You are on page 1of 13

bs_bs_banner

Journal of Applied Philosophy,Vol. 30, No. 1, 2013


doi: 10.1111/japp.12001

Keeping Score for Causal Claims: Causal Contextualism


applied to a Medical Case

CEI MASLEN

ABSTRACT This article investigates how Causal Contextualism applies in a medical context.
It is shown how the correct interpretation of some medical causal claims depends on relevant
alternatives and then argued that these relevant alternatives are determined by standards of
practice and practical limitations (of equipment, personnel, expertise, cost), amongst other
factors. Causal Contextualism has recently been defended by a number of philosophers; however
details of the relevant factors determining content in different contexts have been lacking. It seems
to me that establishing such details of Causal Contextualism goes a long way towards making
the view plausible, and is also necessary for discovering some important consequences of Causal
Contextualism.

1. Introduction

This article investigates how Causal Contextualism applies in a medical context.


Although Causal Contextualism has recently been defended by a number of philoso-
phers, the view has never been developed in a lot of detail. Before introducing the view
I begin with a brief discussion of Epistemic Contextualism, as this is more likely to be
familiar to readers and will be helpful in introducing some features of contextualist
views.
Epistemic Contextualism is the view that the word ‘know’ contributes an element of
context dependence to expressions in which it occurs (in addition to any context-
dependence arising from tense and other context-dependent vocabulary in the expres-
sions). For example, truth conditions of statements of the form ‘S knows that P’ and ‘S
does not know that P’ involve facts about the context in which those statements occur,
such as the interests of the speaker.This view has risen in popularity over the past twenty
years. Defenders have elaborated it in detail, and drawn surprising conclusions, and
detractors have subjected it to vigorous attack.1
Epistemic Contextualism is a general view, but here is a simple version of how this
might work. In a high stakes context (that is, a situation in which the costs of being wrong
are high) the claim ‘S knows that P’ expresses the proposition that S’s belief that P is at
least strongly justified; while in a low stakes context, the claim ‘S knows that P’ expresses
the proposition that S’s belief that P is at least weakly justified. For example, suppose that
in the course of an idle discussion I say ‘Fred knows that this is the way to the hospital’.
In this context, ‘Fred knows that this is the way to the hospital’ expresses the proposition
that Fred has at least weak justification for his belief that this is the way to the hospital.
But suppose now that someone’s life is at stake and we are counting on Fred to get him
to the hospital. In this second context, ‘Fred knows that this is the way to the hospital’

© Society for Applied Philosophy, 2012, Blackwell Publishing, 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main
Street, Malden, MA 02148, USA.
Keeping Score for Causal Claims 13

expresses the proposition that Fred has at least strong justification for his belief that this
is the way to the hospital. Suppose that in fact Fred only has moderate justification for
his belief. Then it follows from this view that the very same knowledge claim is true in
the first context but false in the second context.
It is possible to define a parallel view about causation: Causal Contextualism is the
view that the word ‘cause’ contributes an element of context dependence to expressions
in which it occurs (in addition to any context-dependence arising from tense and other
context-dependent vocabulary in the expressions). For example, truth conditions of
statements of the form ‘c is a cause of e’ and ‘c is not a cause of e’ involve facts about the
context in which those statements occur, such as the standards or interests of the
speaker. Who would hold such a view of causation and why? Causal Contextualism has
not yet enjoyed the same popularity or even unpopularity that Epistemic Contextualism
has; it has simply remained relatively unknown until recently. Versions of Causal Con-
textualism were first presented over a decade ago with compelling presentations by
Hitchcock (1996), Holland (1986), and Horgan (1989), for example.2 More recently,
interest has been increasing with developments from Menzies (2004), Price (2007),
Schaffer (2005a), Woodward (2003) and others.3 The examples and arguments that can
be used to argue for Causal Contextualism are similar to those used by epistemic
contextualists.
In Section 2, I describe a medical example in some detail and argue that Causal
Contextualism applies to it. This is not intended to give a full defence of Causal
Contextualism. For that I refer the reader to the writings just mentioned, which I take to
already have made a strong case for Causal Contextualism. Rather the example is
intended to illustrate one plausible detailed medical application of Causal Contextual-
ism. In Sections 3–5, using the same medical example, I investigate some mechanisms by
which the truth of causal claims is fixed by context. This task has been neglected by
causal contextualists. However, much has been written by epistemic contextualists on
the parallel task of describing mechanisms by which the truth conditions of knowledge
attributions are said to be fixed by context, and it will be helpful to use these as a
comparison in developing a picture for causation.
Finally, in Section 6, I address one general question about Causal Contextualism.This
is the question of whether the state of knowledge or ignorance of the speaker prevents
him or her from making meaningful causal claims. I will argue that it does not, using
what we have learned about how context determines alternatives.

2. A Medical Example4

Suppose that a patient comes to a small public hospital in New Zealand with a non-
malignant tumour blocking the left side of her bowel. The patient will not survive long
without treatment, so the surgeon recommends immediate bowel resection (removal
of the diseased portion of the intestine and suturing of the healthy ends together).
The patient consents and the surgeon successfully performs this operation to remove
the tumour. Unfortunately, the patient dies several days later from complications of the
surgery (leakage from the reconnection of the intestine).
Was the decision to perform bowel resection a cause of this patient’s death? As we shall
see, the correct answer to this question may depend upon the context of enquiry.

© Society for Applied Philosophy, 2012


14 Cei Maslen

Suppose that the surgical team’s internal audit finds that the decision to perform
bowel resection was not a contributing factor to the death, on the grounds that bowel
resection was at least as good as available alternative procedures. Although the death did
arise from complications of this procedure, the decision to perform bowel resection in
particular was not a contributing factor; it was not at fault, and the level of care provided
did meet existing standards.
Let us suppose that the only alternative procedure that the surgical team considered
relevant in their audit was the standard alternative procedure of colostomy (removal
of the diseased portion of the intestine and connecting it through the abdominal
wall to an external bag). Suppose that randomised controlled trials have found colos-
tomy to have a mortality rate of 20% compared to a mortality rate of 12.5% from
bowel resection without colostomy. If they had decided to perform colostomy instead,
then the chance of death would have been even higher. This does seem to indicate
that the decision to perform bowel resection was not a contributing factor to the
death.
However, in a different context of enquiry another procedure may also be considered
relevant. Suppose that a medical researcher comes across this case while collecting data
for her research aimed at designing international standards. In judging the causes of
death in the very same case she may take a wider range of procedures into account.
Standard procedure for treating acute bowel obstructions in Europe and the United
States is endoscopic placement of a self-expanding metal stent (SEMS). Endoscopic
SEMS placement is a relatively new procedure, and SEMSs are very expensive, but
recent studies have shown endoscopic SEMS placement to have considerably lower
mortality rates as compared with bowel resection or colostomy. Citing these lower
rates, the medical researcher comes to a seemingly conflicting conclusion about the
very same New Zealand case. She concludes that the bowel resection was a cause of
the death on the grounds that it was greatly inferior to the alternative procedure of
endoscopic SEMS placement.
Note that endoscopic SEMS placement would not have been an available alter-
native in the small New Zealand hospital for two main reasons. Firstly, no surgeons
in that hospital had the training and expertise to perform this operation, and trans-
porting the patient to a tertiary centre would not have been possible in time. Secondly,
SEMSs are not kept in stock at small hospitals in New Zealand. They need to be
pre-ordered and this would not have been possible in time. Given these practical
limitations and the standard alternatives, the finding of the surgical team in their
internal audit does seem to be justified. Yet, the medical researcher also seems to be
justified in taking another procedure into account, and justified in her subsequent
finding.
The two investigations reached apparently conflicting conclusions about the very same
case. So, which is correct? The Causal Contextualist can say that both investigations are
correct, and explain why the disagreement is only apparent. The different conclusions
reflect a difference in perspective, not a difference in facts. Bowel resection was not a
cause of death, relative to the only other available standard procedure, colostomy.
However, bowel resection was a cause of death, relative to a wider range of known
procedures including endoscopic SEMS placement. One range of alternatives was
relevant in the first context while a different range of alternatives was relevant in the
second context.

© Society for Applied Philosophy, 2012


Keeping Score for Causal Claims 15

3. The Standard Counterfactual Approach to Causation and a Working


Contextualist Analysis

Standard analyses of causation are not contextualist.They do not allow for both medical
investigations to be correct, or for the truth of causal claims to be relative to alternative
procedures. At the end of this section I will present a simple contextual analysis of
causation that does allow these claims to be true.This involves only a small modification
to a standard analysis of causation.
One of the most popular approaches to analysing causation is in terms of counterfac-
tuals, that is, subjunctive conditionals. The basic idea of this approach is that a cause is
anything that if it had not occurred then the effect would not have occurred. Lawyers
sometimes use this approach and call a cause a sine qua non — that without which the
effect would not have occurred.5 A simple probabilistic version of this requires only that
if the cause had not occurred then the probability of the effect would have been much
lower.6 Applying this standard analysis to our example, the decision to perform bowel
resection was a cause of death if and only if had the decision to perform bowel resection
not occurred then the probability of death would have been much lower. Which alter-
native procedure is relevant here? The answer is that it is the alternative procedure that
would have been chosen otherwise — the second-choice procedure, as I will say. An
assumption of the standard analysis is that that there is always a unique second-choice
procedure and that it does not depend on context.7
Our example shows the standard analysis to be unsatisfactory on at least two counts:
in not allowing for context-dependence, and in its crucial dependence on the second-
choice procedure. Firstly, let’s suppose (plausibly) that had the doctors in question not
chosen to perform bowel resection, then the alternative procedure they would have
chosen is colostomy (with a higher probability of death). The standard analysis then
yields the result that the decision to perform bowel resection is not a cause of death, as
it did not raise the probability of death relative to the second-choice procedure. This
works for the first context, but yields a counterintuitive result for the context of the
medical researcher, for the standard analysis also judges that the decision to perform
bowel resection is not a cause of death in this second context. The second-choice
procedure is the procedure that the New Zealand doctors would have chosen if they had
not made their actual choice, and this is the same regardless of whether it is an internal
audit investigating the question or an independent medical researcher. So the standard
analysis does not account for the context dependence in our example.
Secondly, suppose that, as it happens, had the decision to perform bowel resection not
been made, then the doctors in question would have chosen not to operate on this
patient at all — they would have made a negligent decision. The standard analysis
implausibly implies that this fact about the second-choice procedure is relevant to the
question of whether or not the actual procedure is a cause of death. (It is hard to imagine
how we would find firm evidence for such a fact about the psychologies of these doctors.
Evidence might be the consultant later confessing that he was on the verge of a mental
breakdown that day, and was close to making the negligent decision. Though it is not
unreasonable to think that in many other cases there is no determinate fact of the matter
as to the second choice.)
On the supposition that the second-choice procedure was to choose not to operate,
the standard analysis again yields the judgment that the bowel resection was not a

© Society for Applied Philosophy, 2012


16 Cei Maslen

cause of death, which is an intuitive end result in the context of the internal audit.
However, a variation on our original example quickly shows that this dependence on
second-choice procedure is unsatisfactory. As a variation, relocate the example to a
hospital in America where endoscopic SEMS placement is an available procedure.
Again the doctors made the actual choice of bowel resection and the patient died. If we
suppose the second-choice procedure to be endoscopic SEMS placement, then the
standard approach yields the judgment that the actual choice was a cause of death (as
it raised the probability of death relative to endoscopic SEMS placement). On the
other hand, if we suppose the second-choice procedure to be not to operate, then the
standard analysis judges that the actual choice was not a cause of death (as it lowered
the probability of death relative to the alternative of no operation). What a peculiar
result! It seems that mentally choosing a worse second-choice procedure (no operation)
can make a difference as to whether or not something is a cause of death. Also, if in
the case of a second-choice not to operate the doctors are no longer implicated in the
cause of death, the standard approach to causation has the consequence that mentally
choosing a worse second-choice procedure can make all the difference as to whether
someone is to blame for a crime. I think this is an unsatisfactory consequence of the
standard analysis to causation.
We have concluded that the answer to the question of which alternative procedure
would have been chosen otherwise (even supposing that there is a determinate answer to
this question) is sometimes completely irrelevant. One possible lesson to draw from our
example is that causation depends on what should have happened otherwise, rather than
on what would have happened otherwise. ‘Should have happened otherwise’ can be
understood as ‘should have happened, subject to the aims, availability and norms in that
context’, and even if there is some widest sense of ‘should have ideally’, this sense is not
normally relevant. Although ‘should have happened otherwise’ does seem to capture the
contextual shifts in our example that ‘would have happened otherwise’ fails to capture,
we will see in the next section that we need to allow for a dependence on an even wider
range of alternative procedures.
We can easily modify the standard analysis given above in order to fit our contextual
conclusions. Unlike the standard analysis, we will not assume that the appropriate
alternative to the cause depends on a determinate answer to the question ‘What would
have happened without the cause?’ Instead our analysis will explicitly state a dependence
on a set of alternatives implicit in the context.8 A number of contextual features
determine this set, as we shall see later: assumptions, goals, plans, limitations, and
standards — all of which may be partly pragmatic. I expect that the set is frequently only
partly determined by the context — it is left vague — and that this indeterminacy only
occasionally leads to miscommunication.
Here is the analysis:9
Where C* is the set of relevant alternatives to the cause in context C: ‘c is a cause
of e’ is true in context C iff, for every c*i in C*, had c not occurred and c*i had
occurred instead, then the probability of e would have been much lower.
Note that implicit alternatives to the effect can also vary. For example, with the medical
example given, the obvious contrast with dying is staying alive. However, in other
cases, the alternative of never having existed could be a relevant contrast with dying.10
In the main example used throughout this article, alternatives to the effect do not seem

© Society for Applied Philosophy, 2012


Keeping Score for Causal Claims 17

to be relevant, so, for the purposes of this article, I have not stated a more general analysis
incorporating alternatives to the effect.11
This may be a good place to distinguish some closely related concepts. Singular
causation is causation in a single case, for example where smoking causes cancer in a
particular patient; general causation does not concern a particular case, for example,
smoking also causes cancer in general.We have been concentrating on singular causation
with our medical example, and we shall put questions of general causation aside
throughout this article. A second important distinction may be less familiar. This is the
distinction from Hitchcock between component cause along a causal route and net cause.12
A component cause along a causal route is something that contributes to an intermediate
step in one of the pathways to the effect, while a net cause is something that makes an
overall difference to the effect. For example, the surgery contributed to the leakage,
which in turn contributed to the infection, which in turn contributed to the death. Hence
bowel resection is undeniably a component cause along a causal route to the death and
this is true in both contexts described above. I agree with Hitchcock that the word ‘cause’
may be commonly used to mean either of these concepts.We will restrict ourselves to net
causation here, unless otherwise stated.

4. Causal Contextualism and Relevant Alternatives

What determined the relevant alternatives in the different contexts of our medical
example? In the first context (the clinical audit of the surgical team in the small New
Zealand hospital) there seems to be a restriction to procedures that conform to New
Zealand professional standards. This first restriction arises naturally from the accepted
goal of this type of audit, which is to review the quality of surgical care by comparing
recent past practice with recognised standards and criteria. The professional standards
here are related to legal requirements and to typical surgical practice in this country. In
New Zealand, some legal criteria for hospitals are stated by the Ministry of Health and
the local district health boards, and legal requirements on doctors also depend on typical
recent surgical practice in New Zealand, because legal cases may appeal to a panel of
experts describing their own practices.
There also seems to be an implicit restriction to available surgical procedures in the
sense of the procedures that the surgical team had the ability to carry out at that time and
in those circumstances.That is, for the purposes of this audit, the surgical team seems to
accept as fixed the limitations of personnel, technical expertise and equipment that they
had to work with at the time and place that the patient presented.13 Although the surgical
team may have some influence over attainment of more equipment, training and per-
sonnel for a later date, this would not be relevant in this first context. This second
restriction also arises naturally from the accepted goal of this specific type of audit, which
is only to review the surgical team’s own past performance and judge whether they had
provided the best surgical care that they could in the circumstances, or at least whether
they had provided adequate surgical care in the circumstances.
In the second context (of the medical researcher), a wider range of alternative proce-
dures was relevant but there are still implicit restrictions in place. Given norms of
evidence for medical research and the explicitly stated goal of designing international
standards, the contrast class was limited (roughly) to those procedures that have been

© Society for Applied Philosophy, 2012


18 Cei Maslen

used in the past for treating the condition, whose results have been investigated by
well-run studies. Some careful researchers may explicitly state their selection criteria for
alternative procedures, for example a limitation to procedures whose results have been
published in certain peer-reviewed medical journals.
In conclusion, we have found that relevant alternative procedures were determined by
a number of factors: aims, abilities, professional and legal standards, epistemological
norms, and availability of equipment, personnel and expertise.We can group this answer
into three main groups: aims, availability of alternatives, and norms.

5. Conversational Score and Conversational Salience

In the previous section, we looked at contextual factors that determine relevant alter-
natives for Causal Contextualism. This was the main goal of this article, but it is a task
that other causal contextualists have set aside, perhaps thinking it is a tedious linguis-
tics task with not much philosophical import, or perhaps suspicious that such factors
need to be derived from more general and fundamental linguistic principles. We have
stated a general analysis for Causal Contextualism that relies on an intuitive notion of
relevant alternatives. But it seems to me that we need to see the details of how Causal
Contextualism applies to real cases, before we can evaluate Causal Contextualism,
or to investigate its consequences. I also think that discovering the details of how
relevant alternatives can be found in the context is helpful for understanding how
causation can depend on context, and for dispelling some mystery that surrounds
Causal Contextualism.
Lewis was one of the first to see the need for such detailed discussion of contextual
features in the case of Epistemic Contextualism, and we will look at this soon.14 He
also introduced the notion of conversational score, which can help us to fit our project
into a more general linguistic framework.15 Lewis suggests that it is often helpful to
think of a conversation as a language game in which there are features analogous
to the components of the score of the game. Just as a baseball game has a scoreboard
to keep tally of innings, runs, strikes, outs etc., well-run conversations have features
such as presuppositions, standards of precision, and salient objects which can be
thought of as components of an abstract scoreboard. These components persist or
change in an orderly manner as the conversation progresses: as things are said and
events occur. Lewis’s Epistemic Contextualism depends on the notion of a relevant
possibility, which is another component of conversational score. Similarly, our relevant
alternatives to the cause function as a component of conversational score in Lewis’s
sense. They persist throughout a conversation while goals, standards and other limi-
tations persist, and later in this section we will also see how they depend on what is
said in a conversation.
As I said, in a later paper Lewis goes beyond these general comments on Epistemic
Contextualism and scorekeeping to investigate the details of how knowledge claims
depend on context.16 The most important rule he introduces is a rule of conversational
salience, and for the remainder of this section I want to focus on versions of this rule and
how they might apply to Causal Contextualism. As we shall see, conversational salience
plays a part in Causal Contextualism too, but I will argue that it is not the primary factor,
in medical contexts at least.

© Society for Applied Philosophy, 2012


Keeping Score for Causal Claims 19

Conversational Salience

In the case of Epistemic Contextualism, ‘The salience rule — or something like it — is


the most common tool of contextualists’.17 This is the rule that an alternative that is
salient in a conversational context thereby becomes a relevant alternative for the
purposes of interpreting the contextual claim. An alternative may be salient because
someone recently mentioned it in the conversation, or perhaps because the speaker
simply had it in mind. Lewis states the salience rule for Epistemic Contextualism like
this: ‘No matter how far-fetched a certain possibility may be, no matter how properly we
might have ignored it in some other context, if in this context we are not in fact ignoring
it but attending to it, then for us now it is a relevant alternative.’18 (This is his ‘Rule of
Attention’.) For example, suppose someone says, ‘I know that this is a zebra’, but that
same person has recently mentioned or thought of the possibility that the animal in front
of him is a cleverly painted mule.Then it follows from the Rule of Attention that he does
not know that it is a zebra unless he can rule out that alternative. The mere thought or
explicit mention of the alternative makes it a relevant alternative that cannot be properly
ignored. Other epistemic contextualists agree with Lewis, though the formulations of the
rule are quite varied.19
A parallel principle for Causal Contextualism would be a principle that explicitly
mentioning an alternative ensures that it is a relevant alternative in that context. Does
such a principle hold for the medical example we looked at? Suppose that an additional
alternative is explicitly mentioned during the surgical team’s audit meeting — say, the
alternative of endoscopic SEMS placement. Would that be enough to make it a relevant
alternative? If it were enough, then after explicit mention were made the team would be
forced to conclude that the decision to perform bowel resection was a cause of death, and
that adequate care had not been given. But, on the contrary, endoscopic SEMS place-
ment would still be irrelevant, for the same reasons we examined above — because it was
not an available alternative, and was not required by the standards in place. Simply
mentioning it as an alternative, or even seriously entertaining it as an alternative, is not
enough to make it relevant to evaluating the current outcome here.
Turning now to the second context we discussed above — the medical researcher
constructing a meta-analysis — it seems that this context has more flexibility for
constructing and modifying standards and goals by explicit mention. Would explicitly
mentioning an additional alternative procedure be enough to make it relevant in this
context? I think it would, provided that it doesn’t contradict other explicitly stated
selection criteria for alternative procedures. (If the explicitly mentioned procedure is
unusual enough — say based on a new religious ritual with no data as yet on its success
— that would certainly serve to reduce the quality of the study. However, unless it
contradicts other explicit selection criteria for alternative procedures, interpretation of
the causal claim would still unambiguously include it and the causal claim may even
happen to be true, though lacking in evidence.)
Conversational salience is the primary rule for epistemic contextualists not just
because it is their most common tool, but also because they claim it to be a factor that
trumps all other factors. For example, if one of Lewis’s other rules entails that a
possibility is not relevant, but his salience rule (the Rule of Attention) entails that it is
relevant, the salience rule always has priority. For example, one application of Lewis’s
Rule of Reliability entails that possibilities involving hallucination are not relevant

© Society for Applied Philosophy, 2012


20 Cei Maslen

(on the grounds that our process of perception is normally reliable). However, if
someone explicitly mentions the possibility that we are merely hallucinating that there is
a zebra in front of us, then according to Lewis that automatically becomes a relevant
possibility that must be ruled out in order for knowledge to obtain.
In the case of Causal Contextualism, this is not so. In our first context, even explicit
mention of an alternative was not enough to make it relevant, when it was neither
available nor required by medical standards. Hence, in the case of Causal Contextual-
ism, dependence on aims, standards and availability can trump conversational salience.
Note that a dependence on norms and availability of alternatives does not fit with
some understandings of how contextualist views should work. In particular, it involves
a departure from what Stanley calls the intention-based view. This is the view that the
pragmatic features of the context that determine the proposition expressed are the
referential intentions of the attributor.20 In other words, it is the idea that any con-
text shifts in the meanings of words must arise somehow from the speaker himself
intending the meaning to shift. A simple contextualist view based only on conversa-
tional salience fits this idea well. So, if I intend alternative A to be the one relevant
alternative for my utterance of ‘c caused e’, then the meaning of my utterance is ‘c
caused e, relative to alternative A’. This is the simplest caricature of contextualism,
but some still wish to saddle contextualism with it. However, I see no reason why
Causal Contextualism should be limited by the intention-based view. A registrar
presenting the surgical audit may intend her causal claims to be relative to a wide
range of surgical procedures, but this is not enough to make the appropriate inter-
pretation of her claims relative to these procedures, for that is determined by other
considerations.

Vetoes
Epistemic contextualists are divided on whether alternatives can be made salient in a way
that makes it clear that they are not relevant. This could happen when someone finally
says, ‘Forget cleverly painted mules for now, okay?’ and thereby makes it the case that the
possibility that it is a cleverly painted mule is not a relevant alternative. DeRose considers
something like this when; he talks about ‘veto power’ and the ‘Aw, come on!’ response to
the mentioning of a sceptical hypothesis.21
I think that this type of conversational salience does play a role in Causal Contextu-
alism. Causal claims arise commonly in a wide range of medical contexts and thus far we
have only looked at two contexts. Let us consider briefly just one further context — an
informal conversation between medical experts about the very same case in New
Zealand that we have looked at throughout. Suppose that the conversation takes place in
the tea-room, outside of any formal meeting or enquiry. One expert says, ‘The bowel
resection decision wasn’t the cause of death. You know if they had used endoscopic
SEMS placement like we do back home, then the patient would have stood a good
chance.’ The second expert then replies, ‘That’s beside the point.You know that wasn’t
going to happen here.’
It seems to me that in this context, the content of the first expert’s claim is determined
in part by the second expert’s rebuff. The second expert did not agree to the relevance
of the proposed alternative, so the first expert’s mention of it did not succeed in making
it relevant.

© Society for Applied Philosophy, 2012


Keeping Score for Causal Claims 21

We have now looked at causal claims occurring in several different contexts about
the same medical case. I’m not claiming to have listed all the contextual factors
that determine relevant alternatives for Causal Contextualism.22 However, from the
factors we have looked at, it may be puzzling that we have ended up with so many
factors that eliminate relevant alternatives (limitations, standards, vetoes), and only
one factor (conversational salience) that generates relevant alternatives. In cases where
no alternatives are conversationally salient to begin with, aren’t we always going to
be left with no alternatives? This is not a problem for epistemic contextualists
because of their primary emphasis on conversational salience, which generates relevant
possibilities.
I am unsure what the best answer to this question is. One promising suggestion for an
answer is that as a default, interpretation of a causal claim is always relative to the widest
possible range of alternatives, which is then narrowed down by other contextual factors.
So with our medical example, a default interpretation of the claim ‘the actual procedure
is a cause of death’ is to interpret it as relative to all possible alternative procedures to the
actual procedure.23

6. Do We Have to Know What We’re Talking About?

It has been suggested to me that it doesn’t seem right that the truth of a causal claim may
depend on the state of ignorance or knowledge of the speaker.24 Carroll also suggests that
there may be contexts where no suitable alternatives to a cause are salient and a causal
utterance fails to be true as a result. He calls these contexts impoverished contexts.25
For example, we may worry that a patient with very little medical knowledge will be
unable to successfully make meaningful medical causal claims. Certainly someone who
accepted the intention-based view might expect this. If the patient is not familiar with
any alternative procedures then she may not be able to have appropriate referential
intentions, and hence her causal utterances may count as false or meaningless. Note that
what is in question here is not whether such a patient can know that a medical causal
claim is true or meaningful, but whether it can even be true or meaningful coming from
her mouth.
Returning to the same medical example, suppose that our patient (prior to death) had
predicted that bowel resection would cause her death, but that she had no medical
knowledge of alternative procedures. What are the appropriate relevant alternatives with
which to interpret her claim, and would her causal claim be automatically meaningless
or false due to her lack of knowledge?
There are several reasons why it does not follow from Causal Contextualism that an
ignorant speaker is unable to utter true causal claims. Admittedly, when we don’t know
what we’re talking about we may be unable to make salient some more esoteric alter-
natives. But our causal claims can still be interpreted straightforwardly and simply and
may still be true. My answer to the question has several parts, for I think the appropriate
interpretation of the patient’s claim does depend on her intentions.
The first part of the answer is that even the most ignorant patient is aware of the ‘do
nothing’ option.With the ‘do nothing’ option as the only relevant alternative the patient’s
prediction that bowel resection would cause her death is false, as doing nothing had a
higher chance of death than the actual procedure. But at least her causal claim is not

© Society for Applied Philosophy, 2012


22 Cei Maslen

meaningless.26 Now it may be that the ‘do nothing’ option is salient in this context, as
the patient has presupposed it or even explicitly mentioned it. However, this is not the
only possibility.
The second part of the answer is that the relevant alternative for the patient’s utterance
may be determined by what was salient to a particular doctor when the patient discussed
it with him. The patient may have a referential intention to defer to her own doctor,
or simply ‘to the experts’. This is one way in which the relevant alternative can be a
procedure that she has never heard of.
The third part of the answer is that the relevant alternatives may be determined not by
conversational salience but by natural limitations of the case, just as it was in the first
context we considered (the context of the surgical audit). We begin as a default with the
widest possible range of alternative procedures, and then narrow this down by limitations
of the availability of alternatives. The result is that in the absence of specific referential
intentions of the patient, the relevant range of alternative procedures is all of the
alternatives to the actual procedure that were available in the circumstances. This third
part of the answer seems to me to most likely to give the appropriate interpretation of
her claim.
In this article, we have investigated the contextual factors that affect interpretation of
medical causal claims, by looking at one medical example in several different contexts.
The story we have discovered is a complicated one: we have found a dependence not just
on explicitly mentioned alternatives and explicitly vetoed alternatives, but also on alter-
natives determined by goals, standards, availability and other limitations. We have also
briefly looked at the interaction of different contextual factors. We have found that
conversational salience can be trumped by limiting factors such as professional stand-
ards. Finally, we have used some of our conclusions in order to explain why Causal
Contextualism does not entail that a speaker ignorant of alternative procedures is unable
to make true or meaningful causal claims.27

Cei Maslen, Philosophy Programme, School of History, Philosophy, Political Science and
International Relations, Victoria University ofWellington, PO Box 600, Wellington 6140, New
Zealand. Cei.Maslen@vuw.ac.nz

NOTES

1 For defences see, for example, Stewart Cohen, ‘How to be a fallibilist’, Philosophical Perspectives 2 (1988):
91–123; K. DeRose, ‘Contextualism and knowledge attribution’, Philosophy and Phenomenological Research
52,4 (1992): 913–929; D. Lewis, ‘Elusive knowledge’, Australasian Journal of Philosophy 74,4 (1996):
549–67; Ram Neta, ‘S knows that P’, Noûs 36,4 (2002): 663–81. For attacks see, for example, Patrick
Rysiew, ‘Contesting contextualism’, Grazer Philosophische Studien 69 (2005): 51–70; S. Schiffer, ‘Contextu-
alist solutions to skepticism’, Proceedings of the Aristotelian Society 96 (1996): 317–333; J. Stanley, Knowledge
and Practical Interests, (Oxford: Clarendon Press, 2005); P. Yourgrau, ‘Knowledge and relevant alternatives’,
Synthese 55,2 (1983): 175–91; Timothy Williamson, ‘Contextualism, subject-sensitive invariantism and
knowledge of knowledge’, Philosophical Quarterly 55, 219 (2005): 213–35.
2 C. R. Hitchcock, ‘Farewell to binary causation’, Canadian Journal of Philosophy 26,2 (1996): 267–82;
P. Holland, ‘Statistics and causal inference’, Journal of the American Statistical Association 81 (1986): 945–60;
T. Horgan, ‘Mental quausation’, Philosophical Perspectives 3 (1989): 47–76.
3 P. Menzies, ‘Difference-making in context’ in J. Collins, E. Hall & L. A. Paul (eds) Causation and Counter-
factuals (Cambridge, MA: MIT Press, 2004), pp. 139–80; H. Price, ‘Causal perspectivalism’, in H. Price &
R. Corry (eds) Causation, Physics and the Constitution of Reality: Russell’s Republic Revisited, (Oxford: Oxford

© Society for Applied Philosophy, 2012


Keeping Score for Causal Claims 23

University Press, 2007), pp. 250–92; J. Schaffer, ‘Contrastive causation’, The Philosophical Review 114
(2005a): 297–328; J. Woodward, Making Things Happen, (New York: Oxford University Press, 2003).
4 I want to stress that this is a description of a plausible scenario, not a real case study.
5 See, for example, H. L. A. Hart & Tony Honore, Causation in the Law (Oxford: Oxford University Press,
1959).
6 For example, Lewis (1996 op. cit.) proposes a slightly modified version of this.
7 Counterfactuals themselves are often acknowledged to be context-dependent, so if the standard counter-
factual approach to causation is correct one might expect this context-dependence to carry over to
causation. A famous example of the context-dependence of counterfactuals is from Quine. He observed
that the following counterfactuals both seem to be true yet conflict with each other:
If Caesar were in command in Korea, then he would have used catapults.
If Caesar were in command in Korea, then he would not have used catapults – he would have used
the atom bomb.

However, this feature of counterfactuals has usually been ignored by defenders of the counterfactual
approach to causation. An exception is Lewis, the foremost defender of the counterfactual analysis of
causation, who did observe in his first paper on causation that, ‘The vagueness of similarity does infect
causation and no correct analysis can deny it’ (D. Lewis, ‘Causation’, Journal of Philosophy 70 (1973):
556–67). However, he did not expect this element of context-dependence to ever be important in practice
(1996, personal conversation). Another exception is Carroll (J. Carroll, ‘Making exclusion matter less’,
Manuscript (2004)) who defends a contextualist approach to causation and suggests the context-
dependence of counterfactuals as the source of the context-dependence of causal claims. However, I think
that the context-dependence of causal claims goes beyond any context-dependence in counterfactuals. (As
I argue later in this section, it can be legitimate to contrast with alternative procedures that should have
happened if the cause had been absent, or indeed more generally with a range of options that might have
occurred if the cause had been absent.)
8 Alternatives may also be explicitly stated in the context, for example, with claims such as ‘Taking medica-
tion1 rather than medication2 was a cause of her recovery’.
9 Note that the standard analysis fails to account for preemption and overdetermination, a problem which has
often been considered to be of the first importance. I think that Causal Contextualism can provide a
satisfactory treatment of preemption and overdetermination. However, I do not have space to argue for this
claim here, so such cases are not discussed. For discussions of Causal Contextualism and preemption and
overdetermination, see Woodward, op. cit. and C. R. Hitchcock, ‘Prevention, preemption, and the principle
of sufficient reason’, The Philosophical Review 116,4 (2007): 495–531.
10 Woodward and Hitchcock suggest this as a contrast when discussing another example. (Woodward mentions
a remark from Hitchcock about a case where the contrast with never having existed is relevant (J. Woodward,
‘Sensitive and insensitive causation’, Philosophical Review 115,1 (2006): 19, footnote 13).This is a case where
writing a letter of recommendation leads to taking a job, moving city, the conception of a baby, and
eventually the baby’s death many years later.)
11 A dependence on alternatives to the effect could easily be incorporated into an analysis if a more general
account is needed For example: c, relative to a set of contrast events {c*i}, is a cause of e, relative to a set of
contrasts {e*i}, iff for every c*i in { c*i } there is an e*i in { e*i } such that if c*i had happened then e*i would
have happened.’
12 See C. R. Hitchcock, ‘A tale of two effects’, Philosophical Review 110,3 (2001): 361–96.
13 Note that limitations of technical expertise do not just play a part in limiting relevant available procedures,
but also play an important part in the assessment of risk. (Thanks to John Matthewson for pointing this out
to me.) The surgeon’s personal success rate for this kind of operation will be more indicative of the
probabilities relevant to the causal analysis than nationwide statistics. For example, endoscopic SEMS
placement in inexperienced hands will have a lower success rate than the same procedure performed by an
expert, and so bowel resection, relative to inexperienced endoscopic SEMS placement, will not count as a
cause of death on our analysis. Given this dependence on technical expertise already comes into the causal
analysis by influencing probabilities, it may be unnecessary to also include limitations of technical expertise
on relevant alternative procedures.
14 See Lewis 1996 op cit.
15 D. Lewis, ‘Scorekeeping in a language game’, Journal of Philosophical Logic 8 (1979): 339–59.
16 Lewis 1996 op cit.

© Society for Applied Philosophy, 2012


24 Cei Maslen

17 J. Hawthorne, Knowledge and Lotteries (Oxford: Oxford University Press, 2004).


18 Lewis 1996 op cit.
19 For example, Hawthorne argues that serious worrying is required. As he says, merely watching The Matrix
doesn’t rob one of the knowledge that one is at the movie theatre. ‘Entertaining or attending to a state of
affairs is one thing. Taking seriously the idea that things may be actually that way is quite another . . . ’
(Hawthorne op. cit., p. 64).
20 See Stanley op. cit., p. 25.
21 See K. DeRose, The Case for Contextualism (Oxford: Oxford University Press, 2009), chapter 4).
Note that the standard answer the epistemic contextualist gives to the sceptic depends on it being harder
to raise standards (e.g. rule out alternatives by mentioning them) than to lower standards (e.g. to make
alternatives relevant by mentioning them). I think this is one reason that this second type of rule of attention
is rarely mentioned, despite its plausibility.
22 For example, there may also be a probability threshold rule, similar to the rule suggested for Epistemic
Contextualism A suggestion is as follows: an alternative that is very improbable, or below a threshold that is
salient in the context, is not relevant. For example, if expert 2 were to say (sarcastically) ‘Yeah, right. Like
that was really going to happen’ this seems to be both fixing a threshold for how probable an alternative must
be in order to count and implying that it does not meet this threshold. This seems to involve a Rule of
Probability working together with an explicit veto.
This is similar to the main mechanism that Cohen (op. cit.) postulates in his version of Epistemic
Contextualism. Collins (J. Collins, ‘Preemptive prevention’, in J. Collins, N. Hall & L. A. Paul (eds)
Causation and Counterfactuals (Cambridge, MA: MIT Press, 2004), pp. 107–18) seems to have something
like this in mind, although he does not advertise his account as a version of Causal Contextualism.
23 I realise that the phrase ‘all possible alternatives to the cause’ has not been made precise. It will take some
further work to state rigorously what is involved in an event being an alternative to the cause. In some cases,
‘possible’ may mean physically possible and in other cases it may mean logically possible. Perhaps there are
also cases in which logically impossible alternatives are relevant, either because they have been explicitly
mentioned, or because there are no logically possible alternatives to the cause. I have in mind in particular
claims of mental causation but I will not explore such cases in this article. For now, we can let the phrase ‘all
possible alternatives to the cause’ be captured by the might-counterfactual: ‘all the situations that might have
been present if the cause had been absent.’
24 Maurice Goldsmith expressed this worry to me.
25 Carroll op. cit.
26 There is not always such an obvious alternative as ‘do nothing’. Suppose that someone ignorant of
neuroscience is claiming that the firing of neuron535 was a cause of some action. That person may very well
be unable to make any neuroscientific events salient alternatives to the firing of neuron535, either by
explicitly stating them or by having secret neuroscientific thoughts.
27 Thanks to John Carroll for letting me quote his unpublished work here.Thanks also to Tom Crisp, Josh Gert
and John Matthewson for helpful comments on early drafts of this article and to Yasser Salama for surgical
background (although he should not be held responsible for any errors). Thanks also to two anonymous
referees for this journal for helpful comments.

© Society for Applied Philosophy, 2012

You might also like