what is parallism

14
What is parallelism? Robert W. Scotland Department of Plant Sciences, South Parks Road, University of Oxford, Oxford, OX1 3RB, UK Author for correspondence (email: [email protected]) SUMMARY Although parallel and convergent evolution are discussed extensively in technical articles and textbooks, their meaning can be overlapping, imprecise, and contradictory. The meaning of parallel evolution in much of the evolutionary literature grapples with two separate hypotheses in relation to phenotype and genotype, but often these two hypotheses have been inferred from only one hypothesis, and a number of subsidiary but problematic criteria, in relation to the phenotype. However, examples of parallel evolution of genetic traits that underpin or are at least associated with convergent phenotypes are now emerging. Four criteria for distinguishing parallelism from convergence are reviewed. All are found to be incompatible with any single proposition of homoplasy. Therefore, all homoplasy is equivalent to a broad view of convergence. Based on this concept, all phenotypic homoplasy can be described as convergence and all genotypic homoplasy as parallelism, which can be viewed as the equivalent concept of convergence for molecular data. Parallel changes of molecular traits may or may not be associated with convergent phenotypes but if so describe homoplasy at two biological levelsFgenotype and phenotype. Parallelism is not an alternative to convergence, but rather it entails homoplastic genetics that can be associated with and potentially explain, at the molecular level, how convergent phenotypes evolve. INTRODUCTION A perennial issue of unresolved discussion in biology is the distinction between parallel and convergent evolution (Scott 1891, 1896; Hennig 1966; Reidl 1979; Saether 1979; Patter- son 1982, 1988; Rieppel 1988; Donoghue 1992; Sanderson and Hufford 1996; Abouheif 1997, 1999; Abouheif et al. 1997; Wray and Abouheif 1998; Wichman et al. 1999; Abouheif and Wray 2002; Gould 2002; Cooper et al. 2003; Sucena et al. 2003; Cracraft 2005; Christin et al. 2007; Hall 2007; Arendt and Reznick 2008a,b; Rokas and Carroll 2008; Scotland 2010). In a recent article Arendt and Reznick (2008a) claim that parallelism is a superfluous term and that convergence suffices to describe all occurrences of the inde- pendent evolution of the same phenotype. Ironically, this reformulation comes at a time, when the term parallelism is enjoying something of a resurgence in interest among evo- lutionary biologists (Zhang and Kumar 1997; Gould 2002; Cooper et al. 2003; Sucena et al. 2003; Colosimo et al. 2005; Fong et al. 2005; Harrison et al. 2005; Derome et al. 2006; Roberge et al. 2006; Christin et al. 2007; Rokas and Carroll 2008; Liu et al. 2010). Gould’s (2002, p. 1089) excitement epitomized in a triumphalist tone usually shunned in science, but clearly justified in this rare case stated that Parallelism has now, and finally after a century of terminological recog- nition, become an operational subject for evolutionary research. Arendt and Reznick (2008a) distinguish parallelism from convergence on the basis of whether taxa that share indepen- dently acquired phenotypic traits are closely or distantly re- lated. This contrasts with the views of others, that distinguish parallelism from convergence on the basis of whether or not the same genetic mechanisms are involved (e.g., Haas and Simpson 1946; Gould 2002; Hall 2007) whether or not they are the same trait (e.g., Patterson 1982; Patterson 1988) or whether they have the same or different ancestral character states (e.g., Hennig 1966; Rokas and Carroll 2008). One reason this issue is of significance is that under some (but not all) formulations, parallelism is a term that attempts to link the phenotype with the genotype; a link that represents a central theme in contemporary developmental and evolu- tionary biology (Scotland 2010 and references therein). The uncertainty that surrounds parallelism and convergence often results in the same evolutionary phenomena being described as belonging to different categories. For example, two studies entitled C 4 photosynthesis evolved in grasses via parallel adaptive genetic changes (Christin et al. 2007) and Conver- gent sequence evolution between echolocating bats and dol- phins (Liu et al. 2010) interpret what are very similar results, albeit for very different taxa, as parallel evolution of genetic EVOLUTION & DEVELOPMENT 13:2, 214–227 (2011) DOI: 10.1111/j.1525-142X.2011.00471.x & 2011 Wiley Periodicals, Inc. 214

Upload: bishoy-hanna

Post on 06-Mar-2015

68 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: What is Parallism

What is parallelism?

Robert W. Scotland

Department of Plant Sciences, South Parks Road, University of Oxford, Oxford, OX1 3RB, UK�Author for correspondence (email: [email protected])

SUMMARY Although parallel and convergent evolution arediscussed extensively in technical articles and textbooks, theirmeaning can be overlapping, imprecise, and contradictory.The meaning of parallel evolution in much of the evolutionaryliterature grapples with two separate hypotheses in relation tophenotype and genotype, but often these two hypotheseshave been inferred from only one hypothesis, and a number ofsubsidiary but problematic criteria, in relation to thephenotype. However, examples of parallel evolution ofgenetic traits that underpin or are at least associated withconvergent phenotypes are now emerging. Four criteria fordistinguishing parallelism from convergence are reviewed. Allare found to be incompatible with any single proposition of

homoplasy. Therefore, all homoplasy is equivalent to a broadview of convergence. Based on this concept, all phenotypichomoplasy can be described as convergence and allgenotypic homoplasy as parallelism, which can be viewedas the equivalent concept of convergence for molecular data.Parallel changes of molecular traits may or may not beassociated with convergent phenotypes but if so describehomoplasy at two biological levelsFgenotype and phenotype.Parallelism is not an alternative to convergence, but rather itentails homoplastic genetics that can be associated with andpotentially explain, at the molecular level, how convergentphenotypes evolve.

INTRODUCTION

A perennial issue of unresolved discussion in biology is the

distinction between parallel and convergent evolution (Scott

1891, 1896; Hennig 1966; Reidl 1979; Saether 1979; Patter-

son 1982, 1988; Rieppel 1988; Donoghue 1992; Sanderson

and Hufford 1996; Abouheif 1997, 1999; Abouheif et al.

1997; Wray and Abouheif 1998; Wichman et al. 1999;

Abouheif and Wray 2002; Gould 2002; Cooper et al. 2003;

Sucena et al. 2003; Cracraft 2005; Christin et al. 2007; Hall

2007; Arendt and Reznick 2008a, b; Rokas and Carroll

2008; Scotland 2010). In a recent article Arendt and Reznick

(2008a) claim that parallelism is a superfluous term and that

convergence suffices to describe all occurrences of the inde-

pendent evolution of the same phenotype. Ironically, this

reformulation comes at a time, when the term parallelism is

enjoying something of a resurgence in interest among evo-

lutionary biologists (Zhang and Kumar 1997; Gould 2002;

Cooper et al. 2003; Sucena et al. 2003; Colosimo et al. 2005;

Fong et al. 2005; Harrison et al. 2005; Derome et al. 2006;

Roberge et al. 2006; Christin et al. 2007; Rokas and Carroll

2008; Liu et al. 2010). Gould’s (2002, p. 1089) excitement

epitomized in a triumphalist tone usually shunned in science,

but clearly justified in this rare case stated that Parallelism

has now, and finally after a century of terminological recog-

nition, become an operational subject for evolutionary

research.

Arendt and Reznick (2008a) distinguish parallelism from

convergence on the basis of whether taxa that share indepen-

dently acquired phenotypic traits are closely or distantly re-

lated. This contrasts with the views of others, that distinguish

parallelism from convergence on the basis of whether or not

the same genetic mechanisms are involved (e.g., Haas and

Simpson 1946; Gould 2002; Hall 2007) whether or not they

are the same trait (e.g., Patterson 1982; Patterson 1988) or

whether they have the same or different ancestral character

states (e.g., Hennig 1966; Rokas and Carroll 2008).

One reason this issue is of significance is that under some

(but not all) formulations, parallelism is a term that attempts

to link the phenotype with the genotype; a link that represents

a central theme in contemporary developmental and evolu-

tionary biology (Scotland 2010 and references therein). The

uncertainty that surrounds parallelism and convergence often

results in the same evolutionary phenomena being described

as belonging to different categories. For example, two studies

entitled C4 photosynthesis evolved in grasses via parallel

adaptive genetic changes (Christin et al. 2007) and Conver-

gent sequence evolution between echolocating bats and dol-

phins (Liu et al. 2010) interpret what are very similar results,

albeit for very different taxa, as parallel evolution of genetic

EVOLUTION & DEVELOPMENT 13:2, 214 –227 (2011)

DOI: 10.1111/j.1525-142X.2011.00471.x

& 2011 Wiley Periodicals, Inc.214

Page 2: What is Parallism

changes in the case of C4 grasses, but convergent evolution of

genetic changes in echolocating bats and dolphins. Further-

more, some view convergence at the molecular level as not

being possible (Patterson 1988) others regard it as very rare

(Castoe et al. 2009; Liu et al. 2010) others as occurring in-

frequently but differing from parallelism which is regarded as

widespread (Zhang and Kumar 1997; Rokas and Carroll

2008), others that some level of convergence is common if not

universal (Sanderson and Donoghue 1996; Bull et al. 1997)

and by still others as being prevalent at the functional, mech-

anistic and structural levels, but maybe not the gene sequence

level (Doolittle 1994). As a result, this issue constitutes much

more than a semantic debate about appropriate terminology.

Rather, it lies at the core of contemporary evolutionary

biology and in particular, turns on how the patterns produced

by the evolutionary process are described, characterized, and

understood.

This article focuses on several aspects of this issue from the

perspective of homology, to ask whether any pertinent lessons

can be gained from a consideration of convergence and par-

allelism from the perspective of that concept. Homoplasy and

homology are terms that travel together (Wake 1996, p. xvii)

because homoplasy is homology at a more restricted hierar-

chical level (Hall 2007).

HOMOLOGY

Homology as an equivalence relation remains a term in gen-

eral use in contemporary comparative biology (Scotland 2010

and references therein). Core aspects of homology are the

conditional phrase (what is being compared) and the hierar-

chical level at which the comparison applies (Fig. 1). Bock

(1974) suggested that any statement about the homology of

features in different organisms must include a conditional

phrase describing the nature of the relationship. The condi-

tional phrase is partly determined by the variation of the parts

being described and this, in part, is determined by the hier-

archical level of the comparison (Fig. 1). The relation of ho-

mology is described as such because homology is a relational

concept and because the degree of relationship varies, one

must always state the nature or condition of a particular set of

homologues (Bock 1974, p. 387).

Although the conditional phrase and the hierarchical level

can be considered as two separate aspects of homology de-

termination, they are so interlinked as to encompass one idea

or hypothesis. The alternative is to consider the conditional

phrase as being independent and separate of the hierarchical

level (either pre-Darwinian classifications or explicit phyloge-

nies). Such considerations are difficult to imagine from a sys-

tematic perspective, because character concepts for example

carpels, nucleic acid, seeds, legumes, vertebrae, fins, amnion,

etc., have been developed, refined, and conceptualized, hand

in-hand, alongside the context of hierarchy and classification

and thus a specific hierarchical level (e.g., carpels of angio-

sperms, nucleic acids of life, seeds of spermatophytes, legumes

of leguminosae, vertebrae of vertebrates, fins of fishes, amnion

of amniotes) and are therefore closely linked to the construc-

tion of predictive classifications and/or, more recently, explicit

phylogenies.

In contemporary biology the taxic view of homology is

rightly explained in terms of common ancestry and mono-

phyly. Nevertheless the development and reciprocal illumina-

tion of character concepts alongside hierarchical groupings

(taxa) has a creative, dynamic, and productive pre-evolution-

ary history. The taxic view of homology places great emphasis

on the meaning, description, and interpretation of anatomy as

interdependent with the hierarchical level. Other frameworks

that emphasize comparative anatomy and development inde-

pendent of hierarchy (Wagner 1989a, b) or alternatively pro-

mote hierarchies built upon operational character concepts

(Sneath and Sokal 1973), have not been successful because it

is the relationship between the twoFconditional phrase and

hierarchyFthat provides insight and meaning. For example,

Fig. 1 compares two fruits (one open and one closed) of

Afzelia africana, in comparison with: (a) each other (b) fruits

from the same family and (3) fruits from several families of

flowering plants. Although the fruit of A. africana, remains

the same, the conditional phrase and abstracted homology is

determined by the context of the comparisons that is the hi-

erarchical level. Thus, the fruit can be categorized as: a type of

legume relative to other individuals of the same species

(A. africana); a fruit (legume) of the family Fabaceae; and a

carpel of angiosperms. In this way hypotheses of homology

comprise both an abstract conditional phrase and an explicit

hierarchical level, for recent discussion see (Scotland 2010).

HOMOPLASY

Lankester (1870) is credited with introducing the term ho-

moplasy as distinct from homology (which Lankester termed

Homogeny) and provided the following example concerning

the possible origin of the forms of weapons and utensils of

various races of men to illustrate the distinction between

homology and homoplasy. Lankester (1870, p. 41) wrote:

Two races, A and B, without communication, may devise a

stone axe or a canoe of similar forms: the resemblance is in this

case homoplastic. Lankester’s example includes two separate

hierarchical levels (races) independently acquiring similar

traits (axe and canoe). The homoplasy comprises part com-

parison (canoe and axe) and part hierarchical levels (races A

and B). Homoplasies are therefore hypotheses of a corre-

spondent trait that are distributed minimally at two separate

hierarchical levels.

What is parallelism? 215Scotland

Page 3: What is Parallism

All homoplasy creates the same pattern, resultingin the independent occurrence of the same featurerelative to phylogenyIn the phylogenetic specialist literature and textbooks, ho-

moplasy is often described as having a number of possible

explanations including convergence, parallelism and reversal

(Hennig 1966; Page and Holmes 1998; Hall 2007, 2008).

However, reversal of a character state causes the independent

occurrence of the same character state in different places on

the cladogram. As a consequence, reversals can and often are

viewed (Conway Morris 2003, 2010; Sucena et al. 2003) as a

subset of either parallelism or convergence. For example, re-

versal to the plesiomorphic fusiform body shape in whales

and dolphins compared with fishes is a reversal that results in

convergence (Conway Morris 2003) and independent loss of

trichomes in different species of Drosophila are described as

convergence (Sucena et al. 2003). Figure 2 illustrates this rel-

ative to an unrooted tree of seven taxa, three of which have

the character state gray and four of which have an alternative

character state black. The un-rooted tree shows that gray is a

In the comparative context of the objects on the right, it is a particular type of legume of Afzelia africana

what is this?

In the comparative context of the objects on the right, it is a legume of the family Leguminosae

In the comparative context of the objects on the right and below, it is a fruit of an angiosperm.

what is this?

Conditional phrase Hierarchical level

Legume type Afzelia africana

what is this?

Conditional phrase Hierarchical level

Legume Leguminosae

Conditional phrase Hierarchical levelFruit Angiosperm

A

B

C

The Relation of Homology

Fig. 1. With reference to the specimentop left, or any biological specimen, wecan ask, what is it? The answer (condi-tional phrase) depends on the hierarchicallevel of the comparison. (A) In compar-ison to a similar specimen from the samespecies, it is a particular type of legume ofAfzelia africana. (B) In comparison toother fruits from the same family (Leg-uminosae/Fabaceae) it is a legume. (C)In comparison to other types of fruit, itis a fruit of angiosperms. Therefore theidentity and meaning of a biological ob-ject is determined by the hierarchical levelof the comparison which in turn deter-mines the conditional phrase that is nec-essary to describe the object. Drawings ofspecimens from the Oxford UniversityHerbarium (FHO) by Rosemary Wise.Fruits are from Afzelia africana, Glirici-dium sepium, Enterolobium cyclocarpum,Castanopsis tibetana, Entandrophragmaexcelsum, Proboscidea althaeifolia, Jaca-randa mimosifolia.

216 EVOLUTION & DEVELOPMENT Vol. 13, No. 2, March--April 2011

Page 4: What is Parallism

homoplasy. Rooting positions 1 and 2 both result in the in-

dependent evolution of gray relative to the phylogeny. The

scenarios of character evolution leading to these convergences

are different, but nevertheless both share the same pat-

ternFcorrespondent character state distributed at more than

one hierarchical level. Therefore, all homoplasy creates the

same general pattern; the separate occurrence/acquisition of a

correspondent character state at more than one location on a

phylogeny. In this sense, all homoplasy whether it be pheno-

typic or genotypic is equivalent: the recurrent tendency of

biological organization to arrive at the same ‘‘solution,’’

(Conway Morris 2003, p. xii).

Four criteria for distinguishing parallelism fromconvergence

Homoplasy is considered by most biologists to include ex-

amples of both parallelism and convergence, but these terms

are interpreted and used in a number of ways. For some

authors for example Arendt and Reznick (2008a) parallelism

and convergence are indistinguishable and the distinction be-

tween them is unnecessary and irrelevant. For others includ-

ing Arendt and Reznick (2008a) and Scott (1891, p. 363) the

distinction between the two classes of phenomena is obviously

one of degree rather than kind and for Simpson (1945, p. 9)

the two phenomena intergrade continuously and are often

indistinguishable in practice. Still other authors have distin-

guished between them on the basis of: (1) homoplastic char-

acter states that either correspond structurally (parallelism) or

do not (convergence), (see Patterson 1982, 1988); (2) taxa

sharing independent traits are either closely related (parallel-

ism) or not (convergence), (see Scott 1896; Arendt and Rez-

nick 2008a, b; Leander 2008), (3) Independent traits that share

the same ancestral character states (parallelism) or not (con-

vergence), (see Hennig 1966; Reidl 1979; Leander 2008); and

(4) independent traits caused by the same underlying genetics

or an ancestral predisposition (parallelism) or not (conver-

gence), (see Haas and Simpson 1946; Brundin 1976; Gould

2002). These four criteria are listed in Table 1.

The first criterionFwhether the characters are really sim-

ilar (structurally correspondent) or notF, refers, in part, to

the distinction between analogy used in this article in a broad

sense to distinguish all nonstructurally correspondent com-

parisons and all other relations including homology and ho-

moplasy (Fig. 3). In much of the literature, both parallelism

and convergence are regarded as forms of homoplasy in the

sense of Lankester (1870) and consist of structurally corre-

spondent traits distributed between at least two hierarchical

levels (Fig. 3, Table 2). If convergence is restricted to inde-

pendent traits that are noncorrespondent then this would be

at odds with many examples and much of the widespread and

general usage of this term in evolutionary biology.

The second criterionFclose or remote relationships of the

taxa bearing the charactersFis not relevant in the overall

context of a comparative system that emphasizes the condi-

tional phrase and hierarchical levels. This is because an ex-

plicit hierarchical level is much more precise and lacking in

ambiguity compared with close or remote relationships of the

taxa and the inevitable and endless discussion about just how

close is close. Homoplasy and homology can be relevant at all

hierarchical levels.

The third criterionFwhether the characters have the

same or different ancestral character statesFplaces emphasis

not on the phenomenon under scrutiny (evolution of

correspondent traits at two or more separate hierarchical

levels), but on the underlying differences at an altogether

root 1

root 2

root 1 root 2

Fig. 2. Hypothetical unrooted tree showing the independentdistribution of gray as homoplasy. Rooting the tree in two sep-arate positions changes the scenario of character evolution butboth phylogenies contain a convergent phenotype, generated byreversal/loss in rooting position 1 and independent gain in rootingposition 2.

Table 1. Four criteria identified in the literature to distinguish convergence from parallelism

1 Homoplastic phenotypes are structurally correspondent in parallelism but noncorrespondent in convergence

2 Homoplastic phenotypes in closely related taxa is parallelism but in distantly related taxa is convergence

3 Homoplastic phenotypes have the same ancestral character states for parallelism but different ancestral character states for convergence

4 Parallelism comprises homoplastic phenotypes caused by the same underlying genetics resulting from an ancestral predisposition to evolve the

same character states, whereas convergent phenotypes are caused by dissimilar genetics

What is parallelism? 217Scotland

Page 5: What is Parallism

more inclusive hierarchical level, the plesiomorphic precursors

of the independent traits. This criterion is at odds with

some universally agreed examples of convergence that con-

tradict this view because the convergent phenotypes have

the same ancestral character states. For example, in the

classic case of succulent desert plants of Cactaceae compared

with succulent plants of the distantly related genus Euphor-

bia, the plesiomorphic conditions that led to both groups’

sharing structurally correspondent ridged and columnar

stems are similar. Therefore, both the legitimacy and utility

of distinguishing parallelism from convergence using criterion

3 is certainly far from universal. Furthermore, in a book de-

voted to convergence Conway Morris (2003) characterized

hundreds of examples of convergence without any explicit

focus on ancestral character states reflecting the widespread

use of this term without reference to ancestral character

states. Thus, in agreement with many authors (Haas and

Simpson 1946; Zhang and Kumar 1997; Wichman et al. 1999;

Gould 2002; Cooper et al. 2003; Colosimo et al. 2005; Fong

et al. 2005; Harrison et al. 2005; Derome et al. 2006;

Roberge et al. 2006; Shapiro et al. 2006; Christin et al. 2007;

Hall 2007), criterion 4Fwhether similar independently

evolved phenotypic character states have similar developmen-

tal mechanisms and genetics (parallelism) or not (conver-

gence)Flies at the heart of the distinction between these

categories. Criteria 1–3 are sometimes viewed individually

or collectively as reflecting, or acting as a proxy for, criterion

4. However, in the vast majority of cases, the validity of

this assumption is, at best, loose and more typically,

unknown.

Criterion 4, genetics and parallelism

Haas and Simpson (1946, p. 336) reviewed the distinction

between parallelism and convergence and noted that paral-

lelism would be similarity in structure due to a common ge-

netic basis (and so far resembling homology) but not reaching

morphological expression until after the separation of the two

hypothesis of homology

Analogy sensu Owen1843, non-homologysensu de Beer 1971,convergence sensuPatterson 1982,disparate morphologysensu Shubin et al 2009

Homoplasy sensuLankester1870, usuallyinterpreted as includingconvergence, parallelism,reversal and loss.

Homology sensu Patterson(1982) equivalent tosynapomorphy (includessymplesiomorphy) atappropriate level.

more than onehierarchical level one hierarchical level

Non- structuralcorrespondence

structuralcorrespondence

analogy & homology sensu Owen 1843

Convergence HomologyPhenotype Analogy

Non-correspondence Parallelism HomologyGenotype

Fig. 3. Schematic treatment for hypotheses of homology contrasting homology sensu Owen (1843) and homology sensu Patterson (1982).Putative hypotheses of homology are distinguished as analogy or homology in the sense of Owen (1843). Analogy is interpreted in a broadsense to include all nontopographically correspondent features including nonhomology sensu de Beer (1971) and disparate morphology sensuShubin et al. (1997, 2009). Homology sensu Owen (1843) can be further divided into hypotheses of homology that are congruent withphylogeny at a particular hierarchical level and homoplasy that is incongruent with phylogeny. Some hypotheses that are congruent withphylogeny nevertheless have incomplete conditional phrases (ancestral character states) and define paraphyletic groups. Structurally corre-spondent homology propositions that are congruent with phylogeny and define monophyletic taxa are apomorphy (synapomorphy) or taxichomology sensu Patterson (1982). Two lower panels reflect the framework adopted in this article for homoplasy. Convergence describes allphenotypic homoplasy and parallelism describes all genotypic homoplasy.

218 EVOLUTION & DEVELOPMENT Vol. 13, No. 2, March--April 2011

Page 6: What is Parallism

or more lines involved (and in this differing from homology).

Several other terms, (homoiologies, latent homology, under-

lying synapomorphy, unique inside-parallelism) have been

used at various times to describe parallel evolution (Hennig

1966; de Beer 1971; Brundin 1976; Saether 1979, 1983, 1986).

Hennig (1966) used the term homoiologies for corresponding

characters that occur in narrow kinship groups but that nev-

ertheless develop independently in their bearers. Saether

(1983, p. 343) defined underlying synapomorphy as close

parallelism as a result of common inherited genetic factors

within a monophyletic group causing incomplete synapomor-

phy and Saether (1986, p. 5) as the inherited capacity to de-

velop parallel similarities. de Beer (1971, p. 9) when discussing

latent homology, thought that criteria for recognizing homo-

logy were perhaps over exacting because in many situations

the manifestation of a homology was only visible or expressed

in some of the taxa that shared the homology. These

explanations of parallelism seek to explain a homoplastic

morphological expression associated with a genetic mecha-

nismFcommon inherited genetic factors and inherited capacity

and community of inheritanceFthat is a genetic mechanism

distributed at a more inclusive phylogenetic level than its as-

sociated homoplastic morphology.

Gould (2002), in an extensive discussion of parallelism,

makes the case that parallelism is an under-researched and

important phenomenon within evolutionary theory. His ar-

guement, partly rested on the results of evo-devo research that

demonstrate a deep homology of shared genetics that under-

pin a range of phenotypic structures. Gould (2002) uses these

examples to make the case for parallelism being channeled

from within by homologous generators (Gould 2002, p. 1079)

as the result of internal constraint (Gould 2002, p. 1080).

Gould’s treatment of parallelism states that its definition has

been problematic: parallelism is a ‘‘grey zone’’ between homo-

logy and convergence but that criteria for its operational dis-

covery, the operational rescue of parallelism by evo-devo and

the development of genetic and developmental techniques that

established the field of evo-devo have finally allowed biologists

to identify the homologous generators that always specified the

concept of parallelism in theoretical terms. Parallelism has now,

and finally after a century of terminological recognition, become

an operational subject for evolutionary research (Gould 2002,

pp. 1088–1089). Gould was correct to highlight the huge po-

tential of molecular biology to explore the relationship be-

tween the genotype and phenotype. However, this in itself

does not provide a solution per se regarding the ambiguous

use of the term parallelism, as the examples discussed below

clearly demonstrate.

A recurring theme in the discussions cited above is that

parallel evolution of the same phenotypic trait which often

but not exclusively occurs in closely related monophyletic

lineages, represents something distinct from convergence, as

the genetics and phylogenetic context (community of inher-

itance, homologous generators) that underpin parallel phe-

notypic homoplasies are more extensive than for convergent

phenotypic homoplasies. This view would implicitly consider,

analogy, convergence, parallelism, and homology as stages

along a continuous spectrum from less to more shared ge-

netics determined in part by the closeness of phylogenetic

relations. In other words, analogies and convergences have

few, if any genes in common whereas parallelism, and par-

ticularly homology, are both largely underpinned by the same

genetic mechanisms. However, the view that homologous

structures can be underpinned by nonhomologous genotypes

and vice versa (de Beer 1971) coupled with recent discoveries

that homologous genes can underpin widely disparate anal-

ogous morphologies (Panganiban et al. 1997; Shubin et al.

1997, 2009; Panganiban and Rubenstein 2002) calls this over-

simplified view into question. In short, the concept of deep

homology (Shubin et al. 1997, 2009) for examples in which

‘‘the sharing of the genetic regulatory apparatus that is used

to build morphologically and phylogenetically disparate an-

imal features,’’ has developed because analogous and homo-

plastic phenotypes can be regulated and determined by

varying degrees of similar or different underlying genetics.

The concepts of deep homology and parallelism are therefore

closely related but both require precision in their use if they

are to be more than a simple loose description of associated

genotypic and phenotypic change.

A solution for use of the term parallelism

The criteria listed in Table 1 to distinguish parallelism from

convergence may be unsuccessful because they treat these two

ideas as mutually exclusive alternatives. But what if parallel-

ism and convergence are not regarded as alternatives, but

rather that the parallel evolution of genetic traits represents

one of several possible types of explanation of phenotypic

convergence. Under this model the parallel evolution of the

same genetic traits can underpin and explain someFalthough

not allFinstances of phenotypic convergence.

Table 2. Structural correspondence and hierarchical

level distinguish noncorrespondent morphologies,

homoplasy, and homology, but not parallelism

from convergence

Structural

correspondence

Hierarchical

level

Nonhomology No N/A

Homoplasy: convergence Yes At least two

Homoplasy: parallelism Yes At least two

Homology Yes One/unique

What is parallelism? 219Scotland

Page 7: What is Parallism

Biological levels

To appreciate the perceived distinction between parallelism

and convergence it is necessary to understand the significance

of criterion 4Fwhether or not the characters have similar

developmental mechanismsFin the wider context of the dis-

sociation of homology (and homoplasy) propositions at var-

ious biological levels. de Beer (1971) was among the first to

point out that the expectation of homologous phenotypes

being determined/regulated/specified by homologous genes/

genetic mechanisms was not always true. He provided several

examples of homologous phenotypes and nonhomologous

genotypes and vice versa, a view that has been explored

widely (de Beer 1971; Roth 1984, 1988; Dickinson 1995;

Bolker and Raff 1996; Galis 1996; Abouheif 1997, 1999;

Abouheif et al. 1997; Shubin et al. 1997; Wray and Abouheif

1998; Wray 1999; Abouheif and Wray 2002; Hall 2003;

Scholtz 2005; Sommer 2008, 2009; Shubin et al. 2009; Scot-

land 2010). Given the possible dissociation of hypothesis of

homology (and homoplasy) at various biological levels that is

genetic mechanisms, genes, and morphology (see Scotland

2010 for recent discussion), what lessons can be learned rel-

ative to criterion 4 from Table 1 for distinguishing parallelism

from convergence?

Focusing on two biological levelsFphenotype and geno-

typeFwithin a simplified framework of nonhomology,

homoplasy and homology under the assumption that the

hypotheses of phenotypic and genotypic homology can be

disassociated, results in nine possible combinations (Fig. 4).

Eight of these theoretical combinations could be pertinent, the

exception being those of a nonhomologous phenotype and

nonhomologous genotype. The association of a homoplastic

phenotype as shown on the left hand side of Fig. 4, with the

genotype on the right hand side, is reflected by three arrows.

Furthermore, whether the homoplastic phenotype has the

same genetic mechanism (criterion 4), is reflected by the two

arrows pointing to homoplastic and homologous genotypes

respectively. Because homologous traits of any type (genetic

or phenotypic) define monophyletic groups, it is logically im-

possible for the same genetic mechanism to characterize a

clade as well as regulate a phenotypic convergence within that

clade, reflecting the fact that maybe all theoretical combina-

tions in Fig. 4 are not possible. Therefore, the same genetic

mechanism, associated with homoplastic phenotypes, must

also be homoplastic. As a result, the black arrow that points

between the homoplasy of the phenotype and homoplasy of

the genotype fulfills criterion 4.

In this context, the terms convergence and parallelism de-

scribe the relationship between two biological levelsFpheno-

type and genotype. Convergence refers to the association

between a homoplastic phenotype determined by nonhomol-

ogous genotype. Parallelism refers to the association between

homoplastic correspondent genotypes determining homoplas-

tic phenotypes. If this distinction is accepted, one source of

confusion between terms is that they are often used to refer to

homoplastic phenotypes (left hand side of Fig. 4 only) in the

hope that some aspect of the phenotype (not really the same,

close or distance of the comparison, ancestral character states)

can serve as a proxy for the genotypic information on the

right-hand side of the diagram. This is mistaken because if

phenotype and genotype can be dissociated, then no aspect of

the phenotype predicts the condition of the genotype and vice

versa. The obvious solution is to use the terms parallelism and

convergence only when information exists for both phenotype

and genotype, in which case they can then be used to describe

the two biological levels relative to the independent acquisi-

tion of the same phenotype.

The disadvantage of this solution is that it lacks the ability

for either term to be used in a widespread manner when re-

ferring to the independent evolution of correspondent phe-

notypes. This is because the terms will now rarely be able to

be used, as only rarely will the information required to dis-

tinguish them be available. Another solution is as follows.

Parallel evolution has long been associated with the idea of

shared genetic traits that underpin correspondent homoplastic

phenotypes with the expectation that this phenomenon is

more prevalent in closely related taxa (Haas and Simpson

1946). A possible resolution of this aforementioned dilemma,

therefore, is to view parallel and convergent evolution not as

alternatives, but rather parallelism as a possible explanation

of phenotypic convergence. Parallel evolution of the same

Phenotype Genotype Relation /description

Homology

Homoplasy

Non-homologyNon-homology

Homoplasy

Homology

Convergence

Parallelism

Fig. 4. Two biological levels of organizationFphenotype and thegenotypeFviewed relative to nonhomology (noncorrespondentcomparisons), homoplasy, and homology. There are eight possiblecombinations of interest. A black arrow points to homoplasticphenotypes in combination with nonhomologous genetic mecha-nisms and this is considered by some authors (e.g., Haas andSimpson 1946; Gould 2002; Hall 2007) as convergence. A secondblack arrow indicates the combination of homoplastic phenotypesand homoplastic genotypes; this has been characterized as paral-lelism (e.g., Haas and Simpson 1946; Gould 2002; Hall 2007). Theproblem with this framework is that convergence is a widespreadterm referring to convergent phenotypes for which no geneticinformation exists.

220 EVOLUTION & DEVELOPMENT Vol. 13, No. 2, March--April 2011

Page 8: What is Parallism

genetic traits can underpin and explain phenotypic conver-

gence. In this framework (Fig. 5), convergent phenotypes

refer to one biological level of homoplasyFthe phenotype.

Parallel genetic traits refer to one biological level of homo-

plasyFthe genotype. Parallelism of genetic traits associated

with a convergent phenotype refers to two biological levels of

homoplasyFthe genotype and the phenotype (Fig. 5). Con-

vergent phenotypes, however, have other explanations that

do not rely on the independent parallel evolution of shared

genetic traits or mechanisms.

In summary, convergence can be viewed as a concept of

phenotypic homoplasy, and parallelism viewed as a concept

of genotypic homoplasy (see Figs. 3 and 5). These two sep-

arate hypotheses of homoplasy of phenotype and genotype

can be combined such that shared homoplastic genetics can

explain convergent phenotypes and therefore the genetics that

underpin these concepts can be determined empirically rather

than assumed from subsidiary problematic criteria (Table 1).

FOUR EXAMPLES

Example 1: pale and dark coloration in animals

In an article titled ‘‘A single amino acid mutation’’ contributes

to adaptive beach mouse color Hoekstra et al. (2006) iden-

tified a derived, charge-changing amino acid mutation in the

melanocortin-1 receptor (Mc1r) in light-colored subspecies of

Peromyscus polionotus from the Florida Gulf Coast compared

with their dark-colored mainland conspecifics. The same

study reported that similarly light-colored subspecies from the

Atlantic coast were missing the derived Mc1r light-colored

allele, thus implying that different molecular mechanisms are

responsible for the superficially convergent phenotypic evo-

lution in light-colored subspecies from the Atlantic coast

compared with the Florida Gulf Coast. This same gene

(Mc1r) has also been implicated in the evolution of pale or

dark coloration in lizards, several birds, various felids, pocket

mice, black bear and woolly mammoths. In a discussion of

parallel and convergent evolution, Arendt and Reznick

(2008a) use the example from Hoekstra et al. (2006) to show

that the distinction between parallel and convergent evolution

is problematic. These authors reason that the distinction be-

tween convergent and parallel evolution assumes that, when a

given phenotype evolves, the underlying genetic mechanisms

are different in distantly related species (convergent) but sim-

ilar in closely related species (parallel). However, studies of the

Mc1r gene and coloration show that the same phenotype

might evolve among populations within a species by changes

in different genes (light-colored beach mice from the Gulf

Coast compared with the Atlantic Coast), and conversely that

similar phenotypes might evolve in distantly related species by

changes in the same gene (lizards, birds, felids, pocket mice,

black bear, and wooly mammoths). This led Arendt and

Reznick (2008a) to the conclusion that the distinction between

convergent and parallel evolution is a false dichotomy, at best

representing ends of a continuum. They concluded that all

instances of the independent evolution of a given phenotype

can be described with a single termFconvergence.

An interpretation of these data from the perspective of

homology is as follows. In the example of P. polionotus, there

seem to be three reasons why a particular group of individual

mice share the light-colored trait. Either (a) the mice are light

colored because they share a homologous genetic mutation

inherited from light-colored parents (homology of trait and

mechanism), (b) they share a genetic mutation of the same

light-colored trait but both trait and mechanism have arisen

in different individuals independently (homoplasy of both

trait and mechanism), or (c) they share the light-colored trait

but not the mechanism that determines the trait (homoplasy

of trait and nonhomology of mechanism). These three sce-

narios are depicted in Fig. 6. Thus, this case can readily be

accommodated within the framework outlined above, so long

as the various hierarchical levels implied by the comparison

are made explicit. The distribution of light pelage occurs at

three independent hierarchical levels, two in the Florida gulf

coast and one in the Atlantic coast. This phenotypic trait is a

homoplasy. The answer to the convergence/parallelism ques-

tion depends on the same qualifications as any other example

of homology or homoplasy; on the hierarchical and biological

level of the focus. Relative to Fig. 6, the presence of light

pelage is homologous (apomorphic) at three separate

Phenotype Genotype

Homology

Non-homology

Deep homologyHomology

Convergence Parallelism

Non-homology

Homoplasy

Parallel evolution

Fig. 5. Two biological levels of organizationFphenotype and ge-notypeFviewed relative to nonhomology (noncorrespondent com-parisons), homoplasy, and homology. Convergence is restricted toone side of the diagram as it describes phenotypic homoplasy.Parallelism is restricted to one side of the diagram as it describesgenotypic homoplasy. The conditional phrase associated with par-allelism can refer to genotype only for example frequent and wide-spread parallel evolution of protein sequences Rokas and Carroll(2008) or genotype in relation to phenotype, C4 photosynthesisevolved in grasses via parallel adaptive genetic changes Christinet al. (2007), in which case it is an example of parallel evolution.Deep homology as originally proposed by Shubin et al. (1997,2007) has a conditional phrase that combines nonhomologousphenotype and homologous genotype.

What is parallelism? 221Scotland

Page 9: What is Parallism

phylogenetic levels (FG, MN, and W–Z) and homoplastic

within the level of A–Z, (at three independent levels). Within

the level of E–N, light pelage is homoplastic and this ho-

moplasy is determined by the independent evolution (homop-

lasy) of the same genetic mechanism (parallelism of genetic

mechanism in relation to a convergent phenotype). In addi-

tion, comparison of light pelage at levels FG or MN com-

pared with W–Z is homoplasy that is determined by the

independent evolution of a phenotype determined by different

genetic mechanisms (convergence of phenotype).

In describing this example, I was reminded of the classic

textbook example concerning whether the wings of birds and

bats are homologous. The official answer is that the wings of

birds and bats are homologous as forelimbs at the level of

tetrapods, homoplastic as wings at the level of tetrapods, and

homologous as particular types of forelimb at the levels of birds

and bats respectively. These interpretations depend solely on

the hierarchical level of the comparison. This is the case

within any cladeFa species, a genus, or all of life. Therefore,

the evolution of light pelage in Florida gulf coast mice, liz-

ards, birds, felids, pocket mice, black bear, and woolly mam-

moths involving the Mc1r are all examples of parallel

evolution of the same genetic mechanism that has a role in

determining a convergent phenotype. The evolution of white

pelage in various subspecies of mice from the Florida gulf

coast compared with the Atlantic coast can be described as

convergent evolution of a phenotype.

Example 2: echolocation in bats and dolphins

In an article titled Convergent sequence evolution between

echolocating bats and dolphins, Liu et al. (2010) demonstrated

that the motor protein Prestin expressed in mammalian outer

hair cells has a role in echolocation in bats and dolphins. The

ability of some bats and all toothed whales to produce sonar

pulses and process the returning echoes for prey detection and

orientation (echolocation) is described as a spectacular exam-

ple of phenotypic convergence in mammals. These authors

demonstrate that echolocating whales and bats uniquely

shared 14 derived amino acids in the Prestin protein. Using

the full amino acid alignment of the Prestin protein these

authors infer a phylogenetic tree that has the echolocating

whales nested within bats due to homoplastic evolution of the

shared amino acids (Fig. 7).

A

B

C

D

EF

G

H

I

J

K

L

M

N

O

Q

R

S

T

U

V

W

X

Y

Z

Floridagulf coastand main-land

Atlanticcoastandmainlanddark pelage

light pelage

derived Mc1r allele determining light pelage

unknown genetic mechanism

P

Fig. 6. Hypothetical phylogeography ofsubspecies of Beach Mouse Peromyscuspolionotus distributed in the Florida gulfcoast and the Atlantic coast. The treeshows that light pelage (coat color) hasevolved three independent times. Two ofthese occurrences are found in mice inFlorida and it has been shown that this isdetermined by the derived Mc1r allele.Another set of mice on the Atlantic coastdo not share this allele and therefore thelight pelage in these mice is determined byanother, unknown genetic factor. Lightpelage is a convergence (three levels) atthe level of A–Z and homology at theindependent levels of FG, MN, andW–Z. The convergent light pelage at thelevel of B–N is explained by the parallelevolution involving Mc1r whereas theconvergence between light pelage in theFlorida gulf coast and Atlantic gulfcoast remains as convergence. Illustrativephylogeny to capture discussion pre-sented in Arendt and Reznick (2008a)and Hoekstra et al. (2006).

222 EVOLUTION & DEVELOPMENT Vol. 13, No. 2, March--April 2011

Page 10: What is Parallism

There seems little doubt in this example that echolocation

and associated morphological traitsFshorter and stiffer co-

chlear outer hair cellsFare homoplastic in mammals. The

question is whether this represents convergent or parallel

evolution. Following the framework developed above, the

shared echolocating phenotype of bats and dolphins and the

shared amino acids of the Prestin protein are both homo-

plasies and can be described as the same phenotype evolving

independently using the same genetic changes. Replacing

the word convergence with parallelism in the title of the

article would more accurately describe these data. In other

words, parallel evolution of the Prestin protein is associated

with phenotypic convergences involved in echolocation in

mammals.

Example 3: C4 photosynthesis in grasses

In a article entitled C4 photosynthesis evolved in grasses via

parallel adaptive genetic changes, Christin et al. (2007) dem-

onstrated that the same parallel, putatively adaptive genetic

changes were associated with the independent evolution of C4

photosynthesis in grasses. Excerpts from the abstract of this

article are very instructive in relation to the precise use of the

terms parallelism and convergence (Christin et al. 2007,

p. 1241).

Phenotypic convergence is a widespread and well-recognized evo-

lutionary phenomenon. However the responsible molecular mech-

anisms remain often unknown mainly because the genes involved

are not identified. A well-known example of physiological conver-

gence is the C4 photosynthetic pathway, which evolved indepen-

dently 445 times. Here we address the question of the molecular

bases of the C4 convergent phenotypes in grasses (Poaceae) by

reconstructing the evolutionary history of genes encoding a C4 key

enzyme (PEPC). . . . Using phylogenetic analysis we showed that

grass C4 PEPCs appeared at least eight times independently from

the same non-C4PEPC (Fig. 8). Twenty-one amino acids evolved

under positive selection and converged to similar or identical

amino acids in most of the grass C4PEPC lineages.

These authors concluded that C4 photosynthesis evolved in

grasses via parallel adaptive genetic changes using the term

convergence to refer to homoplastic phenotypes and their us-

age of parallelism is restricted to parallel genetic changes as-

sociated with this convergent phenotype. As a result this

usage is in agreement with the framework developed in this

paper.

Example 4: Drosophila trichomes

The abstract of a compelling article entitled Regulatory evo-

lution of shavenbaby/ovo underlies multiple cases of morpho-

logical parallelism (Sucena et al. 2003, p. 935) is as follows:

Cases of convergent evolution that involve changes in the same

developmental pathway, called parallelism, provide evidence that

a limited number of developmental changes are available to evolve

a particular phenotype. To our knowledge, in no case are the

genetic changes underlying morphological convergence under-

stood. However, morphological convergence is not generally as-

sumed to imply developmental parallelism. Here we investigate a

case of convergence of larval morphology of insects and show that

the loss of particular trichomes, observed in one species of the

Drosophila melanogaster species group, has independently evolved

multiple times in the distantly related D. virilis species group. We

present genetic and gene expression data showing that regulatory

changes of the shavenbaby/ovo (svb/ovo) gene underlie all inde-

pendent cases of this morphological convergence. Our results in-

dicate that some developmental regulators might preferentially

accumulate evolutionary changes and that morphological paral-

lelism might therefore be more common than previously

appreciated.

The conceptual framework of Sucena et al. (2003) comprises

morphological convergence and parallel developmental path-

ways to explain those morphological convergences. The au-

thors conclude however that morphological parallelism is

perhaps more common than previously appreciated. These

Dog

CatRat

MouseGerbilRabbit

Pig

Horse

Cow

Toothed echolocatingwhales

Echolocatingbats

Echolocatingbats

Fruitbats

baleenwhales

Human

Correct positionfor toothed echolocatingwhales

Fig. 7. Evidence of sequence convergence in the Prestin genebetween dolphins and bats. Phylogeny based on amino acid data.Tree shows echolocating toothed whales (dolphins) nested withinbats. Gray line indicates the correct position of toothed whales.Echolocation is convergent for some bats and dolphins and evolvesvia parallel changes in sequence evolution of the Prestin protein.Modified and redrawn from Liu et al. (2010).

What is parallelism? 223Scotland

Page 11: What is Parallism

authors seek to reinterpret cases of morphological conver-

gence as morphological parallelism, when it can be shown

that the same genetics underpin the independent evolution

of the same phenotype. A simpler and more straight-

forward interpretation of these data is that the convergent

evolution (loss of trichome) of the same phenotype can be

explained by parallel evolution of svb regulatory mechanism

(Fig. 9).

DISCUSSION AND CONCLUSIONS

Homology and homoplasy are terms used to describe the

distribution of correspondent traits relative to phylogeny. Just

as homology is associated with descriptive terms such as

unique innovation and evolutionary novelty, homoplasy is

associated with a plethora of terms such as loss, reversal,

parallelism and convergence. Relative to homoplasy however,

A V MR

E A FP

A V MR

A V MR

E A LP

A V MR

E A MP

E AP

E W LP

E A LP

E S MP

A V MR

E A FH

E A LP

A V MR

E Q FP

A

S

A

S

S

A

S

S

S

S

S

A

S

S

A

S

A

P

A

A

P

A

P

P

P

P

P

A

P

P

A

P

L

V

L

L

V

L

V

I

I

I

V

L

I

I

L

I

T

C

T

T

C

T

T

C

T

C

T

T

A

A

T

T

A

S

S

S

T

A

S

S

T

S

S

A

S

S

A

S

S

A

S

S

A

S

A

A

A

A

A

S

A

A

S

S

F

V

F

V

V

F

V

V

V

F

V

F

V

V

F

I

R

K

R

R

K

R

K

K

K

K

K

R

K

K

R

K

M

Fig. 8. Grass phylog-eny with bold bran-ches corresponding toC4 species and non-bold branches C3 spe-cies. Colored panel of12 amino acid posi-tions from the C4 keyenzyme PEPC, show-ing parallel evolutionto the same amino ac-ids in the C4 taxa. C4

photosynthesis evol-ved in grasses via par-allel adaptive geneticchanges. Black arrowscorrespond to C4 taxa.Modified and redrawnfrom Christin et al.(2007).

224 EVOLUTION & DEVELOPMENT Vol. 13, No. 2, March--April 2011

Page 12: What is Parallism

this issue can be simplified because the common pattern of all

phenotypic homoplasy is convergence and genotypic homop-

lasy is parallelism the equivalent term for molecular data. Just

as homology has a number of explanations and causes

(Van Valen 1982; Shubin et al. 1997, 2009; Wagner and

Lynch 2010), so too is convergence associated with a number

of possible causes, explanations and associations that is re-

versal, loss, parallel genetic changes, nonhomologous genetic

changes, near and distant relatives, ancestral and non-ances-

tral character states, positive developmental constraint, func-

tional convergent trait shaped by natural selection, etc. The

relative extent of these various associated phenomena can be

determined empirically if the underlying phylogenetic pattern

is discovered independent of anyone particular association or

process. In other words, the description of phenomena and

terms that refer to an explanation of phenomena should be

distinctFdistinguishing the explanans from the explanandum

(Brady 1985)Fand confusion and inconsistency regarding

which of these a term refers to, are at the root of many con-

troversies in biology (e.g., Gould and Vrba 1982).

The view presented in this article provides a simplified

framework for what evolutionary and developmental biolo-

gists have to explain in terms of comparative anatomy: the

unique evolution of correspondent traitsFnovelty/synapo-

morphyFand the evolution of independent traitsFconver-

gence or its molecular equivalent parallelismF(Fig. 3). This

clarification of the use of the terms convergence and paral-

lelism involves restricting convergence to describing the inde-

pendent evolution of correspondent phenotypes and view

parallelism as describing the homoplasy of genotypes (Figs. 3

and 5). These two levels of homoplasy from the phenotype

and genotype can then be combined to describe parallel evo-

lution that is parallel genetic traits that underpin or are at least

associated with phenotypic convergence. Convergence is

about phenotypes, whereas parallelism, just like deep homo-

logy, is a conditional phrase that attempts to describe the

relationship between genotype and phenotype. Convergent

evolution involves one hypothesis whereas parallel evolution

of genetic traits that underpin convergence of the phenotype,

comprises two. When the term parallelism is restricted to all

instances of genotypic homoplasy (5 the genotypic equiva-

lent of phenotypic convergence), the meaning of parallel evo-

lution becomes clear and unambiguous. One possible

objection to this framework is that the use of the term par-

allelism to describe all genotypic homoplasy is too broad and

that parallel genetic traits should be distinguished from

genetic traits due to convergence and reversal. Rokas and

Carroll (2008) distinguish molecular substitutions due to

reversal to the pleseiomorphic state, convergence from differ-

ent ancestral states and identical parallel changes to the same

state, whereas Bull et al. (1997) adopted a broad view of

convergence to describe all genotypic homoplasy. Despite

terminological differences, these studies (Bull et al. 1997;

Rokas and Carroll 2008) are clear in their use of these terms.

If parallelism is a term used to describe all instances of ge-

notypic homolpasy, there is no reason why the shared pres-

ence of the same or different ancestral character states cannot

be explicitly part of that framework where appropriate. The

widely different use of these termsFconvergence, reversal,

and parallelismFto describe different facets of molecular ho-

moplasy at the DNA and amino acid sequence level conveys

the impression that most authors don’t distinguish funda-

mentally different underlying processes.

In conclusion, Arendt and Reznick (2008a) were correct to

highlight the confusing and ambiguous use of the terms par-

allel and convergent evolution that pervades the evolutionary

literature. They were also correct to demonstrate that both

D. virilis

D. kanekoi

D. ezoana

D. littoralis

D. borealis(Eastern)

D. borealis(Western)

D. lacicola

D. montana

D. flavomontana

Fig. 9. A phylogeny of selected species of Drosophila presented bySucena et al. (2003). The authors interpretation was of loss oftrichomes on three separate occasions (black bars) resulting inconvergent phenotypes. The authors demonstrate that parallelchanges in the shavenbaby/ovo gene underly all three independentcases of loss of trichomes. Redrawn from Sucena et al. (2003).

What is parallelism? 225Scotland

Page 13: What is Parallism

similar and different genetic mechanisms can be associated

with very closely related convergent phenotypes. I also agree

with those authors that all instances of the independent evo-

lution of a given phenotype can be described with a single

termFconvergence. However, Arendt and Reznick’s starting

point, that parallelism and convergence were separate and

distinct concepts primarily distinguished on close or remote

relationship of the taxa, is not an accurate nor full charac-

terization of the manner in which these terms have been used

in the literature of evolutionary biology. It was their use of

this incomplete framework that led them to conclude that the

two supposedly distinct ideas could not be distinguished and,

as a result, suggest that one termFconvergenceFsuffices to

describe independent evolution of phenotypes. I consider this

to be mistaken, because the independent evolution of genetic

traits (parallelism) that underpin or are at least associated

with convergent phenotypes, is an increasingly active area of

research and has become a well characterized phenomena

(Sucena et al. 2003; Harrison et al. 2005; Christin et al. 2007;

Liu et al. 2010). Adopting the view that parallel evolution of

genotypes can explain or is at least correlated with some but

not all convergent phenotypes accommodates the ultimate

convergence of Leander (2008), the main arguments presented

by Arendt and Reznick (2008a) and Gould (2002), as well as

providing a conceptual framework for several examples from

the recent literature (Hoekstra and Nachman 2003; Sucena

et al. 2003; Hoekstra et al. 2006; Christin et al. 2007; Liu et al.

2010).

AcknowledgmentsI thank Richard Bateman, Maxim Kapralov, Ian Kitching, NormanMcLeod, Bill Wickstead, Brian Hall, and Rudi Raff for commentson the manuscript. Thanks also to Pascal-Antoine Christin whoprovided help with Fig. 8 and Angela Hay who helped with Fig. 1.Thanks to Miltos Tsiantis for many thought-provoking andinteresting discussions about parallelism.

REFERENCES

Abouheif, E. 1997. Developmental genetics and homology: a hierarchicalapproach. Trends Ecol. Evol. 12: 405–408.

Abouheif, E. 1999. Establishing homology criteria for regulatory gene net-works. In G. R. Bock and G. Cardew (eds.). Homology. Wiley, Chich-ester, pp. 207–222.

Abouheif, E., et al. 1997. Homology and developmental genes. TrendsGenet. 13: 432–433.

Abouheif, E., and Wray, G. A. 2002. Evolution of Development, Encyclo-pedia of Life Sciences. Nature Publishing Group, London.

Arendt, J., and Reznick, D. 2008a. Convergence and parallelism reconsid-ered: what have we learned about the genetics of adaptation? TrendsEcol. Evol. 23: 26–32.

Arendt, J., and Reznick, D. 2008b. Moving beyond phylogenetic assump-tions about evolutionary convergence: response to Leander. Trends Ecol.Evol. 23: 483–484.

Bock, W. J. 1974. Philosophical foundations of classical evolutionary clas-sification. Syst. Zool. 22: 375–392.

Bolker, J. A., and Raff, R. A. 1996. Develomental genetics and traditionalhomology. BioEssays 18: 489–494.

Brady, R. H. 1985. On the independence of systematics. Cladistics 1: 113–126.

Brundin, L. 1976. A Neocomian chironomid and Podonominae–Aphrote-niinae (Diptera) in the light of phylogenetics and biogeography. Zool.Scr. 1: 107–120.

Bull, J. J., et al. 1997. Exceptional convergent evolution in a virus. Genetics147: 1497–1507.

Castoe, T. A., et al. 2009. Evidence for an ancient adaptive episodeof convergent molecular evolution. Proc. Natl. Acad. Sci. 106: 8986–8991.

Christin, P.-A., Salamin, N., Savolainen, V., Duvall, M. R., and Besnard,G. 2007. C4 photosynthesis evolved in grasses via parallel adaptivegenetic changes. Curr. Biol. 17: 1241–1247.

Colosimo, P. F., et al. 2005. Widspread parallel evolution in sticklebacks byrepeated fixation of Ectodysplasin Alleles. Science 307: 1928–1933.

Conway Morris, S. 2003. Life’s Solution: Inevitable Humans in a LonelyUniverse. Cambridge University Press, Cambridge.

Conway Morris, S. 2010. Evolution: like any other science is predictable.Philos. Trans. R. Soc. B 365: 133–145.

Cooper, T. F., Rozen, D. E., and Lenski, R. E. 2003. Parallel changes ingene expression after 20,000 generations of evolution in Escherichia coli.Proc. Natl. Acad. Sci. 100: 1072–1077.

Cracraft, J. 2005. Phylogeny and evo-devo: characters, homology, and thehistorical analysis of the evolution of development. Zoology 108: 345–356.

de Beer, G. 1971. Homology, An Unsolved Problem. Oxford UniversityPress, Oxford.

Derome, N., Duchesne, P., and Bernatchez, L. 2006. Parallelism in genetranscription among sympatric lake whitefish (Coregonus clupeaformisMitchll) ecotypes. Mol. Ecol. 15: 1239–1249.

Dickinson, W. J. 1995. Molecules and morphology: where’s the homology?Trends Genet. 11: 119–121.

Donoghue, M. J. 1992. Homology. In E. F. Keller and E. A. Lloyd (eds.).Keywords in Evolutionary Biology. Harvard University Press, Cambridge,MA, pp. 170–179.

Doolittle, R. (1994) Trends Biochem. Sci. 19: 15–18.Fong, S. S., Joyce, A. R., and Palsson, B. O. 2005. Parallel adaptive

evolution cultures of Escherichia coli lead to convergent growthphenotypes with different gene expression states. Genome Res. 15:1365–1372.

Galis, F. 1996. The evolution of insects and vertebrates: homeobox genesand homology. Trends Ecol. Evol. 11: 402–403.

Gould, S. J. 2002. The Structure of Evolutionary Theory. The Belknap Press,Cambridge, MA.

Gould, S. J., and Vrba, E. S. 1982. ExaptationFa missing term in thescience of form. Paleobiology 8: 4–15.

Haas, O., and Simpson, G. G. 1946. Analysis of some phylogenetic terms,with attempts at redefinition. Proc. Am. Philos. Soc. 90: 319–349.

Hall, B. G. 2008. Phylogenetic Trees Made Easy. Sinauer Associates, Sun-derland, MA.

Hall, B. K. 2003. Descent with modification: the unity underlying homologyand homoplasy as seen through an analysis of development and evo-lution. Biol. Rev. 78: 409–433.

Hall, B. K. 2007. Homoplasy and homology: dichotomy or continuum?J. Hum. Evol. 52: 473–479.

Harrison, C. J., Corley, S. B., Moylan, E. C., Alexander, D. L., Scotland, R.W., and Langdale, J. A. 2005. Independent recruitment of a con-served developmental mechanism during leaf evolution. Nature 434:509–514.

Hennig, W. 1966. Phylogenetic Systematics. University of Illinois Press,Urbana.

Hoekstra, H. E., Hirschmann, R. J., Bundey, R. A., Insel, P. A., andCrossland, J. P. 2006. A single amino acid mutation contributes toadaptive beach mouse color pattern. Science 313: 101–104.

Hoekstra, H. E., and Nachman, M. W. 2003. Different genes underlieadaptive melanism in different populations of rock pocket mice. Mol.Ecol. 12: 1185–1194.

Lankester, E. R. 1870. On the use of the term homology in modern zoology,and the distinction between homogenetic and homoplastic agreements.Annu. Mag. Nat. Hist. 4 (6): 34–43.

226 EVOLUTION & DEVELOPMENT Vol. 13, No. 2, March--April 2011

Page 14: What is Parallism

Leander, B. S. 2008. Different modes of convergent evolution reflect phylo-genetic distances: a reply to Arendt and Reznick. Trends Ecol. Evol. 23:481–482.

Liu, Y., Cotton, J. A., Shen, B., Han, X., Rossiter, S. J., and Zhang, S.2010. Convergent sequence evolution between echolocating bats anddolphins. Curr. Biol. 20: 53–54.

Owen, R. 1843. Lectures on the Comparative Anatomy and Physiology of theInvertebrate Animals, Delivered at the Royal College of Surgeons. Long-man, Brown, Green, and Longmans, London.

Page, R. D. M., and Holmes, E. C. 1998. Molecular Evolution. Blackwell,Oxford.

Panganiban, G., et al. 1997. The origin and evolution of animal append-ages. Proc. Nat. Acad. Sci. USA 94: 5162–5166.

Panganiban, G., and Rubenstein, J. L. R. 2002. Developmental functions ofthe distal-less/Dlx homeobox genes. Development 129: 4371–4386.

Patterson, C. 1982. Morphological characters and homology. In K. A.Joysey and A. E. Friday (eds.). Problems of Phylogenetic Reconstruction.Academic Press, London, pp. 21–74.

Patterson, C. 1988. Homology in classical and molecular biology.Mol. Biol.Evol. 5: 603–625.

Reidl, R. 1979. Order in Living Organisms. John Wiley, New York.Rieppel, O. 1988. Fundamentals of Comparative Biology. Birkhauser, Basel.Roberge, C., Einum, S., Guderley, H., and Bernatchez, L. 2006. Rapid

parallel evolutionary changes of gene transcription profiles in farmedAtlantic salmon. Mol. Ecol. 15: 9–20.

Rokas, A., and Carroll, S. B. 2008. Frequent and widespread parallel evo-lution of protein sequences. Mol. Biol. Evol. 25: 1943–1953.

Roth, V. L. 1984. On homology. Biol. J. Linn. Soc. 22: 13–29.Roth, V. L. 1988. The biological basis of homology. In C. J. Humphries

(ed.). Ontogeny and Systematics. Columbia University Press, New York,pp. 1–26.

Saether, O. A. 1979. Underlying synapomorphies and anagenetic analyses.Zool. Scr. 8: 305–312.

Saether, O. A. 1983. The canalized evolutionary potential: inconsistencies inphylogenetic reasoning. Syst. Zool. 32: 343–359.

Saether, O. A. 1986. The myth of objectivityFpost Hennigian deviations.Cladistics 2: 1–13.

Sanderson, M. J., and Donoghue, M. J. (1996).Homoplasy: The Recurrenceof Similarity in Evolution. M. J. Sanderson and L. Hufford (eds.). Ac-ademic Press, New York, pp. 67–89.

Sanderson, M. J., and Hufford, L. 1996. Homoplasy. Academic Press, SanDiego.

Scholtz, G. 2005. Homology and ontogeny: pattern and process in com-parative developmental biology. Theory Biosci. 124: 121–143.

Scotland, R. W. 2010. Deep homology: a view from systematics. BioEssays32: 438–449.

Scott, W. B. 1891. On the osteology of Mesohippus and Leptomeryx, withobservations on the modes and factors of evolution in the Mammalia.J. Morphol. 5: 301–402.

Scott, W. B. 1896. Palaeontology as a morphological discipline. BiologicalLetters of the Marine Biology Laboratory, Woods Hole Summer Ses-sion, 1895, pp. 43–61.

Shapiro, M. D., Bell, M. A., and Kingsley, D. M. 2006. Parallel geneticorigins of pelvic reduction in vertebrates. Proc. Natl. Acad. Sci. 103:13753–13758.

Shubin, N., Tabin, C., and Carroll, S. 1997. Fossils, genes and the evolutionof animal limbs. Nature 388: 639–648.

Shubin, N., Tabin, C., and Carroll, S. 2009. Deep homology and the originsof evolutionary novelty. Nature 457: 818–823.

Simpson, G. G. 1945. The principles of classification and a classification ofmammals. Bull. Am. Mus. Nat. Hist. 85: 1–350.

Saether, O. A. 1983. The canalized evolutionary potential: inconsistencies inphylogenetic reasoning. Syst. Zool. 32: 343–359.

Sneath, P. H. A., and Sokal, R. R. 1973. Numerical Taxonomy. Freeman,San Francisco.

Sommer, R. J. 2008. Homology and the hierarchy of biological systems.BioEssays 30: 653–658.

Sommer, R. J. 2009. The future of evo-devo: model systems and evolu-tionary theory. Nat. Rev. Genet. 10: 416–422.

Sucena, E., Delon, I., Jones, I., Payre, F., and Stern, D. 2003. Regulatoryevolution of shavenbaby/ovo underlies multiple cases of morphologicalparallelism. Nature 424: 935–938.

Van Valen, L. M. 1982. Homology and causes. J. Morphol. 173: 305–312.Wagner, G. P. 1989a. The biological homology concept. Ann. Rev. Ecol.

Syst. 20: 51–69.Wagner, G. P. 1989b. The origin of morphological characters and the

biological basis of homology. Evolution 43: 1157–1171.Wagner, G. P., and Lynch, V. 2010. Evolutionary novelties. Curr. Biol. 20:

R48–R52.Wake, D. B. (1996). Introduction. In M. J. Sanderson and L. Hufford

(eds.). Homoplasy. Academic Press, San Diego, pp. xvii–xxv.Wichman, H. A., Badgett, M. R., Scott, L. A., Boulianne, C. M., and Bull,

J. J. 1999. Different trajectories of parallel evolution during viral adap-tation. Science 285: 422–424.

Wray, G. A. 1999. Evolutionary dissociations between homologous genesand homologous structures. In G. R. Bock and G. Cardew (eds.).Homology. Novartis Foundation Symposium 222. John Wiley & Sons,Chichester.

Wray, G. A., and Abouheif, E. 1998. When is homology not homology?Curr. Opin. Genet. Dev. 8: 675–680.

Zhang, J., and Kumar, S. 1997. Detection of convergent and parallel evo-lution at the amino acid sequence level. Mol. Biol. Evol. 14: 527–536.

What is parallelism? 227Scotland