Is there synonymy in Ockham's mental language?

Is there synonymy in Ockham's mental language?

David J. Chalmers

Department of Philosophy
University of Arizona
Tucson, AZ 85721.

[[Published in The Cambridge Companion to Ockham, edited by Paul Vincent Spade. Cambridge University Press, 1999.]]

1 Introduction

William of Ockham's semantic theory was founded on the idea that thought takes place in a language not unlike the languages in which spoken and written communication occur. This mental language was held to have a number of features in common with everyday languages. For example, mental language has simple terms, not unlike words, out of which complex expressions can be constructed. As with words, each of these terms has some meaning, or signification; in fact Ockham held that the signification of everyday words derives precisely from the signification of mental terms. Furthermore, the meaning of a mental expression depends directly on the meaning of its constituent terms, as is the case with expressions in more familiar languages.

As one might expect, there are important differences between mental language and everyday languages. For example, mental languages signify their objects naturally rather than conventionally. At a more concrete level, Ockham suggested that numerous features of spoken or written language - participles and pronouns, for example - might not exist in mental language.

Two ubiquitous features of everyday languages are the phenomena of equivocation and synonymy. The first of these is exemplified in English words such as `bank', which have two entirely different meanings. The second is exemplified by pairs of terms such as `bachelor' and `unmarried man', which share a common meaning. The question arises: are these features are also found in mental language? It seems to be commonly accepted that they are not. Ockham himself is not entirely clear on the matter, but Trentman and Spade have argued, based on both textual and theoretical considerations, that the most coherent position broadly compatible with Ockham's work is that neither synonymy or equivocation may occur in mental language.[1] I shall not discuss the status of equivocation in this paper, as I find the arguments on that topic persuasive. However, I wish to argue that the case against synonymy is not as strong.

My argument for synonymy is largely theoretical, although it also has a textual element. The theoretical case has a positive and a negative part. On the positive side, I will argue that there are a number of reasons why mental synonymy might exist, and that a mental language without synonymy would be relatively clumsy, with a number of ad hoc features. On the negative side, I will address various arguments that have been put forward against the possibility of synonymy, and will try to show that they are not conclusive. Textually, I will argue that although Ockham appears to deny the possibility of synonymy in mental language, he also makes remarks that commit him to that possibility. I do not think that my arguments are entirely conclusive, but I hope to demonstrate that the possibility of mental synonymy is not as implausible as has sometimes been thought.

2 Positive Arguments

2.1 Logical Equivalence

The first argument is based on the observation that once certain logical primitives are admitted into a language, then it seems to follow immediately that certain complex expressions are synonymous. Consider, for example, the English expressions `man and (cat or dog)' and `(man and cat) or (man and dog)'. The parentheses are introduced for clarity, but they are not strictly necessary, being straightforwardly replaceable by longer locutions (for instance, `a man with either a cat or a dog' and `either a man with a cat or a man with a dog'). Call the two expressions E1 and E2. It seems apparent that E1 and E2 are synonymous. It is a straightforward logical inference to go from a sentence involving E1 to the corresponding sentence involving E2, and vice versa. It is plain that the two expressions come to exactly the same thing in the matter of signification.

As it is with written language, so it should be with mental language, at least in this case. Mental language presumably has among its simple terms such logical operators as `and' and `or'; if it does not have these, then it presumably possesses others with equivalent power, on the basis of which a similar case could be made. There is no doubt that simple expressions such as `dog' and `cat' exist in mental language (if not these, then any simple nouns will do). And there seems to be no reason why these expressions and logical operators should not be combined to form complex expressions. So mental language contains the expressions `man and (dog or cat)' and `(man and dog) or (man and cat)', or other expressions that make the same point. These expressions, it seems, must be synonymous, for just the reasons given above. Therefore there is synonymy in mental language.

There are two paths that an opponent of mental synonymy might take in countering this argument. She could argue that the two expressions are not in fact synonymous, or she could argue that the two expressions are in fact identical in mental language, so that no synonymy between different expressions need be introduced. I will consider these objections in turn.

2.1.1 The First Strategy

In the Summa logicae,[2] Ockham says that for two terms to be synonymous, they must signify exactly the same things, and they must signify these in the same way. An opponent of mental synonymy might argue that although the two expressions given above signify the same things, they signify them in different ways, and so are not synonymous.

It is not entirely clear what Ockham meant by `to signify in the same way'. Spade argues that two mental expressions signify in the same way if and only if they are syntactically equivalent - that is, if and only if the consist of exactly the same categorematic expressions in exactly the same syntactic constructions.[3] If this were the case, then our two given expressions would certainly not be synonymous. However, it seems to me that this construal of synonymy buys the conclusion entirely too cheaply, by defining mental synonymy out of existence, and I will argue against it below. For now, let us consider a weaker construal of `in the same way'. This construal captures our intuition that what is required of synonymy is not just that two terms signify the same things, but that they must signify the same things. To be more precise, under this construal of synonymy, two terms are synonymous if their equivalence is a priori and necessary.

This yields a strong criterion for synonymy. The coextensive pair of terms "renate" and "cordate" fail the test, for example, as they are not necessarily coextensive. The coextensive pair of terms "Hesperus" and "Phosphorus" fail the test, as they are not a priori coextensive. It is difficult to know whether this criterion corresponds to Ockham's intentions, but it seems to capture a large part of our intuitions behind what is meant by `synonymy, and in particular to capture the extra strength carried by synonymy compared to mere coextensiveness. And the logically equivalent pair of terms given above clearly satisfy this criterion.

(One might also invoke a criterion according to which synonymous terms are those that are substitutable salva veritate even in modal contexts. I do not use this criterion here, as it yields counterintuitive results when combined with the contemporary understanding of necessity: for example, "Hesperus" and "Phosphorus" come out synonymous. But on the medieval understanding of necessity, this criterion may yield the same results as the criterion given above, and less anachronistically.)

An opponent might try to argue that this criterion is not strong enough: although the logically equivalent pair of terms satisfies the criterion, the terms nevertheless signify in subtly different ways. For example, our opponent might argue that the two terms are not intersubstitutable in certain epistemic contexts, and are therefore not synonymous. For example, it might be the case that John believes that to gain admission to the party, one must be accompanied by a man with either a dog or a cat; but John does not believe that to gain admission to the party, one must be accompanied by either a man with a dog or a man with a cat. The reason for this, presumably, would be that John is not very capable at logic, and so has not made the straightforward inference. Now, we might respond to this by arguing that in fact, if John has the first belief, he has the second belief whether he knows it or not. He believes it implicitly, one might say. But for the sake of argument let us accept that John has the first belief but not the second. If this is so, then it is apparent that in this epistemic context, the two terms are not intersubstitutable.

It seems to me, however, that intersubstitutability in such epistemic contexts is too strong a condition to require for synonymy.[5] If we admit this as a criterion, it would seem to follow that there are no synonyms at all. For given any pair of purported synonyms - e.g., `bachelor' and `unmarried man' - we can construct examples like those given above. For example, John might believe that all bachelors are invited for dinner, but might not believe that all unmarried men are invited for dinner - presumably because he is once again too muddled to make the required inference. (Again one might argue that if John has the first belief, he has the second whether he knows it or not. But as we disallowed this line of reasoning for the sake of argument in the paragraph above, we must equally disallow it here.) This conclusion shows that the suggested criterion is too strong. In fact, it seems that even Ockham himself would reject intersubstitutability in epistemic contexts as a criterion for synonymy, as the following passage demonstrates[6]:

Those synonyms are broadly so called even though not all users believe them to signify the same [thing] but rather, under a deception, they judge something to be signified by the one that is not signified by the other.

It follows that this argument provides no reason to reject the synonymy of E1 and E2, or at least no more reason than there is for rejecting the synonymy of such terms as `bachelor' and `unmarried man'. Logical synonymy seems to be at least as strong as common-or- garden conceptual synonymy, so that if we want to retain the notion of synonymy at all, we had best admit the possibility of logical synonymy, at least in written language.

Once we have accepted the possibility that the two written expressions E1 and E2 are synonymous, then it follows that either (a) they are subordinated to synonymous mental expressions, thus establishing my case, or (b) they are subordinated to identical mental expressions, thus saving the same for the opponent of mental synonymy. I will consider possibility (b) next.

2.1.2 The Second Strategy

An opponent's second strategy is to argue that only one expression in mental language is involved here: the two mental expressions that we are labeling `man and (cat or dog)' and `(man and cat) or (man and dog)' are in fact the same expression. It might be claimed that the fact that they look like different expressions is an artifact of English, not mental language: the different written expressions E1 and E2 are in fact subordinated to the same mental expression. To get an idea of how this argument might run, consider a simpler pair of expressions: `man and dog' and `dog and man'. For reasons similar to those given above, these expressions seem to be synonymous. But an opponent of synonymy could very plausibly argue that these are not different expressions in mental language at all, as the only difference between them is one of word order - a feature that need not be preserved in mental language.[7] For example, it could well be argued that the term `and' in mental language is not interpolated sequentially between two terms in a particular order, but rather is bound to them both in some symmetrical fashion, such as a tree structure, or perhaps as a common ending (something like `man-and dog-and', where we take it that the two words are unordered). On this plausible account, the difference in the two expressions `man and dog' and `dog and man' would not persist into mental language.

It is difficult to see, however, how we could apply an argument like this to more complex cases, such as our `man and (dog or cat)' vs. `(man and dog) or (man and cat)'. These two expressions are not just superficially different; they are structurally different. One of them is a conjunction; the other is a disjunction. However we represent `and' and `or' in mental language, they will need to be put together hierarchically on occasion, and it seems clear that the differences in their hierarchical order must be represented in mental language.

One way out for our opponent might be to argue that logical expressions in mental language must be reduced to some common form such as disjunctive normal form. It would then be the case that any two logically equivalent terms or propositions would be represented identically. The trouble with this is that is an ad hoc restriction that places limits on how mental terms may be combined. On the face of it, would take it that if `man', `dog', `cat', `and' and `or' are admissible mental terms, then `man and (dog or cat)' would be an admissible expression in mental language; but according to the present proposal this expression would somehow be debarred from being formed, for no apparent reason other than to preserve a theoretical claim. Indeed, to place such a restriction on combination of terms would seem to reduce mental language's claim to being a language in the first place, one of the key properties of language being that complex expressions can be generated recursively and compositionally, whereas according to this proposal, while `man' and `dog or cat' would be valid mental expressions, `man and (dog or cat)', their conjunction, would not be. This seems arbitrary and implausible.

It seems much more plausible for mental language to consist of a set of simple terms that can be combined and recombined without any limits other than syntactic requirements. In particular, logical operators ought to be able to combine nominal terms in any combination. But once we admit this, no matter what set of logical operators we choose, logically equivalent but non-identical expressions will be able to be formed. We might be able to get around certain superficial differences in structure, such as that in the `man and dog' case, by proposing certain symmetries in our operators, but there will always be propositions with different structures that turn out to be logically identical. One can see this easily by noting, for example, that while there are only four logically distinct truth-functional combinations of two propositions, any compositional system will yield an infinite number of syntactically distinct complex sentences formed by combining these propositions under truth-functional operators. Logical equivalence is thus endemic in any combinatorial system.

The conclusion is that given that logical equivalence implies synonymy, it follows that synonymy will be ubiquitous in mental language, unless we put ad hoc restrictions on the manner of combination of mental terms.

Before passing to the next topic, I should respond to a natural response, based on textual considerations, that may have occurred to the reader. This response notes that according to Ockham, synonymous written expressions are subordinated to the same mental expression,[8] so that no matter what I say above, the expressions E1 and E2 must be subordinated to a single expression in mental language. My response to this objection is of course to point out that the arguments I am giving for the existence of synonymy in mental language are equally arguments for the rejection of the claim that synonymous written expressions are subordinated to the same mental expression. If there is synonymy in mental language, a more plausible criterion is that synonymous written expressions must be subordinated to synonymous mental expressions. Of course, this means that I am going against an explicit claim of Ockham's; but I am giving a theoretical argument, after all. I will address this issue further below.

2.2 Conceptual Change

I now pass to a quite different line of argument in favor of mental synonymy. This argument stems from the observation that many terms in everyday languages, and presumably in mental language, undergo gradual changes in meaning. Many current terms do not mean exactly what they meant fifty years ago. Even within a single individual, the meaning of a term can gradually shift over time. Sometimes it may even occur that two terms that originally had quite different meanings gradually drift until the mean the same thing. Perhaps, for instance, when one was a child one used the term `stone' only for small objects, and used the term `rock' for large ones, but over time one's use of the term has drifted until now they are used in exactly the same way, so that the terms are synonymous. Nothing in my argument turns on the specific example - all that is required for the argument is the possibility of this kind of graduate drift into synonymy taking place, and this seems indisputable.

So let us say that Y and Z are two terms that start out with different meaning, but which through a continuous process over time come to mean the same thing. This might even occur without the individual within whom it is happening being conscious of the drift to synonymy. He might realize at some later point that `stone' and `rock' are synonymous for him, whereas they were not twenty years ago, and the drift might have occurred without his realizing it at the time. Ockham himself endorses the possibility that synonymy might go unrecognized, in the passage quoted above.

Now as these terms gradually drift toward synonymy, what is happening in the mental language? Presumably Y and Z are subordinated to mental terms Y' and Z', and Y' and Z' are also undergoing a slow drift in meaning. Now what happens on the day when Y and Z finally become synonymous? Do the terms Y' and Z' suddenly become the same term? Does one of them suddenly disappear, so that for example Y and Z are now both subordinated to Y', and Z' is gone? Neither of these possibilities are strikingly plausible. The process of change is sufficiently slow that is hard to imagine that any sudden change could be taking place in mental language. In practice, it will be very hard to locate a precisely moment at which Y and Z become synonymous; there may be a long period of approximate synonymy before we can say with confidence that they are definitely synonymous. Are we to suggest that the mental language has hair-trigger sensitivity to the process, so that at the exact moment when the meanings come to coincide, a jump suddenly occurs?

It seems at least as plausible to argue that as the drift takes place, the two mental terms remain distinct, gradually becoming synonymous but nevertheless oblivious, at least for a while, to each other's presence. It may well be for a long period that the subject does not make the expected inferences from propositions involving Y to those involving Z, because the subject does not consciously realize that the two concepts have become synonymous for him. If the corresponding mental concepts had become identical, this inferential gap would be harder to explain. After a time, the subject may consciously realize that the concepts are synonymous, and perhaps the mental terms will then be "fused" into one, but until then there seems no reason in principle why the terms should not be distinct.

The other possibility to be considered is that as Y and Z drift in meaning, they are subordinated to a series of different terms in mental language, Y'1, Y'2, ... and Z'1, Z'2, ... Upon every change in the meaning of Y, no matter how small, Y becomes subordinated to a new Y' in the mental language. Over time, the corresponding Y' terms gradually become more like the Z' terms, until finally the become identical. This account has the virtue of avoiding the sudden change required by the previous story, but it still seems problematic. For a start, it seems to be very extravagant with terms in mental language, postulating a large number of different terms where we previously needed only two. Further, it does not solve the problem mentioned above, of the possible inferential gap in the subject's abilities - one would think that if the corresponding concepts have become identical, any inferences from one to the other would be automatic. This seems in turn to imply that two terms cannot become synonymous without a subject recognizing that fact, which appears to contradict Ockham's own claim quoted above.

On the balance of things it seems to me that conceptual change provides a strong argument but not a knockdown argument for mental synonymy. The opposing story just given seems feasible enough that it could be the case, although it does raise a serious question about compatibility with Ockham's own views on the recognition of synonymy, an issue I discuss further below. However, the story that I sketched of gradual drift into synonymy without identity seems equally if not more plausible, giving more reason why mental synonymy is not an altogether unreasonable idea.

2.3 Efficiency

My final argument in favor of mental synonymy is a pragmatic one. Ockham, in his writings, give very little idea about how mental language functions - how it enables one to make inferences from one proposition to another, for instance, or how mental propositions help determine the actions we take toward the world. We can imagine, however, that this functioning does not come for free, but rather takes some work. Evidence for this can be seen in the fact that we find it harder to make inferences from complex propositions than from simple propositions. The more complex a proposition, it seems, the harder it is for our mental system to deal with. Therefore one might draw the conclusion that in order for our mental system to function as well as it can, mental propositions should be as simple as possible (as long as they are complex enough to express the required meaning).

Now in mental language, presumably, there are certain complex expressions that get used repeatedly.[9] In the interests of efficient functioning of the mental system, it might seem to be advantageous for it to introduce interval abbreviations for these complex expressions. Instead of having to deal with expressions like `man who cuts cloth and makes suits', the system would only have to deal with the simple expression `tailor' instead. We certainly find this kind of abbreviation useful in our external practice; I am here suggesting that it might have a role in internal functioning as well. If, as we have supposed, complex expressions are more difficult for the mind to deal with, then the systematic replacement of these by simple expressions might allow a significant enhancement to our cognitive capacities.

Of course, there would have to be systematic links between a term and its abbreviation, so that thoughts about `tailor' could easily lead back to conclusions about suits and cloth when necessary. But inferences involving tailors would in general be much easier to make in the above form. The only downside is that we would have introduced synonymy into mental language, but it seems to me that in this context this is only advantageous. A typical argument against mental synonymy[10] has the form "Who needs it?" - in other words, what is the point of mental synonymy if synonyms are truth-conditionally equivalent and so have the same expressive power? Any distinction between synonyms would be a difference that makes no expressive difference, and so would be unnecessary in mental language. Here, we have seen that pragmatic considerations show that there might be relevant differences that are not expressive differences. Rather, they are differences that aid mental function.

This is obviously not a knock-down argument. For a start, it may be anachronistic in focusing on the way that mental propositions are processed, a way of thinking that was not so common in Ockham's time. Nevertheless, all that is required to see the point is that complex mental propositions might be more cumbersome to deal with than simple ones, and this does not seem to be a particularly advanced consideration. Of course, the argument is only a plausibility argument demonstrating why mental synonymy might be a reasonable thing, and as such is a weaker argument than two arguments given above, but nevertheless it helps in breaking down the intuition that mental synonymy would be an entirely useless thing.

3 Textual Considerations

I now consider some textual issues. In particular, I must confront what seems to be the strongest argument against synonymy in Ockham's mental language: the fact that on at least on occasion, Ockham appears to deny that mental synonymy exists. This occurs in his Quodlibet V.812:

To the principal argument, I say that everything that is an accident of a mental term is an accident of a spoken term, but not the other way around. For some [things] are accidents of spoken terms because of the necessity of signification and expression, and they belong to mental names. Others are accidents of spoken terms for the sake of the decoration of speech (like synonyms) and for the sake of well-formedness, and they do not belong to mental terms.

Now, it is not entirely clear that this is a blanket denial of the possibility of mental synonymy. It might alternatively be interpreted as a denial that synonymy in spoken language is reflected in mental synonymy, leaving open the possibility of mental synonymy that arises in other ways. Or it might be interpreted as the denial that certain synonyms in spoken language - those that exist merely for the sake of decoration - are reflected in mental synonyms, leaving open the possibility that other (nondecorative) spoken synonyms correspond to mental synonyms. Furthermore, even if we accept this as a denial of the possibility of mental synonymy, it seems to be the only occurrence of such a denial in Ockham's writing, so that even Spade (who argues against synonymy in mental language) concedes that the textual support is not as strong as it could be.13

In any case, my central purpose is to make a theoretical argument about what Ockham should have held, rather than about what he in fact did hold. In other words, insofar as Ockham can be seen to have denied the possibility of mental synonymy, this paper should perhaps be taken more as a criticism of Ockham than as an interpretation. Of course this is a somewhat delicate matter, as if one jettisons too much of Ockham's theory one runs the risk of not so much criticizing him as ignoring him. Nevertheless, I believe that the position I am putting forward is compatible with Ockham's overall thrust, and only requires the rejection of one or two specific claims. And I argue below that Ockham's claims on this matter appear to be inconsistent, so that any interpretation must reject some of them. So this interpretation may not be worse off than any other.

The other specific claim of Ockham that causes problems for mental synonymy is his statement on a number of occasions that synonyms in spoken or written language are subordinated to identical terms in mental language. As Spade points out,[14] it does not necessarily follow from this that mental synonyms cannot exist; all that follows is that if they do exist, they do not have associated synonyms in spoken or written language. However, this seems too weak a ground to base a defense of mental synonymy on. For a start, if our mental synonyms had any written or spoken terms subordinated to them, these terms would be synonyms, in violation of Ockham's claim; and on the face of it it seems reasonable that any given mental term should at least possibly have an associated written or spoken term. If we were to retain Ockham's claim and to argue for mental synonymy, we would be committed to the existence of mental terms that could not have associated spoken or written terms, and while this is not impossible it at least seems ad hoc and unmotivated. Furthermore, my best evidence for the existence of mental synonymy[15] came precisely from the consideration of synonymous spoken or written expressions that I argued had to be subordinated to different mental expressions.

For these reasons, it is best for the defender of mental synonymy to argue that Ockham's claim should be jettisoned along with his denial (insofar as it is a denial) of mental synonymy. The correct criterion for synonymy of spoken or written terms is that they should be subordinated to synonymous mental terms, rather than to identical mental terms. This jettisoning of Ockham's claim may reduce this account's chances of being an interpretation of Ockham, but I believe that it does not reduce its overall plausibility.

I now come to the internal tension within Ockham's text. This casts further doubt on the subordination of synonymous terms to identical mental terms, and indeed suggests that Ockham may be committed to the possibility of mental synonymy, despite what appear to be claims to the contrary. In a passage quoted earlier, Ockham says:[16]

Those synonyms are broadly so called ... even though not all users believe them to signify the same [thing] but rather, under a deception, they judge something to be signified by the one that is not signified by the other.

Ockham goes on to point out that this is the sense of synonymy with which he is dealing for most of his text. According to this passage, then, it is possible that a user can judge that `That object is an X' is true, and at the same time judge that `That object is a Y' is false, even thought the terms X and Y are synonymous. But this implies that X and Y cannot be subordinated to the same term of mental language! For if X and Y were subordinated to the same mental term, then all mental judgments about X's and Y's would coincide. The two propositions `That object is an X' and `That object is a Y' would be subordinated to identical propositions in mental language, and it is impossible that one could be judged to be true and the other simultaneously be judged to be false. Therefore this passage seems to contradict his claim that synonymous terms are subordinated to identical mental terms.

In fact, if we accept the above passage at face value, it is hard not to draw the conclusion that the terms X and Y must be subordinated to different but synonymous mental terms (as it is surely impossible that synonymous spoken or written terms are subordinated to nonsynonymous mental terms), which would directly establish that synonymy exists in mental language.

It is difficult to say how we should deal with this contradiction, or how Ockham would have dealt with it, had it been pointed out to him. He might have chosen to retract the claim made in the passage above; the opposing claim that synonymous terms are subordinated to identical mental terms is certainly made more often. On the other hand, he might have seen the possibility of unrecognized synonymy in spoken or written language as providing grounds to reject the claim about subordination of synonymous terms. In any case, the internal tension revealed here further weakens the plausibility of the claim that synonymous terms are subordinated to identical mental expressions, and indicates that the claim might be rejected without doing too much violence to the rest of Ockham's system.

Furthermore, insofar as Ockham's own claims are inconsistent, it follows that any consistent interpretation must reject one or more of these claims. So Ockham's apparent denials of mental synonymy do not provide any overwhelming reason to reject the view I have offered in favor of any other consistent interpretation. This opens the way to accepting the possibility of synonymy in Ockham's mental language.

4 Other Negative Arguments

The central negative argument against the possibility of mental synonymy is the textual argument I have just discussed. Three other negative arguments deserve consideration, however. These are (1) the argument by analogy with Ockham's treatment of grammatical features; (2) Spade's argument that signification "in the same way" requires syntactic identity in mental propositions11; and (3) the argument from the fact that concepts bear a "natural likeness" to their objects.

4.1 The Analogy With Grammatical Features

A significant argument against the existence of mental synonymy derives from Ockham's criterion for determining which grammatical features do and do not persist into mental language. As Spade puts it[17]:

... grammatical features of spoken or written language that do not serve the "needs of signification" by affecting truth conditions are not present in mental grammar. That this rules out all synonymy inn mental language seems to be the clear intention of Ockham's whole discussion in Quodlibet 5, q. 8, with its repeated statement that what is in mental language is there only because of the "needs of signification", not for the sake of "decoration" or "well-formedness", and that synonymy does not serve the "needs of signification".

For example, such grammatical features as gender of nouns and the difference conjugations of verbs do not persist into mental language, as they are irrelevant to signification - that is, they have no effect on truth conditions. Such features would be unnecessary in mental language. If they existed there, they would have the status of mere ornaments. It seems plausible that the differences between any synonymous expressions in written language would be equally unnecessary in mental language. Such differences would therefore not persist into mental language, implying that that mental language has no synonymy.

As an initial reply, one might point out that it is not unreasonable to suppose that there could be other relevant needs besides the "needs of signification." For example, I argued above that synonymous mental expressions might sometimes serve the needs if efficiency in mental language, by providing a simpler expression for complex thoughts. This would seem to be a more relevant reason for the presence of synonyms than mere "decoration." Grammatical features such as gender and different conjugations, of course, would not serve even this purpose (in fact they would achieve the opposite, by making mental expressions more complex than necessary), and so would not make it into mental language, but there are conceivably other instances of synonymy that would qualify.

As perhaps a more compelling reply, one might argue that certain instances of synonymy in mental language are not there because they serve any particular needs, but rather because they must exist as a byproduct of other properties of mental language. Logical synonymy would be one example of this. If we assume that mental language has a need for certain primitive logical operators, and for the ability to combine any expressions according to these operators, then as we saw above we are forced to the conclusion that some complex expressions must be synonymous. This fact in itself does not serve any particularly useful purpose for mental language; rather, it is a consequence of other facts about mental language.

This leads us to another line of reply to the argument above. We might argue that Ockham's criterion should apply to the determination of those grammatical features that persist into the lexicon - that is, the set of simple terms - of mental language, but that once we have determined the simple terms, then complex expressions should be derived directly from these according to combinatorial principles, just as English sentences are derived from words according to such principles. So although no two lexical items in mental language might be synonymous, it could nevertheless be the case that certain complex expressions could be synonymous. To stipulate that two apparently quite different but synonymous complex expressions should in fact be identical in mental language would require ad hoc tinkering with the structure of mental language, as we saw above. By contrast, the elimination of irrelevant grammatical features from simple terms in mental language is not ad hoc at all, precisely because it can be achieved by a correct specification of the elements of mental language at the basic level.

Of course, if we accept this argument against lexical synonymy, it would follow that the argument from conceptual change, above, might have to be rejected. Although I am less convinced by the conceptual change argument than by the logical equivalence argument, I should nevertheless point out that a similar defense of it could be mounted. As with logical synonymy, the synonymy of two mental concepts that have drifted together is not in mental language to serve any need. Rather, it is there as a simple byproduct of the process of conceptual change, and of the way that conceptual change works. And the distinction between two synonymous terms might not be immediately eliminated from mental language, as irrelevant grammatical features are, because the synonymy might not be immediately apparent. If two independent concepts have by coincidence drifted together, there might be no obvious marker of their synonymy, so we might expect it to take a while before mental language became sensitive to the fact of their common meaning. The existence of synonymous mental terms would not serve any purpose in mental language, and we could expect that the terms might be merged at some future time, for instance when the subject becomes aware of their synonymy. Until that time, however, the synonymous terms might both exist, with their coexistence being a byproduct of the terms' divergent histories.

4.2 Signification "in the same way"

As we saw earlier, Ockham's criterion for the synonymy of two terms is that they not only signify the same things, but they signify those things in the same way. It is not clear what Ockham meant by signification "in the same way," but Spade has presented a construal of this phrase that, if correct, would immediately destroy any possibility of synonymy in mental language.[18] It is therefore vital that we come to grips with this interpretation of Ockham's criterion.

Spade's suggestion is that a "way" of signifying should be interpreted syntactically[19]:

A mental expression or concept signifies a thing x in a given syntactic mode m if and only if x is signified by some constituent non-complex categorematic term occurring within that mental expression in the grammatical or syntactical construction m.

If this is the case, then any two synonymous expressions must be made up of identical simple categorematic concepts, in identical syntactic constructions. In other words, they must be identical. It follows that there can be no synonymy in mental language.

The ease with which this argument buys its conclusion is suspicious. It seems to define mental synonymy out of existence; but we have already seen plenty of cases where the idea of mental synonymy at least seems coherent. It may be for a variety of reasons that there turns out to be no synonymy in mental language, but this does not seem on the face of it to be an analytic truth. The case discussed earlier - that of synonymous, logically equivalent but distinct propositions - seems at least to be a possible example, even if certain facts about the way mental language functions might indicate that it is not actual. Therefore on the face of it it would seen unlikely that mental synonymy can be debarred definitionally.

Spade's argument for this definition is not entirely clear to me. He argues that a "way of signifying" should not be construed as one of Ockham's four "modes" of signification in S. log. I.33.20. He then leaps from this claim to the claim that a "way" should be construed purely syntactically, but the grounds for this transition are not obvious. The closest thing I can find to an argument for this conclusion is the following (in the context of a discussion of why `blind' and `sight' are not synonymous, despite both signifying sight)[21]:

Now, in the case of a nominal definition, or indeed of complex expressions generally, it is relatively clear what it might mean to signify a thing x "negatively". It could mean that the expression as a whole signifies x in virtue of some constituent non-complex categorematic term which signifies x and which occurs within the scope of a negation- sign in that expression. This is, in part at least, a syntactic criterion.

But it does not seem to me that "negative" signification must be construed purely syntactically. It does not seem unlikely that a semantic characterization could be arrived at. For instance, there might be a characterization that exploited the fact that although `blind' and `sight' signify the same things, expressions in which they are embedded have opposite truth-values (for instance, `James is blind' is true precisely when `James is sighted' is not true). This might well be more along the lines of what "signifying in the same way" comes to. Indeed, it is interesting to note that on the later "adverbial" theory of signification (used by Peter of Ailly, among others[22]), the same locution - that is, the locution of signify "in a given manner" - is used for that part of signification (of a proposition, in this case) from which truth-values derive. Perhaps this is not entirely coincidental. If we are to speculate, it does not seem unlikely that Ockham might have had a similar semantic criterion in mind, although it is quite possible that he never formalized the criterion, leaving it at the level of intuition.

Furthermore, even if we accept Spade's argument in the quoted passage, it seems to fall short of establishing that signification "in a given way" is a purely syntactic notion. All that is established is that "negative" signification - just a small aspect of what counts in determining a "way" of signifying - might have a criterion that is at least partly syntactic. This seems too weak to establish the broad conclusion suggested by Spade when he states,[23]: "The considerations of section VII, above, suggest that ... the `ways' at stake here are ... syntactic modes of signification."

It seems to me that there is something wrong with the idea that there are syntactic criteria for a notion such as synonymy, which is a deeply semantic notion. It is more plausible that synonymy ought to be characterized purely semantically, in terms of the relationship between elements of language and their actual or possible referents. Looking at Ockham's criterion for synonymy, it seems to me that this is what he may have been doing, although he left the crucial strengthening clause ("signify in the same way") somewhat vague. Nevertheless it seems intuitively plausible that there is an extra semantic criterion required of synonymous terms, over and mere coincidence of signification. This can be captured by the modern idea that this coextensiveness is necessary and a priori. Ockham may not have thought about the issue in explicitly this way, but the intuitive notion is clear enough. I would therefore suggest that this is what Ockham's strengthening clause in the criterion for synonymy should come to.

It seems to me that if we accept the claim that there is no synonymy in mental language, then Spade's syntactic construal of the strengthening criterion will be correct in practice - but this will be precisely because synonymy does not exist, so that a syntactic check will be sufficient to check for identity of meaning. It seems to me further that there are no compelling grounds, other than the nonexistence of synonymy, for supposing the syntactic construal to be correct. If so, then it follows that the syntactic construal cannot be used to provide support for the nonexistence of synonymy.

A further argument in favor of a semantic construal of the strengthening criterion is that it satisfies our intuition that mental expressions corresponding to `man and (dog or cat)' and `(man and dog) or (man and cat)' are synonymous whether or not they are identical. For all these reasons, I believe that the case for the syntactic construal of "ways" of signification is at best inconclusive, and does not provide compelling evidence against the possibility of synonymy in mental language.

While on this topic, I should note that a semantic criterion for synonymy would answer another argument against mental synonymy, also given by Spade[24]:

What would equivocation or synonymy in mental language amount to? Since there is no supramental language to appeal to in the way one appeals to mental language to account for synonymy and equivocation in spoken and written language, how could it even arise in mental language?

This argument seems to assume that the only real criterion for mental synonymy could be appeal to some higher language25 - that is, the criterion must be a formal criterion. It seems to ignore the possibility of a semantic criterion, in terms for instance of referents and/or truth-values. If we have a semantic criterion for synonymy, the problem posed here by Spade is no problem at all.

4.3 Natural Likenesses

The final argument rests on Ockham's contention that concepts bear natural likenesses to their objects. Spade argues that it is unlikely that two concepts could bear a natural likeness in the relevant sense to their objects, but still be more than numerically distinct.[26] If this is correct, it would follow that mental synonyms could not exist.

The reply to this argument is simply to say that it does not seem too implausible that distinct mental terms could bear natural likenesses to a common object, especially if they are complex expressions rather than simple terms. Going back to our favorite example, it seems plausible that the two concepts `man and (dog or cat)' and `(man and dog) or (man and cat)' both bear a natural likeness to their object, despite the apparent difference in the expressions. These two mental expressions seems to differ in form or structure, but they are nevertheless concepts of exactly the same things. It might be difficult to imagine how two distinct simple terms could bear the relevant likenesses simultaneously, but this difficulty vanishes in the case of complex expressions. So this does not seem to be a compelling objection to mental synonymy.

5 Simple Connotative Terms

Before concluding, I should note the possible effect of the discussion above on the question of whether there exist simple connotative terms in mental language. Spade has argued that simple connotative mental terms cannot exist, as it they did, they would be synonymous with their expanded nominal definitions[27]; but of course there is no synonymy in mental language! Panaccio has pointed out another possibility: that simple connotative terms exist, but in fact are not synonymous with their nominal definitions.[28]

The present discussion raises a third possibility: that simple connotative terms exist, and are synonymous with their nominal definitions. Spade and Panaccio reject this possibility because of the supposed impossibility of mental synonymy; but if we accept the above arguments for the possibility of mental synonymy, then this becomes a live option. Indeed, once we have gotten over the hurdle of mental synonymy, it may even fit certain textual evidence better than either of these claims, as Ockham certainly claimed that simple connotative terms exist, and there are strong reasons to believe that connotative terms are synonymous with their definitions (although Panaccio's argument may have weakened this evidence somewhat).

The discussion I have given here even provides a reason why simple connotative terms might exist: namely, to act as abbreviations for their nominal definitions, for the sake of making mental language less cumbersome. This is perhaps the reason why many connotative terms exist in everyday languages. It does not seem wholly unreasonable that they might exist in mental language for the same purpose.

Of course, I do necessarily wish to maintain that Ockham actually held this position, as the existence of mental synonymy seems to contradict textual evidence, at least as the evidence has traditionally been interpreted. Nevertheless I should point out that this instance of mental synonymy is less difficult to reconcile with textual evidence than the other instances we have looked at. This is because we can maintain the synonymy between simple connotative terms and their nominal definitions in mental language without giving up Ockham's claim that spoken or written synonyms are subordinated to identical mental expressions. Instead, we can hold that a spoken or written connotative term and its nominal definition are both subordinated to the corresponding term (the complex expression) in mental language. The simple connotative mental term is introduced entirely within the mental sphere, for the purposes of abbreviation. No spoken or written term is subordinated to it, as it is solely a construct introduced for mental efficiency, just as connotative terms in spoken or written language are frequently introduced for efficiency.

While this position may or may not be independently plausible, we should note that it seems to be at least as compatible with the textual evidence as Spade's and Panaccio's accounts. The only claim of Ockham that it might explicitly contradict is the passage from Quodlibet V.8,[29] and even that passage might be interpreted in a manner compatible with the position maintained here. In any case, whether or not this position was actually held by Ockham, it seems quite attractive as a theoretical possibility.

6 Conclusion

I have done my best to act as an advocate for the possibility that synonymy could exist in mental language. Now is the time to pass considered judgment.

First, the question: did Ockham believe there is synonymy in mental language? To this I believe the answer is probably no. For a start, he seems to deny it on one occasion. Furthermore, as we have seen, two of my arguments for synonymy have required overturning a claim that Ockham made a number of times, that is, the claim that synonyms in spoken or written language are subordinated to identical terms in mental language. On the other hand, the third argument, the argument from efficiency, is independent of this claim, as we can suppose that abbreviation of complex expressions could be entirely internal to the mental system; and this argument also gains some additional textual support from Ockham's otherwise anomalous claim that simple connotative terms exist in mental language. Still, overall, it seems most plausible to me that Ockham was at least implicitly committed to the nonexistence of mental synonyms, although we have seen that internal tensions arise in his theory due to this commitment.

Next, the question: should Ockham have believed that there is synonymy in mental language? To this, my answer is less clear. I believe that there is a very strong argument for the possibility of mental synonymy, in the argument from logical equivalence. As far as I can tell, Ockham never considered this argument. If he had, it is possible that it might have caused him at least to have some doubts abut the impossibility of synonymy. Even if he had eventually come down against the possibility, he might have been required to modify his theory in some serious way, or at least to make a number of its features more explicit. As for the second and third argument, those from conceptual change and from efficiency, I believe that neither of these is conclusive, but that both provide some added plausibility for the notion of mental synonymy, by demonstrating how it might come abut and even perhaps serve a useful purpose. Finally, Ockham's belief that synonymy in spoken or written language can go unrecognized seems to yield the existence of synonymous mental terms as a natural consequence. This provides further grounds for thinking that whether or not Ockham accepted synonymy, he ought to have.

Of the negative arguments (apart from the textual considerations), I believe that the argument from analogy with grammatical features has been cast into doubt, by showing how certain instances of synonymy might exist without having to serve any need, but rather because they are byproducts of other features of mental language. The argument from signifying "in the same way" seems to me to be quite inconclusive, due to the difficulty in ascertaining just what Ockham meant by his phrase. It does not seem too unlikely that he had a semantic criterion in mind, rather than a syntactic criterion. It is possible that the final argument, from natural likenesses, could be made into a strong argument against mental synonymy with some work. I have been concerned to reply to the argument as it is given by Spade, but it is not impossible that stronger versions of the argument exist. Still, such arguments would have to deal with the apparent existence of complex expressions that are distinct but clearly have the same meaning. If some version of natural likeness theory had the consequence that these expressions are not synonymous, then it might not be unreasonable to suggest that that version of the theory should be thrown out, rather than the possibility of synonymy.

Overall, I must adjudicate the case "not proven," but with some strong evidence in favor of mental synonymy that must be dealt with before a retrial. Whichever way the verdict comes down, it seems to me that the possibility of mental synonymy is not as objectionable as has commonly been supposed.


0 This paper was written when I was a graduate student in Paul Spade's class on Medieval Logic at Indiana University in 1991. I owe an enormous debt to Spade for his insights and encouragement.

1 Trentman 1970; Spade 1980.

2 S. log. I.6 (OPh I.19.8-14).

3 Spade 1975.

4 To be strictly accurate, it is even weaker than this, as for instance a term and its negation are taken to signify the same thing but are not intersubstitutable. DELETE

5 Contrary to what Spade 1996, 109, seems to suggest.

6 S. log. I.6 (OPh I.19.8-13; Ockham 1995b, 16-17).

7 And probably is not so preserved, if one accepts the arguments of Gregory of Rimini and Peter of Ailly. See Spade 1996, 120-127.

8 Quodl. V.9 (OTh IX.513.11-12). Compare Quodl. V.8 concl. 2 (OTh IX.510-511.65-69).

9 For example, expressions corresponding to connotative terms, on Spade's account (Spade 1975).

10 See ?? below.

11 Spade 1975.

12 OTh IX.513.130-136; Spade 1996, 358-359.

13 Spade 1980, 12.

14 Ibid.

15 See II.1 and II.2 above.

16 S. log. I.6 (OPh I.19.8-13; Ockham 1995b, 16-17).

17 Spade 1980, 12.

18 Spade 1975 and 1980.

19 Spade 1975, 68.

20 OPh I.95-96.

21 Spade 1975, 67.

22 See Spade 1996, 180-182.

23 Spade 1975, 68.

24 Spade 1996, 99.

25 Spade acknowledges this, ibid.

26 Spade 1996, 99-100.

27 Spade 1975.

28 Panaccio 1990.

29 See III.1 above.