Some of Bach's most interesting and original writings in the philosophy of language have tended to focus on the problems and puzzles that arise from so-called propositional attitude attributions, in particular belief attributions. Such attributions (or
reports) take the form
A believes that p where
A is the subject to whom a belief is attributed and
p represents the sentence, proposition (or, more vaguely still,
content) that is supposed to be believed by
A.
In
A Puzzle About Belief Reports and
Do Belief Reports Report Beliefs?, Bach argues that there is a false assumption underlying all of the traditional forms of explanation of the nature of belief reports: they all implicitly endorse something that he refers to as the
Specification Assumption. This is basically the idea that the "that"- clauses of belief reports ("that William will take the train tomorrow", "that the sun will rise tomorrow morning") specify (i.e. directly refer to) propositions (or sentences) that the believer believes. Bach suggests that "that"-clauses do not specify but merely "describe" or "characterize" what a person believes.
He argues his thesis by first invoking several classic puzzles which have confounded philosophers of language since the time of Frege. The first type of puzzle is a variation on the classic problem of the substitution of co-referential terms in the context of attitude attributions. A simple illustration is the following pair of sentences:
- Lois Lane believes that Clark Kent is a wimp.
- Lois Lane believes that Superman is a wimp.
The simple substitution of one co-referring term for another would seem to transform the truth-value (and hence the content) of the sentence from true to false. But this should be impossible if we are faithful to a few simple and reasonable assumptions that are commonly accepted among philosophers: direct reference, semantic compositionality and
semantic innocence. Direct reference is the principle that singular terms contribute their referents to the propositions expressed by the sentences which contain them. Compositionality is the ideas that the meaning of a composite expression is derived from the parts which make it up. Semantic innocence is the principle that "embedding" a term or name in a "that" clause should not change its semantic value.
Given these basic assumptions how is it possible that the truth value of a sentence in the context of an attribution can change - that is, how is semantic opacity possible? Bach sketches four historical approaches to the resolving the problem and demonstrates each of their inadequacies. The first approach is that of Frege himself. Frege claimed that the reference of a term in the context of a belief report (or any other attitude attribution) was no longer its "customary" reference but rather its sense (see sense and reference). While this proposal maintains compositionality, however, it obviously violates the aforementioned principle of semantic innocence. Terms do not have the same reference in attitude attribution contexts that they do in ordinary sentence contexts. Bach illustrates
why this is a problem by providing an example sentence that involves anaphora and that leads to serious problems for the Fregean view:
- (An)Lois Lane believes that Clark Kent is a wimp, but he is not.
Here, the pronoun
he is being used, as linguists say, anaphorically: its meaning is derived from the proper name which it is standing in for (in this case,
Clark Kent). It seems clear, then, that the "he" used in this sentence refers (and quite directly) to Clark Kent. Frege's theory would predict that it refers to the
name Clark Kent, the
sense of the term in the terminology of Frege. Consequently, Frege's theory "denies semantic innocence" and this makes it ring somewhat counterintuitive.
Bach next considers what he calls, the
metalinguistic or
sententialist view. On this view, a sentence embedded in a "that"-clause refers to some sort of sentence, whether the sentence itself or a sentence in some language of thought depends on the specifics of the theory. However that may be, this view also violates the principle of semantic innocence: the referents of terms change from ordinary contexts (where they are objects in the external world) to attitude report contexts (where they are linguistic items. Hence, it has the same problem as Frege's, as well as several others.
The so-called
hidden indexical theory maintains that the difference in truth value (and content) between sentences
1 and
2 above has nothing to do with what they say about
what Lois Lane believes, but with what they implicitly say about
how she believes it. The two sentences do not differ in their contents, the singular proposition expressed by the statement that Superman/Clark Kent is a wimp, but by some implicitly referred to way of taking the proposition. In Bach's view this approach violates the principle of compositionality. There is no syntactical place in the sentence
A believes that G is F for some "unarticulated constituent" or "hidden indexical". He also points out that sentences such as "Joe is ready" and "Fred has finished", which are missing an argument, are not necessarily sentences that express propositions with unarticulated constituents. They may simply be semantically incomplete and hence not express propositions at all.
The last position that Bach considers is the so-called
neo-Russelian theory. Neo-Russelians attempt to solve the problem by rejecting the "anti-substitution intuition". They insist that sentences such as
1 and
2 actually have the same contents and that there is no transformation in truth values at all. Similarly, if
the Joker realizes that Bruce Wayne is rich is true, then it is also true that
the Joker realizes that Batman is rich; if
the Joker doubts that Bruce Wayne is a threat is true, then so is the statement that
the Joker doubts that Batman is a threat and so forth. These consequences make the neo-Rusellian theory seem extremely awkward and counterintuitive.
Bach's own, alternative solution is to reject the Specification Assumption discussed at the beginning.He further illustrates the problems associated with this assumption by way of another famous philosophical puzzle: Kripke's
Paderewski puzzle, which does not involve substitution.
- (a)Peter believes that Paderewski had musical talent.
- (b)Peter disbelieves that Paderewski had musical talent.
Kripke's puzzle arises from the fact that Peter takes Paderewski to be two different individuals: one a statesman and the other a pianist. In fact, they are one and the same person. According to Bach's
descriptivist view, sentence
a here
describes Peter as believing something and sentence
b describes him as believing something else. Since the "that"-clauses do not specify what the two things are that Peter believes (they do not refer to one specific object), then they are not necessarily the same thing. The condition for the truth of a belief report is that the believer must believe
something such that the proposition expressed by the "that" clause turns out to be true.
Attempts have been made to resolve the Paderewski puzzle by suggesting that the "that"-clauses involved are not sufficiently specific and that if all contextually relevant information were provided in detail, then we could eventually determine exactly what it is that Peter believes and disbelieves. However, as Bach shows, this leads to an infinite regress. We could add information to the sentences
a and
b which further specifies that
Peter believes that Paderewski "the pianist" has musical talent and
Peter disbelieves that Paderewski "the statesmen" has musical talent. But let us suppose that Peter hears a recording of Paderewski playing Mozart and is impressed with the performance. Later, he hears a recording of Paderewski playing Keith Jarret and is disgusted by the performance. Given that we have the same individual Paderewski and that Peter still does not know that it is the same individual in the two cases, we would have to say that
Peter believes that Paderewski the "classical pianist" has musical talent and that
Peter disbelieves that Paderewski the "jazz pianist" has musical talent. This specification might not suffice either. Suppose that Peter now hears Paderewski play Beethoven and is not impressed. We would have to say
Peter believes that Paderewki "the classical pianist playing Mozart"...".As Bach puts it, "that clauses are not inherently capable of specifying their contents fully".