Abstract: In his recent study, Concepts, Fodor identifies five non-negotiable constraints on any theory of concepts. These theses were all shared by the standard medieval theories of concepts. However, those theories were cognitivist, in contrast with Fodor’s: concepts are definitions, a form of natural knowledge. The medieval theories were formed under two influences, from Aristotle by way of Boethius, and from Augustine. The tension between them resulted in the Ockhamist notion of a natural language, concepts as signs. Thus conventional signs, spoken and written, signify things in the world by the mediation of concepts which themselves form a language of thought, signifying those things naturally by their similarity. Indeed, later thinkers realised that everything signifies itself and what is like it naturally in a broad sense by means of the concept of its natural likeness.
1. Introduction1 The medieval theory of signification underpinned the theory of truth, which in turn fed into a theory of inference. The theory of signification describes generally how words relate to things, and how propositions come to mean what they do. But this general description needs a further account of how a particular occurrence of a word in a particular proposition is related to which things in what way. Only then can one say what has to be the case for the proposition to be true, and so determine how truth is preserved in an inference.
For the medievals in whom I am interested, the signification of words and propositions was made possible by their link to concepts. Vocal signs are seen as imposed by custom as marks or signs for concepts, and written signs are in turn marks or signs for vocal signs, and so indirectly for concepts. Concepts, however, signify or conceive a range of objects naturally, not by any conventional imposition. Concepts are formed by abstraction from sensory cognition. These medieval thinkers inherited from Aristotle, and took further, an elaborate and rich theory of cognitive powers which drew from sensation the whole panoply of cognitive awareness. The common sense discovers shape, motion and other aspects of cognition not present in each particular sense—separate experiences are needed to discern motion, and both sight and touch are needed to learn about shape or figure. An estimative or cogitative sense is needed to recognise the hostility of the wolf or the friendliness of the dog, qualities not immediately evident in sensation. Further composition and division is needed to create further concepts, and abstraction to understand generality. But they were empiricists, following Aristotle in believing that all knowledge is derived from the senses. “The mind is a tabula rasa on which nothing is at first written, but can be written” (De Anima 430a1). The innate powers of cognition were manifold and considerable, but no more than is necessary to the empiricist project of obtaining all real knowledge through the senses.
Concepts, therefore, have a natural epistemological relation to the class of things which they signify. To call it “natural” means that the concept is linked by a law-like causal connection to that of which it is a concept, that causal link being explained by the mind’s cognitive abilities. Conventional signs, the signs of spoken and written language, in contrast, gain their signification only by being linked by custom and practice to those natural signs. They obtain their signification indirectly, in what Simon Blackburn called a “dog-legged” manner.2 Their immediate signification, or what they are primarily attached or subordinate to, is the concept; thereby, their ultimate signification is the range of things to which the concept applies. John Buridan wrote, in the 1350s: “Categorematic words … signify things by the mediation of their concepts, according to which concepts, or similar ones, they were imposed to signify. So we call the things conceived by those concepts ‘ultimate significata’ … but the concepts we call ‘immediate significata’.”3
2. Concepts in Modern Philosophy What are concepts? Nowadays, just as in medieval times, it is common to identify them as constituents of mental propositions or thoughts. Many mental states have content. They consist in representing something as of some character. That content is specified by a proposition. Just as the sentence expressing that proposition has structure and consists of sub-sentential expressions, so it is claimed, propositions can be articulated into components. These propositions have a content which is structured and consists of concepts arranged in a certain conceptual structure, a mental language of concepts.
Concepts are hyperintensional. That is, the criteria of distinctness of concepts is stricter than necessary equivalence. For example, the concepts ‘triangle’ and ‘trilateral’ are necessarily equivalent—anything which is trilateral is necessarily triangular and vice versa. Nonetheless, the concepts are distinct. What makes them distinct is that one can believe that the one applies to something without the other. Concepts map out the fine structure of beliefs.
The central opposition in theories of concepts is between those who treat concepts as mental particulars and those who treat them as abstract objects. What unites concepts is their content, and that content consists of a common conceptual structure. Without the mental acts of judging, believing, thinking, there would be no content, no abstract universal. The nominalist urges that concepts have no existence beyond our minds, yet is willing to treat them as a common medium for the articulation and expression of those thoughts. The realist, in contrast, claims priority for the common medium, construing the private thoughts as derivative therefrom.
On the common view, concepts are a kind of particular. They are not only mind-dependent but also private. We may suggest that different people can share the same concept, since they can think the same thought—that is, their thoughts can have the same content. But all these terms equivocate between a private and a public reference. Each of us has his or her own thoughts and beliefs, and we make our own judgments; we speak loosely when we say that we share and communicate these thoughts, that we can have the same beliefs, and that we make the same judgments.
The hardest question is how a concept is related to its instances, to objects. Again, the popular view grounds this relation in the elaboration of conceptual structure as a mental language. As a language, its constituents are signs, and a concept is related to its instances as a sign to what it signifies. Just as sentences and words are signs, so too mental propositions and concepts are signs, with their own special relation to those objects. What is that relation? Contemporary philosophy of mind is predominantly naturalistic, so that such relations are to be explained in terms of some natural relation, for example, a causal relation. Acquaintance with the object causes the formation of the concept, which then becomes a natural sign for it. An alternative theory claims that the atomic elements of the language of thought are, in some way, innate. Yet that threatens to make their relation to objects mysterious and unanalysable, or at least to require an explanation not grounded in experience. Less mysterious is the normal empiricist resort, which is to posit an (innate) faculty of abstraction, which responds to suitable input by eliciting the appropriate concept. Indeed, it is difficult to see how a naturalistic, for example, causal theory, can work without such an inborn ability.
Concepts play an explanatory role. Intentional explanation of a person’s actions needs to relate those actions to the intentions and goals to which they are directed. To fill out the intentional nature of the explanation, those actions need to be described in terms which relate to the agent’s beliefs and purposes. Thus similar actions can result from dissimilar intents; and different actions can subserve similar purposes. The theory of content articulates this explanatory scheme into a language of the mind. The building blocks of this language are concepts.
Not only is this conceptual language the focus of contemporary research into cognitive science and philosophy of mind. It was also the centre of attention in philosophical thinking in the fourteenth century. There is, indeed, a remarkable parallel between contemporary concerns and theories and medieval conceptions. Perhaps this should not be remarkable—after all, the one preceded the other by some six centuries. But the chain of descent is tangled, and there is assuredly no direct influence of the one on the other. Some may argue that it illustrates the phenomenon of convergence on the truth; others that the same siren voices continue to lead the best thinkers astray.
In a recent study, Concepts, Jerry Fodor presents five theses which together make up what he calls the Representative Theory of Mind, which is to form the bedrock of his theory of concepts.4 Concepts are mental particulars with both causal powers and semantic content. The central task he presents is to reveal the link between these two aspects. The first thesis states:
1. Psychological explanation is typically nomic and is intentional through and through.
Those who deny this are physicalists, who believe that one day all will be stateable in the language of physics. The medieval conception was definitely on Fodor’s side.
2. “Mental representations” are the primitive bearers of intentional content.
That is, the mental realm is prior to, and explanatory of, the vocal and written realm. The medievals would heartily agree.
3. Thinking is computation.
Fodor makes much here of Turing’s analysis of the notion of effective computation. The crux seems to be, however, that Turing treated mental representations as symbols. We will see in §3 that this was the revolutionary insight of the thirteenth century which produced the flowering of semantic theory in the fourteenth century.
4. Meaning is information (more or less).
The intention is that semantic content follows (somehow) from causal relations, that the content of a concept is a result of its causal relationship to what falls under it. This is what the medievals articulated in their talk of cognitions “naturally signifying” those things which were included in them. Nor were they backward in describing the sensory and intellectual mechanisms by which the cognitions were formed, building on Aristotle’s discussion in his De Anima and other biological works.
5. Whatever distinguishes coextensive concepts is ipso facto “in the head”.
The medievals would agree: if the things signified are the same—where then can the difference lie? It must lie in the cognitions—“in the head”. The cognitions have different content, but any abstract notion of content is parasitic on the particular content of particular mental states. So Fodor shares an internalist conception with his medieval precursors.
Fodor’s main target in his study is to show that concepts cannot be definitions, for no such concepts could be acquired—or at least, the primitive basis must be atomic and not definitional. He describes his preferred theory as Informational Atomism, where the concept x is not constituted by reference to xs, but by reference to the response which we humans have to experience of xs. Here he departs radically not just from contemporary cognitive science—as he recognises—but from the medieval picture. For the medieval conception is through and through cognitivist: there is a mechanism by which we acquire the concept of x by experience of xs. The medievals received this model from Aristotle. It is a classical, empiricist model. For it is worth recognising that the contrast between empiricism and rationalism is really one of degree, not of kind. The former announces that all concepts are acquired; the latter that some are innate. But each has to modify its claim, the empiricist admitting that some “innate” capacity or mechanism (abstraction, induction or whatever) is needed in order to acquire concepts from experience; the rationalist that concepts are only latently there at birth, and their overt recognition is triggered by experience through an appropriate “innate” mechanism. The medievals then are seen as Aristotelians, interested in the mechanism of concept acquisition; Fodor as Platonist, interested in the trigger by which we respond to experience by producing certain concepts to which we are by our nature prone. Fodor describes his theory as explicitly “non-cognitivist”: a concept x is constituted not (as the definitional theory claims) by what xs have in common, but by our response to xs—how xs strike us. Only in this way, he believes, can one explain concept acquisition.
Medieval authors show their rejection of such a theory from the start: “First, we say that concepts are cognitions”, writes Thomas of Cleves in the first paragraph of his treatise on Concepts of 1370 or thereabouts.5 This is not just a pun: concepts for the medievals really put us in touch with how things are, and constitute a form of knowledge. Moreover, concepts are definitions. The main casualty of denying this, Fodor notes, is analytic connection. If each concept embraces others, either as parts in a literally combinatorial conception, or as constituted by its inferential connections with others, we have an immediate explanation of analytic connections as connections between concepts. On the other hand, if concepts are atomic, connections between them can only be inessential and by association—a position sympathetic to Quine’s notorious rejection of analyticity.
In fact, both definitional and atomist theories of concepts admit both complex and unanalysable concepts. Only a holistic theory (the so-called “theory theory”6) can claim that every concept is decomposable. Fodor’s so-called “atomist” theory includes molecular concepts like ‘white man’ which conjoins the possibly atomic concepts ‘white’ and ‘man’. Conversely, the definitional theory cannot maintain that all concepts decompose further, on pain of a regress. The real difference between the two accounts lies in their account of concept acquisition and application: what is the link between the concept x and xs? For Fodor, it is contingent and statistical; for the definitional theory, it is essential—causal and natural.
Thomas of Cleves devotes several pages to discussion of such analytic connections. His analysis suggests a compositional conception of concepts and signs. For example, ‘white man’ is superior in signification to ‘man’ for it includes it as part of its signification and so signifies more. But it is inferior to ‘man’ inferentially [in consequendo] for if there is a white man, there is a man but not vice versa. Signs and concepts are constituted by their signification, which results in their inferential connections. But the inferential model which in more recent times has displaced the compositional model as an articulation of the definitional theory is not in evidence. It is the innate capacity to abstract which makes the compositional and definitional conception of concepts viable.
Thomas’ treatment of concepts, a model of its time, is thus very different in its epistemological basis from Fodor’s information-theoretic account. What they share, however, is also important and extensive. Concepts are mental particulars which provide categories for classification, and they can be composed into thoughts or propositions in such a way that the signification of the composite thought derives from that of its constituents. Concepts are—for us humans, at any rate—learned and public: we all learn them and share them. Fodor describes his five conditions on any theory of concepts as “non-negotiable” (p. 23), so it is not too surprising, perhaps, that he shares them with Paul and Thomas. But it is important, for it shows how contemporary are the medievals’ concerns despite the six hundred years which separate us from them.
3. The Boethian and Augustinian Traditions Our conception of concepts and content has, as described in §2, an ancestry in the semantic theories and epistemology of the Middle Ages, which in turn were inspired by a passage in Aristotle’s De Interpretatione as interpreted by Boethius. Boethius translates Aristotle’s remark at 16a3-4 as follows: “spoken [words] are signs (notae) of impressions (passionum) in the soul, and written of spoken”;7 and in his commentary he elaborates on the relation between written inscription (litterae), the spoken sound (vox), the thought or concept (intellectus) and the thing (res) as follows: “for thing, concept, sound and letter are four: the concept conceives the thing, spoken sounds are signs of the concept, and letters signify the sounds.”8 That is, the letters (in writing) signify the spoken utterance, which in turn designates (a literary variant of ‘signify’, perhaps, since in the Aristotelian passage they were both ‘notae’) the concept.9 The function of the concept is to act as an intermediary, relating word (spoken and written) and thing, for the concept is an effect or impression (passio) of the thing in the mind. So, if the word signifies the concept by some human conventional imposition, then on hearing the word, we will come naturally to think of the thing. For the mind forms a likeness (similitudo) of things by a natural capacity. Whereas different peoples, speaking different languages have different words, both written and spoken, the things and the concepts and likenesses by which they conceive them are the same. Boethius’ translation of Aristotle continues: “and just as the letters are not the same, neither are the sounds. But they are signs of the same impressions for all, and those in turn are likenesses of the same things.”10 Thus we inhabit the same world and conceive of it the same way, says Aristotle; but we speak (and so write) of it differently.
This is the dominant semantic scheme of the medieval period: words signify concepts, which are likenesses of things in the world. But there is a further suggestion, indeed, a rather different picture, present in Boethius which was only fully elaborated in the thirteenth and fourteenth centuries. A few pages later in his first commentary, Boethius writes: “for the spoken sound and the concept of the thing signify the same thing.”11 The spoken sound signifies both the concept and the thing in the world. When we say ‘stone’, we signify first the thought or concept of stone and secondarily the stone itself. But in his second commentary, Boethius draws back from this conception, declaring that “although sounds are names of things, but we don’t use sounds in this way to signify things, but [to signify] those impressions of things in the soul which are within us.”12 That is, although spoken sounds are names of things and are so by virtue of signifying (by convention) their corresponding concepts (passiones animae), the vocal utterance does not itself in any way signify the thing conceived. For Aristotle had not expressed it in that way.
Equally influential, however, if not more so, was a passage in Augustine’s De Doctrina Christiana, where Augustine defines ‘sign’: “for the sign is something beyond the object which acts on the senses making something else come into thought.”13 There is explicit reference to this definition in, for example, the Commentary on Priscian Major ascribed to Robert Kilwardby,14 and an implicit one in William of Ockham’s Summa Logicae I 1,15 and many other places. However, the Augustinian account differs from Boethius’ in two important respects. First, what is signified is the thing—as brought out by Augustine’s examples: a footprint is a sign of a foot; and ‘cow’ signifies cows (op.cit. II x 15)—in both cases because the sign brings the object to mind. Secondly, the sign is required to act on the senses, whether visually, or aurally, or through one of the other senses (cf. II iii 4).
The first of these differences takes further the idea that we saw Boethius toyed with in his first commentary, that names signify those things of which they are names. To be sure, they do so through the mediation of a concept or act of mind. That mental act is vital to Augustine’s semiotics; but it is not itself signified by the sign. Rather, he speaks of the sign “showing” (demonstrare) what is in the mind: “Conventional signs are those which living things give to each other, in order to show … their mental state, or what they have sensed or thought.”16 The spoken and written word serves to convey from one mind to another these thoughts or emotions (motus animi). But in the Augustinian account, words signify what they name—the things we talk about.
The question, whether words signify concepts or things, came to the fore in the mid-thirteenth century. Roger Bacon speaks of a great debate on the subject,17 as does John Duns Scotus some years later.18 Besides the authority of Aristotle via Boethius—and what Aristotle meant is open to interpretation—arguments on each side can be given. For example, if words signified things, would empty names not then become impossible?19 On the other hand, when I say that Socrates runs, I mean that Socrates himself runs, not that some image or concept in the mind runs (or images running in the mind).20 Bacon was only one among many, e.g., Peter of Auvergne, Peter of Cornwall and (ps-)Kilwardby, to extend what is signified from concepts to things.21
In tandem with this change in what was signified (things, not concepts) went another, but this time one to which the Augustinian formula was antithetical, in its further requirement that the sign affect the senses in some way. If the word (spoken or written) signifies the thing by virtue of its connection to the concept, what is the right account of the connection of concept and thing? The big idea, which was to produce a huge theoretical development in the fourteenth century, was to suggest that the concept itself is a sign. This was the major original contribution of the thirteenth century, prefigured in Boethius but contrary to the Augustinian inheritance. For it goes directly against Augustine’s suggestion that the concept is invoked as intermediary by the operation of the sign on the senses: concepts are insensible; but in the Boethian tradition, although the concept is what is signified by the spoken sound, there is a suggestion that it itself signifies. Boethius writes: “of which therefore there are these four: letters, sounds, concepts, things, and most closely and principally letters signify verbs and names. These in turn principally and truly designate concepts, but in the second place also things. Concepts however are significative only of things.”22
The new conception of the signification of concepts is found as early as the Logica attributed to Lambert of Auxerre (c. 1240).23 (Ps-)Kilwardby shows how to square the new conception with the Augustinian requirement. ‘Sign’ can be taken in two ways, he says. In one way, it is material and sensible, and works by its action on the senses. But in another way, we can think of it in abstraction from its sensible and material aspects—and this is the real subject of the science of signs (op.cit., p. 4). This new departure found its greatest influence in the work of Scotus and Ockham. It also needs to be squared with Aristotle. Scotus writes: “the concept is a natural sign of its object (De Interpretatione ch. 1: impressions [in the soul] are signs of things and naturally so).”24 Bacon says much the same: “not every sign is offered to the sense as the usual description of a sign supposes, but something is offered to the intellect alone, as Aristotle observed, who says that impressions in the soul are signs of things.”25 Perhaps misquotation is the only solution, for Aristotle goes on to say categorically that every name is so by convention, since no name is a name naturally: “I said, ‘according to convention’, for nothing is naturally a name but only when it is a sign.”26 The way was now clear for Ockham to open his Summa Logicae with the words: “a concept-term is an concept or impression in the soul signifying or consignifying something naturally”.27 In doing so, he refers explicitly to Boethius and to Augustine.