Professional Documents
Culture Documents
354
NEW BOOKS
(155 mid), but I do not think he deals with it. If his particular model of the
situation is to work, he must use something like the notion that any particular
observed set can be expected to have the characteristics possessed by the
majority of the group, and that we are very unlikely to come across an
exceptional set. He refuses to use the notion, which one might think appro-
priate here, of a random selection (133-4), ™r does he appeal to empirical
considerations, but is content to take the rejection rule as a satisfactory
account of the situation. More discussion seems to me to be desirable on
this matter.
The point he does discuss is different, and I am not happy about bis
treatment of it.
He says that the frequency statement "Less than k of all the pos°ible
selections show more than so and so per cent or less than so and so per cent
of males" is a probability statement, whose meaning is still to be explained
(155 top) and he goes on to explain it. But it is a frequency statement of an
entirely diSerent type from the statement that 51 per cent of births are male.
It refers to a perfectly definite class of selections (ioo1000 in number) and
contains no probability element whatever, any more than does the statement
that of the thirty people in this room, just six are males. It is purely arith-
metical and can be taken literally, whereas the statement about births
cannot. It is not a statement about our possible observations, and nothing
so far said in the text justifies passing from it to a probability statement
about our possible observations.
Having slipped into regarding it as a probability statement, however, he
then shows that it needs to be given meaning by a further rejection test,
which itself involves an arithmetical frequency statement of the same sort,
and by treating this frequency statement as a probability statement (without
any justification being given) proceeds to a new rejection test of the same
sort, and so on.
These tests, as he stresses (161 note) are not proposed as practical tests,
but as giving the meaning of the original probability statement. But after all
the tests, are we any nearer an answer than we were after the first test? To
explain the meaning of any probability statement, we have to pass to a test
involving a probability statement of exactly the same kind; and we never
come to an end. This rather bothers me, but I am much more troubled by
his treating the purely arithmetical frequency statement about alternative
possibilities which is the basis of each test as if it were in fact a probability
statement about possible observed cases.
Rejection was stressed for giving the meaning of universal statements on
the ground that no acceptance rule could reach finality. But for probability
statements rejection is equally provisional. Hence we should have expected
him to put provisional acceptance on the same footing with provisional
rejection in this case, and in consequence to have given rules for provisional
acceptance of universal statements at least some part in determining their
meaning. A statement, whether universal or probability, is sometimes
accepted: but if its meaning is given solely by the conditions for its rejection,
what is it we accept?
I have no space to deal with his discussions arising out of the fact that the
tests for rejection of probability statements depend on a decision, for which
there are no absolute logical grounds, as to what value of k to adopt in
particular cases, nor with his extremely interesting treatment in Chap. VII
of the reasons for preferring one statistical hypothesis to another, in which
he expounds the views of A. Wald.
The remaining chapters are concerned with the justification of induction,
355
PHILOSOPHY
•whether laws of nature state necessary connections, teleological explanation,
and the sense in which an explanation can be given of scientific laws them-
selves. In all these discussions his views are consistent with the views I have
already outlined about the meaning of scientific laws, and bring out their
implications.
There is one special feature of the book to which Professor Braithwaite
himself draws attention, viz. the devices of exposition used in it. "I have
tried," he says, "to bridge the gap between the awareness of all educated
persons of the general way in which science proceeds (by the hypothetico-
deductive method) and their vagueness or puzzlement about such scientific
fundamentals as the function of mathematics in science, the nature of
theoretical concepts and the principles of statistical inference by examining
in great detail simple examples specially constructed to display, without
irrelevances, the logically important features." (xi). Among these the
account of two different forms of a calculus, intended to bring out the
nature of scientific deductive systems (Chap. II) is noteworthy and extremely
effective, though I found the use made of them in Chapters III and IV, which
deal with the status of the theoretical terms in a science and with the use
of models, hard to follow. Many readers are likely to lose grip at page 58 and
to find the calculus rather a hindrance than a help during these two chapters. •
I found myself wondering whether a change of example, rather than a
complicating of the original one, would not have proved more enlightening.
The discussion of the use of models is important, and enables him to bring
out into clear relief his rejection of the view that scientific concepts and
theories are attempts at giving pictures of the real.
I noted a few misprints missed in proof correction, but none of any
importance. Something has gone wrong with the first sentence in the paragraph,
on page 330 mid.
This is a noteworthy addition to the small number of recent English books
on the Logic of Science.
L. J. RUSSELL. •