SCIENCE EDUCATION

Understanding phylogenies: Constructing and interpreting phylogenetic

trees
Joseph L. Staton1
1
University of South Carolina Beaufort

Interpretation of phylogenetic trees is fundamental in understanding the relationships between organisms, their traits
or characteristics, their ecology and even their genomic and developmental biology. As trees appear more often in
basic texts, many students, and even their teachers, clearly understand little of how they are constructed and even
less about what can be inferred from them about the history of the representatives analyzed. Not only are these trees
a source of confusion on what they do tell us, often non-specialists infer things wrongly or, worse, others misuse
them in an attempt to negate the validity of evolutionary theory. In this brief introduction, I attempt to give a
synopsis of basic tree-building methods, and more importantly demonstrate interpretation and dispel some common
misconceptions about them.

Understanding a phylogeny, its construction and its

interpretation, is at the core of the modern comparative method in
biology. Life on Earth is diverse and seemingly impossible to
comprehend. Even though objective methods to develop working
hypotheses are central to understanding the evolutionary history
for a group, these methods, or their resulting interpretation, are
not immediately transparent to the majority of students or even
researchers in the broader field of biology. Indeed, many outside
of— or distrustful of— science are dubious that we can actually
derive and study these histories. The use of modern phylogenetic
methods in the life sciences has informed and revolutionized our
Figure 1. Two hypothetical trees. Trees progress from the
understanding of the history of life on the planet and impacted tips (terminal “taxa”) to the base of a tree when “rooted” (see
diverse areas of research in forensic biology, biogeography, section on rooting a tree). Each lineage is represented by a
adaptation, and evolutionary biology. line that joins other lines at “nodes” (=a branch point
There are multiple ways to construct a branching tree (Fig. representing a common ancestral condition; 1-5). This
1) of organisms based on characteristics. Many of the primers progresses further and further until all lines are joined to a
available for learning these methods suffer, if not from single line which is the root of the tree (R). Note that the two
philosophical complexity (not a trivial aspect of different trees are different only in their orientation and the branch
methods employed), then at least from length. Even the more representation; the branching pattern is the same for both.
extensive presentations are geared toward the senior
undergraduate or graduate levels (Baum & Offner 2008, Brooks or taxon (singular of taxa, an operational term for the tips of a
& McLennan 1991, Hall 2011, Page & Holmes 1998). I will tree, either a species, a genus, or higher), a set of characteristics
focus on a brief description of the different methods and a brief (or characters) is compiled into a matrix, consisting of T rows x C
discussion of the relative strengths and weaknesses, but realize columns, where T = the number of taxa and C = the number of
that entire books and even journals have been published dealing characters. Each character has two or more possible types of
with the complex philosophies and intense computer- condition called states. Trees are then constructed to minimize
computational methods to produce these results. The three main the number of changes between all states for all characters
areas of methodology for creating trees rely on characteristic- among all taxa. This is a deceptively simple statement. There is
based (called Cladistics), distance-based (a mathematical index no a priori method for drawing one specific tree based on the
of relative similarity between taxa), or some complex variation among states in character state matrix (see Fig. 2, for
combination of the two (Maximum Likelihood). For example), but for simple datasets the solution can be readily
simplicity’s sake, I will forgo the last category and focus on the apparent, even to non-scientists. An excellent hands-on exercise
details of the first two, as examples to be used to relate the for demonstrating this method was published previously
information that can be and cannot be derived from a phylogeny. (Goldsmith 2003). For large datasets the patterns are often not
obvious, so we construct all possible branching diagrams, count
Cladistics the number of state changes required for each, and retain those
network(s) with the fewest changes. These minimum change
Cladistics arose from the work of Hennig (1950, 1966). A solutions are considered to hold sway over more complex ones
dataset of characteristics (=alternative traits) is accumulated for a (i.e., are more parsimonious, Sober 1981). An analogy that can
number of related organisms to be analyzed. For every organism illustrate this point to students is, as follows: a suspect that is

Journal of the South Carolina Academy of Science, [2015], 13l(1) | 24

 practice for creating a root is by using an outgroup (Wheeler
1990). The outgroup is considered a taxon (or sometimes a group
of taxa) that is closely related to the group of interest (i.e., the
ingroup), but not a part of that group. As a result, a network can
be “rooted” by dividing the outgroup(s) from the remaining
ingroup taxa (see Fig. 1, 2c), with a basal branch or root (R).
One can envision this as “grabbing” the branch between the
outgroup (singular or a cluster of taxa) and remaining network
(the ingroup) and pulling that branch “down” to create a new
basal branch (trunk) of the tree. For example, the network in
Figure 2c was rooted by using “D” as the outgroup for one
unrooted network in Figure 2b (which one was used?).
Rooting the tree provides perspective to the history of
character states and order that new states came into existence. In
a network, one can distinguish how character states transition
along branches of the network, but with no root—there is no
historical perspective of the order in which the traits arose. A
Figure 2. A brief example of a simple matrix with the root identifies those traits common to all descendants in the study
resulting networks. A) An abbreviated data matrix for 9 and those that are shared among only subsets of the groups. With
DNA positions of a gene. B) Three possible unrooted a root, we can ascribe new characteristics to states to identify
networks for four taxa. The top network is the preferred as it their status, either an autapomorphy, a symplesiomorphy, or a
has the least number of changes required when mapping the synapomorphy. In Cladistics, synapomorphies are considered to
state changes on the network. C) That network is redrawn be the “phylogenetically informative” character states in a study,
and “rooted” using taxon D as the outgroup. The root is then
because they support the pattern of branches between taxa. One
a stem introduced to the network that isolates the outgroup
from the remaining ingroup and represents the basal ancestral last group of characteristics is called homoplasy. When the most
condition for the phylogeny. parsimonious (= shortest) tree is found, a small subset of the
characteristics analyzed may not be consistent with the overall
arrested near a murder scene fits the general description of a tree, which is driven by the largest congruent subset of
person who was witnessed committing a murder and has similar synapomorphies. These non-congruent changes are called
fingerprints to those found at the murder scene. Barring further homoplasies. They are not “bad” data, but represent either
evidence of another such person existing (i.e., with similar characters that are thought to be misrecorded (and need to be re-
appearance and fingerprint pattern) and being near the vicinity of evaluated), or that the hypothesis of their homology may be
the murder at the same time, the suspect in custody is the most flawed, as traits can evolve more than once (parallel evolution).
likely culprit. Computers usually work with networks of taxa to This case is especially well known for individual DNA bases over
complete these computations (see Fig. 2), and subsequently a long time periods. Indeed, classic homoplasies are often trivially
network can be “rooted” (that is, shown with an estimated point obvious in many taxa (e.g., the wing of a bird and the wing of a
of origin) when one considers the ancestral condition of the group bat), but many not so obvious in terms of subtle characteristics in
under consideration (see “Rooting a Tree”). a cladistic analysis.
With relatively few taxa, this is trivial. For only three taxa,
there is only one unrooted network (but three possible “rooted” The Use of Cladistic Methods
trees, where the root isolates two taxa from the other one, what
would that look like?). With four taxa, there are only three The use of morphology in cladistics is appealing since it has
possible arrangements (see Fig. 2). After the number of changes been at the core of evolutionary study for over a hundred years.
(=evolutionary events) needed are noted for each, the network The further idea that cladistics methods are objective and can
that requires the fewest evolutionary events (= state changes) is remove researcher bias adds to that appeal. By using a large
chosen as the preferred network, given the notion of parsimony number of character states, cladistics methods can provide a
(i.e., “Ockham’s razor”). It is the shared character states that robust understanding of traits across an entire group with some
support the branching patterns of a tree, whereas those states level of completeness.
unique to a single branch or shared among all taxa (while Additionally, a phylogenetic tree can aid in understanding
evolutionarily of interest) do not shed light on relationships in the processes that yield the patterns in it. One can “map” the
network or tree. character state changes onto the tree to see when specific traits
evolved and in which taxa they occur, or which descendants may
Rooting a Tree and Analyzing Characters have secondarily lost them. For all those descendants possessing
a trait, it must have been present in the common ancestor—
Once the best network is chosen, it is usually rooted at a therefore it evolved at some point prior to the speciation of that
“trunk,” which represents the ancestral condition for all members ancestor (i.e., the node) and from the previous ancestor that gave
within the tree. Several ways have been suggested over the years rise to it (i.e., the next deeper node). In a humorous treatment
of phylogenetic analysis, but by far the commonly accepted elsewhere (Staton 1998), I argued that the characteristic of “tastes

Journal of the South Carolina Academy of Science, [2015], 13l(1) | 25

like chicken” confuses a deep symplesiomorphic trait (here In the recent decades, more phylogenetic work has focused
“chicken flavor”) with a derived condition (the evolution of on DNA data. Although morphological homology may be
adapations that are present in birds, alone). Clearly, a “chicken- difficult to assess between a lobster and a human, both species
like” flavor is present in many other related organisms (the contain genes that are homologous at the molecular level.
Tetrapods, e.g., reptiles, amphibians, some mammals, etc.)— Phylogenetic analyses of DNA is a powerful tool, however its
therefore the phrase should not be “tastes like chicken” but more analysis with cladistic methods can result in a unique problem
aptly “tastes like tetrapod.” Similarly, if we think about biology called “long branch attraction” (described originally by
in a phylogenetic sense of first appearance of a specific Felsenstein 1978). Some DNA datasets contain broadly related
characteristic, we can answer the age-old pseudo-philosophical taxa with highly divergent DNA sequences. Random patterns of
question, “Which came first, the chicken or the egg?” since species formation and extinction may mean that one or a few taxa
several species predating the origin of birds laid eggs. Such might be quite different from those in the rest of the study. After
humorous approaches can get even the most jaded student DNA is aligned at the nucleotide level— unique taxa will differ
engaged in discussions about phylogenetic trees. greatly from all others (> 30-50% of variable bases at all
Sometimes, organisms that share a close biological nucleotide positions). When this happens, these taxa tend to
relationship, like that of parasites and their hosts, can be analyzed branch off at a node near the least-related taxon— not from
separately and compared for similar branching patterns among common ancestry but due to a few similar (convergent) mutations
the host/parasite trees. When there is close agreement of that randomly accumulate between sequences over time (i.e., they
branching pattern, it is taken as evidence of a shared, co- share chance similar homoplasies). This is most common for
evolutionary history (Brooks & McLennan 1991). There are also DNA because an adenine (A) at position 132 looks like another
more complex analyses that relate biogeography to phylogeny adenine at that position (A), whether it was inherited via common
(termed phylogeography; see Avise 1998, 2000) so that ancestry or converged to the same base by a separate mutational
researchers can understand how species evolve within a event (in this case A). Since the algorithms are developed to find
geographical landscape and do or do not spread over the different the shortest tree by making branching patterns, they will make all
biotic zones of the planet. taxa branch with one another, even if badly. The most
Up to this point, I have not discussed the use of phylogenetic maddening fact about this miscalculation is that the addition of
methods with molecular data, but the real growth in these more sequence data will only make the algorithm find— with
analyses has been in their use with DNA data to produce even greater support—that a “bad” tree is the most parsimonious
phylogenies, since the development of new sequencing (see Felsenstein 1978). In such cases, other methods (such as
technologies in late 1980s. Such methods allow for hypotheses distance-based methods) provide a different means of developing
of deeper phylogenies, as well as comparison of morphologic to a phylogeny that do not suffer from these issues.
molecular trees to assess agreement in these results. There is no
succinct way of reviewing the diverse ways that phylogenetic Building Distance-based Trees
trees can be employed in different research programs, here— but
it is without doubt that these methods are in widespread use Oddly enough, the first methods for making trees were
across most biological fields. phenetic methods (Sokal & Sneath 1963) based on some metric
of similarity (or its converse—distance) between datasets to
Issues that Impact Cladistic Methods construct a branching network. This often worked poorly for
morphological data (true phenetics), but worked arguably better
Several difficulties are inherent in application of a cladistic for molecular data, where there was an implied mode of evolution
analysis. I will try to give a very brief listing of them, not as an between pairs of species DNA. These methods are
indictment against their use, but more as an explanation as to why computationally faster and usually yield a single tree (a feat not
there is no single accepted method for producing phylogenies. always accomplished by cladistic methods). Many argue that
Since cladistic methodology works backwards to since the data are transformed into a comparative metric (e.g.,
hypothesize an evolutionary process, we approach the problem by percentage similarity or weighed similarity), the results are not as
constructing all possible networks and saving the shortest reliable as cladistic methods. However, with the aforementioned
network as the best estimate of the correct one. The problem “long branch” problem and with large datasets involving
with this method is that the number of networks possible hundreds of taxa (often the case in many studies), cladistic
increases exponentially as the number of taxa increases linearly. methods are not accurate or practical. As previously noted, to
The formula for the number of possible unrooted networks for n review all possibilities of methods here would not be possible.
taxa is (2n-5)! / [2n-3  (n-3)!] (Eq. 5.1; Li 1997). For example, There are several competing tree/network building algorithms
there are ~5  1094 possible networks for 60 taxa, which is more and many distance metrics that are reviewed elsewhere (Li 1997,
than the estimated number of atoms in the universe. It is, in fact, Nei 1996, for example). For the sake of teaching the essence of
a computational impossibility to search all of these networks these methods, I will focus on the classic analysis of variation in
merely due to time limitations. Researchers have developed the protein cytochrome c (data from Fitch & Margoliash 1967)
shortcuts of searching a reasonable subset many of the most using a simple metric—minimum mutational distance— and
probable networks. Still, the methods are computationally branching algorithm—the unweighted pair-group method based
intensive and can provide misleading results or even miss the on arithmetic means (or UPGMA, Sokal & Michener 1958). In
most parsimonious tree. this study, amino-acid sequences were aligned for each species

Journal of the South Carolina Academy of Science, [2015], 13l(1) | 26

from previously published data, and a minimum mutational
distance was calculated based on the fewest base substitutions
required in the redundant genetic code to change each amino acid
between each pair (Table 1A). The largest difference in the
subset of their data we analyze (Table 1A) is Tuna/Moth pairwise
distance of 41, meaning the minimum number of nucleotide
changes needed to account for the paired Tuna/Moth cytochrome
c differences at the amino-acid level locus is 41 (i.e., each
changed codon could differ by either 1, 2, or all 3 bases in the
translation code). Methods that use pairwise distances seek to
produce a single branching “tree” so that length of branches
between all taxon pairs is proportional to each branch distance for
taxa in the “tree.” In the case that all distances are not strictly
additive, as can happen with distance datasets, UPGMA averages
the non-equal distances to reach a balanced compromise among
values across all pairs. For UPGMA, the smallest difference
between a pair of taxa is used as the starting point—here between
Man (Homo sapiens) and Monkey (a Rhesus monkey [Macaca
mullata])—and is taken to be the complete distance between the
two taxa. In this case, the minimum distance of one nucleotide
between them is considered as one-half of one nucleotide
difference (on average) from their common ancestor (see Fig. 3).
In the next (and all subsequent steps), the original matrix of
Figure 3. A UPGMA tree based on the minimal mutational
pairwise distances is collapsed one level so that the new group (in distance for each pair of taxa listed in Table 1 (after Fitch &
step 2 “Man/Monkey”) becomes a single column (or a combined Margoliash 1967). Note that the distance between Man or
taxon group), and all remaining individual distances are collapsed Monkey to node 1 is half the total distance between Man and
so that each collection of distances (in step 2, those paired with Monkey and represents an average distance from their most
“Man” and “Monkey”) now becomes an average for the collapsed recent common ancestor. Each node is labeled in ascending
group to each remaining taxon (e.g., in this case the “Man-Turtle” order of the calculation from the original matrix. The next
distance of 19 differences gets averaged with the “Monkey- calculation in the iterative process is node 2 joining “Turtle”
Turtle” distance of 18 to make an average difference of and “Chicken”.
“Man/Monkey” and “Turtle” now 18.5). This is continued for all
and reptiles form a group with a common ancestor exclusive of
pairwise distances associated with “Man” or “Monkey” until a
other vertebrates (node 2), and all organisms believed to have
second matrix is complete, which has one less column and row
descended from a common amniote-egg layer cluster at a deeper
(see Table 1B). This procedure is continued with the next
level (node 4) than do the nodes of more recent divergence
smallest distance in the recalculated table (“Turtle/Chicken” in
(nodes 1, 2 & 3). The fact that molecular evolution parallels that
Table 1B). All subsequent averages (means) involve defining the
of other evolutionary theories based on different data (e.g.,
mean for all possible pairwise distances for each new cluster from
morphology) is not surprising, but it is impressive that even a
those in the original distance matrix. While calculating this by
small protein sequence can accurately capture these hierarchical
hand for the first time seems complicated, it is a simple repetitive
patterns, in this case.
algorithm and can be completed in milliseconds by computer
even for large datasets. The end result of this process is a tree
where pairs or groups of taxa will have branches where each pair
Interpretation of trees
of taxa is approximately 0.5 the total distance to the node as they
Certainly, the interpretation of a phylogenetic tree is the
have between one another (i.e., “Man” to “Monkey” = 1, Table
basis for most of comparative evolutionary biology. The inferred
1A; “Man” or “Monkey” to node 1 = 0.5, Fig. 3). With the
pattern of branching is a road map to the understanding of any
averaging process, the deeper (i.e., more distant nodes) are less
other hierarchy of traits that is possessed by those groups. Such
reflective of the original data (“Moth” to “Tuna” = 41, Table 1A;
traits can then be assessed as to relatively when they arose in the
“Moth” or “Tuna” to node 6 = 17 [not 20.5]). Other tree methods
spectrum of ancestor-descendant relationships within the groups
have been developed to circumvent these types of averaging
and their subgroups.
errors (e.g., neighbor-joining, Saitou & Nei 1987), and results
from these methods of tree construction can be demonstrated with
a computer program in the classroom—but are not as amenable to
Misconceptions in tree interpretation
direct calculation by teacher or student at the introductory level.
There are several misconceptions that can arise in the
The resulting tree is close to the evolutionary tree that many
interpretation of phylogenetic trees. Some of these will already
would predict based on morphological, physiological or
seem foolish to you, the reader, if you have followed the narrative
biochemical similarities (Fig. 3). The two primates are shown
to this point, but others may not be as straightforward.
closest together (node 1), mammals form a clade (node 3), birds

Journal of the South Carolina Academy of Science, [2015], 13l(1) | 27

 Table 1. A) The minimal mutational distance between cytochrome c amino acid sequences of different taxa from Fitch &
Margoliash (1967). The top half of the matrix is omitted, as it would be a duplicate of the data shown. The diagonal is the identity
value for each species. B) This matrix shows the recalculation and reduction of the data in A to a six by six matrix, with Man-
Monkey as a single row/column and each pairwise difference now an average of each distance (e.g., [AB + AF]/n or [19 + 18]/2 =
18.5) from the original columns containing “Man” and “Monkey”, respectively (see shaded cells). The next lowest pairwise
distance (here the “Turtle/Chicken” distance of 8) will be the next pair of taxa to be collapsed into a single matrix row/column. Note
that in many cases the next smaller group can be between a single taxon and a previously collapsed row/column.

A: Turtle Man Tuna Chicken Moth Monkey Dog

A B C D E F G
Turtle A --
Man B 19 --
Tuna C 27 31 --
Chicken D 8 18 26 --
Moth E 33 36 41 31 --
Monkey F 18 1 32 17 35 --
Dog G 13 13 29 14 28 12 --
Man to node 1 = 0.5 Monkey to node 1 = 0.5

B: Turtle Man-Monkey Tuna Chicken Moth Dog

A BF C D E G
Turtle A --
Man-Monkey BF 18.5 --
Tuna C 27 31.5 --
Chicken D 8 17.5 26 --
Moth E 33 35.5 41 31 --
Dog G 13 12.5 29 14 28 --
Chicken to node 2 = 4.0; Turtle to node 2 = 4.0 (8 changes total)

The intuition of some students can lead to erroneous the formation of each descendant species. The idea that the study
thoughts that the gene in question is the source or driver of of evolution focuses on finding direct “missing links” is a strange
speciation, since they are loosely taught that “mutation the source hold over from earlier times when people believed in almost
of genetic change.” However, even though every gene has the “alchemistic” transformation of one species into another (see
potential to capture unique mutational/molecular events that Saint-Hilaire 1822). At fine scales of micro-evolution we often
parallel a speciation event within individuals, the temptation is find transitions, but not finding specific transitions between larger
for the individual gene under scrutiny to be the sole focus of the taxonomic groups is not an indictment of the absence of the
speciation process for the novice. Molecular evolution is only a evolutionary process, as has been suggested by some factions
measure of the speciation pattern which has been captured in a critical of evolution.
molecular context of any given stretch of DNA. Students might Lastly, the casual observer often attributes importance to the
aver that “a Dog would be a Monkey, if not for those 12 order of the tips of the tree. While these are grouped in some
changes!” Or, wrongly assert that 12 mutations directly mutated a order by the author, their specific pattern has no inherent
dog ancestor into a monkey, like some perverse science fiction meaning, necessarily. The fact that our example tree has “Dog”
movie. It is a difficult concept to grasp that DNA mutations in positioned next to “Monkey” has no special relevance (Fig. 3),
the cytochrome c record, in parallel, other changes in these nor does “Turtle” being next to “Tuna”— although novices will
lineages, and that only 12 of these changes were accumulated be confused by this pattern. Dogs are equally distant or close to
between dog and monkey cytochrome c, since their last common the Man/Monkey common ancestor, but that is all. Likewise,
ancestor. In this sense, different genes are changing faster or “Chicken” is no closer or more distant than “Turtle” to the
slower in the evolutionary process, and sometimes they give us mammalian ancestor (node 3, Fig. 3). In fact, the final tree could
insight into the process as a whole. be drawn with “Moth” next to “Man,” however the branch of
There is no real expectation that each node represents a “Moth” would still connect at the base of the tree, at node 6 (Fig.
specific known species, or worse is some combined (chimeric) 3). The order of the tips should be considered fluid, like
“monster” (i.e., a “Dog-Man-Key” for node 3, Fig. 3). Each node membrane-bound proteins floating in an unconstrained lipid-
merely represents that some lineage of varying individuals that bilayer, even though they might be connected within the
had a specific cytochrome c sequence that was ancestral prior of “cytoplasm.” Each node is like a frictionless turnbuckle/pivot;

Journal of the South Carolina Academy of Science, [2015], 13l(1) | 28

free to rotate like arms of a mobile viewed upside down. Only would also like to thank Dr. Alan Harvey, who first taught me to
the relative (internal) branching pattern is important, which is understand cladistics.
why trees can easily be represented by Venn diagrams or as a
nesting code (called Newick format)— (((((Man, Monkey) Notes and references
Dog)(Chicken, Turtle) Tuna) Moth)—where the parentheses
*Corresponding author E-mail: [email protected]
represent the clustering of taxa. The ordering of the clusters is
unimportant: (((((Man, Monkey) Dog)(Chicken, Turtle) Tuna)
Avise JC (1998) The history and purview of phylogeography: a personal
Moth) is equal to (Moth (Tuna (Dog (Monkey, Man)) (Turtle, reflection. Molecular Ecology 7: 371-379.
Chicken))). The relative grouping of taxa is the same. Avise JC (2000) Phylogeography: The History and Formation of
Species. Harvard University Press, Cambridge, MA. 484 pp.
Misrepresentations of phylogenetic trees Baum DA and Offner S. 2008. Phylogenics & tree-thinking. The
American Biology Teacher 70:222-229.
Brooks DR & McLennan DA (1991) Phylogeny, Ecology, and Behavior:
Lastly, it is necessary to point out that our phylogenetic a Research Program in Comparative Biology. Chicago: University of
knowledge is not complete and hypotheses about relationships Chicago Press. 441 pp.
change over time. Some groups are well-studied and this is less Dobzhansky T (1973) Nothing in biology makes sense except in the light
likely, but the more research data is gathered on different of evolution. The American Biology Teacher 35:125-129.
Felsenstein J (1978). Cases in which parsimony and compatibility
groups—the clearer the tree of life becomes. As we progress methods will be positively misleading. Systematic Zoology 27: 401–
toward a better understanding of evolutionary patterns, we often 410.
admit that parts of our understanding are incomplete. Texts will Fitch WM, Margoliash E (1967) Construction of phylogenetic trees.
publish trees with representative question marks or dotted lines to Science 155:279-84.
Goldsmith DW (2003) The great clade race: presenting cladistic thinking
represent poorly understood or unstudied/understudied regions of
to biology majors & general science students. The American Biology
a phylogeny. This is not a representation of the ineffectiveness of Teacher 65: 679-682
the method, but an honest assessment of what can being Hall BG (2011) Phylogenetic Trees Made Easy: A How To Manual.
substantiated by the method at present. That is the strength of Sinauer Associates, Inc.; Fourth edition. 255 pp.
phylogenetics, we can list objective criteria for the tree-branching Hennig W (1950) Grundzüge einer Theorie der phylogenetischen
Systematik. Deutscher Zentralverlag, Berlin. 370 pp.
patterns that we uncover. The relative strengths and weaknesses
Hennig W (1966) Phylogenetic Systematics. Urbana, IL: University of
can be objectively presented. Unfortunately, non-scientists that Illinois Press.
object to the principle of evolutionary biology point to this as a Li W-H (1997) Molecular Evolution. Sinauer Associates, Inc. 487 pp.
weakness of the process of evolution or that somehow lack of Nei M (1996) Phylogenetic analysis in molecular evolutionary genetics.
complete resolution, in fact, renders all of evolutionary theory Annual Review of Genetics 30: 371-403
Page DM & Holmes EC (1998) Molecular Evolution: A Phylogenetic
questionable. Nothing could be farther from the truth. Evolution Approach. Wiley-Blackwell. 352 pp.
could be invalidated if an objective team of researchers found a Saint-Hilaire GE (1822) Considérations générales sur la vertebre.
human fossil in the Cambrian, or that land plants predated ocean Mémoires du Muséum d'Histoire Naturelle 9: 89-119
algal species. Mere incompleteness of understanding has never Saitou N & Nei M (1987) The neighbor-joining method: a new method
been a challenge to science or evolution as a process, only a for constructing phylogenetic trees. Molecular Biology & Evolution
4: 406-425.
challenge pointing out the need for better studies. Absence of Sober E (1981) The principle of parsimony. British Journal for the
evidence is not the same as evidence of absence. In fact, modern Philosophy of Science 32: 145–156.
molecular methods and classical understanding of morphology Sokal RR & Michener CD (1958) A statistical method for evaluating
are showing us in ways never before thought possible the unity of systematic relationships. University of Kansas Science Bulletin 38:
1409-1438.
life on this planet.
Sokal RR & Sneath PHA (1963) Principles of Numerical Taxonomy.
Freeman & Co., San Francisco. 359 pp.
Conclusions Staton J (1998) Tastes like chicken? Annals of Improbable Research 4(4):
5-9.
Understanding of phylogenetic trees and the underlying Wheeler WC (1990) Nucleic acid sequence phylogeny and random
outgroups. Cladistics 6:363 – 368.
process of creation has become central to the understanding of
comparative biology at multiple levels. Phylogenies put vast
amounts of biological data into an integrated whole (Baum &
Offner 2008) not unlike GIS databases integrate geography with
demographics and economics of a region. It adds depth and
connectivity to data that to a non-specialist might seem
completely unrelated. As Theodor Dobzhansky (1973) said,
“Nothing in biology makes sense, except in the light of
evolution.”

Acknowledgements: I would like to thank Dr. T.M. (Mac)

James, Dr. Mark Maloney, and others, for comments that greatly
improved this manuscript. All remaining errors are mine. I

Journal of the South Carolina Academy of Science, [2015], 13l(1) | 29