Gene trees, Species trees & systematics: a cladistic perspective

Abstract

The proliferation of molecular data in systematics has opened a Pandora's Box of alternate approaches to inferring hierarchical patterns of realtionship among taxa. In this review, we examine practical and theoretical reasons for employing some methods and avoiding others. We offer a philosophical overview of the relationship between systematics patterns and evolutionary processes, and we discuss the differential emphasis given to each of these areas by opposing methodological camps. We review the sources and types of incongruence between data partitions from different sources and recommend a specific procedure for contending with incongruence. We then focus on inference of relationships among closely related taxa, with particular emphasis on mtDNA as a source of characters, its advantages and potential pitfalls. We conclude with a review of several widely cited empirical studies and suggest that the gene-tree species problem may be less severe than its prevalence in the literature would suggest.

Full text

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
Annu. Rev. Ecol. Syst.1996. 27:423–50
Copyright c
1996 by Annual Reviews Inc. All rights reserved
GENE TREES, SPECIES TREES,
AND SYSTEMATICS:
A Cladistic Perspective
A. V. Z. Brower1, R. DeSalle1, and A. Vogler2
1Department of Entomology, American Museum of Natural History, Central Park West
at 79th Street, New York, NY 10024; 2Department of Entomology, The Natural
History Museum, Cromwell Road, London SW7 5BD, United Kingdom; and
Department of Biology, Imperial College at Silwood Park, Ascot, Berkshire, SL5 7PY,
United Kingdom
KEY WORDS: mtDNA, phylogeny, coalescence, cladistics, empiricism
ABSTRACT
The proliferation of molecular data in systematics has opened a Pandora’s box of
alternate approaches to inferring hierarchical patterns of relationship among taxa.
In this review, we examine practical and theoretical reasons for employing some
methods and avoiding others. We offer a philosophical overview of the relation-
ship between systematics patterns and evolutionary processes, and we discuss the
differential emphasis given to each of these areas by opposing methodological
camps. We review the sources and types of incongruence between data partitions
from different sources and recommend a specific procedure for contending with
incongruence. We then focus on inference of relationships among closely related
taxa, with particular emphasis on mtDNA as a source of characters, its advan-
tages and potential pitfalls. We conclude with a review of several widely cited
empirical studies and suggest that the gene tree–species tree problem may be less
severe than its prevalence in the literature would suggest.
INTRODUCTION
The architects of the New Synthesis (e.g. 54, 114, 164) unified disparate disci-
plinesunder theexplanatory umbrellaof evolutionarytheory,whichtheyargued
provided an empirically sound philosophical framework for all biology (165).
Nevertheless, the day-to-day work of microevolutionists studying changes of
allele frequencies in populations, and of macroevolutionists studying historical
423
0066-4162/96/1120-0423$08.00

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
424 BROWER, DESALLE & VOGLER
patterns of diversity, was carried on in largely separate empirical and intellec-
tual realms. Systematists recognized a logical lower bound for their methods
(85): Belowthepoint whererelationshipsbetween taxabecome nonhierarchical
(i.e. when dichotomously branching trees dissolve into anastomosing networks
connected by interbreeding), systematic methods become philosophically and
practically inappropriate for describing character relations and inferring rela-
tionships among organisms (46, 56, 131). By contrast, the upper bound of
population genetics was limited by the physical constraint of the disjunction
of gene pools between species. In recent years, technological advances have
made it increasingly possible to apply genetic techniques and data to evolu-
tionary questions beyond the traditional species boundary, and the success of
these efforts has led many researchers to view larger-scale patterns of diver-
sity as interpretable with the same methodological tools, or at least within the
same philosophical framework, as population-level patterns (e.g. 4, 12, 111,
189).
In this review, we compare population genetic and cladistic approaches to
interpreting patterns of relationship among closely related taxa. As some of the
most interesting issues in evolution are accessible only by study of such groups,
our main objective is to investigate the problem of discovering hierarchical
patterns around the species level, with respect to the complications of ancestral
polymorphism, lineage sorting, and introgression. Discovery of hierarchical
patterns of relationship among taxa is widely regarded to be the province of
systematics (20, 139). In principle, the cladistic method permits inference
of relationships as soon as characters are discovered that allow diagnosis of
discrete groups of populations (taxa). However, theoretical studies have argued
that the probability of error of cladistic methods remains high for long periods
after gene flow has ceased between divergent populations, if the cladograms are
inferred from molecular data from single genes (38, 89, 127, 138, 178, 181).
To counter and correct for these potential difficulties, numerous alternative
approaches have been devised.
Our essay begins with a brief review of philosophical issues that underlie
the differences between population genetic and systematic approaches to the
gene tree–species tree problem. The fundamental role of philosophy as arbiter
between competing interpretive paradigms is acknowledged only rarely in the
methodologically or empirically focused literature, but it is in this realm that
we believe methods must compete. We hope this discussion provides a start-
ing point for philosophical debate among the various methodological schools.
We then examine ideas of species and speciation, showing how the cladistic
approach circumvents many of the ontological difficulties besetting these com-
plex theoretical issues. Because congruent hierarchical character distributions

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 425
provide the evidence for inferring patterns of phylogenetic relationship, char-
acter incongruence among closely related taxa is theoretically problematical
(47). We review theories, models, and observed patterns of character incon-
gruence in some detail. We draw empirical examples from animal mitochon-
drial data sets, as their interpretation has been paradigmatic in the gene tree–
species tree theories of systematists and population geneticists alike (6, 49,
56, 82, 86, 120, 141). Many readers will disagree with our philosophical
stance, but we hope our exploration of ideas and of empirical studies provides
a framework for better understanding the advantages and limits of alternate
approaches.
EPISTEMOLOGY AND “TRUTH”
Scientific hypotheses are framed in the context of plausible prior theories that
are accepted as background knowledge (assumptions not tested by the data
testing the hypothesis in question, but that are in principle testable). Accepting
a particular body of background knowledge (an ontology) is justifiable only by
recourse to an infinite regressof prior, moregeneral assumptions of background
knowledge (144). Epistemology is the questioning of the adequacy of these
assumptions: Howdowe constructa rational frameworktointerpret andexplain
our observations? Epistemological questioning diminishes when a discipline
reaches a consensus about an ontological paradigm; at that point the discipline
becomes a “normal science” (105, 139).
Both systematics and population genetics depend on generally accepted, un-
derlying ontological tenets; they are thus Kuhnian normal sciences. However,
these tenets differ greatly between the two fields. Population genetics is a de-
terministic, regular science that draws deductive inferences about evolutionary
process from established laws of heredity (109, 147, 193), while systematics
is a historical science that attempts to discover and describe the intricacies of
the evidently hierarchical pattern of nature (70, 139, 149). Population genet-
ics seeks to document and explain the origin and maintenance of diversity in
systems where continuity of process is explicitly postulated (evolution happens
onlybecause of deviationsfrom Hardy-Weinbergequilibrium conditions). Sys-
tematics seeks to document hierarchical patterns among disjunct entities and
needsto postulate littleexcept thata tree-like hierarchy exists andis recoverable
by studying attributes of individual organisms.
Reviewing tree-building algorithms and philosophies lies beyond our scope
here (see 162, 176 for references), but some comments about their assumptions
may highlight the pattern-process dichotomy discussed above. Distance-based
methods such as UPGMA (166) depend on prior knowledge about rates of
evolution. These methods were popular in the 1970s, when optimism about the

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
426 BROWER, DESALLE & VOGLER
existence of a molecular clock ran high. More recent distance methods, such
as neighbor joining (155), add additional model parameters to buffer against
manifest violations of clock-like regularity in DNA evolution (e.g. 73).
Likedistance methods,each ofthe myriadvariations ofphylogenyestimation
via maximum-likelihood (e.g. 64, 66, 74) also relies on an explicit underlying
modelof character transformation. Thesemodels rangefrom simple butunreal-
istic one- or two-parameter approximations to complex models that incorporate
relative frequencies of amino acid changes (103). Extra parameters increase
“realism” by improving the fit of the model to some “true” topology (see be-
low), but adding them sacrifices both explanatory power and computational
speed.
Further, because methods that rely on explicit, a priori models of evolution
are acknowledged to be poor estimators of hierarchical pattern when the as-
sumptions of the models are violated (65, 200), the underpinnings of a chosen
model must be defensible on empirical grounds. Although statistical computer
simulations of model-fit and robustness-to-error (e.g. 60, 92, 153) may provide
quantitative rigor, the plausibility of a particular model of evolution, and there-
fore the verisimilitude of results from its employment in systematic inference,
has no ultimate extrinsic appeal except to congruence with “known” topolo-
gies, derived from real data and real organisms by an independent method (121,
122).
It is frequently stated that all phylogenetic inferences depend on underlying
models of the evolutionary process (e.g. 27, 74). This does not mean that all
models are equally general or entail results with equal explanatory potential.
When population geneticists, evolutionary systematists, and “phylogenetic”
systematists1examine hierarchical patterns of relationship, they operate im-
plicitly or explicitly with the presupposition that evolutionary processes (such
as phylogeny) are responsible for the observed patterns of biotic diversity. The
more explicit the process model, the more precise the estimate of relationships
may be. But precision is not a substitute for accuracy, and the accuracy of
the method (an ontological claim) is only measurable indirectly, by reference
1BothHennig (85) andWiley(199) made theterm “phylogenetic” synonymouswith “genealog-
ical.” As recognized by early evolutionists (e.g. 45, 94), propinquity of descent is not the cause
of, but a theory that explains, our continuing successful discovery of a hierarchical Natural System
of classification that was well established long before Darwin’s or anyone else’s materialist causal
hypothesis was proposed to explain it (see e.g. 17, 173, 195, and discussion in 130, 135, 172). As
pointed out by Brady (20), systematics must exist independent from process theories if evolution
is to avoid tautology. The term “phylogeny” is so vague and metaphysical as to be embraced
by researchers with antithetical systematic philosophies (e.g. 35 contra 161; 129 contra 49). For
clarity, we avoid “phylogeny” and “phylogenetic” when referring to cladograms or trees, but rather
use the term only to describe the process theoretically responsible for the patterns we observe.

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 427
to the plausibility of the background knowledge underlying the model (167,
194).
To a cladist, phylogeny is a compelling and well-corroborated process theory
that provides a causal explanation for the inferred patterns. It is not part of
the framework of background knowledge within which results are interpreted
(except in the sense that some single hierarchical pattern is postulated to exist;
142). Cladistic methods rely on general ontological claims that lie deeper
in the infinite regress of background knowledge than the claims underlying
methods based on more explicit process models, instilling them with generally
greaterepistemological credibility(62). Moreimportantly,cladistic hypotheses
are free from specific evolutionary assumptions (143) and therefore represent
independent explananda for adaptive or phylogenetic process theories (20).
Process theories that do not fit the observed pattern may be rejected empirically,
rather than dictating the pattern discovered.
At the bottom of the ontological problem of understanding phylogeny is the
epistemological question, “How much about the evolutionary process are we
willing to accept as background knowledge when inferring particular patterns
from empirical observations, and constructing causal hypotheses to explain
them?” In our view, a strong justification of the plausibility of underlying as-
sumptionsis thesine quanon ofa particularmethod. It isnot difficultto imagine
myriad possible explanations for patterns we observe in nature, once we aban-
don the position that logical consistency is a necessary underlying feature of
the chosen methodology.
THE PROCESS OF SPECIATION
To proceed with our discussion of pattern inference at the species boundary, we
need to ask, “What are species, how do they cometo be, and howdo we discover
them?” Understanding the gradation of differences between taxa has remained
perhaps the most elusive and intractable challenge to natural historians (and
later, evolutionary biologists) since the beginnings of modern systematics in
the eighteenth century (19). The vexedness of the problem continues to inspire
philosophical tracts, definitions, and discussions of the “reality” of species
(e.g. 12, 13, 48, 49, 70, 77, 110, 111, 196), but we are unconvinced that the
conception of the process has advanced beyond Hennig’s (85) oft-reproduced
cartoon (Figure 1). This diagram represents a simple ontological model of the
speciation process, as the bifurcation of a braided lineage of interbreeding in-
dividuals into two such lineages. Each daughter lineage retains its own genetic
connectedness, yet gains a discrete genetic identity with respect to its sister. As
Hennig recognized, however realistic such a model might seem, our ability to
assess its validity springs from, and is constrained by our access to, empirical

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
428 BROWER, DESALLE & VOGLER
Figure 1 Hennig’s (Figure 6 in 85) diagram of “the total structure of hologenetic relationships
and the differences in form associated with its individual parts.” Development of single individu-
als (ontogenetic relationships), interbreeding between individuals (tokogenetic relationships), and
bifurcation of species (phylogenetic relationships) are depicted as coexisting at nested levels of or-
ganization. Note that (coincidentally) the mtDNA of individuals in each descendant species would
be fixed after one generation for alternate haplotypes from the ancestral species. Copyright 1979
by the Board of Trustees of the University of Illinois. Used with permission of the University of
Illinois Press.

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 429
information. The process is an explanation for the pattern inferred from com-
parative data, which almost always is limited to observed differences in holo-
morphology (“the multidimensional gestalt of the semaphoronts,” p. 66).
Rieppel (149) argued that species have a necessarily dual nature: Depend-
ing on the question being asked, species must be seen as either individuals or
classes. He characterized these alternate views as nominalist and essentialist,
and he argued compellingly that they are both potentially valid, although epis-
temologically incompatible. We suggest that the contrast between population
genetics and systematics developed above reflects these two viewpoints. The
nominalist population-genetic view of species is an inclusive one, of species
made contemporaneously coherent by gene flow and historically connected by
the coalescence of alleles. There are no breakpoints in the continua of lin-
eages, and divisions between species are not recognizable, except after the fact
and by arbitrary criteria, or “statistical distributions” (111). By contrast, many
cladists identify species as exclusive classes composed of individuals united by
sharing unique combinations of characters (34, 130, 131). Metaphysical ques-
tions concerning the true nature of species are not important to the operation
of identifying groups by this criterion. Trying to shoehorn additional informa-
tion from the nominalist perspective, such as cohesion (185) and exclusivity
(13, 49, 88a), into cladistic discovery of patterns serves only to obfuscate the
process-independent, empirical basis of the method, which is its chief strength.
GENE TREES AND INCONGRUENCE
The cladistic method implies the most parsimonious cladograms for the taxa
examined, given a particular data set. Whether these cladograms correspond
to the “true” phylogenetic history of the taxa is not subject to empirical proof
or disproof. Instead, they should be considered as theories, subject to corrob-
oration or refutation by comparison with additional data (144, 198). We do
not address theoretical issues of data partitioning versus simultaneous analysis
in systematic inference in depth, as the topic has been reviewed recently (47,
122, 132). If the simultaneous analysis approach is rigorously applied, there
are no separate data sets or topologies to conflict with one another. However,
many systematists are not satisfied with this view, and all recognize that some
variable traits of organisms have poor potential as characters. A particular set
of characters2may not reflect “organismal history” (or be corroborated by the
bulkof otherevidence thatwe believedoes reflect organismal history). Concern
2The idea that logically separable data sets can be drawn from the same taxa is problematical
(47, 104, 122, 132). For convenience, we refer to data drawn from different sources (e.g. mtDNA
sequence versus morphology, or one gene versus another) as different “data sets,” although they
might more appropriately be called subsets or hypothetical “process partitions” (27, 122).

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
430 BROWER, DESALLE & VOGLER
Figure 2 Flow chart illustrating the classification of congruence types and the procedure for
distinguishing them, described in the text. Differential character weighting of incongruent data
sets is shown as an alternate pathway, with dashed lines signifying our reservations about such
techniques (see also 24).
over data incongruence is increasing as multiple molecular data sets become
available for comparison with one another and with morphology for the same
taxa. The remainder of this paper addresses the sources, hypothesized causes,
and prevalence of this pernicious difficulty. First, we enumerate a hierarchy of
proximate explanations for conflict between alternate data sets (Figure 2).
The idea of assessing congruence or incongruence of a gene tree with the
species tree presupposes the existence of at least two data sets with features
amenable to formal comparison (117). Frequently in the literature (e.g. 23,
53), however, gene trees are compared to traditional classifications based on

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 431
non-explicit interpretations of intuitively analyzed data, or concern about the
genetree’srelation to the“true history”of the taxais expressedin the absence of
any conflicting data at all. A hypothesis of relationships with limited or absent
empirical foundations is easily refuted by the first data that contradict it. Thus,
the first type of “incongruence” we identify is incongruence with non-empirical
expectations, which we arguerepresent nothing more than arbitrary, conjectural
starting points for empirical investigation. Under such circumstances, the gene
tree should be provisionally accepted as the best available hypothesis of taxo-
nomic relationship.
Traditional, informal ideas about relationships are not without value in our
methodological scheme. They can (and should) guide the discovery of charac-
ters for explicit systematic analysis. Some types of comparative data are not
amenableto discoveryof hierarchicalpatterns amongtaxa, however, and should
be ignored as sources of systematic evidence, in our opinion. These include
morphometric data (18), DNA-DNA hybridization data (35), and differential
frequencies of traits (such as allozymes) among polymorphic populations (46).
A second type of incongruence between trees can be a product of incompa-
rability of analytical methods. The same data set can yield different topologies
under alternate optimization criteria (117, 118). If two datasets are analyzed by
different methods, it is difficult to distinguish incongruence as an artifact of the
analyses from incongruence intrinsic to the data. For example, Omland (134)
showed that the topological incongruence between molecular and morphologi-
calanalyses of dabblingduck relationships was largely due to theanalysis of the
mitochondrial DNA restriction fragment data with the UPGMA method. When
both mtDNA and morphology were analyzed using the parsimony method, the
data were found to be almost entirely congruent.
One reason topologies from noncladistic analytical methods should not be
compared is mosaic incongruence (61), which results from unequal amounts
of state change in different characters, yielding either differences in the degree
of resolution or unequal branch lengths between topologies from alternate data
sets. Farris contrasted this to incompatibility incongruence due to homoplasy,
or active disagreement in hierarchical groupings implied by alternate charac-
ters (or data sets). Of course, one of the main criticisms against tree-building
algorithms based on distances, such as the UPGMA method criticized by Om-
land (134), is that they conflate mosaic and incompatibility incongruence. In
cladistic methodology, only incompatibility incongruence is relevant, because
lack of information or symplesiomorphic similarity due to lack of change in
one character has no bearing on the informativeness of another.
Type 3 incongruence is based on minor character incompatibility (homo-
plasy) discovered in separate analyses of partitioned data sets, rather than on

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
432 BROWER, DESALLE & VOGLER
strong conflict between data sets. Farris et al (63) described the incongru-
ence length difference (D), which measures the extra homoplasy entailed by
combining data sets, and a computer algorithm that tests this value against
random partitions of the same data. It is quite unlikely that two data partitions
drawnat randomfrom an identical distribution will imply completelycongruent
topologies, and thus some basic level of incongruence is expected between any
real data partitions. Empirical cases have been described from the Solanaceae
(133), primates (201), and josiine moths (119). Incongruence between the sep-
arate data sets in these cases was ignored, and the cladogram resulting from
analysisof all data was presented as the best current hypothesisof relationships.
It is notable in these cases that data partitions implying a different topology in
separateanalysis addadditional branch support (22) tothe topologyfrom simul-
taneous analysis. By contrast, some recently proposed topological congruence
methods (11, 107, 175) might interpret these results as conflicting because they
conflate character incompatibility with mosaic incongruence, lack of informa-
tion, or weak character support for conflicting nodes in alternate data partitions.
Because trees are summaries of character data, comparing them instead of the
data themselves can add no additional knowledge to the analysis and can result
in erroneous conclusions (104, 197).
As the amount of disagreement between data sets increases, attributing in-
compatibility incongruence to homoplasy (Type 3) becomes more and more
challenging to the assumption that the data reflect a single underlying hierar-
chical pattern (27). Incongruence due to inferred separate histories (Type 4)
is the only kind of incongruence that challenges the validity of the single hi-
erarchical pattern postulate. Most of the current debate over combined versus
separate analysis (see 47, 132) revolves around determining how much homo-
plasy (Type 3) we are willing to allow before invoking separate histories (Type
4). Beyond this threshold, we must decide if and by which criteria the data
should be discounted (e.g. by alternative weighting), and we must invoke ad
hoc explanations to discount homoplasy in one or another of the data sets to
meet the underlying assumption of a single hierarchy.
One possibility is to develop a general model of expected data behavior,
and statistical tests to estimate when the data are misbehaving (47, 108, 174).
How to explain, rationalize, and factor out “significant” differences discovered
by such methods remains problematical. A first step is to identify potential
processes that could systematically bias certain types of data. The next section
describes some possible reasons why different fragments of DNA sequence
from a sample of taxa might strongly imply incongruent hierarchical patterns
of relationship.

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 433
SOURCES OF CONFOUNDING INCONGRUENCE IN
MOLECULAR DATA
In the 1980s, there was a sense of optimism that DNA would be the Rosetta
Stone of systematics, allowing us empirical access to objective phylogenetic
patterns obscured in morphologies warped by adaptation (e.g. 76, 158). More
recently, Patterson et al (141) presented a more pessimistic view, that sequence
data were not contributing much to our knowledge of hierarchical relations,
except in areas where morphological analysis is not applicable, such as the
“tree of life” and patterns of relationship in human mitochondrial DNA. One
of the worst problems with DNA data is that they are noisier than we would
like, yielding cladograms with as much or more homoplasy than morphological
data sets of comparable size (156). This homoplasy is thought to be caused
by drastically different rates of evolution in characters (nucleotide sites) within
and among genes (100, 137). Thus, some sites are invariant, some sites are too
variable, and some sites contain the information we desire—the hierarchical
pattern.
As argued by Farris (62; see above), if homoplasy is random with respect to
the “phylogenetic signal,” then it doesn’t make any difference how homopla-
sious the data are: The hierarchical pattern will emerge from the noise. How-
ever, evidence exists for differential mechanistic constraints on DNA sequences
thatmay produce patternsin thedata that donot reflectthe “truephylogeny.” Nu-
cleotide bias or codon bias in protein-encoding genes varies greatly from gene
to gene and from taxon to taxon (14, 95, 159); it may result in grouping due to
convergentsimilarityratherthan putativeancestry (32,83, 162, 171). Whilenu-
cleotidebias may leadastray evolutionarymodelsbasedon neutralassumptions,
it is not a problem for cladistic analysis unless there are independent, conver-
gent origins of differential bias among the ingroup taxa. To our knowledge,
such a pattern has not yet been demonstrated among closely related taxa (e.g.
112, but see 1). Of course, saturation of variable sites in a particular sequence
may result in poor resolution among taxa, but poor resolution results in Type 3
incongruence between data sets, which should be ignored in light of other data
bearing a strong signal. The latter will predominate in simultaneous analysis.
Of more serious concern are other hypothetical evolutionary processes that
could result in a particular character (or set of characters, such as a gene) being
actively incongruent with the preponderance of other evidence (e.g. morphol-
ogy or other genes) (27). Three such processes are horizontal transfer, intro-
gression, and ancestral polymorphism. Again, we emphasize that detection
of incongruence depends fundamentally on the prior recognition of a domi-
nant pattern of relationships: If, for example, horizontal transfer of DNA were
rampant, the hierarchical pattern of diversity we observe would not exist, and

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
434 BROWER, DESALLE & VOGLER
the inference of horizontally transferred DNA fragments would be impossible.
Our ability to discover well-corroborated cladograms (and instances of homo-
plasy) implies that these phenomena are the exception rather than the norm,
and the attention paid them in the literature may reflect more the unease they
inspire in our phylogenetic paradigm than their prevalence in nature.
Horizontal Transfer
In eukaryotes, DNA generally passes only from the parents’ genome(s) to the
offspring’s genome, and it is this unidirectional pattern of inheritance that pro-
vides the basis for the inference of descent with modification. However, certain
small bits of DNA (transposable elements) are capable of transfer between lin-
eages that are not connected by interbreeding (e.g. P-elements; 44), or even,
it appears, between insect orders (e.g. mariner; 150). Transposable elements
(TEs) tend to occur in multiple copies, and they appear capable of inserting and
excising themselves from particular sites throughout the genome. The mech-
anism for transfer of TEs both within and between genomes is unknown but
may be mediated by retroviruses or other parasitic vectors (reviewed in 99).
Fortunately for hierarchical pattern inference, the 90 or so classes of TEs rec-
ognized to date appear to play an insignificant role in the disruption of taxon
boundaries. However, if horizontal transfer turns out to be a more prevalent
phenomenon than we currently believe it to be, we may need to retool our en-
tire conceptual and methodological battery for inferring patterns of relationship
(e.g. 128, 192).
Introgression
Unlikehorizontaltransfer, introgression isanastomosis ofcladesdue tointrinsic
hybridizationand differential gene flow betweenthe taxa themselves. Evidence
for introgression consists of discovery of particular “foreign” characters (e.g.
electrophoretic alleles, mitochondrial haplotypes) against a mostly “normal”
genetic or morphological background. The classic situation for discovery of
introgressionis inhybrid zones betweenformerly allopatric, closely relatedtaxa
that are capable of interbreeding (114). For many years, it was not considered
possibleto distinguish introgression betweenhistorically separated populations
from primary intergradation of divergent traits across environmental gradients
(59), but cladistic approaches (81, 190) allow distinction between the two, at
least under certain biogeographical circumstances. Introgression may be bal-
anced by selection in tension zones, resulting in a long-term, stable equilibrium
through which unlinked, neutral alleles may flow extensively (10, 183). It may
also lead to fusion of taxa that had been differentiated to some degree in the
past. The latter seems especially prevalent in taxa that have been brought into
contact by human intervention (e.g. 5, 58, 148).

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 435
Ancestral Polymorphism
Closely related taxa may share more than one allele at a locus, due to stabilizing
selection or simply to failure of neutral alleles present at the time of lineage
splitting to drift to fixation. Analysis of hierarchical patterns among such
alleles amounts to an orthology-paralogy problem (68, 75, 136, 140). If we are
interested in inferring relationships among taxa, it is necessary to distinguish
the splitting event that gave rise to the alternate alleles (which may represent
the deepest node in the topology), from the splitting event that gave rise to the
taxa. Of course, these alternate alleles can become fixed in the taxa via lineage
sorting (127), yielding hierarchical patterns that do not reflect the course of
taxon splitting. Such patterns may be further complicated at autosomal loci by
recombination,which canresult in a mosaic pattern of relativedegreeof identity
by descent among alleles (89). A practical recommendation for molecular
systematics is to avoid genes that are suspected of being subject to balancing
selection, such as genes of the major histocompatibility complex (MHC) (e.g.
DQB1; 71).
Doyle (57) elegantly discussed the related problem of heterozygosity in
diploid loci: Two alleles occurring in a single individual need not be closely
relatedin the hierarchyof relationships foralleles at that locus. The nonsensical
outcome of drawing systematic inferences from such data is that, depending on
which allele is sequenced, the same organism will occupy different positions
in the gene tree. Doyle recognized that systematic methods are inappropriate
under such conditions, and proposed a method akin to population aggregation
analysis (46) that delimits the boundaries of the gene pool by discovering all
alleles cohabiting with other alleles in heterozygote combinations among in-
dividuals. This procedure is tantamount to an empirical investigation of the
boundaries of biological species (114, 115), and although rigorous, it is not
likely to provide a feasible solution to most systematic problems at or around
the species level.
This section has briefly described some genetic processes that could cause
genetrees tobe misleadingrelativeto speciestrees. However,anyparsimonious
explanation of data could be wrong, and speculating in the realm of the possible
doesnot advanceourknowledge veryfar. Inorderto knowhowmuch weshould
trust our molecular trees, we wouldlike to know how often incongruenceoccurs
betweena particulargene tree andtrees basedon other data from thesame taxa.3
3Recall that we are not concerned about the correspondence of a particular result with “truth,”
but only with other empirical results. Suggesting that one data set is false because it is incongruent
with another that is true both opens the door for Ockham’s genius malignus (how do we know that
all the data aren’t misleading?) and begs the point of further inquiry (149; see “Epistemology and
Truth,” above).

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
436 BROWER, DESALLE & VOGLER
In the next two sections, we examine theoretical and empirical results based
on hierarchical pattern analysis of mitochondrial DNA in order to address two
questions, “Are gene trees theoretically likelyto be misleading?” and “Are these
theoretical concerns reflected by empirical results?”
THEORETICAL APPROACHES TO GENE TREES AND
SPECIES TREES: THE MTDNA PARADIGM
MtDNA has well-known advantages as a source of characters for systematics
and biogeography (6, 15, 80, 124). It is usually maternally inherited (effec-
tively haploid), lacks recombination, evolves relatively rapidly, and segregates
independently from the nuclear genome (but see 202). These features have
made mtDNA a tractable subject for genetic models as well. Birky et al (16)
pointed out that the effective population size (Ne) of mtDNA is one fourth that
of an autosomal locus in a random-mating diploid population. This means that
fixation due to genetic drift occurs four times more rapidly among selectively
neutral mitochondrial haplotypes than among neutral nuclear alleles, and that
gene flow sufficient to maintain nuclear panmixia may allow differentiation of
mitochondrial lineages in different local demes. These effects are amplified if
dispersal is male-biased. Under ideal circumstances, therefore, mtDNA would
seem to be a superior marker for studying relations among closely related taxa
(6, 123, but see 42, 116).
However, because mtDNA does not undergo the anastomosis of recombining
nuclear loci, unambiguous yet spurious hierarchical relationships exist among
mtDNA haplotypes of individual organisms within interbreeding populations.
Ifpatterns of haplotyperelationship are usedto infer patternsof taxonomic rela-
tionship, it is critical that these patterns agree: Hierarchical structure must exist
at both levels (46). As we noted at the outset, if there is no hierarchical structure
among terminal taxa being compared, systematic methods are inappropriate for
inferring their relations.
Even when hierarchical structure exists among the taxa, an extensive body of
theory suggests that hierarchical patterns from mtDNA (or any other loci) may
not reflect the historical order of divergence events among the examined taxa.
Introgression of unlinked, neutral alleles across hybrid zones is both likely and
potentially far-reaching (183), even if the zone is maintained by quite strong
selective forces acting on nuclear loci. Ancestral polymorphism of mtDNA in
populations is also theoretically likely (but not, of course, heterozygosity of
individuals). Tajima (178) showed that, even for very simple situations such
as two neutral alleles drawn from each of two populations, the probability of
inferring the correct tree is small if the time since the branching event is less

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 437
than 2Ne generations. This is because the mean time to fixation (monophyly)
of neutral alleles in each of the populations is 4Ne (Ne for mtDNA) under
neutral expectations (101), and the daughter populations are therefore likely to
contain alleles older than the population-splitting event. Avise and colleagues
(7, 127) extended this theory to predict patterns of mtDNA evolution under
various demographic scenarios. Pamilo & Nei (138) used genetic distances
to show that increasing the number of alleles sampled did not help solve this
problem.
A neutral model of the genealogical process called coalescent theory has
been increasingly important to incorporating hierarchical information into the
interpretation of genetic diversity (55, 89, 91, 102). Coalescent theory relates
the age of a gene clade to its degree of genetic diversity, given a stochastic
mutation process and temporal constancy of a series of factors that can affect
the rate of diversification (or, looking backwards, coalescence). These fac-
tors include population size, selection, linkage and recombination, geographic
structure, and migration between populations. Takahata (181) used coalescent
theory to argue that larger sample sizes of alleles could increase the probability
of inferring the correct historicalpattern, contrary to Pamilo &Nei’s(138) view.
Recently, Moore (123) used a coalescent argument to suggest that mtDNA is
a substantially more reliable marker, based on its smaller Ne, for estimating
phylogenetic patterns than are nuclear genes.
Templeton et al (188) and Crandall (37) used elements of coalescent theory
to develop a novel cladogram estimation procedure. The Templeton et al algo-
rithm(188)uses astatisticalapproach basedonHudson’s(89) neutral coalescent
model to establish the probability of obtaining a “non-parsimonious” inference
from RFLP data. This “statistical parsimony” procedure determines the prob-
ability of genealogical connections in haplotype networks being parsimonious
(sensu 37, 89) by providing a measure of the probability that connections be-
tween two haplotypes differing by some number of steps are nonhomoplasious.
We stress that coalescent methods so far developed are for single genes (see
for instance 89, 186–188) or for linked clusters of genes such as mtDNA or
small phage genomes (see 123 and 37, respectively). Crandall & Templeton
(38) tested several predictions of coalescent theory using many empirical data
sets from both mtDNA and nuclear DNA. Their tests suggest that Templeton’s
approachto coalescent theoryrepresents avaluable innovationfor studyof gene
genealogies and may offer insights to the various causal factors responsible for
the pattern observed.
Coalescentmodels havealso beenused to arguethat certainmarkers are more
likely than others to recover the species genealogy because of their population
genetics (e.g. 123) and to infer causes of organismic geographical distribution

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
438 BROWER, DESALLE & VOGLER
(189). Crandall (37) examined the performance of the algorithm, using a sub-
samplingtechnique onT7 bacteriophagedata, and showedthatit performs more
efficiently than maximum parsimony at recovering the “true” phylogeny when
very few restriction sites are available for analysis. This led him to suggest that
the choice of discovery method should depend on the level of diversity at which
a gene genealogy is being constructed and the amount of character information
available.
Such extrapolations of coalescent theory are only as plausible as the assump-
tionsunderlying theparticularmodel applied. Mostcommon coalescentmodels
depend upon constant historical population size, lack of selection, panmixia,
and several other assumptions (90, 163). The chance that these assumptions
are met by any real data set is difficult to assess but is probably quite small.
Simulations of the effects of altering single parameters show that current or
historical selection on the gene or a gene it is linked to (98, 113, 182), recom-
bination (84, 90), and fluctuations in population size (126, 151, 179) can each
have major impacts on allelic diversity and times to coalescence. If two or more
of these “constants” are not constant, then it becomes difficult to infer which
process is responsible for observed patterns (180). Unfortunately, actual data
from natural populations often appear to violate at least some of the underly-
ing assumptions of coalescent theory (9, 31, 123, 126), throwing in doubt the
explanatory relevance of the models in those systems.
It is arguable, in fact, that neutral models of allele genealogy must be unre-
alistic. As noted above, if taxa diverge over periods less than 4Ne generations,
then the probability of any particular gene tree reflecting the population history
is low (138, 154). Under such circumstances, most genes will have allele ge-
nealogies that are incongruent with one another and do not reflect the “species
phylogeny.” If this is what actually happens, it is difficult for us to imagine how
divergent evolution can occur, unless it is driven by a special subset of “specia-
tion genes” that drive phylogenesis in the face of homoplasious evolution of the
majority of the genome. If this were the case, studying those genes alone might
providethe keytodiscovering thetrue courseof phylogenetichistory. However,
evidence supporting this proposition is decidedly sparse (but see 33). A more
plausible hypothesis is that Ne is often small at speciation (due to selection or
any number of population subdivision scenarios), and thus all genes are likely
to track population history to someextent, unless times between splitting events
have been very short (in which case, we do not expect to see much character
support for any topology; 24).
In any event, determining regularities of population splitting should start
from an empirical framework: Theoretical speculation, even in the guise of
elegant mathematics, does not advance our knowledge unless it is tested with

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 439
actual data. The T7 studies (37) offer a beginning, but it is notable that while
parsimony recovered the “true” tree with strong resolution from all the data,
neither method performed well on the tiny subsets of data employed in this
comparative test. Further, generalizations drawn from contrived experimental
phylogenies are bound to be oversimplified, and whether they tell us anything
at all about gene trees from more complex taxa under “real” circumstances is
subject to debate (87, 168).
EMPIRICAL APPROACHES TO GENE TREES AND
SPECIES TREES: EVIDENCE FROM MTDNA
To exhaustively review empirical evidence for and against congruence between
mtDNA trees and trees from other data would fill a book (see 4). There are
many cases showing that molecules are congruent with morphology in closely
related taxa (e.g. 3, 8, 23, 25, 119, 125, 184, 191). Nevertheless, the caveats in
quite a few papers list the same, rather small, group of publications as evidence
that gene trees are not congruent with species trees, so we examine some of
these here.4
In a widely cited paper, Powell (145) examined mtDNA RFLP data from
sympatric and allopatric populations of Drosophila pseudoobscura and D. per-
similisin thewestern UnitedStates, Mexico, and Colombia. His title,“Interspe-
cific cytoplasmic gene flow in the absence of nuclear gene flow: evidence from
Drosophila,” suggested that he had observed introgression of mtDNA across
the species boundary. However, this interpretation is based on mosaic (Type 2)
incongruence (61), rather than compatibility (Type 4) incongruence (indeed, no
cladogram or network was presented). Later (146), Powell suggested that the
mtDNAparaphylydue tolineage sorting isan equallyparsimonious explanation
of the observed pattern. Another plausible alternative is that populations of D.
pseudoobscurahave noapomorphies with respect to D. persimilis, a conclusion
that is unproblematic for the cladistic view of species (46, 85, 131). A more
extensive study of mtDNA (79), as well as evidence from RFLP analyses of
the ADH region of the fourth chromosome (157) and the AMY region of the
third chromosome (2), support the species-level paraphyly of D.pseudoobscura
populations with respect to D. persimilis.
Astrongly analogous situationwas described in desert tortoises (106). In this
case, lineage sorting was invoked to explain the apparent paraphyly of mtDNA
amongindividuals of Xerobatesagassizi with respect to X. berlandieri mtDNA,
4Ironically, Avise & Saunders (8) is frequently cited as a case exhibiting mtDNA introgression,
in spite of the authors’ unambiguous statement (p. 252): “in the present study we have no mtDNA
(or allozyme) evidence for introgression between species of Lepomis.”

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
440 BROWER, DESALLE & VOGLER
yet the morphological authority cited (21) viewed the latter as a peripheral
isolate of the former, implying that other characters also support the pattern
observed in the mtDNA. This is another example of Type 2 incongruence.
Another case of apparent introgression of mtDNA was reported between
white-tailed deer (Odocoileus virginianus) and mule deer (0. hemionus) (28).
Mule deer are traditionally held to be conspecific with black-tailed deer, but the
mule deer mtDNA in this study differed from that of black-tailed deer by 17
restriction sites, while sharing haplotypes with white-tailed deer. Further study
(39, 41, 43) corroborated the affinity of mtDNA between the two species. It
is interesting that mtDNA data (39) suggest that the black-tailed deer is more
closely related to the neotropical red brocket (Mazama americana) than to the
other Odocoileus deer. By contrast, allozymes (40, 50, 72) show mule and
black-tailed deer to be phenetically more similar to one another than either is to
white-taileddeer. Allthese papersattributed thediscrepancy betweenthe mark-
ers to historical mtDNA introgression from white-tailed deer to mule deer. If
thisis thecase, itis surprisingthat hybridzones betweenthe putative conspecific
taxa (mule and black-tailed deer) reveal no introgression of mtDNA (41), while
interspecific hybridization has evidently resulted in complete replacement of
black-tail–like mtDNA with white-tail–like mtDNA in the mule deer. Cronin
(42) criticized the reliability of mtDNA for study of closely related taxa, in
general, based on these results. However, granting the validity of phenetically
analyzed allozyme frequency trees, the reported patterns can again be equally
parsimoniously explained by mosaic incongruence (Type 2), because none of
the analyses are rooted, except by the clock-dependent long-branch criterion.
In fact, three-taxon statements do not allow inference of relationships at all,
unless a root is provided. Thus, the white-tailed deer and mule deer may be
sister taxa relative to black-tailed deer, as implied by mtDNA.
Of course, some cases of character incongruence between mtDNA and other
data do seem most easily explained by mtDNA lineage sorting, introgression,
or some other process. For example, directional introgression of mtDNA from
Mus domesticus to M. musculus appears to have occurred across a hybrid zone
in Denmark (67). Gyllensten & Wilson (78) suggested that the observed pat-
tern may reflect the colonization of Sweden by a small number of hybrid mice
from another hybrid zone in north Germany, in conjunction with the spread of
agriculture around 4000 years ago. Although the amount of genetic divergence
between the taxa suggests that M. domesticus and M. musculus diverged at least
a million years ago (67), Danish and Swedish house mouse populations must
have immigrated relatively recently, since the region was under ice during the
Pleistocene. As noted above, observed patterns of genetic introgression are fre-
quently associated with recent, human-mediated dispersal events. Peripatetic

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 441
commensal organisms like house mice and Drosophila melanogaster may be
uniquely poor study taxa for phylogeographic analysis, in spite of their popu-
larity in genetic research.
Spolsky & Uzzell (169, 170) documented the presence of Rana lessonae
mtDNA in Rana ridibunda populations from eastern Europe, suggesting intro-
gressive hybridization. They hypothesized that the mtDNA had leapfrogged
the species boundary via R. esculenta, a hybridogenetic species that eliminates
the R. lessonae component of its nuclear genome during gametogenesis. The
unusual genetics of these species may account for the dramatic discordance of
mtDNA with other markers: Other hybridizing European amphibians, such as
Bombina bombina and B. variegate (177), and Triturus cristatus and T. manno-
ratus (3), exhibit largely congruent patterns of introgression between nuclear
and mtDNA.
Another interesting case occurs in Sri Lankan macaques (Macaca sinica).
Based on RFLP analysis, Hoelzer et al (88) found mtDNA sequence divergence
of over 3% between adjacent social groups in a local population that, based on
allozyme samples, is otherwise apparently genetically panmictic. The strong
differential between these character systems may be maintained by sex-biased
gene flow between troops. Female macaques are mostly philopatric, while
males typically disperse more readily. This behavioral difference seems un-
likely to account for the high degree of differentiation, which suggests the lin-
eages have been separate for 1.5 million years, if the standard clock calibration
applies (26). Based on morphology, Fooden (69) recognized two subspecies
of M. sinica in Sri Lanka, and the contact zone between them evidently falls
quite near the mtDNA study site. The sampled population could thus lie on a
secondary contact zone between two formerly separate groups. Allozyme data
(160), however, suggest that the nuclear alleles are not differentiated strongly
between the subspecies.
Perhaps the most extensively documented example of mtDNA incongru-
ence is drawn from the closely related Hawaiian species quartet of Drosophila
silvestris, D. heteroneura, D. planitibia and D. differens. These four taxa have
been the target of intense morphological, behavioral, and genetic research over
the past 30 years (reviewed in 30, 52, 97). There are no fixed chromosomal or
allozyme differences among the four species, and morphological differences
among them are mostly autapomorphies (29, 36). Using RFLP, DeSalle &
Giddings (51) found the mtDNA pattern of relationship for these taxa (d(p(s,
h))) to be incongruent with the topology implied by dendrograms from DNA-
DNA hybridization (93) and allozymes (36) ((d,p)(s,h)).
DeSalle & Giddings (51; p. 6906) argued that, “the mtDNA phylogeny rep-
resents the more accurate evolutionary history of these species, and the nuclear

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
442 BROWER, DESALLE & VOGLER
DNA phylogeny is indicative of a lack of differentiation of nuclear genetic
components of D. differens and D. planitibia.” Under the criteria laid out in
Figure 2, this was an appropriate conclusion, since the alternative morpholog-
ical, karyotypic, allozyme, and DNA-DNA hybridization data sets each fail at
Stage 1 or 2, in that they contain no cladistic information or are intrinsically
incompatible with the hierarchical pattern discovery operation. However, sub-
sequent sequence data from nuclear genes (ADH, 152; vitellogenin yp-1, 96)
support the ((d,p)(s,h)) topology. Since these data are comparable by cladistic
methods, the data from the various genes may be formally compared for active
(Type 4) incongruence. Using the incongruence length difference test (63), we
find that the two nuclear genes are completely congruent (D =0.0). However,
the mtDNA data are strongly incongruent with the nuclear sequences (ILD with
ADH =0.256; ILD with yp-1 =0.186).
How are these incongruent data sets best interpreted? The weight of the
evidence supports the nuclear gene topology ((d,p)(s,h)), so the position that
the mtDNA pattern is superior must be abandoned, regardless of population
genetic arguments that can be made concerning the higher relevance of the
maternal marker (51, 123). This conclusion does not mean that the current
answer is correct in the ontological sense. Neither does it imply that it is not
interesting to investigate causes for the observed pattern of incongruence. It
simply means that ((d,p)(s,h)) is the best current estimate of the relationships
among these taxa.
CONCLUSION
Should we expect gene trees to conflict with species trees? In our view, any
expectation that comparative biologists bring to their data must be based on
relevant empirical results. First, there must be some reason to hypothesize
the existence of hierarchical relationships among the study taxa. If these do
not exist, using a hierarchy-based discovery technique is simply inappropri-
ate. If we believe there is a hierarchical pattern, then we should expect the
data to reflect that pattern alone. The relevant characters should be evaluated
parsimoniously and interpreted without prejudice derived from false notions
of incongruence (Types 1, 2, and 3) based on process theories. If the data do
not exhibit incongruence in the first place, ad hoc theories explaining incon-
gruence are unwarranted. Topological comparison of gene trees to traditional
classifications or trees based on ill-conceived or inappropriately analyzed data
is likewise an inappropriate measure of “incongruence.” Only character incon-
gruence (Type 4) is problematical for the hierarchy paradigm. Even when
such incongruence is discovered, the best estimate of hierarchical relation-
ships is still derived from parsimonious interpretation of all the data, analyzed

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 443
simultaneously, unless compelling rationalizations are provided for weighting
or otherwise discrediting some of the evidence in favor of the remainder of the
evidence. Explaining incongruence is not a task of systematics.
ACKNOWLEDGMENTS
Wethank J Gatesy,PGoldstein, HRosenbaum, V Schawaroch,BI Vane-Wright,
D yeats, A. Burt, and our AMNH discussiongroup for helpful comments on our
manuscript, and A de Queiroz for providing a preprint of his ARES manuscript.
A Brower was supported by NSF DEB-9303251 and BSR-9106517.
Any Annual Review chapter, as well as any article cited in an Annual Review chapter,
may be purchased from the Annual Reviews Preprints and Reprints service.
1-800-347-8007; 415-259-5017; email: arpr@class.org Visit
the Annual Reviews home page at
http://www.annurev.org.
Literature Cited
1. Anderson CL, Carew EA, Powell JR.
1993. Evolution of the Adh locus in the
Drosophila willistoni group: the loss of
an intron, and shift in codon usage. Mol.
Biol. Evol. 10:605–18
2. Aquadro CF, Weaver AL, Schaeffer SW,
Anderson WW. 1991. Molecular evolu-
tion of inversions in Drosophila pseu-
doobscura: the amylase gene region.
Proc. Natl. Acad. Sci. USA 88:305–
9
3. Arntzen JW,Wallis GP. 1991. Restricted
geneflow in a movinghybrid zone of the
newtsTrituruscristatus and T.marmora-
tusinwesternFrance.Evolution45:805–
26
4. Avise JC. 1994. Molecular Markers,
Natural History and Evolution. New
York: Chapman & Hall
5. Avise JC, Ankney CD, Nelson WS.
1990. Mitochondrial gene trees and the
evolutionary relationship of mallard and
black ducks. Evolution 44:1109–19
6. Avise JC, Arnold J, Ball RM, Berming-
ham E, Lamb T, et al. 1987. Intraspe-
cific phylogeography: the mitochon-
drial DNA bridge between population
genetics and systematics. Annu. Rev.
Ecol. Syst. 18:489–522
7. Avise JC, Neigel JE, Arnold J. 1984.
Demographic influences on mitochon-
drial DNA lineage survivorship in an-
imal populations. J. Mol. Evol. 20:99–
105
8. Avise JC, Saunders NC. 1984. Hy-
bridization and introgression among
species of sunfish (Lepomis): analysis
by mitochondrial DNA and allozyme
markers. Genetics 108:237–55
9. Ballard JWO, Kreitman M. 1995. Is
mitochondrial DNA a strictly neutral
marker? Trends Ecol. Evol. 10:485–88
10. Barton NH, Hewitt GM. 1985. Analysis
of hybrid zones. Annu. Rev. Ecol. Syst.
16:113–48
11. Baum BR. 1992. Combining trees as a
way of combining data sets for phylo-
genetic inference, and the desirability of
combining gene trees. Taxon 41:3–10
12. Baum DA, Donoghue MJ. 1995. Choos-
ing among alternative “phylogenetic”
species concepts. Syst. Bot. 20:560–73
13. Baum DA, Shaw KL. 1995. Genealog-
ical perspectives on the species prob-
lem.In Experimental and Molecular Ap-
proachesto Plant Biosystematics, ed.PC
Hoch, AG Stephenson, pp. 289–303. St.
Louis: Mo. Bot. Gard.
14. BernardiG,BernardiG. 1986. Composi-
tionalconstraints and genomeevolution.
J. Mol. Evol. 24:1–11
15. Birky CW Jr. 1991. Evolution and pop-
ulation genetics of organelle genes:
mechanismsand models. InEvolution at
the Molecular Level, ed. RK Selander,
AG Clark, TS Whittam, pp. 112–34.
Sunderland, MA: Sinauer
16. Birky CW Jr, Maruyama T, Fuerst P.

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
444 BROWER, DESALLE & VOGLER
1983. An approach to population and
evolutionary genetic theory for genes
in mitochondria and chloroplasts, and
some results. Genetics 103:513–27
17. Blyth E. 1836. Observations on the vari-
ous seasonal and other external changes
which regularly take placein birds, more
particularly in those which occur in
Britain; with remarks on their great im-
portance in indicating the true affinities
of species; and upon the natural system
of arrangement. Mag. Nat. Hist. 9:393–
409, 505–14
18. Bookstein FL. 1994. Can biometrical
shape be a homologous character? In
Homology: The Hierarchical Basis of
Comparative Biology, ed. BK Hall, pp.
197–227. San Diego: Academic
19. Bowler PJ. 1989. Evolution, the History
of an Idea. Berkeley: Univ. Calif. Press.
2nd ed.
20. Brady RH. 1985. On the independence
of systematics. Cladistics 1:113–26
21. Bramble DM. 1971. Functional mor-
phology, evolution and paleoecology
of gopher tortoises. PhD. Diss., Univ.
Calif., Berkeley
22. BremerK.1994.Branchsupportandtree
stability. Cladistics 10:295–304
23. Brower AVZ. 1994. Phylogeny of He-
liconius butterflies inferred from mito-
chondrialDNAsequences(Lepidoptera:
Nymphalidae). Mol. Phylogenet. Evol.
3:159–74
24. Brower AVZ, DeSalle R. 1994. Prac-
tical and theoretical considerations for
choice of a DNA sequence region in in-
sect molecular systematics, with a short
review of published studies using nu-
clear gene regions. Ann. Entomol. Soc.
Am. 87:702–16
25. Brown JM, Pellmyr O, Thompson JN,
Harrison RG. 1994. Phylogeny of Greya
(Lepidoptera: Prodoxidae), based on
nucleotide sequence variation in mito-
chondrial cytochrome oxidase I and II:
congruence with morphological data.
Mol. Biol. Evol. 11:128–41
26. Brown WM, George M Jr, Wilson AC.
1979. Rapid evolution of animal mito-
chondrial DNA. Proc. Natl. Acad. Sci.
USA 76:1967–71
27. Bull JJ, Huelsenbeck JP, Cunningham
CW, Swofford DL, Waddell PJ. 1993.
Partitioning and combining data in phy-
logeneticanalysis. Syst. Biol.42:384–97
28. Carr SM, Ballinger SW, Derr JN,
Blankenship LH, Bickham JW. 1986.
Mitochondrial DNA analysis of hy-
bridization between sympatric white-
tailed deer and mule deer in west Texas.
Proc. Natl. Acad. Sci. USA 83:9576–80
29. Carson HL. 1983. Evolution of Droso-
phila on the newer Hawaiian volcanoes.
Heredity 48:3–25
30. Carson HL. 1987. Tracing ancestry with
chromosomal sequences. Trends Ecol.
Evol. 2:203–7
31. Choudhary M, Singh RS. 1987. His-
torical effective size and the level
of genetic diversity in Drosophila
melanogaster and Drosophila pseu-
doobscura. Biochem. Genet. 25:41–51
32. CollinsTM,KrausF,Estabrook G. 1994.
Compositional effects and weighting of
nucleotide sequences for phylogenetic
analysis. Syst. Biol. 43:449–59
33. Coyne JA. 1992. Genetics and specia-
tion. Nature 355:511–15
34. Cracraft J. 1983. Species concepts and
speciationanalysis.In CurrentOrnithol-
ogy, ed. RF Johnston, 1:159–87. New
York: Plenum
35. Cracraft J. 1987. DNA hybridization in
avian systematics. In Evolutionary Bi-
ology, ed. MK Hecht, B Wallace, GT
Prance, 21:47–96. New York: Plenum
36. Craddock EM, Johnson WE. 1979. Ge-
netic variation in Hawaiian Drosophila.
5.Chromosomaland allozymic diversity
in Drosophila silvestris and its homose-
quential species. Evolution 33:137–55
37. Crandall KA. 1994. Intraspecific clado-
gramestimation: accuracy at higherlev-
els of divergence. Syst. Biol. 43:222–35
38. CrandallKA, TempletonAR. 1993. Em-
pirical tests of some predictions from
coalescent theory with applications to
intraspecific phylogeny reconstruction.
Genetics 134:959–69
39. Cronin MA. 199l. Mitochondrial DNA
phylogeny of deer (Cervidae). J. Mam-
mal. 72:553–66
40. Cronin MA. 1991. Mitochondrial DNA
and nuclear genetic relationships in
deer (Odocoileus spp.) in western North
America. Can. J. Zool. 69:1270–79
41. Cronin MA. 1992. Intraspecific varia-
tion in mitochondrial DNA of North
American cervids. J. Mammal. 73:70–
82
42. Cronin MA. 1993. Mitochondrial DNA
in wildlife taxonomy and conservation
biology: cautionary notes. Wildl. Soc.
Bull.. 21:339–48
43. Cronin MA, Vyse ER, Cameron DG.
1988. Genetic relationships between
mule deer and white-tailed deer in Mon-
tana. J. Wildl. Manage. 52:320–28
44. Daniels SB, Peterson KR, Strausbaugh

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 445
LD, Kidwell MG, Chovnick A. 1990.
Evidence for horizontal transmission
of the Ptransposable element between
Drosophila species. Genetics 124:339–
55
45. Darwin C. 1859. On the Origin of
Species. London: John Murray. (Fac-
simile of 1st ed., 1964. Harvard Univ.
Press)
46. Davis JI, Nixon KC. 1992. Popula-
tions,geneticvariation,andthedelimita-
tion of phylogenetic species. Syst. Biol.
41:421–35
47. de Queiroz A, Donoghue MJ, Kim J.
1995. Separate versus combined analy-
sis of phylogenetic evidence. Annu. Rev.
Ecol. Syst. 26:657–81
48. deQueirozK, Donoghue MJ. 1988. Phy-
logenetic systematics and the species
problem. Cladistics 4:317–38
49. deQueirozK, Donoghue MJ. 1990. Phy-
logenetic systematics or Nelson’s ver-
sion of cladistics? Cladistics 6:61–75
50. Derr JN. 1991. Genetic interactions be-
tweenwhite-tailed deer andmule deer in
the southwestern United States. J. Wildl.
Manage. 55:228–37
51. DeSalle R, Gaddings LV. 1986. Dis-
cordance of nuclear and mitochon-
drial DNA phylogenies in Hawaiian
Drosophila. Proc. Natl. Acad. Sci. USA
83:6902–6
52. DeSalle R, Hunt JA. 1987. Molecu-
lar evolution in Hawaiian drosophilids.
Trends Ecol. Evol. 2:212–16
53. Disotell TR. 1994. Generic level rela-
tionships of the Papionini (Cercopithe-
coidea). Am. J. Phys. Anthropol. 94:47–
57
54. Dobzhansky T. 1937. Genetics and the
Origin of Species. New York: Columbia
Univ. Press
55. Donnelly P, Tavar´e S. 1995. Coalescents
and genealogical structure under neu-
trality. Annu. Rev. Genet. 29:401–21
56. Doyle JJ. 1992. Gene trees, and species
trees: molecular systematics as a 1-
character taxonomy. Syst. Bot. 17:144–
63
57. Doyle JJ. 1995. The irrelevance of allele
tree topologies for species delimitation,
and a non-topological alternative. Syst.
Bot. 20:574–88
58. Echelle AA, Connor PJ. 1989. Rapid,
geographically extensive genetic intro-
gression after secondary contact be-
tween two pupfish species (Cyprinodon,
Cyprinodontidae). Evolution 43:717–27
59. Endler JA. 1977. Geographic Variation,
Speciation, and Clines. Princeton, NJ:
Princeton Univ. Press
60. ExcoffierL, Smouse PE. 1994.Using al-
lele frequencies and geographic subdi-
vision to reconstruct gene trees within a
species: molecular variance parsimony.
Genetics 136:343–59
61. Farris JS. 1971. The hypothesis of non-
specificity and taxonomic congruence.
Annu. Rev. Ecol. Syst. 2:277–302
62. Farris JS. 1983. The logical basis of
phylogenetic analysis. In Advances in
Cladistics, ed. NI Platnick, VA Funk,
2:7–36. New York: Columbia Univ.
Press
63. Farris JS, K¨allersj¨o M, Kluge AG, Bult
C. 1994. Testing significance of congru-
ence. Cladistics 10:315–19
64. Felsenstein J. 1982. Numerical methods
for inferring evolutionary trees. Q. Rev.
Biol. 57:379–404
65. Felsenstein J. 1988. Phylogenies from
molecular sequences: inferences and re-
liability. Annu. Rev. Genet. 22:521–65
66. Felsenstein J, Churchill GA. 1996. A
hidden Markov model approach to vari-
ation among sites in rate of evolution.
Mol. Biol. Evol. 13:93–104
67. Ferris SD, Sage RD, Huang C-M,
Nielsen JT, Ritte U, Wilson AC. 1983.
Flow of mitochondrial DNA across a
species boundary. Proc. Natl. Acad. Sci.
USA 80:2290–94
68. Fitch WM. 1970. Distinguishing homol-
ogous from analogous proteins. Syst.
Zool. 19:99–113
69. Fooden J. 1979. Taxonomy and evo-
lution of sinica group of macaques:
I. Species and subspecies accounts of
Macaca sinica. Primates 20:109–40
70. Frost DR, Kluge AG. 1994. A consider-
ation of epistemology in systematic bi-
ology, with special reference to species.
Cladistics 10:259–94
71. Gaur LK, Hughes AL, Heise ER,
Gutknecht J. 1992. Maintenance of
DQB1 polymorphisms in primates. Mol.
Biol. Evol. 9:599–609
72. Gavin TA, May B. 1991. Taxonomic
status and genetic purity of Columbian
white-tailed deer. J. Wildl. Manage.
52:1–10
73. Gillespie JH. 1991. The Causes of
Molecular Evolution. Oxford: Oxford
Univ. Press
74. Goldman N. 1993. Statistical tests of
models of DNA substitution. J. Mol.
Evol. 36:182–98
75. Goodman M, Czelusniak J, Moore GW,
Romero-Herrera AE, Matsuda G. 1979.
Fitting the gene lineage into its species

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
446 BROWER, DESALLE & VOGLER
lineage, a parsimony strategy illustrated
by cladograms constructed from globin
sequences. Syst. Zool. 28:132–63
76. Gould SJ. 1985. A clock of evolution:
we finally have a method for sorting out
homologies from “subtle as subtle can
be” analogies. Nat. Hist. 94:12–25
77. GraybealA.1995. Naming species. Syst.
Biol. 44:237–50
78. Gyllensten U, Wilson AC. 1987. Inter-
specificmitochondrialDNAtransferand
thecolonization of Scandinavia by mice.
Genet. Res. 49:25–29
79. Hale LR, Beckenbach AT. 1985. Mito-
chondrial DNA variation in Drosophila
pseudoobscura and related species in
Pacific Northwest populations. Can. J.
Genet. Cytol. 27:357–64
80. Harrison RG. 1989. Animal mitochon-
drial DNA as a genetic marker in popu-
lation and evolutionary biology. Trends
Ecol. Evol. 4:6–11
81. Harrison RG. 1990. Hybrid zones: win-
dows on evolutionary process. In Oxford
Surveys in Evolutions Biology, ed. D Fu-
tuyma, J Antonovics, 7:69–128. Oxford:
Oxford Univ. Press
82. Harrison RG. 1991. Molecular changes
at speciation. Annu. Rev. Ecol. Syst.
22:281–308
83. Hasegawa M, Hashimoto T. 1993. Ribo-
somal RNA trees misleading? Nature
361:23
84. Hein J. 1990. Reconstructing evolu-
tion of sequences subject to recombina-
tion using parsimony. J. Math. Biosci.
98:185–200
85. Hennig W. 1966. Phylogenetic System-
atics. Urbana: Univ. Ill. Press. 263 pp.
86. Hillis DM. 1987. Molecules versus mor-
phological approaches to systematics.
Annu. Rev. Ecol. Syst. 18:23–42
87. Hillis DM, Bull JJ, White ME, Bad-
gettMR,MolineuxIJ.1993. Experimen-
tal approaches to phylogenetic analysis.
Syst. Biol. 42:90–92
88. Hoelzer GA, Dittus WPJ, Ashley MV,
MelnickDJ. 1994. The localdistribution
of highly divergent mitochondrial DNA
haplotypes in toque macaques Macaca
sinica at Poponnaruwa, Sri Lanka. Mol.
Ecol. 3:451–58
88a. HoelzerGA, Melnick DJ.1994. Patterns
of speciation and limits to phylogenetic
resolution. Trends Ecol. Evol. 9:104–7
89. Hudson RR. 1990. Gene genealogies
and the coalescent process. In Oxford
Surveys in Evolutionary Biology, ed. D
Futuyma,J Antonovics, 7:1–44. Oxford:
Oxford Univ. Press
90. Hudson RR. 1993. The how and why of
generating gene genealogies. In Mech-
anisms of Molecular Evolution, ed. N
Takahata, AG Clark, pp. 23–36. Sunder-
land, MA: Sinauer
91. Hudson RR, Kreitman M, Aguad´eM.
1987. A test of neutral molecular evolu-
tion based on nucleotide data. Genetics
116:153–59
92. Huelsenbeck JP. 1995. Performance of
phylogenetic methods in simulation.
Syst. Biol. 44:17–48
93. HuntJA,Carson HL. 1982.Evolutionary
relationshipsoffourspeciesof Hawaiian
Drosophila as measured by DNA reas-
sociation. Genetics 104:353–64
94. Huxley TH. 1874. On the classification
oftheanimal kingdom. Nature 11:101–2
95. Ikemura T. 1985. Codon usage and
tRNA content in unicellular and mul-
ticellular organisms. Mol. Biol. Evol.
2:13–34
96. Kambysellis MP, Ho K-F, Craddock
EM, Piano F, Parisi M, Cohen J. 1995.
Patternofecologicalshifts in the diversi-
ficationofHawaiianDrosophilainferred
from a molecular phylogeny. Curr. Biol.
5:1129–39
97. Kaneshiro KY, Boake CRB. 1987. Sex-
ual selection and speciation: issues
raised by Hawaiian Drosophila. Trends
Ecol. Evol. 2:207–12
98. Kaplan NL, Hudson RR, Langley CH.
1989. The “hitchhiking effect” revisited.
Genetics 123:887–99
99. Kidwell MG. 1993. Lateral transfer in
naturalpopulations of eukaryotes. Annu.
Rev. Genet. 27:235–56
100. Kimura M. 1983. The Neutral Theory of
MolecularEvolution. Cambridge: Cam-
bridge Univ. Press
101. Kimura M, Ohta T. 1969. The average
number of generations until fixation of a
mutant gene in a finite population. Ge-
netics 61:763–71
102. Kingman JFC. 1982. On the genealogy
of large populations. J. Appl. Probab. A
19:27–43
103. Kishino H, Miyata T, Hasegawa M.
1990. Maximum likelihood inference
in protein phylogeny and the origin of
chloroplasts. J. Mol. Evol. 21:151–60
104. Kluge AJ. 1989. A concern for evidence
and a phylogenetic hypothesis of rela-
tionshipsamong Epicrates (Boidae, Ser-
pentes). Syst. Zool. 38:7–25
105. Kuhn TS. 1970. The Structure of Sci-
entific Revolutions. Chicago: Univ.
Chicago Press. 2nd ed.
106. Lamb T, Avise JC, Gibbons JW. 1989.

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 447
Phylogeographic patterns in mitochon-
drial DNA of the desert tortoise (Xer-
obates agassizi), and evolutionary rela-
tionshipsamongtheNorthAmericango-
pher tortoises. Evolution 43:76–87
107. Lanyon SM. 1993. Phylogenetic frame-
works: towards a firmer foundation for
the comparative approach. Biol. J. Linn.
Soc. 49:45–61
108. LarsonA.1994.The comparison of mor-
phological and molecular data in phylo-
genetic systematics. In Molecular Ecol-
ogy and Evolution: Approaches and Ap-
plications, ed. B Schierwater, B Streit,
GP Wagner, R DeSalle, pp. 371–90.
Basel: Birkhauser
109. Lewontin RC. 1974. The Genetic Ba-
sis of Evolutionary Change. New York:
Columbia Univ. Press
110. Luckow M. 1995. Species concepts: as-
sumptions, methods, and applications.
Syst. Bot. 20:589–605
111. Maddison W. 1995. Phylogenetic histo-
ries within and among species. In Ex-
perimental and Molecular Approaches
to Plant Biosystematics, ed. PC Hoch,
AG Stephenson, pp. 273–87. St. Louis:
Mo. Bot. Gard.
112. Martin AP. 1995. Mitochondrial DNA
sequence evolution in sharks: rates, pat-
terns, and phylogenetic inferences. Mol.
Biol. Evol. 12:1114–23
113. Maruyama T, Birky CW Jr. 1991. Ef-
fects of periodic selection on gene di-
versity in organelle genomes and other
systems without recombination. Genet-
ics 127:449–51
114. Mayr E. 1942. Systematics and the Ori-
gin of Species. New York: Columbia
Univ. Press (1982 ed.)
115. Mayr E. 1963. Animal Species and Evo-
lution. Cambridge, MA: Belknap
116. Melnick DJ, Hoelzer GA. 1992. Differ-
ences in male and female macaque dis-
persal lead to contrasting distributions
of nuclear and mitochondrial DNA vari-
ation. Int. J. Primatol. 13:379–93
117. Mickevich MF. 1978. Taxonomic con-
gruence. Syst. Zool. 27:143–58
118. MickevichMF,Johnson MS.1976.Con-
gruence between morphological and al-
lozyme data in evolutionary inference
and character evolution. Syst. Zool.
25:260–70
119. Miller JS, Brower AVZ, DeSalle R.
1996.Phylogenyofthe neotropical moth
tribe Josiini (Notodontidae: Dioptinae):
evidencefrom DNAsequences and mor-
phology. Biol. J. Linn. Soc. In press
120. Mitton JB. 1994. Molecular approaches
to population biology. Annu. Rev. Ecol.
Syst. 25:45–69
121. MiyamotoMM, Cracraft J.1991. Phylo-
genetic inference, DNA sequence analy-
sis, and the future of molecular system-
atics. See Ref. 121a, pp. 3–17
121a. Miyamoto MM, Cracraft J, eds. 1991.
Phylogenetic Analysis of DNA Se-
quences. Oxford, UK: Oxford Univ.
Press
122. Miyamoto MM, Fitch WM. 1995. Test-
ing species phylogenies and phyloge-
netic methods with congruence. Syst.
Biol. 44:64–76
123. Moore WS. 1995. Inferring phylogenies
from mtDNA variation: mitochondrial-
gene trees versus nuclear-gene trees.
Evolution 49:718–26
124. Moritz C, Dowling TE, Brown WM.
1987.Evolutionofanimalmitochondrial
DNA: relevance for population biology
and systematics. Annu. Rev. Ecol. Syst.
18:269–92
125. Moritz C, Schneider CJ, Wake DB.
1992. Evolutionary relationships within
the Ensatina eschscholtzii complex con-
firm the ring species interpretation. Syst.
Biol. 41:273–91
126. Nei M, Graur D. 1984. Extent of protein
polymorphism and the neutral mutation
theory. Evol. Biol. 17:73–118
127. Neigel JE, Avise JC. 1986. Phylogenetic
relationships of mitochondrial DNA un-
der various demographic models of spe-
ciation. In Evolutionary Processes and
Theory, ed. E Nevo, S Karlin, pp. 515–
34. San Diego: Academic
128. Nelson G. 1983. Reticulation in clado-
grams. In Advances in Cladistics, ed.
NI Platnick, VA Funk, 2:105–11. New
York: Columbia Univ. Press
129. Nelson G. 1989. Cladistics and evolu-
tionary models. Cladistics 5:275–89
130. Nelson G, Platnick N. 1981. System-
atics and Biogeography. New York:
Columbia Univ. Press
131. Nixon KC, Wheeler QD. 1990. An am-
plification of the phylogenetic species
concept. Cladistics 6:211–23
132. Nixon KC, Carpenter JM. 1996. On si-
multaneousanalysis. Cladistics. In press
133. Olmstead RG, Sweere JA. 1994. Com-
bining data in phylogenetic systemat-
ics: an empirical approach using three
molecular data sets in the Solanaceae.
Syst. Biol. 43:467–81
134. Omland KC. 1994. Character congru-
ence between a molecular and morpho-
logical phylogeny for dabbling ducks
(Anas). Syst. Biol. 43:369–86

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
448 BROWER, DESALLE & VOGLER
135. Ospovat D. 1981. The Development of
Darwin’s Theory. Cambridge: Cambrid-
ge Univ. Press
136. Page RDM. 1993. Genes, organisms,
and areas: the problem of multiple lin-
eages. Syst. Biol. 42:77–84
137. Palumbi SR. 1989. Rates of molecu-
lar evolution and the fraction of nu-
cleotide positions free to vary. J. Mol.
Evol. 29:180–87
138. Pamilo P, Nei M. 1988. Relationships
between gene trees and species trees.
Mol. Biol. Evol. 5:568–83
139. Panchen AL. 1992. Classification, Evo-
lution and the Nature of Biology. Cam-
bridge: Cambridge Univ. Press
140. Patterson C. 1988. Homology in clas-
sical and molecular biology. Mol. Biol.
Evol. 5:603–25
141. Patterson C, Williams DM, Humphries
CJ. 1993. Congruence between molec-
ular and morphological phylogenies.
Annu. Rev. Ecol. Syst. 24:153–88
142. Platnick NI. 1977. Cladogams, phyloge-
netic trees, and hypothesis testing. Syst.
Zool. 26:438–42
143. Platnick NI. 1979. Philosophy and the
transformation of cladistics. Syst. Zool.
28:537–46
144. PopperKR.1965. The Logic of Scientific
Discovery. New York: Harper & Rowe
145. Powell JR. 1983. Interspecific cytoplas-
mic gene flow in the absence of nuclear
gene flow: evidence from Drosophila.
Proc. Natl. Acad. Sci. USA 80:492–
95
146. Powell JR. 1991. Monophyly/parap-
hyly/polyphyly and gene/species trees:
an example from Drosophila. Mol. Biol.
Evol. 8:892–96
147. Provine WB. 1971. The Origins of The-
oretical Population Genetics. Chicago:
Univ. Chicago Press
148. Rhymer JM, Williams MJ, Braun MJ.
1994. Mitochondrial analysis of gene
flow between New Zealand mallards
(Anasplatyrhynchos) and grey ducks (A.
superciliosa). Auk 111:970–78
149. Rieppel OC. 1988. Fundamentals of
Comparative Biology. Basel: Birkha-
user
150. Robertson HM. 1995. The Tcl-mariner
superfamily of transposons in animals.
J. Insect Physiol. 41:99–105
151. Rogers AR, Harpending H. 1992. Popu-
lation growth makes waves in the distri-
bution of pairwise genetic markers. Mol.
Biol. Evol. 9:552–69
152. Rowan RG, Hunt JA. 1991. Rates of
DNA change and phylogeny from the
DNA sequences of the alcohol dehy-
drogenase gene for five closely related
species of Hawaiian Drosophila. Mol.
Biol. Evol. 8:49–70
153. Rzhetsky A, Nei M. 1995. Tests of ap-
plicabilityof severalsubstitution models
for DNA sequence data. Mol. Biol. Evol.
12:131–51
154. Saitou N, Nei M. 1986. The number
of nucleotides required to determine
the branching order of three species
with special reference to the human-
chimpanzee-gorilla divergence. J. Mol.
Evol. 24:189–204
155. Saitou N, Nei M. 1987. The neighbor-
joining method: a new method for re-
constructing phylogenetic trees. Mol.
Biol. Ecol. 4:406–25
156. Sanderson MJ, Donoghue MJ. 1989.
Patterns of variation in levels of homo-
plasy. Evolution 43:1781–95
157. Schaeffer SW, Aquadro CF, Anderson
WW. 1987. Restriction-map variation
in the alcohol dehydrogenase region of
Drosophila pseudoobscura. Mol. Biol.
Evol. 4:254–65
158. Selander RK. 1982. Phylogeny. In Per-
spectives on Evolution, ed. R Milkman,
pp. 32–59. Sunderland, MA: Sinauer
159. Sharp PM, Li W-H. 1987. The rate of
synonymous substitutions in enterobac-
terial genes is inversely related to codon
usage bias. Mol. Biol. Evol. 4:222–30
160. Shotake T, Nozawa K, Santiapilai C.
1991. Genetic diversity within and be-
tween the troops of toque macaque,
Macaca sinica, in Sri Lanka. Primates
32:283–99
161. Sibley CG, Ahlquist JA, Sheldon FH.
1987.DNAhybridizationandavianphy-
logenetics. In Evolutionary Biology, ed.
MK Hecht, B Wallace, GT Prance,
21:97–125. New York: Plenum
162. Simon C, Frati F, Beckenbach A, Crespi
B, Liu H, Flook P. 1994. Evolution,
weighting and phylogenetic utility of
mitochondrial gene sequences and a
compilation of conserved polymerase
chain reaction primers. Ann. Entomol.
Soc. Am. 87:651–701
163. Simonsen KL, Churchill GA, Aquadro
CF.1995. Propertiesofstatistical tests of
neutrality for DNA polymorphism data.
Genetics 141:413–29
164. Simpson GG. 1944. Tempo and Mode in
Evolution. New York: Columbia Univ.
Press
165. Smocovitis VB. 1992. Unifying biol-
ogy: theevolutionarysynthesis and evo-
lutionary biology. J. Hist. Biol. 25:1–65

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
GENE TREES VS SPECIES TREES 449
166. SneathP,SokalR.1973. Numerical Tax-
onomy. San Francisco: Freeman
167. Sober E. 1988. Reconstructing the Past.
Cambridge, MA: MIT Press
168. Sober E. 1993. Experimental tests of
phylogenetic inference methods. Syst.
Biol. 42:85–89
169. SpolskyC,Uzzell T.1984.Natural inter-
species transfer of mitochondrial DNA
in amphibians. Proc. Natl. Acad. Sci.
USA 81:5802–5
170. Spolsky C, Uzzell T. 1986. Evolutionary
historyofthe hybridogenetic hybrid frog
RanaesculentaasdeducedfrommtDNA
analysis. Mol. Biol. Evol. 3:44–56
171. Steel MA, Lockhart PJ, Penny D.
1993. Confidence in evolutionary trees
from biological sequence data. Nature
364:440–42
172. Stevens PF. 1983. Augustin Angier’s
“Arbre botanique” (1801), a remarkable
early botanical representation of the nat-
ural system. Taxon 32:203–11
173. StricklandHE.1841. On the true method
of discovering the natural system in zo-
ology and botany. Ann. Mag. Nat. His.
6:184–94
174. Sullivan J, Holzinger KE, Simon C.
1995. Among-site variation and phy-
logenetic analysis of 12S RRNA in
sigmodontine rodents. Mol. Biol. Evol.
12:988–1001
175. Swofford DL. 1991. When are phy-
logeny estimates from molecular and
morphological data incongruent? See
Ref. 121a, pp. 295–333
176. Swofford DL, Olsen GJ, Waddell PJ,
Hillis DM. 1996. Phylogenetic infer-
ence. In Molecular Systematics, ed. DM
Hillis, C Moritz, BK Mable, pp. 407–14.
Sunderland, MA: Sinauer. 2nd ed.
177. SzymuraJM, Spolsky C, UzzellT.1985.
Concordantchangeinmitochondrialand
nuclear genes in a hybrid zone between
two frog species (genus Bombina). Ex-
perientia 41:1469–70
178. Tajima F. 1983. Evolutionary relation-
ship of DNA sequences in finite popula-
tions. Genetics 105:437–60
179. Tajima F. 1989. DNA polymorphism
in a subdivided population: the ex-
pected number of segregating sites in
the two-subpopulation model. Genetics
123:229–40
180. Tajima F. 1993. Measurement of
DNA polymorphism. In Mechanisms of
Molecular Evolution, ed. N Takahata,
AG Clark, pp. 37–59. Sunderland, MA:
Sinauer
181. Takahata N. 1989. Gene genealogy in
three related populations: consistency
probabilitybetweengeneandpopulation
trees. Genetics 122:957–66
182. Takahata N. 1990. A simple genealogi-
cal structure of strongly balanced allelic
linesandtrans-speciesevolutionofpoly-
morphism. Proc. Natl. Acad. Sci. USA
87:2419–23
183. TakahataN,SlatkinM. 1984. Mitochon-
drial gene flow. Proc. Natl. Acad. Sci.
USA 81:1764–67
184. Tegelstr¨om H, Gelter HP. 1990. Hal-
dane’s rule and sex-biased gene flow
between two hybridizing flycatcher
species (Fidecula albicollis and F. hy-
poleuca, Aves: Muscicapidae). Evolu-
tion 44:2012–21
185. Templeton AR. 1989. The meaning of
species and speciation: a genetic per-
spective. In Speciation and Its Conse-
quences,ed.D Otte, JAEndler,pp.3–27.
Sunderland, MA: Sinauer
186. Templeton AR. 1994. Biodiversity at
the molecular genetic level: experiences
from disparate macroorganisms. Philos.
Trans. R. Soc. London Ser. B 345:59–
64
187. Templeton AR, Boerwinkle E, Sing CF.
1987. A cladistic analysis of phenotypic
associations with haplotypes inferred
from restriction endonuclease mapping.
I.Basic theory and an analysisof alcohol
dehydrogenase activity in Drosophila.
Genetics 117:343–51
188. Templeton AR, Crandall KA, Sing CF.
1992. A cladistic analysis of phenotypic
associations with haplotypes inferred
from restriction endonuclease mapping
and DNA sequence data. III. Cladogram
estimation. Genetics 132:619–33
189. Templeton AR, Routman E, Phillips
CA. 1995. Separating population struc-
ture from population history: a cladistic
analysis of the geographical distribution
of mitochondrial DNA haplotypes in the
tiger salamander, Ambystoma tigrinum.
Genetics 140:767–82
190. Thorpe RS. 1984. Primary and sec-
ondarytransitionzones in speciation and
population differentiation: a phyloge-
netic analysis of range expansion. Evo-
lution 38:233–43
191. Vogler AP, Knisley CB, Glueck SB, Hill
JM, DeSalle R. 1993. Using molecular
and ecological data to diagnose endan-
gered populations of the puritan tiger
beetle Cicindela puritana. Mol. Ecol.
2:375–83
192. von Haeseler A, Churchill GA. 1993.
Networkmodels for sequenceevolution.

September 20, 1996 16:5 Annual Reviews BROWCHPT.DUN AR 19-14
450 BROWER, DESALLE & VOGLER
J. Mol. Evol. 37:77–85
193. Weir BS. 1990. Intraspecific differentia-
tion. In Molecular Systematics, ed. DM
Hillis, C Moritz, pp. 373–410. Sunder-
land, MA: Sinauer
194. Wenzel JW, Carpenter JM. 1994. Com-
paring methods: adaptive traits and
tests of adaptation. In Phylogenetics
and Ecology, ed. P Eggleton, RI Vane-
Wright, pp. 79-101. London: Academic
195. Westwood JO. 1840. Observations upon
the relationships existing amongst nat-
ural objects, resulting from more or
less perfect resemblance, usually termed
affinity and analogy. Mag. Nat. Inst.
(NS) 6:141–44
196. Wheeler QD, Nixon KC. 1990. Another
way of looking at the species problem:
a reply to de Queiroz and Donoghue.
Cladistics 6:77–81
197. Wheeler WC. 1991. Congruence among
data sets: a Bayesian approach. See Ref.
121a, pp. 334–46
198. Wiley EO. 1975. Karl R. Popper, sys-
tematics, and classification: a reply to
Walter Bock and other evolutionary sys-
tematists. Syst. Zool. 24:233–43
199. Wiley EO. 1981. Phylogenetics. New
York: Wiley & Sons
200. Yang Z, Goldman N, Friday A. 1994.
Comparison of models for nucleotide
substitution used in maximum-
likelihood phylogenetic estimation.
Mol. Biol. Evol. 11:316–24
201. Yoder AD. 1994. Relative position of
the Cheirogaleidae in strepsirhine phy-
logeny: a comparison of morphological
and molecular methods and results. Am.
J. Phys. Anthropol. 94:25–46
202. Zouros E, Ball AO, Saavedra C, Free-
man KR. 1994. An unusual type of mi-
tochondrial DNA inheritance in the blue
mussel Mytilus. Proc. Natl. Acad. Sci.
USA 91:7463–67