Paleobiology, 31(3), 2005, pp. 354–372

Evolutionary modifications of ontogeny: heterochrony and beyond

Mark Webster and Miriam Leah Zelditch

Abstract.—Consideration of the ways in which ontogenetic development may be modified to give morphological novelty provides a conceptual framework that can greatly assist in formulating and testing hypotheses of patterns and constraints in evolution. Previous attempts to identify distinct modes of ontogenetic modification have been inconsistent or ambiguous in definition, and incom- prehensive in description of interspecific morphological differences. This has resulted in a situation whereby almost all morphological evolution is attributed to heterochrony, and the remainder is commonly either assigned to vague or potentially overly inclusive alternative classes, or overlooked altogether. The present paper recognizes six distinct modes of ontogenetic change, each a unique modifi- cation to morphological development: (1) rate modification, (2) timing modification, (3) heterotopy, (4) heterotypy, (5) heterometry, and (6) allometric repatterning. Heterochrony, modeled in terms of shape/time/size ontogenetic parameters, relates to parallelism between ontogenetic and phylo- genetic shape change and results from a rate or timing modification to the ancestral trajectory of ontogenetic shape change. Loss of a particular morphological feature may be described in terms of timing modification (extreme postdisplacement) or heterometry, depending on the temporal de- velopment of the feature in the ancestor. Testing hypotheses of the operation of each mode entails examining the morphological development of the ancestor and descendant by using trajectory- based studies of ontogenetically dynamic features and non-trajectory-based studies of ontogenet- ically static features. The modes identified here unite cases based on commonalities of observed modification to the process of morphological development at the structural scale. They may be heterogeneous or par- tially overlapping with regard to changes to genetic and cellular processes guiding development, which therefore require separate treatment and terminology. Consideration of the modes outlined here will provide a balanced framework within which questions of evolutionary change and con- straint within phylogenetic lineages can be addressed more meaningfully.

Mark Webster. Department of the Geophysical Sciences, University of Chicago, 5734 South Ellis Avenue, Chicago, Illinois 60637. E-mail: [email protected] Miriam L. Zelditch. Museum of Paleontology, University of Michigan, Ann Arbor, Michigan 48109-1079

Accepted: 3 November 2004

Introduction gier and Vlahos 1988; Raff and Wray 1989; Wray and McClay 1989; McKinney and Mc- Ontogeny holds a pivotal place in evolu- Namara 1991; Sattler 1992; Alberch and Blan- tionary biology. Modification of ontogenetic co 1996; Raff 1996; Zelditch and Fink 1996; development (‘‘developmental reprogram- Reilly et al. 1997; Rice 1997; Klingenberg 1998; ming’’ [Arthur 2000]) forms a critical link be- tween mutation and selection at any life stage. Lovejoy et al. 1999; Arthur 2000; Li and John- Comparative studies of ontogeny provide ston 2000; Sundberg 2000; Smith 2001). In this unique windows into evolutionary mecha- paper we present a novel consideration of nisms, permitting recognition of constraints modes of evolution because those previously on morphological evolution and of the proxi- proposed offer incomplete and sometimes mal processes responsible for evolutionary confusing characterizations of ontogenetic change, as well as helping to elucidate phylo- modifications. Confusion arises partly from genetic patterns (see Wagner et al. 2000). Not application of the same terminology for cases surprisingly, several attempts have been made of developmental reprogramming identified to recognize the modes by which ontogeny by using morphological and non-morpholog- can be modified (e.g., Zimmermann 1959; ical criteria despite evidence that these may be Takhtajan 1972; Gould 1977, 2000; Alberch et biologically incommensurate. Additional con- al. 1979; McNamara 1986a; Atchley 1987; Re- fusion results from poorly defined evidentia-

᭧ 2005 The Paleontological Society. All rights reserved. 0094-8373/05/3103-0002/$1.00 MODIFICATIONS OF ONTOGENY 355 ry criteria demarcating the modes to the point parallels between the stages of ontogeny and that a heterogeneous array of changes can be phylogeny’’ (Gould 1977: p. 2). Gould (1977) interpreted as a single type. Moreover, the fail- and Alberch et al. (1979) considered heteroch- ure to entertain a full range of competing hy- rony in terms of morphological criteria (spe- potheses means that cases failing to meet the cifically, the decoupling of organismal shape or evidentiary criteria of a well-defined mode are size from developmental time) and expressed assigned to poorly conceived or nonspecified the concept in terms of the types of data avail- modes, or even ignored. Such conceptual and able to morphologists. Heterochrony was de- semantic confusion has resulted in a strong tected when, for a given age, the descendant bias in favor of recognition of some modes of has a shape typical of the ancestor at a more change (particularly heterochrony), in the juvenile age (paedomorphosis) or at a more confounding of heterochrony with other kinds mature age (peramorphosis; this could be a of modification to development, and in a ten- hypothetical extrapolation beyond the termi- dency to oversimplify the evolution of ontog- nal morphology of the ancestor), or retains an- eny. cestral shape but differs in size (dwarfism or Some workers contend that attempts to re- gigantism). A nomenclature is in place de- solve the confusion will only further compli- scribing the various ways in which a paedo- cate an already excessively verbose literature morphic or peramorphic descendant can re- (e.g., Klingenberg 1998; McKinney 1999). sult from an increase (acceleration) or de- However, we are concerned that the present crease (neoteny) in the rate at which the tra- framework, if left unrefined, will lead to in- jectory of shape change is followed, or from an adequate summaries of developmental repro- early (predisplacement) or late (postdisplace- gramming and a misleading impression of the ment) onset of the trajectory of shape change, relative contribution of some of the modes to or from the early (progenesis) or late (hyper- evolution. We think that the need for a com- morphosis) termination of the trajectory of prehensive consideration of modes of devel- shape change (Gould 1977; Alberch et al. opmental reprogramming, with explicit cri- 1979). As well as a nomenclature, there is a for- teria by which the various modes can be iden- mal analytic scheme for detecting which of tified and the increased investigative rigor these modes of heterochrony occurs in a par- that this brings, far outweighs the understand- ticular case (Alberch et al. 1979). The fact that able reluctance to stir once again the already the evolutionary and ontogenetic vectors of muddied terminological waters of this re- shape change coincide under this definition of search field. heterochrony suggests a channeling or con- The present paper aims to reduce confusion straint in the direction of morphological evo- about heterochrony and to expand and refine lution, which may be significant in terms of re- the range of hypotheses considered beyond lating such change to underlying developmen- heterochrony. These aims are interrelated, as tal processes. clearly articulating distinctions among con- Heterochrony has been recast in a more cepts and specifying their contrasts helps to mechanistic context, as changes in relative rate reduce confusion. We therefore present a com- or timing of developmental processes (e.g., prehensive summary of natural modes of evo- Raff and Wray 1989). Application of this con- lutionary modification of ontogeny relevant to cept requires direct knowledge of those pro- morphologists. cesses. This was a significant change in em- phasis: any developmental process has a tem- The Confusion about Heterochrony poral aspect to its operation and thus becomes Much of the confusion surrounding heter- susceptible to heterochronic modification ochrony can be traced to the application of in- whether or not the process is directly con- consistent criteria for recognizing the phe- cerned with morphological development. The nomenon. Heterochrony has been defined as two concepts are not commensurate: cases changes in developmental rate or the relative meeting the criteria of one definition might time of appearance of features that ‘‘produce not meet those of the other (e.g., Raff and 356 MARK WEBSTER AND MIRIAM L. ZELDITCH

Wray 1989; Alberch and Blanco 1996; Raff can be quantified in two ways: (1) the onto- 1996). McKinney and McNamara (1991; also genetic change in geometric shape (estimated McNamara 1995, 1997) later defined heteroch- by multivariate regression of geometric shape rony as ‘‘change in timing or rate of develop- on geometric scale; Fig. 1A,B), or (2) the allo- mental events, relative to the same events in metric coefficients of size measurements (also the ancestor.’’ This third definition removes estimated by multivariate regression on body all constraint of morphological parallelism be- size; Fig. 1C,D). Interspecific comparison of tween ontogeny and phylogeny and does not patterns of shape change involves calculation require mechanistic knowledge of the process: of the vector correlation between the two on- the only requirement is that an event (however togenies. On purely intuitive grounds we defined) occurs at a different time or that de- might reject the hypothesis of a shared trajec- velopment proceeds at a different rate in the tory of ontogenetic shape change between descendant relative to the ancestor. When a these piranha species given the relatively class signified by a term is so broadly encom- weak correlation between their trajectories ob- passing and heterogeneous (‘‘serves too many tained from the geometric data (0.332). How- masters’’ [Wake 1996]) it can generate confu- ever, we might find it equally reasonable on sion because its theoretical implications de- intuitive grounds to accept that these species pend specifically on which meaning is intend- are very similar in light of the high correlation ed, but the language is presently too imprecise between their trajectories obtained from the to allow us to specify that meaning. traditional size data (0.995). Not only does heterochrony have various That such a difference in magnitude of cor- meanings, but the criteria for applying it to relation (and in potential inference about the data are inconsistent as well. For example, to underlying biological process) is caused by apply Gould’s (1977) definition to data (as for- differences in approach to measurement is malized by Alberch et al. 1979) we need an disturbing. However, the difference is less- axis of ontogenetic shape change common to ened when the correlations are subjected to all species under comparison. Hypotheses of statistical testing (using a resampling proce- the operation of heterochrony in this sense can dure detailed in Zelditch et al. 2003b): both be accepted only once it is demonstrated that data sets then reveal significant differences be- all interspecific differences are explicable in tween the ontogenies. The striking contrast in terms of rate or timing differences along that the correlations estimated from the two data shared trajectory of ontogenetic shape change. sets results from the different scales on which If the trajectory of shape change was not con- the correlations are estimated. To compare the served, then parallelism between ontogeny values we need to calibrate them, taking into and phylogeny was lost and spatial rather account the values we would expect between than (or in addition to) temporal aspects of random vectors. For the geometric data, 1000 ontogeny must have been evolutionarily mod- random permutations of the ontogenetic vec- ified. Modifications to spatial aspects of de- tor of each species yield average correlations velopment lie within the realm of ‘‘non-het- for the two species of 0.0125 and 0.0047 (for P. erochrony’’ (see Alberch 1985). However, cri- denticulata and P. nattereri, respectively), with teria by which shape is summarized, and by 95% confidence intervals ranging from Ϫ0.322 which hypotheses of shared trajectories of to 0.257 and Ϫ0.265 to 0.257. These are the val- shape change are tested, differ among work- ues we would expect from independent vec- ers (see also ‘‘Dimensionality Bias,’’ below). tors. In contrast, random permutations of the To exemplify the difficulty of testing the hy- traditional morphometric data yield an aver- pothesis of heterochrony in light of inconsis- age correlation of 0.9870 and 0.9921 for P. d e n - tencies in approach, consider an example ticulata and P. nattereri, respectively, with 95% drawn from a comparison between the pira- confidence intervals ranging from 0.9841 to nha species Pygopristis denticulata and Pygo- 0.9920 and 0.9891 to 0.9946. The correlations centrus nattereri (Zelditch et al. 2003a) (Fig. 1). expected from independent vectors for tradi- The ontogenies of shape for the two species tional morphometric data are strong. Thus, for MODIFICATIONS OF ONTOGENY 357

FIGURE 1. Ontogenetic trajectories of shape change in two species of piranha, each depicted by using geometric shape data and traditional morphometric measurements. Locations of 16 landmarks selected to summarize shape of each specimen are shown in top figure (see Zelditch et al. 2003a for details). Differences in the relative position of homologous landmarks between conspecific specimens of different sizes (developmental ages) can be used to quantify patterns of ontogenetic shape change. A, Ontogeny of geometric shape for Pygopristis denticulata shown as a deformation using the thin-plate spline (Bookstein 1991; see Webster et al. 2001 for a nonmathematical sum- mary). B, Ontogeny of geometric shape for Pygocentrus nattereri shown as a deformation using the thin-plate spline. C, Allometric coefficients for traditional length (inter-landmark distance) measurements of Pygopristis denticulata. Va lu e s Ͼ1.0 indicate positive allometry (relative to standard body length) during ontogeny; values Ͻ1.0 indicate negative allometry. D, Allometric coefficients for traditional length measurements of Pygocentrus nattereri. Regres- sion of the full set of partial warps (including the uniform component) from the geometric data, or of the full set of allometric coefficients from the traditional morphometric data, against a measure of geometric scale (centroid size; see Fig. 3) gives the trajectory of ontogenetic shape change for a species. The degree to which two species share similar patterns of ontogenetic shape change is given by the vector correlation between their respective tra- jectories (the dot product of the ontogenetic vectors, normalized to unit length). The inverse cosine of the vector correlation is the angle between the ontogenies. Ontogenies identical in their patterns of shape change have an angle of 0Њ between them; larger angles indicate differences in allometric patterning. 358 MARK WEBSTER AND MIRIAM L. ZELDITCH both data sets, the interspecific correlation is berg 1998; Godfrey and Sutherland 1995, only marginally higher than one expected 1996). One methodological issue (herein from two random vectors. termed ‘‘dimensionality bias’’) is critical to the Without taking into account the very high identification of modes of developmental re- correlations that can be produced solely by programming in general and to the testing of chance when analyzing traditional morpho- hypotheses of heterochrony in particular, and metric data, it is difficult to appreciate the yet has been almost entirely overlooked. The magnitude of the difference between the on- nature and significance of dimensionality bias togenies. Recognizing that these species are are therefore now discussed. very different in their trajectories of shape change is clearly important in light of the def- Dimensionality Bias: Methodologically inition of heterochrony given by Gould (1977) Determined Biological Conclusions and implicit in its formalization by Alberch et The mode(s) of evolutionary changes in de- al. (1979). Of course, were we to use the broad velopment that can be detected in a particular definition that equates heterochrony to any case can be limited by the complexity with change in the slope or intercept of allometric which morphological features are described. coefficients (McKinney and McNamara 1991) When complex morphologies are reduced to we would necessarily accept the hypothesis of single dimensions, the variety of modifica- heterochronic evolution. We would infer that tions that could be discerned is drastically re- P. nattereri is neotenic relative to P. denticulata duced. We therefore term this ‘‘dimensionali- in those measurements in which it displays ty bias.’’ lower allometric coefficients (e.g., in postcra- Whether shape data are interpreted as such, nial depth and posterior lengths), and is ac- or instead reduced to a collection of individ- celerated in development in those measure- ual size measurements that are interpreted ments in which it displays higher allometric separately, is the primary cause of dimension- coefficients (e.g., anterior head depth and ality bias. This bias has its greatest effect on lengths). If we followed an alternative conven- the ability to recognize changes in ontogenetic tion favored by many workers (e.g., Shea 1985; allometries, and therefore on the ability to re- Wayne 1986), we would infer neoteny or ac- ject a hypothesis of heterochrony which, fol- celeration from contrasting changes in posi- lowing the definition of Gould (1977), requires tively and negatively allometric coefficients: that the ancestral pattern of ontogenetic shape neoteny predicts that positively allometric co- change be conserved in the descendant. If the efficients are decreased whereas negatively al- pattern of shape change itself is evolutionarily lometric coefficients are increased. Under a modified, then ancestor and descendant do more process-based definition of heterochro- not share a common ‘‘axis of shape change,’’ ny (e.g., Raff and Wray 1989) we would have and their ontogenetic trajectories cannot be to conclude that these data are immaterial be- plotted on the same shape/size/time plot (Al- cause they contain no direct information about berch et al. 1979; Alberch 1985), irrespective of process: any attempt to draw a conclusion the rate at which the taxa progress along their about the nature of the change in ontogeny respective trajectories or when progressions would be regarded as suspect. along those trajectories are initiated or ter- Heterochrony in one guise or another has minated. However, the degree to which taxa frequently been documented in empirical case are determined to share the same trajectory of studies (indeed, it ‘‘explains everything’’ [Mc- ontogenetic shape change depends on the Namara 1997: p. 46]). However, the inconsis- complexity with which shape is summarized. tencies and ambiguities associated with the Shape is inherently multidimensional, and concept of heterochrony and its evidentiary trajectories of shape change must therefore be criteria, which almost certainly account for quantified by utilizing multidimensional vec- some of its apparent prevalence, have rarely tors (e.g., O’Keefe et al. 1999; Zelditch et al. been stressed except in the context of quite 2000; Webster et al. 2001; Roopnarine 2001; mathematical technical issues (e.g., Klingen- Nehm 2001; also piranha example above and MODIFICATIONS OF ONTOGENY 359 trilobite example below). Reducing the de- lutionary change futile. Studies using the scription of shape to a collection of univariate same methodology are subject to the same po- size measurements, regressing each indepen- tential biases, and their results are directly dently against time, and comparing the re- comparable. When comparing modes of evo- sulting bivariate regression parameters can- lutionary change between studies it is there- not test a hypothesis of heterochrony (nor doc- fore critical to consider dimensionality bias: ument it). ‘‘Heterochrony’’ detected in such a do observed similarities/differences in evo- way is an artifact of the geometric constraint lutionary mode result from biological reality imposed on the data: one-dimensional data or methodological approach? This is especial- are constrained in that taxa can differ only in ly important when compiling cases to assess (1) the point at which progress along that di- the relative contribution of various modes to mension starts or stops, or (2) the rate of pro- evolution (e.g., McNamara 1988). gress along that dimension. Any interspecific difference in such bivariate regression param- Beyond Heterochrony eters can be interpreted only in terms of a dif- Most workers acknowledge that some evo- ference in rate and/or timing, and detection of lutionary modifications to ontogeny lie be- ‘‘heterochrony’’ is methodologically guaran- yond the realm of changes in developmental teed for each measure of size in which the taxa rate or timing, and that the nomenclatural differ at a given age. (Note that the only evo- scheme of progenesis, hypermorphosis, pre- lutionarily conserved aspect of ontogeny im- and postdisplacement, and neoteny and ac- plied by detection of ‘‘heterochrony’’ using celeration is incomplete in its description of this methodology is a homologous length evolutionary phenomena (e.g., Alberch 1985; measurement on ancestor and descendant.) Gould 2000). Various non-heterochronic Restricting the morphological coverage to modes of developmental reprogramming more localized regions to test for conserved have been proposed (see below), but each is patterns of shape change on a local scale can potentially subject to the same problems as produce a similar bias in favor of interpreting the concept of heterochrony. The problems any interspecific differences as heterochrony, plaguing heterochrony are in most immediate in that this often decreases the number of danger of being repeated by the concept of shape variables and thus the dimensionality heterotopy (a change in the topology of de- of the analysis. There is no rule dictating how velopment; Haeckel 1875; Gould 1977; Wray many dimensions an analysis must include and McClay 1989; Zelditch and Fink 1996; Li before hypotheses of constraint in shape and Johnston 2000). Some workers have con- change (local or otherwise) can be considered sidered heterotopy as a modification to spatial adequately tested. However, when the number aspects of morphological development (e.g., of inferred ‘‘local heterochronies’’ required to Zelditch and Fink 1996; Webster et al. 2001). explain the data in terms of modification to Others have regarded heterotopy as any spa- temporal aspects of ontogeny alone approach- tial dissociation in development. This is ex- es the number of variables by which shape is emplified by a statement by Raff (1996: p. 335) summarized, the data may be more parsimo- who included under heterotopy ‘‘such diverse niously explained by an alternative hypothe- events as shifts of gene expression from one sis (such as allometric repatterning, discussed cell type or group of cells to another, homeotic below). changes, production of serial homologues, Dimensionality bias is not unique to the production of repeated structures, changes of modes of evolutionary change recognized location of structures relative to the body axis herein. Rather, dimensionality bias pervading or some other frame of reference, and changes all previous concepts and classifications of in relative proportions of structures.’’ Incom- modes of evolution has been overlooked, ig- mensurability of meaning across these scales nored, or hidden. Acknowledging the exis- is inevitable. The ambiguity in precise defini- tence of a dimensionality bias does not auto- tion means that any modification to ontogeny matically render recognition of modes of evo- failing to meet the criterion of heterochrony is 360 MARK WEBSTER AND MIRIAM L. ZELDITCH likely to be considered a case of heterotopy, typy, homeosis), and amount/abundance which could devolve into a virtually meaning- (heteroposy), as opposed to temporal param- less concept (a wastebasket for cases of ‘‘non- eters. Others (heteromorphy, novelty, and de- heterochrony’’). viation) are defined in terms of what they are To forestall that possibility, we present here not (i.e., wastebasket classes for cases failing a complete consideration of modes of modifi- to meet other criteria). Some workers have al- cation of morphological development (i.e., luded to the existence of additional ontoge- modification to the dynamics of morphologi- netic parameters that may be evolutionarily cal change, and to the distribution of morpho- modified but did not explicitly identify them logical features on the organism). A consid- (e.g., Atchley 1987; Sattler 1992). This leads to eration of modes of modification to more uncertainty as to the number of discrete non- proximal developmental processes might heterochronic modes of ontogenetic modifi- seem preferable, but a morphological ap- cation, with proposals ranging from one proach has merit in that: (1) morphology rep- (‘‘novelty’’ [Gould 1977]) to a potentially in- resents one direct link between the organism finite number (Sattler 1992). and the surrounding environment and is sub- Our characterization of modes (Table 1) ject to selective pressures; (2) shared modifi- achieves comprehensiveness of description cations of morphological development may re- without recourse to wastebasket classes or un- flect shared changes in life history; (3) the specified ‘‘phantom parameters.’’ The defini- types of observed changes in morphological tions are intended to be more restrictive and development can constrain hypotheses re- rigorous, and therefore of more utility in em- garding causes at more proximal (mechanis- pirical studies. Testing hypotheses of the op- tic) levels; and (4) it is applicable to specimen- eration of each mode of change entails exam- based studies, including paleontological ex- ining the morphological development of the amples. ancestor and descendant by using trajectory- For each mode recognized here we include based studies of ontogenetically dynamic fea- a diagnosis that offers unambiguous criteria tures (i.e., aspects of morphology which for its recognition. Hypotheses of the opera- change during ontogeny; Fig. 1) and non-tra- tion of each mode of modification therefore jectory-based studies of ontogenetically static become rigorously testable. We recognize dis- features (i.e., aspects of morphology which tinct modes of ontogenetic modification rather can be considered ‘‘fixed’’ or unchanging dur- than a continuum with ‘‘pure heterochrony’’ ing ontogeny). Although discussed in evolu- and ‘‘pure heterotopy’’ as end-members (con- tionary terms, the modes also account for her- tra the view taken by Zelditch and Fink itable morphological traits differentiating [1996]). subspecies, populations, or dimorphs. Hy- potheses of modification at any taxonomic lev- Modes of Modification to el must be framed within a defendable phy- Morphological Development logenetic context (Fink 1982, 1988; Jaecks and Several modes of evolutionary change of on- Carlson 2001). togeny have been proposed in previous stud- A comprehensive reexamination of empiri- ies, including heterotopy, heterometry, heter- cal studies in light of the modes recognized otypy, homeosis, heteroposy, heteromorphy, here is beyond the scope of this paper. Nev- heteroplasy, novelty, and deviation, in addi- ertheless, it is important to demonstrate the tion to heterochrony (e.g., Zimmermann 1959; applicability of the scheme, and so examples Takhtajan 1972; Gould 1977; Regier and Vla- of each of the modes are here cited. In many hos 1988; Sattler 1992; Wake 1996; Zelditch cases, additional data are required for ade- and Fink 1996; Gellon and McGinnis 1998; Ar- quate testing of alternative hypotheses as a re- thur 2000; Li and Johnston 2000; Sundberg sult of our expanded and refined framework. 2000). These terms refer to modification of on- Rate Modification: Modification of the Rate of togenetic parameters such as location (heter- Morphological Development Relative to Time. otopy), number (heterometry), type (hetero- This can be detected by plotting the magni- MODIFICATIONS OF ONTOGENY 361

TABLE 1. Modes of evolutionary modification of morphological ontogeny recognized herein. Dynamic aspects of morphological development (i.e., relating to features that change during ontogeny) are susceptible to modification through rate or timing modifications and/or allometric repatterning (depending on whether or not patterns of shape change are assessed). Static aspects of morphological development (i.e., relating to features that are consid- ered ‘‘fixed’’ or unchanging during ontogeny) are susceptible to heterotopy, heterotypy, or heterometry. See text for details.

Ontogenetic Aspect of morphology behavior oftrait investigated Potential modes of evolutionary modification Dynamic changes in type, num- No modification ber, location, or size of struc- Rate modification ture during ontogeny Timing modification (including deletion) Dynamic No modification Rate modification (ϭ rate heterochrony) Ontogenetic shape change Timing modification (ϭ event heterochrony) Rate modification ϩ allometric repatterning Timing modification ϩ allometric repatterning Allometric repatterning Location of structure (fixed No modification through ontogeny) Heterotopy Static Type of structure (fixed through No modification ontogeny) Heterotypy Number of structure (fixed No modification through ontogeny) Heterometry (including deletion) tude of change away from the condition ob- parameters have been modified). An in- served at the youngest stage against time. A creased or decreased rate of shape change change in rate is evident as a difference in then results in peramorphosis (acceleration) slope between the regression lines of the an- or paedomorphosis (neoteny), respectively. cestor and descendant when the relationship Morphological parameters such as size are of- between development and time is linear, or by ten used as a proxy for time (developmental changes in the parameters of a nonlinear age), but detection of a rate modification with growth model when the relationship is non- such data does not provide the information linear (see Zelditch et al. 2003b). The aspect of needed to determine which parameter (mor- morphology under consideration may be phological change or size) has been decoupled shape, size, or the location, number, or type of from the ancestral relationship to time. particular elements, but must obviously be Timing Modification: Modification of the Rela- ontogenetically dynamic if ancestor and de- tive Sequence and/or Temporal Spacing of Events in scendant differ in terms of their rate of change Morphological Development. Examples of (Table 1). Evolutionary change in the rate at events would include the appearance of par- which morphological differentiation is ticular anatomical structures or relationship achieved during ontogeny represents a rate between structures, or discrete ‘‘kinks’’ in the modification even when species follow differ- trajectory of ontogenetic shape change (onto- ent ontogenies of shape and so cannot be com- genetic changes in allometric patterning; per- pared by using the Alberch et al. (1979) model haps marking entry into a different phase of (e.g., Zelditch et al. 2003b). Parameters sum- development, including termination of marizing size and net shape change are sug- growth if appropriate) (Table 1). The events gested below. For shape data, a rate modifi- must be conserved between ancestor and de- cation may be further qualified as a rate het- scendant: only their relationship to ontoge- erochrony (sensu Gould 1977 and Alberch et netic time is modified (Alberch and Blanco al. 1979) when the ancestral trajectory of 1996). The ontogenetic trajectory can be mod- shape change is conserved in the descendant eled as a linear series of events, arranged ac- (because parallelism between ontogeny and cording to the order in which they occur dur- phylogeny is then retained, and only temporal ing ontogeny (Fig. 2). The parameter of inter- 362 MARK WEBSTER AND MIRIAM L. ZELDITCH

been modified). A later or earlier termination of an otherwise conserved trajectory of shape change then results in peramorphosis (hyper- morphosis) or paedomorphosis (progenesis) in the descendant, respectively. Conversely, a later or earlier initiation of an otherwise con- served trajectory of shape change results in paedomorphosis (postdisplacement) or pera- morphosis (predisplacement) in the descen- dant, respectively. Timing modification is most evident when the temporal order of events is modified in the descendant relative to the ancestor (Fig. 2A). This can be detected in the absence of absolute temporal information (Smith 2001) and sug- gests developmental modularity, with non- contingency among the switched events. However, modification to the temporal spac- ing between events can occur without dis- placing one event temporally beyond another, conserving the rank order of events (Fig. 2B). It is then conceivable that constraint is oper- ating: later events may be contingent upon earlier events. Detection of this more subtle timing modification does require absolute temporal information. (There seems little rea- son to assume that timing modifications re- sulting in a switch in rank order of develop- mental events are always fundamentally dif- FIGURE 2. Detecting timing modification. In each case, ferent from those that modify timing of events species A is ancestral to species B. A, Timing modifi- cation involving a change in rank order of events. Event without altering their rank order. Thus 4 has been temporally displaced in species B. Note that Smith’s [2001] concept of ‘‘sequence heteroch- no reference to absolute ontogenetic time need be made rony,’’ which she defined as an evolutionary to detect the change. B, Timing modification without a change in rank order of events. Event 4 has been tem- switch in rank order of developmental events, porally displaced in species B but to a lesser extent than is subsumed within the revised concept of in Figure 2A. Absolute ontogenetic time must be known timing modification.) Timing modification to detect such a change. cannot be inferred from comparison of the rel- ative timing of one event alone in the ancestor est is the temporal spacing between these and descendant, as any differences detected events during ontogeny. A given event may could result from rate modification. The cri- occur earlier or later in the descendant ontog- terion that one event is evolutionarily dis- eny, and can be considered to have been pre- placed differentially relative to another event displaced or postdisplaced, respectively, rel- is evidence for a timing rather than a rate ative to other events. For shape data, a timing modification. modification may be further qualified as an Examples of Rate and Timing Modifications. event heterochrony (sensu Gould 1977 and Al- Many empirical studies have found support berch et al. 1979) when the ancestral trajectory for ‘‘heterochrony,’’ but rarely do these ade- of shape change is conserved in the descen- quately test for evolutionary conservation of dant (because morphological parallelism be- the ancestral pattern of ontogenetic shape tween ontogeny and phylogeny is then re- change. The ‘‘shape data’’ in these studies are tained, and only temporal parameters have often one-dimensional size axes and thus are MODIFICATIONS OF ONTOGENY 363 subject to the extreme form of the dimension- ment was modified between morphs, this rep- ality bias discussed above. Modified rate of resented a case of a timing modification with septal insertion relative to shell size in Car- allometric repatterning. Similarly, the loss of boniferous ammonoids (Stephen et al. 2002), proximale formation in Democrinus maximus change in the timing of proximale formation with modification of shape of the basals (Kjaer relative to thecal size in bourgueticrinid cri- and Thomsen 1999) represented a rate or tim- noids (Kjaer and Thomsen 1999), change in ing modification with allometric repatterning. the timing of fusion of the scapula and cora- A case of event heterochrony was recently coid relative to attainment of maturity in croc- documented in marginellid gastropods odilians (Brochu 1995), and modified rates of (Nehm 2001). His study demonstrates that cranidial widening, ocular lobe shortening, or evolutionary conservation of the ancestral tra- glabellar elongation relative to cranidial jectory of shape change can be detected by us- length in trilobites (Nedin and Jenkins 1999) ing multidimensional morphometric tech- therefore support hypotheses of rate or timing niques. The rigorous criteria by which heter- modification but not of heterochrony. Similar- ochrony is defined (above) can be met by em- ly, Alberch and Blanco (1996) and Smith (1997, pirical data. 2001) documented several morphological Several studies document rate or timing ‘‘event heterochronies’’ in vertebrate evolu- modification in dynamic aspects of ontogeny tion, but pending investigation of the degree other than shape. Evolutionary decrease in to which patterns of shape change were evo- thickness (including loss) of the secondary fi- lutionarily conserved these examples are best brous shell layer in thecideide brachiopods qualified as timing modifications. (Jaecks and Carlson 2001) represented a tim- Several studies assess temporal aspects of ing modification, in that the ontogenetic inter- morphological development (often using size val over which the secondary layer was de- as a proxy for age) and include consideration posited was shortened (assuming a constant of patterns of shape change, and can therefore rate of shell deposition). The decrease in size potentially detect heterochrony. Rate modifi- of thecideide brachiopods documented by cation coupled with allometric repatterning these authors represented either a rate or a (see below) of body form has been document- timing modification (pending age data). Fur- ed in piranhas (Zelditch et al. 2000, 2003a) and ther examples of rate modification are docu- between the rodents Mus musculus domesticus mented in the trilobite genus Nephrolenellus be- and Sigmodon fulviventer (Zelditch et al. low. 2003b). The trilobite Aulacopleura konincki ex- Heterotopy: Modification of the Location of a hibited several morphs differentiated by the Particular Morphological Feature. The result is number of thoracic segments in maturity a ‘‘topological shuffling’’ of features such that (Hughes and Chapman 1995). This difference novel proximity relationships are established resulted from delayed termination of segment between features. Heterotopy can be consid- release from the transitory pygidium into the ered ‘‘morphological ectopy without replace- thorax (and of segment generation at the pos- ment,’’ in contrast to heterotypy (below). It terior of the transitory pygidium) relative to differs from heterometry (below) in that het- morphological development: a timing modi- erotopy does not affect the number of ele- fication. However, Hughes and Chapman ments. (This definition of heterotopy is more (1995; Hughes et al. 1999) found that the var- restricted than previous definitions [e.g., Raff ious morphs showed no significant difference 1996; Zelditch and Fink 1996], which incor- in overall mature body proportions despite porate cases that are now classified as heter- the differing number of thoracic segments. In otypy, heterometry, and allometric repattern- morphs with more thoracic segments in ma- ing.) Heterotopy should be applicable only to turity, each segment was therefore relatively morphological features that are considered shorter (longitudinally) than in the morphs fixed in location once formed (Table 1). If a with fewer segments. As the pattern of onto- structure undergoes migration during ontog- genetic shape change followed by each seg- eny, then an evolutionary difference in its lo- 364 MARK WEBSTER AND MIRIAM L. ZELDITCH cation may alternatively be interpretable in pects of trilobite evolution in terms of home- terms of temporal modification of the migra- otic change. However, his criterion for home- tory path, and/or in terms of repatterning of otic change (‘‘distribution changes of charac- the migratory path. ters among body segments’’) is rather liberally Examples. The classic case of heterotopy, applied, and not consistently differentiated used by Haeckel in his original formulation of from his concept of ‘‘heterochronic’’ change. the concept, is the shift in germ layer (from the In our classification, his cases of homeosis are ectoderm or endoderm into the mesoderm) better described in terms of rate modification, from which reproductive organs differentiate, heterometry, or allometric repatterning (see which must have occurred at some point in below). the evolution of triploblasts from diploblasts. Deposition of an acicular calcitic shell layer Heterotopy has also been documented in in thecideide brachiopods (Jaecks and Carlson plants (reviewed in Li and Johnston 2000), 2001) represented heterotypy. The acicular typically involving a change in the site of ini- layer was discretely different from other shell tiation of organ primordia. microstructures, and its deposition was not Heterotypy: Modification of the Type of Morpho- contingent upon first depositing the typical fi- logical Structure at a Given Location. A mor- brous shell layer (e.g., on the ventral valve of phological structure of one type is found in Thecidea radiata). Description of the introduc- the descendant in place of one of a different tion of the acicular layer in terms of pera- type found in the ancestor. Heterotypy should morphic extrapolation of shell microstructure be applicable only to morphological features ontogeny (going ‘‘beyond’’ the secondary lay- that are considered fixed in type once formed. er microstructure) is therefore unsatisfactory. Heterotypy can be considered ‘‘morphologi- Similarly, the transformation of walking leg to cal ectopy with replacement’’ or a ‘‘change in maxilliped during the ontogeny of Porcellio type,’’ in contrast to heterotopy (Table 1). In scaber (Abzhanov and Kaufman 1999) repre- contrast to heterometry (below), heterotypy sents a novel heterotypic change introduced at involves a decrease in number of one type of a relatively late ontogenetic stage of this iso- structure for every increase in number of an- pod (assuming that this transformation was other. Heterotypy therefore excludes cases of not present in the ontogeny of its ancestor). ‘‘homeosis’’ involving a net change in the total Heterometry: Modification of the Number of a number of features. Particular Morphological Feature. Ancestor Examples. Homeotic replacement of one and descendant differ in the number of a par- type of appendage by another in arthropods ticular type of feature, without ectopic re- represents heterotypy. The famous Antenna- placement, and not reached through trunca- pedia and bithorax mutants of Drosophila rep- tion or prolongation of the ancestral ontoge- resent examples of (non-evolutionary) hetero- netic trajectory. Heterometry should be appli- typic change (Lewis 1978). In Antennapedia cable only to morphological features that are mutant flies, walking legs develop on the head considered fixed in number through ontoge- in place of antennae. In bithorax mutants, ny. Supernumerary features may develop in wings develop in place of halteres. Similar ho- novel locations (without replacing existing meotic changes in plants, such as transfor- structures). Reduction in the number of a type mation of stamens into petals or petalodia, of feature is simple deletion (loss) and does represent heterotypy (reviewed in Li and not involve replacement by a different struc- Johnston 2000). However, not all documented ture. cases of homeotic change in arthropods rep- Heterometry differs from heterotopy in that resent heterotypy. Many cases involve devel- heterotopy relates to change in the location of opment of appendages on previously non-ap- a given morphological feature on the organ- pendicular segments or loss of appendages ism, with no increase or decrease in the num- from a segment altogether, and so represent ber of that feature. Heterometry differs from heterometry in the revised scheme (see be- heterotypy in that heterotypy relates to a dis- low). Sundberg (2000) interpreted many as- crete change in type of feature at a given site MODIFICATIONS OF ONTOGENY 365 on the organism (there is no net change in the togenetic shape change followed by a struc- total number of features). Heterometry de- ture or the organism (Fig. 1). The parameter scribes increase or decrease in the total num- of interest is the pattern of shape change fol- ber of features. The requirement that the lowed by a structure or individual over a spec- change in number cannot be described ified ontogenetic interval, and is necessarily through truncation or prolongation of the an- ontogenetically dynamic (Table 1). The dis- cestral ontogenetic trajectory (i.e., the number tinction between allometric repatterning and is ontogenetically fixed rather than dynamic) rate or event heterochrony lies in their differ- serves to differentiate heterometry from rate ent impacts on ratios among allometric coef- or timing modification (Table 1). In the case of ficients: in the case of heterochrony these ra- complete absence of an ancestral morpholog- tios are conserved between ancestor and de- ical feature from the descendant, knowledge scendant, but in the case of allometric repat- of the temporal dynamics of the feature in the terning they are modified (subject to ancestor is therefore critical for distinguishing dimensionality bias; see above) and parallel- (deletive) heterometry from (extreme) timing ism between ontogenetic and phylogenetic modification. shape change is lost. Heterochrony and allo- The term heterometry was used by Arthur metric repatterning are therefore mutually ex- (2000) to describe ‘‘change in amount’’ (simi- clusive categories when assessed for the same lar to Regier and Vlahos’s [1988] ‘‘heteropo- morphological structure (Table 1): they differ sy’’: a change in the abundance of a develop- in terms of constraint or channeling (or lack mental process). His example of a heterome- thereof) along the ancestral ontogenetic trajec- tric change was the production of a higher tory, with implications regarding how devel- concentration of BICOID caused by an in- opment was modified at more proximal levels. creased rate of transcription in the nurse cells. In some cases, difference in form of a ho- This would not be classed as heterometry in mologous structure in the ancestor and de- the present scheme, as the example is con- scendant can be described either in terms of cerned with amount of gene product rather modified ontogenetic allometry (allometric re- than morphology (and so belongs at a more patterning) or in terms of qualitative differ- proximal process level). ence in type (perhaps defined by function; Examples. Results of the experimental ma- heterotypy). This is particularly evident in nipulation of the eyeless gene in Drosophila, cases concerning appendage morphology whereby supernumerary compound eyes de- (e.g., vertebrate forelimb versus hind limb; velop on the antennae, legs, and wings of the some arthropod appendages), but is also seen insect (Halder et al. 1995), represent a (non- in cases of other features (e.g., brachiopod evolutionary) case of heterometry. Develop- lophophore support structures; see below). ment or loss of axial nodes on the glabella Cases where ultimate interpretation is deter- and/or thorax of some trilobites (see also case mined by choice of analytical style are exam- study below), described as homeotic change ples of dimensionality bias in a broad sense by Sundberg (2000), represented heterometry (above). When the form of one structure can- as the nodes were added or deleted rather not be (or is not) described in terms of dynam- than being replaced by another structure. Sim- ic modification of elements present in another ilarly, cases of appendage loss or the devel- (e.g., changes in shell microstructure or ap- opment of appendages on typically non-ap- pendage type), then any evolutionary change pendicular segments in arthropods (e.g., will be interpreted as heterotypy. Heterotypic Akam et al. 1994; Carroll 1994; Gellon and change can be treated as ontogenetically spon- McGinnis 1998; Kettle et al. 1999) are de- taneous, whereas difference in form resulting scribed as heterometry rather than heterotypy from allometric repatterning can gradually in the revised classification. develop over a longer interval of ontogeny. Allometric Repatterning: Modification of the Heterometry and heterotopy are distinct Pattern of Ontogenetic Shape Change. Ancestor from allometric repatterning in that they re- and descendant differ in the trajectory of on- late to modification of the number and spatial 366 MARK WEBSTER AND MIRIAM L. ZELDITCH distribution of structures on the organism (Ta- was itself modified (Stephen et al. 2002: Fig. ble 1). Allometric repatterning describes the 6.1). The rate of increase in shell width was not dynamics of shape change (of a structure or modified to the same degree as the rate of in- the organism) through ontogeny. Of course, crease in shell diameter between the di- rate or event heterochrony, heterotypy, heter- morphs. A hypothesis of allometric repattern- ometry, or heterotopy in local morphological ing (with rate modification) is more satisfac- structures can have a ‘‘shunting’’ effect re- tory than one of independent heterochronic sulting in modification to shape on a more modification to each of the morphological pa- global scale. Wake and Roth (1989; Wake 1996) rameters by which shape was summarized. termed the complex morphological outcome Other examples of allometric repatterning of a mosaic of heterochronic changes on a local coupled with rate or timing modifications scale (with some traits unmodified) ‘‘ontoge- were discussed above. netic repatterning.’’ If allometric repatterning Evolutionary change in the degree of com- on one scale is to be fully accounted for by plexity of brachiopod lophophore support such processes on a more local scale, then it structures (Jaecks and Carlson 2001) high- should be possible to demonstrate conserva- lights the need to consider methodological di- tion of ancestral patterns of ontogenetic shape mensionality bias. If assessed in terms of pat- change at that local scale (subject to dimen- terns of shape change, then evolutionary mod- sionality bias, above). Wake (1996) used the ification would have been either through al- term heteroplasy to describe interspecific dif- lometric repatterning of the ontogenetic ferences in rates of cell proliferation leading to development of the structures (perhaps with differences in patterns of allometry. Identify- rate or timing modification), or through rate ing heteroplasy therefore requires knowledge or event heterochrony. However, if support of process, and the concept lies at a more prox- structure morphology were assessed in terms imal level of investigation. (Not all cases of al- of a linear progression of ‘‘types’’ (i.e., degrees lometric repatterning will ultimately stem of complexity), then evolutionary modifica- from modification to the rate of cell prolifer- tion would likely be described in terms of rate ation: differences in rate of cell death, cell size, or timing modifications (perhaps with heter- or cell arrangement, for example, would also otypy, depending on whether the succession lead to allometric repatterning at the structur- of types follows a conserved order). In the lat- al scale.) ter case, no inference is made regarding con- Examples. Allometric repatterning has strained patterns of shape change, and hy- been documented in the evolution of the ce- potheses of heterochrony or allometric repat- phalon of the Early Cambrian trilobite Ne- terning have not been tested. phrolenellus geniculatus from N. multinodus (Webster et al. 2001; see also case study be- A Detailed Case Study: Morphological low), in the evolution of the piranha genera Evolution in the Trilobite Nephrolenellus Serrasalmus and Pygocentrus (Zelditch et al. An empirical case study will serve to dem- 2000) (allometric repatterning was considered onstrate how the modes of ontogenetic mod- synonymous with heterotopy in both of these ification can be recognized. Trilobites have studies), and between the rodents Mus mus- long been a source of empirical support for culus domesticus and Sigmodon fulviventer heterochrony in paleontology (e.g., McNa- (Zelditch et al. 2003b). mara 1978, 1981, 1983, 1986b, 1988; but see Stephen et al. (2002) documented ‘‘heter- Webster et al. 2001) and it is appropriate that ochrony’’ affecting shell width and diameter the evolution of a Cambrian olenelloid trilo- in dimorphic Carboniferous ammonoids. The bite is used as an example. compressed morph showed slower increase in A strong case can be made for the evolution each of these traits relative to the number of of Nephrolenellus geniculatus from its ancestral septa laid down (a proxy for time) in compar- sister taxon N. multinodus through peramorph- ison to the depressed morph. However, the re- ic heterochrony (Fig. 3) (Webster et al. 2001; a lationship between shell width and diameter systematic revision of the clade plus formal MODIFICATIONS OF ONTOGENY 367

FIGURE 3. A, The ontogenetic development of the cephalon (head shield) of the trilobite Nephrolenellus geniculatus (top) and its ancestral sister taxon N. multinodus (bottom). Development of the cephalon of these animals was di- vided into four discrete phases (see Webster et al. 2001), each represented here (size measurement refers to sagittal cephalic length). The distinguishing features of mature specimens (phase 4) of these species are (1) the fewer axial nodes on the glabella of N. geniculatus; (2) the stronger adgenal angle in N. geniculatus; and (3) the absence of a preglabellar field in N. geniculatus. Note that in the ancestral species the axial nodes are progressively lost, the adgenal angle progressively developed, and the size of the preglabellar field progressively reduced during ontog- eny. These facts can be used to make a strong case for N. geniculatus having evolved from N. multinodus by simple extrapolation of the ancestral ontogenetic trajectory—a case of peramorphic heterochrony (see Webster et al. 2001). B, Morphological terms used in describing the trilobite cephalon. The glabella consists of lobes LO, L1, L2, L3, and LA, separated by furrows. Each of the circles represents a landmark used in the geometric morphometric analysis of cephalic development (Webster et al. 2001). Black circles represent landmarks also used in the analysis of oculo- glabellar development in the present paper. For clarity, landmarks are shown for the right half of the cephalon only. For such geometric data, size of a specimen can be quantified as its centroid size (the square root of the sum of squared distances between each landmark and the centroid of the form [Bookstein 1991]). 368 MARK WEBSTER AND MIRIAM L. ZELDITCH

represents a rate modification, but is far from a complete description of the interspecific on- togenetic differences. Was the increased relative rate of node loss in N. geniculatus part of a more integrated ac- celeration of ontogenetic development? Using geometric data, Webster et al. (2001) demon- strated that the pattern of ontogenetic shape change for the entire cephalon had been evo- lutionarily modified from at least as early as FIGURE 4. Location of the anteriormost axial node on phase 3 of development and (depending on the glabella during the ontogeny of N. multinodus (cir- choice of analytical technique) perhaps as ear- cles, upper regression line) and N. geniculatus (squares, lower regression line). For each specimen, all glabellar ly as phase 1. Clearly the increased rate of lobes posterior to the one indicated bore an axial node. node loss relative to size in N. geniculatus was Note the progressive loss of nodes during development not part of a global (cephalic) acceleration of of each species. See text for details and interpretation. morphological development accounting for all interspecific differences, but is there evidence documentation of the ontogenies of these taxa for integration on a more local scale? This pos- is provided in a forthcoming paper). However, sibility is now investigated by studying the a detailed investigation of the pattern of on- development of the glabella in each species. togenetic shape change for the cephalon of First, let us examine the rate of glabellar each species found that the ancestral trajectory shape change relative to size. Size is quanti- of ontogenetic shape change was not con- fied as the log centroid size of the oculo-gla- served in the descendant. Webster et al. (2001) bellar landmark configuration (Fig. 3B) of therefore argued that the evolution of N. gen- each specimen. Glabellar development has iculatus from N. multinodus was better de- sometimes been quantified by linear measures scribed in terms of ‘‘spatial repatterning,’’ and (e.g., Edgecombe and Chatterton 1987). Rate that ‘‘any heterochronic shifts were localized of glabellar shape change quantified as gla- and of minor importance.’’ Some details of the bella length relative to log centroid size development of the oculo-glabellar region showed no statistically significant differences (Fig. 3) of Nephrolenellus are now examined in between N. multinodus and N. geniculatus (Fig. light of the evolutionary modes discussed 5A; ANCOVA: p ϭ 0.64). However, when gla- above. bellar shape change is quantified as Procrus- An important feature distinguishing N. gen- tes distance away from the average juvenile iculatus from N. multinodus was the number of form, a significant interspecific difference in axial nodes on the glabella through ontogeny. rate of glabellar shape change relative to size Nephrolenellus multinodus initially bore an axial is evident (Fig. 5B; ANCOVA: p Ͻ 0.001). (In node on each lobe of the glabella and pro- this case the average juvenile was specified as gressively lost nodes (in an anteroposterior the consensus of all early phase 3 individuals, direction) during ontogeny (Figs. 3, 4). In ma- although results are robust to reference turity, nodes were retained on lobes LO, L1, choice; data not presented.) Again, modifica- L2, and rarely L3. Nephrolenellus geniculatus tion of the ancestral ontogeny (in terms of rate initially bore axial nodes on all glabellar lobes of development of the glabella relative to size) except the anteriormost (LA) and progressive- is described as a rate modification. However, ly lost nodes in the same fashion, retaining a although the rate of node loss relative to size node only on LO (occasionally also L1, ex- was increased, the rate of glabellar develop- tremely rarely on L2) in maturity (Figs. 3, 4). ment relative to size was decreased between Analysis of covariance (ANCOVA) demon- these taxa. strates that the rate of node loss (relative to log Assessment of the pattern of glabellar shape centroid size) was significantly different be- change in each taxon further complicates the tween the species (p Ͻ 0.001). This difference situation. Vectors of glabellar shape change MODIFICATIONS OF ONTOGENY 369

mirroring results obtained from study of the entire cephalon (Webster et al. 2001). During phase 4 of development high within-species variance rendered a between-species angle of 44Њ statistically insignificant, suggesting con- servation of the ancestral ontogenetic trajec- tory of glabellar shape change during this phase. (This conservation was local to the gla- bella, because the pattern of shape change of entire cephalon differed significantly between the species during this phase [Webster et al. 2001].) The description of the evolution of N. geni- culatus from N. multinodus can now be en- hanced by considering the modes of ontoge- netic modification defined above. The evolu- tionary increase in rate of axial node loss rel- ative to size throughout ontogeny was a rate modification. During phases 3 and 4 of devel- opment the rate of glabellar development rel- ative to size was evolutionarily decreased— also a rate modification but in the opposite di- rection to that describing node loss. The an- cestral pattern of glabellar ontogenetic shape change was modified through allometric re- patterning during phase 3 in N. geniculatus. However, upon entry into phase 4 N. genicu- latus apparently returned to the ancestral pat- tern of glabellar shape change (although the morphological parallelism between ancestor FIGURE 5. Rate of shape change of the oculo-glabellar and descendant was already lost during phase structure as a function of size during phases 3 and 4 of development in N. multinodus (circles) and N. geniculatus 3). It is difficult to determine the extent to (squares). A, Using glabella length to summarize shape which these localized modifications of glabel- of the oculo-glabellar structure. B, Using Procrustes dis- lar development may have been responsible tance (the square root of the summed squared distances between homologous landmarks on two configurations for the overall interspecific differences in ce- following Procrustes superimposition [Bookstein 1991]) phalic development (Webster et al. 2001), al- to summarize shape of the oculo-glabellar structure. A though vectors of landmark movement sug- regression line has been fitted for each species. See text for details and interpretation. gest interspecific differences in nonglabellar features (Fig. 6). One additional noteworthy feature was the for each species were calculated from the absence of an axial node on the anteriormost warp scores following decomposition of a glabellar lobe (LA) in the earliest preserved thin-plate spline analysis (methods outlined developmental stages of N. geniculatus. Given in Webster et al. 2001; software written by H. the presence of the node in the earliest onto- D. Sheets available at http://www.canisius. genetic stages of N. multinodus, this represent- edu/ϳsheets/morphsoft.html; data available ed a case of deletive heterometry (above). from senior author upon request). During However, it has been demonstrated that the phase 3 of development the species showed rate of ontogenetic loss of axial nodes relative significantly different patterns of glabellar to size was higher in N. geniculatus. Absence of shape change (between-species angle of 64Њ, the node on LA in N. geniculatus may conceiv- significantly different to 95% confidence), ably have resulted from this increased rate of 370 MARK WEBSTER AND MIRIAM L. ZELDITCH

provides an unavoidable, preservation-related lower limit to knowledge of their develop- ment. This highlights the fact that identifica- tion of the modes by which morphological evolution occurred is dependent upon the completeness of ontogenetic coverage. Recognition and accurate characterization of such a suite of modifications to ontogeny is a necessary step toward linking patterns of morphological evolution to changes at more proximal developmental levels. Much of the complexity uncovered in this deceptively sim- ple case would have been unrecognized or as- signed to peramorphic heterochrony under previous conceptual frameworks. We consider it likely that such complexity in developmen- tal reprogramming will be found to be typical of morphological evolution, and that case studies fully explained by a single mode of modification will be rare.

Conclusions Consideration of all the distinct modes of evolutionary change in development allows identification of modified and conserved as- pects of ontogeny. As a result, understanding of the patterns and constraints in evolution is enhanced. In particular, we can recognize far more than heterochrony, which is sometimes regarded as the dominant mode of evolution- FIGURE 6. Vectors of cephalic landmark movement dur- ing phase 3 (A) and phase 4 (B) of development of N. ary change in development (e.g., McKinney multinodus (dashed arrows) and N. geniculatus (solid ar- and McNamara 1991; McNamara 1997; Reilly rows). For simplicity, only the landmarks on the right et al. 1997). Recognizing modes of modifica- half of the cephalon are shown (see Fig. 3B). Note that non-glabellar landmarks show species-specific vectors tion to morphological development offers a of movement apparently independent of glabellar land- scheme applicable to specimen-based studies, marks. which includes the vast majority of paleonto- logical studies (as well as the majority of neon- tological studies). node loss. If this were true, it would be pre- We recognize six modes of developmental dicted that developmentally younger individ- modification. Rate modification is a modifica- uals of N. geniculatus (not covered by the sam- tion of the amount of morphological change ple) would have possessed such a node. Dis- achieved over a specified interval. Timing mod- covery of smaller specimens of N. geniculatus ification is a modification of the timing of de- therefore has the potential to change the in- velopmental events and can be independent of terpretation from one of heterometry into one developmental rate. Rate or timing modifica- of rate modification. However, this is unlikely tion to an otherwise conserved ontogenetic to happen as the smallest individuals exam- trajectory of shape change can be further qual- ined are believed to be in the earliest ontoge- ified as rate or event heterochrony, respective- netic stage at which a mineralized exoskeleton ly. Heterotopy is a change in location of a par- was developed. The time at which trilobites ticular morphological structure. Heterotypy is initiate mineralization of their exoskeleton the replacement of a morphological structure MODIFICATIONS OF ONTOGENY 371 of one type by one of a different type at a given script. L. Knox (now L. Webster) helped M.W. location. Heterometry is a change in the num- retain (some) sanity during the effort to make ber of a particular morphological feature (not sense of the heterochrony literature. M.W. was attained through replacement or through funded by National Science Foundation grant truncation/extrapolation of the ancestral tra- EAR-9980372. jectory). Allometric repatterning is a modifica- tion of the trajectory of ontogenetic shape Literature Cited change. Our systematic treatment of these Abzhanov, A., and T. C. Kaufman. 1999. Novel regulation of the modes of evolutionary change in development homeotic gene Scr associated with a crustacean leg-to-max- uses clearly defined and consistent evidentia- illiped appendage transformation. Development 126:1121– 1128. ry criteria, which should help to reduce the Akam, M., M. Averof, J. Castelli-Gair, R. Dawes, F. Falciani, and confusion in the literature. Each of the modes D. Ferrier. 1994. The evolving role of Hox genes in arthropods. of evolution makes unique and explicit pre- Development 1994(Suppl.):209–215. Alberch, P. 1985. Problems with the interpretation of develop- dictions regarding the type of morphological mental sequences. Systematic Zoology 34:46–58. change it engenders, and hypotheses of their Alberch, P., and M. J. Blanco. 1996. Evolutionary patterns in on- operation can therefore be rigorously tested. togenetic transformation: from laws to regularities. Interna- tional Journal of Developmental Biology 40:845–858. Nevertheless, the degree of morphological Alberch, P., S. J. Gould, G. F. Oster, and D. B. Wake. 1979. Size and ontogenetic coverage, as well as the level and shape in ontogeny and phylogeny. Paleobiology 5:296– of detail in which morphology is investigated, 317. Arthur, W. 2000. The concept of developmental reprogramming can affect the type of evolutionary change de- and the quest for an inclusive theory of evolutionary mecha- tected. Caution is therefore required when nisms. Evolution and Development 2:49–57. making generalizations regarding the relative Atchley, W. R. 1987. Developmental quantitative genetics and the evolution of ontogenies. Evolution 41:316–330. contribution of each mode to evolution. Bookstein, F. L. 1991. Morphometric tools for landmark data: ge- Although discrete and independent in ometry and biology. Cambridge University Press, New York. terms of their effects on morphological devel- Brochu, C. A. 1995. Heterochrony in the crocodylian scapulo- coracoid. Journal of Herpetology 29:464–468. opment, the modes of evolutionary change are Carroll, S. B. 1994. Developmental regulatory mechanisms in not necessarily independent in terms of types the evolution of insect diversity. Development 1994(Suppl.): of change to cellular and subcellular process- 217–223. Edgecombe, G. D., and B. D. E. Chatterton. 1987. Heterochrony es. We suggest that evolutionary changes in the Silurian radiation of encrinurine trilobites. Lethaia 20: based on such non-morphological criteria will 337–351. require a different terminology in order to Fink, W. L. 1982. The conceptual relationship between ontogeny and phylogeny. Paleobiology 8:254–264. avoid unwarranted inference of mechanistic ———. 1988. Phylogenetic analysis and the detection of onto- cause and morphological effect (see also Al- genetic patterns. Pp. 71–91 in M. L. McKinney, ed. Heteroch- berch and Blanco 1996). Utilization of the rony in evolution: a multidisciplinary approach. Plenum, New York . scheme presented here will allow a thorough Gellon, G., and W. McGinnis. 1998. Shaping animal body plans and meaningful investigation of the modes of in development and evolution by modulation of Hox expres- morphological evolution. sion patterns. BioEssays 20:116–125. Godfrey, L. R., and M. R. Sutherland. 1995. What’s growth got to do with it? Process and product in the evolution of ontog- Acknowledgments eny. Journal of Human Evolution 29:405–431. Our thanks extend to the ‘‘HeteroCronies’’ ———. 1996. Paradox of peramorphic paedomorphosis: heter- ochrony and human evolution. American Journal of Physical (P. Fitzgerald, R. Guralnick, N. C. Hughes, G. Anthropology 99:17–42. Jaecks, P. Kaplan, D. Lindberg, R. Mooi, R. Gould, S. J. 1977. Ontogeny and phylogeny. Harvard University Nehm, P. Roopnarine, and B. Waggoner), who Press, Cambridge. ———. 2000. Of coiled oysters and big brains: how to rescue the participated in discussions following the De- terminology of heterochrony, now gone astray. Evolution and veloping Paleontology topical session at the Development 2:241–248. Geological Society of America Annual Meet- Haeckel, E. 1875. Die Gastrula und die Eifurchung der Thiere. Jenaische Zeitschrift fu¨ r Naturwissenschaft 9:402–508. ing in Reno, 2000. Morphometrics software Halder, G., P. Callaerts, and W. J. Gehring. 1995. Induction of was written by H. D. Sheets. Comments by N. ectopic eyes by targeted expression of the eyeless gene in Dro- C. Hughes, P. M. Sadler, and M. Foote and re- sophila. Science 267:1788–1792. Hughes, N. C., and R. E. Chapman. 1995. Growth and variation views by S. Carlson, D. Wake, and an anony- in the Silurian proetide trilobite Aulacopleura konincki and its mous reviewer helped improve the manu- implications for trilobite palaeobiology. Lethaia 28:333–353. 372 MARK WEBSTER AND MIRIAM L. ZELDITCH

Hughes, N. C., R. E. Chapman, and J. M. Adrain. 1999. The sta- specific and intraspecific phenomena. Biological Journal of the bility of thoracic segmentation in trilobites: a case study in Linnean Society 60:119–143. developmental and ecological constraints. Evolution and De- Rice, S. H. 1997. The analysis of ontogenetic trajectories: when velopment 1:24–35. a change in size or shape is not heterochrony. Proceedings of Jaecks, G. S., and S. J. Carlson. 2001. How phylogenetic inference the National Academy of Sciences USA 94:907–912. can shape our view of heterochrony: examples from thecidei- Roopnarine, P. D. 2001. Testing the hypothesis of heterochrony de brachiopods. Paleobiology 27:205–225. in morphometric data: lessons from a bivalved mollusk. Pp. Kettle, C., W. Arthur, T. Jowett, and A. Minelli. 1999. Homeotic 271–303 in Zelditch 2001. transformation in a centipede. Trends in Genetics 15:393. Sattler, R. 1992. Process morphology: structural dynamics in de- Kjaer, C. R., and E. Thomsen. 1999. Heterochrony in bourgue- velopment and evolution. Canadian Journal of Botany 70:708– ticrinid sea-lilies at the Cretaceous/Tertiary boundary. Paleo- 714. biology 25:29–40. Shea, B. T. 1985. Bivariate and multivariate growth allometry: Klingenberg, C. P. 1998. Heterochrony and allometry: the anal- statistical and biological considerations. Journal of Zoology ysis of evolutionary change in ontogeny. Biological Reviews 206:367–390. 73:79–123. Smith, K. K. 1997. Comparative patterns of craniofacial devel- Lewis, E. B. 1978. A gene complex controlling segmentation in opment in eutherian and metatherian mammals. Evolution Drosophila. Nature 276:565–570. 51:1663–1678. ———. 2001. Heterochrony revisited: the evolution of develop- Li, P., and M. O. Johnston. 2000. Heterochrony in plant evolu- mental sequences. Biological Journal of the Linnean Society tionary studies through the twentieth century. Botanical Re- 73:169–186. view 66:57–88. Stephen, D. A., W. L. Manger, and C. Baker. 2002. Ontogeny and Lovejoy, C. O., M. J. Cohn, and T. D. White. 1999. Morphological heterochrony in the Middle Carboniferous ammonoid Arkan- analysis of the mammalian postcranium: a developmental ites relictus (Quinn, McCaleb, and Webb) from northern Ar- perspective. Proceedings of the National Academy of Sciences kansas. Journal of Paleontology 76:810–821. USA 96:13247–13252. Sundberg, F. A. 2000. Homeotic evolution in Cambrian trilo- McKinney, M. L. 1999. Heterochrony: beyond words. Paleobi- bites. Paleobiology 26:258–270. ology 25:149–153. Takhtajan, A. 1972. Patterns of ontogenetic alterations in the McKinney, M. L., and K. J. McNamara. 1991. Heterochrony: the evolution of higher plants. Phytomorphology 22:164–171. evolution of ontogeny. Plenum, New York. Wagner, G. P., C.-H. Chiu, and M. Laubichler. 2000. Develop- McNamara, K. J. 1978. Paedomorphosis in Scottish olenellid tri- mental evolution as a mechanistic science: the inference from lobites. Palaeontology 21:635–655. developmental mechanisms to evolutionary processes. Amer- ———. 1981. The role of paedomorphosis in the evolution of ican Zoologist 40:819–831. Cambrian trilobites. U.S. Geological Survey Open-File Report Wake, D. B. 1996. Evolutionary developmental biology—pros- 81;c- 743:126–129. pects for an evolutionary synthesis at the developmental level. ———. 1983. Progenesis in trilobites. Special Papers in Palaeon- Pp. 97–107 in M. T. Ghiselin and G. Pinna, eds. New perspec- tology 30:59–68. tives on the history of life. California Academy of Sciences, ———. 1986a. A guide to the nomenclature of heterochrony. San Francisco. Journal of Paleontology 60:4–13. Wake, D. B., and G. Roth. 1989. The linkage between ontogeny ———. 1986b. The role of heterochrony in the evolution of Cam- and phylogeny in the evolution of complex systems. Pp. 361– brian trilobites. Biological Reviews 6:121–156. 377 in D. B. Wake and G. Roth, eds. Complex organismal func- ———. 1988. The abundance of heterochrony in the fossil re- tions: integration and evolution in vertebrates. Wiley, Chich- cord. Pp. 287–325 in Heterochrony in evolution: a multidisci- ester, U.K. plinary approach. M. L. McKinney, ed. Plenum, New York. Wayne, R. W. 1986. Cranial morphology of domestic and wild ———. 1995. Evolutionary change and heterochrony. Wiley, canids: the influence of development on morphological Chichester, U.K. change. Evolution 40:243–261. ———. 1997. Shapes of time: the evolution of growth and de- Webster, M., H. D. Sheets, and N. C. Hughes. 2001. Allometric velopment. Johns Hopkins University Press, Baltimore. patterning in trilobite ontogeny: testing for heterochrony in Nedin, C., and R. J. F. Jenkins. 1999. Heterochrony in the Cam- Nephrolenellus. Pp. 105–144 in Zelditch 2001. brian trilobite Hsuaspis. Alcheringa 23:1–7. Wray, G. A., and D. R. McClay. 1989. Molecular heterochronies Nehm, R. H. 2001. The developmental basis of morphological and heterotopies in early echinoid development. Evolution 43: disarmament in Prunum (Neogastropoda: Marginellidae). Pp. 803–813. 1–26 in Zelditch 2001. Zelditch, M. L., ed. 2001. Beyond heterochrony: the evolution of O’Keefe, F. R., O. Rieppel, and P. M. Sander. 1999. Shape disso- development. Wiley, New York. Zelditch, M. L., and W. L. Fink. 1996. Heterochrony and heter- ciation and inferred heterochrony in a clade of pachypleuro- otopy: stability and innovation in the evolution of form. Pa- saurs (Reptilia, Sauropterygia). Paleobiology 25:504–517. leobiology 22:241–254. Raff, R. A. 1996. The shape of life: genes, development, and the Zelditch, M. L., H. D. Sheets, and W. L. Fink. 2000. Spatiotem- evolution of animal form. University of Chicago Press, Chi- poral reorganization of growth rates in the evolution of on- cago. togeny. Evolution 54:1363–1371. Raff, R. A., and G. A. Wray. 1989. Heterochrony: developmental ———. 2003a. The ontogenetic dynamics of shape disparity. Pa- mechanisms and evolutionary results. Journal of Evolutionary leobiology 29:139–156. Biology 2:409–434. Zelditch, M. L., B. L. Lundigran, H. D. Sheets, and T. Garland Regier, J. C., and N. S. Vlahos. 1988. Heterochrony and the in- Jr. 2003b. Do precocial mammals develop at a faster rate? A troduction of novel modes of morphogenesis during the evo- comparison of rates of skull development in Sigmodon fulvi- lution of moth choriogenesis. Journal of Molecular Evolution venter and Mus musculus domesticus. Journal of Evolutionary 28:19–31. Biology 16:708–720. Reilly, S. M., E. O. Wiley, and D. J. Meinhardt. 1997. An integra- Zimmermann, W. 1959. Die Phylogenie der Pflanzen, 2d ed. tive approach to heterochrony: the distinction between inter- Gustav Fischer, Jena.