pairs that include the appearance of the obturator process (J, Table 3) as an event have the same event-pair states for these two taxa, reflecting possible convergences, or reversals in Hypacrosaurus, in their developmental sequences.

Cladistic analysis of morphological characters recover Maiasaura and Brachylophosaurus as sister taxa (Horner et al., 2004), so similarity in the developmental sequences between the two is expected. One of the significant events uniting the hadrosaurine taxa is the appearance of the extensor tunnel on the femur (O, Table 3). This morphological feature is unique to hadrosaurids. However, it appears to occur late in the sequences of both Maiasaura and Brachylophosaurus compared to Hypacrosaurus. Given the conservative morphological nature of the hadrosaurid femur, it is somewhat unexpected that the development of femoral characters could be taxonomically informative (Fig. 6). The appearance of the obturator process on the ischium (J, Table 3) also occurs relatively later in the sequence for the hadrosaurines than it does for the other two taxa, as indicated by all of the event-pairs that include the obturator process as a developmental event.

Figure 6.

Comparative ontogenies of the femora of all taxa. Elements are not to scale; scale bars below each specimen represent the relative size of that element in relationship to the adult element, which is shown at the far right.

Several event-pairs that have states shared only by Brachylophosaurus and Hypacrosaurus, which would, at first, suggest a similarity in their developmental histories. However, upon examination of each of these pairs, the state in each is (1), meaning the developmental events occur simultaneously. As above, this is the least informative event-pair state and most likely the result of limited data. Its presence in all of these event-pairs, therefore, does not suggest that there is a closer relationship between Brachylophosaurus and Hypacrosaurus than previously thought.

The 15 event-pairs relating vertebral column development to the other regions of the postcranium do not distinguish any taxa. Ten of the 15 event-pairs have uniform event-pair states across all taxa. The single vertebral developmental event included in the data set (A, Table 3) occurs early in ontogeny for all taxa in the study. Forty-two event-pairs include developmental events in the pectoral girdle; of these pairs, 40% have one event-pair state distributed across all taxa (Figs. 7 and 8). Another 42 event-pairs involve developmental events of the forelimb, specifically the humerus. Of these, 62% have a uniform event-pair state across all taxa. Fifty-four event-pairs include pelvic girdle developmental events, and 65 event-pairs include one or more developmental events that occur in the hind limb. For event-pairs that include developments in the pelvic girdle, 56% had identical event-pair states across all taxa. Event-pairs that included developmental events of the hind limb have a single event-pair state across all taxa in 49% of the event-pairs.

Figure 7.

Comparative ontogenies of the scapulae of all taxa. Elements are not to scale; scale bars below each specimen represent the relative size of that element in relationship to the adult element, which is shown at the far right.

Figure 8.

Comparative ontogenies of the coracoids of all taxa. Elements are not to scale; scale bars below each specimen represent the relative size of that element in relationship to the adult element, which is shown at the far right.

Some of the greatest variability in the distribution of event-pair states among taxa is shown in those event-pairings that compare developmental events from the pectoral and pelvic girdles (Figs. 7–11). In particular, event-pairs involving the development of morphological features of the coracoid are not consistent across taxa. Event-pairs coupling the development of the biceps tubercle of the coracoid (C, Table 3) with the supraacetabular process of the ilium (H, Table 3) and the obturator process of the ischium (J, Table 3) show variation across taxa (Table 4). The hadrosaurid coracoid is unique for its prominent biceps tubercle, but it otherwise has not been shown to be diagnostic at a lower phylogenetic level (Fig. 8). However, its placement in the developmental sequence does appear to vary among taxa. This introduces the possibility that the coracoid can be taxonomically informative, if only from a developmental perspective.

Figure 9.

Comparative ontogenies of the ilia of all taxa. Elements are not to scale; scale bars below each specimen represent the relative size of that element in relationship to the adult element, which is shown at the far right.

Figure 10.

Comparative ontogenies of the pubes of all taxa. Elements are not to scale; scale bars below each specimen represent the relative size of that element in relationship to the adult element, which is shown at the far right.

Figure 11.

Comparative ontogenies of the ischia of all taxa. Elements are not to scale; scale bars below each specimen represent the relative size of that element in relationship to the adult element, which is shown at the far right.

Although the morphology and, in particular, the proportions of the humerus are taxonomically distinctive, the relative timing of humeral features in the developmental sequence does not seem to vary greatly among taxa (Fig. 12). Of the 42 event-pairs that involve a developmental event of the humerus, only 11 show any variation in the distribution of event-pair states among taxa. Several of those 11 pairs show variation only due to their inclusion of the (1) state, which is of limited value.

Figure 12.

Comparative ontogenies of the humeri of all taxa. Elements are not to scale; scale bars below each specimen represent the relative size of that element in relationship to the adult element, which is shown at the far right.


The morphological character analysis of the event-pair data indicates that while the postcranial skeletons of iguanodontians are developmentally conservative throughout the clade, event-pair transformations have occurred during its evolution, perhaps resulting in subtle morphological differences in the postcranial elements. The analysis also suggests that elements containing morphologically significant characters do not necessarily contribute informative developmental data.

Scapula, humerus, and pelvic girdle element morphologies are useful for distinguishing the postcrania of the two hadrosaurid subclades. However, in the humerus, for example, the present analysis does not reveal sequence heterochrony at work in the differing adult morphologies across taxa. It appears that changes in developmental rate, rather than the developmental sequence of events, lead to the large deltopectoral crest of lambeosaurines and the reduced deltopectoral crest of hadrosaurines.

In general, event-pair data from the pelvic girdle appear to be among the most instructive in the analysis. Event-pairings that include the appearance of the obturator and supraacetabular processes are important as components of event-pair synapomorphies for both Hadrosauridae and either Hypacrosaurus or Lambeosaurinae. In contrast to the humerus, differing adult morphologies of the pectoral and pelvic girdles appear to result, at least in part, from sequence heterochrony. The development of different morphologies of the obturator process and the supraacetabular process of the ilium appear to result from changes in timing of those events relative to other developmental events in the skeleton. For example, in Hypacrosaurus, the development of the supraacetabular process of the ilium begins relatively early in the developmental sequence compared to its hadrosaurine cousins. This early appearance of the process could explain, at least in part, the fact that this morphological feature is better developed in Hypacrosaurus than it is in all other taxa being studied. This corresponds to the general trend of lambeosaurines having a better developed, more ventrally extensive supraacetabular process than both hadrosaurines and more basal iguanodontians.

Event-pair data from the hind limb suggest that developmental data from these elements can be more revealing than their morphological characters. Event-pairings that contain femoral events support the Brachylophosaurus and Maiasaura clade.

Taxonomically, the developmental data generated with event-pairing reaffirms taxonomic relationships within Iguanodontia from a different perspective (Table 4). This suggests that the use of a developmental data set can confirm phylogenetic analyses based on morphological data, reinforcing the initial conclusions. The event-pairing analysis strengthens support for a monophyletic Hadrosauridae, helps to illuminate differences between hadrosaurines and lambeosaurines, and supports hypotheses of a close relationship between Brachylophosaurus and Maiasaura. Because of the limited number of taxa used in this analysis, it is difficult to know exactly where in the course of the evolution of the clade the changes in ontogeny occur; if the growth series of related taxa are discovered this will be clarified.

Having uniform event-pair states across all taxa reflects a phylogenetically conservative and stable portion of the developmental sequence, at least in terms of those specific events. Morphologically, certain elements are much more informative phylogenetically than others, with certain elements providing several synapomorphies and others providing few or none. The same trend is true for developmental data. The utility of event-pairs defined by certain regions, such as the forelimb, proved more limited than expected.

Characters that are part of this event-pairing analysis are instructive on multiple levels. Understanding that the developmental sequence of morphological features varies helps to explain the differing morphologies observed in mature individuals. For example, the relatively early development of the biceps tubercle and the coracoid foramen in hadrosaurids may explain the unique morphology of the hadrosaurid coracoid compared to basal iguanodontian taxa (Horner et al., 2004). The comparatively early start of the development of the biceps tubercle allows more time for it to form, resulting in its prominence on the medial border of the coracoid. The coracoid of Tenontosaurus is generally more rounded than the elongate hadrosaurid morphology. Although not specifically tested in this analysis, it is possible that the elongation of the coracoid also results from an earlier start to the development of the medial process of the coracoid.

The coupling of Tenontosaurus and Hypacrosaurus is not intuitive from a phylogenetic perspective; however, this analysis did not simply determine shared derived characters, but rather similarities in the order of development. The similarities in the developmental series of Tenontosaurus and Hypacrosaurus suggest that the robust postcranial morphology of Hypacrosaurus may be the result of convergences in developmental sequence or the retention of more of the symplesiomorphic developmental sequence than the hadrosaurines. An example of this is the development of the ischium. Although the finer morphologies are different, the robust nature of the ischium of Hypacrosaurus is reminiscent of the robust ischium of Tenontosaurus. Both taxa ultimately develop an ischium with a robust obturator process and wide distal shaft. Because the distal “boot” of the ischium is unique to lambeosaurines, its development was not included as a developmental event, and its place in the developmental sequence was not analyzed. However, it is possible that a similarity in the timing of the expansion of the distal shaft between Hypacrosaurus and more basal iguanodontians resulted in the characteristic distal ischial “boot” of the lambeosaurines. The developmental sequences of the hadrosaurines Maiasaura and Brachylophosaurus may then represent a more derived state, with the late onset of developmental events that ultimately resulted in the more gracile postcranial skeleton of the hadrosaurines.


The application of the event-pair method, first developed and utilized for extant metatherian and eutherian mammals (Smith, 1996, 1997), allows for an objective approach to the study of sequence heterochrony even in extinct taxa. This method removes the problems of absolute timing and well-defined growth rates, both of which hinder studies of the ontogenies of fossil taxa. The obvious limitation in using this method with extinct taxa is the comparatively small number of available taxa and specimens to study.

This study had the purpose of answering a narrow question within a narrow range of taxa, specifically to determine the impact of sequence heterochrony on the postcranial development of iguanodontian taxa. It appears that sequence heterochrony does have an impact on the development of the pectoral girdle and pelvic girdle elements in particular, resulting in varied adult morphologies. Synapomorphic postcranial morphological characters of Hadrosauridae, including the prominent biceps tubercle and the development of the extensor tunnel, do appear to be the result, at least to some degree, of changes in developmental sequence. In addition to forming hypotheses about the evolutionary causes behind certain morphologies, this method encourages the inclusion of developmental data in phylogenetic analyses when it is available.

Future possibilities for this methodology include analyses of data from a combination of extant and extinct taxa to answer broad questions about the Archosauria. Event-pairing could be applied to crocodylians or to avian groups with the ultimate goal of utilizing this method within the framework of Extant Phylogenetic Bracketing. This could provide support for hypotheses of developmental sequences of dinosaurian groups for which there are limited or no growth series available. The next step in the utilization of this event-pair methodology is to apply it to an archosaurian group that includes both extant and extinct taxa.


The author thanks Peter Dodson, Hermann Pfefferkorn, and Fredrick Scatena (University of Pennsylvania) for their advice on early versions of this manuscript. He also thanks Carrie Ancell and John Horner (Museum of the Rockies), Philip Currie (University of Alberta), Margaret Feuerstack and Kieran Shepherd (Canadian Museum of Nature), James Gardener (Royal Tyrrell Museum of Palaeontology), Jeff Person (North Dakota Geological Survey), and Kevin Seymour (Royal Ontario Museum) for help with and access to specimens. He also thanks reviewers David Evans (Royal Ontario Museum) and Jerald Harris (Dixie State College of Utah).