I. Fodor’s Naturalization Proposal and the Represented-Disjunction Problem
The specific task which the naturalization project is designed to accomplish is to
construct an information-based semantic theory that articulates (I) an atomistic and (II)
naturalistic — i.e., in non-intentional, non-semantic terms — sufficient
condition for a syntactically primitive predicate to express a property. (1)
For convenience (I) will be referred to as the Atomistic Requirement and (II) as the
Naturalistic Requirement. Now if Direct Informational Semantics were correct,
"tamarao" would express the disjunctive property tamarao or thump
because, under certain conditions, a thump and a tamarao are each sufficient for its
tokening. Thus a tokening of a symbol would be ipso facto true. There would be no
accounting for the fact that meaning is robust:’ "cow" tokens get
caused in all sorts of ways, and they all mean cow for all that’. (2)
I shall refer to this problem (3) as the
Represented-Disjunction Problem, considering that it figures on the represented side.
Fodor proposes to solve the Represented-Disjunction Problem without abandoning the main
thrust of the information-theoretic account. His proposal (FP) is that "X" means
X if: (4)
(1) ‘Xs cause [X]s’ is a law. (Information Condition)
(2) Some [X]s are actually caused by Xs.
(3) For all Y not = X, if Ys qua Ys actually cause [X]s, then Ys causing [X]s is
(synchronically) asymmetrically dependent on Xs causing [X]s. (Asymmetrical
Dependence Condition)
(4) Some [X]s are actually caused by non-Xs. (5)
(Robustness Condition)
(1) should be read as saying that the property in virtue of which Xs cause [X]s is
nomically connected with the property of being the cause of [X]s. (6)
The process of naturalization is constrained to withhold the morphemic specification of
tokens in favour of the more neutral phonological/orthographic reading. I have marked this
distinction by using brackets for phonological/orthographic sequences whilst morphemes or
words shall be indicated in the usual way by using double quotes. Thus [cow] for the
phonological/orthographic sequence, and "cow" for the morphemic entity. But a
warning is in order. The ordinary or natural construal of ‘phonological/orthographic
reading’, involving as it does a presupposition of the notions of convention,
conventional correctness and language, and these being intentional as anything can be,
cannot be the right construal in the present context. The phonological/orthographic
elements that are said to figure in the laws appealed to have to be brute
phonological/orthographic elements, i.e., brute physical shapes.
II. Robustness and the Representer-Disjunction Problem
The way FP handles the Represented-Disjunction Problem looks very straightforward and
has some intuitive appeal. Tokening a [horse] has many causes but it does not follow that
"horse" means (represents or expresses) all of them or a disjunction thereof.
"Horse" means horse because the horse-[horse] connection is the
law on which all the other nomic connections involving [horse] tokens asymmetrically
depend: Cows woudn’t cause [horse]s but that horses do.
It must be noted at the outset, however, that the robustness of meaning has a facet not
emphasized by Fodor. It is that robustness means not only that a symbol’s getting
caused in various ways is compatible with its representing a single non-disjunctive
property, it means — corollarily-that (instantiations of) represented properties
cause various events including but not limited to tokenings of phonological/orthographic
sequences and tokenings of symbols compatible with their being picked out by
non-disjunctive symbols. Properties typically produce in the exercise of their causal
powers a multiplicity of effects some of which can be specified intentionally, some
physically and others phonologically/orthographically. The causal power of X, which may be
either intrinsic or extrinsic, is not exhausted by its causing [X]. I believe that
Fodor’s failure to emphasize this aspect of robustness is due to his contempt for
holism and strong commitment to the Atomistic Requirement, a requirement which favours
adopting the methodological stance of determining the content of a symbol in isolation
from the determination of the contents of other sysmbols.
Our presentation of the phenomenon of robustness highlights what is otherwise concealed
in discussions of the Asymmetric Dependence Condition, the condition which forms the core
of FP’s solution to the disjunction problem. In these discussions, the disjunction
problem is exhibited as something about the represented property, but this portrayal
covers only half of the story. The other half concerns the representing side. Consider
situations like these:
(i) horse—>[horse]
cow—>[horse]
(ii) horse—>[horse]
horse—>[cow]
In (i) two properties are each nomically connected with the same
phonological/orthographic sequence. This is the type of situation which raises the
Represented-Disjunction Problem for Direct Informational Semantics. Now (ii) is
interesting because it raises a different type of disjunction problem. Here we have two
different phonological/orthographic sequences being nomically connected with the same
property. Direct Informational Semantics must say that it is [horse]-or-[cow] (7)
rather than just [horse] which expresses horse. Insofar as it wants to identify
these phonological/orthographic sequences with our English morphemes, it must say that
"horse"-or-"cow", rather than just "horse", expresses horse.
In a way the problem is to look for the semantically relevant difference between the
property of being a cause of [horse] tokens and the property of being a cause of [cow]
tokens. Hereafter I shall refer to this type of disjunction problem as the
Representer-Disjunction Problem, considering that it figures on the representer side of
the semantic relation. Therefore there are are least two types of disjunction problem.
Adding the two together, Direct Informational Semantics and all information-based
naturalistic semantics are saddled with the problem of having to avoid saying, on the
level of primitives, that:
[horse]-or-[cow] expresses horse-or-cow.
The actual world, I suppose, is such that the ‘disjunctions’ on either side
of this relation may be intolerably massive.
Direct Informational Semantics leads not merely to pansemanticism, as Fodor calls the
doctrine that everything means something, but to double disjunctivism. To be sure, Direct
Informational Semantics is open to another possibility, viz., [horse] and [cow]
individually express horse — but only because it cannot give any reason for
preferring one over the other. Since this possibility of acute semantic redundancy, itself
only one remove from disjunctivism, is intolerable, we are well-advised to reject it as
without mitigating potential. It is necessary for a Fodor-type naturalization project to
solve both types of disjunction problems. Naturalizing the representational relation comes
down to naturalizing the property of being a representation of X and naturalizing
the property of being represented by Y. The former property is as intentional as
the latter. This alone should have made it obvious that if there is a disjunction problem
to be dealt with on the represented side, there must at the same time be a disjunction
problem on the representing side, the latter being no less urgent than the former. The
Representer-Disjunction Problem shows its sting because one is not entitled to assume that
orthographic/phonological divisions coincide with morphemic ones. In fact, in large
measure, the problem of the naturalization project is whether, for example, [tamarao] can
be identified with "tamarao".
How might FP’s solution to the Representer-Disjunction Problem and the
Represented-Disjunction Problem go? Prima facie FP has no solution to offer of the
former problem. Remember that the Asymmetric Dependence Condition has something to do
with, for instance, the law involving [horse]s on which all other laws involving [horse]s
asymmetrically depend. It is designed to deal with the latter problem, and its application
presupposes that a successful solution to the former has been found whereby [horse] in (i)
and (ii) is determined to be the right phonological/orthographic sequence. So let us
review some moves that Fodor might make.
(a) He might react with his usual dismissive strategy and simply legislate that [horse]
is the relevant sequence, [cow] isn’t, and that is all there is to it: so nothing
prevents the application of his conditions including the Asymmetric Dependence Condition.
The reason why the move won’t do is because it amounts to an acknowledgment that the
distinction between the semantically relevant and the irrelevant sequences is not
principled. What the relevant sequence is does not lie in the nature of things, it is
rather imposed by the theorist. The Naturalistic Requirement, in consequence, is
compromised. With the Representer-Disjunction Problem, it becomes implausible to insist
that it is enough to have an objective asymmetrical dependence of the cow-[horse]
law on the horse-[horse] law as if there is no need to explain the special status
accorded to the [horse] token or type. It now appears that asymmetrical dependence cannot
explain that, for asymmetrical dependence is given pride of place in part because [horse],
in the situation, is special.
(b) Fodor might devise, if only to pursue an aesthetically satisfying symmetry between
the two disjunction problems, a Second Asymmetric Dependence Condition which runs like
this:
(5) For all [Y] not = [X], if Xs qua Xs actually cause [Y]s, then Xs causing
[Y]s is asymmetrically dependent on Xs causing [X]s.
And to complete the symmetry, in analogy with the idea that "asymmetric dependence
engenders content only if it produces robustness," (8)
a counterpart of the Robustness Condition might be deemed essential:
(6) Some non-[X]s are actually caused by Xs.
(6) merely reaffirms the importance of the other side of robustness — as such
there should not be any qualms about it. The Second Asymmetric Dependence Condition is a
different matter. When applied to situations (i) and (ii), it requires that the
horse-[cow] connection be asymmetrically dependent on the horse-[horse] connection.
Unfortunately, whereas the application of the Asymmetric Dependence Condition takes for
granted the availability of a successful solution to the Representer-Disjunction Problem,
the Second Asymmetric Dependence Condition reverses the direction, as it were, by begging
the Represented-Disjunction Problem. Moreover, there is a manifest impossibility of
satisfying it. Look at situations (i) and (ii) again. Imagine that Mary is a normal human
being (in the actual world) so that the laws in (i) and (ii) are operative. Now it is
physically possible for her to acquire a rare neurological disease which makes her
incapable of tokening [horse] in writing and in speech (or, if you will, tokening [horse]
in the brain). In that possible world, let us suppose, she can still token [cow]. Though
the horse-[horse] law and the cow-[horse] law are in effect broken, the horse-[cow]
law remains intact. Mary’s case can be generalized, showing that the Second
Asymmetric Dependence Condition can never be satisfied.
(c) An alternative strategy which won’t do is to claim that [cow] is to be taken
out of consideration when determining what expresses horse, because its content is
determined independently through the application of FP. Our not recognizing this
alternative strategy, it might be said, is due to our limiting our attention to (i) and
(ii). But, one might continue, there are situations like the following:
(iii) cow—>[cow]
horse—>[cow]
The problem with this, despite its call for more attention, is that it suffers from
partial vision. The Representer-Disjunction Problem appears here as well as shown by this
situation:
(iv) cow—>[cow]
cow—>[horse]
Even ignoring the fact that the strategy does not in any way address the problems
already mentioned, it is definitely not open to Fodor. For one thing, it violates the
Naturalistic Requirement: the condition proposed here is obviously semantic. It can no
doubt be rephrased in a way which can seem to disarm this objection, by saying that [cow]
is ruled out because it participates in another set of relations where the Asymmetric
Dependence Condition holds. But quite independently of whether the redescription will
ultimately go through so as to avoid violating the Naturalistic Requirement, it cannot
help violating the Atomistic Requirement in that the determination of the semantically
relevant token or type is made to rest on the content of another having been determined.
III. Asymmetric Dependence and Phonology/Orthography
But assume, for the sake of discussion, that the Representer-Disjunction Problem has
been fixed for naturalization purposes, (9) we can still ask
directly if there is truth to the assertion that the favoured status falls on the horse-[horse]
connection in the manner required by the Asymmetric Dependence Condition. Robustness, as
we have described and exemplified it, demonstrates the falsity of such claim. Block has
objected to FP along this line:
If you mean by "cow" something like the phonological/orthographic
sequence [cow], (10) then there’s just no reason at all to believe
the claim you’re making. For example, there is surely a possible world in which cows
don’t cause [cows] but trees do, viz., the world in which [cow] means tree. So
... it clearly can’t be nomologically necessary in order for "cow" to mean cow
that nothing causes [cow]s in worlds where cows don’t.
. . . There is no orthographic/phonetic sequence [X] which mightn’t
mean tree in some nomologically possible world or other, whatever [X] happens to
mean here. . . . It follows that there is no orthographic/phonetic sequence [X] the
nomologically possibility of tokenings of which is dependent on [X]s being caused by Xs.
So there is no such sequence that satisfies your sufficient condition for meaning X. (11)
The response is evasive and brings in a substantial modification to FP:
[T]he asymmetric dependence proposal is that all else being equal, breaking the
cow—>[cow] breaks the X—>[cow] for all X.... [W]hat’s wrong with
Block’s argument is that all else isn’t equal in the worlds that he
imagines. To get those worlds, you need to suppose not only that cow—>[cow]
is broken, but also and independently that tree—>[cow] is in force.
It’s this independent supposition that violates the ‘all else equal’
clause.
... If you put in ‘all else equal’, then what the theory requires is not
that cows cause [cow]s in every nomologically possible world where Xs cause
[cows]s. Rather, what’s required is just that there be worlds where cows cause [cow]s
and noncows don’t; and that they be nearer to our world than any world in which
noncows cause [cow]s and no cows do.... [T]he intuitively plausible assumption [is] that
worlds that are just like ours except that it’s the case that cows don’t cause
[cows] are ipso facto nearer to us than worlds that are just like ours except that
it’s both the case that cows don’t cause [cow]s and that trees do. (12)
The supposition in this response is that in the actual world trees never cause
[cows] — unless such connection asymmetrically depends on the cow-[cow]
connection. Which brutally begs the problem. It is Fodor who assumes that tree—>[horse]
is not in force. There is no need to imagine, as Block does, possible but
non-actual worlds in which [cow] means tree. Robustness affirms that trees or other
things, events and processes cause — ’independently’ cause — [cow]
tokens among others. Suppose that the tree-[cow] connection is mediated and
sustained in my own case by a memory of seeing a tree on whose bark a [cow] has been
accidentally or intentionally etched. This guarantees the non-existence of the required
(synchronic) asymmetric dependence, and far from being isolated cases like this are
ubiquitous. Connections between properties and phonological or orthographic sequences are
mediated in all sorts of ways. (13) It matters not at all
whether the ‘all else equal’ clause, as understood by Fodor, is violated, what
matters is that robustness, as described, is the way the world is. I do not stipulate
independent connections; I merely draw attention to how things are. In the actual world
— and so nothing could be nearer — there are [cow]s which don’t mean cow:
the [cow] in [cowage], for instance. Cowage would still cause [cows] even if cows
don’t, because cowage would still cause [cowage] and this orthographic
sequence contains [cow]. Therefore, Fodor’s "intuitively plausible
assumption" stated in his reply is false.
It should have become apparent to the proponent of FP how insurmountable the problems
would be for any causal account couched in terms of tokenings of phonological/orthographic
sequences. The unsurprising thing about them is that they are embedded in many contexts.
Note that the use of "limpkin", for example, depends neither on the use of
"limp" nor on the use of "kin". In general, ‘nomic’
connections implicating the tokenings of semantic or morphemic entities like
"limpkin" do not asymmetrically depend on laws about the tokenings of other semantic/morphemic
entities like "limp" and "kin". Likewise for the corresponding
orthographic sequences. It doesn’t help to add that one is speaking, for example, of
[S]s qua [S]s, (14) purporting thereby to rule out
[SZ]s as involving tokenings of [S]s. "[S]s qua [S]s" simply means [S]s
as brute orthographic tokens. In so far as cases of embedded phonological/orthographic
sequences abound, they pose a serious challenge to FP as a general theory of reference. (15)
There is just no reason to suppose that orthographic individuation coincide with
semantic/morphemic typing; whilst there is every reason to believe that
phonological/orthographic classification cuts across semantic/morphemic classification.
Fodor seems to have presupposed that they do coincide, and so he begged the problem which
lies at the heart of the naturalization program he wants to carry out. (16)
What should really drive one to resist the belief that the Asymmetric Dependence
Condition can be met either in thought (Mentalese) or in public language, are such
examples as: "[cow]s cause [cow]s" is a law (read: the property of being a [cow]
token is nomically connected with the property of being a cause of [cow] tokens);
"[horse]s cause [horse]s" is a law; and so on. That they defy the condition is
so convincing that if one of them behaves in the way the naturalizer wants them to, it can
only be regarded as no more than a fortuitous state of affairs. The [cow]-[cow] connection
does not (synchronically) asymmetrically depend on the cow-[cow] connection, and
one might even be tempted to bet on the proposition that it’s the cow-[cow]
connection which asymmetrically depends on the nomic connection between [cow]s and [cow]s.
I surely do not intend to go that far, but if the proposition were true the naturalizer
would have to say that [cow] means [cow], naturally an unpleasant thing to say. Nor should
it ever occur to us to escape from the present difficulty by considering [cow] to be
ambiguous, by claiming that it means [cow] and cow.
The conclusion holds that Fodorian naturalistic semantics is yet to cure itself of
double disjunctivitis. No indication is felt that relief from within, if not from without,
is forthcoming.
|
