Professional Documents
Culture Documents
Nelson Goodman: First Published Fri Nov 21, 2014
Nelson Goodman: First Published Fri Nov 21, 2014
1. Life
2. Anti-Absolutism
3.1 Nominalisms
3.2 Mereology
5. The Old and the New Riddle of Induction and their Solution
o
6.1 Irrealism
6.2 Worldmaking
Bibliography
o
A. Primary sources
B. Secondary sources
Academic Tools
Related Entries
1. Life
Henry Nelson Goodman was born on August 7, 1906, in Somerville, Massachusetts (USA), to Sarah
Elizabeth (Woodbury) Goodman and Henry L. Goodman. In the 1920s he enrolled at Harvard University
and studied under Clarence Irving Lewis (who later became his Ph.D. supervisor),Alfred North Whitehead,
Harry Scheffer, W.E. Hooking, and Ralph Barton Perry. Goodman graduated from Harvard in 1928. It took
him, however, 12 more years until he finished his Ph.D. in 1941 with A Study of Qualities (SQ). There are
several possible reasons for the lateness of his Ph.D. Maybe the most important was that Goodman was
Jewish, and therefore not eligible for a graduate fellowship at Harvard (Schwartz 1999; Elgin 2000a; Scholz
2005). He had to work outside the university to finance his studies. From 1928 until 1940, Goodman worked
as the director of theWalker-Goodman Art Gallery at Copley Square, Boston. This interest and activity in
the artworld is more frequently cited as a reason for the lateness of his Ph.D. During his graduate studies
Goodman was also a regular participant in W.V. Quine's seminars on the philosophy of the Vienna
Circle (in particular of Rudolf Carnap). Goodman also worked closely with Henry Leonard, who wrote his
Ph.D. at the same time under Alfred North Whitehead's supervision. After military service, Goodman taught
briefly as instructor in philosophy at Tufts College, and was then hired as associate professor (194651)
and later as full professor (195164) at the University of Pennsylvania. He served briefly as Harry Austryn
Wolfson Professor of Philosophy at Brandeis University (196467), finally returning to Harvard in 1968,
where he taught philosophy until 1977. At Harvard, he founded Project-Zero, a center to study and improve
education in the arts. Besides being an art gallery director as a graduate student, and private art collector
throughout his life, Goodman was also involved in the production of three multimedia-performance
events, Hockey Seen: A Nightmare in Three Periods and Sudden Death (1972), Rabbit, Run (1973),
and Variations: An Illustrated Lecture Concert (1985) (Carter 2000, 2009).
Goodman was more interested in solving philosophical problems than in his celebrity as a philosopher. He
authorized only two interviews (Goodman 1980, 2005), did not write an autobiography, and rejected the
invitation to be honored with a volume in the prestigious Schilpp Library of Living Philosophers (Elgin
2000a: 2). Sparse bits of information about his personal life can only be gathered from the autobiographies
of his contemporaries and their published correspondences (e.g., Quine 1985; Creath 1990) or his obituaries
(e.g., Carter 2000; Elgin 1999 (Other Internet Resources), 2000a, 2000b; Elgin et al. 1999; Mitchell 1999;
Scheffler 2001; Scholz 2005; Schwartz 1999). Goodman died on November 25, 1998, in Needham,
Massachusetts, at the age of 92, after a stroke.
2. Anti-Absolutism
Nelson Goodman's philosophy synthesizes German/Austrian Logical Empiricism, as developed and
practiced by philosophers like Rudolf Carnap and Carl Hempel, with American Pragmatism of the kind
practiced and advocated by C.I. Lewis. Goodman, however, departs from both traditions considerably. As
we will see, he departs from Lewis' pragmatism in dismissing the idea of an indubitable given in experience.
He departs from logical empiricism in giving up a principled analytic/synthetic distinction.
The received view, that Goodman's main work, The Structure of Appearancewas intended as an antifoundationalist reconception of Carnap's Der logische Aufbau der Welt (cf. Elgin 2001; Hellman 1977) is
particularly misleading here.
In fact, Goodman was quite aware that Carnap's work was itself anti-foundationalist in the same respect as
his. Already in his dissertation thesis A Study of Qualities (which was later developed into The Structure of
Appearance), Goodman writes:
[] Carnap has made it clear that what we take as ground elements [for a constitutional system] is a matter
of choice. They are not dignified as the atomic units from which others must be built; they simply constitute
one possible starting point. [] In choosing erlebs, Carnap is plainly seeking to approximate as closely as
possible what he regards the original epistemological state [] Yet whether it does so or not is no test of
the system. [] Hence [] argument concerning whether the elements selected are really primitive in
knowledge is extraneous to the major purpose of the system. (SQ: 9698)
The quote makes it obvious that Goodman himself did not consider his constructionalist approach inA Study
of Qualities as an epistemological alternative to Carnap's. Insofar as criticism of a foundationalist
epistemology does play a role in The Structure of Appearance or A Study of Qualities, this criticism was
rather directed at the philosophy of C.I. Lewis, who was Goodman's teacher at Harvard. Lewis indeed held
the view that empiricism must presuppose the incorrigibility and indubitability of what is given in
experience. According to Lewis, I might need to revise, for example, that I saw a plane crossing the sky
when I learn that what I mistook for a plane was Superman. However, nothing can make me revise that
there was a blue and a red spot in the center of my visual field that then led to the (false) belief that there
was a plane.
A Study of Qualities, on the other hand, begins with the argument that even the simplest judgments of this
sortas the one about a blue and a red spot in the center of my visual fieldmight be revised in the light
of new evidence. My judgment that I had a blue spot in the middle of my visual field a few seconds ago
when I looked at a ripe apple under normal conditions might be revised when I now judge that I have a red
spot in my visual field, looking at the same object under the same conditions and know that it could not
have changed its color. However, if such revisions can be made in retrospect, nothing of the given is
indubitable or incorrigible. Judgments about qualia, in this sense, are decrees; which judgments are
accepted is a matter of the overall coherence of my system of beliefs and my other qualia judgments.
The literal unverifiabilty of such quale-recognition is, nevertheless, in the last analysis beyond question. If
I say the green presented by that grass now is the same as the green presented by it at a certain past moment,
I cannot truly verify that statement because I cannot revive that past moment. The statement therefore
constitutes an arbitrary and supreme decree. But a decree, simply because it is arbitrary, is not therefore
necessarily haphazard. My quale-identifications are influenced; I do not feel equally inclined to identify the
color presented by the grass now with the color presented by a cherry a moment ago, though such a decree
if made would be equally supreme and unchallengeable on strict grounds. We are all much in the same
position of absolute but sane monarchs; our pronouncements are law, but we use our heads in making them.
(SQ: 17; cmp. SA (2nd ed.): 134)
Also in this respect Goodman is following Carnap and the logical empiricists. C.I. Lewis emphasizes this
in his Logical Positivism and Pragmatism (Lewis 1941). There he explains that the main difference
between the empiricism of the pragmatists and the empiricism of the logical positivists (especially the
Carnap of Philosophy and Logical Syntax (1935)) is that the latter were ready to analyze empirical
knowledge fully in the so-called formal mode. Accordingly, they analyze empirical knowledge as more
or less coherent systems of accepted sentences, some of which are protocols, some are sentences of
mathematics and logic, some are generalizations, etc. In particular, the formal mode would not distinguish
between statements such as This object looks red and This object is red.
For Lewis, this sort of empiricism is not worthy of the name. After all, the experiential element does not
seem to show up at all in this kind of formal analysis. Lewis claims instead that a proper empiricism must
treat sentences of the form This looks red. as special, indubitable statements. We might err when
classifying things as being red, but we cannot err when it comes to recognizing things as looking red. This
is the given in experience, the phenomenal states we find ourselves in when making experiences. Without
such an indubitable element, Lewis fears that our epistemology would necessarily collapse into a coherence
theory of truth (Lewis 1952). Goodman, on the other hand, is ready to bite that bullet when throwing away
the indubitable given. Lewis, the major advocate of pragmatism, comments on this move by Goodman that
his proposal is, I fear, a little more pragmatic than I dare to be (Lewis 1952: 118).
Indeed, Goodman's early and later philosophy is anti-foundationalist. This is truly a characteristic of his
work on induction, metaphysics, logic and even the languages of art. It should, however, not be interpreted
as a counter program to logical positivism. What Goodman didin all these areasis better understood as
a continuation and enlargement of Carnap's program. This is obvious if we consider Goodman's relativism
and irrealism. It is also apparent when we think about his pluralism in logic and his insistence that there are
more cognitively valuable representation systems than just the sciences, namely the languages of art.
His anti-foundationalism therefore is more than just a restatement that there is no bedrock for
knowledgeas was argued by Karl Popper and Otto Neurath, but also that there are no fundamental
ontological objects, that there are no fundamental logical principles, and that there are no privileged
representation systems. All of these echo Rudolf Carnap's famous Principles of Tolerance (Carnap 1934):
tolerance with regards to ontology, to logical principles, and to representation systems in general.
I don't understand what confirmation is, or let us say projectibility, in the sense that I can't frame any
adequate definition; but give me any predicate and (usually) I can tell you whether it is projectible or not. I
understand the term in extension. But analytic I don't even understand this far; give me a sentence and I
can't tell you whether it is analytic because I haven't even implicit criteria. I can't look for a definition
when I don't know what it is I am defining. (Goodman in a letter to Quine and White, 2 July 1947, in White
1999: 347)
Goodman's remark is instructive, since it undermines a move that Grice and Strawson would later make
against Quine's argumentation in Two Dogma's of Empiricism. In their In Defense of a Dogma (Grice
and Strawson 1956) they argue that Quine's skepticism about the analytic-synthetic distinction as such is
undermotivated in light of our pre-theoretic grasp of the distinction. Goodman's claim is that there in fact
is no such pre-theoretic grasp of the distinction.
The official result of the exchange between White, Goodman, and Quine was that any sharp analytic
synthetic distinction is untenable and should just be abandoned:
I think that the problem is clear, and that all considerations point to the need for dropping the myth of a
sharp distinction between essential and accidental predication (to use the language of the older
Aristotelians) as well as its contemporary formulationthe sharp distinction between analytic and
synthetic. (White 1950: 330)
Goodman's view on the matter had already appeared in print in 1949 under the title On Likeness of
Meaning. In this paper Goodman proposes a purely extensional analysis of meaning, the upshot of which
is that no two different expressions in a language are synonymous. He discusses several objections to
theories of meaning that rely on intensional entities (such as, for example, Fregean senses) to explicate the
notion of synonymy in non-circular ways, such that the question of whether two terms are synonymous
is comprehensible as well as scrutable. Goodman eventually rejects intensional approaches and opts for an
extensional theory for sameness of meaning. According to such an extensional theory, two expressions have
the same meaning if and only if they have the same extension. This criterion is certainly intelligible, but
also scrutable; we can
decide by induction, conjecture, or other means that two predicates have the same extension without
knowing exactly all the things they apply to. (PP: 225)
But an extensional theory is, of course, not thereby free of problems. Consider, for example, the expression
unicorn and centaur, which have the same extension (namely the null-extension) but differ in meaning.
Hence, whereas sameness of extension is a necessary condition for sameness of meaning, sameness of
extension does not seem to be sufficient for sameness of meaning. Goodman proposes an extensional fix to
this problem that gives necessary and sufficient conditions for sameness of meaning. He observes that
although unicorn and centaur have the same extension, simply because of the trivial fact that they
denote nothing, centaur-picture and unicorn-picture do have different extensions. Clearly, not all
centaur-pictures are unicorn-pictures and vice versa. Thus the flight to compounds makes an extensional
criterion possible:
[I]f we call the extension of a predicate by itself its primary extension, and the extension of its compounds
as secondary extension, the thesis is as follows: Two terms have the same meaning iff they have the same
primary and secondary extensions. (PP: 227)
The primary extensions of unicorn and centaur are the same (the null-extension), but their secondary
extensions do differ: the compounds unicorn-picture and centaur-picture differ in extension.
If we allow all kinds of compounds equally, we arrive immediately at the result that by our new criterion
no two different expressions have the same meaning. Consider the expressions bachelor and unmarried
man: is a bachelor but not an unmarried man is a bachelor description that is not an unmarried-man
description. Hence, by Goodman's criterion, the secondary extensions of bachelor and unmarried man
differ because the primary extensions of at least one of their compounds do. Since the same trick can be
pulled with any two expressions, Goodman is left with the result that no two different expressions are
synonymous, but he is ready to bite this bullet. P-descriptions that are not Q-descriptions are easy to
construct for any P and Q (provided these are different terms) and these constructions might well be
relatively uninteresting. If only such uninteresting constructs are available to make a difference in
secondary extension, P and Q, despite being not strictly synonymous, might be more synonymous than a
pair of predicates for which we are able to find interesting compounds (as in the case of centaur and
unicorn). This turns sameness of meaning of different terms into likeness of meaning, and synonymy and
analyticity into a matter of degree.
perhaps by adopting a plural interpretation of second-order logic (Boolos 1984, 1985), or a proof-theoretic
semantics, or in any other wayFrege's original definition (which is not formulated in set theory, but in
his version of second-order logic) can be employed (Rossberg and Cohnitz 2009).
Even though these two particularly pressing gaps appear to be capable of being closed, a general recipe for
recasting platonist statements appears out of reach, in particular, when we consider statements of pure
mathematics itself. Without such a nominalist recasting, Goodman and Quine hold, platonist mathematical
statements cannot be deemed intelligible from a strictly nominalist perspective. The question becomes,
according to Goodman and Quine,
how, if we regard the sentences of mathematics merely as strings of marks without meaning, we can account
for the fact that mathematicians can proceed with such remarkable agreement as to methods and results.
Our answer is that such intelligibility as mathematics possesses derives from the syntactical or
metamathematical rules governing those marks. (Goodman and Quine 1947: 111)
Goodman and Quine construct a theory of syntax for the set-theoretic language and a proof theory based
on the Calculus of Individuals (see section 3.2 below) supplemented with a token-concatenation theory.
The tokens in question are concrete, particular inscriptions of the logical symbols, variable-letters,
parentheses, and the (for set-membership) that are use to formulate the language of set theory. Primitive
predicates are introduced to categorize the different primitive symbols: all concrete, particular inscriptions, for instance, fall under the predicate Ep. Concrete complex formulae, e.g., xy, are
concatenations of concrete primitive symbolsin our case the concatenation of x and and y. Bit
by bit, Goodman and Quine define their way up to which concrete inscriptions count as correctly formed
sentences of the language of set theory, and finally which concrete inscriptions count as proofs and
theorems. Goodman and Quine argue that in this way the nominalist can explain the remarkable
agreement of mathematicians mentioned above.
Since Quine and Goodman not only impose nominalistic strictures, but also finitism in their joint article
(Quine and Goodman 1947: 2), the syntactic and proof-theoretic notions defined still fall short of the usual
platonist counterparts. Even if any given sentence or proof is finite in length, the platonist would hold that
there are sentences and proofs of any finite length, and thus sentences and proofs that are too long to have
a concrete inscription in a given finite universe. Moreover, there are infinitely many (and indeed
uncountably many) truths of mathematics, butin particular, in a finite universethere will only ever be
finitely many inscriptions of theorems. Even if the universe is in fact infinite, perhaps a theory of syntax
and proof should not make itself hostage to this circumstance.
Platonists and nominalists will likely disagree whether Goodman and Quine successfully argue their case
in their joint paper. Goodman and Quine will be able to account for any actual mathematical proof and any
theorem actually proven, since there are at any stage only finitely many of them, each of which is small
enough to fit in our universe comfortably. Thus, arguably, they reach their goal of explaining the agreement
in mathematical practice without presupposing mathematical platonism. Due to its finitistic nature,
however, the account radically falls short of giving explications that are extensionally equivalent to the
platonists' conceptions (see Rossberg and Cohnitz 2009 for discussion and a landscape of possible
solutions). Goodman later (1956) explains that nominalism is not incompatible with the rejection of
finitism; it is
at most incongruous []. The nominalist is unlikely to be a non-finitist only in much the way a bricklayer
is unlikely to be a ballet dancer. (PP: 166; on the question of finitism see also MM: 53; Field 1980; Hellman
2001; Mancosu 2005)
Given the ardent pronouncements in the 1947 article with Quine, the common misunderstanding that
Goodman's mature nominalism encompasses, or is motivated by, the rejection of abstract objects is
understandable. Nonetheless, it is incorrect. Goodman does not reject all abstract objects: in The Structure
of Appearance, he embraces qualia as abstract objects (see section 4 below), some of which (in fact all but
moments) are universals (SA: VII.8). Goodman's mature nominalism, from The Structure of
Appearance onwards, is a rejection of the use of sets (and objects constructed from them) in constructional
systems, and no blanket rejection of all universals or abstract particulars. To be sure, Goodman also refuses
to acknowledge properties and other non-extensional objects, but the reason for his rejection of such entities
is independent, and in fact more fundamental, than his nominalism: it is his strict requirement of
extensionality (WW: 95n3; see alsosection 6 below). Goodman does occasionally include extensionalism
in his nominalism (see LA: xiii, 74; under the entry nominalism the index of LA references some passage
which discusses properties ; see also MM: 51; WW: 10n14). Strictly speaking, however, nominalism for
Goodman is the refusal to use class terms in a constructional systemno more, and no less.
Goodman presents two positive considerations for the rejection of a set-theoretic language (not counting
the remarks in Goodman and Quine 1947: 105). Methodologically, nominalistic constructions have the
advantage that they do not use any resources that the platonist could not accept (Goodman 1958; PP: 171).
The advantage of a nominalistic construction is thus one of parsimony:
As originally presented in A Study of Qualities [] the system was not nominalistic. I feel that the recasting
to meet nominalistic demands has resulted not only in a sparser ontology but also in a considerable gain in
simplicity and clarity. Moreover, anyone who dislikes the change may be assured that the process of
replatonizing the systemunlike the converse processis obvious and automatic; and this in itself is an
advantage of a nominalistic formulation. (SA: Original Introduction, page L of the 3rd. ed.; regarding the
simplicity remark see SA: III.7)
All resources employed by the nominalist are (or should be) acceptable to the platonist, while the converse
may not be the case (see also Goodman 1956: 31 (PP: 171); MM: 50).
By the time he writes The Structure of Appearance, Goodman has come around to a different criterion for
whether or not a system obeys nominalistic structures: the predicates present in the whole system (SA:
II.3). This is as opposed to merely considering the basis of the system in answer to this question as he does
in A Study of Qualities (as mentioned above). In Goodman 1958 (see also SA: III.7), he suggests a
different, perhaps more precise, way to characterize nominalistic systems in terms of the
system's generating relation:
System S is nominalistic iff S does not generate more than one entity from exactly the sameatoms of S.
Goodman describes the criterion as demanding that sameness of content entails identity. Systems that
have only mereological means of generating composite objects (see section 3.2 Mereologybelow) count
as nominalistic according to this criterion. Parthood is transitive, so from atoms a andb only one further
object can be generated, the mereological sum of a and b. A set-forming operation, however, will
distinguish, for instance, between {a, b} (the set of a and b) and {{a, b}} (the set containing the set
of a and b) and {{a}, {b}} (the set containing the singleton set of a and the singleton set of b). None of the
these three are pairwise identical. Membership is not transitive. The first and third contain two members,
but not the same members (both a and b are members of the first set, but not of the third), while the second
set has only one member (namely the first set). All three (and infinitely many others) are generated from
the same atoms, however, or as Goodman might put it, they have the same content: a and b. A system
featuring a set-theoretic generating relation thus does not count as nominalistic.
The sameness-of-content criterion was criticized by David Lewis (1991: 40) as question-begging. Lewis
suggests that the only alternatives for generating relations that Goodman allows are mereological, settheoretical, or a combination of the two, and that only mereological generation passes the test. Unless one
rejects set theory already, Lewis contends, one would not find the criterion plausible. There are, however,
non-extensional mereological systems that violate the sameness-of-content criterion as well (see entry
on mereology). Moreover, the sameness-of-content criterion may be understood as a version of Ockham's
Razor, demanding not to multiply entities beyond necessity.
3.2 Mereology
The Polish logician Stanisaw Leniewski (18861939) must surely count as the father of mereologythe
theory of parts and wholesbut around 1930, Goodman re-invents the theory together with his fellow
graduate student Henry S. Leonard (19051967). Only in 1935 do Goodman and Leonard learn of
Leniewski's work through one of their fellow students, W.V. Quine (Quine 1985: 122). An early version
of Leonard and Goodman's system is contained in Leonard's Ph.D. thesis, Singular Terms (Leonard 1930).
In 1936, Leonard and Goodman present their mature system at a meeting of the Association of Symbolic
Logic; the corresponding paper is published four years later under the title The Calculus of Individuals
and Its Uses (Leonard and Goodman 1940). Subsequently, Goodman uses the calculus in his own Ph.D.
thesis, A Study of Qualities (SQ), and a version of it inThe Structure of Appearance (SA). Little is known
about the nature of Goodman and Leonard's cooperation on the calculus. Goodman attributes the first
thought for the collaborative project to Leonard (PP: 149). Leonard, more concretely, suggests in a (still)
unpublished note:
If responsibilities can be divided in a collaborative enterprise, I believe that it may be fairly stated that the
major responsibility for the formal calculus [] was mine, while the major responsibility for discussions
of applications [] lay with Goodman. (Leonard 1967)
Quine only mentions that he himself was able to help them on a technical problem (Quine 1985: 122).
Leonard's system of Singular Terms is significantly different from, and indeed in philosophically interesting
ways weaker than, the Calculus of Individuals (Rossberg 2009), but the exact extent of Goodman's technical
contribution to the calculus remains unknown.
Perhaps surprisingly, nominalistic scruples were not the driving force behind the development the Calculus
of Individuals. Instead, their goal is to solve a technical problem in Carnap's Aufbau (1928) (see section
4 below), and to this end they employ both set-theoretic and mereological notions. Leonard, in his Ph.D.
thesis (supervised by Alfred North Whitehead), presents his calculus as an interpolation in Whitehead and
Russell's Principia Mathematica between *14 and *20 (Leonard 1967), and makes liberal use of classterms in the formulation (Leonard 1930). Also the joint paper of Leonard and Goodman is formulated using
class terms, as is the system Goodman used in his own Ph.D. thesis, A Study of Qualities (1941, SQ). It is
not until his joint article with Quine (Goodman and Quine 1947) and his Structure of Appearance (1951,
SA) that Goodman eschews the use of set theory to formulate the Calculus of Individuals.
As mentioned above, parthood, as opposed to the set-theoretic notion of membership, is transitive: if a is a
part of b and b is a part of c, then a is a part of c. Neither the system Leonard and Goodman present in their
1940 article, nor the version in Goodman's A Study of Qualities, nor the calculus he uses in The Structure
of Appearance, take part as primitive. Rather, it is in all three cases defined based on the sole primitive
notion adopted: overlap in SA, and discreteness in the other two systems. Overlap can pre-systematically
be understood as sharing a part in common; discreteness as sharing no part in common. All three systems
indeed define parthood so that these two pre-systematic understandings come out as theorems.
The Calculus of Individuals in all its formulations contains principles of mereological summation and
mereological fusion. Mereological summation is a binary function of individuals, so that the sum s of two
individuals a and b is such that both a and b, and all their parts, are parts of sand also all sums of parts
of a and b and are parts of s. Mereological fusion is a generalization of the mereological summation. In
Leonard and Goodman 1940 fusion is defined using sets: all the members of a set are fused in the sense
that they, and all their parts, and all the fusions amongst their parts, end up being parts of the individual
that is the fusion of set .
The technical details of the different versions of the Calculus of Individuals can be found in this
supplementary document: The Calculus of Individuals in its different versions (see also entry
onmereology).
Unrestricted mereological fusion has been widely criticized as too permissive. It allows for so-called
scattered objects (e.g., the sum of the Eiffel Tower and the Moon) and in the case of Goodman's construction
in The Structure of Appearance for sums of radically different kinds of objects, like sounds and colors.
W.V. Quine, after endorsing this principle in a joint paper with Goodman (Goodman and Quine 1947),
becomes one of its first critics in his review of The Structure of Appearance:
part, clear initially as a spatio-temporal concept, is here understood only by spatio-temporal analogy. []
When finally we proceed to sums of heterogeneous qualia, say a color and two sounds and a position and
a moment, the analogy tries the imagination. (Quine 1951b: 559)
Goodman (1956) maintains that the criticism is disingenuous if put forward by a platonist: set-theoretic
composition is as least as permissive as mereological fusion. Whenever there is a fusion of scattered
concrete objects, there is also a set of them (see Simons 1987 or van Inwagen 1990 for prominent criticisms
of unrestricted composition).
within the sets to serve as an explicatum for point as geometry demands. This rather pragmatic criterion
of adequacy for philosophical analyses serves Goodman well for a number of reasons. The most important
is perhaps that Goodman does not believe there was any such thing as identity in meaning of two different
expressions (see our discussion of likeness of meaning above). Thus, if synonymy was made the criterion
of adequacy, no analysis could ever satisfy it. But relaxing the criteria of adequacy for philosophical
analyses to structure-preservation also supports Goodman's more radical theses in epistemology and
metaphysics, especially in his later philosophy. One of his reasons to replace the notion of truth with that
of rightness of symbolic function, the notion of certainty with that of adoption, and the notion of
knowledge with that of understanding is the thought that the new system of concepts preserves the structural
relations of the old without preserving the philosophical puzzles related to truth, certainty and knowledge
(RP: chapter X).
However, we do not yet know that there are such things as colors. In fact the only thing we know about the
erlebs (AF) is that they are part similar as displayed in the following graph (where part-similarity between
erlebs is indicated by a line):
If we take the graph and now group exactly those erlebs together that are mutually part-similar, we get the
following sets:
{A,B,C,F},{A,E,F},{C,D,F}
But, of course, these sets correspond exactly to the extensions of the colors in our example (viz. black,
green, and red). Thus, by knowing about the part-similarity between erlebs alone, we seem to be able to
reconstruct their properties with the method of quasi-analysis.
Goodman observes, however, that in unfavorable circumstances quasi-analysis will lead to the wrong
results. Consider the following situation:
If we use Carnap's rule for quasi-analysis, we will obtain all color classes except {A,E,F}, the color class
for green, because green only occurs in constant companionship with the color black. Goodman calls
this the constant companionship difficulty.
A second problem can be illustrated with the following example of erlebs:
But here {A,C,F} should be a color class resulting from quasi-analysis, although A, C, and Fhave in fact
no color in common. Goodman calls this problem the difficulty of imperfect community. It is
controversial to what extent these problems are devastating for Carnap's project, but Goodman considered
them to be serious.
In order to see the generality of this argument, we have to note that the same problem also arises
for deduction (FFF: III.2). That deduction is in the same predicament is observed by Goodman and
exploited for his solution of Hume's problem of induction. Thus the upshot of Goodman's understanding of
Hume's argument is that there can be no justification of our inferential practices, if such a justification
requires a reason for their justifiedness. Accordingly, the old problem of induction, which requires such a
justification of induction, is a pseudo-problem.
Obviously, it does not. But what makes the difference? Both regularity statements (L1 and L2) are built
according to the exact same syntactical procedure from the evidence statements. Therefore, it does not seem
to be for a syntactical reason that B1 confirms L1 but B2 fails to confirm L2. Rather, the reason is that
statements like L1 are lawlike, whereas statements like L2 at best express accidentally true generalizations.
Lawlike statements, in contrast to accidentally true general statements, are confirmed by their instances and
support counterfactuals. L1 supports the counterfactual claim that if this thing I have in my hand were a
piece of copper, it would conduct electricity. In contrast, supposing that it is indeed true, L2 would not
support that if an arbitrary man were here in the room, he would be a third son. To tell which statements
are lawlike and which statements are not is therefore of great importance in the philosophy of science. A
satisfactory account of induction (or corroboration -- see the discussion of Popper in the entry on the
problem of induction) as well as explanation and prediction requires this distinction. Goodman, however,
shows that this is extremely hard to get.
Here comes the riddle. Suppose that your research is in gemology. Your special interest lies in the color
properties of certain gemstones, in particular, emeralds. All emeralds you have examined before a certain
time t were green (your notebook is full of evidence statements of the form Emerald x found at
place y date z(zt) is green). It seems that, at t, this supports the hypothesis that all emeralds are green
(L3).
Now Goodman introduces the predicate grue. This predicate applies to all things examined before some
future time t just in case they are green but to other things (observed at or after t) just in case they are blue:
(DEF1)x is grue =df x is examined before t and green x is not so examined and blue
Until t it is obviously the case that for each statement in your notebook, there is a parallel statement asserting
that the emerald x found at place y date z(zt) is grue. Each of these statements is analytically equivalent
with the corresponding one in your notebook. All these grue-evidence statements taken together confirm
the hypothesis that all emeralds are grue (L4), and they confirm this hypothesis to the exact same degree as
the green-evidence statements confirmed the hypothesis that all emeralds are green. But if that is the case,
then the following two predictions are also confirmed to the same degree:
However, to be a grue emerald examined after t is not to be a green emerald. An emerald first examined
after t is grue iff it is blue. We have two mutually incompatible predictions, both confirmed to the same
degree by the past evidence. We could obviously define infinitely many grue-like predicates that would all
lead to new, similarly incompatible predictions.
The immediate lesson is that we cannot use all kinds of weird predicates to formulate hypotheses or to
classify our evidence. Some predicates (which are the ones like green) can be used for this; other
predicates (the ones like grue) must be excluded, if induction is supposed to make any sense. This already
is an interesting result. For valid inductive inferences the choice of predicates matters.
It is not just that we lack justification for accepting a general hypothesis as true only on the basis of positive
instances and lack of counterinstances (which was the old problem), or to define what rule we are using
when accepting a general hypothesis as true on these grounds (which was the problem after Hume). The
problem is to explain why some general statements (such as L3) are confirmed by their instances, whereas
others (such as L4) are not. Again, this is a matter of the lawlikeness of L3 in contrast to L4, but how are
we supposed to tell the lawlike regularities from the illegitimate generalizations?
An immediate reply is that the illegitimate generalization L4 involves a temporal restriction, just as L2 was
restricted spatially (see e.g., Carnap 1947). The idea would be that predicates that cannot be used for
induction are analytically positional, i.e., their definitions refer to individual constants (for places or
times). A projectible predicate, i.e., a predicate that can be used for induction, has no definition which
would refer to such individual constants but is purely qualitative (e.g., because it is a basic predicate). The
trouble is that this reply makes it relative to a language whether or not a predicate is projectible. If we begin
with a language containing the basic predicates green and blue (as in English), grue and bleen are
positional. Bleen is defined as follows:
(DEF2)x is bleen =df x is examined before t and blue x is not so examined and green
But if we start with a language that has bleen and grue as basic predicates, green and blue are
positional:
(DEF3)x is green =df x is examined before t and grue x is not so examined and bleen
(DEF4)x is blue =df x is examined before t and bleen x is not so examined and grue
Both languages are symmetrical in all their semantic and syntactical properties. So the positionality of
predicates is not invariant with respect to linguistically equivalent transformations. But if this is the case,
there is no semantic or syntactic criterion on whose basis we could draw the line between projectible
predicates and predicates that we cannot use for induction.
Two lines of argument can be separated in Goodman's writings (Dudau 2002). First, Goodman argues that
there are conflicting statements that cannot be accommodated in a single world version: some truths conflict
(WW: 10916; MM: 3044). If that is the case, we need many worlds, if any, to accommodate the
conflicting versions and bring them in unison with the standard correspondence account of truth, that is,
that the truth of a statement is its being in correspondence with a world. The second line of argument seems
to be that we need no worlds at all if we need many. If we need a world for each version, why postulate the
worlds over and above the versions?
Let us first have a closer look at the first line of reasoning. The Earth stands still, revolves around the Sun,
and runs many other courses as well at the same time. Yet nothing moves while at rest. As Goodman
concedes, the natural response to this is that the sentences
However, according to Goodman, this would be wrong-headed (WW: 112). Consider the following two
historiographical sentences: The kings of Sparta had two votes and The kings of Sparta had only one
vote. The first sentence is part of a report by Herodotus, the second part of a report by Thucydides. There
is, again, an inclination to understand these sentences as ellipses for According to Herodotus, the kings of
Sparta had two votes, and According to Thucydides, the kings of Sparta had only one vote. But
obviously these latter two sentences do not tell us anything about Sparta. They only tell us about what
Herodotus and Thucydides said about Sparta. It is true that according to Herodotus, the kings of Sparta
had two votes, even if they actually had no vote or had three votes. The same holds for the relativizations
to the geocentric and the heliocentric systems: it is true that the Earth is at rest according to the geocentric
system, but that still does not inform us about the world. Thus, if we assume that (S1) and (S2) are both
true, we end up with a contradiction if we take them to be literally true of one and the same world. If we do
not take them to be literally true, but to be elliptical and implicitly relativized, we end up with two truths
that are not about any world. At least, they are not about the parts of the world we were interested in. They
turn out to be truths about versions, but not truths about planets. The solution chosen by Goodman is to
claim that they are about two different worlds. Both state a literal truth about a world, but just not about the
same world.
It is crucial for Goodman's argument that in the conflict between (S1) and (S2) we have (a) an actual conflict
between statements, and (b) no other way to resolve that conflict (like, for example, rejecting at least one
of the two statements in a non-arbitrary way). Of course, other contemporaries of Goodman, such as Quine
and Carnap, also consider the problem that experience alone might underdetermine theory-choice, but
believed that pragmatic criteria will in the long run allow us to arrive at one all-encompassing coherent
version of the world. In Quine's (Quine 1981) and Carnap's (Carnap 1932) philosophies, this is assumed to
be a physical version. But Goodman does not believe in physicalist reductivism. First of all, there does not
currently seem to be any convincing evidence that all truths are reducible to physics (just consider the
problem of reducing mental truths to physical truths), and, secondly, physics itself does not even appear to
form a coherent system (WW: 5). Hence, for Goodman, we are stuck with conflicting world versions we
consider true. Since, as we have seen above, relativism is no option for Goodmanbecause it would make
true statements true about versions onlywe arrive at Goodman's pluralism: conflicting true versions
correspond to different worlds.
The second line of reasoning in Goodman's writings toys with the idea that there are no worlds that the right
versions answer toor at the very least that such worlds are not necessary. The world versions suffice, and
are really the only things that are directly accessible, anyway. The versions can for many purposes be treated
as worlds (WW: 4 and 96; cmp. MM: 3033).
Goodman of course recognizes a difference between versions and worlds. A version can be in English and
comprise words. Worlds are neither in English nor comprise words. However, for a version that is true of
a world, the world has to answer to it in a way. A world that corresponds to (S1), for example, is a world
with planets and spacetime that is so arranged that one of the planets, Earth, is at rest in it. But planet,
spacetime, at rest, and so on are ways to categorize reality that depend on a version. These predicates
are the ones chosen in this very version. There was no world that came ordered in correspondence with
these predicates before this version was constructed. Instead, the world corresponds to the version expressed
by (S1), because the world with that structure was made, when that version was made.
But what are worlds made of? Should we not at least assume Reality to be some kind of stuff that allows
structuring with alternative versions as dough allows structuring with a cookie-cutter? Does there not have
to be some substance for our versions to project structure onto? According to Goodman, this tolerant realist
view that a plurality of worlds can be versions of a unique underlying Reality is also nothing but a useless
addition. A Reality underlying the worlds must be unstructured and neutral, and thus serves no purpose. If
there are many equally satisfactory versions of the world that are mutually incompatible, then there is not
very much left that the neutral Reality could be. Reality would have no planets, no motion, no spacetime,
no relations, no points, no structure at all. One can assume there is such a thing, Goodman seems to admit,
but only because Reality is really not worth fighting for (or fighting against, for that matter). If we can tell
true from false versions and explain why some are true versions and others false versions of worlds without
assuming anything like an underlying reality, why assume it then? Parsimony considerations should lead
us to refrain from postulating it.
6.2 Worldmaking
While Goodman insists that there are many worlds, if any (MM: 127; see also MM: 31), Goodman's
worlds should not be conflated with possible worlds. There are no merely possible Goodman worlds, they
are all actual (WW: 94 and 104; MM: 31). It is Goodman's view that worlds are made by answering to
right versions, but there are no (merely possible) worlds that correspond to false versions. It is important to
note that this view does not collapse into irrationalism or a fancy form of cultural relativism favored by
postmodern thinkers. Making a true version is not trivial. Not surprisingly, it is not any easier than making
a true version is for a realist. How we make true versions is absolutely the same on both accounts; the
difference is only with respect to what we do when we make true versions (see WW and McCormick 1996
for discussion).
The constraints on worldmaking are strict. We cannot just create things; predicates must be entrenched and
thus there must be some close continuity with former versions. Simplicity will keep us from creating new
things from scratch, coherence from making anything in conflict with beliefs with higher initial credibility,
and so on.
A world is made by making a world version. So, according to Goodman, the making of a world version is
what has to be understood. Carnap's Aufbau, as already mentioned, presents a world version, the systems
in A Study of Qualities and The Structure of Appearance are world versions, and so are scientific theories.
The heliocentric and geocentric worldviews are relatively primitive world versions, while Einstein's theory
of general relativity is a more sophisticated one. World versions do not have to be constructed in a formal
language, though; indeed, they do not need to be in a language, formal or informal, at all. The symbol
systems used in the artssuch as in painting, for instancecan be used in the process of worldmaking too.
It is in this sense that philosophy, science, and art are all epistemically significant; they all contribute to our
understanding; they all help to create worlds.
Making a world version is difficult. Acknowledging a great number of them does not make it easier. The
hard work lies, for example, in creating a constructional system that overcomes the problems of its
predecessors, is simple, uses well-entrenched predicates, or successfully replaces them with new ones
(which is even harder), allows us to make useful predictions and so on. For Goodman, scientists, artists,
and philosophers are faced with analogous problems in this.
Goodman's insistence that we make worlds when we make their versions and that we might just as well
replace talking about worlds with talk about versions creates a problem that is not simply solved by
acknowledging that making a true version is very hard. Making a version and making the objects that the
version is about are clearly two different tasks. As Israel Scheffler writes in the abstract to The Wonderful
Worlds of Goodman:
The worldmaking urged by Goodman is elusive: Are worlds to be identified with (true) world-versions or
do they rather comprise what is referred to by such versions? Various passages in [WW] suggest one
answer, various passages another. That versions are made is easy to accept; that the things they refer to are,
equally, made, I find unacceptable. (Scheffler 1979: 618)
Indeed, Scheffler argues, Goodman confusingly uses world and worldmaking both in a versional and
in an objectual sense. As we said above, Goodman's claim is that we make a world in the objectual sense
by making a version of it. The claim is based on his conviction that the only structure of the world to which
all true versions correspond does not exist independently; rather, it is only to be found because we project
this structure onto the world with our conceptualizations. His favorite example is the constellation known
as Big Dipper. Indeed, we made the Big Dipper by picking out one arbitrary constellation of stars and
naming it. (More precisely, it is a so-calledasterism that is part of the constellation Ursa Majorbut the
point still stands.) Which arrangement of heavenly bodies makes up the Big Dipper is purely conventional
and hence only due to our conceptualization. Hilary Putnam (1992a) suggests that this idea may have some
plausibility for the Big Dipper, but it does not, for instance, hold true of the stars that constitute the Big
Dipper. True, star is a concept with partly conventional boundaries; however, that the concept star has
conventional elements does not make it a matter of convention that star applies to something (and thus
merely a matter of making a world version).
Putnam also points out that there is a tension between Goodman's notion of worldmaking and his first line
of thought that leads to his irrealism: the idea that there are conflicting statements in different adequate
world versions. As Putnam argues, to say that a statement of one world version is incompatible with that of
another (such that a single world cannot accommodate both versions) requires that the expressions in the
two versions have the same meaning. However, it is not clear that our ordinary notion of meaning allows
such inter-version comparison of sameness of meaning (a thought that Goodman should sympathetic to,
since he already doubts an intra-version notion of that kind). Moreover, there might be better ways to
compare alternative versions (e.g., by way of homomorphism-relations between versions, as developed by
Goodman in SA and discussed above in 4.1) and to explain how versions relate despite their apparent
incompatibility (for example, by paying attention to the practice of scientists who manage to move from
one version to another).
Bibliography
A. Primary sources
Books
SQA Study of Qualities, Ph.D. dissertation thesis, Harvard University, 1941. First published New
York: Garland, 1990 (Harvard Dissertations in Philosophy Series).
SAThe Structure of Appearance, Cambridge, MA: Harvard University Press, 1951; 2nd ed.
Indianapolis: Bobbs-Merrill, 1966; 3rd ed. Boston: Reidel, 1977 (page numbers in our text refer to
this last edition).
FFFFact, Fiction, and Forecast, University of London: Athlone Press, 1954; Cambridge, MA: Harvard
University Press, 1955; 2nd ed. Indianapolis: Bobbs-Merrill, 1965; 3rd ed. Indianapolis: BobbsMerrill, 1973; 4th ed. Cambridge, MA: Harvard University Press, 1983.
BABasic Abilities Required for Understanding and Creation in the Arts, Final Report (with David
Perkins, Howard Gardner, and the assistance of Jeanne Bamberger et al.) Cambridge, MA: Harvard
University: project no. 9-0283, grant no. OEG-0-9-310283-3721 (010). Reprinted (in part and with
changes) in MM, ch. V.2.
MMOf Mind and Other Matters, Cambridge, MA: Harvard University Press, 1984.
RP(with Catherine Z. Elgin) Reconceptions in Philosophy and other Arts and Sciences, Indianapolis:
Hackett; London: Routledge, 1988; paperback edition, London: Routledge, Indianapolis: Hackett,
1990.