The Builders
Proceedings of the 34th International Wittgenstein Conference, Epistemology, Contexts,
Values, Disagreement, eds. Christof Jager and Winfried Loffler (Verlag, 2011).
Meredith Williams
Johns Hopkins University
Rush Rhees, in his 1970 paper on “Wittgenstein’s Builders”, argues that the
builder’s game Wittgenstein describes in PI §2 is not a language. He cites a number of
reasons for holding this view. The central problem with this game is “to imagine that
they spoke the language only to give these special orders on this job and otherwise never
spoke at all. I do not think it would be speaking a language”1. In elaborating on this
point, Rhees identifies three further shortcomings. These are that the calls of A are only
signals “which cannot be used in any other way”2; the builders game does not show how
speaking is “related to the lives which people live “3; and lastly for A and B “there would
be no distinction between ‘that is not what we generally do’ and ‘that makes no sense.’”4
The builders use of language is rigidifying, indeed as Stephen Mulhall puts it in his
excellent discussion of the builders game, “stultifying the human imagination and
depriving itself of a future” (p. 57).5 Anyone familiar with the builders recognizes the
1
Rush Rhees, “Wittgenstein’s Builders,” Discussions of Wittgenstein (New York:
Schocken Books, 1970), p. 76
2
Ibid., p. 77.
3
Ibid.
4
Ibid.
5
Stephen Mulhall, Inheritance & Originality (Oxford: Clarendon Press, 2001), p. 57.
Mulhall treats PI §2 and the passages following (§§2-46) as a kind of abbreviated natural
1
relevance of these objections. Valuable though they are in allowing us to reflect on what
it is to live in the medium of language, these objections do not focus on the primary use
of the builders game. It plays a methodological role in Wittgenstein’s diagnostic critique
of the Augustinian picture of language, which is a crude or simple expression of the
representational picture of language. What I shall argue is that the critique of this picture
is directed against not only a referential theory of meaning, that is, a certain account of
the semantic relation between words and objects, but also against what Wittgenstein calls
“Frege’s idea”, an idea that implicates Frege’s semantic theory of sense and reference
more generally. In doing so, a certain picture of the systematicity of language is invoked.
The essence of this picture is the primacy of assertoric form. In presenting
Wittgenstein’s argument against this important idea, I shall have the opportunity to recur
to the objections raised by Rhees and others.
Wittgenstein introduces the expression “language-game” several passages after
the builders, which Wittgenstein describes as “a complete primitive language” (PI §2,
my italics). What he means by this phrase “a complete primitive language”, becomes
important in assessing the philosophical work to which this game can be put.
Wittgenstein distinguishes three ways to use the notion of a language-game (PI §§6-7).
The first use came into play in the opening two passages of PI: namely, the
methodological construction of fictional languages (the grocer and the builders). The
builders game is the iconic exemplar of such a fictional language-game.6 The second use
history of the evolution of language users. Anyone interested in Wittgenstein’s builders
should consult this interpretation. See pp. 52-87.
6
Wittgenstein uses this methodological device, after the Tractatus, throughout his
philosophical writings. The Brown Book, of course, is the outstanding example of the
use of artificially constructed language-games to explore concepts and rules, using the
2
is that of a teaching practice or game, particularly the teaching of the young child or
initiate learner. Such simpler teaching practices have a kinship to Wittgenstein’s
“complete primitive” language games. And finally there is the more descriptive, perhaps
even explanatory, use of language, which Wittgenstein calls “the whole, consisting of
language and the actions into which it is woven, a “language-game”” (PI §7).
Certainly in the Investigations, Wittgenstein uses such fictional languages in
pursuing the most powerful and original of the arguments of Part I. In constructing these
language games (the builders, the private diary and the beetle in the box)7, Wittgenstein
imposes two constraints. Any such methodological use of a language-game must be
constructed to “fit” the philosophical theory under investigation. Secondly, no
proprietary language can be used in the description of the language-game. The first
constraint is obvious, given that the task at hand is the critical examination of a
philosophical theory. The second is to ensure that the theory is not itself presupposed or
otherwise insinuated into the characterization of the fictional language-game. Especially
the technical vocabulary of Frege’s semantic theory and the elements of the Tractatus’
picture theory of meaning are avoided: pictorial form and pictorial relation, Fregean
sense and reference, and most importantly proposition. The point at which appeal to the
proposition is reached is the point at which one of Wittgenstein’s most insightful
critiques of the referential theory of language is raised. Surprisingly Wittgenstein uses
artificial game as “an object of comparison” (PI §130). Also see H.-J. Glock,”Language
Game, A Wittgenstein Dictionary (London: Blackwell, 1996).
7
There are many other language games that Wittgenstein uses in a methodological role,
several of these are simple teaching practices that Wittgenstein takes to resemble his
artificial games. These include the color chart argument (48) and the pupil just learning
arithmetic (143). The meter stick argument of PI 50 can plausibly be counted as a
language game, given the important argumentative role it plays.
3
the builders language-game as a tool in examining matters of linguistic systematicity: “is
the call “Slab!” in example (2) a sentence or a word?” (PI §19). Wittgenstein does not
use the builders game in his critique of ostensive definition as fixing the meaning of a
word. It is the language-game as teaching device that dominates the arguments of PI
§§26-38. The topic of this paper is the builders and systematicity: PI §§2-22.
1. The Builders and Systematicity
The builders game is played by two participants using “a language consisting of
the words ‘block’, ‘pillar,’ ‘slab,’ ‘beam’.” Each of these words is associated with a kind
of building-stone. Builder A calls out one of the words and his assistant B brings a stone
of the appropriate sort to him. The words are public labels for objects belonging to clearcut categories. In uttering these words the two builders coordinate their behavior.
Clearly, if words get their use from being names of objects, how this is so should be clear
in this simple language. Yet Wittgenstein does not immediately take up the semantic
relation between word “slab” and a slab. Rather the first contact with a “slab”-slab
relationship is the learning situation that mimics the naming situation in important
respects. In PI §4 Wittgenstein introduces an analogy by which we can understand the
philosophical point of Wittgenstein’s turning from the naming relation between “slab”
and slab to the learning relation between pupil and teacher that also uses the “slab”-slab
relation. The analogy from §4 reads as follows:
Imagine a script in which the letters were used to stand for sounds, and
also as signs of emphasis and punctuation. (A script may be conceived as
a language for describing sound-patterns.) Now imagine someone
interpreting that script as if there were simply a correspondence of letters
to sounds and as if the letters had not also completely different functions.
4
Augustine’s conception of language is like such an over-simple
conception of the script.
Augustine’s simple name theory of language treats language “as if there were simply a
correspondence of letters to sounds and as if the letters had not also completely different
functions.” This script cannot capture sound patterns. Different symptoms of
systemacity are involved with that: “signs of emphasis and punctuation” and the
“completely different functions”. The first place to look for illumination is the learning
situation, in which we can see that the apparent similarity with the Augustinian naming
language is broken.
Augustine takes the naming game to be initiate learning, but this is a grave
mistake. The superficial similarity between Augustine’s child and Wittgenstein’s child is
the teaching of words by the teacher’s saying, e.g., “slab” while pointing to a slab. For
Augustine that is sufficient to endow the word with meaning for the child. For
Wittgenstein, this saying is not a successful ostensive definition, but an ostensive
teaching that can succeed in effecting an understanding of the word “slab” only together
with a particular training. With different training the same ostensive teaching of these
words would have effected a quite different understanding” (PI §6). Two important
features must be noted: First, the results of teaching a “slab”-slab relation is causally
effected. It is not teaching in which the student uses a standard or of a rule or of a
meaning. Second, it is the background training in the use of that word that exploits our
causal situatedness in the world. What is of note is that issues of systematicity are in play
at the outset. The grocer’s game opens with the diversity of signs (PI §1). This is echoed
in the extended builders game of PI §8. There too Wittgenstein introduces a range of
differently regulated signs, including numerals, indexicals, color samples. The different
5
uses show the connections among the signs. This is the background against which
Wittgenstein turns his critical eye on what he calls “Frege idea” (PI §22). This idea is the
hypertrophied form of systematicity in which the Tractatus was steeped and which
persists in various conceptions of language. But what is shared by these conceptions that
rank the systematicity of language as its defining feature is Frege’s idea, the idea that
assertoric form is primary. Closely associated with this in Wittgenstein’s mind is what
he describes as the “chemical” structure of language and language mastery.8 Just as there
are chemical-radicals (sets of structured chemicals that, as a block, interact with other
sets) that are used in constructing more complex chemical structures, so these sentenceradicals that can be used within other more complex sentences. What opens the way to
these matters is Wittgenstein’s raising the question, quite naturally, whether “Slab!” is a
single word or a sentence (PI §19), a holophrastic sentence as are our own uses of single
words in teaching children, one that makes sense against the background of adult
training. Now to Wittgenstein’s criticisms of Frege’s idea.
2. The Development of Frege’s Idea
In PI §22, Wittgenstein identifies three theses that make up Frege’s idea, all of
which he claims are mistaken or misleading. These are:
(1) Frege's idea that every assertion contains an assumption, which is the
thing that is asserted….
(2) … the assertion consists of two actions, entertaining and asserting
(assigning the truth-value, or something of the kind)…
8
See PI p. 11 footnote. “Imagine a picture representing a boxer in a particular stance.
Now this picture can be used to tell someone how he should stand, should hold himself;
or how he should not hold himself; or how a particular man did stand in such-and-such a
place; and so on. One might (using the language of chemistry) call this picture a
proposition-radical. This will be how Frege thought of the “assumption”.”
6
(3) and that in performing these actions we follow the propositional-sign
roughly as we sing from the musical score.9
The first two theses make a principled contrast between sense and force fundamental to
any language. The first thesis identifies the constraints on a core theory of sense: the
proposition (“the thing that is asserted”) is the unit of meaning and so assertoric form is
fundamental to the structure of language. The propositional sign is the sentence-radical.
The second thesis provides the supplementary theory of linguistic action or force. And
the third thesis identifies linguistic understanding with a particular mental mechanism, a
kind of reading from a text of propositional-signs. In Wittgenstein’s analysis, the core
and supplementary theories are taken to support a mentalistic hypothesis according to
which the “chemical” structure of speech acts is mirrored in similarly articulated
psychological processes and states.10 Though Wittgenstein doesn’t think that it is so easy
9
I have added the numbers to the passages taken from the Investigations. I shall take up
thesis (3) in Chapter 3. Let me also note that Frege, strictly speaking, would object to
several features of this characterization of his “idea”. Where Wittgenstein speaks of
“assertion”, Frege would speak of “judgment.” Also, Frege would reject the third thesis
as introducing unwanted psychological content to the discussion of judgment and
propositional content. Indeed assertion itself is something that we do. Wittgenstein’s way
of putting the problem of the content of judgment shows that Wittgenstein rejects the idea
that the psychological can so readily be sidelined by a purely logical investigation.
Finally, Frege holds that his new notational system can provide a description of a logically
perfect language, but not of a natural language.
10
John McDowell in his excellent article “Anti-Realism and the Epistemology of
Understanding,” Meaning, Knowledge, and Reality (Cambridge, MA: Harvard University
Press, 1998) pursues many of the same issues I am developing here. Though I developed
the argument for this chapter independently of McDowell’s essay, I have found it very
useful in refining my discussion. McDowell and I both reach similar conclusions
concerning Dummett, McDowell by way of criticizing Dummett’s anti-realism and
myself by way of developing Wittgenstein’s early critique. As will be seen, there are two
important converging conclusions: That there is an innocuous or “truistic” conception of
truth conditional semantics that is philosophically “innocent,” as it were; and that
Dummett’s motivation for his revival of verificationism lies outside the demands for an
adequate theory of meaning.
7
to put this third thesis to one side, we will do so, to the extent that that is possible, in
order to respect Frege’s antipsychologism.
Wittgenstein invites us to apply Frege’s idea—consisting of a core theory of sense
(linguistic meaning) and a supplementary theory of force (linguistic acts)—to the builders
game.11 The structural complexity that results when a Fregean theory of sense and force
is applied to the builders is hidden in the builders game both behaviorally and
subjectively. It is not shown in the actual behaviors of A and B—in the one-word calls of
A or the stone carrying behavior of B—nor are A and B aware of this complexity. It
cannot be observed, from within the game, either from the third-person point of view or
the first-person point of view. Indeed the extreme simplicity of the game gives us no
reason to read such Fregean complexity back into the game, and some grounds for
thinking that it is not possible. So, the only reason for doing so is methodological. It is a
device for displaying, clearly and economically, the elements of the Fregean picture.
According to Wittgenstein, Frege’s theory of sense and force stands on a mistake, one
that conflates a banal feature of any natural language with a metaphysically robust entity.
The product of this Fregean conflation is the proposition.12
To see the conflation take place, in slow motion as it were, consider the way in
which the first two theses apply to the builders. The first step taken in applying Frege’s
picture to the builders is to treat A’s call “Slab!” as a holophrastic sentence in much the
11
Further, the point of the command might vary from occasion to occasion. It could be
used straightforwardly to get a building-stone moved to A’s location or it could be used
to display the status of builder A or to make a joke (Wittgenstein’s suggestion at PI §42),
and so on. But these contextually based factors are not considered essential to a theory
of meaning and understanding. The sense and force of a linguistic utterance are key in a
way that the point of its utterance is not.
12
This conflation argument can be paired with a paradox argument that is given in PI
§95. Wittgenstein’s critique of Frege’s idea is completed with this “paradox of thought”.
8
way that young children’s vocalizations are heard as holophrastic (PI §5). As
Wittgenstein says later in diagnosing problems in the philosophy of mind, this “first step
is the one that altogether escapes notice…(The decisive movement in the conjuring trick
has been made, and it was the very one that we thought quite innocent)” (PI §308).13
Later the one word call is described as a “degenerate sentence”(PI §19).14
Let us begin with a schematic presentation of a Fregean analysis of A’s call. We
can think of this analysis as presenting the logical structure that constitutes A’s
meaningful call “Slab” and B’s intentional action. We need not think of these as
psychologically real though the demands of the theory are made more vivid if we were to
do so. Let us make the Fregean analysis explicit for A’s call “Slab!”:
A
Slab!
(P1)
Bring me a slab!
(P2)
|-- (B brings A a slab.)
(P3)
Entertain ((B brings A a slab) and Assert (B brings A a slab))
(P4)
Order (Entertain (B brings A a slab) and Assert (B brings A a slab))
(P5)
Order (Entertain (B brings A a slab))
(6)
Order (proposition that B brings A a slab)
13
PI §308 occurs within Wittgenstein’s examination of the Cartesian model of mind. We
should expect there to be similarities in Wittgenstein’s diagnostic argument against the
Cartesian model of mind as we find in his critical examination of the Fregean picture of
language. And we won’t be disappointed.
14
PI §19 transforms a philosophically innocent remark into the opening move in
Fregeanizing the builders. Wittgenstein’s conflation diagnosis will identify just what
goes wrong. Let me note here that we are in the domain of the child when the word is
holophrastic; and we are in the domain of the adult when the phrase is degenerate. This
second way of characterizing the one-word “sentence” indicates the intervention of
Frege’s picture in characterizing the linguistic mastery of the builders.
9
A’s calls are most naturally construed as orders even though, on the Fregean
picture, they cannot be orders any more than names in isolation can refer to objects. The
context within which these calls are orders that mean something, enabling A and B to
coordinate their behavior, must be specified. (It should be noted that this specification
has nothing to do with the kind of background training Wittgenstein appeals in such
circumstances.) (P1) The first step in this specification is to take A’s call “Slab!” to
mean the lengthened form “Bring me a slab.” In §19 Wittgenstein raises the question “if
they mean the same thing—why should I not say: “When he says ‘slab!’ he means
“Slab!”? His interlocutor replies “But when I call “Slab!, then what I want is, that he
should bring me a slab!.” Here we see that even in this primitive case, meaning is tied to
something much more sophisticated than it would seem the builders game could support.
This something is, of course, propositional content. But we do not have a right to appeal
to propositional content this early in the analysis. The point of the analysis is to lead us
to the “discovery” of the proposition. (P2) Thus, the second step in the analysis is to
translate “Bring me a slab!” as having the same meaning as the assertion formally
demarcated by the assertion sign. So: |-- (B brings A a slab). (P3) The logical analysis
of the assertion sign reveals its two components: Entertaining a content (represented by
the vertical dash) and asserting that content (represented by the horizontal dash). (P4)
The requirement that anything meaningful must have assertoric form (the core theory of
meaning) brings with it, as a matter of course, the supplementary theory of force. If A
really means that B brings him a slab when he calls “Slab!,” the order to B must be
affixed externally to the meaningful sentence. This, as we can see, involves embedding
10
the meaningful call within force operators—asserting and entertaining. (P5) Since A is
not asserting or judging that B brings A a slab, the correct analysis reveals that what A
must be ordering is the entertained sentence “B brings A a slab.” The entertained
sentence just is that which means what the asserted sentence means without the assertion.
(6) This is the proposition.
We need an account of assertoric meaning that is fully independent of linguistic
force. To this end, (6) can be viewed as a hypothesis to explain (P5). This hypothesis
that one entertains an assertoric sentence without asserting that sentence is the hypothesis
that the proposition is the content of the entertained assertoric sentence. The proposition
is independent of any particular linguistic act, though it is born out of assertion. It is
independent of the particular syntax of any language; yet it mirrors in its own constituent
structure the syntactic structure of assertoric sentences in any language. It is that which
constitutes the content or meaning of any linguistic act. We can view the primacy of
assertion thesis as ineluctably tied to the discovery of the proposition. To make the
significance of this more vivid, let’s examine the difficulties in separating entertaining
from assertion.
P2 introduces the notation sign for assertion deliberately to indicate a more
technical notion of assertion, one that specifies the two distinct acts, asserting and
entertaining, as required for assertion. Now we can see that the use of “assertion” is
equivocal. It can be used in its ordinary way for any act of asserting, stating, declaring
and the like. All such acts are notable in their use of assertoric sentences. One can say
quite correctly and unproblematically that the sense of any ordinary assertion is that
which is asserted. This is to be understood as a commonplace or truism. But we also see
11
from the introduction of the Fregean assertion sign that the expression “assertion” is used
in a technical sense, one that analyzes assertion into two linguistic acts, (technical)
asserting (or assertingT) and entertaining. We can now examine the primacy of assertion
thesis by first considering the naïve response to the thesis. If assertion is semantically
basic, then many if not most of our linguistic acts are meaningless, because no act of
asserting is performed (as with builder A). If these linguistic acts are meaningful, it must
be because assertion is implicated in the acts in a way that is not marked either by its
syntactic shape or by its apparent meaningfulness. Illumination presumably is to be
found in the analysis of assertion provided by Frege’s assertion sign. Here we see that
even ordinary assertion is to be explicated in terms of two acts, assertingT and
entertaining. If sense is to be secured by assertionT, assertionT, as a kind of assertion,
must itself be explicated in terms of assertionT² and entertainment. Clearly we have an
explanatory regress.
3. Wittgenstein’s Criticisms of Frege’s Idea
The problem for the Fregean picture is to show how the semantic primacy of
assertion and the independence of semantic content from force are compatible. Despite
the independence thesis, assertion is fundamental, he can argue, since it is only through
assertion, understood as an abstract function that assigns truth values to well-formed
assertoric strings, that semantic content is realized. This is a claim as to the metaphysics
of content, not the epistemology of content. The issue at hand is not whether anyone
actually asserts anything, a psychological matter, but rather what they would mean if they
were to make any use of language. Thus, the primacy of assertion is really the primacy
12
of an abstract semantic function that operates only on assertoric sentences. Frege uses a
distinct noun to describe what I have been calling assertionT. That noun is “judgment”
and judging is “…striving for truth that drives us always to advance from the sense to the
reference.”15 The content of this function (this “advancing”) can be expressed by a thatclause. A proposition is a that-clause by another name; or perhaps, alternatively, a thatclause names a proposition.16 This metaphysical use of the proposition is just the device
to ensure the independence of content and force while providing the semantic primacy of
assertoric form. This is just what Wittgenstein criticizes.
Wittgenstein’s diagnostic argument is given in PI §§19-22. We can think of this
argument as the move from Frege’s idea to Frege’s mistake. Frege’s mistake is to
conflate an assertion with a propositional clause, that is, to take a meaningful sentence as
that which fixes meaning. There he aims to reveal the mistakes that keep assertion at the
center of language and cognition. He foreshadows, at this early stage, his alternative
picture by remarking that “to imagine a language is to imagine a form of life” (PI §19).
The Fregean account hypothesizes epicycles that at best ornament the moves in the
language game, but do not contribute to an explanation or a justification of the game in
play. We are thus forewarned that the debate over meaning and meaningfulness concerns
the background to linguistic use, whether it is a complex formal structure or a form of
life—a logic or a grammar. Most language users are blind to both of these, blind because
they are without the means to articulate either. Logical form requires training into the
techniques of logic and other formal systems. So if logical form is indeed the
15
Frege, op.cit., p. 63.
See, for instance, Gilbert Harman, “Logical Form,” Foundations of Language, vol. 9
(1972), where he explicitly takes the that-clause to name a particular proposition.
16
13
background to our language use, it is no more accessible to us than is our
neurophysiology. But with a form of life as the background, there too we are blind, not
because we fail to have the requisite logical or other technical skills but because its very
ubiquity in our lives makes it literally unnoticeable.
PI §19 informs us that the entry-point for complicating the structure of the
builders’ linguistic ability is our natural interpretation of the builder’s call as “Bring me a
slab.” Wittgenstein’s strategy (in PI §§20-21) is to argue that although the lengthened
sentence is a correct translation for us to make of the builder's call, it does not thereby
show that in order for A to command "Slab!" A must have first thought or presupposed
the lengthened sentence "B brings me a slab."17 The reason for our translating the
builder's call in this way has nothing to do with explaining what the builder does or
undergoes in making that call, in short, with what makes the builder’s calls meaningful.
The builder's "language" simply hasn't the syntactic, lexical and logical complexity to
support such a hypothesis. Indeed, if we want to explain the builder's use of the words of
his game, we can appeal to how he was trained in their use and what he goes on to do.
But if we want to explain why we translate "Slab!" in the way that we do, we must look
elsewhere, namely, to features of our more complicated language. It is because we can
say things like "This is a slab," "The slab is broken," "Where is the slab?" that we
translate A’s call "Slab!" as “Bring me a slab.” Our sentences have a subsentential
17
The builders, if they came in contact with us, would no doubt translate our "Bring me a
slab" into their "Slab!". See PI §19 where Wittgenstein asks "Buy why should I not on the
contrary have called the sentence "Bring me a slab" a lengthening of the sentence
"Slab!"."
14
structure created by patterns of substitution within sentences.18 But this does not require
attributing assertoric content to the builders or other language games that are not
asserting games. Yet minimal though it is, it is a normative practice: open to sanctions,
culturally heritable, and informed by the interests of the participants. Meaningfulness is
explained in terms of the surroundings, the history of the practice, and the way children
are trained into it. It cannot be explained in terms of a hidden mental process that mirrors
the demands of Frege’s idea. Nor is the builders’ practical skill in using these words
correctly described using Frege’s idea. The translations that introduce Frege’s idea into
the builders game prove to be at best a Ptolemaic trap, in which Fregean complexities are
as idle as Ptolemaic epicycles, required by the theory but explaining nothing.
The confusion that lies behind Frege’s idea is, according to Wittgenstein,
remarkably simple. The alleged ubiquity of the proposition in all linguistic interactions is
attributed to nothing more than a trivial feature of language:
Frege's idea that every assertion contains an assumption, which is the
thing that is asserted, really rests on the possibility found in our language
of writing every statement in the form: "It is asserted that such-and-such
is the case."--But "that such-and-such is the case" is not a sentence in our
language--so far not a move in the language-game. And if I write, not "It
is asserted that…", but "It is asserted: such-and-such is the case", the
words "It is asserted" simply become superfluous (PI §22).
18
See Robert Brandom, Making It Explicit, (Cambridge, MA: Harvard University Press,
1994), Ch. 6, where he explains subsentential structure in terms of substitution strategies
rather than in terms of semantically autonomous constituent elements. This is part of his
defense of an inferentialist conception of language, one that has much in common with
Wittgenstein's approach. Brandom’s detailed defense of this strategy offers a full
alternative to the traditional approach in terms of reference and truth-conditional
semantics. It shares, however, much with the Fregean picture of language. More
importantly Brandom retains the primacy of assertion. The core of language is tied to a
logically structured and rationalized conception of semantic meaning.
Also see Warren Goldfarb’s excellent discussion of these passages in his article “I
Want You to Bring Me a Slab”, Synthese, vol. 56 (1983), pp. 265-282.
15
In effect, Wittgenstein urges a deflationary treatment of Frege’s assertion sign and the
idea of propositional content that it seems to support. Treating "the thing that is asserted"
as a separable component—the content or sense—of an assertion is not an insight into the
necessary structure of language, but merely a reflection of the fact that it is always
possible to construct a sentence of the form "It is asserted that …" whenever an assertion
is made.19 This is a banal point. As Wittgenstein points out, we could also write every
statement in the form of a question followed by a "Yes," but this would not show that
every statement really contains a question. Equally, from the fact that we can add the
phrase "It is asserted:" to each statement does not thereby show what it is to grasp the
sense of a sentence and then assert it. This banal feature of any language that has the
resources of second-order descriptors takes on a different significance when it is confused
with or taken to reveal the ontological ubiquity of assertoric form.20
19
In section 1 above I pointed out that Frege rejects a certain understanding of the
relation of sense and reference in sentences. It is a mistake, Frege urged, to think that the
relation between these two semantic properties is the relation of subject to predicate. As
Frege puts it: “One might be tempted to regard the relation of the thought to the True not
as that of sense to reference, but rather as that of subject to predicate. One can indeed
say: ‘The thought, that 5 is a prime number, is true.’ But closer examination shows that
nothing more has been said than in the simple sentence ‘5 is a prime number’” (Frege, op.
cit., p. 64). Frege uses this to argue that the relation between sense (or thought) to
reference (the True) is the relation of name to object.
20
As we can see from footnote 28, Frege takes this banal feature to be the grounds for
rejecting a certain conception of the relation between thought and the True. In section 1,
I have given reasons for thinking that Frege needs to use something like the argumentfunction relation. But what I wish to emphasize here is the very different morals Frege
and Wittgenstein draw from the emptiness of moving from “p is true” to “p” (Frege) and
“it is true that p” to “p” (Wittgenstein). For Frege, it eliminates a possible account of the
relation between sense and reference and clears the way for his own theory. For
Wittgenstein, this banal inference helps explain why a certain conflation seems so
unobjectionable. It is part of the mistake that gives us the proposition as the
metaphysically robust intermediary between language and the world.
16
The second stage of the diagnosis reveals that philosophers make a particular kind
of mistake, opening them to the plausibility, if not inevitability, of Frege’s idea. As
Wittgenstein says, the use of Frege’s assertion sign “… is only a mistake if one thinks
that the assertion consists of two actions, entertaining and asserting (assigning the truthvalue, or something of the kind)” (PI §22, my italics). He goes on to say that “Frege’s
assertion sign marks the beginning of the sentence,” not the introduction of a
metaphysically robust content.21 Again, as with the confusion, a deflationary
understanding of Frege’s assertion sign as marking the beginning of a sentence does not
aim at an explanation for how we manage to assert anything, let alone show that assertion
is implicit in every meaningful linguistic act. It only seems to provide such an
explanation when the two components, marked by the assertion-sign, are taken for
distinct actions. With this mistake we arrive at a metaphysically robust conception of
content, propositional content. It is metaphysically robust because it is not to be
identified with psychological acts of asserting. The mistake makes possible the view that
“entertaining an assumption” could be found in every linguistic act, not just acts of
assertion.
The conflation and the mistake combine to support the semantic machinery of
Frege’s idea. This final stage secures the proposition as the robust content of our
sentences. In conflating the two sentences—“it is asserted that such-and-such is the case”
21
We can see that Wittgenstein’s deflationary account of assertion and proposition
(proposition being that which is asserted) is matched by his expressivism in connection
with the mental. Here too Wittgenstein uses the same language: “It would be quite
misleading, in this case [taking the sentence ‘I know how to go on’ to be a ‘description of
a mental state’], for instance, to call the words a ‘description of a mental state’.—One
might rather call them a ‘signal’; and we judge whether it was rightly employed by what
he goes on to do” (PI §180).
17
and “it is asserted: such-and-such is the case”—a new logical space seems to open: a
“sentence” that isn’t a sentence yet has meaning; and an ordinary meaningful sentence
has the very meaning of the sentence-fragment. To treat the two assertion sentences as
the same obscures their differences, which is evident once the phrase “it is asserted” is
removed from each. What is needed to turn this confusion into the semantic machinery
for all of language is the mistake of taking the assertion sign to introduce two acts, an act
of entertaining, which now anchors the sentence fragment, and an act of asserting which
“advances” us from sense to the world, and so anchors the content shared by entertaining
and asserting to the world. In the confusion, we have created shared syntactic structure;
and through the mistake, we have shared semantic sense. In that intersection of ideas
comes the proposition. It is that which shares the content of assertion without being
asserted. And it shares constituent structure with sentences without being a sentence.
In contrast to the primacy of assertoric form, the builders game underscores the
primacy of normativity within our language games. Yet the primitive normativity
displayed in the builders game leads some, not to question Frege’s idea, but to urge that
the builders game is irrelevant to a discussion of language. It is simply too primitive to
bear upon the issue at all. Indeed for many like Rhees, Goldfarb, Brandom and others,
the poverty of the calls and actions show the irrelevance to language. In examining these
objections, I will show that the builders game is just the right instrument to use in
breaking the hold of Frege’s idea. Its simplicity enables us to see normativity without
logicism.
4. The Builders Game: Primitive Language or Animal Signaling?
18
Rush Rhees, Warren Goldfarb and Robert Brandom, all maintain that the builders
game is too simple to count as a language. In a sense their grounds for rejecting the
builders as a primitive language-game is the same. Rhees puts it this way:
What [the builders] have learned are signals which cannot be used in any
other way…It does not show how speaking is related to the lives which
people lead.22
As Brandom puts it, the builders engage in a practice that is vocal but not verbal:23 “It is
not a genuine language game.” For Brandom, the builders may have a primitively
normative practice, but they don’t have a discursive linguistic practice. Goldfarb24 asks
rhetorically, “If this is the whole of their ‘language’, can we take them to be speaking, to
be using words with understanding, to be human?” He urges with Rhees that the
simplicity of the builders’ calls makes it more appropriate to assimilate these calls to the
cries of animals, to be explained in causal terms. The practice fails to be normative in
any sense. The role of the builders game, according to Goldfarb, is to unsettle the reader
and lead to a serious examination of the claims made on behalf of the explanatory work
achieved through analysis. All three hold that the very simplicity of the builders game
counts against its explanatory or methodological relevance to our understanding of
language. But this, I think, is a mistake.
The importance of the builders in challenging Frege’s idea comes with the
primitive normativity of that practice. Rhees and Goldfarb does not see a normative
22
Rush Rhees, “Wittgenstein’s Builders,” Discussions of Wittgenstein (New York:
Schocken Books, 1970).
23
Brandom makes this point explicitly in both Making It Explicit, op. cit., p. 172, and
Articulating Reasons, op.cit., p. 15.
24
Warren Goldfarb, “I Want You to Bring Me a Slab: Remarks on the Opening Sections
of the Philosophical Investigations,” Synthese, vol. 56 (1983), pp. 265-282.
19
dimension in their behaviors at all. For Rhees, its poverty and rigidity highlights how
very different language is:
It has to do rather with what is taken to make sense, or with what
can be understood: with what it is possible to say to people: with what
anyone who speaks the language might try to say.25
Goldfarb allows the usefulness of the builders game can only be that of a
stimulant to reconsider what philosophical analysis achieves. For Goldfarb, the analysis
of “Slab!” into “Bring me a slab” is to be explained in terms of the variant sentences in
which the expression “slab” can occur in our language; it is not be explained in terms of
an underlying structure of thought. In making this point, Goldfarb suggests that the
builders game is left behind, serving only to initiate the inquiry. To be a language-user is
to be adept in the use of such variant sentences. This is to say that language, to count as
language, must have a certain degree of syntactic structure and logical
interconnectedness. Linguistic mastery is minimally a practical competence that respects
this systematicity. On my interpretation the builders play a more direct part in the
argument, a methodological device for criticizing a formalist conception of language tout
court. Brandom locates the defect of the builders elsewhere, though the defect on his
account also concerns systematic features of language. Brandom holds that the builders
do not use concepts, the structure of which must be discursive; that is, to wield concepts
is to be able to ask for and give reasons for claims (Wilfrid Sellars’ “space of reasons”).
These objections fail to acknowledge Wittgenstein’s central point in using the
builders:
25
Rhees, op. cit., p. 84. Rhees uses other tropes in exploring the builders game,
including comparisons to mathematics and learning
20
I want to regard man here as an animal; as a primitive being to which one
grants instinct but not ratiocination. As a creature in a primitive state.
Any logic good enough for a primitive means of communication needs no
apology from us. Language did not emerge from some kind of
ratiocination (OC §475).
The cost of rejecting the relevance of the builders game to language, I shall argue, is to
blind ourselves philosophically to the background of normatively primitive, contextually
bound skills that linguistic meaning, formal systematicity and reason giving presuppose.
Such a background cannot be treated as something merely animal and so reducible to
behaviors amenable to causal explanation. Thus, both Goldfarb and Brandom open a gulf
between language as a normatively and logically complex system and mere behavior that
is part of the causal nexus. This is just the picture Wittgenstein combats.
Brandom denies that the builders game is a genuine language game because
(w)hat makes something a specifically linguistic (and therefore, according
to this view, discursive) practice is that it accords some performances the
force or significance of claimings, of propositionally contentful
commitments, which can serve as and stand in need of reasons. Practices
that do not involve reasoning are not linguistic or (therefore) discursive
practices.26
For Brandom, “what distinguishes specifically discursive practices from the doings of
non-concept-using creatures is their inferential articulation. To talk about concepts is to
talk about roles in reasoning.”27 This is Brandom’s way of endorsing the primacy of
assertion. He sees linguistic mastery as consisting of the practical ability to keep track of
the commitments and entitlements incurred by one’s own assertions as well as those of
others.
26
27
Brandom, Articulating Reasons, op. cit., p. 14.
Ibid., pp.10-11.
21
Thus, Brandom does not deny that the builders game is a language because it fails
to display the requisite formal systematic structure. Failure to show the right kind of
(abstracted) formal systematicity is, for Brandom, a symptom of why the builders game
fails, but it does not explain the failure.
The builders don’t speak a language because they have not entered the space of
reasons. They are unable to engage in the material inferences that constitute the
commitments and entitlements incurred by the exercise of concepts. Their vocalizations
are not open to epistemic challenge. The structure of discursive practices, according to
Brandom, is the structure of argumentation, of epistemic challenge and response.
Wittgenstein would argue that normativity (and so meaningfulness) cannot be explained
in terms of the discursive structure that Brandom argues is fundamental to an explanation
of meaning. So, though Wittgenstein would agree with Brandom’s rejection of formal
systematicity as explanatorily fundamental, he would nonetheless reject Brandom’s
theory of meaning that makes assertoric form primary.
But we need to see more precisely where Wittgenstein could challenge
Brandom’s account. To do this, we need to introduce further points of contact between
them. Brandom offers a more nuanced account of normative practice than has been
suggested thus far. And to see where he makes his mistake with respect to the builders
we need to bring some of this richer account out. Brandom agrees as most everyone does
that there is a fundamental divide between the merely causal order and the normative
order. What is distinctive about our human form of life is that our ways of acting are
normatively constrained.
22
Brandom distinguishes three levels of exercising practical skills, or three levels of
doing as he puts it.28 There is reliably responding differentially to properties in the
world. This could be the result of innate mechanisms or connections acquired through
conditioning, e.g., being conditioned to differentiate red things. The Skinner box
experiments are replete with examples of pigeons, rats, and other animals being trained to
differentiate properties in this way. This is a causally explicable response to the world.29
In contrast to such causally explicable doings in the world, there are two kinds of acting
in the world that cannot be explained wholly in causal terms because they are normative.
They are normative doings because these actions are explicable only in terms of what is
appropriate, correct or warranted. The explanatory apparatus of conditioning theory
cannot capture the very features that individuate and explain these doings. The first and
more primitive kind of doing is that which is appropriate to the situation, the case, for
example, in which red things are appropriately responded to by making a certain call.
Using the vocable “red” in such circumstances is not yet a conceptual matter for
Brandom. To apply the concept of red when saying “red” in the presence of red things is
not just a matter of being conditioned to red stimuli nor to uttering the expression when
appropriate. It is a matter of engaging in a discursive practice in which what is implicit in
the actions of the participants, saying “red” in the presence of red things, becomes
explicit in the application of the concept of red. This occurs when the exercise of that
responsive capacity is used to provide reasons or stands in need of reasons. It’s not just a
28
Brandom, Articulating Reasons, op. cit., p. 17.
This claim can, of course, be challenged and for good reasons. At this point in the
argument, however, it is useful to let Brandom’s distinctions between these three kinds of
doing stand. For careful criticism of Skinnerian conditioning theory and particularly the
claims made on behalf of the Skinner box experiments, see Charles Taylor’s excellent
book The Explanation of Behaviour (London: Routledge & Kegan Paul, 1964).
29
23
matter of saying “red” when appropriate (say when a red traffic light comes on and not
when a cardinal flies by), but being able to use this saying to make claims, give reasons,
correct others, and the like. Only in this discursive practice is the use of the vocable
“red” a linguistic act.
This is a helpful taxonomy. It will prove useful to distinguish first what marks off
merely causal doing (reliable behavioral response) from normative doing (appropriate
response), and second what marks off nonconceptual yet appropriate response from fully
conceptual discursive doing. In general terms, normative practices of any sort can be
distinguished from causally explicable behaviors in terms of four features. First, the use
of linguistic items or signs is an indispensable part of the behavior of the practitioners
and the coordination of their behavior. Second, the behavior is normative in that it is
appropriate or not, correct or not. Third, in being normatively constrained, the behavior
is evaluable and thus subject to sanctions. And lastly such normatively constrained
behavior is culturally heritable; it is not just acquired, it is learned. Nondiscursive
normative practices differ from discursive practices in that these responses to the world
are appropriate and yet have not an argumentative structure, and so are not conceptual.
In Brandom’s taxonomy, nondiscursive responses do not function as claims that can be
used as reasons for other claims or that can be supported by reasons. Nonetheless, as
normative practices, they incorporate sanctions against inappropriate behavior. The
sanctions are not mere reinforcements; nor do they involve referees to adjudicate disputes
(or whatever person is authorized to impose the sanctions within the game).
Let me match kinds of sanctioning to the three levels of doing that Brandom
distinguishes. The pigeon conditioned to respond to a red light by pecking is, in an
24
attenuated sense, subject to something akin to sanctions, namely, positive and aversive
reinforcement. The distinction between the causally reliable behavior of the pigeon and
primitively normative practices is that the selection of behavior and stimulus as well as
the history of conditioning is imposed by the natural environment, or in the experimental
situation, by the scientist who creates the pigeon’s environment. Normative practices are
sustained by participants of the practice.
We need to see what is it about the builders that make their calls normative in a
way that the causally induced peckings of conditioned pigeons are not. It can’t turn, of
course, on the superficial difference between the pigeons and the builders, namely, that
the builders use vocables while the pigeons do not. Pigeons could be conditioned to coo
in a distinctive way whenever a red light goes on in their box. How does the call of
“Slab!” differ from a trained cooing pigeon?
We can see the difference along the four dimensions of a normative practice
introduced earlier. The role of “Slab” in the game expresses a norm, that B ought to
bring a slab when A calls out “Slab”. First, the call “slab!” plays an ineliminable role in
the action. Second, that this is a norm can be seen in the fact that B’s behavior can be
incorrect or inappropriate. Third, incorrect behavior is sanctionable. And fourth, and of
especial importance to Wittgenstein, the children of this tribe can be taught the norm. So
the builders game is structured by norms, which are supported by sanctions, and are
culturally heritable. Skinner box creatures can obscure the difference between causally
conditioned doings and normatively informed doings. This is because the behavior of
Skinner box creatures is informed by norms imposed by the experimenters. They decide
what behavior is correct and impose sanctions accordingly, but here both norms and
25
sanctions are fully external to the behaviors of the animals and it is for this reason that
their behavior is taken to support a causal conditioning theory of behavior modification.
Remove the experimenters and the pigeon’s failure to peck in association with a red
stimulus cannot be said to be incorrect or inappropriate. It is just a different behavior.30
The norms of the builders game, on the other hand, are internal to the behavior of A and
B; theirs is a normative practice. Nonetheless, Brandom insists that it is not a linguistic
practice.
At bottom, Brandom’s mistake in denying the builders the status of a languagegame turns on his use (or misuse) of the two constraints that are important to his
discussion of linguistic practice. One is the contrast between the normative and the
causal, and concern issues of explanatory reduction. The other is the contrast between
discursive and nondiscursive normative practice, and concerns what counts as conceptual
engagement with the world and each other. For Brandom, the builders’ calls are merely
vocal because they are not fully discursive. But surely the key difference is the
normative-causal divide. That is certainly Wittgenstein’s point. Once that key difference
is in place, there is no deep theoretical need to identify just the point at which
nondiscursive becomes the discursive or the nonconceptual becomes conceptual. Rather
it is the causal-normative divide that is crucial. Brandom is correct in distinguishing the
primitive normativity displayed by the builders from the complex set of norms
constituting full-fledged language. But this doesn’t diminish the importance of the
builders since their primitive normativity is the background required for participating in
30
It is interesting to note that Chomsky says something similar about innate grammar,
and that is that there is no such thing as the “correct” natural languages. There is only
what people say, and what they say, of necessity, expresses they grammar they use. A
difference in spoken grammar is just different. See CHECK
26
the complex inferentially articulated space of reasons. To understand the character of
that background requires giving up the primacy of assertion thesis.31
31
It is perhaps useful to note an affinity between Wittgenstein’s use of the simple
builders game and Plato’s methodological use of the simple economic community in the
Republic. In Book II, Plato begins his investigation into the nature of justice by
introducing a thought experiment, that of a rudimentary community. It is a purely
economic community in which there is a natural division of labor such that the physical
needs of the members of the community are met. The skills required are passed on from
one generation to the next. It will be recalled that Glaucon CHECK condemned this
rudimentary community as “a city of pigs” hardly worthy of the name “city” at all.
Given that all naturally, and as a matter of course, carry out their labors, the issue of
justice simply cannot arise, according to Plato. Not until the community becomes
complicated in ways that require a political structure does it become a city.
We find an echo of this charge in Goldfarb’s and Brandom’s claim that the
builders are so primitive that their calls are not worthy of the name “language”. But just
as Plato found his economic community useful though it lacks the political complexity of
class structure required for justice or injustice, so does Wittgenstein find the builders
useful though their game lacks the systemic complexity required for making claims and
justifying them. For both Plato and Wittgenstein, this simplicity is the point of
introducing the thought experiment. The right way to understand the usefulness of both
the economic community and the builders game is to note that they are explanatory
because they lack the complexity of full-fledged practices. They provide the background
against which political complexity and epistemic-logical complexity arises and can be
sustained. See Plato, Plato’s Republic, ed. by A. Anderson and tr.by B. Jowett (Millis,
MA: Agora Publications, 2001).
27