You are on page 1of 3

Philosophy of Science Association

Safety, Strength, Simplicity


Author(s): Nelson Goodman
Source: Philosophy of Science, Vol. 28, No. 2 (Apr., 1961), pp. 150-151
Published by: The University of Chicago Press on behalf of the Philosophy of Science
Association
Stable URL: http://www.jstor.org/stable/185805
Accessed: 23/03/2010 14:22

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ucpress.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

The University of Chicago Press and Philosophy of Science Association are collaborating with JSTOR to
digitize, preserve and extend access to Philosophy of Science.

http://www.jstor.org
SAFETY, STRENGTH, SIMPLICITY

NELSON GOODMAN
University of Pennsylvania

When the evidence leaves us with a choice among hypotheses of unequal


strength, how is the choice to be made?
Caution would counsel us to choose the weakest, the hypothesis that asserts
the least, since it is the least likely to fail us later. But the principle of
maximum safety quickly reduces to absurdity; for it always dictates the choice
of a hypothesis that does not go beyond the evidence at all.
The very opposite proposal has been advanced by Popper': that the strongest
hypothesis not falsified by the evidence should be chosen. But for every
hypothesis strong enough to go beyond the evidence, there is an equally
strong conflicting hypothesis based upon the same evidence. This is easily
shown. Suppose our evidence tells us just that every examined A is a B,
and suppose hypothesis H1 affirms in addition that every (or even some one)
other A is a B. Then hypothesis H2, affirming that every examined A is a B
and that every (or the particular one) other A is not a B, likewise conforms
to the stated evidence. Hence strength is indifferent as between any projection
and its opposite. And to exclude every hypothesis that conflicts with another
equally strong one unviolated by the evidence would be, once more, to exclude
every hypothesis that goes beyond the evidence at all.
Thus although both safety and strength are desirable features of a theory,
they are by themselves incompetent criteria for choice. Another and controlling
factor, simplicity, must be taken into account. Simplicity has sometimes been
mistakenly identified with safety or with strength, but is readily shown to
be distinct from both. Suppose we have examined many and widely distributed
specimens of maple trees and found them all to be deciduous, and suppose
this constitutes our entire evidence. Since we still will not have examined
specimens from every small locality, our evidence may then leave us with
a choice between the following two hypotheses:
(1) All maples, except perhaps those in Eagleville, are deciduous.
(2) All maples are deciduous.
The second is clearly both the stronger and the simpler. We would incorporate
(2) in our theory, and retreat to (1) only if further evidence indicated that (2)
is false. Insertion of the ad hoc exceptive clause both weakens and complicates
the hypothesis. Cases like this, where the preferable hypothesis is the simpler
and stronger one, give plausibility to the view that strength is the measure
of simplicity and is the cardinal principle of choice. But exactly comparable
I In The Logic of Scientific Discovery (London, 1959-from the German of 1935), especially
Chapters VI and VII.
150
SAFETY, STRENGTH, SIMPLICITY 151

cases point to the opposite conclusion. The stated evidence also leaves un-
falsified the hypothesis:
(3) All maples whatsoever, and all sassafras trees in Eagleville, are
deciduous.
Now (3) is stronger than (2) but is less simple and acceptable. The expansion
made in (3) is as unwelcome as the exception made in (1). Hypothesis (2),
although it lies between (1) and (3) in safety and strength, is simpler than
and preferable to either.
This shows that neither safety nor strength is the measure of simplicity,
and that simplicity takes precedence over both as a factor in the choice of
hypotheses. The delicate problem of balancing safety and strength against
each other is significant only as between hypotheses of equal simplicity.
If neither safety nor strength determines simplicity, what does ? Formulation
of general standards for comparing the simplicity of hypotheses is a difficult
and neglected task. Here brevity is no reliable test; for since we can always,
by a calculated selection of vocabulary, translate any hypothesis into one of
minimal length, the simplicity of the vocabulary must also be appraised. I am
inclined to think that the standards of simplicity for hypotheses derive from
our classificatory habits as disclosed in our language, and that the relative
entrenchment of predicates underlies our judgment of relative simplicity;
but spelling this out takes some pains. Merely, to reject unfamiliar predicates
wholesale in favor of familiar ones would be to disallow the introduction
of needed new terms into scientific language. Furthermore, we must ordinarily
decide not merely which of two hypotheses is the simpler but which one
makes for the simpler total theory. I have discussed these matters in Fact,
Fiction, and Forecast2; and the criteria of projectibility I have outlined there
in terms of entrenchment is perhaps essentially a simplicity criterion. But
the whole matter wants more study. What is evident is that adequate canons
of induction must incorporate criteria of simplicity that cannot be given solely
in terms of strength or safety.

2
Harvard University Press, 1955; see especially Chapter IV.

You might also like