The birth of the adjective class as a problem of translation

Abstract This paper seeks to analyze the translation of grammatical terminology. One of the main differences between the Greek-Latin parts of speech theory and that of traditional later European linguistics (from Port Royal onwards) lies in the existence of the adjective as an independent word class. The paper examines the definitions of the categories of noun, verb and epithet/adjective from Dionysius Thrax through the 17th century, with the aim of showing that the birth of the adjective as an independent word class, along with the stabilization of the labels nomen substantivum and adjectivum with reference to the common noun and the adjective, hides a problem in meta-semiotic translation. Specifically, the issue concerns the translationreinterpretation of Aristotle’s metaphysics in light of Neo-platonic ontology during the Middle Ages, as well as its influence on the reinterpretation-translation of the Greek-Latin parts of speech theory between Late Antiquity and the Renaissance.


Introduction
In a seminal article, Hockett (1954) divides the history of (descriptive) linguistics into two main phases. The first, termed "traditional linguistics", spans the period from the Middle Ages to the era of American structuralism; while the second, beginning with Sapir and Bloomfield, can be termed "modern linguistics". Interestingly, Hockett identified traditional linguistics as originating from the Middle Ages rather than from Greek or Latin grammar. He does not overtly explain the reason why this specific starting point was selected, but we can imagine the motive for his choice.
Although traditional linguistics is largely a continuation of the ideas sketched in Greek and Latin grammar, its parts of speech theory differs from the Greek-Latin theory (Alfieri 2006;2014). While the grammars written between the 15 th and the 18 th Centuries display a tripartite division among the noun meaning the substances, the verb meaning the actions and the adjective meaning the qualities (Arnauld and Lancelot 1660: 271), the Greek-Latin theory of grammar is essentially based on only two parts of speech: the noun and the verb. Secondly, it displays a second-level division between two phrasal functions of the noun: 1) the noun used as head; 2) the noun used as a modifier. But in no way does the Greek-Latin parts of speech theory include an autonomous part of speech for the adjective. The birth of the adjective class out of the epithet class thus represents an important step of the history of so-called "traditional linguistics", but published works on the history of the parts of speech are vague in establishing when and under which conditions the adjective started to be considered an independent part of speech.1 This paper aims to fill this gap by showing that establishing the adjective as an independent part of speech between late Antiquity and the Middle Ages hides a complex problem of non-linear meta-semiotic translation (see the Introduction to the topical issue).
On a first glance, it seems that the translation of the Greek-Latin terms referring to the parts of speech is simply a question of translating a few Greek or Latin words into English (or any other language): Gk. ὄνομα ≌ Lat. nomen ≌ Engl. noun; Gk. ῥῆμα ≌ Lat. verbum ≌ Engl. verb; Gk. ἐπίθετον ≌ Lat. appositio or adiectivum ≌ Engl. epithet, adjective or apposition. However, translating grammatical terms is a complex task since each time a grammatical term is used, the whole theory that brought that term into being is activated. Conversely, each time a change occurs in the theory, the meaning of each term making up the theory is altered by the change, although the term remains the same.2 This fact does not affect the translation of the Latin terms nomen and verbum dramatically, but it is crucial for establishing the exact English equivalents of Gr. ἐπίθετον and Lat. adiectivum.
Modern scholars often translate Gr. ἐπίθετον as Engl. 'epithet', Lat. adiectivum as Engl. 'adjective' and Lat. appositio as 'apposition', as if the difference between Gr. ἐπίθετον and Lat. adiectivum were the same as that between the modern notions of epithet and adjective. However, up to the Middle Ages, Lat. adiectivum and appositio were mere synonyms meaning 'epithet', no different from Gr. ἐπίθετον, while after the Modistae, Lat. adiectivum meant only 'adjective' and Lat. appositio meant 'epithet'. Despite some confusion, experts in ancient grammar are aware that the category termed as ἐπίθετον in Greek and adiectivum in Latin does not coincide with the category termed as 'adjective' in English exactly. However, manuals often translate Gr. ἐπίθετον through 'epithet', Lat. adiectivum as 'adjective' and appositio as 'apposition' almost automatically, incorrectly generating the idea that the switch from Greek to Latin in the language of grammar also entails the differentiation between the categories of epithet (or apposition) and adjective. This paper, therefore, aims to show that the formal similarity between Lat. adiectivum and Engl. adjective is misleading in translations, because Lat. adiectivum and Lat. appositio mean 'adjective' or 'epithet/apposition', depending both on the age of the text, as well as on the specific parts of speech theory discussed within it.
Aristotle elaborates upon the difference between substance and accidents as a part of a general theory of the being and adapts the substance-accident contrast to be used as a tool for language analysis.4 He adds σύνδεσμοs 'conjunction' as the third part of speech and refines Plato's definition of noun and verb using a twofold, half-linguistic and half-philosophic definition (Poet. 56.b.20,De int. 16.a.19). On the one hand, a purely linguistic clue is used: the verb is time-bound (it consignificat tempus, as Boethius said), while the noun is not. On the other hand, a philosophical feature is provided: the noun is the sign of what exists as itself and is the subject/substance of the discourse (Gk. ὑποκείμενον 'underlying', which Boethius translated with a calque as subiectum 'thrown under'). The verb is neither a subject nor a substance, but an accident, as it predicates something on behalf of a different entity, i.e. the noun/subject.
In addition to the noun and the verb, Aristotle also mentioned the ἐπίθετον. However, he does not include the ἐπίθετον in the traditional list of the parts of speech in Poet. 1456b20 and employs this term only in the Rhetoric, when the poetic style is analysed (1405a10; 1405b21-23; 1406a10-12, 19-24, 30; 1407b31; 1408b11). Therefore, in Aristotle, the term ἐπίθετον does not have the exact meaning of 'epithet', which became commonplace later, rather it is conceived as any kind of ornamental periphrasis that can be added to another element of the λόγοs irrespective of its being a noun (ὁ πατρὸς ἀμύντωρ "the killer of the father", 1405b21), an adjective (τὸν ὑγρὸν ἱδρῶτα "wet sweet" 1406a19) or a multi-word genitival expression (τῇ τῆς ψυχῆς ὁρμῇ "to the rush of the soul" 1406a19).
The verb is a word without case displaying tense, person and number, and presenting an activity or a state.
2) ὄνομά ἐστι μέρος τοῦ λόγου πτωτικόν, σῶμα ἢ πρᾶγμα σημαῖνον, σῶμα μὲν οἷον λίθος, πρᾶγμα δὲ οἷον παιδεία. The noun is a part of speech with case meaning a concrete thing or an incorporeal item (e.g. concrete such as λίθος 'stone' or incorporeal such as παιδεία 'education').6 The noun and the verb are identified by inflectional clues, while semantics provides the necessary exemplification. The verb has tense, person and number reference, but no case, and means a state or an activity. The noun has case, gender and number, but no tense, and means a thing-like notion such as 'stone' or 'education'. Although the word classes are defined inflectionally, philosophy is not totally disregarded. The noun is divided into three types: 1) ὄνομα κύριον 'proper noun', which refers to the individual substance (τὸ τὴν ἰδίαν οὐσίαν σημαῖνον); 2) ὄνομα κοινόν or προσηγορικόν 'common noun', which refers to the common substance (τὸ τὴν κοινὴν οὐσίαν σημαῖνον); and 3) the epithet as the third εἶδοs 'type' (Tekh. 12): 3) ἐπίθετον δέ ἐστι τὸ ἐπὶ κυρίων ἢ προσηγορικῶν ὁμωνύμωs τιθέμενον καὶ δηλοῦν ἔπαινον ἢ ψόγον. The epithet is added to both the proper or the common noun in the same way to mean laud or blame.
Agreement and comparison are the inflectional clues that define the modern class of the adjective, at least in the inflectional IE languages of the Latin type, but Dionysus does not mention them in the definition of the ἐπίθετον. Although most of the epithets quoted by him as examples are effectively adjectives (e.g. σώφρων 'wise', ἀκόλαστοs 'excessive', ταχύs 'fast', etc.), to Dionysius a noun modifying a noun (Σωκράτηs ὁ γραμματικόs "Socrates the grammarian") and a proper adjective (Σωκράτηs ὁ ἀγαθόs "the good Socrates") are one and the same. In other words, Dionysius does not identify the adjective class, but states that when the noun is the substrate of the discourse, it is a true noun (and can be separated into proper and common nouns), but when it is added to another noun in order to express laud or blame, an epithet is found. The epithet therefore is not an independent part of speech, but rather a special type of or a special use of the noun that can be defined on a combination of philosophy (not a substance), semantics (laud or blame), and 5 The dating and the authenticity of Dionysius Thrax's Τέχνη are highly debated. According to some scholars, the whole text, or at least some parts of it, originate from the 3 rd -4 th century CE. For further discussion on this topic, see Di Benedetto (1958-19591973;1999;2000), Ax (1986: 223 ff.), Robins (1995) and Lallot (1998). 6 Boethius translated πρᾶγμα with Lat. res, but res incorporalis 'abstract thing, incorporeal item' seems to be more adequate after Belardi (1990). Dionysius' teacher Aristarchus presents a similar concept regarding the noun and the epithet, but he exemplifies the word classes rather than defining them, see Matthaios (1999:  syntax (added to the proper or common noun).7 The influence of Aristotle's theory regarding the definition of the parts of speech grew in the first centuries AD. Grammarians made extensive use of the substance/accidents pair to produce a large number of binary oppositions to be used on different layers of language analysis, just as Apollonius Dyscolus' Περὶ συντάξεωs shows (2 nd century AD). The work opens with a short definition of the parts of speech, which mirrors that found in Dionysius (De constr. I.13.1). However, Chοeroboscus (GG II/3, 38.20) attributes an interesting philosophical definition of the noun (GG II/2, 524.8) to either Apollonius or his son Herodianus: 4) ὄνομά ἐστι μέρος τοῦ λόγου πτωτικόν, ἑκάστῳ τῶν ὑποκειμένων σωμάτων ἢ πραγμάτων κοινὴν ἢ ἰδίαν ποιότητα ἀπονέμων.
The noun is a part of speech with case, which attributes an individual or common quality to any concrete or abstract substance.
The term ὑποκείμενον 'substrate noun' belongs to the Aristotelian branch of thought, and it means the subject-substrate of the λόγοs 'discourse' (De constr. I.31.1). However, to Apollonius, the noun attributes a quality to the subject/substrate of the discourse, but it also refers to the specific substance of the subject/ substrate of the discourse (διὰ τῆς ὀνομαστικῆς συντάξεως τὴν οὐσίαν ἐπιζετοῦμεν τοῦ ὑποκειμένου "the substance of the subject is meant through the appellative construction", De constr. I.120.1, cf. also I.101.1).
In other words, to Apollonius the noun ἄνθρωποs refers to the individual substance of the subject/substrate of the discourse such as "man" in the sentence "the man is mortal". However, if the substance "man" is named, automatically the qualities that are naturally connected with that subject/substrate are also named such as "the rational being, having two feet and lacking plumage". From this perspective, the noun refers an individual or a common quality (ποιότηs) to the substrate of the discourse, but it also refers to the specific substance of the referent.8 Conversely, the epithets such as ἀγαθός 'good', δοῦλος 'slave' and γράψας 'writing, who writes' can be added to the noun ἄνθρωποs in a noun phrase for referring to the accidental qualities of the noun heading the phrase, those qualities which can be present or absent without modifying the substance of the noun (epithets are, therefore, also called ἐπιτρὲχοντα πτωτικά "accidental nouns" in De constr. I.135.1).9 In brief, the Aristotelian theory supplied the first philosophical tools for defining the noun and the verb. The definitions derived from the idea that the λόγοs was a σύνολοs 'ensemble' of substance and accidents, no different from any other being. In this view, the noun is the substantial part of the discourse, its (onto)logical substrate and its syntactical subject, and therefore, is called ὑποκείμενον 'substantive'. The verb is the accidental part of the λόγοs, the part that can be taken away without compromising its meaningfulness.10 Grammarians, however, slightly modified the Aristotelian distinction between noun and verb. The noun, which was traditionally referred to as the substrate, the substance and subject of the discourse, begins to also refer to the individual substance of the referent. In turn, the epithet, which had not been identified by the philosophers, was defined as a special use or a special type of noun and was used to add an accidental quality to the noun, but it was not at all considered an independent part of speech.
All these data are well known to Greek philologists, who customarily translate Gr. ἐπίθετον as Engl. 'epithet, apposition'.11 However, especially when the category termed as ἐπίθετον is exemplified by true adjectives, scholars tend to translate it as 'epithet' or 'adjective' indifferently, as if the two terms were interchangeable (e.g. Dickey 2007: 127, 237). This choice may be understood in practice since adjectives are 7 A very similar concept of epithet is found in Dionysius' teacher Aristarchus, see Matthaios (1999: 219, 233-42). 8 This and similar passages were commented on by Priscian (Inst. Gram. XVII.35.15) and Servius (GL IV, 489.21) and represent the predicatio de subjecto termed by Boethius (Arens 1984: 180). A further confirmation of this conceptual framework comes from the definition of the pronoun, which in Apollonius' view, refers to the substance of the substrate noun (τὴν οὐσίαν τοῦ ὑποκειμένου) without addressing any of the specific structural qualities (De constr. I.101.1; see also II.22.1). 9 The link between epithets and accidents is found also in De constr. II.2.143.3 and in the Hamburg papyrus edited by Shenkenveld (1993: II, 128). The label πτωτικόν literally means 'having case', from πτῶσις 'case' (lit. 'the falling' from πίπτω 'to fall'). 10 According to Aristotle any finite verb can be substituted with the participle plus the verb 'to be' (Metaph. 1017a28, An.pr. 51b12). 11 See, for instance, Ax (1991: 285), Matthaios (1999: 233, 240-2 and n. 166, 169) and Lallot (1992). among the most typical instances of ἐπίθετα, but it is not perfectly adequate from a linguistic point of view, as it may incorrectly result in the idea that the Greek grammarians knew of an adjective class parallel to the adjective class defined in modern linguistics.

The definition of the major word classes in Latin grammar
Latin grammarians adopted the Greek theory of language without making significant changes to the definitions of the major word classes. Only those features of Greek grammar that were clearly in contrast with the Latin data were subject to revision but, for the most part, Latin grammarians were satisfied with only making a vulgate of the Greek originals. Their works were completed with a method of literary translation, which according to Robins (1957: 62), is no different from that which the Latin playwright Terentius described in the prologue of the Andria (vv. 8-21) and labeled as contaminatio 'contamination'. Similarly to the playwrights, the Latin grammarians merged partly original notes with semi-literal translations of selected passages of different works belonging to the Greek as well as to the former Latin tradition. The originality of this type of works, therefore, relied in the constant dialogue with the former tradition, rather than in the pretence of an absolute novelty.
Varro (1 st century BC) took up Aristotles' definition of the noun and the verb, but he did not use the label that later became common for the noun, which was called only vocabulum 'word, noun' (De ling. lat. 8, 11), and he did not mention the epithet at all in his description of the parts of speech.
Quintilian (1 st century AD) introduced the Greek concept of epithet in Rome, calquing the Greek term ἐπίθετον through Lat. appositum, although some preferred to call it sequens, calquing the philosophical term ἐπιτρέχον (from ἐπιτρέχω 'to run after, follow'). However, Quintilian did not mention the appositum in his Ars grammatica and considered it only a stylistic ornament of the discourse in the Institutio oratoria, as Aristotle also said (Inst. or. VIII.6.4): 5) cetera iam non significandi gratia sed ad ornandam et augendam orationem assumuntur. Ornat enim epitheton, quod recte dicimus appositum, a nonnullis sequens dicitur.
The remaining [tropes] are not employed for semantic reasons but solely to adorn or enhance the discourse. For the epithet, which we rightly term as appositum, although some call it sequens, is an ornament.
The nouns that take their meaning from the connected noun, such as magnus 'big', fortis 'strong', do not refer to any concept in themsleves, therefore they are termed as adiectiones 'addictions' by some, such as magnus vir 'big man', fortis exercitus 'strong army'. These nouns can be compared.12 Since Charisius mentions comparison as one of the specific features of the adiectiones, one might suppose that he is really defining something very close to the modern adjective class, not that of the apposition. However, the nouns that can be compared represent only a specific semantic class of the noun class, not to mention that some of the nouns belonging to this specific semantic sub-class cannot be compared, as is the case with mediocris 'poor', sobrius 'sober', rudis 'rude' (GL I.156.15 ff. In these passages, Apollonius' philosophical background is almost completely retained. The noun is the sign of a substance-plus-quality (substantiae qualitatem) and is divided into nomina adpellativa, quae substantiam significant "which signify a substance", and nomina adpositiva or adiectiva, which are added to the nomina adpellativa in order to express their accidental quality, which can be present or absent without change in the noun-substance (Inst. gram. II.5.24). However, with respect to the Greek originals, the inflectional features distinguishing the nomina adiectiva and the nomina adpellativa are discussed more extensively (Inst. gram. III.1.2, II.28.1): nomina adpellativa, quae comparari non possunt "appellative nouns that cannot be compared".
As was the case with Charisius, the use of comparison in Priscian's definition of the adjectivum, which is not found in the Greek sources, might lead us to think that the category that Priscian terms as adjectivum really is the same category that modern linguists call 'adjective'. Nevertheless, for Priscian the philosophic difference between the nouns that are used as subjects of the discourse and refer to a substance-plusquality and the nouns that are used as attributes and refer to the accidental qualities of the other nouns is more important for defining the adjectivum than the difference between nouns that can be compared and nouns that cannot. When the examples of adjectiva are discussed, in fact, true adjectives (e.g. iustus 'right', magnus 'big' and niger 'black') are quoted side by side with participles (e.g. sapiens 'wise') and nouns (e.g. fur 'thief' and grammaticus 'grammarian'), as well as with non-gradable adjectives (e.g. medius 'middle', sinister 'left, sinister', primus 'first'), with non-gender inflected adjectives (e.g. capax 'able') and with gender alternating nouns (e.g. grus 'crane') (Inst. gram. III.1.2, II.83.17).14 The examples brought in by Priscian thus demonstrate that the category described under the label of adiectivum is closer to the category traditionally termed as ἐπίθετον in Greek grammar, although this category has already begun to slowly evolve toward the modern adjective class, and therefore, a provisional translation as 'epithet-adjective' was selected in the glosses.
In sum, the work of Latin grammarians can be compared to a literary translation of the Greek originals, a translation which could be improved by adapting the originals to the peculiarities of the Latin language (e.g. the article is disregarded) or by specifying some particular aspects of the originals (e.g. the interjection is added, the relevance of comparison is increased by defining the nomen adjectivum), but was primarily aligned with its sources. Latin grammarians followed the Greeks in identifying only two major parts of speech (the noun and the verb), plus a second-layer distinction, such as the attributive use of the noun (the epithet-adjective). Although the use of comparison in the definition of the adjectivum might suggest that the category described by Priscian coincide with the category termed as adjective in modern linguistics, the grammatical category referred to by Priscian is closer to the category termed as epithet or apposition in modern linguistics. Still, even if scholars of Latin grammar know perfectly well that Lat. adjectivum does not mean exactly the same as Engl. adjective, they typically translate this term through 'adjective' (e.g. Luhtala 2005: 49), as if the similarity of the labels Lat. adjectivum and Engl. adjective granted their equivalence in translations. However, this choice, understandable as it may be in practice, is anachronistic and potentially misleading since, according to Priscian, the major parts of speech were only two (the noun and the verb), and instead the term adjective currently refers to a third major part of speech that differs from both the noun and the verb. Bursill-Hall (1971: 20, 114) and Arens (1984: 489) showed that, just as the Latin scholars did, the Middle Ages philosophers and grammarians used a method of literary translation plus comment, in which the ideas of the original source, those of some later comments and translations, as well as those written by the author of the comment/translation are so deeply fused, that the notion of contaminatio could again be of use.

The word class system theory in the Middle Ages
Apparently, the grammarians of the Middle Ages followed the Greek-Latin tradition almost slavishly: glosses and comments on Priscian and Donate were compiled and the doctrines of the Latin grammarians were copied with few changes. However, the presence of innovation beyond the apparent preservation of the original doctrines of their predecessors was a typical characteristic of medieval thought (the so-called principium auctoritatis), and the science of grammar was no exception. Although the number and type of the parts of speech seem to have been canonized once and for all in Priscian, the definitions underlying these categories have progressively changed. The first reason for this change, however, is not found in the grammatical doctrines themselves, but in the in the new philosophical framework underlying the theory of grammar.

The reinterpretation of Aristotle's metaphysic
The scarce knowledge of the Greek version of Aristotle's work fostered the reinterpretation of his theory in the light of the Neoplatonic ontology. In the original Aristotle, the categories of the being differed from the categories of the languages, and both types of categories differed from the categories of the thought (Belardi 1975: 38 ff., 79 ff., 144 ff.). In the Middle Ages, by contrast, an almost perfect isomorphism among categories of the language, categories of the thought and categories of the being was defended. However, the new isomorphism was supposed to represent a continuation of Aristotle's original theory, not reinterpret it.15 In the 3 rd century, Plotinus stated that any vocal discourse was a copy of a more trustworthy discourse enacted in the immortal soul of everyman (En. I.2.3; V.1.3).16 As a result, he was followed by Ammonius (5 th century), who was the first commentator on Aristotle and the best-known pupil of the Neoplatonic Proclus, who, in turn, was a Plotinus scholar. In their view, every vocal emission mirrored a mental utterance, which was the emanation of a simple and universal concept existing in mente dei "in God's mind". As nouns and things were perfectly joined in the eternal mind of God, they should also be so in the mental discourse enacted in the eternal soul of every man. From this standpoint, the existence of a certain noun in the language entailed the existence of the corresponding substance on the ontological level, as Ammonius overtly said (Busse 1897: 57).17 Boethius, who followed the lessons of Ammonius in Alexandria, added that the ability of the words to attach meaning depended on the substances of the metaphysical world, which literally placed their significance on the words. Therefore, the noun was indeed the sign of a substance, as Aristotle said, but it was because the individual substances of the metaphysical world were literally reflected in the nouns meanings, not because they referred to the subject/substrate of the discourse and to the most important parts of the sentence (Arens 1984: 162 ff., 189). Simplicius, who followed the lessons of Ammonius along with Boethius, clarified the effect of this approach, and précised that, if the noun was the sign of the individual substances, the verb should be the sign of the accidents, as Aristotle had said, since every being, including the discourse, was made of an ensemble of substance plus accidents (Pattin 1971: 66).
To summarize, the scarce knowledge in the original Aristotle led to the reinterpretation of his philosophy in light of Neoplatonic ontology. The Aristotelian categories, which were the predicable that could be attributed to an element in the discourse, were gradually conflated with the categories of the being, the thought and the language, as if language, logic and ontology were only one thing. As a result, the first Aristotelian category, the substance, which had originally been a generic and abstract category, was gradually transformed into a specific and individual substance, not entirely different from a Platonic idea. The parts of speech theory was the domain in which the reinterpretation of Aristotle's philosophy triggered the most relevant outcomes. In fact, if language, logic and ontology had coincided, the same universal notions would be expressed in all languages, and all languages should show the same grammatical categories. These categories were empirically identified with the traditional parts of speech defined by Priscian, but their number and their definitions differed from those found in the Latin originals, and instead comply with the new philosophical framework.18

The parts of speech in the early Middle Ages
In the early Middle Ages, the two major parts of speech, the noun and the verb, were redefined in accordance with the reinterpretation of Aristotle's theory, leading to the addition of the adjective as a third major word class. However, just as the philosophers believed that they were continuing Aristotle's original theory rather than innovating it, the medieval grammarians assumed that they were continuing the Greek-Latin parts of speech theory, not restructuring it.
The first trigger of the renewal of the parts of speech theory was attested in Greek and Latin grammar, when Dionysius, Apollonius and Priscian defined the noun as the sign of individual substances. This definition of the noun, which was not originally that of Aristotle and was rarely used among the classical grammarians, was perfectly consistent with the new substantialist ontology of the Middle Ages and its frequency increased in the scholia. See Choeroboscus' commentary (8 th -9 th century), GG IV/1, 105.219 10) τὸ δὲ ὄνομα προτερεύει τοῦ ῥήματος, ἐπειδὴ τὸ μὲν ὄνομα οὐσίας ἐστί σημαντικός, τὸ δὲ ῥῆμα συμβεβηκότος, αἱ οὐσίαι προτερεύει τῶν συμβεβηκότων. The noun dominates the verb, since the noun signifies the substance, while the verb the accident, [and] the substances dominate the accidents.
The second step in the renewal of the theory was the transformation of the subclass of the nouns used as "epithets" into a different word class, which corresponds to the modern adjective class. This change also involved two stages. The first consisted of the gradual shift of the substance/accident pair from the definition of the noun as opposed to the verb-as it was in the original Aristotle-to the definition of the noun as opposed to the adjective, as in the Scholia Marciana (GG I/3, 386.27): Once the substance/accident contrast distinguished the noun and the adjective, a second change was verified, but this change did not affect the philosophical definition of the word classes, but rather the linguistic clues identifying the epithet. In an anonymous gloss to Donatus' Ars Maior (9 th century), the nomina substantialia are distinguished from the nomina accidentalia or adjectiva (Thurot 1868: 80), since the latter do not refer to any substance, but instead to the accidents that can be augmented or diminished without changes to the substance. The presence of a linguistic clue distinguishing these classes was supposed to mirror their ontological difference: since the nomina accidentalia do not refer to individual substances ontologically, they are not fixed in a single gender, but waver from one gender to another depending on the gender of the noun/substance to which they refer (therefore they are also termed nomina mobilia).
The gloss exhibits a fundamental step in the birth of the adjective class. Concord (in adjunct of comparison) is fundamental for telling apart the adjective and the epithet, since only the adjective agrees (and is graded), while non-adjectival nouns used as epithets do not. Priscian, in fact, mentioned comparison in his definition of the epithet, but he did not use concord at all. Moreover, he used comparison as a typical, non-definitional feature since many of the epithets that he quoted (e.g. grammaticus, fur and rex) can neither be graded nor agree, and some of the nouns that he quoted do agree (e.g. amicus). Therefore, the gloss represents neither the first use of the label adjectivum, which can be traced to Charisius (GL I, 156.15), nor the first use of the term adjectivum in relation with concord, which is found in Abelard (see below), but represents the first case in which the use of concord ensures that the class described is the modern adjective class, although this class is labelled as nomen accidentale.
Thanks to Abelard (11 th -12 th century), who first used the label adjectivum with reference to the modern adjective class, we can be certain that the new adjective class, not the epithet, was described in the scholia. Although, in some passages of his work, Abelard says that the major parts of speech are two, the noun and the verb (Arens 1984: 230), in the glosses to Aristotle, he reports the existence of a polemic issue between grammarians and philosophers on the correct definition of the adjective. The polemic issue shows that under the traditional label of nomen adjectivum, a new word class was born ( The substantive is defined differently by grammarians and us philosophers, and it is the same for the adjective. Grammarians call substantives any fixed noun [i.e. a noun without agreement], also those nouns that are derived from the accidental properties; and they call adjectives those items that are adjoined to other nouns, the substantives, such as homo albus 'white man', animal rationale 'rational animal', but we would rather say that rationalis 'rational' is rather substantial than accidental. In fact, we define as accidentalia 'accidentals, adjectives' all nouns taken from an accidental property, although they are substantives [grammatically], such as vir and mulier. Grammarians disregard the ontological nature of the words without even realizing it, Abelard says. They define the parts of speech depending on the position in the sentence (Geyer 1919: 475), rather than depending on the meaning, when such a linguistic definition does not explain anything about the substantial or accidental nature of the notions to which these words refer. Ontologically, if the quality of "being rational" is predicated of the entity "man", this predication is substantial since the quality of "being rational" is naturally included in the definition of "man" as "rational animal". By contrast, if we say rex or grammaticus, we are not referring to a pure substance, such as royalty or being a grammarian, but rather we are referring to the same substance "man" quoted above, although we further determine it through an accidental quality such as that of being a king or a grammarian. Therefore, grammarians distinguish nouns and adjectives based on their different linguistic properties, but this distinction should be rejected since it does not coincide with the ontological difference between substances and accidents, as grammarians typically believe.20 If a polemic issue on the substantial and the accidental nature of the adjective is found in Abelard (11 th -12 th ), a polemic issue on the importance of the linguistic clues of concord and comparison in defining the adjective is found in the glosses on Donatus (9 th ), one may reasonably infer that the classical category of the epithet evolved into the modern adjective class somewhere between the 9 th and the 12 th century.21 The change passed almost unnoticed by medieval grammarians since the labels identifying the adjective were retained (nomen mobile, appositivum, adiectivum, accidentale), but despite the apparent identity of these labels, a new parts of speech theory was going to arise.

The Speculative Grammar
Despite Abelard's critics, the isomorphism between language and ontology was the dominant view of the Middle Ages. In the Speculative Grammar, which was the most important school active between the 12 th and the 14 th centuries, "the only method of research was to derive and justify rules of grammar from system of logic and metaphysical theories on the nature of reality" (Robins 1957: 75).22 In this framework, the nature of the things determined the nature of the universal concepts in mente Dei, and the nature of the universal concepts in mente Dei determined the nature of the word classes in the human languages. Moreover, given the perfect isomorphism between language and ontology, if the noun was the sign of the substances, the words meaning qualities should have also been the sign of a specific ontological class, that of the accidents. See, Sigier de Courtrai (Sum. mod. sign., p. 188, 97):23 13) nomen significat rem suam per modum substantiae seu entis […] adjectivum est modum significandi accidentalis nominis designans circa rem modum essendi adiacentis.
The noun means its referent through the modality of the substance or of the entity; the adjective is the modality of signifying of the accidental noun, which designates the referent through the modality of what is added.
The modi significandi represent the mediation between the universal and perfect concepts in mente Dei and their imperfect expression in the human language. In this view, the noun is characterized by the modality of the substance and the adjective by the modality of the accident. The ontological difference between these two modalities of the being is mirrored in the syntax, as said by Thomas Aquinas (Sum. theol. P.P. Q. 5, a. 5):24 14) haec differentia inter nomina substantiva et adiectiva, quia nomina substantiva ferunt suum suppositum, adiectiva vero non, sed rem significatam ponunt circa substantivum. This is the difference between substantives and adjectives: substantives bear their suppositum, while adjectives do not, but rather they adjoin the signified thing to the substantive.
Lat. suppositum is a complex term, meaning either the logical and syntactic subject of the sentence or its ontological substrate. It is the closest medieval continuant of the Aristotelian notion of ὑποκείμενον. In modern terms, it could be rendered through the notion of head (as opposed to modifier), although this qualification holds on both the linguistic and the ontological level (Bursill-Hall 1971: 57 ff.). In the 21 Thurot (1868: 161) is imprecise when he attributes the invention of the label of adjective to Abelard: Abelard is clearly taking on a traditional label in the text quoted, although he is criticizing the definition of such a traditional label. Also Robins (1997: 107) is not exactly right when he says that Peter Helias was the inventor of the adjective, since Peter Helias was slightly younger than Abelard. 22 The isomorphism between language and ontology could be sustained by two perspectives. Sigier de Courtrai (late 13 th AD) claimed a direct isomorphism (Sum. mod. sign., p. 93), but Peter Helias said that the isomorphism was mediated through the modi significandi (Pinborg 1967: 48). 23 A very similar definition of the noun is found in Thomas Aquinas (Sum. theol. P.P. Q. 93, a. 3), Thomas of Erfurt, Michel de Marbais, Sigier de Brabant, Alexandre de Villedieu (Thurot 1868: 170, 188, 346), in the manuscripts of the Oxford library (Hunt 1980: 170), and William of Conches (Maierù 1972: 80 n.). Obviously, the classical definition of the noun as the sign of a quality could be juxtaposed to the new definition, as in the Notae Dunelmenses (13 th AD, see Hunt 1980: 19 ff.). 24 A very similar definition of the difference between the noun and the adjective is found in Gosvin de Marbais (13 th -14 th , see Thurot 1868: 351), and in the late medieval logicians (Maierù 1972: 191 n., 207).
Middle Ages, therefore, the nouns referred to the single substances of the referents and, therefore, could be the subjects of the sentence, whereas the adjectives copulated the nouns-supposita with their accidental qualities, but they could not be used as supposita.
Finally, once the noun became the sign of the substances and the adjective was the sign of the accidental qualities, the traditional Aristotelian definition of the verb as the sign of the accidents was likely confusing. However, since openly countering Aristotle's theory was not typical of Modistae, the verb was redefined as the sign of the actions and the passions but, since actions and passions were also accidents, the verb was also the sign of the accidents. See Sigier de Courtrai (Sum. mod. sign.,p. 108 In conclusion, the difference between the Latin parts of speech theory and that found during the Middle Ages is greater than once thought. To Priscian, the noun was, by and large, the sign of a quality, the verb the sign of the accidents, and the epithet a noun used as a modifier to mean an accidental quality of the noun. For Modistae, the noun was the sign of the single substances, the adjective was the sign of the accidental qualities and the verb was the sign of the actions (or of the accidents). The change in the theory of the parts of speech was triggered by the reinterpretation of Aristotle's metaphysics into a new philosophical system--a system in which the categories of the being, of the thought and of the languages were isomorphic and coincided with the universal categories of substance, quality and action. The changes in the philosophical definitions of the parts of speech were parallel with the changes in the linguistic clues defining the epithet. Concord, comparison and the inability to head (noun) phrases formed the definitional bundle, which defined the adjective as a new part of speech. This new part of speech was not a special use or a special type of the noun, as it was the epithet, but rather the sign of a totally independent modality of the being, and therefore, of a totally independent word class given the coincidence between language and ontology as claimed during the Middle Ages. The fixation of the adjective class occurred between the gloss to Donatus' Ars Maior (8 th -9 th century) and the Abelard's glosses (12 th century), but the new word class was not included in a complete parts of speech theory until the Speculative Grammar (13 th -14 th century). Therefore, while between the 1 st century BC and the 6 th -7 th century the Latin label adjectivum had a meaning which was closer to that of 'epithet' than to that of 'adjective', from the 8 th -9 th century, the English term adjective was the only possible translation of Lat. adjectivum.

The parts of speech theory in the "traditional linguistics"
As noted by Padley (1976: 15-39), Grammatical studies were no exception to the general Humanist aim of restoring the lost classical past in that they sought to go back beyond medieval conceptions to the grammar of Donatus and Priscian. However, besides the apparent restoration of the past, the two most important innovations of the medieval tradition: 1) the isomorphism between language and thought, and 2) the difference between the substantive and the adjective, were incorporated into the Humanist grammars where they were adopted without mention of their novelty. As in the Middle Ages, during the Renaissance, the parts of speech mirrored the ontological structure of the world. From this perspective, the linguistic difference between nouns and adjectives could not be argued since it replicated the ontological difference between substances and accidents; however, Priscian teaching could not be openly challenged, since it represented the grammatical auctoritas. The contrast between the new ontology of the Middle Ages and the traditional teachings of Priscian created a complex situation in which the tripartite word class system arising from the Middle Ages ontology was kept, but the adjective was seen as a sub-class of the noun, rather than as an independent part of speech.
To Scaliger, the major parts of speech were only two, the noun and the verb (Scaliger 1540: 220, 288), although the nouns were divided into two ontologically different sub-classes: 1) the substantive, which means the substance, and 2) the adjective, which means the accidents (Scaliger 1540: 187). Despite the confusion between language and ontology, the role of agreement was established, seeing that Ramus (1576: 88) defined the substantives as nouns of a single gender and the adjectives as nouns of three genders, and Sanctius defined the adjective as a noun showing gender marking terminations (Sanctius 1587: app. f. i v ). Obviously, if there are only two major parts of speech (the noun and the verb) but the adjective is defined by different terminations, by a different use of concord and comparison, and it is also an ontologically different class with respect to the noun, the contradiction is patent.26 Finally, the Grammaire générale et raisonnée de Port-Royal (Arnauld & Lancelot 1660) made explicit the philosophical assumptions underlying the parts of speech theory of the preceding centuries. They openly claimed that the categories of the Latin language were the same as the categories of the thought, and these categories were universal logic-ontological classes (Arnauld & Lancelot 1660: i). Also in this case, however, the intended word classes were the parts of speech defined in the Latin Middle Ages, not those defined in Greek-Latin grammar: the nouns referring to the substances, the adjectives referring to the qualities and the verbs referring to the actions (Arnauld & Lancelot 1660: 271 ff.).
In sum, the tripartite parts of speech system developed in the Middle Ages was canonized and became the standard doctrine of all of the descriptive grammars of the Renaissance. However, this new word class system was canonized as if it continued the Greek-Latin theory, not as if it resulted out of the modification of the Greek-Latin parts of speech theory through the two most important innovation of the Middle Ages: the birth of the adjective as an independent word class and the isomorphism between language and ontology.

The birth of the adjective class and the problems of translation
The main difference between traditional linguistics and the Greek-Latin theory of grammar consists in the treatment of the adjective class and, more generally, in the parts of speech theory to be accepted in the theory of grammar. In turn, the acceptance of the adjective as a third, major word class hides a complex metasemiotic problem of translation, which concerns the translations, the comments and the re-interpretation of both the work of the Greek and Latin grammarians as well as that of Aristotle. The problem can be summarized as follows.
Plato defines the noun-subject (ὄνομα) and the verb-predicate (ῥῆμα). Aristotle redefines these classes using a linguistic clue (the verb has tense-reference) and adds a philosophical definition for both: the noun is what exists by nature and is the subject/substrate of the proposition (ὑποκείμενον); the verb is predicated of another entity and does not exist by nature, and thus is the accident. In addition to introducing the concept of the epithet (ἐπίθετον), Dionysius and Apollonius rephrase Aristotle's definitions of the noun and the verb by using case marking and tense reference. However, the epithet is not a third, independent part of speech, but rather a special use or form of the noun, which is verified when one noun is added to another in order to express an accidental quality.
The Roman grammarians translate the Greek parts of speech theory into Latin. They render ὄνομα as nomen, ῥῆμα as verbum, ἐπίθετον as appositio or adjectivum and ὑπαρκτικόν as subjectum or substantivum. To Priscian, the noun is the substantial part of the sentence, though it usually refers to a substance-plusquality; on occasion, it may also be the sign of an individual substance, although this definition is rare.
26 The vernacular grammars confirm that the adjective is treated as a single part of speech following the Modistae, although, as Priscian said, it is a subclass of the noun. In Melanchthon's Greek and Latin grammars, the major parts of speech are only two, the noun and the verb (Melanchthon 1533: 21;1558: 6), but the noun is divided into two linguistic-ontological classes: the substantive, which exists for itself and has a single gender, and the adjective, which inheres in a subject and shows terminations of three genders (Melanchthon 1533(Melanchthon : 25, 1558 Hence, in Latin grammar the parts of speech theory was also based on two major classes, the noun and the verb, while the adjectivum was a special use or a special type of the noun and not a single part of speech, although this special type of noun begins to slowly evolve toward the modern adjective class. The classical parts of speech theory undergoes a change between the Late Antiquity and the Middle Ages, and this change is triggered by the reinterpretation of Aristotle's philosophy through the Neoplatonic ontology. While the original Aristotle never claimed the isomorphism between language, logic and ontology, medieval grammarians, despite not reading Greek, not only adopt this idea, but also project this isomorphism onto Aristotle's original theory. In their view, the categories of the language, the thought and the being coincide, and these categories are identified within the major word classes: the noun indicating the substances, the adjective indicating the qualities and the verb indicating the actions. This theory is supposed to continue through to both the parts of speech theory of the Greek-Latin grammarians and Aristotle's philosophy, although Aristotle never claimed the isomorphism between language and ontology, and the Greek-Latin grammarians never considered the adjective an individual part of speech. The new word class theory arises from the glosses in Donatus' grammar (9 th century approx.) and Abelard's work (11 th -12 th century approx.), but its first coherent presentation is found in the Speculative Grammar (13 th -14 th century).
The Renaissance canonizes the new parts of speech theory, although scholars consider it to be a simple continuation of Priscian's teaching as seen through the lens of Aristotle's philosophy. Port Royal's grammar confirms that logic and language coincide in the definitions of the major word classes, the noun/substance, the adjective/accident and the verb/action.
To conclude, the translation of the grammatical terms referring to the parts of speech is one of the best examples of the meta-semiotic problems that may occur in translation. On one hand, the literal translation of the Greek terms referring to the parts of speech into Latin is found; on the other, the translation of the concepts conveyed by the technical terms is verified. The two plans do not necessarily coincide so the same theory can be translated from one language to another, while keeping its philosophical background intact. Alternatively, two theories can be conveyed by the same terms within a single language, changing only the philosophical background of the theory and the linguistic meanings of the terms, while the terms themselves remain unchanged. This was roughly the case with the adjective class. The Roman grammarians translated the Greek term ἐπίθετον into the Latin term adiectivum meaning 'epithet', but nowadays Lat. adjectivum is translated into the English term adjective almost automatically. However, a deep change in the grammatical theory occurred between the 9 th and the 13 th century, so that the class of items called ἐπίθετον in Greek is more similar to the class of items called adjectivum by the Latin grammarians than to the class of items called adjectivum by the Modistae, regardless of the identity of the label adjectivum by the Latin grammarians or by the Modistae. Although contemporary scholars are aware that the modern adjective class and Priscian's adiectivum class do not coincide, they quite often favour the formal similarity of the labels over the conceptual differences encoded in the labels. As a result, Gk. ἐπίθετον is typically translated through 'epithet' and Lat. adiectivum through 'adjective', often without noting that, after the 9 th century, Lat. adiectivum really means only 'adjective', whereas before that, it is a synonym of appositio, both terms being calques used for translating the Greek concept of ἐπίθετον.