2. Synonymy and a Priori Knowledge

The linguistic view, as Giaquinto understands it, holds that one route to a priori knowledge is to deduce it from synonymies and logical truths that are independently known a priori. Here is one example he considers (I will refer to this general scheme as ‘the synonymy model’):

The idea is that this explains how we can know a priori that every cow is a female bovine (given that we already have some a priori knowledge). As Giaquinto is quick to point out,Footnote ⁶ there are a number of potential issues with such explanations: First, he says, we may doubt whether the logical truth is available a priori.Footnote ⁷ Second, we may doubt whether the relevant semantic knowledge is available a priori.Footnote ⁸ And third, we might worry that claims of the form ‘X means the same as Y’ are typically not strictly true because there is no fact of the matter.Footnote ⁹

Giaquinto discusses the first of these worries in the second part of his paper (which I shall not consider here), and is willing to disregard the second for the sake of argument. His objection in the first part of the paper concerns the third worry.

3. Giaquinto's Objection

In order for synonymy-based explanations of a priori knowledge to work, there must be semantic facts underlying claims such as (2). Giaquinto claims, however, that if we reflect upon certain questions about how to individuate meanings, we shall have to conclude that this requirement is typically not met, and that it is simply indeterminate whether two expressions are synonymous.

To substantiate this, Giaquinto considers the question: Does ‘cow’ (in English) mean the same as ‘vache’ (in French)? Well, we first observe that ‘cow’ is used in certain ways that differ from how ‘vache’ is used. For instance, ‘cow’ is applied to female elephants and whales, whereas ‘vache’ is not.Footnote ¹⁰ This by itself, though, does not establish that the synonymy doesn't hold, for we might well say that ‘cow’ is used in several (somewhat similar) senses in English, but that one of these is synonymous with the French word ‘vache’. What this response makes clear, however, is that:

Giaquinto considers several ways of tackling this question, but argues that neither lexicographyFootnote ¹² nor cognitive semanticsFootnote ¹³ provides any reason to think that there is a determinate answer here.

Though the problem is initially presented in terms of synonymy across languages, this is not essential. In particular, the same worries arise if we replace the French ‘vache’ with the English ‘female bovine’, and if so, premise (2) above is in trouble. Of course, if this phenomenon were peculiar to the word ‘cow’ and perhaps a few others, then it wouldn't be all that significant, but Giaquinto argues that the issue will affect most other – but perhaps not all – cases of (purported) synonymy, including that favourite example: ‘bachelor’/‘unmarried man’.Footnote ¹⁴ The upshot is that it is rarely going to be unproblematically true that one expression is synonymous with another, and so synonymy-based explanations are rarely going to be available.

4. Is Moderate Indeterminacy a Problem for the Linguistic View?

I shall not contest Giaquinto's reasons for thinking that moderate indeterminacy of meaning is a widespread phenomenon. My point concerns, rather, what we should conclude if Giaquinto's claims about this are correct. He thinks it shows that there is something wrong with the linguistic view of a priori knowledge:

But at this point we should ask: ‘examples of what, exactly?’ Surely, if this is to be an effective objection against the linguistic view – which is, after all, a claim about how to explain the a priori – we must be dealing with examples of a priori knowledge. The problem, however, is that insofar as Giaquinto is right about the cases he considers, these aren't examples of a priori knowledge after all.

Take the claim that all cows are female bovines. That might indeed seem like a candidate for a priori knowledge. But if Giaquinto is right, and it is indeterminate whether ‘cow’, as it occurs in the above claim, can be correctly applied to female elephants, then it is equally indeterminate whether all cows are female bovines. Since we can only know what is true, that disqualifies this from being an instance of a priori knowledge. Similarly for the other cases Giaquinto considers: If the meanings are indeterminate in the way he suggests, the appropriate lesson appears to be, not that these are counterexamples to the linguistic view, but that these aren't candidates for a priori knowledge in the first place. What we need is a case of genuine a priori knowledge that cannot be explained via the linguistic route due to semantic indeterminacy, but no such case is presented.Footnote ¹⁶

It is worth noting that Giaquinto is not arguing against the very possibility of a priori knowledge:

The thesis of moderate indeterminacy, then, only suggests that there is less a priori knowledge that we might initially have thought; it gives us no reason to reject the idea that all the a priori knowledge there is can be explained by appealing to knowledge of semantic facts.

It might be objected here that while the moderate indeterminacy thesis does not tell against the linguistic view in general, it does tell against the more specific idea that there is a subset of a priori knowledge that can be explained in terms of knowledge of synonymies.Footnote ¹⁸ That would be true if the thesis established that there are no synonymy facts whatsoever, but Giaquinto is explicit that this claim is too strong. There are some synonymy facts, such as when a term is introduced as abbreviating another expression and perhaps a few others,Footnote ¹⁹ and in these cases, synonymy-based explanations are not threatened by moderate indeterminacy of meaning.

Giaquinto concludes, therefore, that ‘[…] a priori knowledge via knowledge of synonymies is at best very restricted’.Footnote ²⁰ However, this assessment is potentially misleading. It makes it sound as if the synonymy model is at fault for being less explanatory than it ought to be – i.e. that there are instances of a priori knowledge which it ought to account for, but cannot. But, again, this has not been established. Giaquinto certainly considers cases where we might initially have thought that we are dealing with a priori knowledge and the synonymy model should be applicable (e.g. ‘all bachelors are unmarried men’). As we have seen, however, this assumption must be revisited if it turns out that these cases are lexically indeterminate and so cannot constitute a priori knowledge after all.

The upshot, then, is not that the synonymy model fails to explain something it ought to explain, but that the phenomenon it is supposed to account for is less prevalent than we thought. This can hardly be held against the synonymy model. Analogously, the hypothesis that all men are mortal does not suffer if it turns out that there are fewer men than we initially believed.

I maintain, then, that the thesis of moderate indeterminacy of meaning does not give us reason to reject either of the following:

1. i) All a priori knowledge can be explained by appealing to knowledge of semantic facts.

2. ii) All a priori knowledge of a certain kind can be explained by appealing to knowledge of synonymy facts.

5. The Correlation between a Priori Knowledge and Semantic Facts

In this section I shall briefly suggest that, if anything, Giaquinto's considerations speak in favour of the linguistic view since they suggest that there is a tight connection between the possibility of a priori knowledge and the presence of semantic facts.

We saw above that moderate indeterminacy does not pose a threat to the linguistic view since it establishes, in any particular case, not just that there is no semantic fact of the matter, but also that we are not dealing with a candidate for a priori knowledge. In these cases, then, it appears that the semantic facts and the possibility of a priori knowledge stand and fall together. That things turn out this way is no accident, and I want to make it clear that the relevant connection is between determinate semantic facts and a priori knowledge specifically, not just knowledge in general.

When considering the case of ‘all cows are female bovines’, I argued that if the semantic facts are indeterminate in the way Giaquinto suggests, then we cannot even know that all cows are female bovines. However, we can easily tweak things so that we could know this, but not know it a priori due to the semantic indeterminacy. If, for example, all species except humans and bovines became extinct, then it would be true and knowable that all cows are female bovines – even if the meaning of ‘cow’ is indeterminate in the ways Giaquinto suggests. However, given this alleged semantic indeterminacy, this would still not be knowable a priori since, presumably, we would have to ascertain empirically that there are no female whales, elephants etc. What the lack of a semantic fact precludes is therefore not knowledge, but a priori knowledge specifically. Similar considerations apply to the other cases Giaquinto considers.

This suggests a conclusion quite opposed to that drawn by Giaquinto, namely that we can know a priori that all cows are female bovines if and only if there is a relevant semantic fact of the matter. I am certainly not suggesting that this confirms the linguistic view – firstly, because this single case would be weak inductive evidence indeed, but more importantly, because this only suggests a correlation between semantic facts and the possibility of a priori knowledge, and this does not suffice to establish the explanatory relation claimed by advocates of the linguistic view. Nevertheless, if there is such a correlation, this is surely a point in favour of the linguistic view, rather than a point against it.

We should note, finally, that other remarks by Giaquinto provide further evidence for a correlation here. Since he is concerned to argue for a moderate kind of indeterminacy, he is explicit that there are semantic facts:

It is a semantic fact, that is, that ‘cow’ cannot be correctly applied to female spiders. Now, if this is correct (and making the concessions that Giaquinto makes about other problematic aspects concerning the linguistic view), then it would seem that we can know a priori that no cow is a female spider. In general, then, reflection on the issue of moderate indeterminacy of meaning suggests that a priori knowledge requires the presence of relevant semantic facts, and that when such semantic facts are present, so is the possibility of a priori knowledge.