Cladistics 13, 67–79 (1997)
WWW http://www.apnet.com
Vicariance Events, not Areas, Should be Used in
Biogeographical Analysis
P. Hovenkamp
Rijksherbarium/Hortus Botanicus, PO Box 9514, NL-2300 RA Leiden, The Netherlands
Accepted 15 December 1996
Biogeographical analysis addresses two distinct problems: the reconstruction of Earth history and of taxon
history. For the reconstruction of Earth history, an analogy between taxa and areas is often assumed. Uncritical
acceptance of this analogy has led to methods for the
reconstruction of taxon history being inappropriately
used for the reconstruction of Earth history. A biogeographical protocol is proposed which does not rely on
this analogy. This protocol resolves the historical
sequence of vicariance events, not the relations between
areas. The protocol is illustrated with the analysis of
three examples. © 1997 The Willi Hennig Society
analysis (Wiley, 1988; Brooks, 1990; Brooks and
McLennan, 1991). Curiously, opponents of this
approach use the same analogy in the opposite direction, applying three-item analysis, apparently first
developed as a biogeographical tool (Nelson and Ladiges, 1991a;b), to taxon analysis (Nelson and Platnick,
1991).
The analogy between taxa and areas has been so
pervasive that it has obscured the fact that historical
biogeography has two very distinct aims that should
be clearly separated, both conceptually and methodologically.
INTRODUCTION
EARTH HISTORY VERSUS TAXON HISTORY
PROTOCOLS
Much of the recent work on biogeographical methods can be summarized as a search for a “protocol
corresponding in as many ways as possible to phylogenetic systematics” (Kluge, 1988). Just as phylogenetic
analysis reconstructs the relationships of taxa, biogeographical analysis has been used to reconstruct the
relationships of areas. Thus, cladograms of areas are
constructed using taxa as characters. In this approach,
it has been taken for granted that there is a strong
analogy between phylogenetic and biogeographical
One of the aims of historical biogeography is to elucidate Earth history on the basis of biological evidence.
The other is to elucidate the history of one or more
groups of organisms on the basis of an established theory of Earth history. This second aim corresponds most
closely to “traditional” biogeography, and also to
Brooks Parsimony Analysis (BPA), as it is formulated
in, for example, Brooks and McLennan (1991). The two
approaches require different data and different
methodologies.
0748-3007/97/010067+13/$25.00/cl970032
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
67
68
Hovenkamp
For the first approach (Earth history) we need at least
two phylogenies of groups inhabiting the same areas.
We then search for any correspondences between the
two phylogenies, and explain these by reference to a
common element, which is usually interpreted as Earth
history. A unique pattern in a phylogeny is no evidence for a common element, and is therefore
irrelevant for this type of analysis. This approach corresponds closely to the approach developed by Rosen
(1978), and to what is usually termed “component
analysis” (Nelson and Platnick, 1981; Brooks, 1990;
Page, 1990). A variety of methods are available to identify the common element of a set of cladograms
(reduced area cladograms, Rosen, 1978; “consensus”
area cladograms; “reconciled trees”, Page, 1990).
The second approach (taxon history) can operate on
one phylogeny at a time, first identifying those elements in the phylogeny which can be explained by
reference to an established Earth history. The remaining elements can then be explained by reference to
other phenomena, such as dispersal (in the form of area
expansion or migration), extinction (resulting in area
contraction), or stasis (lack of speciation). BPA (Brooks
and McLennan, 1991) implements this approach using
methods identical to those used in phylogenetic
analysis.
The “Earth history” approach compares different
cladograms, and identifies the element common to all
cladograms. The “taxon history” approach also compares different cladograms and identifies the elements
common to both, but here one is a taxon cladogram,
and the other one is a representation of area history. At
first sight, therefore, the two methods are superficially
similar. In fact, the methods are so similar that the distinction has not been appreciated by most authors.
Rosen (1978) opposes his own (Earth history) approach
to Mayr's (taxon history) aim of “explaining the separate history of each (faunal) element”, (Mayr, 1965,
cited by Rosen). Humphries, in turn, (1992: 147) condemns Rosen (1978) for “deleting non-congruent
elements of different patterns”. In this exchange of
views, Rosen judges a taxon history approach by
standards derived from an Earth history approach and,
in turn, is judged by standards derived from a taxon
history approach. As the methods are not equivalent,
both criticisms are misguided. A “taxon history”
approach seeks to fit biotic diversity to the matrix of
known Earth history. Although it is dependent on
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
what is known about Earth history at any one time,
that does not make it “ad hoc” or unscientific, but
merely subject to normal scientific progress. On the
other hand, for an “Earth history” approach, incongruent elements occurring in only one cladogram are
simply irrelevant, and when they are deleted, no information is “thrown away”. A unique event is poor
evidence for a general process and, when an alternative explanation is available (for example, dispersal),
its evidential value is negligible. It is only in a “taxon
history” approach that all elements are equally important and need to be explained. As the two methods
have fundamentally different aims, the perceived
opposition is spurious.
Confusion is compounded by Brooks and McLennan
(1991), who, in their exposé of BPA (p. 206) clearly
describe it as a “taxon history” approach, and, at the
same time, as a logical continuation of methods developed for an “Earth history” approach (pp. 195–197).
Most of the examples they cite use an “Earth history”
approach (for example, p.226: Mayden, 1988; p.230:
Cracraft and Prum, 1988; p.238: Brooks, Thorson and
Mayes, 1981).
In this paper, I want to specifically question the
validity of the Earth History protocol. This protocol
uses the analogy between taxa and areas in a way that
obscures the real issues and makes inappropriate use
of the data. Although the analogy has been widely
used, its applicability has been doubted by several
authors (Cracraft, 1988; Sober, 1988; Page, 1990) who
have drawn attention to a number of problems.
AREAS AS TAXA?
In the first place, the taxon/area analogy presupposes the existence of “areas of endemism”. That,
however, should be a subject of investigation; it should
not be assumed. In phylogenetic analysis, both the
existence of taxa and their hierarchical relationship are
predicted (and therefore explained) by evolutionary
theory. As yet, there is no corresponding general
theory of biogeography which predicts the existence of
“areas of endemism” which are characterized by taxa.
It is even doubtful whether we have sufficient
observations that support the existence of such
areas. Whenever we try to match different taxon
69
Vicariance Events in Biogeographical Analysis
distributions, we usually find that no two taxa show
exactly the same distribution. Axelius (1991) has
pointed out that the correct way to code partially overlapping distributions is to consider the overlapping
region as a separate area. However, the number of such
“areas” rapidly increases when more than two different distributions are involved, and the separate “areas”
will be increasingly fragmented. How to define areas
that are relevant to more than one taxon is therefore
not a trivial problem, and will also be subject to scaling
problems.
represented as a cladogram. It is perhaps a theoretical
possibility that the entire history of the Earth is
divergent; nevertheless, geological data are overwh elmin gly in favour of reticulation as a real
phenomenon (although a world map realistic at first
sight can be derived solely by simulation of a large
number of “vicariance” events [Voss, 1988, pl. 1–6],
that is most likely no more than a coincidence). Reticulation of units cannot be expressed in the form of a
branching diagram.
TAXA AS CHARACTERS?
EVOLUTION OF AREAS?
A second point is that areas do not evolve like taxa.
Evolution of taxa is usually divergent, as full hybridization of taxa (with disappearance of the parental
species) is a rare process. If it occurs, then it is only for
taxa at or below the species level. This divergent history above the species level is reflected in the
hierarchical structure of relationships between species
and higher taxa, and can be graphically represented in
a branching diagram. This correspondence between
the shape of the true history and the shape of a cladogram is the reason we are confident that cladistics can
help us recover the history of taxa.
In contrast, the history of geographical areas is rarely
exclusively divergent. In many cases, the very opposite
is true, and we have good reasons to expect that reticulation involving formerly isolated areas is a common
process. We would be able to maintain that Earth history is divergent only if we assumed that even after a
collision of two land areas both areas retained their
separate identities for ever after. This would be a methodological fiction rather than a realistic assumption.
So far, cladistic biogeographers have failed to defend
convincingly the point that Earth history should be
represented as a branching diagram. Rosen's (1978)
defense is “…the underlying assumption that the history of all life has some generality with respect to the
history of the Earth's geography”. The fact that this
generality is not necessarily reflected in a cladogram of
areas is not addressed by Rosen nor by most other
investigators. Just because Earth history influences
life's history, and life's history is divergent, it is not necessarily true that Earth history is divergent and can be
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
Lastly, it is doubtful whether taxa can be treated the
same way in a biogeographical analysis as characters
are in a phylogenetic analysis. That they can be treated
in the same way is taken for granted in most accounts
of biogeographical theory. The statements corresponding most closely to an explicit defence are in Brooks
(1990) and Brooks and McLennan (1991). Brooks (1990:
29) states that “macro-evolutionary interpretation of
homology is the same, common history, at all levels of
analysis”. Brooks and McLennan (1991: 199) use a
slightly different wording: “the explanation for homology, common history, is the same for phylogenetic
analyses at all levels of biological organization”. By
adding “biological”, Brooks and McLennan make
more sense, but reveal that the defence misses the
point. It is not the use of parsimony analysis for different levels of biological organization that has to be
defended, but its use in a non-biological context, viz.
the history of areas.
When two taxa share a character, homoplastic similarity can be distinguished from synapomorphic
similarity by congruence with other characters. In a
phylogenetic analysis, the resulting hypotheses of
homoplasy or synapomorphy can then be tested independently by more detailed character analysis. In a
biog eog rap hica l an alysis, a similarity may be
non-homologous and yet irreducibly identical, the
identity deriving from horizontal transmission instead
of shared inheritance (Sober, 1988). In such cases, no
amount of further analysis will ever be able to disprove
the homology of the taxon-as-character. Whereas in a
phylogenetic analysis shared similarities are independently testable proof of shared ancestry, in
70
Hovenkamp
biogeographical analysis shared taxa are not proof of
shared ancestry and they are not independently testable as homologies.
Thus, the fundamental differences between characters in a taxon analysis and taxa in a biogeographical
analysis derive partly from the much greater vagility of
taxa as compared to characters.
The taxa that are used as characters for areas may
move freely between areas. Taxa may spread to other
areas. At each moment in history, they are constrained
only by the then-operative barriers, and not by the historical origin of the areas they inhabit. As there is no
necessary connection between the origin of the areas
and the existence or location of the current barriers,
therefore, there is no absolute correspondence between
the origin of a taxon and its current distribution. Moreover, as far as there is a connection, it may weaken as
time progresses: barriers once formidable may disappear. In contrast, characters of taxa are much more
strongly restricted to the taxon in which they originated and all its direct descendants; thus, there is a
very strong connection between the historical origin of
a character and its current distribution. As time
progresses, the barriers preventing the flow of characters to other taxa do not weaken, but are likely only to
become stronger.
In a cladistic analysis of relationships among taxa,
we are confident that we can use characters as indicators of history. As indicators of the history of areas,
taxa are inherently less reliable.
IN SEARCH OF AN ALTERNATIVE METHOD
OF BIOGEOGRAPHICAL ANALYSIS
If we want to reconstruct the history of areas on the
basis of the current distributions of taxa, we will need
to address the specific problems of “Earth history” biogeography that are not encountered in phylogenetic
analysis.
One way to do that would be to devise a method that
produces the complex type of pattern that corresponds
to the complexities of “area evolution”, thus allowing
both divergent and reticulate patterns. In view of this
complexity, a "'cladistic analysis” of areas is certainly
not sufficient, resulting, as it necessarily does, in a relatively simple branching diagram.
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
In this paper, I investigate a different approach.
Earth history is not to be resolved in terms of areas, but
in terms of vicariance events. Vicariance events are the
traditional core business of phylogenetic biogeography. They correspond to emerging barriers that
function as dispersal barriers. Compared to the reticulated complexities of the history of areas, which can
only be properly represented in a reticulated diagram,
a history of events is much simpler to represent. Once
all events have been identified, they can be aligned in
linear temporal sequence.
RECONSTRUCTING THE SEQUENCE OF
VICARIANCE EVENTS
To reconstruct the sequence of vicariance events
requires two types of data:
1. geographical information, specifying the location
of each vicariance and event; and
2. historical information, specifying the temporal
relation of each vicariance event to other vicariance
events.
For both types of data, we have the information
available in a cladistic analysis. The necessary
geographical information can be found in the distributional data for lineages. Each pair of mutually
exclusive sets of areas can be taken as evidence for a
vicariance event. Whenever a node in a cladogram
specifies two sister groups which do not overlap in distribution area (or the combined areas of all their
members), we have evidence for a vicariance event.
The position of the vicariance event is specified by the
two combined ranges: it is located in the area between
the two. Note that this position is not dependent on the
extent of the two ranges in all directions: away from
the putative vicariance event, the ranges may be extensive without influencing its position. The position of a
vicariance event is only ill defined under two conditions: if the two ranges overlap; and if the two ranges
are separated by wide gap which possibly corresponds
to a number of different vicariance events. Thus, this
method is relatively insensitive to widespread species.
It uses the best part of the information contained in a
71
Vicariance Events in Biogeographical Analysis
large distributional range; the part where a range is
constrained by an effective barrier to dispersal.
As evidence for the sequence of the vicariance
events, we have the information of the sequential order
of the cladogram nodes. This leads to a procedure
whereby a reconstruction of the sequence of vicariance
events can be attained by going through the following
steps:
1. Extract the Geographical Information from
the Cladograms
For each cladogram to be analysed, the taxa are
replaced by their areas, expressed as sets of recognized
unit areas. It is not necessary that these unit areas correspond to either areas of endemism or areas of single
species. For convenience’s sake it can often be arranged
that the boundaries between unit areas correspond
with boundaries between species areas, but again, this
is not necessary. In some cases it may even be convenient to define the unit areas as units of a completely
arbitrary grid system. The geographical information is
then extracted from the cladograms, by enumerating
all “traceable vicariance events” (TVEs). A traceable
vicariance event corresponds to a node of which both
descendants occupy mutually exclusive areas. Nodes
of which both descendants occupy overlapping areas
are considered as not traceable, and are excluded.
2. Extract the Historical Information from the
Cladograms
If only one traceable vicariance event is present in a
cladogram, there is no historical information present,
and the cladogram has to be discarded as historically
uninformative. The same applies if the TVEs in a cladogram occur on different branches. In that case, their
relative sequence cannot be specified. If two or more
TVEs are present as nodes on the same branch, the
order in which the nodes occur specifying these TVEs
is noted.
In large, mainly vicariant groups of organisms this
may result in a series of three or more TVEs, but in
smaller groups, the TVEs in a cladogram will be mostly
in ordered pairs.
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
3. Search for Independently Confirmed
Vicariance Events
To identify the vicariance events that are confirmed
by more than a single sister-group relationship, the
TVEs extracted from the different cladograms must be
compared and grouped together when they specify the
same vicariance events. Two TVEs can be grouped
together if two conditions apply:
1. there is no contradiction between the two nodes in
the TVE;
2. there is at least some overlap on both sides.
To group two TVE nodes, it is not necessary that the
area sets on each side of the nodes are completely identical. Range contractions and expansions may have
occurred, but most do not influence the result. However, to identify two nodes as specifying the same TVE
there has to be an overlap between the area sets on both
sides of the nodes.
The two TVEs are then combined into a “supported
vicariance event” (SVE), by combining the area sets of
the TVEs. Other TVEs can then be added to the SVE,
using the same criteria. As support for a SVE increases,
the number of areas in its sets may also increase,
thereby again attracting support from other cladogram
nodes.
The inherent vagility of species is here acknowledged in that two sets need only to overlap partially to
be included in a single, larger set. Thus, the range of
either set may have expanded or contracted without
much influence on the results.
After this has been done, TVEs which are left over
can be discarded. There is no reason to assume that
they refer to vicariance events, instead of to freak dispersal events involving only a single taxon.
4. Order the Supported Vicariance Events
On the basis of the relative ordering of cladogram
nodes the SVEs can be ordered in directed pairs or
series, corresponding to all the directed pairs or series
extracted from the cladograms. To do this, in all the
directed pairs or series of cladogram nodes, the nodes
72
Hovenkamp
can be substituted by the SVE which they are taken to
support.
Based on the sequences of SVEs, we can then try to
arrange the SVEs in a single sequence, which complies
with all individual sequences.
PH NG
BO
NG PH
3
CA
BO MP SU
2
4
1
(A)
(B)
LS JA SU
EXAMPLES
The possibilities of application of this protocol will
be demonstrated by three examples.
5
BO
CE PH
8
CA
7
6
(C)
Example 1: Pyrrosia
The three-area cladograms in Fig. 1 are taken from
Hovenkamp (1986), and represent three monophyletic
groups in the fern genus Pyrrosia with roughly allopatric distributions.
In the cladogram of Fig. 1A there are three TVEs.
Corresponding to node 1 is an event separating CA
from (BO PH NG). Corresponding to node 2 is an event
separating BO from (PH NG). Corresponding to node
3, there is an event separating PH and NG. The relative
order of these events corresponds with the order of the
nodes 1, 2 and 3. Table 1 lists the geographical information extracted from these cladograms; the relative
order of the nodes is listed in Table 2.
Comparing the cladograms of Fig. 1A and B, we can
combine the TVE nodes 3 and 5, which both specify a
separation between PH and NG. We can also combine
nodes 2 and 4, as there is a full overlap and no contradiction between the sets BO and BO+MP+SU on one
side, and PH+NG and PH+NG on the other. Translated back into actual events, the combination of nodes
2 and 4 implies that either range contraction (from
BO+MP+SU to BO) has occurred in Fig. 1A, or range
expansion (from BO to BO+MP+SU) in Fig. 2B. At this
stage of the analysis, there is no way to decide between
these two possibilities.
Comparing the cladograms of Fig. 1B and C, we can
see that nodes 4 and 7 are contradictory, as the area SU
is combined with BO in node 4, whereas in node 7 it
occurs in opposition to BO.
The supported vicariance events (SVE) from cladograms 1A, B and C are listed in Table 3. Substitution of
supported vicariance events (Table 3, leaving out the
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
FIG. 1. (A) A cladogram with four endemic taxa and one widespread taxon. (B) A cladogram with two endemic taxa and one
widespread taxon. (C) A taxon cladogram with two widespread and
two endemic species. (A, B, C after Hovenkamp, 1986). Abbreviations: BO, Borneo; CA, Continental Asia (regarded as a single area);
CE, Celebes; JA, Java; LS, Lesser Sunda Islands; MP, Malay Peninsula; NG, New Guinea; PH, Philippines; SU, Sumatra. Explanation
in text.
unsupported nodes), into Table 2 yields Table 4. The
shortest single series of events which is compatible
with all the sequences in Table 4 is:
B>C>A
Translated back into geographical information this
yields the following series of events:
1. (B): a split between Continental Asia and Malesia;
2. (C): a split between West Malesia (areas BO MP
SU LS JA) and East Malesia (Areas PH CE NG);
3. (A): a split between the Philippines and New
Guinea.
TABLE 1
Geographical Information from Cladograms in Fig. 1
Node
Areas of
Descendant 1
Descendant 2
1
2
3
CA
BO
PH
BO PH NG
PH NG
NG
4
5
BO MP SU
NG
NG PH
PH
6
7
8
CA
BO
LS JA SU
BO LS JA SU CE PH
LS JA SU CE PH
CE PH
73
Vicariance Events in Biogeographical Analysis
TABLE 2
TABLE 4
Historical Information from Cladograms of Fig. 1
Cladogram
Historical information
1A
1B
1C
1 > 2, 2 > 3, 1 > 2 > 3
4>5
6 > 7, 7 > 8, 6 > 7 > 8
Historical Information Contained in Fig. 1 Showing Supported
Vicariant Events only
Node 7, in this case, has no geographical explanation. As it is not supported by another node, there is no
evidence that it is part of a general historical pattern.
However, there is more than a single way to combine
the nodes into sets. Instead of combining three nodes
from Fig. 1 into the set (2, 4, 8), we could also combine
nodes 2 and 7, thus forming two sets, (4, 8) and (2, 7).
In this case, node 2 would seem to fit better in the set
(4, 8), as it is completely subsumed by node 4.
Example 2: Amazonian Biogeography
Cracraft (1988) presented an example where a
number of conflicting area cladograms result in two
conflicting general area cladograms. He noted that neither parsimony analysis nor consensus analysis could
resolve the conflict between these two general area
cladograms in a non-trivial way.
The five different area cladograms from his figures 2
and 3 are presented in Fig. 2. Tables 5–7 present an
analysis following the protocol outlined above. The
shortest single sequence which is compatible with all
the historical information is:
TABLE 3
Supported Vicariant Events in Fig. 1 and their Support
SVE
Supported
by nodes
Areas of
Descendant 1
A
B
PH
CA
C
D
BO MP SU LS JA
BO
Descendant 2
NG
BO PH NG LS JA
SU CE
PH NG CE
LS JA SU CE PH
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
3, 5
1, 6
2, 4, 8
7
Cladogram
Order of VEs
1A
1B
1C
B>C, C >A, B >C >A
C >A
B>C
2>3>1>2
This sequence of vicariance events is presented in
Fig. 3. This is closely similar to the solution presented
by Cracraft (his figure 7).
Example 3: Mesoamerican Biogeography
Rosen (1978), in his analysis of the Middle American
fish genera Xiphophorus and Heterandria, has presented
the classical benchmark for biogeographical analysis.
His biogeographical data are here reproduced as Fig. 4.
An analysis following the protocol described here is
presented in Tables 8–10.
When the vicariance events are then arranged
according to the order in Table 10, with the earliest at
bottom, a pattern of area splittings emerges which
closely corresponds to Rosen's pattern of residual
congruence:
Number
Events
1
8 < > 2, 3
2
9, 10 < > 2, 3, 7, 8
3
4–6 < > 2, 3, 7, 8–10
4
1 < > 2–5, 6, 7–10
The apparent differences between this sequence of
events and that implied by Rosen's solution are the
consequence of the different methods, and do not correspond to disagreement between the two sequences.
For example, in Rosen's solution area 6 does not occur.
The appearance of area 6 in the second vicariance event
here does not imply that areas 4, 5, and 6 are “closely
related”. It merely implies that area 6 is on one side of
vicariance event 2. There is no information available
about the moment area 6 was separated from the other
areas.
74
Hovenkamp
NE
SE
SW NW, Im
NE
SE
SW, NW
SE
Im
SW
3
8
2
5
1
7
4
(A)
NW
6
(B)
NE
SW
10
SE
NW
NW
11
(C)
NE
SW
SE
15
14
9
NE
13
12
(D)
(E)
FIG. 2. Five basic area cladograms, after Cracraft (1988), with areas occurring only once omitted. (A) His figure 2a, b; (B) his figure 2c; (C) his
figure 2d; (D) his figure 3a, b, c; (E) his figure 3c.
FORMALIZATION
The rough implementation of the protocol presented
here can resolve a variety of biogeographical problems
in a way that suggests that it is not only a solution for
problems like the one presented by Cracraft, but that it
TABLE 5
Nodes in the Area Cladograms Presented by Fig. 2 and the Corresponding Geographical and Historical Information
Node
Geographical information
Descendant 1
Descendant 2
Historical
information
could be useful as a general protocol for biogeographical analysis.
Attempts to formalize the protocol into a more
explicit procedure would have to address the following problems:
1. During the extraction of geographical information
from the cladograms, any overlap between the combined areas of two branches of a cladogram node
disqualifies that node as a TVE. This position more or
less takes the middle ground between a very restricted
approach where a vicariance event would be recognized only if it can be located very precisely between
two nearly contiguous areas, and a relaxed approach
which recognizes vicariance events also when there is
some degree of overlap between the areas.
Overlapping nodes may correspond to vicariance
events, but then some of the information required to
1
2
3
NE
SE
SW
SE SW NW Im
SW NW Im
NW Im
1 >2
2 >3
1 >2>3
4
5
NE
SE
SE SW NW
SW NW
4 >5
6
7
8
SE
Im
SW
SW NW Im
SW NW
NW
6 >7
7 >8
6 >7>8
TABLE 6
9
10
11
NW NE
NW
SW
SW SE
NE
SE
9 >10
SVE
13
14
15
NW NE
NW
SW
SW SE
NE
SE
13>14
13>15
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
Supported Vicariant Events in Fig. 2 and their Support
9 >11
Areas of
Descendant 1
1
2
3
SE
NW NE Im
NE
Supported by
nodes
Descendant 2
SW NW Im
SW SE
SE SW NW Im
2 5 6 11 15
3 8 9 13
1 4 10 14
75
Vicariance Events in Biogeographical Analysis
TABLE 7
6
1
9
4,5
10
7
8
2,3
Historical information in Fig. 2
Nodes
Vicariant events
9 >10, 13 >14
9 >11, 13 >15
1>2 >3
6 >8
2 >3
2 >1
3 >1 >2
1 >2
G
F
E
locate these events has been lost by subsequent dispersal of taxa. Similarly, information is also lost by local
extinction of taxa.
A formalized protocol would have to include a quantitative measure which specified how much overlap or
distance between areas is allowed in order to include
the node in the list of TVEs. It is doubtful whether such
a quantitative measure can be developed using objective criteria. It may be more practical to locate the
putative vicariance events on the basis of inspection of
maps, thus combining geographical information with
the information present in the cladograms.
2. During the combination of TVEs into sets specifying SVEs, different combinations of TVEs can be made
in many cases. A optimality criterion would be necessary to choose between the various possibilities.
Alternatively, one could decide to evaluate all possible
combinations. In larger analyses, this option could be
time consuming.
D
C
B
A
(a)
1
3
4,5,6
9,10
8
L
K
LIMITATIONS
J
There are some obvious limitations which apply to
this protocol.
2
2
>
2
I
H
3
3
>
3
>
>
1
1
>
2
2
FIG. 3. The sequence of vicariance events resulting from the analysis of Cracraft’s example (1988).
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
(b)
FIG. 4. The two area cladograms for Xiphophorus and Heterandria,
as given by Rosen (1978).
It is obvious that any n umber of con flictin g
two-event statements are compatible with a sequence
in which all the events are set in one order and its
76
Hovenkamp
TABLE 8
TABLE 10
Nodes in the Area Cladograms Presented in Fig. 4 and the
Corresponding Geographical and Historical Information.
Areas Numbered as in Rosen (1978)
Historical Information in Fig. 4 with Regard to Nodes and Vicariant
Events
Node
Geographical information
Descendant 1
Descendant 2
A
6
1–5, 7–10
B
C
D
E
F
G
1
9
4, 5
10
7
8
2–5, 7–10
2–5, 7, 8, 10
2, 3, 7, 8, 10
2, 3, 7, 8
2, 3, 8
2, 3
H
I
J
K
L
1
3
4–6
9, 10
8
2–6, 8–10
2, 4–6, 8–10
2, 8–10
2, 8
2
Historical
information
Nodes
Vicariant events
A>B >C >D>E>F >G
H>I>J>K>L
3 >4>2>1
3 >4>2>1
A >B>C >D >E
>F >G
H >I>J>K>L
reverse. For example, the sequence A > B > C > D >
E> D >C >B> A is compatible with both simple statements A > E and E>A, A >B and B> A, etc. But similarly
compatible is the series E>D > C > B>A> B>C>D > E.
Thus, when only simple statements relating two areas
are found in the analysis, the internal order of the
repeating series can not always be elucidated unequivocally. This requires longer statements. The statement
B>E> C, for instance, can distinguish between the two
different sequences above. Thus, the best results are
obtained with cladograms which contain longer
sequences of informative nodes. These may be difficult
to find. Crisci et al. (1991) list 18 cladograms, of which
only five provide sequences of three or more strictly
vicariant nodes (p. 158, their cladograms b, e, g, m, n).
As a consequence, when there is rampant dispersal
and sympatry in all lineages in the cladogram, analysis
becomes impossible. This is a limitation which applies
to all methods of biogeography, and it is important to
be aware that the most highly resolved results are not
necessarily the best. Consensus methods will produce
unresolved results even when a better resolution is
possible (when cladograms are conflicting), whereas
parsimony analysis will find highly resolved results
more often than warranted (as it makes no distinction
between supported vicariance and possibly non-vicariance events). A sequence protocol as outlined above
will not recognize vicariance events if they are not supported by more than one node in the study group.
At first sight, the results obtained by this protocol
may seem much simpler in form and less well resolved
than the area cladograms obtained by other methods.
However, in an area cladogram, the relative timing of
nodes (vicariance events) on different branches is not
reconstructed, even if, potentially, the information is
present in the data. Thus, this protocol may resolve
some aspects of geographical history that are not
resolved by cladogram-oriented protocols.
WIDESPREAD SPECIES AND ASSUMPTIONS
0, 1 AND 2
The protocol outlined here differs significantly from
other methods for biogeographical analysis in the way
that widespread species are treated. Consider the
example given by Nelson and Ladiges (1991a, their figure 1a, here reproduced as Fig. 5). In this example, we
TABLE 9
CA
Supported Vicariant Events in Fig. 4 and their Support
SVE
1
2
3
4
Areas of
Descendant 1
8
9, 10
1
4–6
Descendant 2
Supported by
nodes
2, 3
2, 3, 7, 8
2–5, 6, 7–10
2, 3, 7, 8–10
G, L
E, K
B, H
D, J
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
BO
SU BO
10
9
FIG. 5. A three-taxon cladogram with two endemic and one widespread species (after Nelson and Ladiges, 1991b), abbreviations as
in Fig. 1.
Vicariance Events in Biogeographical Analysis
have to deal with the conflicting evidence of one species occupying two areas (BO SU).
Other protocols for biogeographical analysis ask the
single question “how are the three areas related?”, and
then run into problems regarding the evidence for the
relationship of area BO. The protocol outlined here
asks two separate questions: “What evidence do we
have for vicariance events?”; and “What evidence do
we have for the history of vicariance events?”.
With regard to the first question, there are two possibilities by which this taxon-area configuration could
have been derived: taxon 1 and the ancestor of 2 and 3
could have originated after vicariance of areas BO and
CA+SU, while taxon 3 dispersed afterwards to area
BO.
Thus, nodes 1 and 2 would be listed:
(1) BO< > CA, SU
(2) CA < > SU
This corresponds to a resolution of this area cladogram with assumptions 1 and 2 (new implementation,
Nelson and Ladiges, 1991a). Alternatively, it is possible that the ancestor of 2 and 3 was still inhabiting areas
CA, BO and SU, and then the first node of the cladogram does not specify a vicariance event. In that case,
node 1 is not informative, and only node 2 would be
listed:
(1) CA < > SU, BO
This corresponds to the resolution of this area cladogram with the aid of assumption 0. In this case, we also
would have no reason to assume that SU represents a
relevant area at all. Under this interpretation, the
cladogram gives no historical information, and is here
regarded as uninformative.
Thus, the different solutions for this area cladogram
correspond to different ways in which the observed
distributions may have been reached. A choice for
either of the solutions (assumption 1, 2 or 0) requires
empirical data about the probability that dispersal has
occurred in a relatively unlikely direction (i.e. back
across the vicariance barrier that triggered an earlier
speciation event, which it would not have done if it
were not a formidable barrier to the species involved),
compared to the probability that non-vicariant speciation has occurred. Which of the two possibilities is
more likely in a particular situation can only be
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
77
decided by relevant data. The resolutions of nodes
with sympatric species thus involve empirical questions, and it is impossible to express a general
preference for one way of dealing with these problems.
When the results of this protocol agree with the solution offered by assumption 0, this does not mean that
this protocol takes an “assumption 0” position.
Assumptions 0, 1 and 2 are only meaningful in a protocol which takes the predefined areas for granted, and
then searches for a way to order the areas in a cladogram. This protocol does not assume the existence of
areas, nor does it result in a cladogram of areas. Any
correspondence between this protocol and assumption
0 is due to the fact that, like assumption 0, it does not
assume vicariance when there is no evidence for it in
the data.
In this protocol, it is accepted that any two vicariant
taxa can be informative about a vicariance event, even
if both are widespread. However, when there is overlap between areas of the two taxa (sympatry), support
for a vicariant speciation decreases as the amount of
overlap increases. When there is only a small overlap,
it will usually be acceptable to ignore the overlap if it
can be explained by chance dispersal of a few dispersal
units across the vicariance boundary. As the overlap
increases, such an explanation becomes increasingly
more difficult to accept. It will become increasingly difficult to decide whether the observed pattern is to be
explained by vicariance followed by dispersal or by
small-scale vicariance not resolved with the chosen
size of unit areas. Preference for one of these alternatives cannot be decided solely by inspecting the area/
taxa matrix. It must be based on some knowledge
about the relative position and distance of the areas,
and preferably also on knowledge about the dispersal
capacities of the species involved.
A third theoretical possibility which could resolve
this area cladogram is that the conflicting area (here
BO) is composed of two distinct smaller areas (here one
with Exemplum borneense, the other one with E. sumatranum). In that case, the unit areas chosen for the
analysis are too large. The speciation may have been
vicariant, but not at the level resolved by the analysis.
In order to assess the degrees of sympatry between
overlapping species, it is therefore important to use
basic units with sufficient resolution. An important
advantage of this protocol is that at one stage, the areas
of nodes supporting the same vicariance event are
78
Hovenkamp
1
2
3
4
5
6
B
1
2
B
7
13
A
A
19
(A)
(B)
FIG. 6. The effect of using a grid on the representation of the distribution of taxa. (A) Distribution of taxa A, B plotted on a representation of
areas 1, 2. Both taxa will be coded as occurring in areas 1 and 2, suggesting complete sympatry. (B) Distribution of taxa A, B, plotted on an arbitrary grid system with units numbered 1, 2 … 24. Taxon A now will be coded for areas 2, 7–9, 13–16; taxon B for areas 9–12. The sympatry is
now immediately seen to be very limited.
combined. Thus, the initial units need not correspond
with units that are recognized after the analysis. In
principle, the initial units can be arbitrarily small. This
makes it possible to use a grid for the representation of
the basic distribution data. The increased resolution of
a grid system with small units over a system with predefined geographical areas is demonstrated in Fig. 6.
The recognition that not all vicariance events may be
traceable renders much of the discussion about widespread species superfluous. It is simply not possible to
extract information about past vicariance events where
no information has survived. Once a species has dispersed substantially across the barrier involved in its
vicariant origin, information about this vicariance
event is irrecoverably lost.
original drive for the development of phylogenetic
biogeography was the insight that sister-group relationships in a cladogram may reflect vicariance events.
An appropriate method for the reconstruction of the
biogeographical history must therefore concentrate on
geographical boundary lines instead of the areas which
they delimit.
ACKNOWLEDGEMENTS
I thank Yde de Jong, Hubert Turner and Peter van Welzen for their
comments at various stages of the preparation of this manuscript.
REFERENCES.
CONCLUSIONS
The emphasis on the construction of area cladograms
in phylogenetic biogeography is misguided. It is based
on an inappropriate analogy between taxa and areas.
The uncritical application of phylogenetic methodology to biogeographical analysis results in attempts to
order elements that are irrelevant in a pattern which is
irrelevant, using criteria that are invalid. As a result,
biogeographical analysis is bogged down in discussions about widespread species, attempting to extract
information which simply is no longer present. The
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
Axelius, B. (1991). Areas of distribution and areas of endemism.
Cladistics 7, 197–199.
Brooks, D. R. (1990). Parsimony analysis in historical biogeography
and coevolution: Methodological and theoretical update. Syst.
Zool. 39, 14–30.
Brooks, D. R., and Mclennan, D. A. (1991). “Phylogeny, Ecology and
Behavior”. University of Chicago Press, Chicago and London.
Brooks, D. R., Thorson, T. B., and Mayes, M. A. (1981). Freshwater
stingrays (Potamotrygonidae) and their helminth parasites: Testing hypotheses of evolution and coevolution. In: “Advances in
Cladistics: Proceedings of the First Meeting of the Willi Hennig
Society”, (V. A. Funk, and D. R. Brooks, Eds), pp. 147–175.
Cracraft, J. (1988). Deep-history biogeography: Retrieving the historical pattern of evolving continental biotas. Syst. Zool. 37, 221–236.
Vicariance Events in Biogeographical Analysis
Cracraft, J., and Prum, R. O. (1988). Patterns and processes of diversification: Speciation and historical congruence in some
neotropical birds. Evolution 42, 603–620.
Crisci, J. V., Cigliano, M. M., Morrone, J. J., and Roig-Juñent, S. (1991).
Historical biogeography of Southern South America. Syst. Zool. 40,
152–171.
Hovenkamp, P. H. (1986). A monograph of the fern genus Pyrrosia.
Leiden Bot. Series 9.
Humphries, C. J. (1992). Cladistic biogeography. In “Cladistics. A
Practical Course in Systematics”, (Forey, P. L. et al.), Chapter 9.
Clarendon Press, Oxford.
Kluge, A. G. (1988). Parsimony in vicariance biogeography: A quantitative method and a greater Antillean example. Syst. Zool. 37,
315–328.
Mayden, R. M. (1988). Biogeography, parsimony and evolution in
North American freshwater fishes. Syst. Zool. 37, 329–355.
Mayr, E. (1965). What is fauna? Zool. Jahrb., System. 92, 473–486.
Nelson, G., and Ladiges, P. Y. (1991a). Standard assumptions for biogeographic analysis. Aust. Syst. Bot. 4, 41–58.
Copyright © 1997 by The Willi Hennig Society
All rights of reproduction in any form reserved
79
Nelson, G., and Ladiges, P. Y. (1991b). Three-area statements:
Standard assumptions for biogeographic anlaysis. Syst. Zool. 40,
470–485.
Nelson, G., and Platnick, N. I. (1981). “Systematics and Biogeography: Cladistics and Variance”. Columbia Univ. Press, New York.
Nelson, G., and Platnick, N. I. (1991). Three-taxon statements: A
more precise use of parsimony? Cladistics 7, 351–366.
Page, R D. M. (1990). Component analysis: A valiant failure?
Cladistics 6, 119–136.
Rosen, D. E. (1978). Vicariant patterns and historical explanation in
biogeography. Syst. Zool. 27, 159–188.
Sober, E. (1988). The conceptual relationship of cladistic phylogenetics and vicariance biogeography. Syst. Zool. 37, 245–253.
Voss, R. F. (1988). Fractals in nature: From characterization to simulation. In “The Science of Fractal Images” (Peitgen, H.-O., and D.
Saupe, Eds), Chapter 1. Springer, New York, Berlin, Heidelberg,
London, Paris, Tokyo.
Wiley, E. O. (1988). Parsimony analysis and vicariance biogeography. Syst. Zool. 37, 271–290.