Roman Grice

Dialectic, they said, is indispensable and is itself a virtue, embracing other particular virtues under it. [...] Without the study of dialectic, they say, the wise man cannot guard himself in argument so as never to fall; for it enables him to distinguish between truth and falsehood, and to discriminate what is merely plausible and what is ambiguously expressed, and without it he cannot methodically put questions and give answers. (Diogenes Laertius 1925, 7: 46–47) 

Almost all of the original writings of the Stoics have been lost and what is known about them is reconstructed from secondary sources compiled 500 to 600 years later. 

The school started with Zeno (335–263 BCE), a philosopher of Phoenician origin who taught in the Stoa in Athens – whence the name Stoics. A leading light was Chrysippus (282–206 BCE), perhaps the greatest logician of ancient times and author of 750 books1 (Diogenes Laertius 1925, 7: 189–202). There was also Diogenes of Babylon (240–150 BCE) and Antipater of Tarsos (mid-second century BCE). Much of the information we have about the Stoic contribution to grammar comes from Sextus Empiricus (160–210 CE), a sceptic and critic of Stoicism; there are two relevant works by him: Adversus Mathematicos “Against the professors” (hereafter AM), Books 8 and 10, and Purrh?neoi Hupotup?seis “Outlines of Pyrrhonism”, Book 2 (hereafter PH 2).2 Most of the rest of our knowledge of Stoicism is to be found in Diogenes Laertius’ (fl. c. 225–50 CE) life of Zeno in Lives and Opinions of Eminent Philosophers Book 7, Ch.1 (hereafter Lives 7). De Lingua Latina “On the Latin language”, by the Roman polymath Marcus Terentius Varro (116–27 BCE), can be counted as reflecting Stoic views on grammar, and we still have Books V–X out of the original 25. 

Because he wrote specifically about language, and Latin at that, his work will be discussed separately. 

The Stoics Posidonius [135–50 BCE] defines dialectic as about truth, falsehood and that which is neither. Chrysippus takes it to be about signs and things signified. Such then is the gist of what the Stoics say in their theory of language. (Lives 7: 62) The Stoics seem to have established the basis for what became the ‘trivium’ of the scholastic educational system in the middle ages (see Chapters 7 and 8): * Dialectic – logic, sound reasoning; * Grammar – correct language use; and 1. Many of these books would be about the length of a modern academic book chapter rather than a modern academic book. A better comparison is with the 39 books in the Old Testament and 27 books in the New Testament of the King James Bible. 2. In Sextus Empiricus 1955, AM 8 is ‘Against Logicians’ II, in Vol. 2; AM 10 is ‘Against Physicists II’, in Vol. 3; and PH 2 is in Vol.1. See Sextus Empiricus 1998: 69 for discussion of the term math?matikos. 60 The Western Classical Tradition in Linguistics * Rhetoric – a persuasive way of expressing oneself.3 As with Plato and Aristotle, linguistic analysis by the Stoics was partly a means to an end, and the revelations about Stoic grammatical insights are presented within the discussion of logic (Lives 7: 43–83). 

Grammar achieved independence from logic when the Alexandrian grammarians established a pedagogical motivation for its study and linguistic analysis was the means to a different end. However, another very strong motivation for the Stoics and later the Alexandrians was to formulate the canons for ‘correct Greek’, for Hell?nismos. For Sextus’ very sceptical account of Hell?nismos see AM I: 176–240, where there is a commonsensical discussion of what should count as the standard for a language. He favours eclecticism and common usage, citing Aristophanes with approval: Speaking like middle-class citizens all, Not with the fop’s effeminate drawl, Nor with the rustics’ vulgar bawl. (AM 1: 228, Sextus Empiricus 1955, Vol.4, p.131) The Stoics had a semantically oriented theory of language. 

They divided language into what can be said, lekton (PL lekta), and the way in which things can be expressed – as ph?n? (PL ph?nai), literally “sound”, but better interpreted as “utterance”. If this seems to neatly match the distinction in modern linguistics between content and form, it is not quite so. Form was taken to be a function of meaning such that word forms are not arbitrary symbols, but reflect the characteristics of the thing denoted. In other words, the Stoics were naturalists (see Chapter 2 and the upcoming discussion of Varro). 

Perhaps because of their preference for naturalism and the belief that a true name reveals the essential nature of the referent, the Stoics used the verb d?loun “reveal, indicate” where today we would use denote or refer. 

True and false have been variously located in what is signified [to s?mainomenon], in speech [ph?n?], and in the motion of thought [kin?sei t?s dianoias]. The Stoics opted for the first of these, claiming that three things are linked together, what is signified, that which signifies [to s?mainon] and the referent [to tunchanon]. That which signifies is speech, e.g. ‘Dion’, what is signified is the specific state of affairs [auto to pragma] indicated by the spoken word and which we grasp as existing dependent on [paruphistamenon] our thought but which the barbarians do not understand although they hear the sound; the referent is the external existent, that is, Dion himself. Of these, two are somatic, speech and the referent. But the state of affairs signified [to s?mainomenon pragma], the lekton, is asomatic, and this is true or false. (AM 8: 11–12 [translation indebted to Long 1996: 76f]) 

The Stoics established relationships that can be seen as a forerunner to the semiotic triangle found in Ogden and Richards 1923: 11; see Figure 4.1.4 S?mainon is literally translated “sign”; given the naturalist beliefs of the Stoics, the English translation “sign” is preferable to “symbol”. 

The s?mainon is the product of the ph?n?; it is corporeal or somatic 3. In the early seventh century St Isidore of Seville quotes Varro saying in the (now lost) Disciplinae ‘Dialectic and rhetoric are as in man’s hand the closed fist and the open palm, the former drawing words together, the latter scattering them’ (Etymologies II.23.1, Brehaut 1912: 116). 4. See Chapter 13, also Lyons 1977: 96, for alternative terms at the vertices of the semiotic triangle. The Stoics and Varro 61 (s?matikos). 

The Stoics had a materialistic metaphysics in which anything that can act upon another thing (i.e. have some effect) or can itself be acted upon, is somatic. 

The tunchanon “what is there, the referent” is also somatic in the Stoic view. S?mainomenon “that which is signified” is the general term for what we now call sense or intension. 5 The lekton “that which is sayable” seems to be applied to what we now call illocutionary types. A lekton may be either ‘complete’ (autoteles “well-formed, expressing a complete thought”) or defective (ellip?s “elliptical, partially-formed”); see Figure 4.2. 

In the Greek sentence H?mera esti ‘day is’ “It is day”, h?mera and esti are both phonetically realized signs and each has a sense; this combination of them makes a statement, the lekton “it is day”, which has a truth value. The ‘defective’ lekta are the meaningful aspect of a pt?sis (named for its nominative case) “subject, thing spoken about” or a kat?gor?ma “predicate”. These are not parts of speech or sentence constituents but semantic functions: without the other, each of the two 5. Sense is decontextualized meaning such as is found in a dictionary. The earliest recorded use is in the dictionary of Palsgrave 1530. Sense describes intension, i.e. the characteristic property or set of properties of a typical denotatum. Both are abstract, in Stoic terms ‘asomatic’. On the lekta see Mates 1961; Graeser 1978; Itkonen 1991; Householder 1995a; Long 1996. Figure 4.1. The Stoic semiotic triangle and that of Ogden and Richards. Figure 4.2. Categories of the lekton (after Mates 1961: 16). 62 The Western Classical Tradition in Linguistics components is a defective lekton because a complete articulated thought requires that anything mentioned be predicated (notionally, not necessarily lexically); each of the defective lekta makes a semantic contribution to the statement, H?mera esti. The name Dion, for instance, only forms a lekton if Dion is being named or called or is predicated as walking, or the like. The predicate writes is defective without the notion that someone or something writes because a predicate is what is said of something; in other words a thing associated with one or more subjects; or, again, it may be defined as a defective expression which has to be joined on to a nominative case in order to create a judgment. (Lives 7: 64) The ‘complete’ lekton is an illocutionary type. That is, the Stoics recognized the (partial) correlation between mood or clause-type and illocutionary force:6 

There is a difference between judgment [axi?ma], yes-no question [er?t?ma], and wh– question [pusma], as also between imperative, adjurative, optative, hypothetical, vocative, whether these terms are applied to a thing or a judgment. For a judgment is that which, when we set it forth in speech, becomes an assertion, and is either false or true;7 a yes-no question is a thing complete in itself (like a judgment) but demanding an answer, e.g. Is it day? and this is so far neither true nor false. Thus It is day is a judgment, Is it day? a yes-no question. A whquestion is something to which we cannot reply by signs, as you can nod yes to a yes-no question; but you must express the answer in words, He lives in this or that place. An imperative is something which conveys a command. […] Yes-no questions, wh- questions, and the like are neither true nor false, whereas judgments are always either true or false. (Lives 7: 66–67, 68) Judgments, questions, etc. as discussed here are illocutionary types rather than illocutionary acts. Additional complete lekta are the vocative (prosagoreutikon, kl?tikon – O King Agamemnon), the exclamatory (thaumastikon – How clever he is!), the rhetorical question (epapor?tikon), the wish or prayer (eutikon), the imprecative (aratikon – May he rot in hell), the oath (omotikon – I swear by Zeus), the propositive (ekthetikon – Let n be the number of chairs in the room), and the hypothetical (hupothetikon – Assume that the world is flat). The Stoics employed the technical term phantasia (often translated “presentation”) for the mental counterpart to things perceived or conceived of; it is their alternative to Plato’s Idea. 

Whereas the lekton is asomatic (incorporeal, abstract), the phantasia, being a mental affect, is somatic. It is a sort of imprint stamped upon the mind, a cognitive model; and a given perception of the real object is matched with this to determine whether or not it is an illusion (Lives 7: 46, 50). A phantasia varies between people: a statue is viewed in a totally different way by the trained eye of a sculptor and by an ordinary man. (Lives 7: 51; cf. Plato’s Theaetetus 160a). When someone sees and recognizes Dion walking, they recognize Dion through a phantasia of him. If they say Dion is walking, this is distinct from the phantasia and from Dion himself. And what is meant by the proposition Dion is walking, the lekton, is an affirmation about Dion. The lekton is something that subsists (huphistamenon) in conformity with a 6. See Allan 2006a for a discussion that focuses on English. 7. Greek axioun denotes “accept or reject” (Lives 7: 65). The Stoics and Varro 63 rational phantasia (Lives 7: 63; AM 8: 70); furthermore, it is something conveyed in a sentence (logos) (AM 8: 70). There is a difference between ph?n? [“a sound”] and lexis [“a word”], because while ph?n? may be mere noise, lexis is always part of a language system [de to enarthron monon]. Lexis is different from logos [“a sentence”], because the latter always signifies something, whereas lexis may be unintelligible as in [the onomatopoeic] blituri [the sound of a twanging harpstring, cf. English boing], which logos never is. And speaking [to legein] is to do more than merely utter ph?nai, because things are meant [lit. ‘spoken’], and these are lekta. (Lives 7: 57) 

A problem with translating ph?n? as “phonetic form”, lexis as “lexical item”, logos as “sentence”, and lekton as “illocutionary type” is that these modern terms carry with them theoretical associations (connotations) that may be inappropriate to Stoic grammar. This problem of the interpretation of terms in a metalanguage is ubiquitous: it applies not only to the terms used by Aristotle and the Stoics, but also when looking at different linguistic theories in the twenty-first century. Chrysippus identified five parts of speech (Lives 7: 58): ‘Indeed, according to the Stoics his parts of speech were five: proper noun, common noun, verb, pronoun or article, and conjunction’ (Priscian 527 II: 16). A proper name (onoma) such as Diogenes or Socrates signifies a quality peculiar to a particular individual. Qualities are not something that an entity partakes of, as was the case in Aristotle; it is characteristic of Stoic materialism that qualities are present within the entity. A common noun (pros?goria), e.g. man, horse, is the part of speech signifying a shared quality. Adjectives were included among pros?goriai because they are inflected like nouns; this classification remained part of the Western Classical Tradition until the twentieth century. 

A proper or common name may function as a pt?sis (propositional argument). A verb (rh?ma) functions as a kat?gor?ma (predicate). An article (arthron) is a declinable part of speech distinguishing the genders and numbers of nouns (cf. Greek ho “the.M.SG.”, hai “the.F.PL.”; Lives 7: 58). Pinborg 1975: 99 suggests that being a formal rather than semantic definition, this one is Alexandrian rather than Stoic; however, formal definition of the verb is regularly attributed to Chrysippus, so perhaps this classification has the same origin. The atomic (haploun) judgment may be definite, indefinite, or intermediate. Diogenes Laertius (Lives 7: 70) is unclear on this, and the following account is from Sextus Empiricus AM 8: 96–98. Definite judgments are those paralinguistically marked by pointing and usually include houtos “that; he” indicating the existence (ousia) of a uniquely identified individual: this is very much like Kaplan’s ‘Dthat’ or ‘Dhe’ (glossed “demonstrative [i.e. kinesically indicated] that” and “[kinesically indicated] he”; Kaplan 1978). An indefinite (aoriston) judgment has a subject such as tis “someone” or ekeinos “the person or thing mentioned”, both of which can be used of any person and are not bound to a uniquely and incontrovertibly identified individual. 

An intermediate judgment is something like Socrates is walking: this is not ‘indefinite’ because it identifies the individual; on the other hand, it is not ‘definite’ because no demonstrative is used. The Stoics recognized that when Houtos peripatei “That [man] is walking” is true, so is Tis peripatei “Someone is walking”. Clearly the domain of definiteness is indicated by the subject noun phrase, and in particular by the article. A conjunction (sundesmos) is defined 64 The Western Classical Tradition in Linguistics as an indeclinable part of speech binding various parts of an utterance together. A sixth part of speech, identified by Antipater of Tarsos, is merely named mesot?s “in-between” (Lives 7: 57), a term later used for middle voice. It is assumed to mean “adverb”, and perhaps gets the name “in-between” because it was recognized as a kind of adjective (pros?goria) modifying a verb. Another hypothesis (Dionysius 1987: 188) is that the term was applied only to adverbs in –?s (e.g. kal?s “nobly”, soph?s “wisely”) which related to a genitive plural form shared by all three genders (e.g. soph?n “of the wise”). Pinborg 1975: 101 links the Stoic parts of speech with the four Stoic categories (Mates 1961: 18; Graeser 1978: 78f): arthron to the first category of hupokeimenon “subject” (presumably via houtos); both proper and common nouns belong to the second category poion “quality” (they are defined in terms of qualities); verbs are divided among the third and fourth categories of p?s echon “state” and pros ti p?s echon “relation” (presumably oneand two-place verbs respectively). The four categories are such that the fourth is dependent on the third, the third on the second, the second on the first. And there was a prior category: ti “the indefinite something” (Mates 1961: 18). Nominals are marked by case; but in the Stoic view (Frede 1978: 32) the cases identify functions of their referents (to use today’s terms). Pinborg 1975: 84 suggests that the orthos pt?sis “upright case, nominative” is so-called because the typical subject is an actor who causes an effect and is therefore quintessentially somatic within Stoic metaphysics. 

This suggestion correlates with Apollonius Dyscolus (see Chapter 6) contrasting the energeia “active force” of the orthos with the pathos of the other cases (plagiai pt?seis). The genitive, dative, and accusative (Lives 7: 65) indicate other aspects of affecting or being affected. For example, the accusative (aiti?tik?) is typically the affected or effected object.8 Aiti?tik?, which means “that which is caused”, was translated into Latin as ad + cusativus “to + having a passive tendency” and so gave rise to the adjective accusativus, whence our term accusative – which has no connection with the verb accuse. Our grammatical term case is from Latin casus “falling”, which is a direct loan translation from Greek pt?sis. The Stoic notion was of cases – or more accurately, case-roles – falling from the upright actor to the acted-upon patient, as shown in Figure 4.3. upright case [nominative] (actor) genitive (possessor) dative (recipient) accusative (patient) 8. In Joe made a table ‘a table’ is the Effected object; in Ed scratched the table ‘the table’ is the Affected object. Figure 4.3. Cases falling from the upright. The Stoics and Varro 65 A verb (rh?ma) signifies an isolated predicate (note the difference from Aristotle) or, according to some Stoics, a part of a sentence without case (apt?ton) signifying something that can be attached to one or more subjects (pt?sis), e.g. Greek leg-? (RH?MA-PT?SIS speak-I) “I speak”. Molecular judgments are judgments combined by sundesmoi “conjunctions, logical connectives”. There are three kinds of molecular judgments: the conditional (sun?mmenon) “if ... then”, the conjunction (sumpeplegmenon) “and”, and the disjunction (diezeugmenon) “or” (Figure 4.2 and Lives 7: 71ff). To these the Stoics added negation, formulaically expressing all negations by placing the word not in front of the negated sentence. Thus Not it is day (ouchi h?mera estin) means “It is not day”. 

Of the negative propositions one kind is the double negative. By double negative is meant the negation of a negation, e.g. It is not not-day. This presupposes [tith?si] It is day. (Lives 7: 69) So begins the tradition that double negatives cancel each other out. While true in logic, the situation is more complex in normal language usage; e.g. English I cannot not go is synonymous with I must go, which implies, but is not implied by, I can go. Like Aristotle, the Stoics distinguish denials like No one is walking from privatives: This man is unkind (the Stoic system places ‘unkind’ first in sequence). Whereas Aristotle analysed the structure of propositions to develop a deductive logic of terms, the Stoics developed propositional logic, recognizing five basic schemas (PH 2: 157–58): (1) If p then q. P holds, therefore q. 9 (2) If p then q. Not-q, therefore not-p. (3) Not(p and q). P holds, therefore not-q. (4) P or q. P holds, therefore not-q. (5) P or q. Not-p, therefore q. Notice the conclusion that ‘not(p and q)’ is equivalent to ‘p or q’. And notice that the truth of the molecular proposition ‘If p then q’ is guaranteed under (1) and in (2) but falsified if p holds but not-q; see Lives 7: 73. From these five basic schemas others can be inferred, e.g. (6). (6) If (p and q) then r. Not-r, but p holds, therefore not-q. Therefore not(p and q). There is controversy over whether or not the Stoics believed that this system is complete (see Mueller 1978 for discussion), but this is a matter of concern to the history of logic rather than of linguistics. Relevant for us is the effect that these philosophical concerns had on linguistic analysis. 

The development of propositional logic is important in linguistic semantics, though less significant than the development of predicate logic in the nineteenth century. Egli 1983 says that Kripke’s modal logic (Kripke 1963) was directly influenced by that of Diodorus Kronos (died c. 282 BCE): p is possible now iff p is true now or will be true later (Diodorus) p is possible in our world iff p is true in a world accessible from ours (Kripke) 9. Instead of p the Stoics used ‘the first’, instead of q ‘the second’; AM 8: 227. 66 The Western Classical Tradition in Linguistics Kripke replaced points of time by possible worlds and the relation “to be now or later” by the accessibility relation. (Egli 1983: 79)10 The Stoics identified a number of verbal categories: tense, aspect, mood, transitivity, and voice. These secondary grammatical categories were all identified on the basis that they affect the meanings and implications of propositions, and not on morphosyntactic form. For instance, a true assertion in the past tense such as There was a tree standing there has a present tense counterpart that was true in some earlier world: keeping the locality constant, it was once true that There is a tree standing there (AM 10: 91). If I see something or I am looking at something, then I have seen it; but if I am building something, I have not yet built it. 

If Helen has a husband, then she has had at least one husband; if Helen has had three husbands, it is not the case that she has three husbands (AM 10: 98). If a woman is pregnant today, then she will either abort or give birth in the future. If a woman gives birth today, she was pregnant yesterday. These are the kinds of considerations that underpin the Stoic analysis of language. It led them to distinguish the tense-aspect system shown in Table 4.1 (Pinborg 1975: 94; Robins 1997). There also existed a future perfect in Ancient Greek; but because it was restricted to the Attic dialect and the analysis is semantic rather than morphological, it was ignored even though it could fit into the top right cell in Table 4.1. The Stoic recognition of aspect as a grammatical category distinct from tense was ignored by the Western Classical Tradition until the twentieth century. Table 4.1. The Stoic tense-aspect system. PAST PRESENT FUTURE COMPLETE PLUPERFECT egegraphei “had written” PERFECT gegraphe “has written” NEUTRAL AORIST egrapse “wrote” FUTURE grapsei “will write” INCOMPLETE IMPERFECT egraphe “was writing” PRESENT graphei “is writing” We have already seen evidence that the Stoics recognized the correlation between mood or clause-type and illocutionary type. In parallel with the binary division of cases, it is reported that the indicative was classed as orthos and the other moods plagiai. The division is analogous to (but not identical with) our modern distinction between unmarked and marked, respectively. Voice and transitivity were discussed under the head ‘types of predicate’ (Lives 7: 64). A transitive verb is a two-place predicate (suntassomenon) which takes a nominative subject and also an oblique case. Transitives divide into active (orthon) and passive (path?tikon) – sometimes called huption “back to front, reversed”. 

A one-place predicate is intransitive 10. On the development of modal logic among the Stoics see PH 2: 110; AM 8: 115; Kneale and Kneale 1962: 128–38. p iff q means “p if and only if q”, i.e. iff is biconditional. The Stoics and Varro 67 (oudeteron). Reflexive and reciprocal predicates (antipeponthota) have passive form but active meaning, e.g. keiretai “he gets his hair cut”, sunethento “they made compacts with each other”. All these were known as sumbamata because they co-occur with the nominative. There are also parasumbamata, such as impersonal verbs which have no explicit subject or one in an oblique case. Egli 1987 persuasively argues that an infinite number of lekta can be derived by a finite number of ‘combination’ and ‘inclusion’ rules, which are similar to phrase structure rules. For example, a nominal in the oblique case with an incomplete predicate together form a complete predicate (one roughly equivalent to in categorial grammar). A negation sign and a statement form a statement. 

There are ‘lexical’ insertion rules, e.g. Dion is insertable as an onoma; and there are ‘transformations’ such as that a preterite enklisis to a statement results in a statement. Egli claims that enklisis ‘was the historical antecedent of parag?g?, declinatio, “inflection”, etc.’ (ibid. 131). According to Diogenes Laertius, the part of dialectic that deals with language is concerned with written language and the parts of speech, with a discussion of errors in syntax [solecism] and in single words [barbarism], poetical diction, verbal ambiguities, euphony and music, and according to some writers chapters on terms, divisions, and style. (Lives 7: 44) This covers not only the study of language as used in reasoning (dialectic), but also its rhetorical and poetic uses. The Stoic views on rhetoric are very similar to those of Aristotle: what is important is the invention of arguments, their expression, arrangement, and delivery; the delivery requires an introduction, replies to opposing points of view, and summing up (Lives 7: 43). 

This strategy reappears in twentieth century descriptions of the structure of academic texts, such as Swales 1990. 

The Stoics identify five excellences (aretai) of speaking (Lives 7: 59). 

* Use a faultless grammar of the standard language; this was known as Hell?nismos. 

* Be lucid: ‘present the thought in a way easily understood.’ 

* Be concise: ‘employ no more words than are necessary for setting forth the subject in hand.’ 

* Be appropriate: use a style ‘akin to the subject’. 

* Avoid colloquialism. As with Aristotle, there are some similarities with the Gricean maxims, but these are much closer to Aristotle than to Grice. 

A good speaker also needed to be aware of ambiguity in order to exploit it gainfully or avoid it where there was a possibility of being misinterpreted. Championing Hell?nismos required the avoidance of ‘barbarism’ (mistakes with words) and ‘solecism’ (mistakes with morphology and syntax). And it gave importance to good diction, sparking a study of spoken language. 

In their theory of dialectic most of them see fit to take as their starting point the topic of sound [ph?n?. ...] While the sound an animal makes is just a percussion of air brought about by natural impulse, that of a human being is part of a language system [anthr?pou d’estin enarthros] and, as Diogenes [of Babylon 240–152 BCE] puts it, is an utterance of reason. (Lives 7: 55) 68 The Western Classical Tradition in Linguistics Furthermore, it produces a (perlocutionary) effect ‘as it proceeds from those who utter it to those who hear it’. Sounds (ph?nai) get written as words (lexeis). An utterance (logos) is a meaningful ph?n? that is issued from the mind. Diogenes goes on to report that there are 24 letters in the alphabet and each letter consists of a sound as element of speech (stoicheion), e.g. /a/; a symbol (charakt?r) ; and a name (onoma), alpha. Seven of the letters are vowels, and six are mutes (stops). This is implicit in Aristotle’s Poetics but is made explicit for the first time here (Lives 7: 57). It recurs in the Techn? Grammatik? attributed to Dionysius Thrax (see Chapter 5). Hell?nismos was also the reason for Stoic studies in etymology; Sextus, on the other hand, thought it entirely wrong-headed to seek to establish correct usage on the basis of etymology instead of commonality (AM 241–47). Perhaps the materialistic metaphysics of the Stoics was responsible for their belief in the natural origin of words; but whatever the motivation, given such beliefs it was important to know whether the original form of a word was Greek or not in order to decide whether its continued use was Hellenistic or not. It is ironic, therefore, that our principal source for Stoic etymologies is not Greek but Latin, in the work of Varro. Varro [V]oluntary affixation is a matter of usage, and natural affixation is a function of a logical system. (De Lingua Latina 10: 15; Varro 1938) 

During the third and second centuries BCE, the Greek world fell progressively under Roman domination. The Roman view was that others could excel in the liberal arts while Rome kept peace in the world (Aeneid 6, 847–53; Virgil 1968); this benefited the Greeks by giving them political stability and freedom of movement within the Roman Empire. There had long been Greek settlements in South Italy disseminating Greek intellectual ideas among the Romans; and in the eastern provinces of the Hellenistic world, Greek was widely spoken by Roman officials, and Greek literature and philosophy were held in high esteem. Romans often had Greek slaves to educate their children. 

The Roman writing system derived from that of the western Greeks, and this doubtless contributed to the belief (widely held until the nineteenth century) that Greek was mother to the Latin language. From the third century BCE Greek literature was systematically translated into Latin; but if this activity encouraged schools of language study, nothing is known of them. So prestigious was Greek literature, that Latin abandoned its indigenous poetic forms and adopted the metres and stanzas of Greece. Thus it is no surprise that all known Roman grammatical studies were parasitic on Greek forebears. 

It is said that linguistic study was introduced to Rome by Crates of Mallos around 170 BCE. Crates was head of the library and literary school in Pergamum11 and an expert on Homer. 

He went to Rome as ambassador for King Eumenes II of Pergamum and (according to Suetonius in De Grammaticis, Suetonius 1914: 2) broke his leg when sightseeing; while waiting for it to mend, Crates gave lectures on literature and Stoic philosophy. 

However, in view of the wide dissemination of Greek ideas within the Roman 11. Modern Bergama in Turkey. The Stoics and Varro 69 Empire, the fact that Stoic doctrine was directly introduced to Rome would not necessarily have given it undue prominence. 

Roman grammarians were not seeking refinements of logic or rational thought but rather how to speak correct Latin and to properly interpret Latin literature. 

This Latinitas was the counterpart to Hell?nismos among the Greeks. 

The first significant Roman writer on language was Marcus Terentius Varro, a man of diverse interests who wrote 76 books on topics as varied as agriculture, geography, the history of Rome, law, mathematics, philosophy, rhetoric, the history of literature, as well as grammar. De Lingua Latina “On the Latin language” (hereafter DLL), dedicated to Cicero, was composed between 47 and 45 BCE in 25 books of which V–X and a few fragments survive. 

Following the Introduction, Books II–VII were on etymology: Book II on objections to it, Book III on its validity, Book IV on its nature; Book V is on names of places and things associated with those places; Book VI on ideas connected with time (including the origins of verbs); Book VII on the etymology of rare and poetical words. Books VIII–XIII deal with inflectional and derivational morphology as matters of regularity (analogy) versus irregularity (anomaly) in language. Books XIV–XXV were on syntax, probably along Stoic lines.

 In DLL VI: 2 Varro displays the Stoic foundation for his approach to etymology by claiming Chrysippus and Antipater of Tarsos (among others) as his authorities. Varro’s Stoicism appears in his claim that: all things in general are divided into four [interconnected] parts and these are eternally present. 

There is never time without there being motion (for even an intermission of motion is time). There is no motion where there is no place and body; because the latter is what is moved, and the former is where [the movement takes place]. [Finally,] wherever there is motion, there is action. [Therefore] place and body, time and action are the four-horse team of elements. (DLL V: 12) 

His etymologies are said to be presented in the same sequence as the four elements are listed here, though the evidence is unconvincing. The same four elements are used in a similar schematic fashion in other works by Varro (Dahlmann 1963: 121–23); they inform his world view and were not specifically characteristic of language. Whereas the Greek etymologists were motivated by Hell?nismos, Varro was not particularly focused on Latinitas. He recognized that language changes over time: Many words indicate one thing now, but formerly meant something else, as is the case with hostis [“enemy”]: for in olden times by this word they meant a foreigner from a country independent of Roman laws, but now they give the name to him whom they then called perduellis. (DLL V: 3) Investigating the vocabulary of Latin to discover how it grew so large and yet could be used with such ease, Varro did not seek to find the original forms of words for some metaphysical purist purpose, but to explain the history and relationships between words. Thus he distinguished between primitive word forms which have to be learned independently of one another – which he says are ‘imposed’ on language learners – and inflections on these which allow for an infinite number of new words to be created (VIII: 5–6). 70 The Western Classical Tradition in Linguistics He who shows that equitatus [“cavalry”] is from equites [“cavalrymen”], equites from eques [“cavalryman”], eques from equus [“horse”], even though he does not give the source of the word equus, still gives several lessons and satisfies an appreciative person. (VII: 4) 

Nevertheless, the method Varro chose was that of the naturalists that we critically examined in Chapter 2 when discussing Plato’s Cratylus. 

The naturalist method of analysis was: describe the denotatum of the word under analysis; cast around for an appropriate phrase that bears some resemblance to the description, and which has a form bearing some resemblance to the word under analysis; if this fails, the word must be either primitive or of foreign origin and is therefore unanalysable. 

Varro himself writes: [A]lterations come about by the loss or addition of single letters and on account of the transposition or change of them, and likewise by the lengthening and shortening of syllables, and their addition or loss. (V: 6) [W]ords are divided into three groups: those which are our own, those which are of foreign origin, and those which are obsolete and of forgotten sources. (V: 10) He identifies many Latin words as being of Greek origin whereas, in fact, most derive not from Greek but from an Indo-European ancestor to both languages. We saw two of his false etymologies in Chapter 2. Here are some more examples: [The word] pater “father” [arose] because patefacit “he makes evident” the seed [semen]; for then patet “it is evident” that conception has taken place, when that which is born springs forth from it. (V: 65) Because Juno is wife to Jove, and he is Heaven, she is Terra or Tellus, [mother] Earth. Because una iuvat cum Jove “together with Jove she helps [people]”, she is called Juno. (V: 67) Ignis “fire” is named from gnasci “to be born”, because from it springs life. Fire enkindles everything which is born; consequently [living things are] warm, and when they die, they lose the fire and become cold. From the fire’s vis ac violentia “force and violence” Vulcan gets his name. (V: 70) [L]olligo “squid” was originally volligo because volat “it flies” up from the deep; now it has one letter changed. (V: 79) None of these is correct. Sometimes Varro got the fact of the relationship right, but its direction wrong; e.g. ‘Mundus is a woman’s toilet set, named from munditia “neatness”’ (V: 129). In fact munditia derives from mundus. But to be fair to Varro, he didn’t get it all wrong. For instance, the quote above from DLL V: 3 concerning hostis and: From manus “hand” comes manupretium “worker’s wage”; mancipium “possession of property” derives from capitur manu “it is taken by hand”; manipulus “maniple, troop of soldiers” [is so named] because it unites several hands; manipularis “soldier in a maniple”; manica “sleeve”. Manubrium “handle” because it is grasped by the hand. Mantelium “towel” on which the hands terguntur “are wiped”. (VI: 85) Here we see a lexical network revealed by the pursuit of lexical derivation. Manipulus derives from manus and pleo “fill” and was originally “a handful or bundle”. Units in the Roman army were led by a standard to which a manipulus of corn was tied, and the name transferred from the standard to the troop of soldiers. Another derivation-based network is found in the following; only the etymologies marked [?] are accurate. The Stoics and Varro 71 Terra “earth” is [...] named thus because teritur “it is trodden”. [...] From this, the place reserved near a town as common land for farmers is territorium [?], because it is trodden most. Then there is the linen garment extermentarium12 because teritur “it is rubbed” by the body. Again, at harvest time tritura “threshing” [gets this name] because the grain is rubbed out [from the wheat] [?], and tribulum “the threshing-sledge”, too [?]. The boundaries of the fields are called termini because those parts are trodden most on account of the boundary lane – therefore this word is pronounced with ‘i’ in some places in Latium, not terminus but terimen, and this form is found in Accius: it is the same word which the Greeks call term?n. Perhaps the Latin word comes from the Greek. (V: 21) We can see from the quotations that Varro was attempting to plot lexical relations through derivation. It led him to propose unattested reconstructed items such as *esum to link sum “I.am” to es “you.SG.are” and est “3.SG.is”. Indeed, he contrasts the ‘sterility’ of words allowing of no affixation such as et “and”, iam “now”, vix “hardly” with the ‘fecundity’ of a root (radix, a term he was the first to use) such as lego “I read”, which can be inflected for person, number, tense, and aspect and derivationally related to lectio “collection”, lector “reader”, legens “reading”, lecturus “about to read”, lecte “choicely (VIII: 9; VI: 36). Varro’s important legacy was to take the matter beyond trying to identify the original meanings of names, as Socrates does in Cratylus, and to make the earliest significant study of morphology. 

It was Varro who first distinguished between ‘inflectional’ and ‘derivational’ morphology. He did this in the course of discussing the so-called ‘analogy versus anomaly controversy’,13 which is more appropriately described as a focus on regularity versus a focus on irregularity in language structures. For some ‘analogists’, language structure and lexical meaning were originally, and ideally certainly ought to be, regular: all grammatical paradigms (case, number, tense, etc.) should be regular, and lexical meaning should be as natural and predictable as possible (e.g. females and feminine things should be represented by female gender). There is some overlap between the views of analogists and naturalists. Many analogists regarded irregularity as a sign of degenerateness, and suggested language reforms. ‘Anomalists’ exaggerated the irregularity to be found in languages: language is conventional, large parts of it are irregular but some parts happen to be regular. One of the Byzantine scholars quoted by Bekker writes: since grammar is not in all places governed by consistent principles [analogia], but often just by the tradition of usage, and we often find grammar unprincipled, they say that this is how the definition holds: grammar is a matter of practical knowledge [empeiria] with many words, but for the most part it is science [techn?]. (Bekker 1965: 730; see Colson 1919). Varro adopted the commonsensical view that regularity is to be preferred, but only when it is sanctioned by normal usage (consuetudo). It is normal usage that sanctions most 12. This is the only recorded use of extermentarium anywhere. 13. Lloyd 1996: 59f reports that the Stoics first distinguished inflectional from derivational morphology; given Varro’s Stoic orientation this is possible, but nonetheless the documentary evidence favours Varro. Blank 1982: 1–4 argues that these remarks of Varro and reports of them (e.g. Aulus Gellius, c. 123–80 CE, Gellius 1927, II.xxv) are the only evidence of an analogy vs anomaly controversy in the ancient world. 72 The Western Classical Tradition in Linguistics irregularities. In several places he points out that like must be compared with like: the similarity between Priamus (a king’s name) and Hecuba (his wife) is much less than that between the inflectionally related pairs lego “I read” and legi “I have read” or Priamus (nominative) and Priamo (dative) (VIII: 3). 

Masculine nominatives praetor and consul add –i to make the dative, so the nominative supplies the stem for case inflection; but socer and macer go different ways, and in the accusative one becomes socerum, the other macrum; in fact it is these forms minus the –um suffix which supply the stem for the other singular oblique cases, and all plurals (X: 28). Varro also says it is not the nominative singulars dux [? *ducs] and trabs which serve as the stem forms for their declension, but the nominative plurals duces and trabes that, minus the –es suffix, give the stem for all other case inflections, including the nominative singular (X: 56f). Therefore the oblique forms can sometimes be determined from the nominatives and sometimes the nominatives from the oblique forms; sometimes the plural from the singular forms, and sometimes the singular forms from the plural. (X: 59) Varro partially accounted for the split between regularity and irregularity in language by recognizing two kinds of declinatio. This Latin term, which is the source for English decline/declension, was a semantic extension of the verb meaning “bend to one side”, and there seems to be a conceptual link with the notion of pt?sis-casus-case sketched in Figure 4.3. Varro recognized the usefulness, necessity, and ubiquity of declinatio: Inflection [declinatio] has been introduced not only into Latin speech, but into the speech of all men, because it is useful and necessary; for if this system had not developed, we could not learn such a great number of words as we have learned – for the possible forms into which they are inflected are numerically unlimited – nor from those which we have learned would it be clear what relationship existed between them so far as their meanings are concerned. (VIII: 3) 

Varro uses declinatio both of the affixing of inflectional morphemes (noun declension, verb conjugation) and the affixing of derivational morphemes; so perhaps the ideal translation of declinatio would be “word-form variation” or “affixation”. There are two kinds of affixation: voluntary and natural. Voluntary affixation is that which is the product of an individual person’s volition. [He gives examples of lexical derivation.] On the other hand, natural derivation is based not on the volition of individuals acting singly, but on general agreement. [He gives examples of case inflection.] (VIII: 21f) Varro gives declinatio naturalis “inflectional morphology” the attributes of natura “natural”, constantia “consistent”, ars “known through skill”, ratio “systematic”, analogia “regular”.14 

Typically, inflectional morphology is regular because it adds a morpheme (form and meaning) to the stem of a large subset of words from one lexical class such that the combination of stem plus inflection is semantically transparent. An inflection never changes the lexical class of the stem word. For example, the English plural is an inflectional morpheme that can be attached to most nouns such that N+PLURAL normally means “more than one N”, as in cat~cats, sheaf~sheaves, mouse~mice, ox~oxen, criterion~criteria, datum~data. Despite the variation of forms, which can be explained on grounds of 14. 

The Latin is in every case a noun, but the meaning is better rendered by an English adjective. 

The Stoics and Varro 73 phonotaxis or the stem word originating in another language, there is an obvious regularity or set of regularities in the system as a whole; irregularities are readily dealt with by means of a small set of exception rules. Similarly with the tense or the aspect system, both of which are inflectional. By contrast, declinatio voluntaria “derivational morphology” has the attributes of impositio or voluntas “creation by an individual’s will”, inconstantia “inconsistency”, historia “known by listing [as a special case]”, consuetudo “known by convention, because unsystematic”, anomalia “irregularity”. In Varro’s etymology, the adjective voluntaria is lexically related to voluntas “will” and volatus “flight” ‘because the mind flies instantly wherever it wishes’ (VI: 47), see Taylor 1974/5: 27; this is what allows the mind to ‘impose’ names (word forms) on things. For instance, at his master’s will a slave might be named Ephesius after his town of origin, or Artemis after someone called Artemidorus. The original names (what Kripke 1972 called, using semantic extension, ‘baptisms’) were all imposed in this way. Derivational morphology is comparatively irregular, and sometimes subject to individual rather than dialectal variation, e.g. appropriacy/appropriateness, suitableness/suitability. Derivational morphemes apply to only a small subset of members of the class of the stem word, and often they cause its lexical class to change. For instance, -en attaches to some adjectives to turn them into causative or inchoative verbs, black~blacken, red~redden, sad~sadden, but there are no verbs *greenen, *bluen, *happien. Varro illustrated similar phenomena with the examples in Table 4.2 (adapted from DLL VIII: 54– 55). In derived words the semantic content of the stem plus the derivational morpheme is rarely transparent except in the case of new coinings: contrast well-established, widely used cooperate with the newer, less frequent co-occur; or consider the various semantic effects of the so-called “causative” –ize suffix in legalize, atomize, computerize, womanize, fraternize. Such differences between inflectional and derivational morphology are universal. Table 4.2. Derivational irregularity. S T E M D E R I V A T I O N sus “pig” ovis “sheep” bos “ox” avis “bird” suile “pigsty” ovile “sheepfold” *bovile *avile avis “bird” ovis “sheep” aviarium “aviary” *oviarium vinum “wine” creta “chalk” (for cleaning) unguentum “perfume” vinaria “wine shop” cretaria “chalk shop” unguentaria “perfume shop” caro “meat” pelles “hides” calcei “shoes” *carnaria (laniena) “butcher’s shop” *pelliaria (pellesuina) “leather shop” *calcearia (sutrina) “shoe shop” 74 The Western Classical Tradition in Linguistics Irregularity may be a function of declinatio voluntaria “derivational morphology”, but some of it is not due to derivational morphology. For instance, Varro draws attention to the fact that while most nouns distinguish singular from plural (puella “girl”, puellae “girls”), faba “bean [F.SG.]” was also used to mean “beans collectively” (IX: 38) – for the same reason that sugar in English is collectively used of grains of sugar. 

A distinction is made between the sexes of creatures that are significant in the speech community, such as equus “stallion” and equa “mare”; but no such distinction is normally made between a male and female crow (compare crow with hen and rooster in English). Varro adds that all pigeons were once known as columbae [F.PL], but once they began to be bred as domestic animals it became necessary to distinguish between male columbus and female columba (IX: 56). And further: The meaning may naturally allow for three genders to be distinguished by language users, such as with doctus, docta, doctum [“learned+M., l.+F., l.+N.”]; for learning is compatible with all three. Use has taught us to differentiate a learned thing from human beings; and among the latter to distinguish the male from the female. But in respect of a male or a female or what is neither, the nature of the male does not shift, nor that of the female, nor the neuter nature; and for this reason there is no feminus, femina, feminum [“woman+M., w.+F., w.+N.”]; similarly with the others. (IX: 57) We turn now to what we know of Varro’s syntactic assumptions. As procedural methods for analysing language, Varro recommends not only taking into account the characteristics of the denotation (IX: 40; IX: 42), but also a variety of formal diagnostics. For example, to determine the gender of a noun he suggests (a) formal analogy with a noun of known gender (VIII: 41; VIII: 81), which can be facilitated by setting up paradigms (X: 43–44); (b) cooccurrence with a demonstrative such as masculine hic, feminine haec (IX: 41); and/or (c) use of the diminutive form (Fragments 9, 10). He divided the parts of speech into four on a purely formal basis. 

He says there are two criteria for classification: (i) the kind of inflection that a stem takes, and (ii) its form (X: 11). The primary categories are identified using the first criterion, their subcategories by the second (VIII: 44; X: 17). * Appellandi (because they typically name): inflected for case but not tense (quae habet casus neque tempora). Exemplified by noun and adjective; but in a construct like homo doctus ‘man learned’ the noun is notionally prior and, normally, first in sequence. * Dicendi (perhaps because a single verb may constitute a sentence): inflected for tense but not case (exemplified by verbs). * Iungendi (because they subordinate propositions): inflected for both case and tense (exemplified by participles, and in X: 34 named participalia. According to Priscian 527 II: 16, the Stoics had classified participles along with verbs.) * Adminiculandi (because of their supporting function): inflected for neither case nor tense. Exemplified by conjunctions and adverbs. In the verb?adverb construct scribit docte ‘writes knowledgably’ the verb is prior and normally first in sequence. As shown in Table 4.3, this use of two inflectional characteristics [±case, ±tense] yields a very neat distinctive feature analysis which uniquely distinguishes each category – as Varro himself recognizes (X: 17). Notice that this system stands apart from declinatio voluntaria, The Stoics and Varro 75 Table 4.3. Varro’s use of inflectional characteristics to distinguish parts of speech. – T E N S E + T E N S E +CASE nominals/articles participles –CASE conjunctions/adverbs verbs yet each inflectional category can be exemplified by a set of derivationally related words (VI: 36; VIII: 58): [+case, –tense] appellandi, e.g. lector “reader”, lectio “collection” [–case, +tense] dicendi, e.g. leges “you will read”, lego “I read” (for Varro, the root form) [+case, +tense] iungendi, e.g. legens “reading”, lecturus “about to read” [–case, –tense] adminiculandi, e.g. lecte “choicely, selectly” The appellandi [+case, –tense] class divides into nominals and articles (including pronouns). Articles may be definite pronomina such as hic, haec “this M.SG., F.SG.”, or indefinite provocabula such as quis interrogative “who”, quae relative “who/what F.SG.” (VIII: 45). The nominals divide into vocabulum “common noun or adjective” and nomen “proper name” (X: 20). Varro points out that Latin has six cases whereas Greek has only five, and (in Stoic fashion) he offers a semantic description of them (VIII: 16): NOMINATIVE: what something is called (e.g. Hercules) VOCATIVE: the form you use when you call someone (e.g. Hercul?) ACCUSATIVE: what you are calling about (e.g. Herculem) ABLATIVE: by whom the calling is done (e.g. Hercule) DATIVE: to or for whom you call (e.g. Herculi) GENITIVE: of whom you call (e.g. Herculis) 

He identifies three genders (sexum) as in doctus “learned” masculine, docta feminine, and doctum neuter; and number: ‘multitudinem, unum an plura significet, ut hic hi, haec hae’ “number, signifying more than one, as in this [M.SG.] these [M.PL.], this [F.SG.] these [F.PL.]” (VIII: 46). He describes a matrix of six cases intersecting with six genders (three singular and three plural), yielding the distinctive feature matrix in Table 4.4 (next page). The text which discusses and exemplifies the cells in the matrix is unfortunately missing, and the example is invented. In the extant text Varro makes no provision for the five noun declensions of Latin. The dicendi [–case, +tense] class has six subcategories (species): tense (tempus), person (persona), interrogative (rogandi), declarative (respondendi), subjunctive (optandi), imperative (imperandi) (X: 31). Varro identifies three tenses: past, present, and future; the past and future forms of verbs derive from the first person singular present form. He continues: 

The persons of the verb are also of three kinds: the one who is speaking, the one addressed, and who or what is spoken about. Every verb has three persons, which [to some extent] explains why there are so many verb forms. (VIII: 20) 76 The Western Classical Tradition in Linguistics The last four subcategories of the dicendi are moods, but no category of mood or illocutionary type is recognized in the extant texts. Varro groups them to identify the impersonal passive as containing mood and tense but no person (X: 32). Aspect is recognized in IX: 96–100, X: 33, and X: 48, giving rise to the tense-aspect paradigm exemplified by a first conjugation verb in Table 4.5. Despite his Stoic inspiration, Varro prefers formal contrastive analysis to semantic or notional classes. Aspect in Latin is very different from aspect in Greek; there is no aorist and Varro is the only Roman grammarian to identify the future perfective indicative (X: 48), which is often indistinguishable in form from the present perfect subjunctive (Taylor 1990; 1995). Table 4.5. Varro’s tense~aspect paradigm. PAST PRESENT FUTURE ACTIVE COMPLETE amaveram “I had loved” amavi “I (have) loved” amavero “I shall have loved” PASSIVE COMPLETE amatus eram “I had been loved” amatus sum “I have been/was loved” amatus ero “I shall have been loved” ACTIVE INCOMPLETE amabam “I was loving” amo “I love” amabo “I shall love” PASSIVE INCOMPLETE amabar “I was being loved” amor “I am loved” amabor “I shall be loved” Varro recognizes that many verbs have the binary opposition (contraria) active~passive as in amo, amor “I love, I am loved”. Some ‘intransitive’, namely deponent, verbs have passive form but active meaning, e.g. loquor “I speak”, venor “I hunt”. These exemplify irregularity in the system because they have full participial forms, e.g. loquens “speaking”, locuturus “about to speak”, locutus “having spoken”. But the other intransitives such as curro “I run” have no passive (VIII: 59). 

Varro’s approach to linguistic analysis was the most similar to a modern linguist’s of any of the ancients in the Western Classical Tradition. But although he had some slight influence on Priscian, his methods were not generally adopted by practitioners within the tradition, and his work did not have the effect that it merits. Varro moved beyond the Stoics, though he was clearly influenced by them. Judging from the text still extant, DLL was a wide-ranging analysis of language for its own sake, whereas the Greek Stoics analysed language as a tool for dialectic, or rhetoric, or poetry. Varro was as close as the Romans got to a theoretical linguist. His principal insights arise from his work on lexical networks, in Table 4.4. A restoration of Varro’s case~gender matrix (see DLL X: 22, 44). NOMINATIVE GENITIVE DATIVE ACCUSATIVE VOCATIVE ABLATIVE M.SG. bonus boni bono bonum bone bono F.SG. bona bonae bonae bonam bona bon? N.SG. bonum boni bono bonum bonum bono M.PL. boni bonorum bonis bonos boni bonis F.PL. bonae bonarum bonis bonas bonae bonis N.PL. bona bonorum bonis bona bona bonis The Stoics and Varro 77 which he sought to determine how a language such as Latin can generate a boundless vocabulary from finite elements (Langendoen 1966). He divided words into a sterile class and a productive class from which new words are generated by affixation (declinatio). The latter divides into declinatio voluntaria, which is irregular and has to be learned by rote, and declinatio naturalis, which is a regular system applicable to much of the vocabulary in a language. 

Declinatio naturalis further identifies lexical categories that are [–case, +tense], [+case, –tense], [+case, +tense], [–case, –tense]. Varro’s use of distinctive feature analysis was rediscovered and further developed in the twentieth century (see Chapters 9 and 11). He gave a rational account of the fact that languages contain both regularities and irregularities in morphology (and, for all we know, syntax too). The learnability of language is explained by the application of finite systematic processes to a finite set of primitives. At the same time, declinatio voluntaria allows for new growth provided there is a referent for the new word and the word is susceptible to declinatio naturalis; this constrains an individual’s will to ‘impose’ (i.e. create) names, and enables a new word to be interpretable by other people. Varro sought to explain both regularities and irregularities by taking a commonsense approach to the functions and use of language, given the nature of human perceptions of the world. Today this approach is echoed in the work of cognitive and functionalist grammarians. The abstract nature of his theory of declinatio is demonstrated by the discussion of proportional relations between numbers in X: 43–46, which is explicitly analogical with morphological relations (X: 44, 47). For instance, he compares what he calls ‘the disjoined proportion’ of 1 is to 2 as 10 is to 20 with noun paradigms, e.g. rex “king [NOM.SG.]” is to regi “king [DAT.SG]” as lex “law [NOM.SG] is to legi “law [DAT.SG] The ‘conjoined proportion’ of numbers, e.g. 1 is to 2 as 2 is to 4 [in which 2 is a common element] is analogical with the conjoined proportion of verb paradigms, e.g. legebam “I was reading” is to lego “I am reading” as lego is to legam “I shall read” Here he has shown how a system may operate within both mathematics and language. Varro had a very commonsensical approach to notions of variation from ‘correct’ language use. 

As to the man who uses as ablatives monti [“hill”] and fonti [“spring”] where others say monte and fonte, along with other words which are used in the two forms,15 one form is correct and the other is wrong; yet the person who errs is not destroying the regularities [analogias], although the one who speaks correctly is strengthening it. He who errs in using these words (where there are two forms used) is not destroying the logic in the system when he uses the wrong form; and even with words that are not pronounced in two ways, someone who decides to pronounce them in an unusual manner, does not falsify knowledge of the language but merely exposes his own lack of knowledge. (IX: 112) 15. See VIII: 66 for additional examples. 78 The Western Classical Tradition in Linguistics It is a great pity that Varro’s cool theoretical approach to language was not adopted and developed during the next two millennia. 

The development of a linguistic theory As was said at the outset of Chapter 2, for the ancient Greek philosophers the study of language was a means to several ends, including the development and understanding of dialectic (sound reasoning), rhetoric (persuasive use of language), and poetry. 

With the Stoics, there were significant advances in the analysis of propositional types, propositional structure, the forms of propositions and combinations of propositions, and valid inferences that may be drawn from propositions (single or combined). This led to the development of propositional logic on the one hand and grammatical analysis on the other. Hell?nismos and the studies of rhetoric and of poetry benefited from, and therefore strengthened the significance of, linguistic analyses. The Stoics considered language to be an abstract entity with mental and physical correlates, but they concentrated on the logical system of language as an abstract entity. They greatly expanded grammatical analysis and contributed significant innovations. Thinking about how negation fits into propositional structure led the Stoics to investigate negative scope to a more sophisticated conclusion than Aristotle. They explicated the deduction that a predicate which applies to a class of objects applies also to any member of the class. They postulated rules of inference, and came to a notion of quantifier scope. The manner in which propositions are connected to one another can affect the inferences to be drawn, so a class of connectives needs to be identified. At the same time there is a need to identify the language forms that represent these notions and to relate them to onoma and rh?ma; consequently, the parts of speech and their subcategories came to be identified one by one, as the need arose. For the Stoics, logos was supplemented with the lekton – a clause classified into an illocutionary type. Of lekta the most frequently discussed is the axi?ma “judgment, statement”. And although the predicate (or verb) was accorded the same features as before, nominals were divided into proper nouns and common nouns (which included adjectives). Aristotle’s sundesmos was reinterpreted as “sentence connective”. The Stoic ‘article’ includes prepositions but not conjunctions. And the Stoics separated out the adverb. 

They also recognized a number of secondary grammatical categories: case, gender, number, mood, tense, aspect, voice, and transitivity – though Aristotle had earlier made some remarks upon gender and also number. In semantics they identified illocutionary type (or something very like it), and they clearly distinguished between what we would now call intension and extension.16 In phonology (probably resulting from studies of poetry) they distinguished vowels from consonants, and distinguished some classes of consonants. Although little is known for sure of Stoic syntax, they seem to have developed rules such as the following (Egli 1987). A lekton consists of a subject and a predicate. A subject may be 16. The extension of a language expression designates something that exists in a particular world, usually the world being spoken of. The Stoics and Varro 79 definite, indefinite, or neither. It may be singular or plural. A predicate may be one-place or two-place; if two-place it may be active or passive. A predicate and the lekton that contains it may be in present, past or future tense; and it may be negated. Lekta may also be conjoined, disjoined, or in a conditional relation. Perhaps these properties were once to be found in Varro’s syntax; we don’t know. Of the scholars we have considered so far, only Varro can truly be said to have had a theory of language. Steeped in Stoic ideas, Varro was more of a linguist than a philosopher. His conception of language as an abstract object was based on the empirical investigation of language in which he recognized recursive processes for generating new terms out of old ones – a notion that was Stoic in motivation if not origin. His naturalist etymologies can be seen in this light: as a search for the sources of word meanings, the relationships between words, and the connection between meaning and form. He was further motivated by an interest in the (pre)history of his language. In the extant text of De Lingua Latina Varro investigated morphological affixation, recognized the generative properties of a regular system of affixation as making the unbounded potential of language learnable, and he developed formal criteria for grammatical categorization in a thoroughly modern manner without losing sight of semantic criteria. Varro identified the major clause constituents by their inflectional characteristics in terms of the features [±case], [±tense]. He also developed feature-based paradigms of nominal and verbal suffixes. It is a great pity (a) that Varro’s analysis of syntax has been lost, and (b) that his work did not become a direct foundation for language scholarship in the centuries that followed.