Two Dogmas of Rationalism

Two Dogmas of Rationalism

We may suppose the rationalist to hold two theses: (a) there is an analytic-synthetic distinction, and (b) the meaning of statements can be reduced to the innate ideas expressed or denoted by the terms in them. Since rationalism accords priority to a priori statements, the odd member of the pair in (a) will be the class of synthetic statements (we will suppose for the sake of argument that the distinction between a priori and a posterioristatements maps onto the distinction between analytic and synthetic statements). What kind of truth do they have? Answer: the kind in which the predicate is not synonymous with, or contained in, the subject—that is, it is different in meaning. But how is this notion to be defined? We know well enough what synonymy is—we recognize it easily when we see it and dictionaries are full of it—but it is more problematic to say what it is for words to differ in their meaning. Is it a matter of the words not being intersubstitutable salva veritate? Or is it a matter of there being different ideas in the minds of speakers who use the words? Neither option is very appealing, but nothing else suggests itself. So the rationalist is unable to define the analytic-synthetic distinction, because of the difficulty of saying what difference of meaning amounts to. Maybe this notion should be abandoned and replaced by the idea that all words really have identical meaning: that is a simpler hypothesis and nothing in science contradicts it. Why suppose distinctions of meaning when no such differences can be defined or discerned? So this dogma is best abandoned: there are only analytically true statements, because the notion of differences of meaning cannot be made satisfactory sense of. The concept of synthetic truth is without scientifically definable content.

            Second, the rationalist is committed to the thesis that statements are isolated constructions built from symbols that express innate concepts. This suggests that they can be verified individually, by reference to the individual concepts they contain. For example, we can verify “Bachelors are unmarried males” by consulting our innate ideas of bachelor, unmarried, and male. We need not stray beyond this statement to take into account other statements. But this account of verification is too atomistic; in fact, we can only verify analytic statements by making reference to other connected statements. Thus their meaning is more holistic than has been supposed, given the connection between meaning and verification. For example, when I verify that bachelors are unmarried males I have to understand such statements as “Marriage is a legal contract between consenting adults containing such and such commitments” and “A male is one sex of a species with two sexes characterized by a certain type of anatomy”. And these statements themselves contain words that I have to understand by reference to other statements such as “A legal contract is a binding agreement between people with such and such penalties attached”. And so it goes on indefinitely in a network of interconnected statements. There is no isolated verification of a single analytic statement but rather a holistic verification of a whole set of connected statements. Really it is whole theories (sets of sentences) that are confirmed not statements taken in isolation. So the rationalist dogma of reduction to a fixed set of innate ideas has to be wrong, with each idea attached to the individual words of the statement. Rationalist reductionism must be rejected.

            What lessons can we draw from this little exercise in parallels? First, the corresponding arguments against empiricist “dogmas” are about as plausible as these arguments, which is to say not very plausible (I will spare you the details). Second, the problems with empiricism alleged in “Two Dogmas of Empiricism” have precise counterparts in relation to rationalism, so they in no way favor rationalism over empiricism (not that the author of “Two Dogmas of Empiricism” claimed as much). What we left with, then, if we accept these arguments is an epistemology that is neither empiricism nor rationalism but exists in some undefined epistemological hinterland. In effect, we are left without any viable conception of sentence meaning and without any theory of how knowledge is acquired—except vague talk of networks, holism, and assent behavior. We are left, that is, with the desiccated picture later defended by the author of the original “Two Dogmas”, consisting of physical stimuli eliciting responses holistically without any viable conception of meaning (indeterminacy plus “naturalized epistemology”). Admirers of “Two Dogmas” need to take this consequence into account. I myself think that the considerations sketched above fail to refute any so-called dogmas in either empiricism or rationalism. On the contrary, the analytic-synthetic distinction is not cast into doubt, and no difficulty in sentence-by-sentence meaning has been detected (whatever “holism” exists is consistent with sentences having their meaning determined by the words they actually contain). At any rate, we can construct entirely parallel arguments designed to undermine the “dogmas” of rationalism.  [1]

  [1] It is particularly noteworthy that verification holism (whatever that exactly comes to) can be applied to the acceptance of analytic truths. Holism is not a unique feature of empirical verification. In other words, beliefs always come in groups, more or less extensive, whether empirically justified or based on rational insight (including grasp of what words mean).