NELSON GOODMAN
283
at some of these that have to do with varieties of rightness other than
truth.
Validity of inductive inference, though a property of a relation
among statements, requires truth neither of premises nor of conclu–
sion; a valid inductive argument may even yield a false conclusion
from true premises . What, then, is required for inductive validity?
Certain formal relationships among the sentences in question
plus
what I shall call right categorization. Now a category or system of
categories-a way of sorting-is not sentential, is not true or false;
but use of wrong categories will make an induction invalid no matter
how true the conclusion. For example, if an emerald is said to be
grue just in case it is either examined before a given time and deter–
mined to be green or is not so examined and is blue, then the same
formal rules that lead from evidence statements about green
emeralds to the hypothesis "All emeralds are green" will also lead
from evidence-statements about "grue emeralds" to the hypothesis
"All emeralds are grue"; but the former inference is valid, the latter
not. For although the evidence-statements are true in both cases,
and the truth of both hypotheses is as yet undetermined, "grue" picks
out a category wrong in this context, a nonrelevant kind. Valid in–
duction runs within - is constrained by - right categories; and only
through distinguishing right categories from among classes in
general can we distinguish valid from invalid induction . But what
makes a category right? Very briefly, and oversimply, its adoption
in inductive practice , its entrenchment, resulting from inertia
modified by invention.
7
Why some categories rather than others
have become entrenched - a subject of avid philosophical debate–
does not matter here; the entrenchment, however achieved, provides
the required distinction. Rightness of categorization, in my view,
derives from rather than underlies entrenchment.
Inductive validity is not only an example of rightness other
than truth but is also one of the criteria applied in the search for
truth: a hypothesis validly inferred is favored over an alternative in–
validly inferred from the same evidence. Yet how is acceptability as
determined by such considerations related to truth, and does this
7. The matter is more complex than can be made clear here. Some outlines are of–
fered in my book
Fact, Fiction, and Forecast,
IV.
