Introduction 1
1. Inductive Arguments 4
2. The Nature of Explanation 23
3. The Justification of Explanation 52
4. Complete Explanation 73
5. The Intrinsic Probability of Theism 93
6. The Explanatory Power of Theism: General
Considerations 110
7. The Cosmological Argument 133
8. Teleological Arguments 153
9. Arguments from Consciousness and Morality 192
10. The Argument from Providence 219
11. The Problem of Evil 236
12. Arguments from History and Miracles 273
13. The Argument from Religious Experience 293
14. The Balance of Probability
1. Inductive Arguments
An argument starts from one or more premisses, which are propositions taken for granted for the purpose of the argument, and argues
to a conclusion. An argument is a valid deductive argument if it is
incoherent to suppose that its premisses are true but its conclusion
false. For example, the following argument is a valid deductive
argument:
(Premiss 1) No material bodies travel faster than light.
(Premiss 2) My car is a material body.
(Conclusion) My car does not travel faster than light.
In a valid deductive argument the premisses make the conclusion
certain. There are arguments that are not deductively valid, but in
which the premisses in some sense ‘support’ or ‘confirm’ or ‘give
strength to’ the conclusion, and some or all arguments of this general
kind are often characterized as ‘good’ or ‘correct’ or ‘strong’ inductive arguments. However, we need here to distinguish carefully between two different kinds of argument. There are arguments in which
the premisses make the conclusion probable, that is, more probable
than not—for example:
P1: 70% inhabitants of the Bogside are Catholic.
P2: Doherty is an inhabitant of the Bogside.
C: Doherty is Catholic.
The conjunction of the premisses makes the conclusion probable.
However, many arguments that are called ‘correct’ inductive arguments are hardly to be regarded as of this type. Take the following
argument:
P: All of 100 ravens observed in different parts of the world are black.
C: All ravens are black.
The normal way to construe this conclusion, in the context of a
discussion of inductive arguments, is to suppose that it is about all
ravens at all moments of time and points of space—and, even if you
suppose that nothing on a distant planet would count as a raven, that
means all ravens at all times in the earth’s history and at all places on
its surface. But, when the conclusion is interpreted this way, it
becomes implausible to suppose that P makes C more probable
than not. For it is not improbable to suppose that the blackness of
observed ravens arises from a particular feature of modern ravens, a
particular feature of their make-up not present in older ravens. To
suppose that all ravens are always black seems to go a long way
beyond the evidence recorded in P. C may, however, be true; and,
most of us suppose, P increases the probability that it is true, but P
does not make C probable.
Most of the arguments of scientists from their observational evidence to conclusions about what are the true laws of nature or to
predictions about the results of future experiments or observations
are not deductively valid, but are, it would be generally agreed,
inductive arguments of one of the above two kinds. (I do not mean
that they have the simple pattern of the easy examples given above,
but only that they are arguments that have the defining characteristics
of one of the two kinds.) The various astronomical observations made
by Tycho Brahe, Kepler, Galileo, and other men of the seventeenth
century were observations that favoured Newton’s theory of motion,
in the sense that they made it more likely to be true, more probable,
than it would have been otherwise. The various botanical, geological,
and breeding data described by Charles Darwin in The Origin of
Species added to the probability of his theory of the evolution of
animal species by natural selection of variations. It is an interesting
question, to which I shall need to allude at a later stage, whether, in a
typical scientific argument from various data of observation and
experiment to a conclusion about what are the fundamental laws of
physics or chemistry, the premisses make the conclusion probable or
merely add to its probability. Laws of nature are normally supposed to
be generalizations that not merely hold at all times and places, but
would continue to hold under unrealized or unrealizable circumstances (for example, however humans interfere with the universe).
Newton’s theory of motion consists of his three laws of motion and his
law of gravitational attraction. Did the various observations of the
seventeenth century make it more probable than not that his theory was true? I pass no judgement on this matter at this stage. However,
on our normal way of looking at these matters, clearly observational
evidence often makes more probable than not a particular prediction
about the future. All the observational evidence about the past behaviour of sun, moon, planets, etc. makes it more probable than not that
the earth will continue to spin on its axis for the next twenty-four
hours and so that the sun will rise over the earth again tomorrow.
Let us call an argument in which the premisses make the conclusion probable a correct P-inductive argument. Let us call an argument in which the premisses add to the probability of the conclusion
(that is, make the conclusion more likely or more probable than it
would otherwise be) a correct C-inductive argument. In this latter
case let us say that the premisses ‘confirm’ the conclusion. Among
correct C-inductive arguments, some will obviously be stronger
than others, in the sense that in some the premisses will raise the
probability of the conclusion more than the premisses do in other
arguments.
The point of arguments is to get people, in so far as they are
rational, to accept conclusions. For this purpose it is not sufficient
that their premisses should in some sense necessitate or probabilify
their conclusion. It is also necessary that the premisses should be
known to be true by those who dispute about the conclusion. There
are plenty of valid arguments to the existence of God that are quite
useless, because, although their premisses may be true, they are not
known to be true by those who argue about religion—for example:
P1: If life is meaningful, God exists.
P2: Life is meaningful.
C: God exists.
This argument is certainly valid. If the premisses are true, the
conclusion must be true. The premisses may be true; but atheists
would deny either the first premiss or the second one. Since the
premisses are not common items of knowledge to those who argue
about religion, they do not form a suitable jumping-off ground for
such argument. What are clearly of interest to people in an age of
religious scepticism are arguments to the existence (or non-existence)
of God in which the premisses are known to be true by people of
all theistic or atheistic persuasions. I therefore define arguments
from premisses known to be true by those who dispute about the conclusion which are valid deductive, correct P-inductive, or correct
C-inductive arguments, respectively good deductive, good P-inductive, and good C-inductive arguments. In investigating arguments for
or against the existence of God, we need to investigate whether any of
them is a good deductive, good P-inductive, or good C-inductive
argument.
I take the proposition ‘God exists’ (and the equivalent proposition
‘There is a God’) to be logically equivalent to ‘there exists necessarily
a person1 without a body (i.e. a spirit) who necessarily is eternal,
perfectly free, omnipotent, omniscient, perfectly good, and the creator of all things’. I use ‘God’ as the name of the person picked out by
this description. I understand by God’s being eternal that he always
has existed and always will exist. There is an alternative understanding of ‘eternal’ in the Christian tradition as ‘timeless’ or ‘outside time’.
This understanding did not, however, arrive in the Christian tradition under the fourth century ad; it is very difficult to make any
sense of it, and, for reasons that I have given elsewhere,2 it seems quite
unnecessary for the theist to burden himself with this understanding
of eternity. By God’s being perfectly free I understand that no object
or event or state (including past states of himself) in any way causally
influences him to do the actions that he does—his own choice at the
moment of action alone determines what he does. By God’s being
omnipotent I understand that he is able to do whatever it is logically
possible (i.e. coherent to suppose) that he can do. By God’s being
omniscient I understand that he knows whatever it is logically possible that he know. By God’s being perfectly good I understand that
he always does a morally best action (when there is one), and does no
morally bad action. By his being the creator of all things I understand
that everything that exists at each moment of time (apart from
himself) exists because, at that moment of time, he makes it exist,
or permits it to exist. The meaning of this claim that there is a
God will be developed in somewhat greater detail at points in later
chapters, especially in Chapter 5.3 The claim that there is a God is
called theism. Theism is, of course, the core belief of the creeds of
Christianity, Judaism, and Islam. In the course of human history many people have taken for
granted the existence of God, and many others no doubt have
taken for granted his non-existence. They have not had consciously
formulated reasons for their beliefs. They have just believed. However, others who have believed have had reasons for their beliefs. As
with most people’s reasons for most of their beliefs, these reasons
have often been very vague and inchoate. Sometimes, however,
people have formulated some of their reasons for belief in a sharp
and explicit form. Then we have something clearly recognizable as an
argument for or against the existence of God. Those arguments that
have been frequently discussed have been given names—and thus we
have ‘the cosmological argument’, or ‘the argument from religious
experience’. Other arguments exist that have not been discussed
frequently enough to gain a name. And people have had other
reasons for belief or disbelief that have never been formulated explicitly enough to constitute an argument.
In the course of this book I shall discuss various of the reasons that
people have had for believing in the existence of God, or in the nonexistence of God, some of which have received a sufficiently precise
form already to be codified in named arguments and others of which
will need to be knocked into a clear shape. I shall discuss only
arguments in which the premisses report what are (in some very
general sense) features of human experience—for example, evident
general truths about the world or features of private human experience. Such arguments I shall term a posteriori arguments. They claim
that something that humans experience is grounds for believing that
there is a God or that there is no God. I shall not discuss a priori
arguments—these are arguments in which the premisses are logically
necessary truths—namely, propositions that would be true whether or
not there was a world of physical or spiritual beings. Among logically
necessary truths are the truths of mathematics or logic. Hence I shall
not discuss the traditional ontological argument4 for the existence of
God, or any variants thereof. Nor shall I discuss arguments against the existence of God that claim that there is something incoherent or
self-contradictory in the claim that there is a God. I think that
ontological arguments for the existence of God are very much mere
philosophers’ arguments and do not codify any of the reasons that
ordinary people have for believing that there is a God. The greatest
theistic philosophers of religion have on the whole rejected ontological arguments and relied on a posteriori ones.5 Arguments against
the existence of God that claim that theism is incoherent do, however, I admit, have some basis in the thought of ordinary people.
I shall not, however, of course be able to discuss all the a posteriori
reasons that people have had for believing that there is or that there is
not a God. But I shall consider those that, in my view, are the most
plausible and have had the greatest appeal in human history. In
reaching my final conclusion about how probable it is that there is
a God, I assume that no a priori arguments of either species,6 and no
a posteriori arguments other than those that I discuss, have any
significant force.
Although my theme is arguments for and against the existence of
God, it will seem that I concentrate on arguments for the existence
of God. I do discuss in a separate chapter the main argument against
the existence of God—the argument from evil, which claims that the
existence of pain and suffering in the world shows that there is no
perfectly good and all-powerful being. But, apart from that argument
(and the associated argument from hiddenness, which I also discuss
there), the main reason that atheists have for believing that there is
no God has been their claim that there is insufficient evidence, that
the theist’s arguments do not make the existence of God probable to
any significant degree. The atheist’s arguments, apart from the argument from evil, have been largely in the form of criticisms of the
theist’s arguments. I therefore discuss such arguments in the course
of discussing each of the main arguments for the existence of God. In
discussing arguments for the existence of God, I shall consider forms
of cosmological and teleological argument, the argument from the
existence of consciousness, the moral argument, arguments from
miracle and revelation, and the argument from religious experience.
A cosmological argument argues that the fact that there is a universe needs explaining and that God’s having made it and kept it in being
explains its existence. An argument from design argues that the fact
that there is design in the world needs explaining, and that God’s
action provides that explanation. There are various forms of argument from design, according to the kind of design to which it draws
attention. I discuss two different genera of the argument under the
headings ‘teleological arguments’ and ‘the argument from providence’, and different species of each genus. The argument from the
existence of consciousness argues that the fact that there are conscious beings is mysterious and inexplicable but for the action of
God. Arguments from miracle and revelation cite various public
phenomena in the course of human history as evidence of God’s
existence and activity. The argument from religious experience
claims that various private experiences are experiences of God and
thus show his existence.
Some of the issues that I discuss are ones that I have treated at
greater length elsewhere; but the discussion in this book is, I hope,
adequate—given the constraints imposed by the length of the
book—to support the conclusions drawn here. For example,
I discussed the problem of evil at book length in my book Providence
and the Problem of Evil;
7 but I hope that the discussion of it in
Chapters 10 and 11 of the present book suffice to make it plausible
that the kind and amount of evil that we find on Earth do not count
significantly against the existence of God. Yet there is one respect in
which my discussion in this book is manifestly incomplete. When
I discuss arguments from miracles, I have space only to discuss which
strange public phenomena (for example, a dead man coming to life)
if they occurred would be evidence for the existence of God, but I do
not have space to discuss the historical evidence for and against the
occurrence of particular public phenomena. So in effect I discuss
here only the form of an argument that needs filling out with detailed
historical material.8
Kant produced a threefold classification of arguments for the
existence of God that has had a permanent and to my mind far
from beneficial influence on the subsequent discussion of this
topic. He wrote:
There are only three possible ways of proving the existence of God by means
of speculative reason. All paths leading to this goal begin either from
determinate experience and the specific constitution of the world of sense
as thereby known, and ascend from it, in accordance with the laws of
causality, to the supreme cause outside the world; or they start from
experience which is purely indeterminate, that is from experience of existence in general; or finally they abstract from all experience, and argue
completely a priori, from mere concepts, to the existence of a supreme
cause. The first proof is the physico-theological, the second the cosmological,
the third the ontological. There are, and there can be, no others.9
The distinction is made in terms of the nature of the premiss. Either
you start from a conceptual truth—in which case you have the
ontological argument; or from ‘existence in general’—in which case
you have the cosmological argument; or from the details of what
Kant calls ‘determinate experience’, how things are in the world—in
which case you have the ‘physico-theological’ argument.
My reason for claiming that this doctrine of Kant has had a far
from beneficial influence on discussion of this topic is that by his use
of the word ‘the’ Kant tends to assume that there can be only one
argument of each type—whereas in fact there can quite clearly be
many different arguments under each heading that are so different
from each other that it would be misleading to call them forms of the
same argument at all. There is, for example, no reason to suppose
that all arguments to the existence of God in which the premisses are
in some sense conceptual truths need have the form of the traditional
ontological argument. Above all, there is no reason to suppose that
all arguments from how things are in the world need have the form of
the argument that Kant calls ‘physico-theological’, and has elsewhere
been called the argument from design. This latter argument may
itself have many forms. It may argue, for example, from the regular
behaviour of objects in the world codified in laws of nature, or from
the ready availability in the world of the things that humans and
animals need to survive. In both cases there is an argument from a
very general order in nature. But there are arguments too, as we have
noted, from particular miracles, from the development of human
history, or from particular religious experiences. Not all of these may
be particularly good arguments but they deserve to be considered on
their merits—Kant’s classification obscures their existence.
So then we shall consider the worth of various a posteriori arguments, not merely two, as listed by Kant. When we have our
arguments in clear form, we shall need to ask—are they good
deductive arguments, or good P-inductive arguments, or good Cinductive arguments? Sometimes the proponents of such arguments
have not been clear whether the arguments were intended to be
deductive or inductive, let alone about the kind of inductive arguments that they were intended to be.
One unfortunate feature of recent philosophy of religion has been a
tendency to treat arguments for the existence of God in isolation from
each other. There can, of course, be no objection to considering each
argument initially, for the sake of simplicity of exposition, in isolation
from others. But clearly the arguments may back each other up or
alternatively weaken each other, and we need to consider whether or
not they do. Sometimes, however, philosophers consider the arguments for the existence of God in isolation from each other, reasoning
as follows: the cosmological argument does not prove the conclusion,
the teleological argument does not prove the conclusion, etc., etc.,
therefore the arguments do not prove the conclusion. But this ‘divide
and rule’ technique with the arguments is admissible. Even if the only
kind of good argument was a valid deductive argument from premisses known to be true, it would be inadmissable. An argument from p
to r may be invalid: another argument from q to r may be invalid. But,
if you run the arguments together, you could well get a valid deductive
argument; the argument from p and q to r may be valid. The argument
from ‘all students have long hair’ to ‘Smith has long hair’ is invalid,
and so is the argument from ‘Smith is a student’ to ‘Smith has long
hair’; but the argument from ‘all students have long hair and Smith is
a student’ to ‘Smith has long hair’ is valid.
That arguments may support and weaken each other is even more
evident, when we are dealing with inductive arguments. That Smith
has blood on his hands hardly makes it probable that Smith murdered Mrs Jones, nor (by itself) does the fact that Smith stood to gain
from Mrs Jones’s death, nor (by itself) does the fact that Smith was
near the scene of the murder at the time of its being committed, but
all these phenomena together (perhaps with other phenomena as
well ) may indeed make the conclusion probable.10
In order to consider the cumulative effect of arguments, I shall
consider them one by one, starting with the cosmological argument
and including the arguments from evil and from hiddenness against
the existence of God, and ask how much the premisses of each
argument add to or subtract from the force of the previous arguments. To give advance notice of some of my conclusions, I shall
argue that (neither separately nor in conjunction) are any of the
arguments that I consider for or against the existence of God good
deductive arguments. There are, of course, as I have pointed out,
valid deductive arguments to the existence of God, but they start
from premisses that are far from generally accepted. On the other
hand, I shall argue that most of the arguments (taken separately and
together) for the existence of God are good C-inductive arguments—
that is to say, their premisses make it more probable (or likely) that
God exists than it would otherwise be. Some of these arguments of
course confirm the existence of God much more strongly than do
others. I shall allow that the argument against the existence of God
from evil is a good C-inductive argument of very limited force. I shall
claim that the argument from hiddenness to the non-existence of
God is not a good C-inductive argument. The crucial issue, however,
is whether all the arguments taken together make it probable that
God exists, whether the balance of all the relevant evidence favours
the claim of theism or not. For clearly, in so far as the probability of a
hypothesis is relevant to whether or not we ought to act on it, we
ought to act on a hypothesis in so far as it is rendered probable by the
total evidence available to us—all we know about the world, not just
some limited piece of knowledge. The religious person claims that his religious viewpoint makes sense of the whole of his experience; and
his atheistic rival is liable to make a similar claim. In the final chapter
I shall reach a conclusion on whether or not the balance of all the
relevant evidence favours theism. I shall be fairly brief in dismissing
the suggestions that any of the arguments separately or all the
arguments taken together constitute a good deductive argument.
I shall be fairly brief because many other philosophers have devoted
their technical skills to this task, and relatively few philosophers
today would accept that there are good deductive arguments to be
had here. I shall devote most of my time to assessing the inductive
strength of such arguments. I shall consider of each argument
whether it is a good C-inductive argument, but only when we have
all the arguments shall I ask whether, taken together, they make a
good P-inductive argument. I proceed in this way because, as will
appear, it is a lot easier to see when we have a good C-inductive
argument than when we have a good P-inductive argument.
It will be useful to introduce at this stage the symbols of confirmation theory that I shall use from time to time in subsequent chapters. I represent by lower-case letters such as e, h, p, and q
propositions. P(pjq) represents the probability of p given q. Thus p
might represent the proposition: ‘The next toss of this coin will land
heads’, and q might represent the proposition: ‘505 of the last 1,000
tosses of this coin have landed heads’. Then P(pjq) represents the
probability that the next toss of the coin will land heads, given that
505 of the last 1,000 tosses have landed heads. (The value of P(pjq)
would then generally be supposed to be 0.505.) However, the relation
between p and q may be of a much more complex kind; and clearly
we normally assess the probability of claims on evidence other than
or additional to that of relative frequencies. p may be some scientific
hypothesis—say, Einstein’s General Theory of Relativity—and q may
be the conjunction of all the reports of the evidence of observation
and experiment that scientists have collected relevant to the theory.
Then P(pjq) represents the inductive probability of Einstein’s General Theory given all the reports of relevant observations and experiments. Inductive probability is thus to be distinguished from
statistical probability, which is a property of classes of things (for
example, inhabitants of a certain town, say Tunbridge Wells) and is
a measure of the proportion of things in the class that have
some other property (for example, voting Conservative in the 2001
Election). The probability of an inhabitant of Tunbridge Wells voting
Conservative in 2001 is just the proportion of inhabitants of Tunbridge Wells who voted Conservative in 2001. (In English, the
indefinite article—for example, ‘the probability of an inhabitant . . . ’—often indicates that the probability is statistical.) The
classes may be of actual things (for example, inhabitants of Tunbridge
Wells), or of hypothetical things, things that would be generated by a
certain process (for example, tosses of this coin, if we were to toss it for
a very long time).
Inductive probability is also to be distinguished from physical
probability. The physical (or natural ) probability of an event (and
so of the proposition that records it) is a matter of the extent to
which at some earlier time the event is predetermined by its causes.
An event that is made inevitable by the preceding state of the world
has a physical probability of 1—its occurrence is physically necessary;
an event whose non-occurrence is made inevitable by the preceding
state of the world has a physical probability of 0—its occurrence is
physically impossible. An event has a physical probability between 1
and 0 if it is not predetermined that it will happen or that it will not
happen, but the preceding state of the world is biased in favour of its
happening to the degree measured by the value of the probability:
larger values of the probability indicate a greater bias in favour of its
happening.11 Physical and statistical probabilities may themselves
constitute evidence that makes some hypothesis inductively probable; or other evidence may make it inductively probable that they
have a certain value.
My concern with inductive probability is a concern with how
probable q makes p, quite apart from who is doing the calculation,
how clever he is, and his degree of confidence in the evidential force
of q. Clearly in science and history and all other empirical inquiries
we think that there are correct ways to assess whether and (within
rough limits) how much some evidence supports some hypothesis.
I shall set out these criteria in Chapter 3. In order to emphasize the
objective character of the value P(pjq) with which I am concerned
and to distinguish it from measures of evidential support that measure subjects’ degree of confidence or are in part functions of
subjects’ abilities to work out the true measure of evidential
support,12 I shall in future call P(pjq), the logical probability of
p on q. This is clearly an a priori matter. If q represents all the relevant
evidence, the value of P(pjq) cannot depend on further evidence—it
measures what the evidence you have already got shows. It is an a
posteriori matter whether, in 1,000 tosses, 505 have landed heads;
but an a priori matter whether that evidence gives a probability of
0.505 to the next toss landing heads.
A hypothesis up for investigation is often represented by h. Then
P(hje &k) represents the probability of a hypothesis h given evidence
(e &k).13 It is often useful to divide the evidence available to an
observer into two parts—new evidence and background evidence;
if this is done, the former is often represented by e and the latter by k.
Background evidence (or background knowledge, as it is sometimes
called) is the knowledge that we take for granted before new evidence
turns up. Thus, suppose that detectives are investigating a murder. h
could represent the hypothesis that Jones did the murder; e could
represent the proposition that reports all the new evidence that
detectives discover—for example, that Jones’s fingerprints were
found on the weapon, that he was near the scene of the murder at
the time it was committed, etc., etc. k could represent the proposition
reporting the detectives’ general knowledge about how the world
works—for example, that each person has a unique set of fingerprints, that people who touch metal and wood with bare hands
usually leave their fingerprints on them, etc., etc. Then P(hje &k)
represents the probability that Jones did the murder, given detectives’
total evidence.
For all propositions p and q P(pjq) ¼ 1 if (and only if) q makes p
certain—for example, if q entails p (that is, there is a deductively
valid argument from q to p); and P(pjq) ¼ 0 if (and only if) q makes
p certain—for example, if q entails p.
14 P(pjq) þ P( pjq) ¼ 1.
So if P(pjq) > 1=2, then P(pjq) > P( pjq) and it is on q more probable that p than that p. So (for background knowledge k) an
argument from e to h will be a correct C-inductive argument if (and
only if) P(hje &k) > P(hjk), and a correct P-inductive argument if
(and only if) P(hje &k) > 1=2. The division between new evidence
and background evidence can be made where you like—often it is
convenient to include all evidence derived from experience in
e and to regard k as being what is called in confirmation theory
mere ‘tautological evidence’, that is, in effect all our other irrelevant
knowledge.
My strategy will be as follows. Let h be our hypothesis—‘God exists’.
Let e1, e2, e3, and so on be the various propositions that people bring
forward as evidence for or against his existence, the conjunction of
which form e. Let e1 be ‘there is a physical universe’. Then we have
the argument from e1 to h—a cosmological argument. In considering
this argument I shall assume that we have no other relevant evidence,
and so k will be mere tautological evidence. Then P(hje1 &k) represents the probability that God exists given that there is a physical
universe—and also given mere tautological evidence, which latter can
be ignored. If P(hje1 &k) > 1=2, then the argument from e1 to h is a
good P-inductive argument. If P(hje1 &k) > P(hjk), then the argument is a good C-inductive argument. But, when considering the
second argument, from e2 (which will be the conformity of the
universe to temporal order), I shall use k to represent the premiss of
the first argument e1; and so P(hje2 &k) will represent the probability
that God exists, given that there is a physical universe and that it is
subject to temporal order. And, when considering the third argument,
from e3, k will represent the premiss of the second argument (e1 &e2).
And so on. In this way all relevant evidence will eventually be fed into
our assessment. I shall consider some eleven arguments. I shall claim
that for most of these en, where n ¼ 1, ... 11, P(hjen &k) > P(hjk)—
that is the argument is a good C-inductive argument for the existence
of God, that two of the arguments (one for and one against) have no
force (P(hjen &k) ¼ P(hjk) in these cases) and that one argument
against has force (P(hjen &k) < P(hjk)) where en is the occurrence of
evil. The crucial issue to which we will eventually come is whether
P(hje11 &k) > 1=2.
In using the symbols of confirmation theory I do not assume that
an expression of the form P(pjq) always has an exact numerical value.
It may merely have relations of greater or less value to other
probabilities, including ones with a numerical value, without itself having a numerical value—P(hje1 &k), for example, might be greater
than P(hje2 &k) and less than P(hjk) and less than 1/2 without there
being some number to which it is equal. Clearly, for example, we may
judge one scientific theory to be more probable than another on the
same evidence while denying that its probability has an exact numerical value; or we may judge a prediction to be more probable than
not and so to have a probability of greater than 1/2, while again
denying that that probability has an exact numerical value.
Now it is sometimes said that the different arguments for the
existence of God show different things. The cosmological argument
shows at most the existence of some sort of necessary being; the
argument from design shows at most some sort of arch-architect;15
the argument from miracles shows at most some sort of poltergeist—
so what have they in common? This objection gets things back to
front. There is no one thing that premisses show. In a deductive
argument there are many different conclusions that can be drawn
from a set of premisses. And in inductive arguments the premisses
support different conclusions with different degrees of force.
What does ‘there is a print in the shape of a human foot on the
sand’ show? It shows with different degrees of force many things—
that sand is shapeable, that some creature has been on the sand, that a
man has walked on the sand. The evidence makes probable the
different propositions to different degrees. Our concern is with
the effect of various pieces of evidence on the proposition in which
we are interested—‘God exists’. Does each confirm it (that is, increase
its probability)? Does it make it probable? Our concern is for various
pieces of evidence en (including any k) and for h ¼ ‘God exists’ with
the value of P(hjen). This may well be for some en less than the value
for some other interesting proposition h1, say, ‘there exists an impersonal cause of the universe’, of P(h1jen). That is, en may make h1 more
probable than it makes h. However, even though, say, P(h1je1) >
P(hje1), it certainly does not follow that P(h1je1 ...e7) >
P(hje1 ...e7). That is, ‘God exists’ may gain only a small amount of
probability from e1, a small amount from e2, a small amount from e3,
and so on. For each of e1, e2, e3, there may be some other proposition h1, h2, h3, which is in some sense a rival to ‘God exists’ for which P(hnjen) > P(hjen); but, nevertheless, on the total evidence
h may be more probable than each of the rivals.
A similar situation normally arises with any far-reaching scientific
or historical theory. Each separate piece of evidence does not make
the theory very probable, and indeed taken on its own makes some
narrower theory much more probable. But the cumulative force of
the evidence taken together gives great probability to the wide theory.
Thus each of the various pieces of evidence that are cited as evidence
in favour of the General Theory of Relativity do not by themselves
make it very probable, but together they do give it quite a degree of
probability. Each by itself (given the general background knowledge
available in the early twentieth century) was evidence in favour of
some rival but far less wide-ranging hypothesis than General Relativity. Thus the movement of Mercury’s perihelion taken by itself
would suggest only that there was a hitherto unknown planet lying
between Mercury and the sun or that the sun was of an odd shape,
rather than that General Relativity was true. Taken by itself it would
not have given much probability to General Relativity; but taken with
other pieces of evidence it did its bit in supporting the latter. It is
along these lines that the theist may wish to answer the accusation
that an argument such as the cosmological argument does not show
the existence of the God of Abraham, Isaac, and Jacob. Not by itself,
he may reply, but it does its small bit together with some very diverse
arguments that do their small bit, to get to this conclusion.
Note that it is no objection to a P-inductive or C-inductive
argument from e to h that some contrary hypothesis h is also
compatible with e, as some writers on the philosophy of religion
seem to think. They seem to think that if, for example, the order in
the universe is compatible with ‘God does not exist’, then there is no
good argument from it to ‘God exists’. But one has only to think
about the matter to realize that this is not so. In any non-deductive
argument from e to h, not-h will be compatible with e; and yet some
non-deductive arguments are good arguments.
Note also a further interesting feature of good C-inductive arguments. In such an argument from e to h, P(hje &k) > P(hjk). It may
be the case that also for some contrary hypothesis h there is a good
C-inductive argument from e—that is, also P(hje &k) > P(hjk).
The fact that certain evidence confirms a hypothesis does not mean
that it does not also confirm a rival hypothesis. Once again, this
should be immediately clear if one thinks about it. Suppose that a detective has background information k, that either Smith, Brown, or
Robinson did the crime, and that only one of them did. Then
evidence (e) turns up that Robinson was somewhere else at the
time the crime was committed. e adds to the probability that
Brown did the crime, and it also adds to the probability that Smith
did the crime. Despite this, one sometimes reads writers on the
philosophy of religion dismissing some consideration that is adduced
as evidence for the existence of God on the grounds that it supports a
rival hypothesis equally well.
So then our task will be to assess the worth of different arguments
to the conclusion ‘God exists’. How are we to do this? In the case of
deductive arguments, philosophers have a moderately clear idea of
what makes a valid argument, and so are in a position to look at
various arguments and see if they are valid. But our main concern
will be with inductive arguments. How are we to set about assessing
the probability of ‘God exists’ on different pieces of evidence? To do
this we need to know for what fillings of p and q P(pjq) becomes high
or low. There is, however, fortunately no need to undertake any very
general examination of this question. This is because all important a
posteriori arguments for the existence of God have a common characteristic. They all purport to be arguments to a (causal ) explanation
of the phenomena described in the premisses in terms of the action
of an agent who intentionally brought about those phenomena.
A cosmological argument argues from the existence of the world to
a person, God, who intentionally brought it about. An argument
from design argues from the design of the world to a person,
God, who intentionally made it thus. All the other arguments are
arguments from particular features of the world to a God who
intentionally made the world with those features.
Not all inductive arguments are arguments to an explanation.
When we argue from the sun having risen at intervals of approximately twenty-four hours over the last many thousand years to the
claim that it will rise tomorrow, we are not arguing to an explanation.
Its rising tomorrow does not explain its previous rising. Yet when the
geologist argues from various deformations to the occurrence of an
earthquake millions of years ago, he is arguing to an explanation; he
is arguing from phenomena to an event that brought those phenomena about. However, not all arguments to an explanation are arguments to the intentional action of an agent. An intentional action is
an action that some agent does, meaning to do it. It is one, therefore that the agent has some reason or purpose for doing—either the
minimal purpose of doing it for its own sake or some further purpose
that is forwarded by doing the action. Since he acts for reasons or
purposes on which he chooses to act, we may term such an agent a
rational agent. Persons are rational agents;16 but they are not the only
ones—animals too often perform intentional actions. By contrast,
however, inanimate objects and events do not have purposes on
which they choose to act and which they seek to fulfil, but rather
they bring about their effects unthinkingly. The geologist’s argument
from deformations to the occurrence of an earthquake is an argument to an explanation of the deformations, but not an argument to
an explanation in terms of the intentional action of a rational agent.
However, when a detective argues from various bloodstains on the
woodwork, fingerprints on the metal, Smith’s corpse on the floor,
money missing from the safe, Jones’s having much extra money, to
Jones’s having intentionally killed Smith and stolen his money, he is
arguing to an explanation of the various phenomena in terms of the
intentional action of a rational agent. Since persons are paradigm
cases of rational agents, I will term explanation in terms of the
intentional action of a rational agent personal explanation. In
Chapter 2 I shall analyse the nature of personal explanation more
fully and I shall contrast it with the other accepted pattern of
explaining mundane phenomena, which I shall call scientific explanation.17 In Chapter 3 I shall go on to consider when it is right to
invoke personal explanation and when it is right to invoke scientific
explanation. A crucial issue that arises there is when is it reasonable to
suppose that phenomena do have a (causal ) explanation, and when,
by contrast, is it reasonable to suppose that phenomena are just brute
facts, things that explain other things, but do not themselves have an
explanation. This issue of what is the proper terminus for explanation will be discussed in Chapter 4. It is one that is crucial for theism. For
the theist claims that the various phenomena that constitute his
evidence—for example, the existence of the world and its conformity
to order—need explanation; and that this is provided by the action of
God, whose existence and action need no explanation. So Chapter 2
will bring out the nature of the theist’s explanations, and Chapters 3
and 4 will provide essential tools for answering the question of when
it is right to invoke them. With these tools we shall then be in a
position to look in detail at the theist’s arguments.
2. The Nature of Explanation
General considerations
When the theist argues from phenomena such as the existence of the
universe or some feature of the universe to the existence of God, he is
arguing, we have seen, to a causal explanation of the phenomena in
terms of the intentional action of a person. Explanation in terms of
the intentional action of a person is the normal case of what I termed
personal explanation. We give a personal explanation of my being in
London by my having gone there in order to give a lecture; or of the
letter’s being on the table by my wife’s having put it there in order to
remind me to post it. However, as we have seen, not all explanations
are personal explanations. Other explanations of the occurrence of
phenomena seem to have a distinct common structure and these I will
call scientific explanations. This chapter will be concerned with
analysing the structure of explanations of the two kinds; and the
next chapter will consider when each is to be invoked.
When someone is said to have provided an explanation of the
occurrence of some phenomenon (that is, an event or state of affairs),
this is ambiguous. What is meant may be that he has provided a true
explanation of the phenomenon, or it may be merely that he has
suggested a possible explanation of the phenomenon. Our interest in
explanations is interest in true explanations. What is it to provide a
true explanation of the occurrence of a phenomenon E? It is to state
truly what (object or event) brought E about (or caused E ), and why
it was efficacious. To explain the occurrence of the high tide is to state
what brought about the tide—the moon, water, and the rest of the
earth being in such-and-such locations at such-and-such times, and
why the moon etc. had that effect—because of the inverse square law
of attraction acting between all bodies. We can thus detect two
components of an explanation of a phenomenon E—the ‘what’
that made E happen and the ‘why’ that made E happen. The
‘what’ will be what I may term some other independent actual
factors—other events, processes, states, objects, and their properties
at certain times. By these factors’ being independent I mean that the
‘what’ is not the same event or process as E nor part of it; nor is it an
object that is a participant in E at the time of E, nor is it a state or
property of E or the objects that participate in E at the time of E’s
occurrence. Only something different from E can make E happen. By
the factors’ being actual I mean only that any events, processes, and
states cited occurred; that any object cited existed and had the
properties cited.
To say that certain factors A...D brought about E entails at least
that each, in the conditions of its occurrence, made it more physically
probable that E would occur; it influenced E’s occurrence. Normally,
perhaps, each of the factors is necessary, given the others and the
world being in other ways the same, for the occurrence of E—that is,
without any one of them, the world otherwise remaining the same, E
would not have occurred. Normally, perhaps, too the set of factors
together is sufficient for the occurrence of E—that is, given their
occurrence, E must necessarily occur. We may call all the factors
together that make up the ‘what’ the cause of E. Alternatively, or
more usually, we distinguish one as the ‘cause’ of E (the effect), and
call the others the conditions that were necessary for the cause to
have its effect (or at least made it physically probable that it would
have the effect); which we call the cause is sometimes a somewhat
arbitrary matter. Normally it will be the most unexpected member of
the set of factors, or the one, the occurrence of which involves the
sharpest change from the previous state of the world. Thus, suppose
someone lights a match close to petrol at a certain temperature and a
certain pressure, and all of this produces an explosion. We may
describe the ignition of the match and the petrol’s being at that
temperature and pressure as jointly the cause of the explosion. But
it would be more natural to describe the ignition of the match as the
cause of the explosion, and the petrol’s being at that temperature and
pressure conditions necessary for the cause to have its effect. My
terminology will be as follows. I shall call a set of factors that together
were sufficient for the occurrence of an event E a full cause of E. Any
member of a set of factors that contributes towards bringing E about
I shall call a cause of E.
24 The Nature of Explanation
To set out the ‘why’ of an explanation is to say why the cause,
under the specified conditions, had the effect that it had. Thus it
might be to cite a law of nature that all events of a certain kind
exemplified by the cause bring about events of a certain other
kind exemplified by the effect. To cite the ‘why’ is to cite what
I shall call the reason why the cause under the conditions of its
occurrence had the effect that it had. I am thus using the word
‘reason’ in a wide but natural sense—in a wider sense than the
sense in which a reason for something is always someone’s reason
for bringing it about. In saying that something was the reason for
some effect I do not necessarily imply that it was someone’s reason
for bringing about the effect.
Now, if there is a full cause C of E and a reason R that guarantees
C’s efficacy, there will be what I shall call a full explanation of E. For,
given R and C, there will be nothing remaining unexplained about
the occurrence of E. In this case, the ‘what’ and ‘why’ together will
deductively entail the occurrence of E. But, if there is no full cause of
E (for example, there occur factors that facilitate the occurrence of E,
but do not necessitate it) or no reason that ensured that the cause
would have the effect that it did, there will be at most what I shall call
a partial explanation of E. Any explanation involving factors or
reasons that did not make the occurrence of E physically necessary
but made it physically more probable than it would otherwise have
been, I will term a partial explanation. E may be given a partial
explanation because there is no full explanation of E. Alternatively,
it may well be the case that, even if a full explanation exists, people
are in no position to provide it, yet they can give some explanation—
they can state some of the causes that make up the ‘what’ and some of
the reasons for their efficacy. In that case they are providing an
explanation, but only a partial one.
Also, of course, people may take for granted or not be interested in
certain aspects of a full explanation and for that reason give only a
partial explanation. A geologist interested in the history of geological
formations may explain a present formation by telling the historical
story of successive stages in its evolution. In telling this story, he may
not bother to cite the physico-chemical laws that are responsible for
one stage succeeding another, simply because he is not interested in
these. For that reason his explanation is only partial. The context
often determines which answers to our questions about ‘the explanation’ of some phenomena will satisfy us. But, while in other contexts
The Nature of Explanation 25
of discussion we may not need to give full explanations even if they
are available, in the contexts of scientific and metaphysical discussion
it is often of crucial importance to know whether there is a full
explanation of some phenomenon and what its character is.
Scientific Explanation
Explanations are of different patterns according to the different kinds
of cause and reason that feature in them. Explanation of the kind
used in science I shall call scientific explanation. The classical account
of the nature of scientific explanation is that set out carefully by C. G.
Hempel and P. Oppenheim, and subsequently championed by
Hempel.1 On the Hempelian account the causes are a group of events
(states of affairs or changes thereof) C, known as the ‘initial conditions’, one of which we may arbitrarily select as ‘the’ cause. The ‘why’
is a set of natural laws L. In the normal case these will be universal
generalizations, having the form ‘all A’s are so-and-so’ or ‘all A’s do
so-and-so’—for example, ‘all copper put in nitric acid dissolves
under such-and-such conditions of temperature and pressure’.
C and L then fully explain E if E follows deductively from them.
We explain a particular explosion by the ignition of a particular
volume of gunpowder in certain conditions of temperature, pressure,
and humidity, and the generalization that under such circumstances
ignited gunpowder explodes. We explain a particular piece of litmus
paper’s turning red by its having been immersed in acid and the
generalization that litmus paper being immersed in acid always turns
red. Sophisticated scientific explanations invoke many laws or generalizations and a complex description of previous events, of which it
is a somewhat remote deductive consequence that the event or state
to be explained occurs. It is a consequence of Newton’s laws and
arrangements of the sun and planets thousands of years ago that they
are in the positions in which they are today, and the former explain
their being in those positions.
This normal pattern of scientific explanation is called by Hempel
deductive–nomological explanation, or D–N explanation—
‘deductive’ because E is deduced from L and C, and ‘nomological’,
from the Greek nomos, ‘law’, because laws are involved in the explanation. A D–N explanation of an event is a full explanation. However,
sometimes the law involved may be a probabilistic law—that is,
claim that ‘n per cent A’s are B’, where n is intermediate between
100 and 0. It may be a law of genetics that ‘90 per cent offspring of
such-and-such a mating have blue eyes’ (or ‘there is an 0.9 probability
of an offspring of such-and-such a mating having blue eyes’.
The probability in this case is a statistical probability.) In such cases,
according to Hempel, a law L together with initial conditions C will
explain E if L and C make it highly probable that E. (The high
probability is in this case an inductive probability, a measure of how
much evidence supports some hypothesis, in this case that E occurs.)
Thus, if an individual a is an offspring of the stated mating, this
together with the law suggested above makes it probable that a
has blue eyes; then Hempel holds, the law and the initial conditions
together explain a having blue eyes. However, the notion of
the inductive probability being ‘high’ is very vague; and plausibly
the law and initial conditions may provide some sort of explanation of
an event even if the probability is not very high—so long as the
law and initial conditions make the occurrence of the event more
probable than it would otherwise be. So, following others,2 I shall
amend Hempel’s account of statistical explanation as follows: a law L
and initial conditions C explain an event E if they increase the
probability of the occurrence of E. Clearly explanation that involves
probabilistic laws is only partial explanation. There is still something
unexplained in why the initial conditions were on this occasion
efficacious.
Science does not explain only particular events, but it may also
explain laws. If it is a consequence of L1 that, perhaps under particular conditions C, L2 operates, then L1 (together with C) explains
the operation of L2. (If the consequence is deductive, the explanation
is a full one; if L1 only makes the operation of L2 probable, the
explanation is only partial.) More fundamental laws explain the operation of less fundamental laws. Given a certain assumption
about the constitution of gases, Newton’s laws of motion explain
the operation of the Van der Waals gas law. One set of laws is often
said to explain another also when a slightly looser relation holds. L1
(perhaps together with some C) may entail and render it probable
that phenomena will be as predicted by L2—to a high degree of
approximation. It then follows that the true laws of nature in the
realm of L2 are very slightly different from L2, but that L2 is a very
close approximation to them. Newton’s laws of motion have the
consequence that, given the distribution of sun and planets through
space, Kepler’s laws of planetary motion will hold to a high degree of
approximation. I shall follow common usage and say that in such
circumstances L2 operates to a high degree of approximation, and
that L1 explains the operation of L2.
It is Hempel’s claim that explanation that does not at first
sight seem to fit into this scientific pattern can really quite easily
be so fitted. Thus we use this scientific pattern of explanation
not only when doing science of any degree of sophistication but in
much everyday explanation of happenings. We explain the cheese’s
being mouldy by its having been left in a warm place for two weeks
and by the generalization that almost always cheese turns
mouldy within two weeks if it is in a warm place. Our explanation
may often take the form of explaining some phenomenon as brought
about, not by an event, but by an object. We may say that the
breaking of the window was brought about by a brick, but what
we are saying here, it is urged, is that the breaking was caused by
some event involving the brick—for example, its fast motion; and
this reduction to the scientific pattern seems initially plausible
enough.
This account, however, needs amplification in order to distinguish
between merely accidentally true universal or probabilistic generalizations and true laws of nature that intuitively involve some sort of
physical necessity or probability. A universal generalization ‘all ravens
are black’ and ‘this is a raven’ would not explain ‘this is black’ unless
the generalization were a claim that there is some sort of causal
connection between being a raven and being black—namely, that
ravens must be black—of physical necessity. Similarly, we need to
add that a statistical generalization ‘n per cent of A’s are B’ (a
statistical probability of n=100 of an A being B ) does not explain
a particular A’s being B unless it asserts some sort of causal connec28 The Nature of Explanation
tion between being A and being B. This will be so if it is claiming that
each A has a n=100 physical probability of being A. By the physical
probability of an event, it will be recalled, I mean a certain bias or
tendency in nature. If nature is deterministic, the only physical
probabilities in nature are probabilities of 1 (physical necessity) or
0 (physical impossibility). But, if there is a certain amount of indeterminism in nature, then there are physical probabilities between 1
and 0. When probabilistic generalizations are concerned with these,
then we may call them probabilistic laws—most interpreters of
Quantum Theory, for example, claim that the basic formulae
of Quantum Theory are fundamental laws of this kind. In the latter
case ‘n per cent of A’s are B’ together with ‘this is an A’ would
(partially) explain ‘this is a B’ if its being an A made it physically
probable to degree n=100 that that thing would be a B. Only so would
there be some sort of causal connection between being A and being B,
which we need if ‘n per cent of A’s are B’ is to explain an A’s being
B. By contrast, John’s voting Conservative is not to be explained by
the fact that his name appears on page 591 of the telephone directory
and 70 per cent of those on that page vote Conservative. For the latter
generalization just states how things happen to be; it is not to be
understood as stating that being on that page pushes people in the
direction of voting Conservative.
I shall in future call Hempel’s account amended in respect of
explanation by probabilistic laws in the way described and amplified
somehow or other so as to make a distinction between true generalizations and laws that involve physical necessity or probability the
amended Hempelian account. But what this model amounts to
depends on how we spell out the notion of a law of nature and so
of the physical necessity or probability involved in a law. One view,
originating from Hume, is the regularity view. On this view, ‘laws of
nature’ are simply the ways things behave—have behaved, are behaving, and will behave. ‘All copper put in nitric acid dissolves
under such-and-such conditions of temperature and pressure’ is a
true universal law of nature if and only if all bits of copper when put
in nitric acid under those conditions always have dissolved, now
dissolve, and will dissolve. ‘50 per cent of atoms of C14 decay
within 5,600 years’ is a true statistical law if and only if, taking
the whole history of the universe, half the atoms of C14 have
decayed within 5,600 years. We do, however, need the distinction
between laws of nature, and accidental generalizations that are true
The Nature of Explanation 29
merely by accident.3 ‘All spheres of gold are less than one mile
in diameter’ may be a true universal generalization, but it holds
only in virtue of the accident of no civilization anywhere in the
universe having put or going to put enough effort into constructing
such a sphere. Regularity theory has reached a developed form
that tries to take account of this distinction, in the work of David
Lewis.
For Lewis, ‘regularities earn their lawhood not by themselves, but
by the joint efforts of a system in which they figure either as axioms
or as theorems’.4 The best system is the system of regularities, which
has (relative to rivals) the best combination of strength and simplicity. Strength is a matter of how much it successfully predicts (that
is, whether it makes many actual events, past, present or future—
whether observed or not—probable; and very few actual events
improbable); simplicity is a matter of the regularities fitting together,
and no doubt, each having internal simplicity in a way that Lewis
does not, but no doubt could, spell out. The true laws are
the regularities of the best system. Accidental generalizations are the
regularities that do not fit into such a system. They float loosely
without being derivable from more fundamental regularities. So ‘all
spheres of gold are less than one mile in diameter’, even if true, is
probably not a law, because it does not follow from the best system—as is evidenced by the fact that it certainly does not follow
from our current best approximation to the ultimate best system—a
conjunction of Relativity Theory and Quantum Theory. Similarly
with probabilistic laws—if and only if ‘90 per cent of A’s are B’ were
a consequence of the best system of regularities would it be a law of
nature. If (and only if) it follows from such a best system that a
particular A will be followed by a particular B (and certain other
complicated conditions hold), then that A causes that B. Lewis’s
account of laws of nature is part of his campaign on behalf of
‘Humean supervenience’, that everything there is supervenes (logically) on ‘a vast mosaic of local matters of particular fact’, which he
interprets as a spatio-temporal arrangement of intrinsic properties, or ‘qualities’.5 Laws of nature and causation are for Lewis among the
things thus supervenient.
There seem, however, to be overwhelming objections to any
Humean account, including Lewis’s, if laws of nature are supposed
to explain anything—and in particular to explain whether and why
one thing causes another, as Humeans suppose that they do. For,
since whether some regularity constitutes a law depends, on this
account, nor merely on what has happened but on what will happen
in the whole future history of the universe, it follows that whether
A causes B now depends on that future history. Yet, how can what is
yet to happen (in maybe two billion years’ time) make it the case that
A now causes B, and thus explain why B happens? Whether A causes
B is surely a matter of what happens now, and whether the world
ends in two billion years’ time cannot make any difference to whether
A now causes B? What is yet to happen can make no difference
to what is the true explanation of why B occurs (namely, that A
occurred and caused B )—though, of course, it might make a difference to what we justifiably believe to be the true explanation. (Put
another way, that some proposed explanation is the simplest explanation of the data, past and future is evidence that it is the true
explanation; but it does not constitute it being the true explanation.)
Further, it is because of their role in causation that laws of nature are
said to generate counterfactuals. Suppose that ‘all copper expands
when heated’ is a law of nature, but that I do not heat a certain piece
of copper; it is all the same fairly evidently the case that, ‘if that
copper had been heated, it would have expanded’. But, if a law simply
states what does (or did or will ) happen, how can it provide any
ground for asserting the counterfactual? It would do that only if there
were some deeper kind of necessity built into it than that provided by
fitting into a system. Fitting into a system could be evidence only of
that deeper kind of necessity.
So, dismissing Humean accounts of laws for good reason, let us
consider alternative accounts of the physical necessity (and physical
probability) involved in laws of nature that do not analyse it away in
terms of patterns of actual events. Physical necessity may be thought
of either as separate from the objects that are governed by it, or as a
constitutive aspect of those objects. The former approach leads to a picture of the world as consisting of events (constituted by substances
having, gaining or losing properties), on the one hand, and laws of
nature (involving physical necessity or probability), on the other
hand; and it can be developed so as to allow for the possibility of
there being universes in which there are no events, but merely laws of
nature.6 Laws of nature are thus ontologically concrete entities.
The version of this account that has been much discussed recently
is the version that claims that laws of nature are relations between
universals.7 (Universals are properties that can be fully instantiated in
many different objects. Thus ‘brown’ is a universal, because innumerable different things can be brown.) It being a fundamental law of
nature that ‘all photons travel at 300,000 km/sec. relative to every
inertial reference frame’ consists in there being such a connection
between the universal ‘being a photon’ and the universal ‘travelling at
300,000 km/sec. relative to every inertial reference frame’. These
universals are tied together, but the tie is not a logically necessary
one—that is, it is not on this view part of what it is to be a photon
that it travels at that speed. But it is physically necessary, and the
physical necessity is a matter of the two universals being tied together. One can perhaps begin to make sense of this suggestion if one
thinks of the causing of states of affairs (for example, the bringing
into existence of a photon) as making properties, which are universals, to be instantiated; and this involving the bringing of them down
to Earth from an eternal Heaven, together with whatever is involved
with those universals—namely, other universals (for example, travelling at 300,000 km/sec.) connected thereto. But why should we
believe that there is such a Platonist heaven in which universals are
tied together? And how can universals act on the world? This is a very mysterious causal relation between the non-spatio-temporal world
and our world for which we have no analogue.
The alternative to thinking of the physical necessity involved in
laws of nature as separate from the objects governed by it is to think
of it as a constitutive aspect of those objects. The way in which this is
normally developed is what we may call the substances-powers-andliabilities (S–P–L) account of laws of nature. The ‘objects’ (the
‘what’) that cause are individual substances—this planet, those molecules of water. They cause effects in virtue of their powers to do so
and their liabilities (deterministic or probabilistic) to exercise those
powers under certain conditions, often when caused to do so by
other substances. Powers and liabilities (the ‘why’) are thus among
the properties of substances. Laws of nature are then just regularities—not of mere spatio-temporal succession (as with Hume), but
regularities in the causal powers (manifested and unmanifested) of
substances of various kinds. That heated copper expands is a law is
just a matter of every piece of copper having the causal power to
expand, and the liability to exercise that power when heated. As a
matter of contingent fact, substances fall into kinds, such that all
objects of the same kind have the same powers and liabilities. The
powers and liabilities of large-scale things (lumps of copper) derive
from the powers and liabilities of the small-scale things that compose
them (atoms; and ultimately quarks, electrons, etc.). And, given a
satisfactory theory integrating all science, all ultimate particulars will
have exactly the same powers and liabilities (for example, the power
to cause an effect proportional in a certain way to their mass, charge,
spin, etc., and the liability to exercise that under conditions varying
with the mass, charge, spin, etc., of other objects).
This account of the ultimate determinants of what happens as
merely substances and their causal powers and liabilities provides
explanation of what happens in familiar terms. As I shall consider
more fully shortly, we ourselves have causal powers that we, unlike
inanimate objects, can choose to exercise. The S–P–L way of explaining things was the way familiar to the ancient and medieval world,
before talk of ‘laws of nature’ became common in the sixteenth
century. It was revived by Rom Harre´ and E. H. Madden in Causal
Powers.
8 When talk of ‘laws of nature’ became common in the
sixteenth century, they were supposed to be God’s laws for nature, and so such talk has its natural place in a theistic world view. But if
there is a God and he makes things in the world behave as they do, he
surely operates not directly, but by sustaining the laws of nature—
which means, on this account, by determining which powers and
liabilities substances have, and conserving those powers and liabilities in substances. The basic structure of explanation in terms of
substances, powers and liabilities does not presuppose that there is a
God who operates in this way.
With the S–P–L account, unlike with the Humean account and the
universals account, we have moved away from the Hempelian structure of scientific explanation in a crucial respect. For ‘laws of nature’
no longer play any causal role in explaining particular phenomena.
What causes the expansion of a particular piece of copper is that
piece of copper, its power to expand and its liability to exercise
that power when heated. The regularity involved in other pieces of
copper having similar powers and liabilities is no part of the explanation. While causation is essentially involved in laws, laws are not
essentially involved in causation. The S–P–L account of laws of
nature and of the explanation of particular events seems to me
more satisfactory than the other accounts. The regularities in the
causal powers and liabilities of particular substances, and so in their
behaviour, which constitute the ‘laws of nature’, do entail that particular substances will have particular powers and liabilities; and so,
any evidence that makes it probable that such and such (e.g. ‘all A’s
do so-and-so in circumstances C’) is a law of nature is evidence that
makes it (inductively) probable that a particular instance of it holds
(for example, that this A has the power to do such-and-such and the
liability to exercise it in circumstances C). But the law does not
explain why these substances have those powers and liabilities. And
so the S–P–L account raises the question of why so many substances
have similar powers and liabilities to each other (why does each
substance in the universe have the power to attract each other
substance in the way stated in, for example, Newton’s ‘laws’), and
we will return to that question in Chapter 8. But, as we shall see, a
question the same in essentials arises also on the other accounts of
laws of nature; and the argument of this book does not depend on my
preferred account of laws of nature and so of scientific explanation.
So I shall normally operate simply with the amended Hempelian
model without presupposing how it is to be spelled out, or corrected.
However, at crucial points I shall draw the reader’s attention to
34 The Nature of Explanation
alternative accounts of laws of nature and so of scientific explanation,
and especially to the S–P–L account. I now move on to contrast
scientific explanation with personal explanation
Personal explanation
The other pattern of explanation that we use all the time in explaining mundane phenomena is what I shall call personal explanation. In
personal explanation the occurrence of a phenomenon E is explained
as brought about by a rational agent P doing some action intentionally. The central case of this, with which we shall be primarily
concerned, is where P brings about E intentionally—that is, brings
E about, meaning so to do. The other case is where P brings about E
unintentionally in consequence of doing something else intentionally—we shall come to this case briefly later. In the central case E
occurred because P meant E to occur through what he was doing.
What an agent meant to occur through his agency may be called the
intention (or purpose—I shall use these terms interchangeably) J in
the agent’s action, for example, that E occur. E is then explained by P
having intention J. E may be the motion of my hand, P be myself, and
J my intention that E occur. E is then what I shall call the result of an
intentional action A of bringing E about.9 In the example cited, A is
my moving my hand. However, E is only partially explained by P’s
having intention J. For a person may have the intention to bring
about some effect and yet fail to do so. I may mean my hand to move
through my agency, and yet the hand may fail to move because
someone is holding it down; in consequence the only action that
I perform is that of trying to move my hand. If E does result from P
and J, a full explanation will tell us why, how it was that P’s intention
was efficacious.
This leads us to the well-known distinction10 among intentional
actions between basic actions and mediated actions. Roughly speaking, a basic action is something that an agent just does, does not
do by doing anything else. A mediated action is an action that is not a
basic action, one that an agent does by doing something else. I signal
by moving my hand. I break the door down by giving it a kick. The
former is a mediated action; the latter a basic action. Now, if bringing
about E is a basic action, the answer to the question how it was that
P’s intention was efficacious will simply be that bringing about E was
among the basic powers or capacities X that P had at that time—that
is, was among the basic actions that P could do at will (that is, would
succeed in doing if he formed the intention to do them). Bringing
about the motion of our arms or legs, lips, eyes or eyebrows, etc., is
for most of us most of the time among our basic powers. E is fully
explained when we have cited the agent P, his intention J that E occur,
and his basic powers X, which include the power to bring about E;
for, given all three, E cannot but occur. Of course, often in such cases
it is so obvious why E occurred that we do not bother to give the
explanation, but the explanation is true nevertheless. We may not
bother to comment, when someone is walking along, that his legs
moved because he moved them (that is, that he brought about their
motion, meaning so to do), but it is true nevertheless. Sometimes,
however, this sort of explanation is not at all obvious—it may on
occasion be the explanation of a person’s ears wiggling or her heart
stopping beating that she brought about these things intentionally.
If bringing about E is a mediated action, the answer to the question how it was that P’s intention was efficacious will be more
complicated. It will be that E was the intended consequence of
some basic action of P’s, A—that is, a consequence that P meant to
occur through his performing A, which consists in bringing about
some state of affairs S. P has the intention J that E occur as a
consequence of the occurrence of A (and so J contained within it
the intention that S occur). For P to have this latter intention, he
must believe that his doing A will (no less probably than his doing
any other basic action) have his bringing about E as a consequence
(normally perhaps by S causing E ). The explanation of how P’s
intention was efficacious is that bringing about S is among P’s basic
powers X, and that the bringing-about of S had as a consequence the
occurrence of E. There will often be a scientific explanation of
the latter. S may cause E in accord with natural laws L, because it is
a consequence of L that in circumstances D (which in fact hold) S is
followed by E. So, in this case, P, J, X, D, and L fully explain the
36 The Nature of Explanation
occurrence of E. E is brought about by P having a certain intention
J, which in consequence of his basic powers X brings about some
state of affairs S in circumstances D, which laws of nature L then
ensure will bring about E. Thus a full explanation of the door being
flat on the ground is that I, exercising my basic powers, brought it
about that my foot moved quickly into contact with the door,
meaning this to occur and it to cause the door to be flat; the
door hinges, the mass of the door, the mass and velocity of my foot
were in fact such that it was a consequence of the laws of mechanics
that the impact of my foot with the door was followed by the
flattening of the door. In the above analyses I use the word ‘consequence’ in a wide sense. The connection between A and E may be
either causal or logical. This may be as in the above example, because
S, the result of A, causes E. It may also be because, given current
circumstances D, the performance of A constitutes bringing about E.
Thus, given conventions in banking and motoring, my writing my
name in a certain place has as a consequence that a cheque bears my
signature, and my putting my arm out of the car window has as a
consequence that a signal indicating a turn to the right is made.
So, to summarize, in the central case of personal explanation we
explain a phenomenon E as brought about intentionally by a rational
agent P. If the bringing-about of E is a basic action A, we need to cite
further an intention J of P that E occur and to state that bringing
about E is among the things that P is able to do at will—namely,
among P’s basic powers X. P, J, and X provide a full explanation of E.
Of course, we can often go further and explain how it is that P has
intention J (for example, by stating that he formed this intention in
order to forward some wider intention, as when we explain that he
formed the intention to sign a cheque because he had the intention to
pay you money). Or we can explain how it is that P has those powers
(for example, by stating which nerves and muscles need to be operative for P to have these powers). But P, J, and X suffice to explain
E—whether or not we can explain how it is that J and X hold. If the
bringing-about of E is a mediated action, things are more complicated. We cite P and his intention J to bring about E as a consequence
of a basic action A; we explain that the performance of A was among
P’s basic powers, and we explain how the performance of A had E as a
consequence. Again, the occurrence and operation of the factors
cited here may themselves be explained further; but they do not
need to be for us to have a full explanation. When there is only a
The Nature of Explanation 37
basic action involved, the agent P is the cause of the effect; his
intention and powers provide the reason for the efficacy of the
cause. Where the action is a mediated action, further factors
are added. The two figures at the top of the diagram on p. 39
summarize these results for basic actions, and for mediated actions
in cases where a natural law L brings it about that S has E as a
consequence. Causes and the conditions for their operation (the
‘what’) are shown to the left of the arrows: reasons (the ‘why’) are
shown above the arrows; effects are shown to the right of the arrows.
There is, I claimed earlier, a second kind of personal explanation.
Here we explain the occurrence of E as brought about unintentionally by a rational agent P bringing about something else intentionally;
E is an unintended consequence of an intentional action. For example, in standing up I may unintentionally knock over a cup. Here
the knocking-over of the cup is caused by my occupying a certain
standing position, which was a state of affairs brought about intentionally by me. I did not mean the cup to be knocked over, but, given
the circumstances (the original position of the cup, etc.), my occupying the standing position causes the knocking-over of the cup in
virtue of the laws of mechanics L. My concern henceforward will be
only with the central case of intentional action where the effect is
brought about intentionally.
Personal Explanation Unanalysable in Terms
of Scientific Explanation
Personal explanation looks very different from scientific explanation.
In scientific explanation in the amended Hempelian model we explain an event E by past events or states C and natural laws L. In
personal explanation we explain E as brought about by an agent P
(not by an event or state) in order to realize intentions for the future.
Despite the apparent difference, it has, however, been argued by some
philosophers, seminally by Donald Davidson11 and by many others
at greater length, that really personal explanation conforms to the
scientific pattern. In my terminology and using the Hempelian model of scientific explanation, a Davidson-like suggestion amounts
to the following.
Suppose, first, that E is the result of a basic action. Then, to say
that P brought E about intentionally is just to say that an event
involving P—that is, P’s intention that E occur—J, brought it
about. To say that P had the power to bring about E is just to say
that P’s bodily condition Y (brain states, muscle states, etc.) and
environmental conditions Z (no one having bound P’s arm, etc.) and
laws L1 are such that an intention12 such as J is followed by the event
intended, E. We then have a scientific explanation as set out in the
diagram.
(fijarse página 48 del pdf)
P -J and X E
Structure of the central
case of personal explanation
of E, when E is the result of
a basic action.
Structure of the central case of
personal explanation of E, when
E is the result of a mediated
action. (One scheme).
Attempted analysis of the
above in the scientific pattern.
Attempted analysis of the above
in the scientific pattern.
Suppose, next, that E is the result of a mediated action. Then to say
that P brought it about is to say that an event involving P—that is,
P’s intention J—under the current bodily and environmental conditions Yand Z brought about (in accordance with laws L1) the result of
the basic action S, which had as a consequence E. We have seen that
there are different ways in which S may have E as a consequence. One
way is that S may bring about E in accord with the normal scientific
pattern of causation—that is, in virtue of some law of nature L. This
is the scheme depicted in the diagram. The other way in which S may have E as a consequence can also, it is suggested, easily be fitted into
the scientific pattern of explanation. So, on this reductionist view,
personal explanation is in essence really scientific explanation. There
are not explanations of events of two kinds—only explanations of
one kind. Events brought about by actions are just those that include
intentions among their causes.
In order to show what is wrong with this, I wish to make two
points—first, that the intention in an action that an agent is performing is not the same as any brain event that might be connected
with it; and, secondly, that having an intention (in the sense with
which we are concerned13) is not a passive state of an agent, but just
is the agent exercising causal influence (which will cause the effect
intended if and only if the agent has the requisite power).
I understand by a substance a thing (other than a property) that
has properties; tables, planets, atoms, humans, and other persons
are all substances. (Sometimes, when there is no danger of misunderstanding, I use ‘object’ as a synonym for ‘substance’.) Substances
have properties—that is, characteristics that can characterize
them and other substances as well. In this sense all properties are
universals; brown is a property, and different things can be brown.
Properties include both monadic properties, which characterize
individual substances, and relational properties, which link two or
more substances. Being square, weighing 10 kilos, or being-tallerthan are properties, the former two being monadic properties, the
latter being a relational property that relates two substances
(one thing is taller than another thing). On these definitions there
is no more to the history of the world than substances coming into
existence, gaining and losing properties (including relations to other
substances), and then ceasing to exist. It is useful to have a word for
these things such that there is no more to the history of the world
than all these things; and a natural word to choose for that category
of thing is the word ‘event’. I propose to use it in this sense: that an
event consists in the instantiation of a property in a substance (or
substances, or in events) at a time or the coming into being, or ceasing to exist, of a substance. Events include the table being
square now, or John being taller than James on 30 March 2001 at
10.00 a.m., or me coming into existence on 26 December 1934. In
order to fulfil the purpose of the definition of ‘event’, we need so to
individuate properties that, if you knew which properties had been
instantiated in what when, you would know (or could deduce)
everything that had happened. This will involve, for example, counting being red and reflecting light of such-and-such a wavelength as
different properties—for you could (just by looking at it in normal
light) know that something was red without knowing (or being able
to deduce) that it reflected light of such-and-such a wavelength, and
conversely.
It follows immediately that having an intention cannot be the
same event as having any brain event, for you could know that
someone was intending to do such-and-such in his action without
knowing that he was in a particular brain state or any brain state, and
conversely. These are two different events connected with a subject ,
even if perhaps of physical necessity they always go together. It is true
that other criteria for two events being the same event might yield a
different result—that the two events were the same; but then, to tell
the whole history of the world on those other criteria, it would not be
enough to know that some event (for example, some brain state) had
taken place; you would need to know that it had two different
somethings, say ‘characteristics’—a brain characteristic, and an intention characteristic—associated with it. Some sort of dualism is
unavoidable here, and I suggest that my proposed use of the word
‘event’ provides a neat system of categories by using which we can
describe the world fully, a system of categories not too distant from
ordinary usage.
So intentions are not brain events, even if closely connected
with brain events. In the sense to be defined in Chapter 9, they are
mental events. The next issue is what sort of mental events are they.
Is having an intention a passive state, some state in which the
agent finds himself—like having a sensation or a belief ? Davidson
thinks of intentions as ‘desires’ and let us read him (despite what
he writes elsewhere) as supposing that these are mental events
distinct from brain events. These desires may need some event such
as a perception or a neural event, to make them cause other events.
Then, he claims, actions are events that have passive mental
states, desires for their occurrence, among their causes. Personal
The Nature of Explanation 41
explanation is analysable in terms of the production of effects by such
desires.
Despite the fact that it is the most plausible form of reductionist
theory, like all the others, Davidson’s theory is open to a fatal
objection. The basic idea of all such theories is that an agent’s
bringing about an effect intentionally—that is, meaning so to do—
which is how we defined the agent’s bringing about an effect having
an intention so to do—is to be analysed as the causing of that
effect by some passive state of the agent or some event involving
him. But all such analyses fail because, if an intention (or wish or
desire) of P to bring about E is some passive state or event, it could
bring about E without P’s having intentionally brought about
E. Causation by an intention (so understood) does not guarantee
intentional action.
The classic objection to the reductionist theory was formulated as
follows by Richard Taylor. Here the causal factor is termed a ‘desire’,
but it could equally well be termed a ‘want’ or an ‘intention’.
Suppose . . . that a member of an audience keenly desires to attract the
speaker’s attention but, being shy, only fidgets uncomfortably in his seat and
blushes. We may suppose, further, that he does attract the speaker’s attention
by his very fidgeting; but he did not fidget in order to catch the speaker’s
attention, even though he desired that result and might well have realized
that such behaviour was going to produce it.14
Here we have a case of a desire for E causing E, and yet there is no
action. The basic point is that desires, wants, etc. may occur and yet
the agent for some reason may not act to fulfil his desire or want.
Nevertheless, in such a case, possibly without the agent’s knowledge,
the desire may bring about the intended effect—without the agent’s
bringing about the effect intentionally. An agent’s bringing something about intentionally is not analysable as his intention bringing
that thing about, if an intention is supposed to be a passive mental
event or state. The same applies if we substitute, for ‘intention’,
‘desire’, ‘want’, or any similar term.15 So a Davidson-type analysis seems to fail. To say that P brought something about intentionally is
not to say that some passive state of P or event involving P, such as an
intention, brought that thing about. There seems to be no other
plausible way in which personal explanations can be analysed into
the scientific pattern, and so it would appear that personal explanation is of a distinct type from scientific explanation (on the amended Hempelian model of the latter). (Note that in future I shall
understand by a ‘desire’ to do some action a causally influential
inclination to do the action, which may or may not coincide with a
judgement by the agent that it is overall good to do the action. If it
does not coincide, the agent has to choose whether to resist his desire
or yield to it.)
If intentions are not states or events that happen to an agent, they
must themselves be actions. Having an intention is not something
that happens to an agent, but something she does. For me to have the
intention in acting of moving my hand is to do what (if I were to fail
or find it difficult to move the hand) would be called ‘trying’ to move
my hand. In the past, having such an intention has been given the
technical name of making a ‘volition’ to move my hand. The basic
mistake that reductionist analyses make is (in the terminology introduced at the beginning of the chapter) to treat intentions as belonging to the ‘what’ rather than to the ‘why’ of explanation. When
one explains an occurrence as brought about by an agent having
some intention, one is not by the word ‘intention’ describing some
occurrent state or event that caused the occurrence, but one is stating
that the agent brought about that occurrence and did so because he
meant to do so. To act intentionally is to exercise causal agency in a
certain direction, which will succeed in producing the intended effect
if the agent has the requisite power. An intention—to avoid a puddle,
say—explains why at a certain time a man with normal basic powers
(and that involves, physically, a normal brain and the operation
of normal psycho-physical laws) behaved as he did, made such movements as in fact led to his feet bypassing the puddle. That this
account is correct is brought out by the linguistic fact that explanations in terms of intentions can easily be paraphrased in terms of
explanations in which there occur no nouns that could conceivably
be regarded as denoting occurrent states or events. To say that
a man’s intention in making certain movements was to avoid the
puddle is to say that he made them in order to avoid the puddle, or so
as to avoid the puddle. But no such paraphrase is possible for the
initial conditions which are cited in normal scientific explanations.
Although intentions, like laws of nature, belong to the ‘why’, the
reasons, of explanation, there are, of course, vast differences between
laws of nature and intentions. Intentions are such that necessarily the
agent whose they are ‘goes along’ with them, is aware of them, and
has privileged access to them in the sense that he is in a better
position than outsiders to know about them. Laws of nature are
not necessarily known to anyone, nor necessarily does any person
‘go along with them’ or have privileged access to them. But that
the ‘why’ is here known and adopted by an agent is one of the
differences between personal and scientific explanation. The other
main difference is that, in personal explanation, talk about a substance which explains, namely a person, is not reducible to talk about
occurrent states of or events involving that person. The contrast
between scientific and personal explanation remains even on the
S–P–L account of the former, although the two patterns are much
closer to each other on this account. That is a reason for preferring
the S–P–L model; it brings out that both personal and scientific
explanation are species of the same genus, causal explanation. In
both, the cause is a substance or substances. In both, the substance
has certain powers, and produces the effect in virtue of its powers.
The difference is that, in scientific explanation, the substance has
liabilities to exercise its powers under certain circumstances; it is
either physically necessary or probable that it will, and it has no
intention or purpose in doing what it does; whereas in personal
explanation the substance (the person) acts intentionally, doing the
action that—given its beliefs—will most probably fulfil its intentions. There is no parallel for that in the scientific case. In consequence, even if scientific explanation can be expressed in terms of an
event (the substance being in certain circumstances) rather than
the substance itself causing the effect, personal explanation cannot
be expressed in this way. A person causing an effect is not analysable
44 The Nature of Explanation
as a passive state of that person or an event involving that person
causing the effect.
Can there be Two Explanations of a Phenomenon?
So far in this chapter I have been concerned to characterize the
structures of the two types of explanation that we use in explaining
the occurrence of phenomena, and to show how they differ from
each other. I now turn to the question of whether there can be only
one true explanation of some phenomenon. For, if so, then, if there is
a personal explanation of some phenomenon, there cannot be a
scientific one, and conversely. I suggest that there can be two true
distinct explanations of some phenomenon E, if one or other of three
conditions is satisfied, but that otherwise there cannot be.
Clearly there can be two true distinct explanations of E, when one
or other or both are partial explanations of E. For the one may
combine with the other to make a fuller explanation. Thus a man’s
death from cancer may be explained by (1) his smoking and a law
about the proportion of smokers who die from cancer, and by (2) his
parents’ having died of cancer and a law about the proportion of
those whose parents die of cancer who themselves die of cancer. Since
(1) and (2) only make probable but do not necessitate the man’s
death from cancer, they are only partial explanations. Clearly they
can be combined into a fuller explanation in terms of the man’s
smoking and his parents having died of cancer and the proportion of
those who smoke and whose parents have died of cancer who die
of cancer.
But can there be two different full explanations of a phenomenon?
The answer is still yes—if the occurrence of the causes (the ‘what’)
and the operation of the reasons (the ‘why’) cited in one explanation
are to be explained at least in part by the occurrence of the causes and
the operation of the reasons cited in the other explanation. For
example, the present position of Mars is explained by its position
in the last few days and the laws of planetary motion, formulated
more or less correctly by Kepler. Where it has been recently and the
laws stating how planets move determine where Mars will be today.
Yet the present position of Mars is also explained by its position and
velocity last year and those of all other heavenly bodies, and Newton’s
laws of motion. Newton’s laws state how material bodies change their
The Nature of Explanation 45
velocities under the influence of other bodies. Both are full explanations, and yet they are clearly compatible. This is because Newton’s
laws and the positions and velocities of the planets explain their
(approximate) conformity to Kepler’s laws. Kepler’s laws operate
because Newton’s laws operate and the sun and the planets have
the initial positions and velocities that they have, and are far distant
from other massive bodies. It is for this reason that the motion of a
human hand is often explicable by both personal and scientific
explanation. The motion of my hand may be fully explained by
goings-on in the nerves and muscles of my arm, and physiological
laws. It may also be fully explained by me bringing it about, having
the intention and power so to do. Yet in this case the causes and
reasons cited in each explanation provide a partial explanation of the
occurrence and operation of the causes and reasons cited in the other.
The goings-on in my nerves and muscles are brought abut unintentionally by my bringing about the motion of my hand intentionally.
Also, the operation of physiological laws provides part of the explanation of my having the power to move my hand—only because
nervous discharges are propagated as they are, am I able to move
my hand. So there is here a twofold reason why two full explanations
can each fully explain the motion of my hand.
But can there be two distinct full explanations of some phenomenon E, when neither in any way explains the occurrence or operation of the causes and reasons involved in the other? Yes, again, so
long as there is overdetermination. In overdetermination each of the
full explanations gives causes and reasons sufficient for the occurrence of the effect, but neither pair on its own is necessary since the
other pair would have produced the effect on its own. If someone
dies as a result of being poisoned by A at the same time as he is shot
by B, we have such overdetermination. But such coordination will be
a coincidence, barring a common cause of the actions of A and B (for
example, C who employed both A and B to murder the same victim
in order to ensure that he really died). It cannot be necessary for the
production of the effect to have two distinct full explanations, when
neither in any way explains the occurrence or operation of the causes
and reasons involved in the other; unless the occurrence and operation of the causes and reasons involved in both are explicable, at
least in part, by the causes and reasons of a third full explanation (a
common cause). It follows, given that scientific and personal explan46 The Nature of Explanation
ation are the only two possible kinds of explanation,16 and barring
accidental overdetermination, that there can be a full personal explanation and a full scientific explanation of some phenomenon only
if one in part explains the occurrence and operation of the components of the other—either the scientific explanation at least in
part explains the causes and reasons in the personal explanation, or
conversely; or there is a further full explanation (either personal
or scientific) that explains the causes and reasons operative in both
the other explanations.
Explanation by the Action of God
In this chapter so far I have been concerned to analyse the structure
of personal explanation, and to show its relation to scientific explanation. I have done this because, when the theist claims that the action of God explains various phenomena, such as the existence and
orderliness of the world, he is proposing a personal explanation of
these phenomena. However, personal explanations of phenomena by
the action of God differ from most mundane personal explanations
in two important respects, on which I must now comment in conclusion of this chapter.
The first is that a personal explanation of the occurrence of a
phenomenon E in terms of God’s bringing it about, meaning so
to do, cannot be even in part explicable scientifically. We have seen
that a personal explanation may often, at any rate in part, be explained by a scientific explanation—and conversely. Thus a person
having the powers that she has may be explained in part by her having
nerves and muscles and by the operation of various physiological
laws. Her having the intentions that she has may also be given a
scientific explanation, and perhaps a human’s existence may also
be explained in this way. The fact that personal explanation cannot
be analysed in terms of scientific explanation does not mean that
its operation on a particular occasion cannot be given a scientific
explanation. However, it seems coherent to suppose that there should
be a personal explanation of the occurrence of some event E by
the agency of an agent P having the intention J to bring about E
and the power so to do, without all this being in any way susceptible
of a scientific explanation. To start with, an agent might have the
power to perform certain basic actions without his having that
power being dependent on any physical states or natural laws.
His capacity to perform these actions might be an ultimate brute
fact (or only explicable by another personal explanation). Likewise,
an agent’s having the intentions in actions that he does, his
choice of intentional actions, may not be susceptible of scientific
explanation.
To see the above, note that there is at present with respect to
some of the intentions that we form no plausible scientific explanation of why we form these intentions, rather than any other ones.
And yet our explanations of other things in terms of these intentions
would still be explanations even if there was no explanation of
why we formed these intentions. Then we have basic powers to
bring about mental images of different geometrical shapes. There
might be a partial scientific explanation of my having this power in
terms of my brain being in a certain state giving me this power.
Yet there is no contradiction in supposing that powers of visualiza48 The Nature of Explanation
tion are not dependent on the brain, or on anything else. Maybe we
just have such powers. But that would not affect the fact that my
having a certain mental image could be explained by my basic power
to produce such images. And, although we normally suppose (correctly) that there is a scientific explanation of the existence of this
body that is mine, there is no scientific explanation of how it
comes about that this body is mine (rather than someone else’s)
and so no scientific explanation of my existence at all. For this world
could have been the same in all its physical aspects, and yet a different
person could have operated through this body. (I develop this point
more fully in Chapter 9.) And yet the fact that science cannot explain
my existence does not mean that there is no true explanation of
things in terms of me bringing them about. Personal explanation
may explain without there being a scientific explanation of the
occurrence and operation of the factors involved in it.
When the theist claims that the existence of the world and its
various features is to be explained by the action of God bringing these
things about meaning so to do, he will claim that God’s action cannot
be explained scientifically, even in part. God is supposed to be
perfectly free. God’s existence and powers do not depend on
the states of the physical world or the laws of its operation—rather,
vice versa. Nor are God’s intentions scientifically explicable. But all
this, as we have now seen, does not in any way weaken the explanatory value of the personal explanation. God’s bringing about some
event may be explicable by a wider personal explanation. He may
bring about E in order thereby to bring about F; F may be an event
that takes a considerable period of time, and E may be the first stage
of F. But the theist claims that this kind of explanation is the only
kind of explanation of God’s actions that can be provided. God’s own
intentions alone explain his doing what he does. God’s basic actions
are supposed to include creating the universe e nihilo (that is, not out
of existing matter), keeping it in existence, making things behave in
accord with natural laws, and occasionally intervening in the universe (sometimes by setting those laws aside). Creating matter e
nihilo is not something that humans are able to do, but it is easy
enough to conceive of their doing it. It is logically possible that
I could just find myself able as easily to make appear before me an
inkwell or to make a sixth finger grow, as I am at present able to move
my hand. Various tests (for example, sealing off the room and
keeping its contents carefully weighed) could show that the inkwell
The Nature of Explanation 49
or finger were not made of existing matter. Creating e nihilo is a
perfectly conceivable basic act.
The other important respect in which personal explanations of
phenomena by the action of God differ from most mundane personal explanations is that God is supposed to be a person without a
body—that is, a spirit. It is important to make clear at this stage what
it is for a person not to have a body. We can best do so by asking a
different question—what is it that I am saying when I say that this
body, the body behind the desk, is my body? First, that I can move, as
basic actions of mine, many parts of this body, whereas I can make a
difference to anything else only by moving parts of this body. To
move the arm over there (your arm), I have to grasp it with this arm,
but I can move this arm straight off. Secondly, my having a mental
life of thought and feeling and intention depends causally on the
operation of this body, and in considerable part which mental events
I have (in particular my sensations, feelings, and perceptual beliefs)
are caused by events in this body. In so far as these events are caused,
it is events in this body that cause them; and other events (for
example, occurrences in the room) cause my mental events only by
causing events in this body that cause the mental events. In consequence, thirdly, while I am aware of goings-on in this body without
causal influences from outside the body impinging on it (I know the
position of these limbs and feel the emptiness of this stomach), I can
come to know about things outside the body only through their
effects on this body. I see the desk and so know where it is only
because light rays from the desk impinge on these eyes. I learn what
you tell me only because by talking you set up air vibrations that
impinge on these ears. And, fourthly, I look out on the world from
where this body is. It is things around this body that I see well, things
further away that I see less well.
Clearly a person has a body if there is a physical object (that is, a
substance) to which he is related in all of the above four ways. And
clearly a person does not have a body if there is no physical object to
which he is related in any of the above ways. If a person is related to
different physical objects in each of these ways, we shall have to say
that his body is of a different kind or more widely extended than
ours. And if he is related to a physical object only (or only to some
50 The Nature of Explanation
degree) in some of these ways, we shall have to say that he is
embodied only to some degree.17
Now, on the traditional account of God, God is supposed not to be
embodied in any of these ways. There is no physical object, not even
the whole universe, through which he has to act in order to make
differences to other things. He could abolish this physical universe at
a stroke and create another one, and he can exert causal influence on
non-embodied creatures without needing to operate through anything physical in order to do so. Nor is God dependent on anything
physical or anything else for his life of thought. And he knows about
everything without being dependant on any physical process for the
acquisition of his knowledge. And he does not have any particular
perspective on the world. He knows how things are without being
dependent for his knowledge on a particular pattern of sensations
arising from a particular viewpoint. So God is in no way embodied.
He can, of course, move any part of the physical universe as a basic
action, and knows without inference about the state of every part of
the universe; but that does not make the physical universe his body,
because he is not dependent on the universe for this ability and
knowledge.
So then in the arguments to the existence to God the theist argues
from the existence and order of the world and various features of it to
a person, God, who brought these things about, meaning so to do. In
this chapter I have been concerned to analyse what it is to explain an
event as brought about by some person meaning so to do; and in
conclusion I have drawn attention to two special features of personal
explanations in terms of the action of God.
Having investigated in this chapter the structure of personal explanation, in the next chapter I will consider the evidence that justifies
us in putting it forward, the evidence that makes it probable that an
explanation of the personal type rather than one of the scientific type
is the true explanation of some phenomenon. We will then be in a
position to see whether the evidence recorded in the premisses of
arguments to the existence of God constitutes such evidence.
