Goodman, Nelson (1906–98)

DOI: 10.4324/9780415249126-M045-1
Version: v1,  Published online: 1998
Retrieved December 04, 2021, from

3. Induction

Symbol systems are artefacts. Their syntactic and semantic features are not dictated by the domain but result from decisions about how the domain is to be organized, which can be done in an enormous number of divergent ways. Acceptability of a symbol system depends on its suitability for the purposes at hand.

In empirical science and elsewhere the purpose is often inductive – a matter of projecting from limited evidence to a wider class of cases. The traditional problem of induction is to say when and to what extent a limited evidence-class warrants such an inference. How many emeralds need be examined and from what sources need they be drawn before we are justified in inferring that all emeralds are green? In Fact, Fiction, and Forecast, Goodman shows that the problem runs deeper. It is a matter not just of the composition of the evidence class, but also of its characterization. He defines novel predicates ‘grue’ and ‘bleen’ as follows:

  • x is ‘grue’ = x is examined before future time t and is found to be green, or x is not so examined and is blue.

  • x is ‘bleen’ = x is examined before future time t and is found to be blue, or x is not so examined and is green.

All emeralds examined to date have been grue as well as green, for the extensions of the two predicates do not yet diverge. Yet we confidently expect emeralds found after t to be green, not grue. What, if anything, is our justification?

We can’t dismiss ‘grue’ as derivative from ‘green’, for ‘green’ can be defined in terms of ‘grue’ and ‘bleen’, just as ‘grue’ is defined in terms of ‘green’ and ‘blue’:

  • x is ‘green’ = x is examined before future time t and found to be grue, or x is not so examined and is bleen.

Which predicate is basic and which is derivative depends entirely on where we start. Nor can we dismiss ‘grue’ on the grounds that it makes essential reference to a specific time t. For whether ‘grue’ or ‘green’ makes reference to t again depends on which is taken as primitive. One might argue that ‘green’ marks a more natural kind than ‘grue’ does. Then it is, in some sense, essentially more primitive. But in the absence of an acceptable standard of naturalness that does not presuppose the very differences in projectibility we are trying to account for, this claim rings hollow. For we neither know what it means, nor how to tell whether one predicate is more natural than another.

The solution, Goodman maintains, lies in entrenchment. What favours ‘green’ over ‘grue’ is the brute fact that ‘green’ and its cognates have been successfully projected far more often than ‘grue’. The fact that up to now ‘grue’ would have worked as well is irrelevant. The decision favours the predicates that were actually successfully used.

Induction provides no guarantees. Goodman recognizes that we currently have no way of knowing whether future emeralds will be grue, green or something else entirely. The problem as he sees it is how to proceed in the absence of such knowledge. He argues that entrenched predicates are to be preferred, not because they have any lien on the future, but because they have served us well so far, and their continued use enables us to make efficient use of available cognitive resources and habits of thought. But the presumption in favour of entrenched predicates evaporates as soon as counterexamples emerge. When the first nongreen emerald is found, ‘All emeralds are green’ loses its claim on our epistemic allegiance.

The emphasis on entrenchment does not preclude innovation. Novel predicates can be projected when entrenched hypotheses are violated. Thus, for example, the Michelson–Morley experiment, by violating Newtonian generalizations, opened the way for the projection of novel, relativistic predicates. New predicates can also be introduced at interstices where no entrenched predicate prevails. A term like ‘quark’ can be introduced to denote phenomena that previously lacked a label. Such terms, Goodman maintains, derive their projectibility from related terms such as ‘subatomic particle’. Novel predicates thus become projectible by fitting into working inductive systems or into replacements for systems that do not work.

Citing this article:
Elgin, Catherine Z.. Induction. Goodman, Nelson (1906–98), 1998, doi:10.4324/9780415249126-M045-1. Routledge Encyclopedia of Philosophy, Taylor and Francis,
Copyright © 1998-2021 Routledge.

Related Articles