The Middle Ages
THE MIDDLE AGES
Most of what St. Augustine (354–430) had to say about language and meaning was said not for its own sake but in support or elucidation of some theological doctrine. Partly for that reason, perhaps, his semantic doctrines had less effect on philosophy of language in the Middle Ages than might be expected, considering his enormous influence on medieval philosophy in general.
The short treatise Principia Dialecticae (probably written around 384, when Augustine was a professor of rhetoric) contains what may be the only instance of a semantic inquiry pursued by Augustine without a motive. In it he distinguishes four principal semantic elements: (1) the word (verbum ), a spoken articulate sound, classifiable as a vocable of some language; (2) the expressible (dicibile ), "whatever is sensed in the word by the mind rather than by the ear and is retained by the mind"; (3) the ordinary use of the word (dictio ), (opposed, for instance, to the use of the word as a sign for itself), which involves "both  the word itself and  that which occurs in a mind as the result of the word" when the word occurs "not on its own account but on account of something else that is to be signified"; (4) the signified thing (res ), which may be "something understood, or sensed, or inapprehensible"—the last category reserved for, for instance, God and formless matter (Ch. 5). Of these four elements, (2) and (3) together seem to represent different aspects of the Stoic lekton; but whatever their origin, their inclusion here indicates a level of sophistication in semantics that was not to be attained again for at least eight hundred years.
Chapter VII of the Principia Dialecticae is devoted expressly to the "import" (vis ) of words. In it Augustine maintains that "the import of a word is that whereby the extent of its efficacy is recognized [qua cognoscitur quantum valeat ], and it is efficacious to the extent to which it can affect a hearer." Import is a broader notion than signification and includes several sorts of effects a given word may have because of its sound, its degree of familiarity to the hearer, its degree of admissibility into polite conversation, its being recognized by the learned hearer as a dactylic foot or as some particular part of speech, and so on. The paradigm case of signification is described as occurring "on an occasion when a sign has been comprehended through a word [and] the mind regards [intuetur ] nothing other than that very thing the sign it comprehends is a sign of. Suppose, for example, that Augustine has been named and someone to whom I am known thinks of nothing other than myself, or that some other man named Augustine comes to mind if the name happens to be heard by someone who does not know me but knows that other man." The most remarkable and apparently novel features of this brief account are (a ) the extension of the notion of meaning in Augustine's doctrine of "import" and (b ) the orientation of his account of meaning around the effects words have on their hearers. The remainder of the treatise deals with simple and conjoined words and sentences, etymology, and various types of ambiguity and obscurity.
The longest of Augustine's discussions of semantic questions occurs in the dialogue De Magistro (389), which is designed ultimately to support the Augustinian doctrine of "divine illumination" as the sole genuine source of truth. Thus the first 11 chapters are supposed to show that "we learn nothing through those signs that are called words" (Ch. X), while chapters XI–XIV develop the thesis that Christ, the truth, teaches us inwardly while men by their use of outward signs merely prompt us to raise questions. The argument in support of the negative conclusion is an outstanding example of overemphasis on the word as the unit of signification. "When words are uttered, either we know or we do not know what they signify. If we know, then we do not learn but are reminded. If, on the other hand, we do not know, then we are not even reminded (though we may be prompted to ask)" (Ch. XI). Therefore, "we learn nothing through those signs that are called words"—as if one's knowing what the words mean in "armadillos are mammals" precluded one's learning anything through hearing that sentence uttered. At this crucial juncture in the dialogue Augustine's ulterior motive seems to have distorted his judgment.
Perhaps the most interesting point in the early chapters of De Magistro is one that bears on the best-known Augustinian passage on language, the description of his learning to speak in Confessions (397), 1.8, made famous by Ludwig Wittgenstein's use of it in Philosophical Investigations (Sec. 1). The passage in the Confessions can hardly be considered a theoretical statement at all, since Augustine's main aim in it is to describe a milestone on his descent "into the stormy fellowship of human life," but it does contain a brief, uncritical account of one way in which a child might be shown "the things of which words are signs."
That this account cannot be considered important in the context of Augustine's own views on language is plain from the fact that he had already criticized just such an account on theoretical grounds in De Magistro. In an attempt to refine the original suggestion of the dialogue that a sign cannot be a sign "unless it signifies something" (Ch. II), Augustine asks to be shown "that one thing itself, whatever it is, which is signified by these two words," ex and de, Latin prepositions there taken to be synonymous. After several obviously unsuitable suggestions, the tentative conclusion is reached that not only in these problematic cases but also in every case of attempting to show "the thing signified," all that can be shown is further signs, such as other words, pointing, pantomiming. This criticism is of course not the same as Wittgenstein's, nor is it particularly far-reaching in its own right, since it is soon modified to allow that we can in certain cases display the very thing signified without the use of further signs—for instance, if the thing signified is something we are able to do (such as walk) and we are not in the act of doing it when asked to display the thing signified (Ch. III).
A rather fully developed semantic theory appears in the De Trinitate (399–419), especially in 9.7–12 and 15.10–16, although it is presented no more for its own sake than is the theory in De Magistro. The theory appears as the explanatory half of an ingenious analogy designed to clarify (a ) the relation between the First and Second Persons of the Trinity, (b ) the two natures of the Second Person, and (c ) the identification of the Second Person as the Word. The analogical points may be ignored for present purposes and the semantic doctrine sketched as follows. "Word" has at least two senses. "In one sense, those things are called words that occupy intervals of time with syllables, whether they are pronounced or only thought. In another sense, everything that is known is said to be a word impressed on the mind as long as it can be brought out of memory and defined" (9.10). (Augustine actually introduces a third sense involving the love of what is known, but it seems pointless except for purposes of the analogy.) The second kind of "word," which Augustine describes more generally as a "locution" when the demands of the analogy are not uppermost in his mind, occupies the central position in the doctrine. "The word that sounds outwardly is a sign of the word that gives light inwardly, and the name 'word' is better suited to the latter; for what is uttered by the mouth of the flesh is the articulate sound of the word [vox verbi ]; … [thus] our word becomes an articulate sound … by taking on [articulate sound], not by consuming itself so as to be changed into it" (15.11).
The doctrine of the inward locution sometimes bears a striking resemblance to Plato's doctrine of ideal names in the Cratylus, although a direct historical connection seems unlikely. "For of necessity, when we say what is true—i.e., say what we know—the knowledge itself, which we retain in memory, gives birth to a word that is altogether of the same kind as the knowledge from which it is born. For the thought formed by the thing that we know is a word that is neither Greek nor Latin nor of any other language. But since it is necessary to convey it into the knowledge of those with whom we speak, some sign is adopted by which it is signified" (15.10; cf. Sermo 225.3). According to this doctrine, then, it seems that one's saying "armadillos are mammals" embodies in sounds one's inward locution to that effect, which itself differs from one's knowledge that armadillos are mammals only in being brought out of memory into conscious thought. Augustine sometimes suggests that the inward locution, then, is not itself verbal; words used in the mind are not essentially different from words outwardly pronounced, as Augustine's first division claims. Indeed, the inward locution is evidently less a mental entity than the state of consciousness into which a mental entity, namely, a known truth, must be brought if it is to be given verbal expression.
As an original contributor to semantics, Boethius (480–524) is much less interesting than Augustine. Since, however, his translations and commentaries constituted the sole source of Aristotelian logic for the medievals until the twelfth century, Boethius's influence over the development of semantics in the Middle Ages is powerful where Augustine's is slight.
Most medieval semantic theories take as their starting point Boethius's translation of the rudimentary account in Aristotle's De Interpretatione, Chapter 1. No doubt the traditional misreading of those passages during and after the Middle Ages is largely the result of the fact that in his otherwise faithful rendering Boethius obliterated the Aristotelian distinction between symbols and symptoms, translating both σύμβολα and σημι̑α as notae. Another of the principal difficulties in Aristotle's account—the apparent interposition of "mental modifications" between words and things—had been discussed at least as early as the third century by Alexander of Aphrodisias, whose confusing resolution of the difficulty was transmitted to the medievals in Boethius's second commentary on the De Interpretatione. Alexander had asked whether Aristotle's account forces us to consider the mental modifications as names of things. In order to avoid that consequence he had developed the view that although "a name is imposed on a thing" and "although spoken words are names of things, nevertheless we use spoken words not in order to signify things, but in order to signify those mental modifications that are produced in us as a result of the things. Therefore, since spoken words are uttered for the purpose of signifying those entities, he [Aristotle] was right to say that they are primarily the signs [notas ] of those entities" (413a–b; all references in this section are to Patrologia Latina, edited by J. P. Migne, Vol. 64).
Perhaps the most influential doctrine (at least in the late Middle Ages) that can be traced directly to Boethius's treatment of De Interpretatione, Chapter 1, is that of the three discourses: written, spoken, and mental. Citing Porphyry (c. 233–c. 305) as his authority, Boethius reported that "among the Peripatetics there were said to be three discourses [orationes ]—one written in letters, another uttered in speech, and a third put together in the mind. Now if there are three discourses, the parts of discourse are no doubt likewise threefold; for since the noun and the verb are the principal parts of discourse, there will be some nouns and verbs that are written, others that are spoken, and still others that are silent and employed by the mind" (407b–c). Here, as in his transmission of the Aristotelian account itself, the vagueness of Boethius's presentation is as important historically as its content. Are there two completely different sets of nouns and verbs, one for writing and one for speech? And is this mental discourse nothing more than silently running over a sentence in Latin or English, Or is it a nonverbal operation, reminiscent of Augustine's "inward locution"? The fact that mental discourse is said to have nouns and verbs of its own suggests the former view, if either; but since Aristotle had maintained that the mental modifications were the same for all (regardless of their native tongue), and since Boethius offers this doctrine of the three discourses in explanation of Aristotle's account, there is some basis for the second view as well. These were among the difficulties discussed in the medieval development of the doctrine.
The medieval distinction between words of first and second "imposition," a genuine prefiguring of the twentieth-century distinction between object language and metalanguage, also has its roots in Boethius's transmission of older doctrines. In his commentary on Aristotle's Categories he presents the distinction very much as he found it in Porphyry's Expositio of the same work (A. Busse, ed., pp. 57–58). "The first imposition [positio ] of a name is made with respect to the signification of the word, the second with respect to its form " (159c). Thus, whenever some extralinguistic entity is called a man, it is a case of first imposition. "But when the word 'man' itself is called a noun, no reference is made to the signification of the word [Boethius has 'noun'], but to its form, in virtue of which it admits of inflection by means of [grammatical] cases" (159b–c). Thus "noun" in its ordinary use is a word of second imposition.
In this primitive form the distinction seems to apply only to the grammarian's kind of interest in discourse. Boethius, however, took the position that "the whole art of logic is concerned with discourse" (161c–d). How does the philosopher's interest in language differ from the grammarian's? Very much as the economist's interest in money differs from the numismatist's, for Boethius compares the signification of a word to the buying power of a coin and its grammatical form to the "bronze stamped with a design." Consider "an utterance that designates nothing, such as 'gargulus.' Although the grammarians, considering its form, contend that it is a noun, philosophy does not recognize it as a noun unless it is imposed in such a way as to designate a conception belonging to a mind (in which same way it can signify some real thing)" (408c–d). Apparently, then, second imposition needs to be more broadly conceived, or a philosopher's kind of second imposition must be added to the kind described by Boethius. The resolution of such difficulties was among the goals of the later doctrine of the impositions and "intentions" of words.
By far the most influential of Boethius's bequests to the Middle Ages was his formulation of the problem of universals in his second commentary on Porphyry's Isagoge. Needless to say, a great many semantic issues were discussed in the long controversy over universals, and a few of the more important ones will be noted below. Boethius's formulation of the problem, however, was oriented around questions of metaphysics rather than of semantics and so may be passed over here.
One of the semantic problems recognized by the early medievals in the few logical works of Aristotle available to them was the problem of paronyms, or denominatives. Its principal source is the following passage in the Categories, Chapter 8 (10a27 ff.). "These, then, that we have mentioned are qualities, while things called paronymously because of these or called in some other way from them are qualified. Now in most cases, indeed in practically all, things are called paronymously, as the pale man from paleness, the grammatical from grammar, the just from justice, and so on."
St. Anselm (1033–1109) remarks at the end of his dialogue on denominatives—De Grammatico —that the semantics of denominatives was a favorite topic among dialecticians of the eleventh century, evidently because of the difficulty of developing a satisfactory account of denominative words that occur both as concrete nouns and as adjectives. (Anselm's chief example is grammaticus, but because the English word "grammatical" is not a denominative of this sort, "illiterate" will be used here.) Thus, the opening question of Anselm's dialogue is whether "illiterate" signifies a substance or a quality. This seems to be a narrow, perhaps artificial problem, but under his characteristically ingenious treatment it leads to results of general importance.
The superficially most plausible solution to the problem is that such a word sometimes signifies a substance—as in "not every illiterate is stupid"—and sometimes a quality (illiteracy)—as in "not every illiterate person is stupid." This solution is shown to fail, however, at least in its second half, for if we tried to use "illiterate" alone in speaking about the quality—as in "illiterate is a deplorable condition"—"not only the grammarians would be upset, but even the peasants would laugh" (Ch. XI). "Illiterate," we must recognize, "does not signify a person and illiteracy as a unit [ut unum ] but signifies illiteracy directly [per se ] and a person indirectly [per aliud ]" (Ch. XII). Another way to put the distinction between the two kinds of signification is to say that "illiterate" is significative of illiteracy and appellative of a person. "I now describe a name as appellative of each thing itself that is called [appellatur ] by that name in the speaker's usage; for there is no speaker's usage in which 'illiteracy is illiterate' occurs, … but rather 'the person is illiterate'" (Ch. XII).
The remainder of the dialogue refines and generalizes this account in the course of dealing with various objections to it. Anselm's most original and important contributions seem to be those developed mainly in the last two chapters (where the discussion centers around albus —"white"—rather than grammaticus ). The Master of the dialogue has suggested that "white" signifies (rather than appellates) nothing but being in possession of whiteness (habens albedinem ). This is disturbing to the Student, who feels the need of a signified thing. "White," he is willing to grant, "does not determinately signify this or that possessing entity, such as a body," but he wants to insist that it "indeterminately signifies something possessing whiteness." His principal argument is that "'white' signifies either something possessing whiteness or nothing; but one cannot conceive of nothing as possessing whiteness; therefore it is necessary that 'white' signify something possessing whiteness" (Ch. XX).
In reply Anselm takes the position that while it may always be the case that what is signified somehow depends for its being on some real thing, it cannot always be the case that what is signified is a thing. His arguments for this position display an interesting use of the principle of substitutivity.
If "white" signified a thing at all, it would signify something white. Now the signification of a word is what its definition presents, and what is presented by the definition may be substituted for the word itself. "So wherever 'white' is used it is taken correctly as 'something white'" (Ch. XXI). Then "Socrates is white" may be rewritten as "Socrates is something white." But "wherever 'something white' is used it is also correctly said twice—'something something white'—and wherever it is said twice, there also three times, and so on indefinitely" (Ch. XXI). Thus the plausible "Socrates is something white" would become the nonsensical "Socrates is something something white" and would ultimately lose all semblance of a statement.
Instead, in "Socrates is white," "white" appellates something white—Socrates himself—but what it signifies is being in possession of whiteness. Nor will it do to introduce a signified thing at this point, for if we take something in possession of whiteness to be what "white" signifies, we shall have to grant that something in possession of whiteness is that which is white. "If, therefore, 'white' is 'that which is white,' it is also 'that which is that which is white'; and if it is that it is also 'that which is that which is that which is white,' and so on indefinitely" (Ch. XXI). The nonsense-engendering substitutions cannot be made within "being in possession of whiteness," however, since the denominative "white" does not itself occur in it. Thus "it is clear enough that 'white' does not signify something in possession of whiteness …, but only being in possession of whiteness—i.e., [the categories] quality and possession [and not the category substance ]—and quality and possession by themselves make up no something " (ibid.). This argument is described as holding good for all single words that, like "white," signify "a plurality [of categories] out of which no one thing is made up" (Ch. XXI).
Although the special consideration of denominatives apparently lost its vogue soon after Anselm, many of the problems dealt with in his De Grammatico remained current and can be found two centuries afterward at the center of the theory of the properties of terms (see below).
The extensive logical writings of Peter Abelard (1079–1142) are best known for the theory of universals developed in them. That theory is important in the history of semantics because (a ) it explicitly approaches the problem of universals as a semantic rather than a metaphysical problem and because (b ) in doing so it introduces many of the elements of the semantic theories developed by the terminist logicians of the thirteenth and fourteenth centuries.
Regardless of how the problem of universals is approached, it involves a consideration of the semantics of words, especially of common names. Nevertheless, many of the countless medieval theories, in their preoccupation with the Porphyrian-Boethian questions about the existential status of genera and species, slighted or ignored the semantic issues. Abelard, on the other hand, began by adding a new semantic question to the three traditional metaphysical questions. "Could a universal consist of the signification of a concept [significatione intellectus ] when the things named were destroyed, as [in winter] when there are no roses to which the name 'rose' is common?" (Logica "Ingredientibus," edited by B. Geyer, p. 8).
Having associated universals with words, Abelard asked "whether they are associated only with words or with things as well" (p. 9). Applying the Aristotelian criterion predicability of more than one thing, he showed in a series of elaborate arguments that a universal cannot be identified as (1) a single thing or (2) a collection of things (pp. 10–16). His principal objection really avoids the issue of whether or not it makes sense to speak of a thing or a collection as predicable at all and concentrates instead on the impossibility of predicating a thing or a collection of more than one thing. Thus "it remains for us to ascribe universality of that sort to words [vocibus ] alone" (p. 16). As Abelard came to realize, words considered as utterances or inscriptions are themselves things. Accordingly he eventually distinguished between utterances [voces ] and words [sermones ] and organized his theory of universals around words in this strict sense—sermo = vox + significatio —which he described as products of human arrangements rather than mere natural effects (Logica "Nostorum Petitioni Sociorum," edited by B. Geyer, p. 522).
The only kind of word to which universality can conceivably be ascribed is the kind of word apparently predicable of more than one thing, that is, a common name in the nominative case. But that ascription cannot mean that the common name has some universal thing as its bearer, for, as he had shown, "universal thing" is a contradiction in terms. Nor can some particular thing be picked out as its bearer, for although it may be Socrates alone of whom the statement "a man is sitting in this house" is true, we cannot infer from it that Socrates is sitting in this house (p. 18). These considerations led Abelard to base the ascription of universality not on what the words name (nominare )—for example, "man" names each and every individual thing that is a man—but on their "mode of signification"; for although they name things that are discrete, they do so not "discretely and determinately" but "confusedly" (p. 29).
Abelard's explanation of this notion of confused naming, which was to play an important part in thirteenth- and fourteenth-century theories of the properties of terms, seems incomplete but runs along the following lines. "To signify is to establish [constituere ] a concept" (p. 136), and "when I hear 'man' … I do not recall all the natures or properties that are in the things subject [to that name]; instead, as a result of [hearing] 'man' I have a conception of animal, rational, and mortal, though not of subsequent accidents as well, [a conception that is] confused rather than discrete" (p. 27). Thus Abelard's answer to his additional semantic question is a qualified "yes." In winter the name "rose" lacks universality in that there are no things of which it is predicable, that is, "it is devoid of nomination" (nominatione ). "Nevertheless, it is still significative then in virtue of the concept [ex intellectu ]; … otherwise there would not be the proposition 'no rose exists'" (p. 30).
Other medieval theories of universals, such as William of Ockham's, center on semantic doctrines; but Abelard's "sermonism" was perhaps the most important medieval influence on the development of semantics during the succeeding two centuries of the high Middle Ages. Topics and terminology remained relatively stable in that remarkable period in the history of semantics, although many philosophers and every logician contributed to the discussions. For that reason the remaining material on the Middle Ages is oriented mainly around topics in medieval semantics, and no attempt is made to mention every man who discussed them.
Impositions and Intentions
The pervasive medieval distinctions between two levels of signification have attracted some attention in the twentieth century because of their resemblance to the object language–metalanguage distinction. Historically there were two such distinctions, both based on the observation, found already in Porphyry, that while some signs signify nonsigns, others are signs of signs.
The original distinction was drawn with respect to conventional signs, specifically, with respect to names (nouns and adjectives) in a natural language. Such signs acquired their signification only as a result of having been imposed by the users of the language. The primary, or first, imposition was on extralinguistic entities, and names such as "man" and "white" were classified as names of first imposition. As the language developed, other conventional signs were imposed on conventional signs as such; thus, names such as "noun" and "plural" are of second imposition. Names such as "utterance" and "mark" do signify conventional signs, but not as such (since there are of course nonsignificant utterances and marks); they are therefore names of first imposition. Most medieval logicians, presumably in avoidance of an infinite regress, were careful not to define names of second imposition as names of names of first imposition. Thus "name of second imposition" is itself a name of second imposition.
But even those who, like Abelard (Logica "Ingredientibus," edited by B. Geyer, p. 112), did define second imposition in terms of first seem never to have recognized a "third" imposition. (The imposition distinction, therefore, cannot reasonably be described as prefiguring a hierarchy of types.) The use made of the imposition distinction was apparently rather meager. Aristotle's categories were, for example, often said to be names of first imposition, while the subject matter of his De Interpretatione was described as names of second imposition. The distinction, although it was refined and discussed well into the fifteenth century, seems to have acquired what importance it had mainly from its connection with the later and better known of the two distinctions between levels of signification.
Concepts in their capacity as natural signs were called intentions and described in the doctrine of the three discourses as mental terms. It was only natural, then, to distinguish levels of signification among intentions as among conventionally significant extramental terms. This distinction, probably stemming from Avicenna (see Carl Prantl, Geschichte der Logik im Abendlande, 2.328), classified as first intentions all those naturally significant of entities other than intentions as such, while those that did naturally signify intentions as such were second intentions. The concept humanity is of course a first intention, but so is the concept mental entity. The concepts of the predicables—genus, species, differentia, property, accident —are second intentions, as is the concept predicable itself; no "third" intentions were ever recognized.
Thus first and second intentions and impositions were fundamentally parallel distinctions in separate domains. However, their development in the thirteenth and early fourteenth centuries was complicated (and sometimes confused) by two factors. First, there were, of course, extramental terms imposed on first and second intentions, such as humanity and genus. Such names were all of first imposition (since no intention was a conventional sign), but they were sometimes further described as names of first or second intention. Second, even more complicating was the fact that the first and second intentions themselves were considered to be terms in mental propositions. Thus, while in the written proposition "animal is a genus" the subject and predicate terms are both of first imposition, in the corresponding mental proposition that animal is a genus the subject term is a first intention and the predicate term is a second intention.
Of the two distinctions between levels of signification, the intention distinction had much more philosophical importance. The confusing interrelations of the two distinctions are perhaps best exhibited in William of Ockham (d. 1349), particularly in Summa Logicae, 1, 11–12. Logicians after William—for instance, Albert of Saxony (d. 1390), Pierre d'Ailly (1350–1421), Paul of Venice (d. 1428), and Paul of Pergula (d. 1451)—exhibited a tendency to simplify them by reverting to the treatment of impositions and intentions as strictly separate, parallel distinctions. Postmedieval scholastics—for example, John of St. Thomas (1589–1644)—were inclined to apply the intention distinction indifferently to extramental as well as to mental terms and to ignore the imposition distinction; it is in this simplified form that the "medieval" distinction between levels of signification is usually discussed in recent literature.
Almost everything genuinely novel in medieval logic is to be found in the theories of the properties of terms and of the functions of syncategorematic words developed by the logicians of the high Middle Ages. One reason why logic set off along that line of logicosemantic inquiries is that medieval logicians, especially through the formative period ending about 1250, thought of their subject as the science of language (scientia sermocinalis ).
That classification itself marked a break with the Aristotelian-Boethian tradition in that it was precise where the tradition had been vague. The notion of predication was unquestionably an essential part of the subject matter of logic, but Aristotle and Boethius had treated it in ways that often suggested that predicates might be extralinguistic and even extramental entities. This crucial vagueness, which was to some extent also the source of the medievals' concern with universals, left open the possibility that logic might be essentially a science of reality, resembling or subsumed under metaphysics.
However, in the earliest complete European logic we have after Boethius—the Dialectica of Garland the Computist (d. before 1102)—that possibility was already noted and explicitly ruled out. Predication, Garland maintained, occurs only in a proposition, and the only constituents of propositions are utterances; thus, only utterances may be predicated. The five predicables (genus, species, differentia, property, accident), the elementary subject matter of medieval logic, are, in virtue of being predicables, utterances and no more; and the ten categories (substance, quality, and so on) are likewise categories of utterances only—for instance, noun and adjective (Dialectica, edited by L. M. de Rijk, p. 3).
The attempt to establish logic as a science of linguistic entities only may be called sermocinalism. During the years 1150 to 1250, when medieval logic was acquiring its distinctive character, sermocinalism held undisputed sway as the philosophy of logic, but it did so in the refined and strengthened form given it in the writings of Abelard. Garland had attempted to make utterances (voces ) the elements of logic, which he thought of as the science of language. Abelard, recognizing that utterances are physical events that are, as such, of no interest to logicians, replaced the overly simple utterance with what he called the sermo, defined as the utterance taken together with its signification. Logicians in the second half of the twelfth century seem to have been unanimous in their adoption of this refinement. An anonymous Dialectica seu Logica supported the rejection of utterances as the elements of logic with the following interesting argument, somewhat reminiscent of Aristotle's doctrine of complex names. "Some utterances are significant; some are not.… This division … is exhaustive but seems not to be exclusive, since the same utterance may be both significant and not significant. For example, the utterance 'king' [rex ] is significant as a word, but since it is also part of a word, a syllable of a word—as in 'smoking' [sorex, shrew]—it is in that case and on that account not a significant utterance" (Martin Grabmann, Bearbeitungen, Berlin, 1937, p. 30).
Having more precisely identified the elements of logic as linguistic entities, Abelard suggested that logic as the science of language should determine significations on the basis of the application of utterances, determining the proper application of utterances on the basis of the investigations of the natural sciences (Dialectica, pp. 286–287). One reason for this suggestion seems to have been his concern with propositions true gratia terminorum, analytically true on semantic rather than on syntactic grounds—for instance, "if there is paternity, there is filiation" or "if it is a body, it is corporeal" (see pp. 284–286).
To most medieval philosophers Abelard's emphasis on the importance of signification as well as of utterance might have suggested that mental entities of some sort were to be considered the elements of logic. He explicitly rejected this possibility, however, and in doing so made his most important contribution to sermocinalism. He argued that a proposition true gratia terminorum could not be verified by an appeal to the status of mental entities. "When we say 'if it is man it is animal,' if we refer to the connection of the understanding of the propositions, as if we were concerned with the concepts, there is no truth to the conditional, since the one concept may occur entirely without the other" (p. 154). What we are concerned with, Abelard maintained, is the connection between the term animal and the definition of the term man —namely, the inclusion of the term animal within the string of terms making up the definition of the term man. As a result of this move, sermocinalism was directed not only against the notion of logic as a science of reality (scientia realis ) but evidently also against the notion of it as the science of reason (scientia rationalis ).
The philosophy of logic that eventually challenged sermocinalism concentrated its opposition on this last point. Since it was explicitly drawn from the philosophy of Avicenna, the rival doctrine may conveniently be called Avicennianism. Although as many of Avicenna's writings as were available to the medievals had been translated into Latin around the middle of the twelfth century, Avicennianism as a philosophy of logic seems not to have come into prominence until Albert the Great (1193–1280) adopted it around the middle of the thirteenth century. By that time, however, medieval logic was firmly committed to its distinctive line of development as the scientia sermocinalis. As a result, the main impact of Avicennianism as an alternative to sermocinalism was felt less on the work of the logicians than on the metaphysicians' discussions of the nature of logic.
The central doctrine of Avicennianism is presented in the frequently quoted passage from Avicenna's Philosophia Prima : "The subject matter of logic, as you know, is intentions understood secondarily, which are applied to intentions primarily understood" (I, 2, f70vA). Logic was the science of reason, Avicenna claimed, for "the relation of this doctrine [logic] to internal thought, which is called internal speech, is just like the relation of grammar to outward signification, which is called speech" (Logica f3rA). Thus grammar, not logic, was the sermocinal science, according to Avicennianism, and the rise of speculative grammar that was to follow may in part be attributed to this point of view.
The Properties of Terms
Until about the middle of the twelfth century the subject matter of medieval logic was drawn from Aristotle's Categories and De Interpretatione, together with a set of books by Porphyry and Boethius that were centered more or less closely on those two books of Aristotle. Later in the Middle Ages this collection of books, or the kind of logic these books contained, became known as logica vetus, the old logic. When the remaining four books of Aristotle's Organon began to circulate in western Europe during the twelfth century, they, or their contents, became known as logica nova, the new logic. The only completely new kind of material in the logica nova was the treatment of fallacy in Aristotle's Sophistical Refutations, which excited a tremendous interest in sophismata, fallacies resulting from the misuse of or natural ambiguities in various devices of ordinary discourse. Largely because of this lasting interest, medieval logicians of the late twelfth and early thirteenth centuries gradually developed an original logicosemantic inquiry. In order to distinguish this genuinely medieval contribution from Aristotle's contributions to logic, thirteenth-century philosophers began to speak of it as the logica moderna, lumping the logica vetus and logica nova together as logica antiqua. Perhaps the logica moderna was aimed originally at nothing more than providing ad hoc rules of inference to cover problematic locutions in ordinary discourse, but, although it retained that aim throughout its three-hundred-year history, its principal aim soon became the development of a reasonably general account of the different ways in which words are used to stand for things and to operate on other words.
The earliest known fully developed productions of the "modernist" or "terminist" logicians are the logical treatises of William of Sherwood (d. 1266/1272), Peter of Spain (d. 1277), and Lambert of Auxerre (fl. 1250), evidently written at Paris about the middle of the thirteenth century. At that time the logica moderna seems to have been thought of as having two branches, an account of "the properties of terms" (proprietates terminorum ) and an account of the signification and function of "syncategorematic words" (syncategoremata ). The two branches naturally differed in detail, but both accounts employed the same principles of explanation and had the same aims. Most nouns, pronouns, verbs, participles, and adjectives were considered to be categorematic words, words capable of serving as terms (that is, as subjects or predicates); and the syncategorematic words were those which can occur in a statement only together with categorematic words. The two branches of the logica moderna were thus theoretically exhaustive of the kinds of words occurring in various roles in statements.
The modernists of the thirteenth century regularly recognized four properties of terms: (1) signification—the word's meaning, broadly conceived, or the range of conventional uses of the word (a property of every categorematic); (2) supposition—the conventional interpretation of a word on a particular occasion of its use, a modification of its signification resulting from its syntactic context, if any, and other considerations (a property only of nouns, pronouns, and "substantive expressions," that is, other categorematics employed as substantives and particularly as subjects); (3) copulation—virtually the same as supposition, except that it is a property only of verbs, participles, and adjectives, especially when they occur as predicates; (4) appellation—"the present correct application of a term" (Sherwood), a property only of nouns, adjectives, and participles; for instance, in 2004 Chicago was an appellatum of "city" but Nineveh was not.
Obviously these four properties are not on an equal footing. The supposition, copulation, or appellation of a term was considered a function of its signification; Vincent of Beauvais (d. 1264) even designated signification the genus of which the other three are species (Carl Prantl, Geschichte der Logik, Vol. III, p. 83, n. 319). Moreover, copulation and appellation are of distinctly secondary importance. By the middle of the fourteenth century only signification and supposition were regularly recognized as properties of terms, and throughout the history of the logica moderna it was the supposition (suppositio ) of terms on which the inquiry centered. For that reason the best way of quickly acquiring a broad but accurate idea of the modernists' account of the properties of terms is to examine their divisions of supposition. (The recognition and treatment of the divisions of course differed from one modernist to another, but the following selection includes all the major divisions and many of the more interesting minor divisions.)
The supposition of a term was divided initially into proper and improper supposition. A term had improper supposition when it was used figuratively, and several varieties of improper supposition were distinguished: antonomastic, synecdochic, metaphoric, ironic, and metonymic. The proper supposition of a term was divided into formal and material supposition, the latter being the use of a term to refer to itself, either as type or as token—for instance, "man is a noun," "'man is an animal ' is a true statement," "man is a monosyllable," "man has three letters." Formal supposition was personal if the term was used to refer to individuals bearing the form signified by the term, simple if the reference was to the form itself, as in "man is a species." The initial division of personal supposition was sometimes based on the division of terms as common or discrete, depending on the possibility of using them to refer to more than one individual at a time. Thus, the subjects of the statements "Socrates is running" and "that man is running" have discrete personal supposition.
The portions of supposition theory dealing with the divisions of the personal supposition of common terms were more fully developed than the rest, not only because they were intrinsically more interesting but probably also because they provided the most points of contact between the logica antiqua and the logica moderna. In those portions of supposition theory, far more than in the others, the emphasis lay on the application of the theory to the evaluation of inferences, especially such inferences as involved Aristotle's four categorical propositions (or near relatives of them) but could not be adequately evaluated within the logica antiqua.
A common term was said to have determinate personal supposition when it was used to refer to some one individual without identifying the supposited individual, as in "a man is running" or in "some man is running." A statement including a "distributive sign" (such as "every" or "no") was, on the other hand, bound to include one or more common terms having confused personal supposition, terms used to refer to more than one individual at once or to one individual many times (as in "every man sees a man," where the second occurrence of "man" has confused supposition even if it is being used to refer to only one individual). If the confused supposition included each and every individual bearing the form signified by the common term, it was designated distributive, as in "every man is an animal," "no man is an ass." If the confused supposition did not plainly include that totality, it was designated merely confused (confusa tantum ), as in "every man is an animal," "every man sees a man."
The modernists observed that in many cases of distributive confused supposition it was possible to make a "descent" under the term having such supposition, instantiating as in "every man is an animal, therefore this man is an animal"; "no man is an ass, therefore no man is this ass." They described such cases as mobile but paid at least as much attention to the immobile distributive supposition produced by the use of "exclusives" or "exceptives" together with distributive signs. Thus, in "only every man is running" the distributive supposition of the common term is immobilized by the exclusive "only," so that one cannot infer "therefore only this man is running." It was also recognized that the inclusion of the distributive sign within the scope of the exclusive or exceptive was not always dependent on their relative positions in the statement, since from "every man except Socrates is running" one cannot infer "therefore this man except Socrates is running" (although this unacceptable inference is uninterpretable rather than invalid).
Supposition theory was an attempt to develop a unified treatment of a great number of semantical and logical topics that are still of interest, although now for the most part they are treated in separate inquiries. It is therefore especially intriguing and difficult to discover just what that unifying notion—supposition—amounted to. One broad description that plainly holds good for most of the divisions of supposition is that they are syntax-dependent referential functions of a term's signification. Any description in terms of syntax and semantics will, however, fail to cover all the divisions of supposition and will miss what is distinctive in it. In the case of improper supposition, for example, while the circumstances under which a term is used clearly do determine whether or not it is being used figuratively, it will not do to limit those circumstances to the syntactic context of the term's use.
Again, supposition theorists frequently remarked on the fact that the supposition apparently determined by the syntax often differs from the supposition intended by the framer of the statement. The man who visits his friend's garden and says "this plant grows in my garden" says what is false "with respect to discourse" (de virtute sermonis ), but the circumstances of his utterance show that he intends to use the word plant as in the statement "such a plant grows in my garden." The correct analysis of the supposition of "plant" takes it to have the supposition determined for it not by that syntactic context but by the clearly discernible "intention of the framer" (intentio ponentis ). Finally, among later supposition theorists it was a matter of controversy whether terms occurring in statements written in a closed book had any supposition at all, and those few who held that they did then have supposition seem to have been motivated by a misguided concern that otherwise certain true statements in a closed book, such as "God exists," might cease to be true while the book remained unread.
The consensus of the modernists seems to have been that a term had one or another kind of supposition only on an occasion of its actually being used in referring (or understood to have been used in referring) to some entity or entities, the particular kind of supposition being determined by a number of the circumstances of the occasion and its syntactic context being the most important but not in itself the decisive circumstance.
Within the logica moderna the investigation of syncategorematic words complemented the investigation of the categorematic words under the doctrine of "the properties of terms." Something closely resembling the modernists' notion of syncategoremata seems to have been operative in Aristotle and the Stoics, but the medievals evidently acquired the technical term and the rudiments of the notion directly from Priscian (fl. 500), who had reported that "according to the dialecticians [not identified], there are two [principal (?)] parts to a sentence—the noun and the verb—since they alone and of themselves make a sentence; but they called the other parts syncategoremata —that is, consignificants" (Institutio de Arte Grammatica, M. Hertz, ed., in H. Kiel, Grammatici Latini, Leipzig, 1855, Vol. II, p. 54).
Interest in the syncategoremata as such began in connection with the twelfth-century interest in fallacies of ambiguity as it became plain that the crucial ambiguity was often located elsewhere than in subjects and predicates. The grammatical basis of distinction provided by Priscian seems to have been adopted at first and occasionally even narrowed so that only the "indeclinables"—prepositions and conjunctions—were considered to be syncategoremata (see the anonymous late twelfth-century Fallacie Parvipontane, edited by L. M. de Rijk, in Logica Modernorum, Vol. I, p. 559; also see Abelard, Dialectica, edited by L. M. de Rijk, pp. 118–121). The notion of syncategoremata that became important in the logica moderna, however, was not founded on a strictly grammatical distinction. Abelard's treatment of "alone" in "a man alone is capable of laughter" (Logica "Ingredientibus," edited by B. Geyer, p. 483) prefigured the pattern that was to be followed by the modernists. He pointed out that if "alone" is taken to be part of the subject, the statement is about a man who happens to be by himself, while if "it is attached to the predication it denies the capacity for laughter to all non-men, as if to say: 'a man is capable of laughter in such a way that nothing else is capable of laughter.'"
When the investigation of syncategorematic words appeared as a separate inquiry in the treatises on syncategoremata by William of Sherwood and Peter of Spain (first half of the thirteenth century), the distinguishing characteristic of syncategorematic words was the fact that they had some effect on the relation between the categorematic words—that is, the predication or the "composition"—or on the relation between two predications or compositions. Thus Sherwood's inventory of syncategoremata, most of which became standard, included the verbs "begins" and "ceases" and the noun "nothing" as well as grammarians' syncategoremata such as "every," "both," "except," "alone," "is," "not," "necessarily," "if," "unless," "and," and "or." The standard syncategoremata, then, cannot be completely described as a selection of logical operators, although they plainly included such a selection. Nor was the investigation of them aimed primarily at uncovering their strictly formal properties. Many of the rules put forward in connection with one or another syncategorema were rules of inference—such as "When there are two distributions over one and the same part of a locution, the first immobilizes the second" (Sherwood)—but just as many were semantic rules—such as "The sign 'every' or 'all' requires that there be at least three appellata [for the term to which the sign is attached]" (Sherwood)—and there seems to have been no clear distinction drawn between the two sorts of rules.
The modernists' treatises on syncategoremata presupposed the doctrine of the properties of terms, as is shown by the rules given just above, and much of their discussion of the function of such words is in terms of the various modifications of supposition produced within the scope of one or another syncategorema. (The problem of determining the scope of syncategoremata, especially in contexts including more than one, was particularly important in these investigations.) In this way the syntactic and semantic questions about syncategoremata were essentially interconnected. Sherwood at least among the older modernists was sometimes concerned to discuss the signification of syncategoremata—"every," for example, was said by him to signify universality—but that seems to have been a feature of his unusual doctrine that in order to be significant a word had to signify some form. Most writers on syncategoremata took up Priscian's really unjustified translation of the Greek syncategoremata as "consignificants" and used it as the basis for their view that "strictly speaking, a syncategorema signifies nothing, but when added to another word it makes that word signify something, or makes it supposit for something or some things in some definite way, or exercises some other function having to do with a categorema" (William of Ockham, Summa Logicae, I, 4; cf. the remarks of John of Salisbury in Metalogicon, Book I, Ch. 16).
The initial impetus to the study of syncategoremata came from the twelfth-century interest in fallacies, and the investigation continued to be associated with fallacies or with sophismata throughout its development. Observations about syncategoremata were only incidental in the twelfth-century treatises on fallacies, but the novel emphasis of the logica moderna on the syncategoremata themselves is evident in the development of a new sort of treatise—"On Exponibles"—in the first half of the thirteenth century. "An exponible proposition is a proposition having an obscure sense that stands in need of exposition because of some syncategorema located in it explicitly or implicitly or in some word" (Tractatus Exponibilium, doubtfully ascribed to Peter of Spain, in The Summulae Logicales of Peter of Spain, edited by J. P. Mullally, p. 104). The exponibles did not involve fallacious arguments nor were they, strictly speaking, ambiguous. They were simply subjects for analysis, an analysis that was to explicate the force of some syncategorema in some particular context. For example, "'Man inasmuch as [inquantum ] he is rational is not capable of braying'—that is,  no man is capable of braying and  every man is rational and  no rational entity is capable of braying and  because an entity is rational it is not capable of braying" (p. 115).
The sophismata that played an increasingly important role in investigations of syncategoremata from the thirteenth through the fifteenth centuries may be characterized as falling somewhere between fallacies of ambiguity and exponible propositions. In the independent treatises on syncategoremata prevalent in the thirteenth century, the sophismata served as the illustrations of the principles uncovered in the investigation and characteristically took the form of an assertion (the sophisma proper) followed by a proof, a counterargument, and an adjudication of the apparent paradox by an appeal to the principles. For example, "Suppose that exactly one individual of each species of animal is running. Then [a ] every animal is running. (Proof: a man is running; a lion … ; a goat … ; and so on with respect to the individuals; therefore every animal is running.) But [b ] every man is an animal; therefore every man is running. [Solution:] [a ] is ambiguous since [because of 'every'] the word 'animal' can distribute for the remote parts (or the individuals belonging to genera)—in which case it is false, since it is then distributed for all its individuals—or for the genera of the individuals (or for the proximate parts)—in which case the minor [b ] is plainly not accepted" (William of Sherwood, Syncategoremata, edited by J. R. O'Donnell, in Medieval Studies, Vol. III, p. 49).
The continuity of the development of the logica moderna was enhanced by the fact that from the twelfth century through the fifteenth century the same sophismata were treated from varying points of view, but at the same time the number and intricacy of the sophismata were constantly increasing. As a result the modernists of the fourteenth century frequently produced treatises titled Sophismata in which large numbers of them were grouped according to the syncategoremata at issue in them, and the investigations that had begun in separate treatises on syncategoremata were pursued in the Sophismata and Exponibilia of the late Middle Ages.
The notion that grammar and philosophy were intimately related was one of the most pervasive of the assumptions that determined the character of medieval thought. It is probably to be explained by the facts that grammar was one of the very few inquiries to survive antiquity intact and that the only ancient philosophy available during the early Middle Ages was Aristotle's Categories and De Interpretatione, works of a decidedly grammatical cast. The usual view of the connection between the two subjects was the one expressed most memorably by John of Salisbury—"Grammar is the cradle of all philosophy" (Metalogicon, Book I, Ch. 13)—and the logica moderna, by far the most impressive medieval contribution to semantics, is a clear example of the influence of grammar on philosophy. The influence ran the other way, however, in the development of "speculative grammar" (grammatica speculativa ), or the doctrine of the "modes of signifying" (modi significandi ), a movement that began somewhat later and subsided somewhat earlier than the logica moderna. Although there were some connections between the two movements—for instance, Roger Bacon (1214/1220–1292), one of the first of the speculative grammarians, or "modists," also contributed to the logica moderna —they tended to be mutually independent and to some extent theoretically opposed developments in the history of semantics.
The most important single factor in the rise of speculative grammar in the early thirteenth century was the enthusiasm for the notion of a science, then being rediscovered in the Posterior Analytics of Aristotle and in his Arabic commentators. For a time it was the aim of every study to achieve the status of an Aristotelian science, a body of necessary knowledge deductively demonstrated, and two facts seemed to stand in the way of certifying grammar as a science. For one thing, as it had been presented by Priscian and Donatus, grammar was simply a set of observations about correct constructions without any attempt at explanation of the correctness; but only knowledge "by causes" qualified as scientific. For another, even Peter Helias (fl. 1150), who in his commentaries on Priscian had been the first medieval to attempt explanations of grammatical facts, had maintained that there were as many grammars as there were languages; but a unified subject matter was a prerequisite of a science.
Thus a science of grammar was not to be found in the grammatical authorities, and it seemed one never would be found as long as grammar was conceived of as something to be discerned only in the investigation of actual languages. Robert Kilwardby (d. 1279) set the stage for speculative grammar when he argued that "since a science remains the same for all men and its subject matter remains the same, the subject matter of grammar must remain the same for all men. But grammatically ordered speech or articulate utterance that can be put into a grammatical pattern is not the same for all men, and for that reason it will not be the subject matter of grammar [as a science]" (Commentary on Book I of Priscian's Ars Minor ). No science of languages was possible, but grammar might become the science of language, the scientia sermocinalis, if the variable external trappings were ignored and one concentrated on the conceptual underpinnings—which, as Aristotle had pointed out, were the same for all men. Thus Roger Bacon was led to proclaim that "with respect to its substance grammar is one and the same in all languages, although it does vary accidentally " (Grammatica Graeca, edited by E. Charles, p. 278), and this became the often repeated fundamental assumption of the speculative grammarians.
As it developed, speculative grammar took the form of an attempt to provide an Aristotelian ontology of language, finding analogues in the various parts of speech for matter, form, substance, process, and so on. As Siger of Courtrai (d. 1341) put it, "grammar is the scientia sermocinalis, which considers discourse and its properties [passiones ] in general for the purpose of expressing principally concepts of the mind by means of interconnected discourse" (Summa Modorum Significandi, edited by G. Wallerand, p. 93). Siger then cited Avicenna for the Aristotelian doctrine that concepts are the same for all men because they are the result of experiencing extramental entities, which are the same for all men. Thus, the ontology that applies to the extramental entities must apply as well to the concepts derivative from them (if they adequately copy the extramental entities) and, in turn, to the discourse employed to express those concepts (if it is to be adequate for that purpose). "Therefore modes of being, or properties of things …, precede a mode of understanding as a cause precedes an effect" (ibid.). In the same way a mode of designating (modus signandi ) follows a mode of understanding, "since a thing is understood and also conceived of before it is designated by means of an utterance [vox ], for utterances are signs of passions, as is said in De Interpretatione, Ch. 1" (p. 94). When the understanding assigns a given concept to an utterance, the merely physical utterance becomes a word (dictio ).
Up to that point the semantic theory underlying speculative grammar might fairly be described as a technical restatement of Aristotle. It was only with the introduction of its "modes of signifying" that the theory acquired its novelty and notoriety. (It was repeatedly attacked and ridiculed by logicians and grammarians of the late Middle Ages and even more strongly assailed by the Renaissance humanists.) As an utterance becomes a word by means of a mode of designating, so a word becomes one or another part of speech by means of a mode of signifying. The modes of signifying, however, are not modes of the utterance of the word but are "certain concepts of the understanding itself" (ibid.). The kind of concept in question seems to be one that links the word to some Aristotelian mode of being. Thus, the kind of concept involved in the mode of designating is the kind that supplies a significatum for the utterance "horse," transforming it from a mere sound into a word, while the mode of signifying consists in the recognition that it is substance that is signified by the word horse. And when the understanding adds to that general mode of signifying—substance—the specific mode of signifying—quality—then horse has been transformed in turn from a mere word to a substantive and from a mere substantive to a noun (pp. 94–95). Along these same lines, the utterance horse will eventually be accounted for as a common concrete noun, and similar patterns of modes of signifying are invoked in order to account for the other parts of speech.
Aristotelian ontology was employed by the speculative grammarians not only as the link between grammatical forms and modes of extralinguistic being but also as a picture of intralinguistic relations. Thus, verbs stood at the pinnacle of the linguistic microcosm because just as the other animals are submissive to man, so the inflections of the other parts of speech in a sentence are ultimately submissive to the verb. The infinitive of a verb, however, was analogous to primary matter in substances. And just as the organisms capable of fewest adaptations are ranked lowest in the kingdom of nature, so the indeclinables, the syncategoremata, are the most inferior parts of speech (pp. –).