Studies in History

and Philosophy of
Biological and
Biomedical Sciences
Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498
www.elsevier.com/locate/shpsc

Block Fitness
Grant Ramsey
Department of Philosophy, Duke University, 201 West Duke Building, Box 90743, Durham, NC 27708, USA

Received 21 October 2005; received in revised form 29 January 2006

Abstract

There are three related criteria that a concept of tness should be able to meet: it should render
the principle of natural selection non-tautologous and it should be explanatory and predictive. I
argue that for tness to be able to fulll these criteria, it cannot be a property that changes over
the course of an individuals life. Rather, I introduce a tness conceptBlock Fitnessand argue
that an individuals genes and environment x its tness in such a way that each individuals tness
has a xed value over its lifetime.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Environment; Fitness; Natural selection; Propensity

1. Introduction

Nothing seems less controversial than the fact that the tness of organisms changes
over time. If an organism becomes ill, infertile, or otherwise compromises its chances at
reproduction, its tness will decrease. Similarly, the birth of an organisms ospring
increases its tness. Fitness also seems to uctuate through interactions between individ-
uals. An altruistic interaction, for example, is characterized as one involving such a change
of tness: Sober & Wilson (1998, p. 17) hold that A behavior is altruistic when it increases
the tness of others and decreases the tness of the actor.
In this paper I argue thatcontrary to our intuitionstness cannot uctuate with
time: conceiving of tness as uctuating is problematic because such a tness concept does
not do a good job explaining and predicting evolutionary and ecological phenomena and
cannot play a central role in the principle of natural selection. The tness concept I argue

E-mail address: [email protected]

1369-8486/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.shpsc.2006.06.009
 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498 485

forBlock Fitnessis a block property, constant over an individuals lifetime, implying

that no events in an individuals life, altruistic or otherwise, can change its tness. Fitness,
then, is not a function of the actual life that an organism lives. Rather, I will argue, it is a
function of all of the possible lives that the organism could have lived. Although this
sounds epistemologically problematicdiscovering the tness of individual organisms
might seem impossiblewe will see below that this conception poses no such diculties.
This time-invariant tness concept is in opposition to recent arguments for a Fisherian
(that is, realized) conception of tness (for example, Matthen & Ariew, 2002; Walsh, Lew-
ens, & Ariew, 2002). Below, in arguing for Block Fitness and against realized tness, I
show that realized tness renders the theory of natural selection tautologous. Moreover,
realized tness is not explanatory of evolutionary change.1
Additionally, the tness concept argued for here implies that an individual does not
simultaneously have more than one real tness. This is in opposition to some recent plu-
ralistic treatments of the concept of tness. Walsh, Lewens, & Ariew (2002, p. 467), for
example, argue that [t]here may be no single correct way to calculate trait tnesses; often
enough it is simply a matter of our explanatory project.
Although previous conceptions of tness have not explicitly treated tness as time-invari-
ant, at least one of these conceptions is amenable to being time-invariant, namely the pro-
pensity interpretation of tness (Brandon, 1978; Mills & Beatty, 1979).2 Although a full
defense of the propensity interpretation of tness is beyond the scope of this paper, I will give
partial arguments in favor of it and show how it can ll the requirement of time-invariance.
I will begin by arguing against the Fisherian conception as an adequate notion of t-
ness. I then argue for the constancy of tness over an organisms lifetime. In order to
do this, I introduce the notion of the tness environment.

2. Fitness as actualization

The most straightforward conception of ospring-based tness is the following: tness

is the number of ospring an individual3 produces over its lifetime. This type of tness,
called realized tness or Fisherian tness,4 is generally conceived of in either of two ways,
Ratchet Fitness or Aristotelian Fitness (Figure 1a,b). With Ratchet Fitness, each organism
1
Matthen (personal commentary) feels that predictive tness can be understood as time-invariant. Although it
might be possible to interpret predictive tness in such a way that it is time-invariant, the way in which Matthen &
Ariew (2002, p. 56) describe predictive tnessas being based on a statistical measure of evolutionary change
: : :implies that predictive tness does vary with time: because the rate of evolutionary change can uctuate
over a relatively short time scale, predictive tness (which is based on this change) would reect such a
uctuation.
2
Propensities can be understood as time-invariant or as changing with time. Thus a propensity interpretation of
tness is not automatically a version of Block Fitness. However, I hold that if the propensities in a propensity
interpretation of tness are taken to be time-invariant, then it is compatible with Block Fitness.
3
In the literature, realized tness is often cast in terms of average tnessthe average contributions of an
ensemble of individuals (organisms, alleles, etc.), not a single individual. For example, Futuyma (1986, p. 552)
denes tness as The average contribution of one allele or genotype to the next generation or to succeeding
generations, compared with that of other alleles and genotypes. For the purposes of this argument, the
distinction between realized tness conceived as either the contribution of one individual or as the average
contribution of many is immaterial. Both fail as adequate conceptions of tness for the same reasons.
4
The predictive tness of Matthen & Ariew (2002) is also a variety of realized tness. It is a prospective
tness that is based on the actual (realized) evolutionary change. Because evolutionary change is a property of
ensembles, not individuals, predictive tness is a property of ensembles, not individuals.
486 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498

a
x

o
t
b
x

c t

t
d

o x

Fig. 1. a. Ratchet Fitness; b. Aristotelian Fitness; c. Flux Fitness; d. Block Fitness. x = tness; t = time; o
denotes birth; denotes death.

is born with zero tness and its tness is ratcheted up with each successful act of reproduc-
tion. With Aristotelian Fitness, on the other hand, an organism has a determinate tness
only upon death. Upon death, its tness is the number of its descendents. There is a great
epistemological benet to these conceptions. One can observe a population of organisms
through their lives and determine, absolutely, the tness of each. Despite this benet, the
problems with realized tness5 are many.
First, realized tness allows stochastic factors to modify an individuals tness. If one of
two identical6 organisms is struck by lightning and killed but the other goes on to have
three ospring, they will, under this conception, have dierent tnesses. But this tness dif-
ference is not due to the constitution of the organism. Rather, it is solely due to a stochas-
tic environmental factor. This is problematic since tness, it seems, should be identical for
identical individuals in the same environment. This modication of tness by chance
occurrences in organisms lives is the central problem with this concept of tness, and
the second and third problems are a consequence of this problem.
Second, realized tness is not explanatory. One desideratum of the concept of tness is
that the tness of an organism (or organism type) can be invoked to explain why it had
more or fewer ospring than one of its conspecics (or an alternate conspecic type).

5
I will use realized tness to refer generically to Ratchet Fitness and Aristotelian Fitness.
6
Identical in all properties save spatial position. There of course cannot be two organisms identical in all
properties.
 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498 487

But if tness is the number of ospring actually produced, then it cannot explain the num-
ber of ospring actually produced. This is the notorious tautology problem or explan-
atory circle that has convinced some that the theory of natural selection is empirically
empty, a mere tautology.
Third, realized tness is not a good predictor.7 If an organism dies with a tness of six (o-
spring) this fact does not accurately predict how many ospring another organism of the
same kind will (or is likely to) bear. The organism might have gotten lucky and produced
many more ospring than others of the same type will (or are likely to) bear. This is
analogous to the fact that the outcome of one roll of a die does not predict the outcome
of rolling other dice of the same type. Rolling a three might carry the information that
it is possible to roll a three. But it does not tell you the crucial information, for example, that
there is a 1/6 chance of rolling a three or that there are six possible outcomes. To predict the
outcome of the die, we need information about the die, such as the fact that it has six faces
each with an equal probability of landing face-up, that one face of the die has one pip,
another has two, and so on. This information is absent from the concept of realized tness.
Given these three problems, it seems that we need a dierent concept of tness. Does
this mean that realized tness is of no use? No. Realized tness is very important for ecol-
ogy and evolutionary biology. Realized tness is a measure of the theoretically adequate
conception of tness. But realized tness should not be confused with the tness concept
that is central to the theory of natural selection.

3. Fitness as lifetime propensity

The propensity interpretation of tness as expressed by Brandon (1978) and Mills &
Beatty (1979) purports to break the explanatory circle by conceiving of tness as a propen-
sity to produce ospring.8 The circle is broken since the actual number of ospring an
organism produces need not equal its propensity to produce ospring. The propensity
to produce some number of ospring is like the propensity gold has for dissolving in mer-
cury. If some ingot of gold never happens to contact, and thus dissolve in, a pool of mer-
cury, this does nothing to weaken our claim that it has this propensity. It is a version of
this account of tness that will be argued for below. Fitness, I will argue, is best conceived
as a function9 of the probability distribution of all the possible numbers of ospring the

7
An average realized tness, however, can be a good predictor.
8
Ospring can refer to ospring of genes, organisms, groups, etc. The view presented here is neutral with
respect to these entities, that is, it is not making a claim in the levels of selection debate. It is a further question
(and not one that I will address here) which entities should be thought of as having a tness. In what follows I
will, for the sake of simplicity, refer only to organisms. This should not be taken as an assumption that organisms
are the only kind of entity that can properly be said to have a tness. Rather, organism should be read as a
placeholder for whatever entity that can properly be said to have a tness.
9
Specifying what, precisely, this function is is not a trivial task. The original formulation of Brandon (1978)
was inadequate because it did not incorporate variance in the distribution of possible numbers of ospring
produced. See Brandon (1990) for a revised version. For a discussion of some of the diculties with the
propensity interpretation of tness see Beatty & Finsen (1989) and Sober (2001). It is beyond the scope of this
paper to attempt to specify what, exactly, this function is. Even if one feels that the latest versions of the
propensity interpretation of tness are problematic, and that this is reason for thinking that tness cannot be
understood in terms of a propensity, my arguments for an unchanging tness still hold: for tness to be
explanatory and play a role in the theory of natural selection, it cannot uctuate over the course of an organisms
lifetime.
488 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498

a 1 x

Pr(H)
o

0
t

b 1
Pr(H)

0 x
t

Fig. 2. a. The coin lands heads-up; b. The coin lands tails-up. Pr(H) denotes the probability of landing heads-
up; t = time; o denotes birth; denotes death.

individual might produce. Including the indirect component of tness, the full character-
ization is this: tness is some function of (1) the probability distribution of all the possible
numbers of ospring the individual might produce plus (2) the probability distributions of
all the possible numbers of ospring all of the individuals relatives might produce multi-
plied by their degree of relatedness to the individual. It is standard to refer to (1) as direct
tness, (2) as indirect tness, and the sum of (1) and (2) as inclusive tness. In what fol-
lows, I will, for the sake of simplicity, speak of tness as being about the number of o-
spring produced. But it should be understood as implicit that it is inclusive, not direct,
tness that is being discussed.
There are two chief ways in which tness as a propensity might be conceptualized, what
I will call Flux Fitness and Block Fitness (Figure 1c,d). With Flux Fitness, the tness of an
organism at time t is a function of the number of ospring it is expected to produce based
on its circumstances at time t. Since an organisms circumstances are always changing, its
tness is in ux. But as an organism approaches the end of its life, its possibilities of engen-
dering more ospring are reduced. As this occurs, its tness is damped. When it can no
longer have any more ospring, its tness becomes xed.10
Many of the problems with realized tness also plague Flux Fitness. To see why this is
the case, and to see why Block Fitness is immune to these problems, lets turn to an anal-
ogy. Consider a coin ipping device in a box. The box is perforated by dozens of small
holes and air is blown in and out of the holes. The air is suciently strong as to change
the outcome of a coin ip. Furthermore, the air is blown in and out of the holes in a ran-
dom11 fashion. Thus, the outcome of each ip of the coin is random. If we plot the prob-
ability that a given coin will land heads-up given its current situation, we might get a graph
like that of Figure 2a. Figure 2a corresponds to the Flux Fitness graph in Figure 1c: at
each time t, the probability of landing heads-up is based in its circumstances at time t.
The coin represented in Figure 2a happened to land heads-up.

10
Actually, because of indirect tness eects, it is possible to have post-infertility tness uctuations. An
organism can, for example, devote itself to caring for its still-reproducing siblings.
11
Random, here, meaning uncorrelated with states of the coin. The addition of air is meant to free the outcome
from the inherent biases of coin ipping machines (see Diaconis, 2004).
 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498 489

Given the outcome of the coin ip in Figure 2a (heads), or the instantaneous proba-
bility some time prior to landing heads, are we able to predict the outcome of another
ip of the same coin or a ip of a dierent coin of the same type? That the coin happened
to land heads-up merely tells us that heads is a possible outcome of ipping a coin of this
type. It does not tell us that it is rare or common. And it certainly does not tell us what
we want to know, that is, that coins of this type have a probability of 0.5 of landing
heads. We want to know the probability of landing heads in this environment for this
type of coin. The instantaneous probability of a coin during its ip gives us no such
information. Thus, the instantaneous probabilities represented in Figure 2a are not very
predictive. Likewise, Flux Fitness is not very predictive. Knowing that a four-year-old
organism has a probability of 0.7 of producing four ospring does not enable us to pre-
dict that another organism of the same type is likely to have four ospring. All we learn
is that four is a possible number of ospring for individuals of that type, and that for this
particular individual at this point in its life it has a high probability of having this num-
ber of ospring. It could be the case that the most likely number of ospring for individ-
uals of this type is two. That Flux Fitness is not very predictive should be no surprise,
since Flux Fitness and realized tness converge as the individuals approach death (that
is, as t approaches ).
Running identical coins through the coin ipping box will give dierent curves. If we
were to ip another coin in the same box, its curve might be something like 2b (in which
the coin lands tails-up). Thus, identical coins can precipitate very dierent probability
curves. The dierence in curves is not explained by dierences in the coins, since there
are no such dierences. Instead, the dierences are due to random variations in the envi-
ronment, which shows that random factors are incorporated into the probability curves.
Likewise, the tness curve for Flux Fitness is contingent upon random occurrences in
the organisms past.
How might we remedy this conception of tness? With the coin example, what would be
ideal for predicting the outcome of a toss of the coin? We do not want to know what the
instantaneous probability of the coin is during its ip. Rather, we want to know this: given
this coin and this box, what is the probability that the coin will land heads-up? The
answer, assuming it is a fair coin, is 0.5. It is this probability that we could use in predict-
ing the outcome of coins of the same type. This probability is given by the constitution of
the box and the coin. It is thus xed and does not uctuate with time. The tness analogue
is Block Fitness. Block Fitness is based on the environment and genes of the individual
and does not uctuate with time. This drives a wedge between random lucky or
unlucky events that happen to occur during its life, and its tness. By purging individual
random events from an individuals tness, survival of the ttest becomes meaningful.
An individual, or type of individual, can survive and out-reproduce another because it is
more t, that is, because it has a higher probability of having more ospring.

4. Objections

It might be objected that Block Fitness couldnt be right, since organisms can clearly
alter their tness. In the case of humans, a possible (rather extreme) example of increasing
ones tness would be a male who decided to donate to as many sperm banks as possible,
thereby greatly enhancing his genetic representation in the next generation. Has he thereby
increased his tness? He has not. The reason is that the probability distribution for the
490 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498

sperm donor (and, perhaps, others in the population) includes the unlikely possibility that
the individual will be enamored of the idea of donating sperm and donate as much as pos-
sible. One might have in their probability distribution a Pr(100 ospring) = 1 108. Vis-
iting the sperm banks and thereby having 100 ospring may be incredibly unlikely, but
exhibiting this behavior and obtaining this outcome does not alter the probability distri-
bution. Analogously, rolling a six each time in ten rolls of a die is an unlikely event,
but it does not prove that the die is biased. And even if the die is biased, for example,
if the probability of rolling a 6 is other than 1/6, the rolling of the die does not engender
the bias. The bias preexists the rolling of the die.
Although one might concede that an organism cannot alter its own tness, it might still
be objected that it is nonetheless possible to raise or lower an organisms tness through,
for example, experimental manipulation. Basolo (1990, 1995) attempted to establish the
fact that in the sh family Poeciliidae the female preference for males with caudal swords
preceded the evolution of swordtails from their swordless ancestors. One of the key exper-
iments involved attaching articial swords to swordless congeners (platysh in the genus
Xiphophorus) and observing whether female platysh prefer males with swords. She found
that female platysh do indeed prefer males with swords. From this, it might seem reason-
able to infer that if one took a male platysh from a wild population, surgically attached a
sword, then released it into its original population, its tness would thereby have increased
(assuming, of course, that the articial sword does not signicantly increase the probabil-
ity of predation or disease). But this inference is not warranted. If the researcher drew the
male at random from all of the males in the population, then the acquisition of a sword is
random with respect to the traits of individual males. Each of the males is equally likely to
acquire a sword. Thus, receiving a sword is just a matter of good luck, not explainable by
reference to specic traits or dispositions.
But there is another possibility. Perhaps the researcher does not just attach swords to
random males, but instead selects males of a certain kind, that is, males with a non-uni-
versal trait. Lets say the researcher attaches swords only to males that are in the top
10% with respect to body size. Furthermore, lets assume that the researcher does this gen-
eration after generation. Assuming there is a genetic basis for body size dierences, there
will be a response to this selection: average body size will increase. How should we inter-
pret this situation? Consistent with Block Fitness, this scenario would thus be described:
the platysh that have the greatest probability of having a body size in the top 10% will
thereby have higher tness than conspecics with a lower probability, other things being
equal. The fact that some particular platysh happens to have a sword added by a
researcher does not thereby raise its tness.
A third objection that could be raised is that the coin in the box is not properly anal-
ogous to organisms because the coin remains morphologically unchanged throughout its
life whereas organisms undergo radical change. To see that this is not a problem, we can
add morphological change to the coin. Imagine that not just air, but also droplets of mol-
ten metal stream in through the holes in the box. During the coins trajectory it accretes,
changing its physical characteristics. Because of this accretion, the probability that a coin
will land tails-up might reach zero. This is analogous to an organism becoming sterilized
prior to reproduction and having its chance of reproducing reach zero. Does this show
that tness can become zero during an organisms lifetime? No, just as the probability
of the coin landing on a particular side is 0.5 in this molten metal bombarded box, so
too the tness of an organism does not change. It may have more or fewer ospring
 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498 491

depending upon how it happens to encounter the ne-grained heterogeneity (Levins, 1964)
of its environment. But this randomness is not tness-aecting.
A fourth objection is that it is unwarranted to claim that the environment in which one
coin is ipped is the same environment in which another coin is ipped. Perhaps each coin
is ipped in a unique environment. Likewise, each individual organism might be thought
to live in its own unique environment. And since tness is a function of environment and
genes, identical coins or organisms can have dierent tnesses. To answer this objection, a
discussion of the concept of environment is necessary. The answer to this objection will be
addressed in the following section.

5. Fitness and the environment

If tness is Block Fitness, what xes tness? The short answer is that an organisms t-
ness is xed by its genes and environment. It is more or less clear what genes is referring
to,12 but what is an environment? Consider these two statements:

(1) An organisms genes and environment x its tness.

(2) Fitness is a relational property between organisms of the same species in the same
environment.

There is one sense of the term environment that I will call an individual environment. An
individual environment comprises all of the environmental inuences that happen to
impinge upon an organism during its life, that is, everything an individual encounters in
its space-time trajectory. An organisms development is a function of its individual envi-
ronment and its genes. But is an individual environment the environment referred to in
(1) and (2)? It is obvious that the environment referred to in (2) cannot be an individual
environment because, since each individual environment is unique, there is no sense in
which dierent organisms can occupy the same individual environment. And since
environment in (1) and (2) need to co-refer, a dierent concept of environment is
needed. Why does environment in (1) and (2) need to co-refer? If it did not, tness
would refer to a dierent property in (1) and (2). But in order to have a theoretically uni-
ed concept of tness that we can use to compare dierent individuals, tness in (1) and
(2) must be co-referential.
What concept of environment will accord with both (1) and (2)? It must be an environ-
ment that is expanded beyond an individual environment. But how far this environment is
to be expanded and what constitute its limits is dicult to discern. One contender for an
expanded concept of environment is a species environment. A species environment com-
prises all of the environmental inuences that impinge upon all of the individuals of a spe-
cies during their lives. Although attractive, for many species a species environment is far
too broad to t our criteria. If we are asking, for example, which of two Virginia opossums

12
Although it might be relatively clear what we mean by gene, it is not clear that genes should be thought to
exhaust the important hereditary material. In sexually-reproducing organisms, not only are genes inherited from
the parents, cytoplasmic material is inherited from the mother. If there are signicant dierences in cytological
material among individuals in a population, these dierences might be tness-aecting. But if the population lacks
signicant cytoplasmic heterogeneity, or if the heterogeneity makes little dierence in the cells functioning, then
the cytoplasm can be ignored. Henceforth, gene should be read as a placeholder for the hereditary material
that constitutes signicant dierences between individuals.
492 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498

living in a suburb of Memphis, Tennessee is the ttest, facts about a rainforest in Brazil,
where members of the species also reside, should not bear on the calculation. The subur-
ban opossums are in no signicant way causally interacting with individuals in the rainfor-
est in Brazil; it makes no dierence to them whether or not individuals in the rainforest
exists. It may be the case that some of the genes from the suburban opossums might even-
tually make it south to Brazil, but if this were to occur, it would likely take centuries. And
since what may occur centuries later has no bearing on how well these opossums are cur-
rently surviving and reproducing, this long-term gene ow does not bear on their tness.
Thus we need a concept of environment that is larger than an individual environment
but smaller than a species environment (for at least some species). One possibility for such
an environment is the selective environment (as introduced by Antonovics, Ellstrand, &
Brandon, 1988, and elaborated by Brandon, 1990). Although in some cases, the selective
environment would be the environment that xes tnesses, it is not always the case. The
reason is that an individuals tness might be xed by more than one selective environ-
ment. Thus a concept of environment that can encompass multiple selective environments
is needed. I will call this environment the tness environment. The question of how this
environment is delimited will be taken up in the next section.

6. Operationalization

A conception of tness and environment that cannot be applied to the natural world
that does not predict or explain the fates of biological entitiesis of little use. Because of
this I will now turn to the question of how Block Fitness can be operationalized.
Block Fitness, as mentioned above, is not based on the actual life that an individual
lives. It is based instead on all of the possible livespossible paths through space and
timeit could have lived. Some of these paths would include contracting a disease or
being killed while young, while other paths will be long and include numerous instances
of successful reproduction. Which path is realized is a matter of chance, but features of
the paths make some reproductive outcomes more likely than others. For example, con-
sider some component of tness, leaf waxiness. We might examine plants in a xeric envi-
ronment and observe that individuals with one allele always have very waxy leaves while
the individuals with an alternate allele have less waxy leaves. These components of tness
add up to make one variant tter than another in the following way: the waxy variant will
be tter if the number of successful reproductive events in the sum of its possible lives is
greater, or the variance smaller, than for the less waxy variant.
Given that Block Fitness is based on all of the possible life-paths an individual might
realize, it would seem impossible to measure an individuals tness. To know the tness
of an individual, one might object, we would have to know all of the innite number of
ways it might live its life. But since this is unknowable, an individuals tness is unknow-
able. To see why this objection is not a problem, we could make the analogous objection
about the coin: the probability that a coin will land heads in the box is not knowable since
there are an innite number of paths the coin might take through the box. But this objec-
tion cuts no ice since we can know what the probability of landing heads is for the coin.
One way to learn this is by examining the coin and the box and observing that the coin is
symmetrical, that the box has no way of recognizing and acting dierentially with respect
to either side of the coin, etc. From these data we can say with some condence that the
probability of the coin landing heads is 0.5. Another way to get at this information is to
 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498 493

ip the coin multiple times (or ip multiple coins of the same type) and see what trend
emerges. These two methods can be combined for more assurance. We can do the same
thing with organisms. We might examine the plants in the xeric environment and reason
that the cost of building the wax is less than the benet accrued from decreased water loss.
We could then judge that the waxy individuals are tter than the less waxy individuals. In
addition, we could conduct eld observations and see whether individuals of one kind tend
to have more ospring than the other kind. We need not consider every possible life a
plant might live. We just need a general understanding of the sum of these lives. Similarly,
we need not map out all of the possible paths a coin might travel through the box. We
merely need to know that the sum of the paths that end heads-up equals the sum that ends
tails-up.
As was evinced by the previous section, the concept of environment is central to the
concept of tness. Fitness comparisons are only properly made between individuals in
the same environment. But what delimits such an environment and how can it be deter-
mined whether two individuals occupy the same environment? I suggest that organisms
carve out their tness environment in the following way: an organism Os tness environ-
ment is the subset of its species environment that is likely to aect O, all of Os possible
mates, all of Os possible ospring, all of Os possible grandospring, etc. How do we ll
out the etcetera, that is, how many generations later are counted? This is not a question
that can be answered in the abstract. Rather, the number of generations depends on a vari-
ety of factors including O and Os possible mates traits, genetic constitution, and ecolog-
ical setting. For example, at one extreme might be a species in a very homogeneous
environment that lacks fertility-compromising mutations, in which one generation might
suce. At the other extreme are cases of progressive telomere shortening mortal germline
mutants that can take more than a dozen generations for sterile phenotypes to occur (see
Ahmed & Hodgkin, 2000). Since biologists often want to obtain estimates of tness from
eld observations, they need to know how many generations ahead to look in estimating
tness. Because this is an important point, I will illustrate some of the ways in which the
number of generations is determined.
One condition for Os tness environment is that O and Os descendents13 have a high
probability of covering the tness environment. What does it mean to cover an environ-
ment? For O to cover an environment is for O to be aected by each environmental factor
in roughly the proportion that each factor eectively exists in the environment. For exam-
ple, consider Figure 3a. The tness environment of O is illustrated. The shaded spots rep-
resent the presence of environmental factor f. In 3a it is apparent that O is likely to cover
the environment by itself, that is, O is in the presence of f in about the proportion that f is
in the tness environment. Because of this, a good tness estimate could be obtained for O
without having to follow its progeny through many generations or observe the reproduc-
tive output of many individuals similar to O. If O had a much shorter life or moved much
more slowly, however, it would be rather unlikely that it covered the tness environment.
This is represented by Figure 3b. In this case, the tness environment is likely to be cov-
ered only by O plus Os descendents.

13
Actually, the full characterization is Os descendents, relatives, and relatives descendents. But as mentioned
at the beginning of Section 3, for expositional simplicity, relatives (and their descendents) should be taken as
implicit.
494 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498

x o

Fig. 3. o denotes birth; denotes death. The line connecting o and denotes Os space-time trajectory. The
shaded spots represent the presence of environmental factor f.

In the above description of covering an environment I referred to the proportion that

each factor eectively exists. Why add eectively? In 3a and 3b O acted indiscriminately
with respect to factor f. But this does not have to be the case. It could be that O has an
aversion to f and that when it encounters the parts of its environment exhibiting f, it soon
moves to an f-free region, as depicted in 3c. In 3c, there is a disparity between the area
exhibiting f and the amount of time O is in the area exhibiting f. Because of Os disposi-
tion, f eectively constitutes a much smaller part of Os environment than might be
assumed.
It should be noted that this understanding of organisms (plus descendents) covering the
tness environment implies that the organism (plus descendents) and environment are
mutually constituting. To illustrate, the structure of the tness environment is an eect
of how many generations of descendents count toward creating Os tness environment.
The more generations, the broader the tness environment. But the number of genera-
tions of descendents that count is to a great extent determined by the environment
 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498 495

(a heterogeneous environment would need more iterations, a homogeneous fewer). This in

no way undermines the truth of my claims. Rather, it shows that modeling such situations
would be dicult. This is similar to the way in which an organisms genes and individual
environment are mutually constituting: the environment controls, to some extent, which
genes are expressed, but which genes are expressed determine, to some extent, the individ-
ual environment.
What other factors might determine the number of generations that one would have to
consider in estimating Os tness? If O were disposed to choose mates with bad genes,
for example, individuals with a higher than average number of deleterious recessive alleles,
and if O were disposed to pass this disposition on to its ospring, then several generations
later traits due to these alleles might emerge. This problem does not (phenotypically)
appear if one looks just one generation ahead. Clutch size in birds provides an example
of the need to look more than one generation ahead: a large clutch might be most produc-
tive in producing ospring during moist years, but unsustainable during dry years. Thus,
the tness would consist in performance in both moist and dry years. Both moist and dry
years would be covered by the descendents.
From these reections, it might seem that the number of generations that count toward
an organisms tness is likely very many, and that this militates against its operationaliza-
tion. While it is true that as the number of generations of descendents that count toward
an organisms tness increases, the probability that the tness environment is covered
increases. But there is good reason to think that these later generations do not count
toward an organisms tness. Assuming O reproduces sexually, its ospring are related14
to it by a factor of 0.5, its grandospring by 0.25, its great grandospring by 0.125, and so
on. It is quite apparent that ones descendent a dozen generations later bears little relation
to it relative to other conspecics. Also, since evolution is occurring, Os descendents many
generations later may not closely resemble O. This is why one ought to reject long-term
notions of tness.15
It might be thought that the way I dened a tness environment makes it no dierent
from a species environment. Recall that an organism Os tness environment is the subset
of its species environment that is likely to aect organism O, all of Os possible mates, all
of Os possible ospring, all of Os possible grandospring, etc. One might form this objec-
tion: if an organisms tness is in part constituted by its possible mates and the tness of its
possible mates is in part constituted by their possible mates, then the chain of possible
mates might reach every reproductively viable individual in the species. The opossums
in Memphis might be in the same tness environment as the opossums in Brazil after all.
The reason that this is not the case is the following. There are two components to Os
tness that are constituted by its possible mates, one genetic and another phenotypic. The

14
Related, that is, relative to random individuals in the population.
15
A conception of tness that consists in the fate of descendents many generations later was argued for by
Thoday (1953). His concept of tness takes into account 108 generations. More recently, Cooper (1984) takes
tness to be the expected time to extinction, ETE: individual x is tter than y if ys lineage is expected to go extinct
rst. The main problem with these long-term approaches is that selection is not future-sighted. Selection acts on
organisms as they currently are, not what might happen to their progeny thousands of generations later. Also,
since evolution is taking place, each ospring is somewhat dierent from its parents and these dierences can
accrue. What goes extinct 108 generations from now might not even belong to the same species whose tness we
want to determine. Thus these long-term notions do not seem to be the kind of concepts that can explain current
reproductive success.
496 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498

genetic component travels only vertically (from parent to ospring). Lets assume that the
individuals in Os group mate with multiple individuals and (at least occasionally) bear o-
spring from multiple mates. How well Os ospring fare is due in part to the genes that the
ospring received from Os mate. And how well all of Os mates ospring fare is in part a
function of the genes Os mates ospring received from Os mates mates. But there is no
gene ow from Os mates mates to Os ospring, that is, there is no horizontal or oblique
genetic transmission. Because this is the case, the pool of individuals that aect Os tness
via direct genetic means extends no further than Os possible mates. Thus, with this genetic
component there is no possibility that the opossums in Brazil will be within the tness
environment of an opossum in Memphis.
But what about the phenotypic component? If Os tness is in part constituted by its
phenotypic interaction with its possible mates, and if the tness of each of Os possible
mates is in part constituted by its phenotypic interaction with their possible mates, than
it seems that one could follow this possible mates chain until every reproductively viable
individual has been included. From this it would seem to follow that individuals in Brazil
and Memphis might inhabit the same tness environment. The reason that this is not the
case is that ones possible mates only partially constitute ones tness. Having the possible
mates that O has (relative to some other set of possible mates) is tness altering by some
fraction 1/x. Having the possible mates that Os possible mates have (relative to some
other set of possible mates) is tness altering by some fraction 1/y. The fractional change
to Os tness of a change in possible mates of Os possible mates is 1/xy. A change in the
next shell of possible mates would alter Os tness by factor 1/xyz, and so on. So as one
follows the possible mate chain, the extent to which the possible mates aect Os tness
undergoes a geometric decay. Thus after a small number of links the eect essentially goes
to zero.
We have seen that although an individuals tness consists in a vast number of environ-
mental factors, it is nonetheless possible to gain knowledge of an organisms tness. In
fact, I would argue that Block Fitness is often implicitly assumed in biology: if, for exam-
ple, a Drosophila researcher thinks an individual y of type A is tter than an individual of
type B in some common environment E, she will not take this to be disproved if a type A
individual produces fewer descendents than a type B individual. Instead, she will test her
hypothesis by looking at the number of ospring produced by a large number of A and B
individuals in E and take the average. Although some biologists erroneously dene tness
as this average (and thus mistake tness for the measurement of tness), this process
assumes Block Fitness. Because of this, I hold that the operationalization of Block Fitness
is not problematic, and in many cases would not prescribe biologists to make dierent
observations or conduct dierent experiments (though it might prescribe a dierent way
of talking about what they are doing).

7. Conclusions

We have seen that (1) an individuals tness is xed over its lifetime, that is, its tness
cannot increase or decrease with time, and (2) tness comparisons can only occur between
conspecics that occupy the same tness environment. What implications does this have
for ecology and evolution? The view implies that nothing that an organism does or has
done to it aects its tness. The honeybee that attacks and stings an interloper will not
thereby alter its tness. A meerkat disposed to spend a disproportionate amount of time
 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498 497

on sentinel duty may thereby have a lower tness than another that is disposed to shirk its
duty. But no meerkat lowers its tness by acting as a sentinel. Similarly, a soldier does not
lower her tness by charging forth in battle. Indeed, the soldier may have a low tness and
this low tness may be an eect of her disposition to act selessly, but her low tness is not
an eect of how she actually acts. Her seless actions might be evidence for her low tness,
but they do not constitute her tness. It could be the case that all of the individuals in bat-
tle have the same tness, but that some act selessly due to the ne-grained heterogeneity
of their environment (Levins, 1964). Similarly, interactions between individuals do not
alter tness. What game theory shows us is not how individual tness is increased or
decreased through behaving in certain games in certain ways. Rather, via game theory
we can learn why organism O, being disposed to participate in certain games in certain
ways, is more or less t than other organisms with diering dispositions.
This conception of tness thus has a profound eect on how we should think about the
consequences of particular occurrences in an individuals life. It also changes the way we
ought to think about the interactions between individuals (or groups). These interactions
are often characterized as being, for example, mutualistic, parasitic, commensalistic, or
altruistic. And these interactions are characterized in terms of tness changes. For exam-
ple, an altruistic act is often said to be one in which the recipient gains (or increases in)
tness while the altruist (or donor) decreases in tness.16 But, as we have seen, tness is
not something capable of change. Does this mean that the above view of tness does
not allow for altruism (or mutualism or parasitism or any such relationship) to obtain?
No. An altruistic act could be characterized as an act such that (1) an individual disposed
to perform the act is less t than another individual lacking this disposition and (2) an
individual disposed to be a recipient of such an act is tter than another individual lacking
this disposition.
A xed tness also bears on the levels of selection debates. Those advocating multi-level
selection often speak of within-group tness and overall tness. The inference from
individuals with trait x having a higher overall or average tness than individuals
lacking trait x (other things being equal) to the claim that x is promoted by individual
(and not group) selection is rightly argued against by Sober & Wilson (1998). The mistake
they and others make is to think that individuals have multiple tnesses (say, one overall
and another group relative). This is a mistake because O cannot have tness x within
group G, but have an overall tness of y, where x 6 y. The tness xed by Os tness envi-
ronment is its unique tness. And it is an empirical fact whether or not Os tness environ-
ment transcends G. If Os environment does transcend G and includes groups H and I,
does it make any sense to speak of its within-group tness, that is, its tness within G?
It does, but only hypothetically. To say that Os within-group tness is x is to say that
if Os tness environment included only individuals in G, its tness would be x. What good
is this counterfactual within-group tness? It sheds light on the process of selection. Selec-
tion may be occurring at various levels of organization, and an organisms tness environ-
ment may include more than one selective environment, but there is only one non-
counterfactual tness. One remedy to the within-group and overall conceptions of tness
is to speak, instead, of components of tness. Os being in group G relative to some other
group H is in part constitutive of its tness. This part can be described as a component of
Os tness due to its residence in that group.

16
See the quote by Sober & Wilson (1998) in Section 1.
498 G. Ramsey / Stud. Hist. Phil. Biol. & Biomed. Sci. 37 (2006) 484498

Acknowledgements

I thank Robert Brandon and Alexander Rosenberg for reading and commenting on
multiple drafts of this paper. I also thank Marshall Abrams, Heather Heying, David Kap-
lan, Stefan Linquist, John Longino, Mohan Matthen, Donald Morisato, Maya Parson,
and Bret Weinstein for their helpful comments and suggestions. A previous version of this
paper was presented at the 2003 meeting of the International Society for the History, Phi-
losophy, and Social Studies of Biology.

References

Ahmed, S., & Hodgkin, J. (2000). MRT-2 checkpoint protein is required for germline immortality and telomere
replication in C. elegans. Nature, 403, 159164.
Antonovics, J., Ellstrand, N. C., & Brandon, R. N. (1988). Environmental variation and genetic variation:
Expectations and experiments. In L. D. Gottlieb, & S. K. Jain (Eds.), Plant evolutionary biology (pp. 275303).
New York: Chapman & Hall.
Basolo, A. L. (1990). Female preference predates the evolution of the sword in swordtail sh. Science, 250,
808810.
Basolo, A. L. (1995). Phylogenetic evidence for the role of a pre-existing bias in sexual selection. Proceedings of
the Royal Society of London B, 259, 307311.
Beatty, J. H., & Finsen, S. (1989). Rethinking the propensity interpretation: A peek inside Pandoras box. In M.
Ruse (Ed.), What the philosophy of biology is (pp. 1730). Dordrecht: Kluwer.
Brandon, R. N. (1978). Adaptation and evolutionary theory. Studies in History and Philosophy of Science, 9,
181206.
Brandon, R. N. (1990). Adaptation and environment. Princeton: Princeton University Press.
Cooper, W. S. (1984). Expected time to extinction and the concept of fundamental tness. Journal of Theoretical
Biology, 107, 603629.
Diaconis, P. (2004). The search for randomness. American Association for the Advancement of Science Annual
Meeting, February 14, Seattle.
Futuyma, D. (1986). Evolutionary biology (2nd ed.). Sunderland, MA: Sinauer.
Levins, R. (1964). Evolution in changing environments: Some theoretical explorations. Princeton, NJ: Princeton
University Press.
Matthen, M., & Ariew, A. (2002). Two ways of thinking about tness and natural selection. The Journal of
Philosophy, 49, 5583.
Mills, S. K., & Beatty, J. H. (1979). The propensity interpretation of tness. Philosophy of Science, 46, 263286.
Sober, E. (2001). The two faces of tness. In R. Singh, D. Paul, & J. Beatty (Eds.), Thinking about evolution:
Historical, philosophical, and political perspectives (pp. 309321). Cambridge: Cambridge University Press.
Sober, E., & Wilson, D. S. (1998). Unto others. Cambridge, MA: Harvard University Press.
Thoday, J. M. (1953). Components of tness. Symposia of the Society for Experimental Biology, 7, 96113.
Walsh, D. M., Lewens, T., & Ariew, A. (2002). The trials of life: Natural selection and random drift. Philosophy
of Science, 69, 452473.