Biology and Philosophy 18: 513527, 2003.

2003 Kluwer Academic Publishers. Printed in the Netherlands.

Explanatory coherence and empirical adequacy: The

problem of abduction, and the justification of
evolutionary models

SCOTT A. KLEINER
Department of Philosophy, University of Georgia, Athens, GA 30606, USA

Received 10 December 2001; revised 1 November 2002; accepted in revised form 2 August 2002

Key words: Abduction, Coherentism, Darwinian histories, Foundationalism, Speciation

Abstract. Foundationalist theories of justification for science were undermined by the theory-ladeness
thesis, which has affinities with coherentist epistemologies. A challenge for defenders of coherentist
theories of scientific justification is to specify coherence relations relevant to science and to show how
these relations make the truth of their bearers likely. Coherence relations include characteristics that pick
out better explanations in the implementation of abductive arguments. Empiricist philosophers have
attacked abductive reasoning by claiming that explanatory virtues are pragmatic, having no implications
regarding truth. However, empiricists basic beliefs are subject to the same challenges facing abduction,
both of which can be met by citing causally coherent etiologies, which are commonplace in biological
explanations, and by demonstrating the relevance of causal coherence to truth.

Introduction

Epistemic progress in science was once taken to be an accumulation of epistemically

justified statements, where justification occurs either as causally induced affirmation
of basic statements (observation statements), or by inference from basic premises to
non-basic statements (theoretical statements) using selected argument schemata
(deduction, induction, hypothetico-deduction). In most scientific disciplines, includ-
ing evolutionary theory, both classes of statement have certain semantic properties:
they are intended by their authors and are understood by audiences to be about parts
of the world other than the content of human consciousness. According to justifica-
tional foundationalism, truth value is introduced into our thought through justified
basic beliefs and transferred to theoretical beliefs with various degrees of reliability
by standard inferences sanctioned by deductive and inductive logic. In the 1960s
historicist philosophers (e.g. Hanson (1961), Kuhn (1962), Feyerabend (1962))
attacked this conception of basic statements and epistemic progress with claims that
observation statements are theory laden. No single set of basic sentences can serve
as the foundation of all science, nor do basic beliefs accumulate in the history of
science. Foundations appear as such only from local perspectives in the history of
science, so basic beliefs may stand at the foundations of Darwinism, but they differ
from those that stand at the foundations of pre-Darwinian creationism. The histori-
cists and many others have taken a foundationless science to imply an epistemic
514

cultural relativism, according to which epistemic standards are cultural norms that
in no way embody the need to represent external reality. I shall assume that this need
to achieve truth about external reality is fundamental in the lives of humans and
other sentient organisms and manifest themselves in the semantic properties of
fundamentally important human beliefs and statements. They also place an im-
portant constraint upon epistemic standards-that worthy epistemic standards for
justification render beliefs meeting those standards probably true.
Epistemological coherentism denies that there are basic beliefs underlying human
knowledge but does not deny the external reference of the knowledge that humans
seek. Coherentists hold that the justificatory relation between an observational belief
P and a non-observational belief H is symmetrical, so no belief is privileged in not
requiring further beliefs for its justification. This means that H can justify P as well
as be justified by P. Also H can undermine P as authoritatively as P can undermine
H. Thus the thesis that distinct theoretical schemata support distinct empirical
foundations is part of coherentism. Coherentism faces two problems which must be
dealt with before it can be taken seriously as a theory of scientific justification, (i)
explaining what the coherence relation should be in science-it cannot just be logical
consistencyand (ii) showing that propositions bearing the relation are likely to be
truelogical consistency clearly falls short in this respect.
Thagard (1992) proposes that superior explanatory coherence is the primary
justification for conceptual change in science. He addresses the first of these
questions (i) and the relativism issue in the history of science by claiming that the
history of science shows a succession of conceptual systems that progress epistemi-
cally by possessing ever increasing explanatory coherence. Thagard offers specific
principles for assessing explanatory coherence and its bearing upon the choice of
propositions to accept, which he then implements in a computer simulation of
certain episodes in the history of science. This simulation generates historical
outcomes that we all know: Copernican astronomy supercedes Ptolemaic as-
tronomy; Darwinian natural history supercedes creationist natural history, and so on.
This outcome suggests that the history of science warrants a specific kind of
coherentist theory of scientific justification, the claim that explanatory coherence is
of primary epistemic importance, a view that I shall call abductivism: Abductive
argumentation is argumentation to the best, or to progressively better explanations,
and the history of science contains a succession of ever better explanatory con-
ceptualizations. Darwinian natural history is epistemically superior to creationist
natural history because the former conceptualization offers superior explanations of
biological diversity, functional organization, embryological and taxonomic rela-
tions, and biogeographical distribution.
Abductive reasoning is now widely received by philosophers as the standard
means of justifying theories referring to unobservable processes. It is also the
principal argument offered in efforts to justify scientific realism: the reality of
unobserved processes is the best explanation of the success that scientists have in
designing and implementing interventions into natural processes. One rationale
behind abductivism is that we can trade off the epistemic assurance of direct
 515

empirical access for understanding via productive mechanisms, even though ele-
ments of the mechanisms are hidden from view due to their small size or their
remote location in space or time.
More generally, what renders one explanatory hypothesis superior to another are
explanatory virtues, features of an explanatory hypothesis that make it superior to its
rivals in the explanations it provides. The explanatory virtues cited by recent
philosophers (e.g. Laudan (1977), Kleiner (1985), Sober (1993), Ruse (1979,
1999)) include simplicity in its various forms, predictive power, the power to unify
diverse kinds of fact, coherence within the preferred explanatory theory, coherence
with received belief, and fruitfulness in generating new research problems. I regard
all of these virtues as subject to the two steps of epistemic vindication for
coherentism mentioned above, specification and demonstrated relevance to truth.
In demanding of explanatory virtues an epistemic link to truth one assumes that
ultimately it is truth that matters in scientific justification, and any epistemic virtue
that does not directly attribute truth must be vindicated by demonstrating that it is a
reliable means to the achievement of truth. Kitcher (1993) would call this position
robust realism. However, Kitcher brings up a caveat: We can have too much
justified information about the world, too many uninteresting true beliefs. If we
simply gather information about the world without limiting ourselves to the
information that is relevant to specific epistemic goals, we shall be overwhelmed
with trivial facts about pebbles on the beach, ant species in the kitchen, and so forth.
Scientific inquiry whose only goal is truth can easily be channeled into endless
unimportant investigations at considerable societal cost. Kitcher suggests that
another important epistemic regulator for scientific inquiry is the high valuation that
scientists place on unity and patterns of dependence in the world, which can be
considered among the list of explanatory virtues. In basic research, the empirical
problems worth solving, the facts worth establishing are those relevant to the
discovery and explanatory utilization of such patterns and unifying schemata.
It is possible to interpret explanatory virtues pragmatically, that is, as reflecting
human values, such as intelligibility and predictive power. Analogies with familiar
processes foster a sense of understanding the unfamiliar; predictive power enables
humans to benefit from reliable estimates of future happenings; Kitchers explanat-
ory unification enables humans to benefit by anticipating happenings they have not
experienced before. The robust realist can claim that embracing beliefs just on the
grounds that they are intelligible or usefully predictive is a form of wishful thinking.
We would wish that the world is intelligible and predictable, but the world may not
fulfill these wishes. A pragmatist caveat to robust realism similar to Kitchers
recommends that the whole truth be compromised in the effort to achieve intel-
ligibility or predictive power by introducing unrealistic simplifying assumptions
into our models of natural processes. Theoretical models that are completely
realistic may be inaccessible to humans, given their limited cognitive capacities and,
for example, organic structures that are bio-chemically enormously complex. Thus
theoretical models may contain simplifying assumptions that are not intended to
refer to actual processes, e.g. that variation at just one genetic locus is relevant to
516

natural selection at one stage in the evolution of a metazoan. Accordingly, pragmatic

values that either do not, cannot or need not have epistemic justification as means to
truth may have regulative force in scientific justification.
An extreme form of robust realism is that such pragmatic values never bear upon
and should never compromise the concern for achieving truth. Explanatory virtues
cannot be given epistemic justification and should have no bearing upon the
justification of models ultimately accepted for their explanatory power. Explanatory
models must be justified as semantically adequate, that is, their components must be
externally referential and empirically realized, before they can be considered
justified and utilized in addressing why-questions. Elisabeth Lloyd argues for this
position in her analysis of Darwins arguments in the Origin (1983): the usual
reading of Darwins defense of evolution by natural selection in the Origin
(including Thagards) is that his theory is abductively consilient and supported by
analogies with familiar processes by which domestic species are produced. Lloyd
interprets Darwins arguments as non-abductive appeals to empirical adequacy.
Elements of Darwinian patterns of explanation, such as heritable variation and the
effects of continual selection, are realized in circumstances accessible to observa-
tion, as evident records Darwin cites concerning the production of domestic breeds.
These empirical realizations provide the fundamentally important semantic link to
reality not provided by realizations of explanatory virtues. The semantic link
justifies beliefs that refer to the world by observable realizations that presumably
assure us of their truth. But Darwins theory is not limited to the observable
processes by which domestic varieties are produced. It also refers to unobservable
processes in the prehistoric past by which species and adaptations are produced.
Lloyd must concede that the justification of models referring to prehistoric processes
requires some kind of inference. Thus Lloyd transfers the burden of justification
from explanatory virtues to inductive or analogical argumentation. Her arguments
suggest that we return to an inductivism like that put forth by Newton. Explanations
must be in terms of vera causae (true causes), that is, causal processes that can be
directly observed or deduced from direct observations and rendered general by
induction. According to this position, abduction-the feigning of explanatory hypoth-
eses-puts the epistemic cart before the horse in assuming we can justify a belief by
showing that it has explanatory virtues. However, according to this form of
inductivism, such beliefs have no explanatory power without deductive or inductive
support from direct observation.
A defense of abduction by providing epistemic warrant to explanatory virtues will
require some clarification of what counts as an explanation in biology. Kitcher calls
the principal explanatory patterns in Darwinian biology Darwinian Histories, and
has acknowledged that these are essentially the same as Lloyds models. These
histories are not just chronicles, the recitation of events in temporal sequence with
little attention to their generation or causal entrainment. Darwinian histories are one
of a class of explanatory structures that I shall call causal etiologies. Causal
etiologies are sequences of events that are linked together much as the steps of a
computer program, a series productive elements (Dennet (1993), compares them to
algorithms) that possess specific starting conditions and outputs. If a series contains
 517

productive components so linked that the terminating conditions of one stage are the
initiating conditions of the next then the series will be an effective causal mecha-
nism for the generation of the output of its final stage. I shall call such etiologies
causally coherent. Citing coherent causal chains is common in biological explana-
tion, as in the progression of a disease, the replication of DNA, the expression or
inhibition of a molecular gene, as well as the generation of adapted form and
diversity in evolutionary theory. Explaining biological diversity for bisexual organ-
isms is applying to known initial conditions one or various etiological schemata or
model types which abstractly represent causally linked steps in a history whose
outcome is the occurrence of a genetically isolated population. Ideally, boundary
conditions under which a sequence of events forms a causally coherent etiology
should be specified. In practice, many such etiologies include ceteris paribus
qualifications that simply register that many of these conditions are unknown. Still,
only under the actual realization of these conditions will a causal etiology work, that
is, actually generate its output, such as a new species.
Causal coherence fills in a gap in Thagards account of scientific justification. He
does claim that explanatory relations are coherence relations, but he does not tell us
explicitly what explanatory relations are, beyond the deduction of actual happenings
from a law with initial and boundary conditions. It is well known that this deductive
nomological account of explanation is not applicable to biology, which is focused
upon organizational structures of entities having specific dispositions to action, not
the application of universal laws to initial and boundary conditions. In this essay I
shall illustrate the importance of causal coherence in scientific practice in this case
the discussion of allopatric models of speciation. I shall show how these cases of
explanatory coherence realize Thagards principles for explanatory coherence in an
epistemically compelling way. An essential part of showing that Thagards princi-
ples are justificatory is demonstrating linkage between causal coherence and truth.
Causal coherence does have clear links with probable truth, and hence escapes the
criticism that explanatory virtues are only pragmatic. Finally, I shall argue that
certain types of causally coherent mechanism are also essential in justifying the
kinds of claims of empirical adequacy that Lloyd thinks are sufficient for establish-
ing that models are verae causae. This I take to be the epistemic significance of the
theory-ladeness thesis.

Explanatory coherence

Thagards interpretation of Darwins argument as abductive seems appropriate

because Darwins theory refers to processes that occur in the prehistoric past, and
thus their actuality cannot be established by direct observation. It must be inferred
from what can be observed, e.g. adaptive organization, biological diversity, bio-
geographical distribution, etc. Also Darwin took pains to point out that in spite of
empiricist skepticism, hypotheses are acceptable to the scientific community if they
are consilient, that is, if they explain more than one class of facts, as does the wave
theory of light (see, e.g., Darwin, letter to Asa Gray, 1919, 2:80, cited in Lloyd
518

(1983)). In short, among other arguments, Darwin appeals to abductive consilience

in defending his theory of transmutation.
Thagard points out that creationism explains several kinds of facts, such as
adaptive organization, the adaptiveness of instincts, and co-adaptive biological
diversity. Darwins thesis that species gradually change (his transmutation thesis)
can explain much more, including shared embryological forms among distinct
species, bio-geographical distribution of extant and fossil forms on continents and
nearby islands, extinction, systematic affinities among species, rudimentary and
atrophied organs, and much else. Darwins transmutation thesis is thus more
abductively consilient than creationism. Also, Darwin also provides a mechanism,
natural selection, that explains transmutation. Furthermore, natural selection itself is
explained by Malthusian population growth, limited resources, and heritable vari-
ation among offspring at birth. Both transmutation and natural selection are thus
supported by converse abduction: The credibility of an explanandum, whether or
not observed, is enhanced by its being explained at a deeper explanatory level. There
is no comparable explanatory hierarchy of mechanisms explaining creation, though
both theories gain explanatory power from analogies with presumably well known
processes, e.g. the production of watches and the production of domestic breeds
(Principle 3, 1992, p. 66). Thagard is not explicit on what these analogies contribute
to explanations, though it may be the demonstration of similarities to familiar
processes in the source of the analogy (in Darwins case, the production of domestic
species) that gives the audience addressed by the argument the conviction that they
understand what is going on in its target (speciation in nature). Alternatively, the
analogies can be interpreted as Lloyd prefers; they provide inductive support for the
actual occurrence of heritable variation and adaptive divergence in prehistory.
According to Lloyd, demonstrating that processes are familiar contributes nothing to
the credibility of their actual occurrence. Hence this explanatory virtue the
possession of analogical relations with familiar processesmay satisfy human
desires for familiarity, but it has no epistemic import. Nature can act in ways that are
inescapably unfamiliar.
Thus the principal factors contributing to the explanatory superiority of Darwins
theory is (according to Thagard) its explanation of a wider range of evidence and a
two-level explanation of the mechanism of evolution, the latter of which is absent in
creationism. These considerations establish the greater (internal) coherence of
Darwinian theory on these principles: (i) Coherence and incoherence are symmetri-
cal relations, so if P and Q cohere (incohere), so does Q and P (Principle 1, Thagard
(1992), Chapter 4, p. 65). Because coherence relations are symmetrical, every
abductive argument for an explanatory hypothesis H is accompanied by a converse
abductive argument for its explanandum P. (ii) Several propositions that explain
something gain coherence with what they explain and coherence among thesmelves
(Principle 2, p. 65). Thus Darwinian principles gain internal coherence by explain-
ing shared embryological forms, more such coherence by explaining biogeographi-
cal distribution, etc. Bio-geographical distribution gains coherence with shared
embryological form and taxonomic relations by being explained by transmutation.
Also transmutation coheres with natural selection because natural selection explains
 519

it. Natural selection coheres with Malthus hypothesis, limited carrying capacity and
individual variation at birth because the latter three propositions explain the former,
and they also thereby gain coherence among themselves (Principle 2). However,
Thagard does not require that Darwinian or Malthusian principles have prior
justification before entering these explanatory relations, such as prior empirically
accessible realizations or inductive or deductive support from such realizations, in
order that they have explanatory coherence.
In advocating Principle 4, propositions describing experimental evidence are
justified independently of coherence relations (p. 66), Thagard takes one step away
from the extreme coherentist wing and makes a concession to foundationalism by
admitting the credibility of observation-based propositions without coherence
relations to other propositions. Principle 4 thus suggests agreement with the
foundationalist that beliefs causally produced by sensory inputs are thereby justified
without appeal to justificatory relations with other beliefs. This concession is
problematic because under some circumstances some beliefs referring to natural
events and causally evoked by sensory processes are not likely to be true. Everyday
observation has evoked in many the false beliefs that terrestrial topography is
unchanging and that organic natural kinds always breed true and thus never evolve.
The same point applies to Lloyds empirical realizations. Sensory production cannot
establish semantical adequacyreference to actual occurrences-unless certain con-
ditions of epistemic reliability are met. In short, foundational beliefs are as much
subject to epistemic warrant as are explanatory virtues.
Also, empiricist philosophers (among whom I would include Lloyd) would
consider Thagards theory of explanatory coherence wanting because, even though
he grants a concession to empiricism by allowing that propositions reporting actual
observations gain credibility without necessarily cohering with other propositions,
there is no requirement that explanatory hypotheses have independent empirical
support. Though Thagards examples all contain explananda (facts requiring
explanation) that presumably are reports of observation that presumably have initial
empirical support without coherence relations, his principles permit the justification
of systems of propositions composing the explanation (the explanans) just on their
mutual coherence relations and their explanatory relation to the explananum. There
is no reason among Thagards principles that would exclude the explanatory relation
to an empirically based explanandums being one of explaning away that explanan-
dum. The hypotheses that explain away the explanandum may enjoy several levels
of explanatory relations and thus be more coherent than rival hypotheses that
explained the explanandum without being explained by anything else. One may
argue for transmutation by citing its explanation by natural selection, and natural
selections explanation by Malthus thesis, etc., without providing any evidential
connections to reports of observation, particularly if one is just comparing Darwins
theory with creationism. The conclusion would be that Darwins theory is superior
because transmutation is explained by a deeper mechanism and creationism is not.
What is missing is evidence that this explanatorily coherent mechanism actually
caused species adaptation and diversity. Certainly coherence with empirical claims
would enhance the coherence of this system on Thagards principles, but so also
520

would coherence with further empirically unsupported hypotheses, such as ana-

logies with other coherent fictions. For Thagard such analogies would undermine
the belief that any of the coherent propositions are fictions. Hence, though Thagard
departs from the radical fringe of coherentism, the departure does not obstruct
empiricist criticism because a non-tautologous belief can be justified entirely a
priori by its superior explanatory coherence with other a priori beliefs.
This objection applies to any coherentist epistemology that maintains that
justificational relations between all beliefs is symmetrical. If P can justify (under-
mine) H, then H can justify (undermine) P by coherence (incoherence) relations
between P and H. Accordingly, a hypothesis H that refers to some process
independent of human cognition may increase its coherence relations by undermin-
ing all seemingly relevant observation statements P. Such justification of H should
not be considered acceptable science because it provides no grounds for thinking H
or P correspond to the actual world.

Appraising allopatric models of speciation: causal coherence and empirical

adequacy

Mayrs allopatric model of speciation is a good example of a Darwinian history and

also a model containing a specific causal etiology. According to Mayr, the specia-
tion process begins with a founder event in which one or very few actually or
potentially fertile individuals enter a geographical region at the periphery of their
normal habitat. The boundary conditions for the peripheral region include a barrier
to migration. Colonizing organisms that cross such barriers are often described as
accidental migrants. Thus such crossings are not to be expected to occur frequently
or regularly. This part of the model provides little practically useful predictive
power in the sense that in most localities such events will not occur during a human
lifetime, both in fact and by prediction from the model. However, this point does not
undermine the explanatory power of founder events. If the belief that these events
occur can be justified empirically, by direct observation or by some reliable
inference from present evidence, then founder events do have explanatory power
even if their occurrence and the speciation processes they initiate are extremely rare.
In this case there is a gap between explanatory virtues, viz. evidence for truth, and
some pragmatic values, viz. predictability.
According to Mayr (1954, 1963), the founder event isolates a very small
population in an environment that is unusual for that species because the founder is
no longer subject to the influx of genes from immigrants. The parent population
contains genes or systems of genesbalanced polymorphisms maintained by
heterozygote superiority and balanced epistatic systems. These ancestral genetic
systems are good mixers in the sense that they can absorb immigrant genes and yet
maintain their viability. Gene flow also prevents a complete response of the genome
to local selection pressures and thus acts as a stabilizing influence for the population
as a whole. Mayr (1954) cites evidence for this in small mammals living in desert
 521

areas in the Southwest intersected with lava flows. Cutting off the source of
immigrants, according to Mayr, opens the possibility of local adaptation and
changes the genetic background and thus the selective value of alleles that make it
into the founder population, leading to the elimination of some and thus another
change in background. Thus there occurs a cascade of changes in fitnessa genetic
revolution. Since these founding populations often occur beyond the periphery of
the ecological boundaries of a species, they will be subject to differing and likely
unfavorable selection pressures. Further pressure on the equilibrium comes from
increasing homozygosity in the founders because of sampling from the parent
population, inbreeding and the likely loss of variation through drift in the small
population. Mayr cites observed adaptive instability in many island populations as
evidence for this reduction of variability in founders.
In rare cases the disturbed genetic system will eventually regain its co-adapted
equilibrium. Since in this time there will be some mutations, these are very unlikely
to parallel mutations in the parent population, and if they are maintained they
become part of a changed genetic background for further divergent mutations,
further propagating the cascade. Any re-assembly of co-adapted genes is likely to be
very different from the system in the parent population, and will take the founder
population to an adaptive peak that is quite different from the parents. Thus we can
expect from the speciation process significant and relatively rapid phenotypic
changes in the founder population. This re-organization effects a divergence of
phenotypic and physiological traits that influence niche occupation and pre-mating
and post-mating capacities for forming hybrids with the parent population, among
other phenotypic traits. The emergence of these isolating factors set the initiating
conditions for genetic isolation between the parent and the daughter populations
should the geographical barrier that initiated the process disappear. Re-colonization
sets initiating conditions for the occupation of new niches, which provides oppor-
tunity for expanding populations and increasing variation. Increased variation
initiates the restoration of balanced genetic diversity in the new species.
As to the initiating conditions of this process, Carson and Templeton (1984) point
out that the geological history and biogeography of a number of oceanic island
archipelagos, such as the Galapagos and the Hawaiian islands, provide ample
circumstantial evidence for repeated accidental colonization events by land organ-
isms, such as picture winged Drosophila and honeycreepers. In the case of the
Drosophila, there is also evidence for repeated isolation of populations by lava floes
on Hawaii providing geographical conditions for subsequent separation and re-
colonization. Knowing the motility of the organism and the nature of the barrier, the
best explanation of the distribution, the number of species and their endemic
character on these archipelagos is that the colonization events happened rarely, most
probably involving one gravid individual or a very small breeding population, thus
producing an extreme genetic bottleneck. The abduction here is unproblematic
because it appeals to well known causal capacities and boundary conditions, viz.the
migratory capacities of particular land organisms and the barriers presented to actual
migration by oceanic straights or volcanic incursions of various widths. Geology
provides the relative ages of the islands of a given archipelago, such as Hawaii, so
522

there are also available empirically based estimates of time intervals over which
improbable events can happen, and even happen frequently.
Carson and Templeton point out that Mayrs model lacks specificity concerning
the architecture of the genetic system taking part in the genetic revolution. Without
some specificity in this regard, there is no clear and causally coherent mechanism
producing the genetic revolution and its consequent isolating mechanisms and
morphological changes. Also in supposing that the founder event increases
homozygosity by sampling, inbreeding, and drift, his model leaves few if any
resources for the generation of a new co-adapted genetic system. Selection requires
variation as an initiating condition, but the model fails to provide such a condition
until after speciation is supposed to take place. In effect, they are saying that Mayrs
model lacks causal coherence and therefore cannot be effective in the production of
new species.
Carson and Templeton offer a specific proposal and some evidence for what they
call type II genetic architecture, which consists of a major segregating locus with
minor epistatic modifiers. One empirical exemplar of this architecture differentiates
the head shape of D. Silvesteris from D. heteroneura. This difference has been
shown experimentally to be associated with an X-linked segregating unit interacting
with several minor autosomal genes. They argue that the maintenance of this
contrasting character is due to strong sexual selection on the highly epistatic
complex controlling the head shape because head shape is stable in spite of evidence
for genetic variation in other traits influencing mating success. Thus a breakdown of
this complex is likely to have significant effects on mating success giving it a direct
bearing on the production of new species. Also experiments in which laboratory
populations of Hawaiian Drosophila were subjected to extreme bottlenecks resulted
in the appearance of strong pre- and post-mating barriers.
In sum, natural and laboratory evidence support the claim that bottlenecks cause
breakdown of these type II epistatic systems which stably maintain a strongly
selected character in ancestral populations in spite of polymorphisms among the
minor epistatic genes therein. These polymorphisms give rise to the potential for
genetic changes from recombination, but the minor epistatic genes have been
selected so that recombination will not disturb functional characters in the ancestral
population, which include characters bearing upon mating success and thus subject
to strong sexual selection. Carsons founder flush model departs from Mayrs in
that he adds a stage in which there occur reduced selective pressures in the founding
population immediately following the founding event. As a consequence, the
founding population rapidly expands into the colonized environment. Carson and
Templeton also cite population genetic arguments to the effect that population
bottlenecks do not significantly reduce genetic variation, contrary to an element in
Mayrs model. The relaxed selective environment allows multiple recombinant
offspring in a still variable and growing population, which would have been
removed by strong selection in the ancestral population. This recombination is what
disrupts ancestral epistatic balances and multiplies variation in these traits in the
daughter population. Subsequently stabilized characters will, by a cascade of
recombinations in ever changing genetic environments, be very likely to be different
 523

from the corresponding ancestral characters. Thus a daughter population can be

driven to a new adaptive peak, which includes new stable characteristics influencing
mate selection. Thus speciation is the outcome of this process.

Conclusions

We have seen that Mayrs model for speciation supposes that the founder bottleneck
produces increased homozygosity as a consequence of sampling, drift and inbreed-
ing. Accordingly, there is little to draw from for the production of a significantly
novel genome, and thus the initial conditions for the outcome of the genetic
revolution are absent. Mayrs model thus lacks causal coherence. If it were realized
in nature, it would fail to produce new species. Therefore it cannot represent actual
causal histories that are producing presently observed biological diversity. Causal
coherence in evolutionary etiologies is in this way linked as a necessary condition to
truth and to the attribution of explanatory power to an etiological model.
However causal coherence alone is insufficient for attributing explanatory power
to a model, even if the models outcome is known to have a natural realization.
Abduction understood as the inference of truth from a coherent history that
terminates in some actual happening is not epistemically compelling because
biologists seem to explain by citing actually occurring causal etiologies. Adding the
premise that there is no alternative account of the happening that the happening is
surprising (Sober 1993; Peirce 1960)does not enhance the explanatory power of
the account, nor does it justify credence in the account. There must be other prior
justification that the components of the account have occurred in nature and that the
initial and boundary conditions for their occurrence are realized in the present case.
Evolutionists have gone to great lengths to try to provide such justification. Mayr
(1963) cites studies of presently extant organisms as analogs of prehistoric stages
that supposedly appeared in causal histories underlying present diversity. Mayrs
procedure can be regarded as an elaboration of Darwins reasoning from knowledge
of domestic variation, as interpreted by Lloyd. The process of speciation in its
various stages can be observed in presently occurring natural and laboratory
populations. Also presently observable in these populations would be starting and
terminating conditions of these stages. The allopatric model of speciation has many
elements that are thus empirically realized, e.g. conditions of geographical isolation
of small peripheral populations and each of the various stages of behavioral,
ecological and physiological isolation are manifest in a wide range of cases studied
in natural and laboratory environments. There are also empirically supported cases
that exemplify the occurrence of variation in isolating mechanisms and various
stages of the transition from allopatry to sympatry. Mayr also cites evidence that
hybrids are generally selected against after the two species have drifted apart in
geographical isolation, an important component in his model for the intensification
of isolating mechanisms by natural selection. Carson and Templeton similarly seek
empirical realizations for the several components of their model, e.g. type II
epistatic systems in sexually selected characters in Hawaiian Drosophila, and the
524

production of isolating mechanisms in populations subjected to bottlenecks in the

laboratory. Carson and Templeton thus cite present experimental realizations of
genetic systems that provide greater causal coherence than Mayr is able to achieve
for the production of outcomes or terminating conditions for these links in the causal
chain. Inferring by induction that the same stages can occur in the past under the
same starting conditions, a continuous series can be constructed in prehistory by
arranging like stages in chronological order. According to Mayr, such arrangement
and interpolation is commonly accepted practice in science, as in accounts of cell
division and disease prognoses. Fitting components with known initial and ter-
minating conditions into a causally coherent etiology is like putting together pieces
of a pre-cut puzzle.
However, these inductive inferences are still not sufficient to establish that some
prehistoric process has actually occurred in the production of presently occurring
species, say in Hawaiian Drosophila. Adding that there are some present analogs to
elements of a causally coherent story does not completely eliminate the suspicion
that the history is plausible fiction. Evolutionists can carry the justification process
one step further: Geological evidence cited by Carson and Templeton concerning
the formation of volcanic islands and ancient lava floes support the claims that the
initiating conditions and some of the boundary conditions for intermediate stages
have actually occurred. Combining these claims with the assembly of a causally
coherent history from stages whose inputs and outputs are justified inductively gives
a more convincing case for the actuality of the history, even though there are bound
to be prehistoric background conditions for the occurrence of some stages that are
unknown. This illustrates how far evolutionists can go in fulfilling Lloyds demands
for empirical adequacy.
Kitchers caveat, viz. that the truths that matter are those relevant to the discovery
of explanatory patterns, does not undermine what he has called robust realism.
Richard Owens schema for unifying comparative anatomy might be taken to have
some sort of explanatory import by demonstrating how vertebrates occur as
variations on a common theme even though it appears to be a-historical and
a-causal. As such it might be considered a rival explanatory scheme to Darwinian
histories that provides a form of unity to the diversity among vertebrates without
explicit causal etiologies. Similar organizing relations might be found within the
other Cuvierian embranchements, the mollusca, radiata and articulata. Darwinian
histories, on the other hand, are explicitly causal and historical. Darwins argument
in the origin as interpreted by Lloyd follows the argument schema Mayr recom-
mends for defending etiologies with prehistoric elements that are thus not directly
observable.Variation among sexually produced offspring observed at birth occurs in
every generation of one or another domestic species, but does not occur in asexually
produced individuals, so sexual reproduction is an initiating condition for such
variation. Conditions of domestication may accelerate the variation, but they do not
produce it ex nihilo in asexual reproduction, so they are neither initiating nor
boundary conditions for variation. Thus variation is likely to occur in wild sexual
species as well. Uniformitarians such as Darwin assume that the natural world is
causally uniform over time and space. Accordingly, if the initiating and boundary
 525

conditions for variation are realized in the past, as they should be in sexual
organisms, variation should occur there as well.
Both Mendelian and molecular genetics support the assumption of causal uni-
formity that Darwin was committed to. There is no essential difference in the
production of variability between domesticated and wild organisms. Both are
equally subject to point mutations, genetic duplications, and recombination. The
same uniformity underlies all taxa of presently living organisms, so by cladistic
criteria, this uniformity would have to be ancestral. Evolutionary developmental
genetics should eventually provide a causally coherent mechanism for the replica-
tion and differentiation of developmental cycles that are shared by domestic and
wild metazoa. This uniformity exemplifies Thagards converse abduction: The
uniformity of patterns of variation and inheritance in domestic and wild species is
explained by deeper genetic and molecular mechanisms, and thus Darwins uni-
formitarian assumption gains credibility. The gain in credibility is partially due to a
causally coherent history that fills an explanatory gap in Darwins thinking, that is
Darwins assumption that generation in domestic is like generation in wild species
without further justification. Also components of the mechanisms-replication,
segregation, recombination, regulation, etc.-have many empirical realizations, so
this explanatory mechanism can be considered to be a vera causa. Darwins
argument from domestic to wild to prehistoric production of variants thus would be
judged reliable from the external viewpoint of contemporary biology-Darwin
himself was in no position to appreciate or offer such a defense. This line of
reasoning can be considered a local epistemic justification of an inductive pro-
cedure: a causally coherent and empirically realized model explains the assumption
of uniformity required by the induction. Thus the induction in these circumstances is
reliable.
I have brought out two cases in which Lloyds worries that explanatory virtues are
just pragmatic are unfounded: First, causal coherence is instrumental to discovering
truth insofar as demonstrably causally incoherent mechanisms can be discarded as
candidate etiologies. Secondly, Kitchers caveat demands that the search causally
coherent mechanisms guide our search for factual truth. But Kitchers hope for
explanatory unification is not wishful thinking that nature realize conditions for
satisfying human practical needs. Kitchers Darwinian histories contain elements
that are realized in presently observable states of affairs, as Lloyd suggests, and thus
she would grant them explanatory power as vera causae. Lloyds avoidance of
explanatory virtues leaves open the possibility that she would be satisfied with
historical chronicles whose elements have presently observable analogs but lack
causal coherence. However, such chronicles lack explanatory power because there is
no reason to think that their early stages actually generated the ultimate output, the
explananda. The chronicle could be a sequence of epiphenomena produced by some
unknown etiology.
Lloyd and other empiricists should also be worried about the reliability of claims
about the empirical realization of elements of explanatory models as well as of
inductive and analogical arguments-they are as much subject to the requirements of
epistemic vindication as are the explanatory virtues. These considerations bring
526

evolutionists to cite further causally coherent histories. The progression of the fossil
record is one source of empirical support for Darwinian histories. However,
evolutionists also need to justify the belief that progression of fossil forms in
observable geological formations correspond to actual histories in their continuity,
locality and temporal order. History has shown us that justification comes in the
form of a causally coherent history that links the deposition, preservation and
erosion of imprints of dead organisms with what is presently observed in fossil
bearing strata. Similar histories underlie the case that can be made for the origins of
geographical regions and barriers, which include initiating conditions for evolution-
ary processes.
Foundationalists worry that demands for this kind of justification will regress to
infinity or carry us into vicious circles, neither of which provide plausible justifica-
tion. However, the history of science contains many occasions in which inquiring
into the justification of apparently foundational perceptions have been justly
occasioned by controversy and the effort to transcend the spatiotemporal locality of
unaided human senses. Recent controversies over the tempo and mode of evolution
raise questions about the causal history behind the absence of intermediates between
marine species represented in the fossil record. We need to be assured that the
protocols for the amplification of DNA samples give us accurate copies of the
original sample to be assured that lines in four tracks in a gel represent the actual
sequence of nucleotides in that sample. The assurance comes from representing the
protocols as a coherent causal history whose elements have credible empirical
realizations. Rather than being circular, the lineages of justification in science are an
expanding net that terminate in elements for which there is practically successful
consensus, dogma or ignorance, any of which may be only temporarily insulated
from further inquiry. These termini can be externally justified beliefs provided that
they are actually connected with reality by reliable informational channels even
though the inquirer is ignorant of what the justification is, as we have seen happen in
Darwins case.

Acknowledgements

The author is indebted to helpful suggestions from Kim Sterelny and an anonymous
referee.

References

Carson H.L. and Templeton A.R. 1984. Genetic Revolutions in Relation to Speciation Phenomena: The
Founding of New Populations. Annual Review of Ecology and Systematics 15: 97131.
Dennet D.C. 1993. Darwins Dangerous Idea. Touchstone Books, New York.
Feyerabend P.K. 1962. Explanation, Reduction and Empiricism. In: Feigl H. and Maxwell G. (eds),
Minnesota Studies in the Philosophy of Science Vol. III. University of Minnesota Press, Minneapolis.
Hanson N.R. 1961. Patterns of Discovery. Cambridge Univ. Press, Cambridge.
Kitcher P. 1993. The Advancement of Science. Oxford University Press, New York.
 527

Kleiner S.A. 1985. Darwins and Wallaces Revolutionary Research Programme. British Journal for the
Philosophy of Science 36: 367392.
Kuhn T.S. 1962. The Structure of Scientific Revolutions. University of Chicago Press, Chicago.
Laudan L. 1977. Progress and Its Problems. University of California Press, Berkeley.
Lloyd E. 1983. The Nature of Darwins Support for the Theory of Natural Selection. Philosophy of
Science 50: 112129.
Mayr E. 1954. Change of genetic environment and evolution. In: Huxley J.S., Hardy A.C. and Ford E.B
(eds), Evolution as a Process. Allen & Unwin, London, pp. 156180.
Mayr E. 1963. Animal Species and Evolution. Harvard Univ. Press, Cambridge, MA.
Peirce C. 1960. The Collected Works of C.S. Peirce. In: Hartshorne C. (ed.), Harvard University Press,
Cambridge, MA.
Ruse M. 1975. Charles Darwins Theory of Evolution: An Analysis. Journal of the History of Biology
8: 219241.
Ruse M. 1979. Philosophical Factors in the Darwinian Revolution. In: Wilson F. (ed.), Pragmatism and
Purpose. University of Toronto Press, Toronto.
Ruse M. 1999. Mystery of Mysteries. Harvard University Press, Cambridge, MA.
Sober E. 1993. The Philosophy of Biology. Hackett, Indianapolis, IN.
Thagard P. 1992. Conceptual Revolutions in Science. Princeton Univ. Press, Princeton.