The first two statements are called the premisses of the inference,

while the third statement is called the conclusion. This is a

deductive inference because it has the foilowing property: ifthe
Sq:$*s's*$##w rwm x,ww'x $ ru gry
premisses are true, then the conclusion must be true too. In other
words, if it's true that all Frenchman like red wine, and if it's true
that Pierre is a Frenchman, it follows that Pierre does indeed like
red wine. This is sometimes expressed by saying that the
premisses of the inference entail the conclusion. Of course, the
premisses of this inference are almost certainly not true - there
are bound to be Frenchmen who do not like red wine. But that is
not the point. What makes the inference deductive is the
existence of an appropriate relation betr,veen premisses and
Scientists often tell us things about the world that we would not conclusion, namely that if the premisses are true, the conclusion
otherwise have believed. For example, biologists tell us that we are must be true too. Whether the premisses are actuaily true is a
closely related to chimpanzees, geologists tell us that Africa and different matter, which doesn't affect the status of the inference as
South America used to be joined together, and cosmologists tell us deductive.
that the universe is expanding. But how did scientists reach these o
unlikely-sounding conclusions? After all, no one has ever seen one Not all inferences are deductive. Consider the following example:
species evolve from another, or a single continent split into two, or o
the universe getting bigger. The answer, ofcourse, is that scientists The first five eggs in the box were rotten
arrived at these beliefs by a process ofreasoning or inf'erence. But it G
Al1 the eggs have the same best-before date stamped on them
would be nice to know more about this process. What exactly is the
nature ofscientiflc reasoning? And how much confidence shouid we Therefore, the sixth egg u'ill be rotten too
place in the inferences scientists make? These are the topics ofthis
chapter. This looks like a perfectly sensibie piece of reasoning. But
nonetheless it is not deductive, for the premisses do not entail the
conclusion. Even ifthe first five eggs were indeed rotten, and even if
Deduction and induction
all the eggs do have the same best-before date stamped on them,
Logicians make an important distinction between deductive and this does not guarantee that the sirth egg witl be rotten too. It is
inductive patlerns of reasoning. An example of a piece of deductive quite conceivable that the sixth egg will be perfectly good. In other
reasoning, or a deductive inference, is the following: words, it is logically possible for the premisses of this inference to be
true and yet the conclusion false, so the inference is not deductive.
All Frenchmen like red u ine
Instead it is known as an inductive inference. In inductive
Pierre is a Frenchman
inference, or inductive reasoning, we move from premisses about
objects we have examined to conclusions about objects we haven't
Therefore, Pierre likes red r.vine
examined - in this example, eggs.
 Deductive reasoning is a much safer activity than inductive
reasoning. When we reason deductively, we can be certain that if
we start with true premisses, we will end up with a true conclusion.
But the same does not hold for inductive reasoning. On the
contrary, inductive reasoning is quite capable oftaking us from
true premisses to a false conclusion. Despite this defect, we seem at ,..
{d*
-*, -^i c t
to rely on inductive reasoning throughout our lives, often without
even thinking about it. For example, when you turn on your
Ii
ttL. ?.l''
ii *"
Ii
*e **
t ?
s:
!!
ll*
computer in the morning, you are confldent it will not explode in
your face. Why? Because you turn on your computer every
Er
-+
*o .|'A
t* ,'^Jc.q
3*
;tb -lt il
-rf,
{ d
morning, and it has never exploded in your face up to now. But the
inference from'up until now, my computer has not exploded when
I turned it on'to'my computer will not explode when I turn it on
this time'is inductive, not deductive. The premiss of this inference
s C-a l .5*a
does not entail the conclusion. It is logically possible that your ,18.
Ae? ;;.3 r.r *ti ". * l-
Tt !*l.
ET t ? t i '1 t ! r
o
c
computer will explode this time, even though it has never done so
previously.
;i!
*t !i !l
tfl
5{} FT
t*' i;. Bf
6 dF
Sl
o
q
o
Other examples of inductive reasoning in everyday life can readily 7 I '1
1
'1
2
o be found. When you turn the steering wheel of your car
4 anticloclrvise,
you assume the car will go to the left not the right. :$ r'
Whenever you drive in traffic, you effectively stake your life on this &ft
** 6* ,t 8{* t{ ia*
lJl '{: #t t8
assumption. But what makes you so sure that it's true? If someone
asked you to justify your conviction, what would you say? Unless
tf *s
13 l(+ 16 17 {o
you are a mechanic, you would probably reply: 'every time I,ve
turned the steering wheel anticlockwise in the past, the car has gone
to the left. Therefore, the same will happen when I turn the steering
&
wheel anticlochrise this time.'Again, this is an inductive inference,
tr . J-
!l *l}r:'
not a deductive one. Reasoning inductively seems to be an :
indispensable part of everyday life. 1S 2A 21 22 {
a

X
Do scientists use inductive reasoning too? The answer seems to be
yes. Consider the genetic disease known as Down's sptdrome (DS
for short). Geneticists tell us that DS sufferers have an additional 5. A representation ofthe complete set of chromosomes - or
chromosome - they have 4,7 instead of the normal a6 (Figure 5). karyot5pe - of a person with Dou'n's syndrome. There are three copies
How do they know this? The answer, of course, is that they of chromosome 21, as opposed to the trvo copies most people have,
giving 4f chromosomes in total.
 examined a large number of DS sufferers and found that each had for humans. The word 'proof should strictly only be used when we
an additional chromosome. They then reasoned inductively to the are dealingwith deductive inferences. In this strict sense ofthe
conclusion that all DS sufferers, including ones they hadnt word, scientific hypotheses can rarely, if ever, be proved true by the
examined, have an additional chromosome. It is easy to see that this data.
inference is inductive. The fact that the DS sufferers in the sample
studied had 47 chromosomes doesn't prove that all DS sufferers do. Most philosophers think it's obvious that science relies heavily on
It is possible, though unlikely, that the sample was an inductive reasoning, indeed so obvious that it hardly needs arguing
unrepresentative one. fbr. But, remarkably, this was denied by the philosopher Karl
Popper, who we met in the last chapter. Popper claimed that
This example is by no means an isolated one. In effect, scientists use scientists only need to use deductive inferences. This would be nice
inductive reasoning whenever they move from limited data to a if it were true, fbr deductive inferences are much safer than
more general conclusion, which they do all the time. Consider, for inductive ones, as we have seen.
example, Newton's principle of universal gravitation, encountered
in the last chapter, which says that every body in the universe exerts Popper's basic argument was this. Although it is not possible to
a gravitational attraction on every other body. Now obviously, prove that a scientific theory is true from a limited data sample, it is
c Newton did not arrive at this principle by examining every single possible to prove that a theory is fblse. Suppose a scientist is
o
body in the whole universe - he couldn't possibly have. Rather, he considering the theory that all pieces of metal conduct electricity. D

sau' that the principle held true for the planets and the sun, and for Even if every piece of metal she examines does conduct electricity,
q objects ofvarious sorts moving near the earth's surface. From this this doesnt prove that the theory is true, for reasons that we've seen. 0
o
data, he inferred that the principle held true for all bodies. Again, But if she finds even one piece of metal that does not conduct f
E
d
this inference was obviously an inductive one: the fact that electricity, this does prove that the theory is false. For the inference
Newton's principle holds true for some bodies doesnt guarantee from'this piece of metal does not conduct electricity'to'it is
that it holds true for all bodies. false that all pieces ofmetal conduct electricity'is a deductive
inference - the premiss entails the conclusion. So if a scientist is
The central role of induction in science is sometimes obscured by only interested in demonstrating that a given theory is false, she
the way we talk. For example, you might read a newspaper report may be able to accomplish her goal without the use of inductive
that says that scientists have found'experimental proof that inferences.
genetically modified maize is safe for humans. \A/hat this means is
that the scientists have tested the maize on a large number of The weakness of Popper's argument is obvious. For scientists are
humans, and none of them have come to any harm. But strictly not only interested in showing that cerlain theories are false. When
speaking this doesnt proae that the maize is safe, in the sense in a scientist collects experimental data, her aim might be to show that
which mathematicians can prove Pl,thagoras'theorem, say. For the a particular theory - her arch-rival's theory perhaps - is false. But
inference from'the maize didn't harm any of the people on rvhom it much more likely, she is trying to convince peopie that her own
was tested'to'the maize will not harm anyone'is inductive, not theory is true. And in order to do that, she will have to resort to
deductive. The newspaper report should really have said that inductive reasoning of some sort. So Popper's attempt to show that
scientists have found extremely goo d eoidence that the maize is safe science can get by without induction does not succeed.

23
 Hume's problem But how do we know that the UN assumption is actually true,
Hume asks? Can we perhaps prove its truth somehow (in the strict
Although inductive reasoning is not logically watertight, it sense of proof)? No, says Hume, lve cannot. For it is easy to imagine
nonetheless seems like a perfectly sensible way of forming beliefs a universe where nature is not uniform, but changes its course
about the world. The fact that the sun has risen every day up until randomly from day to day. In such a universe, computers might
now may not prove that it will rise tomorrow, but surely it gives us sometimes explode for no reason, water might sometimes intoxicate
very good reason to think it will? If you came across someone who us without u'arning, billiard balls might sometimes stop dead on
professed to be entirely agnostic about whether the sun will rise colliding, and so on. Since such a'non-uniform'universe is
tomorrow or not, you would regard them as very strange indeed, if conceivable, it follows that we cannot strictly prove the truth of llN.
not irrational. For ifwe could prove that UN is true, then the non-uniform
universe would be a iogicai impossibility.
But what justifies this faith u,e place in induction? How should we
go about persuading someone who refuses to reason inductively Granted that we cannot prove IJN, we might nonetheless hope to
that they are wrong? The 1Sth-century Scottish philosopher David find good empirical evidence for its truth. After all, since IIN has
Hume (t7tt-t776) gave a simple but radical answer to this always held true up to now, surely that gives us good reason for
o question. He argued that the use of induction cannot be rationally thinking it is true? But this argument begs the question, says
justified at all. Hume admined that we use induction all the time, HumeJ For it is itself an inductive argumen4 and so itself depends
o in everyday life and in science, but he insisted this was just a on the UN assumption. An argument that assumes tIN from the
o
ro matter of brute animal habit. If challenged to provide a good outset cleariy cannot be used to show that UN is true. To put the o
o
o reason for using induction, we can g:ve no satisfactory answer, he point another way, it is certainly an established fact that nature has
A @
thought. behaved largely uniformly up to now. But we cannot appeal to this
fact to argue that nature will continue to be uniform, because this
How did llume arrive at this stafiling conclusion? He began by assumes that what has happened in the past is a reiiable guide to
noting that whenever we make inductive inferences, we seem to what will happen in the future - which fs the uniformity of nature
presuppose what he called the'uniformity of nature'(UN). To see assumption. If we try to argue for f)N on empirical grounds, we end
what Hume means by this, recall some of the inductive inferences up reasoning in a circle.
from the last section. We had the inference from'my computer
hasn't exploded up to now'to'my computer wont explode today'; The force of Hume's point can be appreciated by imagining how you
from'all examined DS sufferers have an extra chromosome'to'all would go about persuading someone who doesn't trust inductive
DS sufferers have an ex-tra chromosome'; from'all bodies observed reasoning that they should. You would probably say: 'look, inductive
so far obey Newton's law of gravifz'to'all bodies obey Newton's law reasoning has worked pretty weil up until now. By using induction
ofgravity'; and so on. In each ofthese cases, our reasoning seems to scientists have split the atom, landed men on the moon, invented
depend on the assumption that objects we havent examined will be computers, and so on. Whereas people who havent used induction
similar, in the relevant respects, to objects of the same sort that we have tended to die nasty deaths. They have eaten arsenic believing
have examined. That assumption is what Hume means by the that it would nourish them, jumped offtatl buildings believing that
uniformiq' of nature. they would fly, and so on (Figure 6). Therefore it will clearly pay you
 to reason inductively.' But of course this wouldn't con'uince the
doubter. For to argue that induction is trustworthy because it has
worked weli up to now is to reason in an inductive way. Such an
argument would carry no weight with someone who doesn't aiready
trust induction. That is Hume's fundamental point.

So the position is this. Hume points out that our inductive

inferences rest on the UN assumption. But we cannot prove that
UN is true, and we cannot produce empirical evidence for its truth
r'vithout begging the question. So our inductive inferences rest on an
assumption about the world for which we have no good grounds.
Hume concludes that our confidence in induction is just blind
faith - it admits of no rational iustification whatever.

This intriguing argument has exerted a powerful influence on the

phiiosophy ofscience, and continues to do so today. (popper's
unsuccessful attempt to show that scientists need only use o

deductive inferences was motivated by his belief that Hume had

o
shown the total irrationalif' ofinductive reasoning.) The influence o
o
of Hume's argument is not hard to understand. For normally we
a
think of science as the very paradigm of rational enquiry. We place

EI
great faith in what scientists tell us about the world. Every time we

H
F LI
t"

r'
travel by aeroplane, we put our lives in the hands ofthe scientists
who designed the plane. But science relies on induction, and
Hume's argument seems to show that induction cannot be
rationallyjustified. If Hume is right, the foundations on which
L
science is built do not look quite as solid as we might have hoped.
This puzzling state of affairs is known as Hume's problem of

t-- -
induction.

Philosophers have responded to Hume's problem in literally dozens

of different ways; this is stiil an active area of research today. Some
people believe the key lies in the concept ofprobability. This
suggestion is quite plausible. For it is natural to think that although
the premisses of an inductive inference do not guarantee the truth
6. What happens to people rvho don't trust induction. of the conclusion, they do make it quite probable. So even if
 scientific knowledge cannot be certain, it may nonetheless be highly lnference to the best explanation
probable. But this response to Hume's problem generates
difficulties of its own, and is by no means universally accepted; we The inductive inferences we've examined so far have all had
will return to it in due course. essentially the same structure. In each case, the premiss of the
inference has had the form'all x's examined so far have been y',
Another popular response is to admit that induction cannot be and the conclusion has had the form'the next x to be examined
rationally justified, but to argue that this is not realiy so problematic will be y', or sometimes,'all x's are y'. In other words, these
after all. How might one defend such a position? Some inferences take us from examined to unexamined instances of a
philosophers have argued that induction is so fundamental to how given kind.
we think and reason that it's not the sort of thing that could be
justified. Peter Strawson, an influential contemporary philosopher, Such inferences are widely used in everyday life and in science, as
defended this view with the following analogy. If someone worried we have seen. However, there is another common f,pe of non-
about u,hether a particular action was legal, they could consult the deductive inference that doesnt fit this simple pattern. Consider the
law-books and compare the action with what the law-books say. But lbllowing exampie:
suppose someone worried about whether the law itself was legal.
o This is an odd worry indeed. For the law is the standard against The cheese in the larder has disappeared, apart from a
o
which the legaliq' of other things is judged, and it makes little sense few crumbs 6

o to enquire whether the standard itself is legal. The same applies to Scratching noises were heard coming from the larder last night
o
c induction, Strawson argued. Induction is one ofthe standards we o
o
o use to decide whether claims about the world are justified. For Therefore, the cheese was eaten by a mouse
G
example, we use induction to judge whether a pharmaceutical
company's claim about the amazing benefits of its new drug are It is obvious that this inference is non-deductive: the premisses do
justified. So it makes iittle sense to ask whether induction itself is not entail the conclusion. For the cheese could have been stolen
justified. by the maid, who cleverly left a few crumbs to make it look like
the handiwork of a mouse (Figtre 7). And the scratching noises
Has Strawson really succeeded in defusing Hume's problem? Some could have been caused in any number ofways - perhaps they
philosophers say yes, others say no. But most people agree that it is were due to the boiler overheating. Nonetheless, the inference is
very hard to see how there could be a satisfactoryjustification of clearly a reasonable one. For the hlpothesis that a mouse ate the
induction. (Frank Ramsey, a Cambridge philosopher from the cheese seems to provide a better explanation ofthe data than do
192os, said that to ask for a justification of induction was 'to cry for the various alternative explanations. After all, maids do not
the moon'.) Whether this is something that should worry us, or normally steal cheese, and modern boilers do not tend to
shake our faith in science, is a difficult question that you should overheat. Whereas mice do normally eat cheese when they get the
ponder for yourself. chance, and do tend to make scratching sounds. So although we
cannot be certain that the mouse hypothesis is true, on balance it
looks quite plausible: it is the best way of accounting for the
available data.
 different tlpes of non-deductive inference. Nothing hangs on which
choice of terminology we favour, so long as we stick to it
consistently.

Scientists frequently use IBE. For example, Darwin argued for his
theory of evolution by calling attention to various facts about the
living world which are hard to explain if we assume that current
species have been separately created, but which make perfect sense
if current species have descended from common ancestors, as his
theory held. For example, there are close anatomical similarities
between the legs of horses and zebras. How do we explain this, if
God created horses and zebras separately? Presumably he could
have made their legs as different as he pleased. But ifhorses and
f^ zebras have both descended from a recent common ancestor, this
provides an obvious explanation of their anatomical similari[,.
o Darwin argued that the abiliq' of his theory to explain facts of this
o o
sort, and of many other sorts too, constituted strong evidence
o for its truth.
o
o B
o
Another example of IBE is Einstein's famous work on Brownian
l G
motion. Brownian motion refers to the chaotic, zig-zagmotion of
microscopic particles suspended in a liquid or gas. It was discovered
/. The mouse h]?othesis and the maid h)?othesis can both account for in 1827 by the Scottish botanist Robert Brown (1713-1858), while
the missing cheese.
examining pollen grains floating in water. A number of attempted
explanations of Brownian motion were advanced in the 19th
century. One theory attributed the motion to electrical attraction
Reasoning ofthis sort is known as 'inference to the best between particles, another to agitation from external surroundings,
explanation', for obvious reasons, or IBE for short. Certain and another to convection currents in the fluid. The correct
terminological confusions surround the relation between IBE and explanation is based on the kinetic theory ofmatter, which says that
induction. Some philosophers describe IBE as a f,pe of inductive liquids and gases are made up of atoms or molecules in motion. The
inference; in effect, they use'inductive inference'to mean suspended particles collide with the surrounding molecules,
'any inference which is not deductive'. Others contrast IBE wiih causing the erratic, random movements that Brown first observed.
inductive inference, as we have done above. On this way of cutting This theory was first proposed in the late 19th century but was not
the pie,'inductive inference'is reserved for inferences from widely accepted, not least because many scientists didn't believe
examined to unexamined instances of a given kind, of the sort we that atoms and molecules were real physical entities. But in 1905,
examined earlier; IBE and inductive inference are then two Einstein provided an ingenious mathematical treatment of

31
 Brownian motion, making a number of precise, quantitative that has been gained through ordinary induction. Thus it is
predictions which were later confirmed experimentally. After incorrect to regard IBE as a more fundamental mode of inference.
Einstein's work, the kinetic theory was quickly agreed to provide a
far better explanation of Brownian motion than any ofthe Whichever of these opposing views we favour, one issue clearh
alternatives, and scepticism about the existence of atoms and demands more attention. If we want to use IBE, we need some way
molecules rapidly subsided. of deciding which of the competing hypotheses provides the best
explanation of the data. But what criteria determine this? A popular
One interesting question is whether IBE or ordinary induction is a answer is that the best explanation is the simplest or the most
more fundamental pattern of inference. The philosopher Gilbert parsimonious one. Consider again the cheese-in-the-larder
Harman has argued that IBE is more fundamental. According to example. There are two pieces of data that need explaining: the
this view, whenever we make an ordinary inductive inference such missing cheese and the scratching noises. The mouse hypothesis
as 'all pieces of metal examined so far conduct electriciff, therefore postulates just one cause - a mouse - to explain both pieces of data.
all pieces of metal conduct electricif" we are implicitly appealing to But the maid hl,pothesis must postulate two causes - a dishonest
explanatory considerations. We assume that the correct explanation maid and an overheating boiler - to explain the same data. So the
for why the pieces of metal in our sample conducted electriciSi, mouse hypothesis is more parsimonious, hence better. Similarly in
whatever it is, entails that all pieces of metal will conduct eiectriciq'; the Darwin example. Darwin's theory could explain a very diverse
u
that is why we make the inductive inference. But if we believed, for range offacts about the living world, notjust anatomical o
o example, that the explanation for why the pieces of metal in our
similarities between species. Each ofthese facts could be explained
co sample conducted electricif was that a laboratory technician had in other ways, as Darwin knew. But the theory of evolution c
o tinkered with them, we would not o
infer that all pieces of metal explained all the facts in one go - that is what made it the best
4 conduct electricity. Proponents ofthis view do not say there is no G
explanation of the data.
difference between IBE and ordinary induction * there clearly is.
Rather, they think that ordinary induction is ultimately dependent The idea that simplicity or parsimony is the mark of a good
on IBE, explanation is quite appealing, and certainly helps flesh out the idea
of IBE. But if scientists use simplicity as a guide to inference, this
However, other philosophers argue that this gets things backwards: raises a problem. For how do we know that the universe is simple
IBE is itselfparasitic on ordinary induction, they say. To see the rather than complex? Preferring a theory that explains the data in
grounds for this view, think back to the cheese-in-the-larder terms of the fewest number of causes does seem sensible. But is
example above. Why do we regard the mouse hypothesis as a beter there any objective reason for thinking that such a theory is more
explanation of the data than the maid hypothesis? Presumably, likely to be true than a less simple theory? Philosophers of science
because we know that maids do not normally steal cheese, whereas do not agree on the answer to this diflicult question.
mice do. But this is knowledge that we have gained through
ordinary inductive reasoning, based on our previous observations of
the behaviour of mice and maids. So according to this view, when Probability and induction
we try to decide which of a group of competing hypotheses provides The concept ofprobabitity is philosophically puzzling. part ofthe
the best explanation of our data, we invariably appeal to knowledge puzzle is that the word'probability'seems to have more than one

33
 meaning. Ifyou read that the probabiiif of an Englishwoman living believe. If I say that the probability of finding life on Mars is high
to 1OO years of age is 1 in 1O, you would understand this as saying and you say that it is very low, neither of us is right or wrong - we
that one-tenth of ali Englishwomen live to the age of roo. Similarly, are both simply stating how strongly rve believe the statement in
if you read that the probability of a male smoker developing iung question. Ofcourse, there is an objective fact about whether there is
cancer is 1 in 4,, you would take this to mean that a quarter of all life on Mars or not; there is just no objective fact about how
male smokers develop lung cancer. This is known as the frequency probable it is that there is life on Mars, according to the subjective
interpretation of probability: it equates probabilities with interpretation.
proportions, or frequencies. But what ifyou read that the
probability of flnding life on Mars is 1 in 1,OOO? Does this mean The logical interpretation ofprobabiliq' rejects this position. It
that one out ofevery thousand planets in our solar system contains holds that a statement such as'the probabiliqr of life on Mars is
life? Clearly it does not. For one thing, there are only nine planets in high' is objectively true or false, relative to a specified body of
our solar system. So a different notion of probability must be at evidence. A statement's probabiliq' is the measure of the strength
work here. of evidence in its favour, on this view. Advocates ofthe logical
interpretation think that for any two statements in our language,
One interpretation of the statement'the probabilif of life on Mars we can in principle discover the probabiliq' of one, given the
@
is 1 in 1,O0O'is that the person who utters it is simply reporting a other as evidence. For example, we might want to discover the
subjective fact about themselves - they are telling us how likely they probability that there will be an ice age within to,ooO years, o

o think life on Mars is. This is the subjective interpretation of given the current rate ofglobal warming. The subjective fl
probability. It takes probability to be a measure of the strength of o
interpretation says there is no objective fact about this o
our personal opinions. Clearly, we hold some of our opinions more probabiiity. But the logical interpretation insists that there is: the
4 G
strongly than others. I am very confldent that Brazil will win the current rate of global warming confers a definite numerica]
World Cup, reasonably confident that Jesus Christ existed, and probabiiity on the occurrence of an ice age within IO,OOO years,
rather less coniident that global environmental disaster can be say O.g for example. A probability of o.g clearly counts as a high
averted. This could be expressed by saying that I assign a high probability - for the ma-ximum is I - so the statement'the
probabiliq' to the statement 'Brazil will win the World Cup', a fairly probabilif that there will be an ice age within 1o,ooo years is
high probability to 'Jesus Christ existed', and a low probability to high'would then be obiectively true, given the evidence about
'global environmental disaster can be averted'. Of course, to put an global warming.
exact number on the strength of my conviction in these statements
would be hard, but advocates ofthe subjective interpretation regard Ifyou have studied probability or statistics, you may be puzzled by
this as a merely practical limitation. In principle, we should be able this talk of different interpretations of probabiliSr. Hor,v do these
to assign a precise numerical probability to each of the statements interpretations tie in with what you learned? The ans\ rer is that the
about which we have an opinion, reflecting how strongly we believe mathematical study of probability does not by itself tell us what
or disbelieve them, they say. probabiliq' means, which is what we have been examining above.
Most statisticians would in fact favour the frequency interpretation,
The subjective interpretation of probability implies that there are but the problem of how to interpret probabiliq,, like most
no objective facts about probability, independently ofwhat people philosophical problems, cannot be resolved mathematicailv. The
 mathematical formulae for u,orking out probabilities remain the However, matters are not quite so simple. For we must ask what
same, whichever interpretation we adopt. interpretation of probability this response to Hume assumes. On
the frequency interpretation, to say it is highly probable that aii
Philosophers of science are interested in probability for tr,vo main objects obey Newton's law is to say that a very high proportion of
reasons. The first is that in many branches of science, especially all objects obey the law. But there is no way we can know that,
physics and biology, we find laws and theories that are formulated unless we use induction! For we have only examined a tiny fraction
using the notion of probability. Consider, for example, the theory of all the objects in the universe. So Hume's problem remains.
known as Mendelian genetics, which deals with the transmission Another way to see the point is this. We began with the inference
ofgenes from one generation to another in sexually reproducing from'all examined objects obey Newton's law'to'a11 objects obey
populations. One of the most important principles of Mendelian Newton's law'. In response to Hume's worry that the premiss of
genetics is that every gene in an organism has a 50% chance of this inference doesn't guarantee the truth ofthe conclusion, we
making it into any one of the organism's gametes (sperm or egg suggested that it might nonetheless make the conciusion highly
cells). Hence there is a 5o% chance that any gene found in your probable. But the inference from'all examined objects obey
mother will also be in you, and likewise for the genes in your Newton's law'to'it is highly probable that all objects obey
father. Using this principle and others, geneticists can provide Newton's larv'is still an inductive inference, given that the latter
o detailed explanations for why particular characteristics (e.g. eye means 'a very high propodion of all objects obey Newton's law', as
o
colour) are distributed across the generations ofa family in the it does according to the frequency interpretation. So appealing to o

o way that they are. Now 'chance' is just another word for the concept of probability does not take the sting out of Hume's
c probability, so it is obvious that our Mendelian principle makes argument, if we adopt a frequency interpretation of probability. o

essential use of the concept of probability. Many other examples For knowledge of probabilities then becomes itself dependent on
G
could be given of scientific laws and principles that are expressed induction.
in terms of probability. The need to understand these laws and
principles is an important motivation for the philosophical study of The subjective interpretation ofprobabiliry is also powerless to
probabili4'. solve Hume's problem, though for a different reason. Suppose John
believes that the sun will rise tomorrow and Jack believes it will not.
The second reason why philosophers ofscience are interested in the They both accept the evidence that the sun has risen every day in
concept of probabiliq' is the hope that it might shed some light on the past. Intuitively, we want to say that John is rational and Jack
inductive inference, in particular on Hume's problem; this shall be isn't, because the evidence makes John's beiief more probable. But if
our focus here. At the root of Hume's problem is the fact that the probability is simply a matter of subjective opinion, we cannot say
premisses ofan inductive inference do not guarantee the truth ofits this. All we can say is that John assigns a high probabiliq' to 'the sun
conclusion. But it is tempting to suggest that the premisses ofa wr'll rrse tomorrow'and Jack does not. I{there are no o b;'ective facts
about probability, then we cannot say that the conclusions of
[,pical inductive inference do make the conclusion highly probable.
Although the fact that all objects examined so far obey Newton's law inductive inferences are objectively probable. So we have no
of gravity doesn't prove that all objects do, surely it does make it explanation of why someone like Jack, who declines to use
very probable? So surely Hume's problem can be answered quite induction, is irrational. But Hume's problem is precisely the
easily after all? demand for such an explanation.

37
 The logical interpretation of probabiliq' holds more promise of a the interpretation of probability, and more. Like most philosophical
satisfactory response to Hume. Suppose there is an objective fact questions, these questions probably do not admit offinal answers,
about the probability that the sun will rise tomorrow, given that it but in grappling with them we learn much about the nature and
has risen every day in the past. Suppose this probabili[' is very limits of scientific knowledge.
high. Then we have an explanation ofwhy John is rational and
Jack isnt. For John and Jack both accept the evidence that the sun
has risen every day in the past, but Jack faiis to realize that this
evidence makes it highly probable that the sun will rise tomorrow
while John does realize this. Regarding a statement's probability
as a measure ofthe evidence in its favour, as the logical
interpretation recommends, tallies neatly with our intuitive
feeling that the premisses of an inductive inference can make
the conclusion highly probable, even ifthey cannot guarantee
its truth.

o Unsurprisingly, therefore, those philosophers who have tried to

o 6
solve Hume's problem via the concept of probability have tended to
favour the logical interpretation. (One of these was the famous
o
o
economist John Mayrrard Keynes, whose early interests were in o
0
logic and philosophy.) Unfortunately, most people today believe that
4 G
the logical interpretation ofprobability faces very serious, probably
insuperable, diffrculties. This is because all the attempts to work out
the logicai interpretation ofprobability in any detail have run up
against a host of problems, both mathematical and philosophical.
As a result, many philosophers today are inclined to reject outright
the underlying assumption ofthe logical interpretation - that there
are objective facts about the probabilily of one statement, given
another. Rejecting this assumption leads naturally to the subjective
interpretation ofprobabiliql but that, as we have seen, offers scant
hope of a satisfactory response to Hume.

Even if Hume's problem is ultimately insoluble, as seems likely,

thinking about the problem is still a valuable exercise. For reflecting
on the problem of induction leads us into a thicket of interesting
questions about the structure of scientific reasoning, the nature of
rationality, the appropriate degree of confidence to place in science,