Integrative and Comparative Biology

Integrative and Comparative Biology, pp. 1–11

doi:10.1093/icb/icy076 Society for Integrative and Comparative Biology

SYMPOSIUM

Biodiversity Assessment, DNA Barcoding, and the

Minority Majority
Julia D. Sigwart1 and Amy Garbett
Queen’s University Marine Laboratory, 12-13 The Strand, Portaferry BT22 1PF, Northern Ireland
From the symposium “Measuring Biodiversity and Extinction: Present and Past” presented at the annual meeting of the
Society for Integrative and Comparative Biology, January 3–7, 2018 at San Francisco, California.

1
E-mail: j.sigwart@qub.ac.uk

Synopsis The majority of species on Earth are in “under-studied” groups, and indeed probably the majority of species
remain undiscovered and undescribed. Species are natural units of evolution, and they are formed from branching phyloge-
netic processes that have a mathematical structure. So it follows that we should be able to develop a set of general principles
that describe global patterns of species groups, like genera. Understanding such patterns would lend considerable power to the
approach of “taxonomic surrogacy.” In environmental assessments, ecology, and paleontology, it is common to substitute
genus-level or family-level identification where definitive species identification is impractical. Clarity and confidence in fun-
damental patterns, based on a robust null model for species and genus level diversity, can accelerate species discovery: there
are more species in the tropics, species-poor genera are very common, large genera are rare. Much hope has been placed in
DNA barcoding as an effective tool to increase the pace of species discovery, but it is abundantly clear that certain mito-
chondrial DNA (mtDNA) markers are more or less variable in different clades and universal threshold values are impractical
to delimit species. This study further examines the patterns of divergence in one common mtDNA barcode fragment,
cytochrome c oxidase subunit 1at the genus level. We compared pairwise divergence in this fragment between two animal
clades that have similar species richness but different evolutionary histories: birds and bivalves. We analyzed quality controlled
alignments of over 39,000 published sequences in 1223 genera. Median pairwise differences at the genus level are positively
correlated with the species richness of a genus, and this is not dependent of the number of sequences sampled. Unsurprisingly,
sequence divergence in vertebrates was far more constrained than in evolutionarily more ancient non-vertebrate clades.
Differences among the groups examined highlight the need for DNA barcode approaches to be considered in the context
of specific biological groups. Vertebrates are better studied, but not necessarily representative of the majority of biodiversity. A
technique that provides powerful insights for vertebrate species may be ineffective for the majority of organisms.

Introduction Skewed distributions have been observed in tax-

Many natural and human systems follow highly
skewed distributions. Skewed distributions are so
prevalent, it has even been proposed that this may
be an emergent property of the universe, attributable
to entropy (Grönholm and Annila 2007). The distri-
bution of wealth, frequency of the sizes of cities, or
the sizes of corporations all form similar patterns
where smaller units are very frequent and larger units
are increasingly less common. There are many small
towns but a few major cities; there are many poor
people and a few mega-rich. The same type of pattern
emerges in ecology, such as in the relative abundance
of co-occurring species: most species are rare.
onomy, where smaller groups are very frequent and
larger groups are less common (Fig. 1). Monotypic
genera, those with only one genus, are the largest
fraction of global genera; large genera are globally
rare. This was first observed by Yule (1925), but
many authors considered the skewed frequency dis-
tribution within taxonomic ranks to be an artifact
of human preferences in classification (Strand and
Panova 2015). Recent work has now shown that the
skew distribution in taxonomy is concordant with
evolution, and phylogenetic simulations produce
similar taxonomic patterns to real world data
(Sigwart et al. 2018).

ß The Author(s) 2018. Published by Oxford University Press on behalf of the Society for Integrative and Comparative Biology.
All rights reserved. For permissions please email: journals.permissions@oup.com.
Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946
by New Mexico State University Library user
on 28 July 2018
 2 J. D. Sigwart and A. Garbett
60%
50%
percent of genera
40%
30%
20%
10%
0%
1 5 10
genus size (species richness)
Fig. 1 The distribution of genus size in terms of species richness follows a well-described right skew pattern. Among global genera in
animals and plants, monotypic (one species) and small genera are most frequent, and large genera are a very small proportion of
genera. Figure modified from Sigwart et al. (2018).
The relationship between phylogeny and taxon-
omy is important because measures of biodiversity
frequently depend on the substitution of higher-
ranked groups wherein species identifications are
not available (Gaston and Williams 1993; Bertrand
et al. 2006). This approach of “taxonomic surrogacy”
is used in estimations of both present biodiversity
and long-term patterns in the fossil record. Major
shifts in the diversity of genus-level or family-level
groups are the primary evidence for past mass ex-
tinction events (Raup and Sepkoski 1986; Hendricks
et al. 2014). Identifying some organisms to genus,
family, or perhaps even coarser groupings is a tech-
nique that is also routinely used in ecology and en-
vironmental biology. In biodiversity assessment it is
common to identify a specimen to the best available
level, based on the expertise of the observer or the
quality of information. It is not clear whether mor-
photaxonomic approaches (using key characters or
synapomorphies for a diagnosis to genus or family
level, as appropriate) are fully transferable to modern
molecular approaches used to delimit species.
Identifying species often involves using mitochon-
drial DNA (mtDNA) fragments. There are several
advantages of mtDNA over nuclear markers. It is
estimated that mtDNA has a mutation rate perhaps
10 times higher than that of nuclear DNA, and this
makes it appropriate for identifying recent evolu-
tionary divergences, at the population or species level
(Brown et al. 1979). There are vast numbers of
Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946
by New Mexico State University Library user
on 28 July 2018
All marine invertebrates Birds
n = 29,269 genera n = 2,278 genera
Odonata Rhodophyta
n = 688 genera n = 1,005 genera
Reptiles Bryophyta
n = 1,176 genera n = 561 genera
Fish Pteridophyta
n = 4,915 genera n = 1,309 genera
Mammals
n = 1,242 genera
50 100 500 1000
mitochondria available from each cell, and mtDNA
is also known to be more durable than nuclear DNA
(e.g., Schwarz et al. 2009), although this varies
among different organisms perhaps with metabolic
rate (e.g., Kazakova and Markosian 1966). This
makes mtDNA a useful tool for cases where environ-
mental and weather conditions may have degraded
the DNA, including in subfossil remains, and mu-
seum material in chemical preservation (Friedman
and DeSalle 2008; Higgins et al. 2015). Work by
Hebert et al. (2003) proposed that most animal spe-
cies can be rapidly and correctly identified by exam-
ining the DNA sequence of a portion of the
cytochrome c oxidase subunit I (COI) mitochondrial
gene)—a “DNA barcode.”
In order for DNA sequence data to provide useful
species-level identifications, there must be a well-
developed reference library to match the genetic se-
quence of the individual of concern to species that
was previously identified (Gotelli 2004). If there is
no reference sequence that matches the species being
examined, then comparing a new sequence to the
database will either return no answer or will be,
worse, an incorrect identification based on the clos-
est available match. Such publicly accessible reference
databases covering taxa from around the globe are
continuously being built, but remain very incomplete
(Federhen 2011; Ratnasingham and Hebert 2007). In
fungi, this was identified as a significant problem in
2008, when the number of sequences produced from
 The biodiversity minority majority
amplifying environmental samples had already begun
to outpace specimen-based voucher sequences with
robust identification (Blackwell 2011).
In cases where a species-level match is not possi-
ble, molecular operational taxonomic units can be
analyzed as interim units for ecological studies
(Blaxter et al. 2005). In the most automated approach,
often used in microbiology, operational taxonomic
units are identified based on a similarity threshold
of 1–3% difference in pairwise comparisons among
sequence fragments. There are, however, limitations
associated with barcoding where recently diverged or
hybridizing species may be overlooked, or specimens
from one species with naturally high variation in that
region are falsely scored as multiple species. Over a
large spatial scale the separation between intraspecific
versus interspecific divergences can be reduced, limit-
ing the power of the analysis to observe distinct spe-
cies (Bergsten et al. 2012). It is clear that a generalized
threshold to separate species is not appropriate, and
may not even be transferable across groups or regions.
Importantly, higher taxonomic ranks are not as-
sumed to represent any fixed level of diversification.
Different groups evolve at very different rates (Langley
and Fitch 1974). Therefore, a genus, family, or class
can neither represent an absolute span of diversity in
terms of species richness nor morphological disparity.
That would imply some ultimate authority and a rig-
idly fixed process of macroevolution. Instead, ranks
are applied in a relativistic framework, grouping spe-
cies based on comparisons with their nearest similar
or evolutionarily related counterparts. Nonetheless,
rates of diversification of lineages that are morpho-
logically recognised as species and clades should be
linked to rates of evolution of the genome.
For many reasons, the question of genus level di-
versity is crucial to understanding global biodiver-
sity. Binominal taxonomic names include species
and genus epithets, and fundamentally similar spe-
cies are classified in the same genus. This article
asked a simple question of whether the genetic “size”
of a genus, measured as divergence in the COI bar-
code region, had any recognizable trends or patterns
that correlate with species richness of the sampled ge-
nus. For the sake of argument, there are several poten-
tial patterns that may link the variation in species
richness of genera to genetic divergence. For example,
there could be a fixed threshold distance that univer-
sally separates species. In this case, higher species rich-
ness in a genus would result in a larger genetic size of
the genus. Alternatively, a genus may be a fundamental
unit with a fixed genetic size. In this scenario, the di-
vergence among species contained in a genus would be
constant regardless of species richness, and remain
Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946
by New Mexico State University Library user
on 28 July 2018
3
unaffected by rapid radiations. Finally, species may
be placed in a genus by taxonomists at random, which
would therefore result in entirely random genetic di-
vergence across genus-level groups.
In order to address this question, we sampled all
available published COI barcode sequences for two
major animal groups with contrasting taxonomic
and evolutionary contexts. Birds are the most taxo-
nomically complete major group of animals; although
there are ongoing arguments about the partitioning of
species and subspecies units, the number of named
species of birds closely matches with the projected
total global species richness (Barrowclough et al.
2016). Birds also present a wealth of natural history
data and extensive knowledge of their evolution and
their major radiation that began shortly after the end
Cretaceous, 66 million years ago (Claramunt and
Cracraft 2015). The rates of evolution of bird mito-
chondrial genomes are correlated with species diver-
sification (Eo and DeWoody 2010). By contrast,
bivalves are a broadly distributed invertebrate group
with a deep and rich fossil record dating to the ear-
liest hard-shelled animals in the Cambrian (Stöger
et al. 2013). Yet there is an ongoing and high fre-
quency of discovery of new species of bivalves, and
the basic shape of the bivalve phylogenetic tree has
only recently been resolved (Gonzalez et al. 2015).
Systematic studies of many other groups of hard-
shelled marine invertebrates with long fossil records
are similarly dominated by discovery of new taxa over
phylogenetic revision.
There are clearly different taxonomic traditions in
different organismal groups; the technical specifics of
what differentiates species- or genus-level diagnoses
cannot be directly transferred from one animal group
to another. Still, emergent taxonomic pattern of
genus-size leads to statistically equivalent frequency
distributions in birds and marine invertebrates
(Sigwart et al. 2018; Fig. 1). This sets up a comparison
of contrasts to examine the role of a DNA barcode
marker in distinguishing genus-level groups.
Methods
The analytical workflow for this project was man-
aged in R (R Core Team 2017). This followed a
process, described below, of extracting data from
the online BOLD Systems (Ratnasingham and
Hebert 2007; http://www.boldsystems.org/, accessed
March 2018), applying quality control filters to the
data, aligning sequence files via the CIPRES Science
Gateway (Miller et al. 2010), and applying further
quality controls by inspection, before assessing
 4 J. D. Sigwart and A. Garbett
patterns of genetic distance. In these steps, data were
managed at the taxonomic ordinal level.
Taxa were selected for the analysis to represent
animal groups with deliberately different taxonomic
completeness and taxonomic practice. We extracted
COI barcode data for all living species of birds,
bivalves, and three other groups of marine inverte-
brates with hard skeletons and deep fossil records:
polyplacophoran molluscs (chitons) in two orders:
Lepidopleurida and Chitonida, bryozoans in the or-
der Cheilostomatida, and brittlestars in the order
Ophiurida. The taxonomic names of orders were
used as the “taxon” search string via the “bold”
package in R (Chamberlain 2017). Alternative ver-
sions of the names were added whenever available
(e.g., combining results for both “Venerida,” and
“Veneroida” for Venus clams). The extracted data
were subset to include only COI fragments, filtered
by the database entry for genetic marker being either
“COI-5P” or blank. The resulting filtered dataset was
exported in fasta format. Each sequence file was then
aligned in CLUSTALW (Larkin et al. 2007) using a
high gap penalty to prevent internal gaps. Resulting
alignment files were then inspected, and any mis-
aligned sequences were discarded (these were as-
sumed to be reverse compliment sequences or
alternative fragments that were not correctly identi-
fied in the source database, but no attempt was
made to fit them to the alignment). Where a se-
quence was fit to the alignment via internal gaps
or violated a conserved sequence region, it was re-
moved. The alignment was then manually trimmed
to a central region of 400 bp that maximized overlap
of the available sequences in that taxonomic order.
All sequence records that were incomplete for ge-
nus and/or species identification were discarded.
Species names in online genetic data repositories
also include unidentified species (often called
“sp.”), hybrids and varieties, and mis-spellings (e.g.,
one species of parrot represented as both
Orthopsittaca manilatus and O. manilata). We did
not attempt any intervention to correct the syntax,
so completeness must be seen as approximate, and
may be inflated. The number of species names per
genus may exceed the number of valid names, where
there are additional variant spellings, varietals, or
subspecies. Where species coverage exceeded 100%
(more species in our analysis than that are taxonom-
ically accepted as valid species in that genus), these
genera were included in the analysis but excluded
from trends in completeness. We examined the
data for ambiguous taxa (those that were identified
with “aff.,” “cf.,” “sp.,” or other indication of uncer-
tainty in the species-level identification), and the
Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946
by New Mexico State University Library user
on 28 July 2018
whole containing genus for any ambiguous taxon
was excluded from the analysis. If there was only
one sequence fragment for a genus, then that genus
was discarded.
Finally, we assessed the taxonomic completeness
of genus-level groups with available sequence data.
We compared the number of species within each
included genus to established lists of genus species
richness from authoritative taxonomic databases as
used in our previous research compiled in 2015
(Sigwart et al. 2018). These figures for species rich-
ness may exclude a small number of new species
named in recent years, but this is unlikely to have
any systemic impact on results.
Alignment files for each taxonomic order were
then split into subsets for each genus (based on
the first element of the sequence name, the genus
ID inherited from the original record data). We cal-
culated a pairwise-distance matrix comparing all
fragments in each genus using the function dist.a-
lignment in the SeqinR package (Charif and Lobry
2007). Gaps (i.e., missing data at either end of these
alignments) were not counted in the distance iden-
tity measure. Note that this function returns a ma-
trix of square root values for percent difference;
these were squared to report percent-differences.
This returned the following data for each genus-
level group: genus name, number of species in-
cluded, number of sequences in analysis, median
percent difference from pairwise comparisons, and
maximum pairwise difference.
We compared genetic distance per genus into four
metrics: genus size (species richness), taxonomic
completeness (proportion of valid species included),
species sampled (the number of species included per
genus), and sequence coverage (number of available
sequence fragments per genus). Comparisons were
visualized using the R package GGally (Schloerke
et al. 2017). We examined patterns of genetic dis-
tance data into four partitions: bivalves, other sam-
pled invertebrates (not a natural group), passerine
birds, and non-passerine birds (as the order
Passeriformes comprises more than half of living
bird species richness). Spearman’s rank correlation
coefficient was calculated for each comparison, us-
ing log-10 transformed data for species richness,
species sampled, and sequences sampled.
Additional statistics were not pursued because of
low correlation values in almost all cases, inappli-
cability of normality criteria, and intrinsic inter-
dependencies in the potential predictors (e.g., num-
ber of species sampled and taxonomic completeness
are capped by the species richness of a sampled
genus).
 The biodiversity minority majority
Table 1 Comparative information for data used in the present analysis in birds (data for Passeriformes are given separate to other
orders, as they are illustrated separately in Fig. 3), bivalves, and additional hard-shelled invertebrate taxa (chitons, bryozoans, and
ophiuroids, see the “Methods” section)
Sequences Species Genera
Animal group sampled sampled sampled
Non-passerine birds 4644 861 276
Passerine birds 8052 1640 517
Bivalves 20,928 1334 290
Chitons 1505 100 32
Bryozoans 530 61 30
Ophiuroids 1038 88 41
The number of genera sampled is based on genus identification provided by the data repository (BOLD), this number excludes genera removed
because they were represented by a single fragment (and therefore no distance metric could be calculated) or included species with ambig-
uously identified species.
Results
This analysis used a total of 1223 animal genera,
represented by 39,644 DNA sequence fragments
(Table 1). As expected, the availability of sequence
fragments was very heavily right skewed, with the
majority of genera having few sequences available
and a few taxa with many sequences. On inspection
of selected data, we observed that the distribution of
pairwise comparison values is often multimodal, per-
haps resulting from a combination of data sources. A
dataset at the genus level could include multiple spe-
cies and/or sites, and local population-level studies
that may produce a cluster of lower distance values
with many measurements from a single species or
locality. An arithmetic mean is not an appropriate
measure of central tendency for the genus genetic
size, and the median value of the distance matrix
was taken as the primary descriptive statistic; al-
though the median does not capture any description
of frequency distribution, it is less biased.
Pairwise distances per genus occupied a much
broader range of values in bivalves and other inver-
tebrate taxa when compared with those of birds
(Figs. 2 and 3). Pairwise distance values were not
strongly correlated with the other factors examined,
although these were nominally statistically signifi-
cant in all cases of birds, the highest correlation co-
efficient was <70% which cannot be considered as a
strong indicator of a potential causative relationship.
The data do suggest a strong positive relationship
between genus-level genetic distance and both spe-
cies richness and sampling density; larger genera and
larger datasets have greater maximum pairwise dis-
tances. This does not account for taxonomic
completeness—here, a genus is only present or
Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946
by New Mexico State University Library user
on 28 July 2018
5
Approximate global species
richness (nearest 1000)
Total valid living Sequences per
genera (approx.) genus (median) Described Estimated
980 7 4000 4000
1300 7 6000 6000
500 12 9000 10,000–20,000
70 12.5 1000 –
400 3 4000 –
300 4 2000 –
absent, without considering the fraction of its known
species that are present in the dataset. Genera with
fewer species are inherently more likely to be
completely sampled (e.g., a genus with only one spe-
cies must have 100% of species sampled, if it is in-
cluded at all). Thus the non-correlation or even
slightly negative correlation of these distance metrics
with taxonomic completeness does not refute the
larger pattern of larger genera with greater distances.
Importantly, maximum value for pairwise distances
among sequences in a genus could easily be skewed
by the inclusion of even a single misidentified sam-
ple. Median values for these data are probably more
informative.
Median values for pairwise distances indicate two
important findings. First, median genetic distances
are larger in genera that contain more species, and
increase with both species richness and the number
of species sampled. Second, median distance meas-
urements are not affected by increasing sample sat-
uration; the number of sequences sampled is not
significantly correlated with median genetic distance
per genus in bivalves (P ¼ 0.70) or other inverte-
brates (P ¼ 0.64) and only weakly correlated with
increasing sampling in birds (Spearman’s q ¼ 0.46).
Discussion
Diversity, divergence, and expectations
It was not our expectation that birds and bivalves
have very much in common, and these two major
groups were deliberately selected as contrasting ani-
mal groups with coincidentally similar species and
genus richness. The pitfalls of dependency on DNA
barcodes have been highlighted in many previous
 6 J. D. Sigwart and A. Garbett

Maximum pairwise distance 50%

40%
per genus (COI)

30%

20%

10%

0%

50%
Median pairwise distance

40%
per genus (COI)

30%

20%

10%

0%
1 10 100 0 0.50 1 1 5 10 1 10 100 1000
species richness taxonomic completeness species sampled sequences sampled
per genus
Maximum

non-passerines: non-passerines: non-passerines: non-passerines:

ρ = 0.65 (P < 0.001) ρ = -0.22 (P < 0.001) ρ = 0.75 (P < 0.001) ρ = 0.60 (P < 0.001)
passerines: passerines: passerines: passerines:
ρ = 0.61 (P < 0.001) ρ = -0.22 (P < 0.001) ρ = 0.78 (P < 0.001) ρ = 0.64 (P < 0.001)
per genus

non-passerines: non-passerines: non-passerines: non-passerines:

Median

ρ = 0.62 (P < 0.01) ρ = -0.20 (P = 0.0012) ρ = 0.72 (P < 0.001) ρ = 0.46 (P < 0.001)
passerines: passerines: passerines: passerines:
ρ = 0.55 (P < 0.01) ρ = -0.19 (P < 0.001) ρ = 0.72 (P < 0.001) ρ = 0.44 (P < 0.001)

Fig. 2 Measures of genetic distance per genus, as correlated to other taxonomic attributes of the genus (i.e., each plotted circle
represents one genus). The metrics are either the maximum (top) or median (lower) pairwise genetic distance for COI fragments
available for all genera of birds. The distance metrics are compared with species richness (the number of taxonomically valid species
per genus), taxonomic completeness (number of species sampled per genus, as a percent of valid species), species sampled (count of
species included per genus in the analysis), and sequences sampled (number of data points as barcode sequences per genus). Genera in
the order Passeriformes are solid, darker circles; other non-passerine genera are lighter, open circles. Spearman’s rank correlation
coefficients for each plot are given in the lower half of the figure.

studies, in particular that specific markers perform
differently across animal groups, and that is illus-
trated by the present results (Rubinoff 2006; Kress
et al. 2015). The standards developed in the context
of one group of organisms may not be transferable
to any other group.
There are very different patterns in genetic dis-
tance at the genus level between birds and the inver-
tebrate animals sampled, and the difference is
sufficiently stark that it is useful to consider similar-
ities as well as differences between the clades and the
data analyzed. Genetic distances in bird genera fol-
low striking and well-resolved patterns with correla-
tions between genetic distances and other factors.
The measured genetic distances per genus are more
constrained in birds, with a maximum pairwise dif-
ference of around 20% per genus compared with up
to 50% divergence within bivalves genera. In light of
the lack of any tidy correlations of our measured
factors with genetic distances in bivalves or other
invertebrates, it is tempting to consider explanations
that would excuse the pattern, based on an assump-
tion that the more constrained patterns observed in
birds is the “correct” result. Perhaps the identifica-
tion or definitions of genera were incorrect in these
invertebrates, species were incorrectly identified,
sequences were contaminated, too few genera were
included, or species richness within genera insuffi-
ciently sampled. We will consider each of these issues
below.
On the issue of sampling available species or ge-
nus diversity, representation appears to be largely
comparable and equivalent between birds and the
invertebrate sequences analyzed (Table 1); in both
cases, they represent mostly low numbers of sequen-
ces per species or per genus and low sampling

Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946

by New Mexico State University Library user
on 28 July 2018
 The biodiversity minority majority 7

Maximum pairwise distance

50%

40%
per genus (COI)

30%

20%

10%

0%

50%
Median pairwise distance

40%
per genus (COI)

30%

20%

10%

0%
1 10 100 0 0.50 1 1 5 10 1 10 100 1000
species richness taxonomic completeness species sampled sequences sampled
per genus
Maximum

bivalves: bivalves: bivalves: bivalves:

ρ = 0.29 (P < 0.01) ρ = 0.14 (P = 0.19, N.S.) ρ = 0.54 (P < 0.001) ρ = 0.4 (P < 0.01)
invertebrates: invertebrates: invertebrates: invertebrates:
corr = 0.4 (P < 0.001) corr = -0.15 (P = 0.18, N.S.) corr = 0.67 (P < 0.001) corr = 0.36 (P < 0.001)
per genus

bivalves: bivalves: bivalves: bivalves:

Median

ρ = 0.42 (P < 0.001) ρ = -0.10 (P = 0.37, N.S.) ρ = 0.47 (P < 0.001) ρ = -0.03 (P = 0.70, N.S.)
invertebrates: invertebrates: invertebrates: invertebrates:
ρ = 0.22 (P = 0.046) ρ = -0.03 (P = 0.81, N.S.) ρ = 0.53 (P < 0.001) ρ = −0.05 (P = 0.64, N.S.)

Fig. 3 Measures of genetic distance per genus, as correlated to other taxonomic attributes of the genus (i.e., each plotted circle
represents one genus) and metrics are all as in Fig. 2). Bivalve genera are solid (teal) circles; select other invertebrates are black, open
circles. Spearman’s rank correlation coefficients for each plot are given in the lower half of the figure.

density of known species. Higher density sampling is
unlikely to change the pattern recovered in bivalves.
Other explanations for the divergent patterns in
bivalves in birds—misidentification, contamination—
would imply substantial, systemic problems across a
very large scale of collective research on bivalves and
marine invertebrates (which some invertebrate work-
ers may suspect is the case) and also require the un-
realistic absence of equivalent errors from anyone
working on birds.
The source data for these analyses were gleaned
from public databases, and the species names on
sequences in these repositories often lead later work-
ers to identify their own samples by comparison.
Early errors lead to incorrect sequence-based identi-
fication, false confidence, and a cascade of later
problems (Morton 2018). It should be noted that
this problem is not isolated to sequence data, but
voucher specimens are prerequisite for later compar-
isons and often not available for published sequences
(Bortolus 2008). The present analysis excluded
genera that contained only a single sequence, or
any sequences that had ambiguous species-level iden-
tification. Uncertainty at the species level indicates
neither uncertainty of genus-level identification nor
the identification of other species in the genus, but
our conservative approach removed relatively few
genera. We did examine data from excluded genera,
and there was no evident pattern of genetic distance
with respect to the presence of ambiguous taxa in a
genus, except that the broadest invertebrate samples,
all genera with more than 15 species sampled in a
genus, contained uncertain identifications. The same
was not true of birds. This could indicate that
among the sampled invertebrates, larger samples of
species richness are an indicator of research concen-
tration on areas of taxonomic or phylogenetic
uncertainty.
Resources like GenBank are anecdotally rife with
misidentifications, at the species level, genus level,
and above, and data integrity protocols or user mo-
tivation may create barriers to correcting

Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946

by New Mexico State University Library user
on 28 July 2018
 8 J. D. Sigwart and A. Garbett
misidentifications discovered after publication
(Morton 2018). It is reasonable to assume that these
errors occur more frequently for invertebrate taxa
than for birds. However, it is our view that misiden-
tification represents a minority (possibly a large mi-
nority); certainly not all of the sequences in GenBank
are misidentified. It is relevant, then, that there is no
pattern in our analysis of bivalve genera that suggests
the presence of identifiable outliers (Fig. 3), which
would be expected if some genera included misiden-
tified sequences, and thus diverged from an underly-
ing pattern. Although data gaps and uncertainty
problems are substantial barriers for understanding
non-vertebrate diversity, a larger issue may be uncon-
scious expectations of equivalency among highly di-
vergent groups.
Genus size and the biodiversity minority majority
DNA barcoding techniques are increasingly used in
combination with classic taxonomy, adapting to the
growing needs for species classification (Purty and
Chatterjee 2016). Species identification and delimita-
tion are often the primary focus to tackle the added
demand for measuring and monitoring global biodi-
versity. Although species level identification is un-
doubtedly a vital measure of diversity, it is often
not feasible to refine identifications to species level;
for example, in fossil taxa, or due to financial con-
siderations, prioritizing efforts in large scale studies,
or differentiating morphologically similar species in
the field (Heino and Soininen 2007; Mazaris et al.
2008; Bertasi et al. 2009). Working toward cataloging
all living global diversity may require resorting to
higher ranked taxonomic levels, such as genera, as
a substitute for species.
Genus level genetic variety, which is relevant to
the overarching goals of biodiversity inventories,
remains understudied. Species-level genetic thresh-
olds have been widely examined across a wide range
of evolutionary groups; for example, a 2% genetic
variance is widely acknowledged as an “acceptable”
species threshold. Although this may be adequate for
specific markers in specific groups, there is no foun-
dation to expect the same threshold even for one
particular marker that would have transferability
across organismal groups, or that one threshold
would apply to different markers. Methods for
assessments are disproportionately focused on taxo-
nomically well-known groups, and later adapted for
application to other less well-studied groups.
Variation among clades is an important factor:
what may work for one may not necessarily work
for another. The results presented here demonstrate
Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946
by New Mexico State University Library user
on 28 July 2018
varying genetic sizes at genus level among the differ-
ent evolutionary groups as well as different effects of
some metrics of interest.
Molecular phylogenetic studies frequently reveal
hidden diversity by identifying evolutionarily distinct
lineages. Yet the appropriate cut-off between what
constitutes distinct populations within a species ver-
sus distinct species is not always readily apparent. As
a result, the act of defining a population (or group
of closely related populations), and what warrants
species status is not clear cut. Divergence in genetic
markers may indicate separate thresholds at pop-
ulation or species level; the additional results here
suggest that median distances at the genus level in
COI are not affected by sample saturation and may
be useful in future studies. Traditionally species de-
limitation has relied heavily on morphological char-
acterization to distinguish taxa. However, this
method can fail to adequately reflect distinct evolu-
tionary lineages and may result in underestimates of
biodiversity (Knowlton 2000). Species delimitation is
especially challenging for organisms with highly con-
served morphology, the so-called cryptic species
(sensu Bickford et al. 2007). Consequently, modern
methods of species delimitation use objective criteria
and incorporate data from multiple sources includ-
ing ecological, geological, morphological, phyloge-
netic, and behavioral attributes (Yang and Rannala
2010).
Global species richness for birds is broadly ac-
cepted as around 10,000 species. Recent work posited
that the number could be much higher based on the
genetic identification (Barrowclough et al. 2016), but
is based much on the finer-scale analysis than could
be currently attempted on any other group of organ-
isms. Global species richness estimates for poorly-
studied groups are notoriously inaccurate, because
it is difficult to predict the data gap of undiscovered
species (Bebber et al. 2007), so it is very difficult to
predict the total global living species richness for
bivalves with any confidence. The broad estimate
of 10,000–20,000 bivalve species (e.g., Gonzalez
et al. 2015) is still in line with the broadest interpre-
tation of bird species richness, from 10,000 to 18,000
(Barrowclough et al. 2016).
Birds and bivalves represent similar extant species
richness but very different evolutionary histories.
Although the sampled invertebrates are all clades
that have deep histories and persisted through mul-
tiple past mass extinction events and birds radiated
more recently, they have similar numbers of living
species. Genera of bivalves likely represent a more
heterogeneous collection of rates of genotypic and
phenotypic evolution, differing level of extinction,
 The biodiversity minority majority
and more variety than found among a similar num-
ber of bird species. Nonetheless certain emergent
patterns in species richness are universal (Sigwart
et al. 2018; Fig. 1). Previous studies have attempted
to impose standard metrics on species and species
groups (e.g., Avise and Johns 1999). The results of
this study are further evidence as to why it is not
viable to extend such standardization across broadly
separated groups with different evolutionary histo-
ries, and emergent patterns in species richness and
clade size may not show straightforward correlation
with genetic diversity or divergence times.
The COI mtDNA fragment was championed as an
identification “barcode” because it is relatively fast-
evolving and thus reveals species-level differences.
This also means that divegences may become satu-
rated, especially in comparing individuals within
groups above the species level. Saturation could be
a factor in the high divergences seen in our sampled
invertebrates, but this does not necessarily mean that
those species or genera are misclassified.
Species evolution is a continuous, ongoing pro-
cess, and the process of speciation—one
population-lineage splitting into daughter lineages
with independent evolutionary trajectories—is not
instantaneous. At any given point in time, different
lineages are de facto in different points of their on-
going cyclical process of speciation and equilibrium.
It is naturally often difficult to tell whether a species
complex is a group of incipient species that are re-
ticulate or overlapping, or rather a collection of very
distinct species at equilibrium but observed without
enough knowledge to differentiate them
(Dobzhansky 1935). This is not to say that species
are not real, but only that the species we observe at
this moment in time are not at all evolutionary equi-
librium. Much hope has been placed in species bar-
coding; however, given this macroevolutionary
paradigm there may be fundamental limitations to
the utility of barcoding to differentiate difficult or
cryptic species. Although thresholds for a percent
divergence in a given marker may be perfectly ap-
propriate for certain constrained clades, the same
approach may not be equally informative in other
organisms (Rubinoff 2006). In this context, it should
probably be not expected that an expanded ap-
proach, the genetic size of genera, would follow a
clear trend. It is rather remarkable that there are
trends like this for birds, as we found here, and these
should probably not be seen as the “correct” answer
or an aspiration for results from research on other
organisms.
In addition to the theoretical and practical impor-
tance of species status, issues of taxonomy and
Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946
by New Mexico State University Library user
on 28 July 2018
9
nomenclature also play a critical role in other areas
of biology. Estimates suggest that only about 10% of
the world’s biodiversity has been described (Mora
et al. 2011). Many taxonomic groups also have un-
dergone revisions in light of molecular phylogenetic
studies or additional discoveries, or are in need of
revision on the basis of the new information from
ongoing systematic research. This has absolutely be-
come routine in most invertebrate animals, whereas
in a well-studied clade like birds, the discovery of
new species or dramatic phylogenetic revision is in-
creasingly infrequent. However, among under-
studied groups, the minority–majority that repre-
sents most animal life, the pace of new species
descriptions has been described as glacial (Scotland
et al. 2003). Many species remain in a taxonomic
limbo due to the lack of formal description, and
this noted problem has been termed as the
“Linnean shortfall” (Brown and Lomolino 1998).
Although the seemingly arbitrary designator of a
name should not dictate management or conserva-
tion of a species, in reality name recognition can
have social, political, and legal influence
(Beheregaray and Caccone 2007). Ambiguous species
status can severely hamper conservation efforts for
threatened cryptic taxa (e.g., Niemiller et al. 2013).
Taxonomy is essential for conservation, but it is
equally important that species names are recognized
as scientific hypotheses subject to the same rigorous
testing and ongoing refinement as any other branch
of science (Thomson et al. 2018).
Morphological diagnoses remain important in the
field, as the ultimate non-destructive sampling ap-
proach. In terms of the process of taxonomic iden-
tification and description, species are identified via a
total evidence approach, which reflects the totality of
their evolution and there is no single easy solution.
Even if an easy solution such as a barcoding marker
is effective in one group, it is unlikely to be a taxo-
nomic silver bullet.
Conclusions
Taxonomically well-studied groups provide aspira-
tional goals for systematics, phylogneteics, and evo-
lutionary biology. The wealth of knowledge on the
natural history of birds, for example, contributes to
the deep understanding of delimiting species and
genera in that clade. However, it may be that the
reason there is so much information and such com-
plete taxonomy of birds is that birds are relatively a
very constrained group that follows unusually pre-
dictable evolutionary patterns. These patterns may
not be a norm that could form any basis for
 10
expectations in the evolution of other more diverse
organisms. Similar patterns may be found in other
vertebrates, but that represents a tiny fraction of real
animal diversity. Although there are emergent uni-
versal patterns in systematics, such as global size-
frequency of genera constrained across animals and
plant groups, these mathematical phenomena are
driven by deeper issues of evolution and not the
specific evolutionary trajectory of individual lineages
at the species or genus level. There are many evolu-
tionary pathways that lead a clade to a small genus—
combinations of both extinction and/or a lack of
diversification—but only one pathway to a large ge-
nus, through rapid radiation. The genetic patterns in
species evolution in one group thus do not serve as a
robust predictor for other groups; as evidenced here
by diversification in a single “barcode” fragment.
The utility of barcode fragments should be con-
sidered with caution—divergence in a single frag-
ment is not equally applicable to all animals, and
different species evolve at different rates so thresh-
olds should not be expected to be universally con-
sistent. Moreover, genetic distance approaches are
evidentially even less consistently applicable to spe-
cies groups, yet groups such as genera or families are
frequently the appropriate level of identification in
large scale biodiversity surveys. Any approach to
assessing biodiversity should work for most organ-
isms. Most genera are small, and moreover most
animal groups are under-studied. These are the mi-
nority majority of biodiversity. The most fascinating
questions in understanding evolution lie in the
source of comparative diversity and diversification.
Acknowledgments
This project would not have been possible without
the resources made freely available by the Barcode of
Life Data System (http://www.boldsystems.org/), and
we are grateful to all the contributors and partners to
that project. Four reviewers provided comments that
improved an earlier version of this article.
Funding
This work was supported by the European
Commission Horizon 2020 research and innovation
program under grant agreement No. H2020-MSCA-
IF-2014-655661, and the government of Northern
Ireland (Department of Employment and Learning).
References
Avise JC, Johns GC. 1999. Proposal for a standardized tem-
poral scheme of biological classification for extant species.
Proc Natl Acad Sci U S A 96:7358–63.
Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946
by New Mexico State University Library user
on 28 July 2018
J. D. Sigwart and A. Garbett
Barrowclough GF, Cracraft J, Klicka J, Zink RM. 2016. How
many kinds of birds are there and why does it matter?
PLoS One 11:e0166307.
Bebber DP, Marriott FH, Gaston KJ, Harris SA, Scotland RW.
2007. Predicting unknown species numbers using discovery
curves. Proc R Soc Lond B Biol Sci 274:1651–8.
Beheregaray LB, Caccone A. 2007. Cryptic biodiversity in a
changing world. J Biol 6:9.
Bergsten J, Bilton DT, Fujisawa T, Elliott M, Monaghan MT,
Balke M, Hendrich L, Geijer J, Herrmann J, Foster GN,
et al. 2012. The effect of geographical scale of sampling
on DNA barcoding. Syst Biol 61:851–69.
Bertasi F, Colangelo MA, Colosio F, Gregorio G, Abbiati M,
Ceccherelli VU. 2009. Comparing efficacy of different tax-
onomic resolutions and surrogates in detecting changes in
soft bottom assemblages due to coastal defence structures.
Mar Pollut Bull 58:686–94.
Bertrand Y, Pleijel F, Rouse GW. 2006. Taxonomic surrogacy
in biodiversity assessments, and the meaning of Linnaean
ranks. Syst Biodivers 4:149–59.
Bickford D, Lohman DJ, Sodhi NS, Ng PK, Meier R, Winker
K, Ingram KK, Das I. 2007. Cryptic species as a window on
diversity and conservation. Trends Ecol Evol 22:148–55.
Blackwell M. 2011. The fungi: 1, 2, 3 . . . 5.1 million species?
Am J Bot 98:426–38.
Blaxter M, Mann J, Chapman T, Thomas F, Whitton C, Floyd
R, Abebe E. 2005. Defining operational taxonomic units
using DNA barcode data. Philos Trans R Soc Lond B
Biol Sci. 360:1935–43.
Bortolus A. 2008. Error cascades in the biological sciences: the
unwanted consequences of using bad taxonomy in ecology.
Ambio 37:114–8.
Brown JH, Lomolino MV. 1998. Biogeography. Sunderland
(MA): Sinauer Associates.
Brown WM, George M, Wilson AC. 1979. Rapid evolution of
animal mitochondrial DNA. Proc Nat Acad Sci 76:1967–71.
Claramunt S, Cracraft J. 2015. A new time tree reveals Earth
history’s imprint on the evolution of modern birds. Sci
Adv 1:e1501005.
Chamberlain S. 2017. Bold: interface to bold systems API. R
package version 0.5.0.9213. (https://github.com/ropensci/
bold)
Charif D, Lobry JR. 2007. Seqin{R} 1.0-2: a contributed pack-
age to the {R} project for statistical computing devoted to
biological sequences retrieval and analysis. In: Bastolla U,
Porto M, Roman HE, Vendruscolo M, editors. Structural
approaches to sequence evolution: molecules, networks,
populations. Berlin: Spinger. p. 207–32.
Dobzhansky T. 1935. A critique of the species concept in
biology. Philos Sci 2:344–55.
Eo SH, DeWoody JA. 2010. Evolutionary rates of mitochon-
drial genomes correspond to diversification rates and to
contemporary species richness in birds and reptiles. Proc
R Soc Lond B Biol Sci 277:3587–92.
Federhen S. 2011. The NCBI taxonomy database. Nuc Acids
Res 40:D136–43.
Friedman M, DeSalle R. 2008. Mitochondrial DNA extraction
and sequencing of formalin-fixed archival snake tissue.
DNA Seq 19:433–7.
Gaston KJ, Williams PH. 1993. Mapping the world’s
species—the higher taxon approach. Biodivers Lett 1:2–8.
 The biodiversity minority majority
Gonzalez VL, Andrade SC, Bieler R, Collins TM, Dunn CW,
Mikkelsen PM, Taylor JD, Giribet G. 2015. A phylogenetic
backbone for Bivalvia: an RNA-seq approach. Proc R Soc
Lond B Biol Sci 282:20142332.
Gotelli NJ. 2004. A taxonomic wish-list for community ecol-
ogy. Phil Trans R Soc Lond B 359:585–97.
Grönholm T, Annila A. 2007. Natural distribution. Math
Biosci 210:659–67.
Heino J, Soininen J. 2007. Are higher taxa adequate surro-
gates for species-level assemblage patterns and species rich-
ness in stream organisms? Biol Conserv 137:78–89.
Hendricks JR, Saupe EE, Myers CE, Hermsen EJ, Allmon
WD. 2014. The generification of the fossil record.
Paleobiology 40:511–28.
Hebert PD, Cywinska A, Ball SL. 2003. Biological identifi-
cations through DNA barcodes. Proc Roy Soc Lond B
Biol Sci 70:313–21.
Higgins D, Rohrlach AB, Kaidonis J, Townsend G, Austin
JJ. 2015. Differential nuclear and mitochondrial DNA
preservation in post-mortem teeth with implications for
forensic and ancient DNA studies. PLoS One
10:e0126935.
Kazakova TB, Markosian KA. 1966. Comparison of physico-
chemical properties of mitochondrial and nuclear deoxyri-
bonucleic acid from rat liver cells. Nature 211:79–80.
Knowlton N. 2000. Molecular genetic analyses of species
boundaries in the sea. In: Sole-Cava AM, Russo CAM,
Thorpe JP, editors. Marine genetics. Berlin: Springer. p.
73–90.
Kress WJ, Garcıa-Robledo C, Uriarte M, Erickson DL. 2015.
DNA barcodes for ecology, evolution, and conservation.
Trends Ecol Evol 30:25–35.
Langley CH, Fitch WM. 1974. An examination of the con-
stancy of the rate of molecular evolution. J Mol Evol
3:161–77.
Larkin MA, Blackshields G, Brown NP, Chenna R,
McGettigan PA, McWilliam H, Valentin F, Wallace IM,
Wilm A, Lopez R, et al. 2007. ClustalW and ClustalX ver-
sion 2. Bioinformatics 23:2947–8.
Mazaris AD, Kallimanis AS, Sgardelis SP, Pantis JD. 2008.
Does higher taxon diversity reflect richness of conservation
interest species? The case for birds, mammals, amphibians,
and reptiles in Greek protected areas. Ecol Indic 8:664–71.
Miller MA, Pfeiffer W, Schwartz T. 2010. Creating the
CIPRES Science Gateway for inference of large phylogenetic
trees. In: Proceedings of the Gateway Computing
Environments Workshop (GCE), 14 November 2010,
New Orleans (LA): IEEE. p. 1–8.
Mora C, Tittensor DP, Adl S, Simpson AG, Worm B. 2011.
How many species are there on Earth and in the ocean?
PLoS Biol 9:e1001127.
Downloaded from https://academic.oup.com/icb/advance-article-abstract/doi/10.1093/icb/icy076/5058946
by New Mexico State University Library user
on 28 July 2018
11
Morton B. 2018. Fake new. Mar Pollut Bull 128:396–7.
Niemiller ML, Graening GO, Fenolio DB, Godwin JC, Cooley
JR, Pearson WD, Fitzpatrick BM, Near TJ. 2013. Doomed
before they are described? The need for conservation
assessments of cryptic species complexes using an amblyop-
sid cavefish (Amblyopsidae: Typhlichthys) as a case study.
Biodivers Conserv 22:1799–820.
Purty RS, Chatterjee S. 2016. DNA Barcoding: an effective
technique in molecular taxonomy. Austin J Biotechnol
Bioeng 3:1059.
Ratnasingham S, Hebert PD. 2007. BOLD: the barcode of life
data system. Mol Ecol Notes 7:355–64.
Raup DM, Sepkoski JJ Jr. 1986. Periodic extinction of families
and genera. Science 231:833–6.
R Core Team. 2017. R: a language and environment for sta-
tistical computing. R Foundation for Statistical Computing.
Vienna, Austria (https://www.R-project.org/).
Rubinoff D. 2006. Utility of mitochondrial DNA barcodes in
species conservation. Conserv Biol 20:1026–33.
Schloerke B, Crowley J, Cook D, Briatte F, Marbach M,
Thoen E, Elberg A, Larmarange J. 2017. GGally: extension
to ‘ggplot2’. R package version 1.3.2 (https://CRAN.R-proj-
ect.org/package¼GGally).
Schwarz C, Debruyne R, Kuch M, McNally E, Schwarcz H,
Aubrey AD, Bada J, Poinar H. 2009. New insights from old
bones: DNA preservation and degradation in permafrost
preserved mammoth remains. Nucleic Acids Res
37:3215–29.
Scotland R, Hughes C, Bailey D, Wortley A. 2003. The Big
Machine and the much-maligned taxonomist. Syst
Biodivers 1:139–43.
Sigwart JD, Sutton MD, Bennett KD. 2018. How big is a
genus? Towards a nomothetic systematics. Zool J Linn
Soc 183:237–52.
Stöger I, Sigwart JD, Kano Y, Knebelsberger T, Marshall BA,
Schwabe E, Schrödl M. 2013. The continuing debate on
deep molluscan phylogeny: evidence for Serialia
(Mollusca, Monoplacophora). BioMed Res Int 2013:1.
Strand M, Panova M. 2015. Size of genera—biology or tax-
onomy? Zool Scr 44:106–16.
Thomson SA, Pyle RL, Ahyong ST, Alonso-Zarazaga M,
Ammirati J, Araya JF, Ascher JS, Audisio TL, Azevedo-
Santos VM, Bailly N, et al. 2018. Taxonomy based on
science is necessary for global conservation. PLoS Biol
16:e2005075.
Yang Z, Rannala B. 2010. Bayesian species delimitation using
multilocus sequence data. Proc Natl Acad Sci U S A
107:9264–9.
Yule GU. 1925. A mathematical theory of evolution, based on
the conclusions of Dr. J. C. Willis, F.R.S. Philos Trans R
Soc Lond B Biol Sci 213:21–87.