Assistant Professor, Department of Philosophy, University of Alabama at Birmingham
(Information for UAB students)
(Note: Most links below are to PDF files.)
Areas of specialization: philosophy of science, philosophy of biology, philosophy of mind/cognitive science.
Areas of competence: metaphysics, epistemology, philosophy of language, symbolic logic, decision theory.
Email: marshall at logical dot net
Does my research have anything to do with normative issues or public policy?
What Determines Biological Fitness? The Problem of the Reference Environment,
forthcoming in Synthese. The original publication will be available at
www.springerlink.com.
Organisms' environments are thought to play a fundamental role in
determining their fitness and hence in natural selection.
Existing intuitive conceptions of environment are sufficient for
biological practice. I argue, however, that attempts to produce a
general characterization of fitness and natural selection are
incomplete without the help of general conceptions of what
conditions are included in the environment. Thus there is a
"problem of the reference environment"--more particularly,
problems of specifying principles which pick out those
environmental conditions which determine fitness. I distinguish
various reference environment problems and propose solutions to
some of them. While there has been a limited amount of work on
problems concerning what I call "subenvironments", there
appears to be no earlier work on problems of what I call the "whole
environment". The first solution I propose for a whole
environment problem specifies the overall environment for natural
selection on a set of biological types present in a population
over a specified period of time. The second specifies an
environment relevant to extinction of types in a population;
this kind of environment is especially relevant to certain kinds
of long-term evolution.
How Do Natural Selection and Random Drift Interact?,
Philosophy of Science, 74(5), December 2007 (© PSA)
One controversy about the existence of so called evolutionary forces
such as natural selection and random genetic drift concerns the
sense in which such "forces" can be said to interact. In this
paper I explain how natural selection and random drift can
interact. In particular, I show how population-level
probabilities can be derived from individual-level probabilities,
and explain the sense in which natural selection and drift are
embodied in these population-level probabilities. I argue that
whatever causal character the individual-level probabilities have
is then shared by the population-level probabilities, and that
natural selection and random drift then have that same causal
character. Moreover, natural selection and drift can then be
viewed as two aspects of probability distributions over
frequencies in populations of organisms. My characterization of
population-level probabilities is largely neutral about what
interpretation of probability is required, allowing my approach
to support various positions on biological probabilities,
including those which give biological probabilities one or
another sort of causal character.
Fitness and Propensity's Annulment?,
Biology and Philosophy,
22(1), January 2007
Recent debate on the nature of probabilities in evolutionary biology
has focused largely on the propensity interpretation of fitness, which
defines fitness in terms of a conception of probability known as
"propensity". However, proponents of this conception of fitness
have misconceived the role of probability in the constitution of
fitness. First, discussions of probability and fitness have almost
always focused on organism effect probability, the probability
that an organism and its environment cause effects. I argue that
much of the probability relevant to fitness must be organism
circumstance probability, the probability that an organism
encounters particular, detailed circumstances within an environment,
circumstances which are not the organism's effects. Second, I
argue in favor of the view that propensities either don't exist or
are not part of the basis of fitness, because they usually have values
close to 0 or 1. More generally, I try to show that it is possible
to develop a clearer conception of the role of probability in biological
processes than earlier discussions have allowed.
Teleosemantics without Natural Selection,
Biology and Philosophy 20(1), 2005
Ruth Millikan and others advocate theories which attempt to
naturalize wide mental content (e.g. beliefs' truth conditions)
in terms of function in the teleological sense,
where a function is constituted in part by facts concerning past
natural selection involving ancestors of a current entity.
I argue that it is a mistake to base content on selection.
Content should instead be based on functions which though
historical, do not involve selection. I sketch an account of
such functions, which defines "function" in terms of changes in
objective probabilities due to changes in ancestral traits.
Infinite Populations and Counterfactual Frequencies in Evolutionary Theory,
Studies
in History and Philosophy of Biological and Biomedical Sciences, 37(2), June 2006
One finds intertwined with ideas at the core of evolutionary
theory claims about frequencies in counterfactual and
infinitely large populations of organisms, as well as in sets of
populations of organisms. One also finds claims about frequencies
in counterfactual and infinitely large populations--of
events--at the core of an answer to a question concerning the
foundations of evolutionary theory. The question is this: To
what do the numerical probabilities found throughout evolutionary
theory correspond? The answer in question says that evolutionary
probabilities are "hypothetical frequencies" (including what are
sometimes called "long-run frequencies" and "long-run
propensities"). In this paper, I review two arguments against
hypothetical frequencies. The arguments have implications for
the interpretation of evolutionary probabilities, but more
importantly, they seem to raise problems for biologists' claims
about frequencies in counterfactual or infinite populations of
organisms and sets of populations of organisms. I argue that
when properly understood, claims about frequencies in large and
infinite populations of organisms and sets of populations are not
threatened by the arguments. Seeing why gives us a clearer
understanding of the nature of counterfactual and infinite
population claims and probability in evolutionary theory.
The Unity of Fitness
According to the original version of the propensity interpretation of
fitness, fitness is a mathematical function of probabilities and
numerical values associated with reproductive outcomes. In
particular, the function was thought to be the expected or arithmetic
mean number of offspring. In response to work by Gillespie in the
70's, some authors have argued that fitness might sometimes be
defined in terms of geometric mean number of offspring, or a linear
combination of the mean and variance of number of offspring, or some
other function (Beatty & Finsen 1989, Brandon 1990, Sober 2001).
While Brandon (1990) argued that fitness therefore merely satisfies a
common schema instantiated by different mathematical functions, Ariew
& Ernst (2007) have gone further, arguing that Gillespie's work shows
that no coherent definition of fitness is possible.
Similar conclusions have been drawn from arguments that fitness must
sometimes be characterized by an even wider variety of mathematical
functions because of conspecifics' mutual influence on reproductive
success (Ariew & Lewontin 2004, Krimbas 2004). For example,
different functions might be needed to deal with sexual vs. asexual
reproduction, frequency-dependent and density-dependent fitness,
maternal effects, and some kinds of niche construction.
Despite the heterogeneity of mathematical functions needed to model
fitness, I argue that fitness is nevertheless a common property of
types in populations, and that: (1) It's plausible that fitness is
constituted by one very complex, parameterized, mathematical function
of probabilities, numbers of descendants, and other factors, of which
different mathematical functions are specializations. (2) Whether or
not (1) is correct, the fact that fitness involves different
functions in different contexts is not in itself problematic, but is
merely an extension of the common idea that fitness is determined by
environment. (3) Though fitness must sometimes be defined in terms
of probabilities of reproductive effects over several generations,
this does not mean that fitness does not have to do with influence in
each generation. Since probabilities of long-term effects can be
derived from probabilities of short-term effects, the former are
simply mathematical properties of causes acting in the short term.
This removes a motivation for Brandon's schema account of fitness.
Functions, Altruism, and Conditional Fitness (available on request)
It's recently been argued that biological fitness cannot change over
the course of an organism's life. However, many
characterizations of biological function and biological altruism
tacitly or explicitly assume that an effect of a trait can produce a
positive change in the fitness of an organism. In the first
half of the paper, I explain how the effects of behaviors on fitness
can be understood in terms of conditional probabilities defined over
sequences of events in an organism's life. The result is a
notion of "conditional fitness" which is static but which captures
intuitions about apparent behavioral effects on fitness. The
second half of the paper investigates the possibility of providing a
systematic foundation for conditional fitness in terms of spaces of
sequences of states of an organism and its environment. I
argue that the resulting "organism-environment conception" helps
unify disparate biological perspectives.
Lewontin's Conditions and the Units of Evolution (available on request)
I present a generalized version of Lewontin's (1970) conditions
for evolution by natural selection, and a generalization of Maynard
Smith's (1987) "unit of evolution". A unit of evolution in
this sense is an entity which determines an probabilities concerning
inheritance. More specifically, I characterize evolution as a
change in the distribution of a set of properties, and a
distribution of a set of properties as the sort of thing which can
be an evolutionary effect. The same set of objects in the
world can instantiate different sorts of properties simultaneously,
allowing one population to be involved in different effects.
These different effects, in turn, may have distinct causes.
Thus, for example, a distribution of alleles at certain loci in a
population of organisms is one evolutionary effect; a distribution
of phenotypic characters--perhaps continuously varying--in the same
population at the same time is another evolutionary effect.
The same population might also have a distribution of properties of
groups within the population--a third evolutionary effect. A
set of properties thus defines a unit of evolution and a kind of
evolutionary effect. This way of defining a plurality of units
of evolution means that their existence does not depend on our
choices, descriptions, theories, etc. It is consistent with
changes in distribution of units of evolution in a population being
caused by properties at a variety of higher and lower levels of
selection, or more generally, contexts of selection. I provide
further reasons to doubt that Hull's (1980) replicator/interactor
distinction is fundamental to natural selection, and further reasons
to doubt that a clear distinction between group selection and
individual selection can be drawn. The present perspective
clarifies one kind of bookkeeping problem (e.g. Ariew and Lewontin
2004), which arises because there is no one obvious way to
understand natural selection in cases, for example, when plants
reproduce clonally or send out underground runners. On my
view, proposed solutions such as taking selection to operate on
number of ramets, number of meristems, amount of biomass, resources
reserved, etc. are all potentially legitimate, since each may define
a different evolutionary effect.
Mechanistic
Probability and the Causal Structure of Fitness
(available on request)
I propose the "mechanistic conception of
biological fitness" as an alternative to existing
conceptions. Mechanistic fitnesses depend on the causal
structure of what I call a "causal map", which is defined
by properties of an organism and its environment, and on
certain general facts about many populations of organisms.
Mechanistic fitnesses are objective, consistent with both
determinism and indeterminism, and give fitness
differences a causal role in evolution. The mechanistic
conception of fitness can be viewed as a descendant of the
propensity interpretation of fitness, but depends on a new
interpretation of probability, "mechanistic probability",
which I sketch. I also propose a resolution to the problem
of specifying environments' temporal and spatial
extents. Longer
description
How Can a Trait Be both Advantageous and Disadvantageous? (available on request)
It's often claimed that a genotype or phenotype is both
advantageous and disadvantageous, or both beneficial and costly, or
that it involves tradeoffs between competing resources, or needs,
or useful effects. Though obviously legitimate, such claims have a
puzzling aspect. They seem to imply that a type both increases and
decreases fitness relative to alternative types. I explain that
by decomposing fitness into probabilities which contribute to
fitness, we can understand claims about such "fitness tradeoffs"
as claims about differences in these component probabilities. I
argue that a fitness tradeoff usually involves a certain pattern
of relationships between probabilities, and illustrate this
pattern using examples based on recent work on the evolution of
animal communication.
Environmental Complexity and the Evolution of Cognition
The environmental complexity thesis (ECT) is the claim that
natural selection for complex cognitive abilities is the result
of living in complex environments. Peter Godfrey-Smith has
argued that the ECT plays a role in several accounts of the
evolution of human cognition. I discuss some variants of ECT and
argue the kind of complexity that matters is complexity in the
determinants of biological fitness. I then argue that careful
application of existing notions of environmental complexity would
lead to the view that nearly every environment is a complex one.
Thus there seems to be no distinction between complex and simpler
environments which might explain the evolution of cognition. I
explain, however, that there is one sort of environmental
complexity which should be particularly favorable to the
evolution of cognition, namely rapid intra-generational
variation in determinants of fitness. I point out that some
variants of the so-called Machiavellian Intelligence Hypothesis
describe such situations.
Short-Run Mechanistic Probability
This paper sketches a concept of higher-level objective probability
("short-run mechanistic probability", SRMP) inspired partly by a style
of explanation of relative frequencies known as the "method of
arbitrary functions". SRMP has the potential to fill the need for a
theory of objective probability which has wide application at higher
levels and which gives probability causal connections to observed
relative frequency (without making it equivalent to relative
frequency). Though this approach provides probabilities on a space of
event types, it does not provide probabilities for outcomes on
particular trials. This allows SRMP to coexist with lower-level
probabilities which do govern individual trials.
Conference on The Evolution of Cognition: Niche Construction, Culture, and Environmental Complexity, April 23-24, 2005 (I organized this with students at Duke).