H. P. Grice, "Pars Orationis"

As anyone knows who has learned Latin, it is one thing to know a great many words, their definitions and inflections, and quite another to be able to construe complex Latin sentences, much less compose them. Modern textbooks of Latin customarily bridge this gap – between a literally elementary knowledge of the language and facility in reading and composition – with systematic instruction in the structure of Latin sentences, the use of the cases, government, the form and function of clauses, and their idiomatic deployment. Reference grammars typically treat similar information in a separate section on syntax. It is a puzzle that the Anglo-Saxons, pioneers in the study of Latin as a foreign language, managed for centuries with grammars from which such aids were apparently all but absent. They inherited from late antiquity a system of elementary grammars that were in many ways inadequate to the needs of non-native speakers, and were enterprising in developing supplements to and commentaries on these grammars to adapt them to their own needs.1 The early Middle Ages also inherited grammars that did deal systematically with syntax – most notably books 17 and 18 of Priscian’s Institutiones grammaticae – but these were in limited circulation and, where they were known in the seventh and eighth 1 For an inventory and typology of Roman grammars known in the British Isles in the seventh and eighth centuries and of grammars composed during this period, see Vivien Law, The Insular Latin Grammarians . (Woodbridge: The Boydell Press, 1982), and also her “Late Latin Grammars in the Early Middle Ages: A Typological History,” in her volume of collected essays, Grammar and Grammarians in the Early Middle Ages . (London and New York: Longman, 1997): 54-69. 2 centuries, seem not to have attracted much attention as potential teaching texts.2 Some have argued that the non-Latin-speaking cultures of early medieval Europe needed, collectively, to acquire a certain expertise in the formal study of Latin before they were ready to make use of Priscian’s massive reference grammar and to engage in their own speculation and creative adaptation of his work.3 Evidence for this adaptation will be surveyed in Part III of this study. But this explicable delay in engagement with the more sophisticated syntactical studies of antiquity still does not explain how pre-Carolingian English speakers fully mastered Latin. Beyond elementary Latin: The state of our knowledge of pre-Carolingian grammar The lingering lacuna in our understanding of this aspect of the Anglo-Saxon curriculum is due in part to a paucity of texts from the period that can clearly be seen to address the needs of intermediate Latin-learners. In part, though, it is also due to priorities of the last two decades’ research on early medieval grammar. It has been the project of the last 2 See Law’s comments in Vivien Law, The Insular Latin Grammarians . (Woodbridge: The Boydell Press, 1982) 20-21. On Irish knowledge of Priscian, see J.F. Kenney, The Sources for the Early History of Ireland, vol. I: Ecclesiastical . 1929. repr. 1966, 1979), 674-677, and for the glosses to the Irish group of Priscian manuscripts, see Rijklof Hofman, The Sankt Gall Priscian Commentary. Part 1 . (Münster: Nodus Publikationen, 1996). Evidence for early Irish engagement with “Priscianus minor”, the two books on syntax, is lacking, because of defective manuscripts and/or flagging glossators. Aldhelm, whose early training was presumably Irish, knew Priscian’s Institutiones grammaticae (IG), and I suggest some implications of this knowledge for his linguistics in Part I.1, below. However, adaptation of Priscianus minor for syntactical instruction in schools did not begin until Alcuin began work on the text of the IG in the last decade of the eighth century. See Part III, below. 3 E.g. Vivien Law, “Late Latin Grammars in the Early Middle Ages: A Typological History,” Grammar and Grammarians in the Early Middle Ages , (London and New York: Longman, 1997) 54-69, at p. 60. 3 twenty years’ work on the medieval curriculum to identify, characterize, filiate, edit and publish the grammars of late antiquity and the early Middle Ages. This crucial work has, justifiably, focused on texts that are unambiguously grammatical, and that had the widest circulation: Donatus’s Ars minor and maior, above all, and the commentaries, expansions, and supplements to his texts that proliferated in early medieval Europe.4 Also, a confluence of scholarly interests has resulted in an emphasis on word-level linguistics in early medieval grammar. Glossaries and lexical glossing have attracted interest from those engaged in the lexicography of medieval vernaculars, and the prejudice in 19th- and earlier 20th-century philology in favor of word-level and sub-word-level analysis has reinforced this tendency to focus on lexical and morphological information in the early grammars.5 When Latin glosses and glossaries from Anglo-Saxon England have been studied as evidence for language teaching, as they increasingly have in recent years, the emphasis has still been 4 The seminal study and edition of Donatus is Louis Holtz, Donat et le tradition de l'enseignement grammatical: études sur l'Ars Donati et sa diffusion (IVe - IXe siècle) . (Paris: Centre National de la Recherche Scientifique, 1981); for Insular reception of late antique grammars, the pioneering work is Law, The Insular Latin Grammarians . These studies, appearing a year apart, made possible the study of early medieval grammar as it is now being practiced. 5On the general problem of medieval grammatical materials being appropriated by other disciplines, see Vivien Law, “The Historiography of Grammar in the Early Middle Ages,” History of Linguistic Thought in the Early Middle Ages , ed. Vivien Law. Amsterdam Studies in the Theory and History of Linguistic Science, Series III: Studies in the History of the Language Sciences 71 (Amsterdam: John Benjamins, 1993), 1-23. Reflections on the uses of glossography from a number of scholarly perspectives are gathered in R. Derolez, ed., Anglo-Saxon Glossography: Papers Read at the International Conference Held in the Koniklijke Academie voor Wetenschappen Letteren en Schone Kunsten van Belgie , (Brussels: Paleis der Academiën, 1992). Particularly apropos to the reclamation of glossography by historians of the grammatical curriculum are Gernot R. Wieland’s articles “Latin Lemma-Latin Gloss: The Stepchild of Glossologists,” Mittellateinisches Jahrbuch 19 (1984): 91-99, and “Interpreting the Interpretation: The Polysemy of the Latin Gloss,” The Journal of Medieval Latin 8 (1998): 59-71. 4 overwhelmingly lexical, for the very good reason that ostentatious display of unusual vocabulary is the most conspicuous feature of Anglo-Latin style.6 Thus the pressing need for editions of the basic texts, the desire to make grammatical materials serve other fields of investigation, and the dominance of lexical studies tended, until relatively recently, to distract historians of grammar from ways in which linguistic analysis was taught at levels larger than the single word. An exception to this focus on word-level linguistics has been the study of syntactical glossing (marks in manuscripts designed to elucidate the structure of the Latin text). Even here, though, early investigations of syntactical glossing focused on the value of such glosses as evidence for the vernacular languages of the countries where such manuscripts were glossed. For example, Maartje Draak’s study of what she called “construe marks” in a number of ninth-century Irish or Irish-Continental manuscripts focused on the relationship of Old Irish to Latin syntax.7 Similarly, the scholarly back-and-forth between Fred Robinson, Michael Korhammer, and Patrick O’Neill over Anglo-Saxon syntactical glosses 6 At least, vocabulary is the feature that to date has been most remarked upon. The seminal study of this style is Michael Lapidge, “The Hermeneutic Style in Tenth-Century Anglo-Latin Literature,” Anglo-Saxon England 4 (1975): 67-111. For the relation of glossaries to the cultivation of this style, see for example Lapidge’s “The Study of Latin Texts in Late Anglo-Saxon England [1]: The Evidence of Latin Glosses,” Latin and the Vernacular Languages in Early Medieval Britain , ed. Nicholas Brooks. (Leicester: Leicester University Press, 1982) 99-140, and Scott Gwara’s University of Toronto dissertation, “Literary Culture in Late Anglo-Saxon England and The Old English and Latin Glosses to Aldhelm's Prosa de Virginitate ,” (1993). David Porter looks at the evidence for Latin vocabulary acquisition in David W. Porter, “The Latin Syllabus in Anglo-Saxon Monastic Schools,” Neophilologus 78 (1994): 463-482. 7Maartje Draak, Construe Marks in Hiberno-Latin Manuscripts . Mededelingen der koninklijke nederlandse Akademie van Wetenschappen afd. Letterkunde, Nieuwe Reeks, Deel 20, No. 10 (Amsterdam: N.V. Noord-Hollandsche Uitgevers Maatschapplij, 1957). 5 centered on the question of whether the glosses could be read as evidence for vernacular (Old English) word order.8 Our understanding of the nature and purposes of syntactical glossing has become more and more nuanced, but it is only recently that the glosses have been read in conjunction with the modes of analysis taught in grammatical treatises.9 This tendency to see medieval grammar as evidence for something else rather than as evidence of itself is pervasive. The more philosophical grammars of antiquity have suffered less from this neglect, perhaps because their self-conscious systematizing authorizes the modern reader to take their terminology seriously.10 The same is true of speculative grammar in the later Middle Ages.11 But school grammars of late antiquity and the early Middle Ages have perhaps been seen as too simplistic, too tradition-bound, or just too 8 Fred C. Robinson. “Syntactical Glosses in Latin Manuscripts of Anglo-Saxon Provenance.” Speculum 48 (1973): 443-475; Michael Korhammer. “Mittelalterliche Konstruktionshilfen und altenglische Wortstellung.” Scriptorium 34 (1980): 18-58; Patrick P O'Neill. “Syntactical Glosses in the Lambeth Psalter and the Reading of the Old English Interlinear Translation as Sentences.” Scriptorium 46 (1992): 250-56. 9 E.g. Suzanne Reynolds on twelfth-century glosses to Horace in Medieval Reading: Grammar, Rhetoric and the Classical Text . (Cambridge: Cambridge University Press, 1996); and Carin Ruff, “Misunderstood Rhetorico-Syntactical Glosses in Two Anglo-Saxon Manuscripts,” Notes and Queries , June (1998): 163- 166. I discuss further possibilities for this approach to syntactical glossing in Part III. 10 David L. Blank’s edition and commentary of Apollonius Dyscolus and Daniel J. Taylor’s of Varro are excellent examples: David L. Blank, ed., Ancient Philosophy and Grammar: The Syntax of Apollonius Dyscolus , (Chico, CA: Scholars Press, 1982); Daniel J. Taylor, Varro De Lingua Latin X . Amsterdam Studies in the History of the Language Sciences 85 (Amsterdam: John Benjamins, 1996). 11 On speculative grammar, see Irène Rosier, La Grammaire spéculative des modistes . (Lille: Presses universitaires de Lille, 1983) and G.L. Bursill-Hall, Speculative Grammars of the Middle Ages: the Doctrine of Partes orationis of the Modistae . (The Hague: Mouton, 1971); on modistic syntax, see Michael A. Covington, Syntactic Theory in the High Middle Ages: Modistic Models of Sentence Structure . (Cambridge: Cambridge University Press, 1984). 6 muddled to merit serious attention. In collecting her earlier articles for publication in 1997, Vivien Law could still write that Reading grammars of the early Middle Ages, not as documents of a major cultural change, but for the ideas contained within them, is something which very few scholars have as yet undertaken... Although the language of linguistic writing is a well-known source for underlying patterns of thought – for metaphors, buried or living, reveal a good deal more about an author’s assumptions than was necessarily intended – the language of early medieval grammar has scarcely been examined from this point of view.12 There are important exceptions to this tendency. Law’s recent study of Virgilius Maro Grammaticus is an inspiring example.13 Mark Amsler founded his study of “etymological discourse” on the principle of taking the methods of late antique and early medieval grammar seriously on their own terms.14 Most important as an inspiration for this investigation is Marc Baratin’s study of the development of syntax in Roman grammar, La naissance de la syntaxe à Rome. 15 Baratin takes as his starting point the question of how a doctrine of syntax emerged within the Roman grammar when that 12 Law, Grammar and Grammarians in the Early Middle Ages , 249. 13 Vivien Law, Wisdom, Authority and Grammar in the Seventh Century: Decoding Virgilius Maro Grammaticus . (Cambridge: Cambridge University Press, 1995). 14 Mark Amsler, Etymology and Grammatical Discourse in Late Antiquity and the Early Middle Ages . Studies in the History of the Language Sciences 44 (Amsterdam: John Benjamins, 1989). 15 Marc Baratin, La naissance de la syntaxe à Rome . (Paris: Les Editions de Minuit, 1989). 7 tradition, in the centuries before Priscian, largely excluded syntax as a branch of linguistics. Baratin’s method was to look for the potential for syntactical description – the germs of syntactical thought – in the terms and concepts that were used in the teaching of grammar. In approaching the question of how syntax was taught to learners of Latin as a second language after Priscian but in the absence of wide circulation of his work, we can begin from the same point, by taking the terms and concepts of early grammars seriously and considering what kinds of linguistic description they support. Anglo-Saxons as Second Language Learners In describing the Anglo-Saxons’ Latin-learning project as one of second language acquisition, I do not mean to assert that we can say anything definite about the psycholinguistic parameters of the Latin learning experience in the early Middle Ages. Rather, I want to emphasize that the Anglo-Saxons lived in a world where Latin was the second language of many but the mother tongue of none.16 In this context, theories of second language acquisition based on studies of how children acquire spoken bilingualism may be of only limited applicability, since this field as currently constituted does not normally address itself to the situation of those acquiring a new language primarily or exclusively as a written language.17 16 Michael Herren makes an interesting analogy to the status of World Englishes in “Latin and the Vernacular Languages,” Medieval Latin: An Introduction and Bibliographical Guide , eds. F.A.C. Mantello, and A.G. Rigg. (Washington, D.C.: Catholic University of America Press, 1996) 122-129, at 124. 17 Scott Gwara makes an interesting attempt to apply second language acquisition theory to the Colloquies of Ælfric Bata in “Second Language Acquisition and Anglo-Saxon Bilingualism: Negative Transfer and 8 The actual Latin-learning experience of any given Anglo-Saxon student would have varied, of course, both as to the age at which he or she began the study of Latin, and in the extent to which that learning approximated the “immersion method”. The child oblate immersed in the acoustical world of liturgical Latin, starting his grammatical studies at the age of seven, is, perhaps, the best-case scenario. Through singing the liturgy and memorizing the Psalter, he or she would, as Vivien Law has pointed out, have been exposed to the aural shapes of Latin words even before coming to the formal study of Latin grammar.18 Bede, who entered Monkwearmouth-Jarrow as a boy and may have been trained in Roman chant by John the Archcantor,19 would have come as close as any of his compatriots to learning the language orally from native speakers, but even there it is debatable to what extent we can describe a seventh-century Italian as a native speaker of Latin.20 The students of the school of Canterbury described by Bede as speaking Latin Avoidance in Ælfric Bata's Latin Colloquia, ca. A.D. 1000,” Viator 29 (1998): 1-24. Gwara includes much useful bibliography on recent work on second language acquisition; another particularly accessible overview is Ellen Bialystok and Kenji Hakuta, In Other Words: The Science and Psychology of SecondLanguage Acquisition . (New York: Basic Books, 1994). The Colloquies , which are edited by Gwara with a translation by David W. Porter in Anglo-Saxon Conversations: The Colloquies of Ælfric Bata . (Woodbridge, Suffolk: Boydell Press, 1997), deserve further study as evidence of an oral/aural component in Latin instruction. 18 Vivien Law, “The Study of Grammar,” Grammar and Grammarians in the Early Middle Ages . (London and New York: Longman, 1997) 129-153, at 129. 19 Bede, Historia ecclesiastica (HE) 4.18 and Historia abbatum 6. Bede probably entered the monastery in 679 or 680, which would have been at just about the same time that Benedict Biscop returned from Rome with John. On these dates see Venerabilis Baedae Opera Historica , ed. Charles Plummer. (Oxford: Clarendon Press, 1896), pp. xi, 369 and note. 20 According to Roger Wright’s controversial hypothesis on the distinction between Latin and Romance, the seventh and early eighth centuries – the period of Anglo-Saxons’ first contact with Latin – is a pivotal period in the emergence of the Romance vernaculars. (Roger Wright, Late Latin and Early Romance in 9 and Greek with native fluency probably also came close to this experience, in that they were exposed to teachers from the truly bilingual (Greek and Latin) culture of the eastern Mediterranean.21 But Aldhelm, the only product of this school whose own work survives, did not benefit from Theodore and Hadrian’s school at Canterbury until he was an adult.22 Beyond these first two generations of Anglo-Latin studies, it becomes increasingly unlikely that Anglo-Saxons would have learned Latin from anyone – whether English, Irish, or other – who had not learned Latin in school. The authors I will be considering in this study were all non-native speakers writing for the benefit of other non-native speakers. Spain and Carolingian France . (Liverpool: Francis Cairns, 1982).) Wright suggests that Carolingian spelling reforms were motivated by the distance between what Alcuin had been trained in York to think of as Latin and what he found being spoken in Francia in the last years of the eighth century. If Wright is right, we should at least question the relationship between the liturgical and literary Latin being taught by Romans in Northumbria a hundred years earlier, and what those Romans would have spoken at home. On the emergence of Italian as distinct from Latin in the seventh century, see Dag Norberg, “Le Développement du Latin en Italie de Saint Grégoire le Grand à Paul Diacre,” Settimane di studio del Centro italiano di Studi sull'alto medioevo 5 (1958): 485-503, at 495. 21 HE 4.1, 5.23. Theodore came from Tarsus in the Greek-speaking eastern Mediterranean, studied in Antioch, and probably lived in a Greek-speaking community at Rome before coming to England late in life; Hadrian was most likely from Greek-speaking Libya and spent most of his career in Naples. For a hypothetical reconstruction of Theodore’s travels, see Michael Lapidge, “The Career of Archbishop Theodore,” Archbishop Theodore , ed. Michael Lapidge (Cambridge: Cambridge University Press, 1995) 1-29; on Hadrian see Lapidge’s supplementary note, p. 505, to Michael Lapidge, “Aldhelm's Latin Poetry and Old English Verse,” Anglo-Latin Literature 600-900 , (London: The Hambledon Press, 1996 (1979)) 247-269. 22 On Aldhelm’s experience of the School of Canterbury, see further the discussion of his metrics in Part I, below. 10 At the other end of the spectrum from the life-long monastic student is the eager but harried layman, pressed with other business and coming to Latin literacy late in life. Alfred is the most famous example of this pattern. However ambiguous Asser’s report of the king’s developing literacy, there is no suggestion that he learned Latin unmediated, by the “immersion method”, or that he achieved (or desired to achieve) spoken fluency in the language.23 Whether or not spoken and liturgical language had a significant effect on the Latin-learning process for most Anglo-Saxon students, it is clear that their introduction to grammar meant encountering the fundamental linguistic categories by which their experience of Latin texts would be organized. The processes that lay between the acquisition of fundamental concepts and the mastery of written Latin that many AngloSaxon authors ultimately achieved are my concern. 23 Asser, Asser's Life of King Alfred, together with the Annals of Saint Neots erroneously ascribed to Asser , ed. William Henry Stevenson. (Oxford: Clarendon Press, 1959). Two recent, influential approaches to Alfred’s experience and to the meaning of literacy in Anglo-Saxon England are Seth Lerer, Literacy and Power in Anglo-Saxon Literature . (Lincoln: University of Nebraska Press, 1991) and Katherine O'Brien O'Keeffe, Visible Song: Transitional Literacy in Old English Verse . Cambridge Studies in Anglo-Saxon England 4 (Cambridge: Cambridge University Press, 1990). Scott Gwara, incidentally, has a very poor estimation of the oral Latin competence of even the most highly-educated Anglo-Saxons: “To my mind, Latin was probably acquired during the pre-conquest period with prodigious labor and spoken at marginal competence, aided by a census of memorized centos and formulae.” (“Second Language Acquisition,” p. 8.) 11 The scope and purpose of this study This study will be devoted to examining closely some of the major “paragrammatical” texts24 of the Anglo-Saxon curriculum with an eye to illuminating how this mastery of Latin might have taken place, most especially in the seventh and eighth centuries when syntactical studies had not, anywhere in European schools, been established as a part of the curriculum with enough status to be recognized in the standard treatises. To this end, I will focus on the didascalia of Aldhelm and Bede: Aldhelm’s De metris ac de pedum regulis and Bede’s De arte metrica in Part I, and Bede’s De schematibus et tropis and De orthographia in Part II.25 In Part III, I will briefly survey recent work on Carolingian developments in language study, and then consider to what extent Ælfric’s Grammar and Byrhtferth’s Enchiridion are in continuity with the approaches of Bede and Aldhelm, and in what ways they reflect the Carolingian innovations to which their authors had access. Focusing on the works of four known authors has the advantage of allowing an assessment of what methods, terms, and concepts were indisputably being written about 24 The term is Law’s (Insular Latin Grammarians , xiii). 25 On my reasons for including metrical treatises in this study, see below. In focusing on the works of Aldhelm and Bede I do not mean to slight two other Anglo-Saxon grammarians: Bede’s contemporaries, Boniface and Tatwine. Neither adds substantially to the major trends in early medieval grammar that I outline here and neither was as influential as Bede, although both have their points of interest: Tatwine’s treatment of the preposition includes information on case rection, and Boniface’s prefatory letter to Sigebert is interesting for what it discloses of the grammarian’s methods. Boniface’s grammar is edited by G.J. Gebauer and Bengt Löfstedt in CCSL 133B:13-99, and Tatwine’s by Maria de Marco in CCSL 133: 1- 93. On their sources and methods, see Law, The Insular Latin Grammarians , 64-67 and 77-80, and eadem, “The Study of Grammar in Eighth-Century Southumbria,” Anglo-Saxon England 12 (1983): 43-71. 12 at identifiable times and places in pre-Conquest England. To this end, my emphasis throughout will be on what we can learn from these four authors’ own works about their conceptual frameworks. This approach necessarily leaves out many of the most exciting textual issues in current studies of the early medieval grammatical curriculum: which late-antique grammars were known at which centers; the dating, localization, and sources of the many anonymous grammars and not a few whose authors, while named, are not securely localized; and the evidence of glossing in manuscripts written or owned in Anglo-Saxon England. I will engage all of these subjects to some degree as they arise, and the last in some detail in Part III. Nonetheless, I regard my main purpose here as providing a touchstone for related studies and filling a gap in the efforts that have so far been taken to understand Latin teaching in the early Middle Ages. My intention is to read with and against one another a group of key texts whose circumstances of composition are reasonably well documented but which are not typically examined together or read from the perspective of what they might have offered the struggling intermediate Latin-learner. I hope that if I can unpack through these texts the conceptual categories under which Latin was understood and attempt an analysis of the pedagogical utility of those categories, I will have provided a framework in which we may more clearly read the elements of the Anglo-Saxons’ “hidden curriculum”. By “hidden curriculum”, I mean those subjects that we can deduce must have been studied but for which the evidence is indirect or circumstantial. That is, we know from direct evidence of large numbers of grammars that learning the parts of speech and 13 their attributes, and testing that knowledge by parsing forms, were central tasks of the Latin student.26 The grammars that present this information and model these skills give little evidence that the order of elements in a Latin sentence was also a matter of concern to masters, students, and readers, but “ordo est” glosses and other sequence-markings in Anglo-Saxon manuscripts make it clear that this was a widely-used concept.27 Glosses are perhaps the strongest evidence, if not always unambiguous, for what went on in Anglo-Saxon schools outside the grammars per se. But evidence for terms and concepts that would have been common currency in the classroom is also available in treatises on language that do not appear, at first glance, to concern themselves with elementary reading instruction. I include in this category handbooks on metrics, which we might think would be appropriate only to a student who was already beyond needing help with basic syntax; orthographical handbooks, which treat deviations from usages presumably already learned; and manuals of “rhetoric-in-grammar” inspired by the third part of Donatus’s Ars maior.28 These last, as developed in a monastic setting, might be seen as 26 Donatus’s Ars minor and part two of his Ars maior are only the best known of the Schulgrammatik type, which takes as its central subject the classification of the parts of speech. On the development of this type in the Roman traditions, see Karl Barwick, Remmius Palaemon und die römische ars grammatica . (Leipzig: 1922), and Louis Holtz, Donat et la tradition de l'enseignement grammatical: études sur l'Ars Donati et sa diffusion (IVe - IXe siècle) . (Paris: Centre National de la Recherche Scientifique, 1981), 75- 216. On parsing grammars, see Martha Bayless, “Beatus Quid Est and the Study of Grammar in Late Anglo-Saxon England,” History of Linguistic Thought in the Early Middle Ages , ed. Vivien Law. Amsterdam Studies in the Theory and History of Linguistic Science, Series III: Studies in the History of the Language Sciences, (Amsterdam: John Benjamins, 1993) 71: 67-110. 27 On syntactical glossing and the idea of ordo , see Part III, below. 28 The term rhetoric-in-grammar, “Rhetorik in der Grammatik”, is borrowed from Gabriele Knappe, Traditionen der klassischen Rhetorik im angelsächsischen England . (Heidelberg: Universitätsverlag C. Winter, 1996). On the origins and development of rhetorical schemes and tropes as part of the standard 14 guidebooks for beginning exegetes – that is, for competent readers who are embarking on the serious task of reading, copying, and interpreting Scripture.29 Treatises of this sort, aimed at the intermediate student, disclose elements of the hidden curriculum in two ways. First, they repeatedly reveal assumptions on the part of their authors about what their readers already know. Thus, a writer on metrics who wishes to discuss end-stopping and enjambment will draw on the idea of a syntactically complete unit (as opposed to a metrically complete unit), despite the fact that notions like “sentence” are cursorily and contradictorily defined, if at all, in the grammars to which their readers would have been exposed. That the metricist uses this concept need not imply that whole lessons of which we have no record were devoted to explaining what constitutes a complete sentence, but it does suggest that such concepts were implicitly if not explicitly present in the conceptual arsenal of the intermediate reader. Similarly, when the orthographer30 treats deviations from expected case regimen, we can assume at worst that the user of such irregularities would learn to extrapolate from a discussion of the unusual to what was usual. A more optimistic interpretation of late-Roman school grammar, see Marc Baratin, and Françoise Desbordes, “La 'troisième partie' de l' ars grammatica ,” The History of Linguistics in the Classical Period , ed. Daniel J. Taylor. Studies in the History of the Language Sciences 46, (Amsterdam: John Benjamins, 1987) 41-66; and Louis Holtz, Donat et le tradition de l'enseignement grammatical: études sur l'Ars Donati et sa diffusion (IVe - IXe siècle) . (Paris: Centre National de la Recherche Scientifique, 1981) 183-216. Knappe covers the reception of this material in Anglo-Saxon England. 29 Anna Carlotta Dionisotti makes this suggestion about Bede’s De orthographia in “On Bede, Grammars and Greek,” Revue Bénédictine 92 (1982): 111-141. See also my discussion in Part II below. 30 Orthographer: a writer on orthographia , which in the early Middle Ages refers broadly to “correct usage”, rather than simply to spelling. See Part II below on Bede’s De orthographia . 15 such evidence would be that the user of an orthographic handbook would have been taught what verbs governed what cases, even if our manuscript evidence would not suggest that he had access to one of the ancient grammars that explicitly treat these rules. A middle road would be to suggest that the intermediate reader had read, heard, and sung enough Latin to recognize a deviation from the norm when he encountered one; resorting to a handbook of exceptions31 to see whether his manuscript needed emending, he would find the deviation stated explicitly, localized historically, and grouped with similar phenomena. A list of exceptions would, therefore, help to impose some analytical shape on a rule unconsciously learned. Again, when a writer on schemes and tropes calls attention to variation, repetition, and other forms of rhetorical patterning on the level of the clause or sentence, he is working from an assumed base of knowledge of normal or unmarked patterns. As with the grammatical rules engaged by the orthographer, these may or may not have been learned consciously, but a discussion of variations from them will crystallize them in the mind of the reader so that they become active tools of analysis. The second way in which paragrammatical treatises disclose aspects of the hidden curriculum is when they repeatedly apply similar terminology, metaphors, or modes of analysis in different linguistic settings. While the technical vocabulary of ancient and 31 See my discussion of the De orthographia in II.2, below. 16 medieval Latin grammar is notoriously both conservative and slippery,32 the recurrence of certain terms, images, and modes of explanation throughout the grammatical curriculum points to a small number of persistent streams in the way language was conceived in the early Middle Ages, and these ingrained ways of thinking about language would have affected Latin learners as they struggled to advance from those aspects of the language for which they had good technical descriptions to those for which they did not.33 What lies between Donatus and mastery of Latin? This study is concerned with the questions: what did an Anglo-Saxon student need to know about Latin beyond what Donatus could teach, and to what extent can we see those needs being addressed in other parts of the curriculum? In her study of preCarolingian English and Irish grammars, Vivien Law characterizes those Insular grammars that deal with Latin accidence as “elementary” and those that take the form of continuous commentaries on Donatus as “exegetical”. In drawing this distinction, she remarks, “The elementary grammarians provided the student with sufficient knowledge to make sense of a Latin text,” and describes the exegetical grammars as “advanced 32 On this see Vivien Law, “The Terminology of Medieval Latin Grammar,” Grammar and Grammarians in the Early Middle Ages , (London and New York: Longman, 1997) 260-269. 33 Elmar Siebenborn offers an example of this approach, exploring a technical term through its etymology and range of application, in “Herkunft und Entwicklung des Terminus technicus perivodoß: ein Beitrag zur Frage der Entstehung von Fachterminologien,” The History of Linguistics in the Classical Period , ed. Daniel J. Taylor. (Amsterdam: John Benjamins, 1987) 229-249. 17 textbooks”.34 I have serious reservations about the former statement. As for the latter description, while it may be true that exegetical grammars were what advanced students read, I am not convinced that they would have learned substantially more about how to put Latin sentences together from these obsessively-detailed inquiries into why Donatus said what he said the way he said it.35 An example of the inadequacy of these commentaries for language learners is the frequently-noted fact that when they discuss the order of the parts of speech, they are actually talking about the metalinguistic rationale for listing the parts the way they are conventionally listed in grammars.36 To put it another way, Donatus alone does not teach enough Latin to enable students to read the commentaries on Donatus. Teachers teach Latin, using Donatus and whatever other resources are available to them, and we must look for clues as to what they might have made of those resources. The term “syntax” I have already used the term “syntax” as a catch-all for “what Donatus could not teach”. To use this term is, admittedly, to impose certain preconceptions on the way language was conceived in the early Middle Ages. It is not my intention to offer a theory 34 Law, Insular Latin Grammarians , 81. 35Law herself expresses doubts elsewhere about the results of the elementary grammars’ obsessive focus on accidence: “Indeed, there is a marked tendency to eliminate all information not directly relating to accidence, even at the expense of facts essential to the construction of a correct Latin sentence, like noun gender.” (Insular Latin Grammarians , 54). On the methods of the exegetical grammarians, see Law, Insular Latin Grammarians , 81-98. 36 E.g. Ars Ambrosiana , ed. Bengt Löfstedt. CCSL 133C (Turnhout: Brepols, 1982) 3. 18 of syntax, or to choose among competing definitions offered by modern linguists or grammarians. Too narrow a definition would leave no space for medieval definitions to emerge and would run the risk of blinding the reader to ways of thinking about language that do not fit neatly into modern linguistic pigeonholes.37 On the other hand, we need a working definition clear enough to provide a sense of the kind of teaching we are looking for. Further, my first concern is with syntax as a pedagogical tool. There is often – and rightly – some distance between definitions of syntax appropriate to a theory of linguistics and those that would be useful in teaching a specific language. I expect that contemporary theories of how language works will emerge from medieval curricular texts, and it would be misleading to begin with a definition crafted to express a twentiethcentury understanding of the nature of language. Avoiding modern preconceptions need not mean failing to test medieval understanding of language against our own conceptions. It will, however, necessitate adopting a flexible attitude towards the proper objects of syntactical study. Modern linguists typically insist on the sentence as the limit and proper object of syntactical inquiry.38 The Latin grammatical tradition did not privilege the sentence as we would now understand it, and instead emphasized the continuity of linguistic combinations from 37 Marc Baratin similarly rejects the notion of looking for modern concepts in ancient texts: “Et d’abord, qu’est-ce que la syntaxe? Il n’est pas trop urgent de chercher à répondre à cette question.” ( Naissance , 8.) 38 So, influentially, Chomsky as cited above; also P.H. Matthews, Syntax . (Cambridge: Cambridge University Press, 1981), p. xix. 19 the smallest to the largest units.39 It therefore seems counterproductive to insist on a terminus above which we will not notice descriptions of linguistic combinatory systems. The resources and methods of such fields as stylistics and discourse analysis offer a model for understanding (and, in some respects, an analogy to) ancient and medieval modes of linguistic analysis. Critics working in these fields are more ready to see the boundary of the sentence as fluid, and to see grammatical, syntactical, and semantic markers of coherence operating both within sentences and across whole texts.40 Topics of 39 Priscian’s is the clearest exposition of this view of language in the Latin tradition: Institutiones grammaticae , Grammatici Latini 2:1-597, 3:1-377 ed. Heinrich Keil. (Leipzig: Teubner, 1855-1880). For the view of language as a continuous combinatory system, see my discussion of the smallest units of language, below. 40Thus Harald Weinrich: “...linguistics is necessarily textual. Syntax, thus, should be conceived of as textual or macro-syntax. That means that there is no reason to stop syntactic research at the magic border of the sentence.” (Harald Weinrich, “The Textual Function of the French Article,” Literary Style: A Symposium , ed. Seymour Chatman. (London and New York: Oxford University Press, 1971) 221-240, at 221-222). For discussion of and extensive bibliography on intersentence coherence, see for example the discussion in Nils Erik Enkvist, “On the Place of Style in Some Linguistic Theories,” Literary Style: A Symposium , ed. Seymour Chatman. (London and New York: Oxford University Press, 1971) 47-61, especially pp. 56-57. Particularly congenial to older grammatical methods is the use in stylistics of the traditional inventory of rhetorical figures and tropes, without undue anxiety about adding these tools to the armory of linguistic analysis. Enkvist notes this interdisciplinary sympathy with approval at p. 50. For an example of the incorporation of traditional rhetorical categories into a descriptive grammar whose aims are shaped by twentieth-century stylistics, see the categories employed under the heading Stylistik by in J.B. Hofmann, Lateinische Syntax und Stylistik . (Munich: C.H. Beck'sche Verlagsbuckhandlung, 1965) – though the interest in “affect” and the syntax of the spoken language that especially characterizes Hofmann’s work is less applicable to this study. For an overview of applications of discourse/coherence perspectives on syntax and word order to Latin pedagogy, see Glenn M. Knudsvig, and Deborah Pennell Ross, “The Linguistic Perspective,” Latin for the 21st Century: from Concept to Classroom , ed. Richard A. LaFleur. (Glenview, IL: Scott ForesmanAddison Wesley, 1998) 25-35. Some applications of these perspectives to particular problems in Latin grammar are found in Harm Pinkster, Latin Syntax and Semantics . (London: Routledge, 1990); Deborah 20 interest to stylistics and discourse analysis have suggestive parallels in early medieval grammar, including logical connectors between sentences and variations in element order conditioned by emphasis, topic, or focus. 41 Moreover, the discussion in early medieval grammars of rhetorical patterning across small groups of sentences, or within units that might contain several sentences (such as the biblical verse), would be among the most helpful to a student learning to negotiate continuous Latin prose.42 Such patterning would help the student to delimit clauses and sentences, but we will not see what patterns are being taught if we eliminate the rhetorical analysis of texts from consideration.43 In order to appreciate how this kind of analysis functions pedagogically, we will need to admit to our study aspects of language usually thought of as the province of stylistics or discourse analysis. It seems practical, then, to remain alert to modes of description that may cross the modern boundaries of syntax, and to examine precisely what Donatus’s Ars minor and maior (and the two other grammatical texts of near-universal circulation, Isidore’s Etymologiae, Book I, and Priscian’s Institutio de nomine pronomine et verbo) do teach Pennell Ross, “The Order of Words in Latin Subordinate Clauses,” Dissertation (University of Michigan, 1987); and Deborah Pennell Ross, “The Role of Displacement in Narrative Prose,” New Studies in Latin Linguistics , ed. Robert Coleman. (Amsterdam: John Benjamins, 1991) 453-466. 41 On the terminology and methods of discourse analysis, see Gillian Brown, and George Yule, Discourse Analysis . (Cambridge: Cambridge University Press, 1983). On early medieval analogues to this methodology, see the discussion of ordo in Part III below. 42 I discuss this kind of teaching in Part II.1, on Bede’s De schematibus et tropis . 43 As noted above, I am using “rhetorical” here in the sense implied by Knappe’s “rhetoric-in-grammar” formulation, and do not want to claim that rhetoric was taught as a separate subject in Anglo-Saxon England. 21 about how smaller linguistic units combine into larger ones. We can then suggest what might be missing from their accounts by comparison to what modern grammars include under the rubric “syntax”.44 The result will be a working inventory of those subjects that we might look for as part of the hidden curriculum. What can one learn from Donatus? The status of Donatus as the sine qua non of medieval grammatical instruction is attested by the way in which grammatical writing in from the fifth century through the eighth, at least, overwhelmingly comments on, responds to, or supplements his works, as well as by comments by later grammatical writers who assume that their students have already mastered Donatus. Isidore, for example, gives pride of place to Donatus throughout Book I of the Etymologiae,45 and Ælfric designed his own Grammar for boys who had already read at least Donatus minor, clearly situating his grammar as an intermediate work: “Ego Ælfricus...has excerptiones de Prisciano minore uel maiore uobis puerulis tenellis ad uestram linguam transferre studui, quatinus perlectis octo partibus 44 On the status of Donatus, Isidore, and Priscian’s short work on the noun, verb, and pronoun as the grammarian’s basic library, see Law, Insular Latin Grammarians , 23-24. 45 Isidore of Seville, Etymologiae sive origines , ed. W.M. Lindsay. (Oxford: Clarendon Press, 1911), and see the comments in Vivien Law, Insular Latin Grammarians , 14-16 and passim. 22 Donati in isto libello potestis utramque linguam, uidelicet latinam et anglicam, uestrae teneritudini inserere interim, usque quo ad perfectiora perueniatis studia.46 I, Ælfric...have been eager to translate these excerpts from Priscian minor and maior into your language for you tender little boys, so that, having read through the eight parts (sc. of speech) of Donatus, you might in this little book for the time being implant in your tenderness both languages, namely Latin and English, until you arrive at more perfect studies. The format of the Ars maior is threefold: Ars maior I covers the smallest phonological and graphic elements of language: letters, syllables, metrical feet, accents, and punctuation. Ars maior II covers the parts of speech: nouns, pronouns, verbs, adverbs, participles, conjunctions, prepositions, and interjections. Ars maior III covers barbarisms, solecisms, and “other faults”, metaplasms, schemes, and tropes. Donatus’s Ars minor covers the material of Ars maior II in a condensed, question-and-answer format. The contents of Donatus’s grammars can be understood in the context of Roman education. Beginning grammars were propaedeutic to the literary and linguistic criticism of the poets, which was the chief aim of grammar in antiquity47 – and ultimately to training in rhetorical production. To these ends, the grammar was designed to introduce native speakers to the analytical categories applicable to the written form of their language, and to the conventional headings under which deviations – whether mistakes or 46 Ælfric, Aelfrics Grammatik und Glossar , ed. Julius Zupitza. trans. Introduction by Helmut Gneuss. 2nd ed. (Berlin: Max Niehans, 1966), 1. I discuss Ælfric’s Grammar in Part III, below. 47 See e.g. Quintilian, Institutiones oratoriae libri XII , ed. Michael Winterbottom. (Oxford: Clarendon Press, 1970), at I.iv.2. 23 poetic licenses - from “ordinary” usage were classified. For native speakers of Latin, there was no need to specify that ordinary usage, in the sense of the vocabulary, forms, and syntax of the language. But that gap in the Roman grammars, between the conceptual categories for describing the smallest units of language and lists of deviations from ordinary usage, is precisely what second-language learners needed to have supplied. The partes orationis As noted above, the basic text on the parts of speech was Donatus’s Ars minor (A.m.), which deals briefly and only with the partes, in question and answer format. The treatment in the A.m. would most often have been supplemented with Donatus’s longer treatment of the partes in the Ars Maior (A.M.), with Priscian’s Institutio de nomine pronomine et verbo,48 and with Book I of Isidore’s Etymologiae.49 Here I survey the treatment of the parts of speech, in their conventional order (nomen, pronomen, uerbum, aduerbium, participium, coniunctio, praepositio, interiectio), as presented in these four widely-used texts.50 Donatus’s discussion of each pars, both in the A.m. and in the A.M., proceeds predictably and systematically: he defines each part, lists its accidents, and then inventories and exemplifies each accident in turn.51 Isidore offers a characteristically 48 GL 3: 443-456. 49 Isidore of Seville, Etymologiae sive origines , ed. W.M. Lindsay. (Oxford: Clarendon Press, 1911). 50 My discussion has been influenced by Vivien Law’s essay, “The Terminology of Medieval Latin Grammar,” Grammar and Grammarians in the Early Middle Ages , (London and New York: Longman, 1997) 260-269. 51 On Donatus’s methodology and the structure of his grammars, see the discussion in Holtz, Donat , 49- 74. 24 etymological spin on Donatus’s doctrine, while Priscian offers a more detailed and systematic treatment of the inflecting parts of speech. The Noun In the Ars minor, Donatus defines a noun as “pars orationis cum casu corpus aut rem proprie communiterue significans,” “a part of speech with case signifying, either properly or commonly, a body or thing.” Its accidents are “qualitas, comparatio, genus, numerus, figura, casus.” The qualitates are propria – proper – and appellatiua, which embraces both common nouns and adjectives. The adjective was not defined as a separate part of speech in antiquity and the Middle Ages. Because the category nomen includes adjectives, comparison – comparatio – is an accident of some nomina, that is, “appellative” nouns signifying quality or quantity. In his discussion of comparatio, Donatus gives the rules for the syntax of cases with comparatives and superlatives, with examples: Comparatiuus gradus cui casui seruit? Ablatiuo sine praepositione...Superlatiuus cui? Genitiuo tantum plurali. (Holtz, Donat, 585-86) What case does the comparative grade serve? The ablative without preposition...What case does the superlative serve? The genitive plural only. Genera, genders, are masculinum, femininum, neutrum, and commune, the latter including nouns that can be masculine or feminine (sacerdos) and adjectives of one termination (felix). Also “epicoenon, id est promiscuum, ut passer, aquila.” Numeri, 25 numbers, are singular and plural. The figura of a noun, as of the other parts of speech, is its status as either simplex or compound, composita. Donatus specifies four ways in which compounding can take place. The discussion of casus is an opportunity to give sample declensions. The cases are “nominatiuus, genetiuus, datiuus, accusatiuus, uocatiuus, ablatiuus. Per hos omnium generum nomina pronomina participia declinantur hoc modo:” “Through these nouns, pronouns, and participles of all genders are declined in this way.” Donatus declines one noun of each gender: magister (masculine), musa (feminine), scamnum (neuter), sacerdos (common of two genders), and felix (common of three genders). The chapter on the noun ends with hints on which genitive and dative plural terminations correspond to which ablative singular terminations, but Donatus does not otherwise classify the nouns in declensions. In the Ars maior, (A.M.) this discussion is fleshed out with further subdivisions and occasional allusions to disputes within the grammatical tradition, but mainly with additional semantic and derivational information. For instance, under qualitas in the Ars maior, Donatus treats derivative morphology, Greek declensional forms, homonyms and synonyms, and the like. Under comparatio, we have a few more examples of the use of the comparative and the superlative, but no substantially new information. The discussions of genus, numerus, and figura are expanded with reference to exceptions and unusual instances. The discussion of casus includes nouns that are indeclinable or defective, but also introduces the construction of cases with certain adjectives: 26 Sunt praeterea nomina, quorum alia genetiuum casum trahunt, ut ignarus belli, securus amorum; alia datiuum, ut inimicus malis, congruus paribus; alia accusatiuum, sed figurate, ut exosus bella, praescius futura; alia ablatiuum, ut secundus a Romulo, alter a Scylla; alia septimum casum,52 ut dignus munere, mactus uirtute. (A.M. II.9; Holtz, Donat, 625) There are nouns, too, of which some take the genitive case, as “ignorant of war”, “free of loves”; others the dative, as "hostile to the evil", "suitable for equals"; others the accusative, but figuratively, as “detesting war”, “predicting the future”; others the ablative, as “second from Romulus”, “next after Scylla”; and others the seventh case, as in “worthy of office”, “honored for virtue”. The rest of the A.M. section De nomine is taken up with further observations on declensional patterns, although Donatus still does not group the nouns into the five declensions we are familiar with. This job is left to Priscian, and it is undoubtedly his rational presentation of Latin inflectional morphology which helped make his short text on the noun, pronoun, and verb such a popular supplement to Donatus. The Institutio de nomine pronomine et verbo opens: Omnia nomina quibus Latina utitur eloquentia quinque declinationibus flectuntur, quae ordinem acceperunt ab ordine uocalium formantium genetiuos. (GL 3: 443) 52 Donatus has just explained that some consider the ablative without aba separate, seventh case. 27 All the nouns that Latin eloquence uses are inflected in five declensions, which have taken their order from that of the vowels forming the genitives. He then identifies the characteristic genitive singulars of the five declensions, and lists the nominative terminations that appear in each declension (with comparisons to Greek forms: Priscian taught Latin in Greek-speaking Constantinople). He relates derivational morphology to the declensions.53 He lists the endings for all cases and numbers for each declension, pointing out which endings resemble one another in which paradigms. He closes the section on the noun by noting that this brief summary will suffice “ad instituendos pueros”, for instructing boys, and recommends his seven books on the noun (in the Institutiones grammaticae54) for a fuller treatment. None of these basic texts on the noun mentions the most common uses of the cases. Syntax of cases is mentioned only in the context of unusual uses. The Pronoun According to Donatus, a pronoun is “pars orationis, quae pro nomine posita tantundem paene significat personamque interdum recipit,” “a part of speech which, used 53 Derivational morphemes are those that create new words out of preexisting words, as for example endings like –tas , –tio , and –tudo , which create new third-declension feminine abstract nouns. Latin derivational affixes have predictable consequences for the inflectional class of the newly-produced word, as in the examples above. 54 GL 2:1-597, 3:1-377. 28 in place of a noun, means almost the same thing, and sometimes has ‘person’.”55 Pronouns have six accidents: qualitas, genus, numerus, figura, persona, casus. The qualitates of the pronoun are finita and infinita. Finita are the personal pronouns and infinita are the others. Gender, number, and figura are the same as for nouns. Personae are first, second, and third. Pronouns have the same cases as nouns, and here, too, Donatus gives sample declensions of pronouns of each type: ego, tu, ille; ipse, iste (“minus quam finita”); hic (“articulare praepositum uel demonstratiuum,” “the preposed articular or demonstrative pronoun);56 is (“subiunctiuum uel relatiuum,” “the subjoined or relative pronoun); quis (“infinitiua”, “indefinite”); the possessive adjectives of all persons and numbers; and finally a list of compound pronouns. (Holtz, Donat, 588- 91) Although Donatus describes the pronoun as substituting for a noun, he does not, as Holtz notes, interest himself in the mechanics of that substitution.57 Correlation between pronouns and their referents does not concern Donatus at all, nor does the structure of relative clauses. Although Donatus lists correlative adjectives (talis/qualis, 55 Holtz devotes a chapter to the doctrine of Ars minor 3: Holtz, Donat , 127-135. As Holtz notes (p. 128 and n. 25), Donatus does not concern himself with the mechanics of the substitution of pronouns for nouns. This becomes a topic for commentary, although the commentators emphasize the use of pronouns to avoid fastidium .) 56 In the Greek grammatical tradition, the article and the pronoun were treated as a single part of speech, the a[rqron. The traces of this doctrine that survive in the Latin tradition understandably cause some confusion for a language which has no article per se . See Holtz, Donat, 125-126, 131-33. On the grammatical convention of treating the demonstrative hic / haec / hoc as if it really were an article, see Holtz, Donat , 131, and Suzanne Reynolds, Medieval Reading: Grammar, Rhetoric and the Classical Text . (Cambridge: Cambridge University Press, 1996), 68-72. 29 tantus/quantus) as pronouns, there is no discussion of the syntax of correlative clauses. Nor is there of the syntax of relative clauses; indeed, Donatus does not distinguish interrogative and relative pronouns, but lists them as alternate forms of one another. The suggestive term subiunctiuum has nothing to do with the use of a pronoun in a subordinate clause: is and idem are “subjunctive” only in that they follow another pronoun or noun to which they refer, while quis, in contrast, is “prepositive” because it always comes at the head of a clause.58 This terminology does presuppose a notion of the relative order of constituents, but does not articulate a rationale for that order, and does not describe that order in relation to hypotaxis. In the interest of categorizing every form by morpho-semantic criteria, and of preserving a strict parallelism between the chapter on the noun and that on the pronoun, Donatus suppresses all other kinds of information on the pronoun. The Ars maior does not materially improve on this situation. Priscian has a more rational treatment of the morphology, distinguishes the interrogative pronoun, and casts doubt on the pronominal status of the correlatives, but he adds nothing of syntactical import. (GL 3: 449-450) The Verb Like the noun, the verb is defined in the Ars minor by morpho-semantic criteria: the verb is “pars orationis cum tempore et persona sine casu aut agere quid aut pati aut neutrum59 significans,” “a part of speech with tense and person, without case, signifying 57 Holtz, Donat , 130-131 58 ibid. 59 A verbum neutrum is a verb that is always intransitive; see discussion below. 30 either doing something or suffering something or neither.” Its accidents are qualitas, coniugatio, genus, numerus, figura, tempus, persona. (A.m. 4; Holtz, Donat, 591) Qualitas uerborum consists in moods (modi) et in “forms” (formae). The former are, more or less, moods in the modern sense, although the correspondence is not complete. The indicative and infinitive are straightforward. The imperative includes both second and third person imperative forms in the modern sense and jussive subjunctives. The optative is what we would call a main-verb subjunctive with optative force; Donatus’s example is an imperfect subjunctive, “utinam legerem”. The “conjunctive” mood is the subjunctive used in subordinate clauses, although it is not explained this way: the example given is “cum legam.” The “impersonal” is also a mood: “legitur”. Formae uerborum, on the other hand, are aspectual categories: “perfecta, ut lego; meditatiua, ut lecturio; frequentatiua, ut lectito; inchoatiua, ut feruesco, calesco.” (Holtz, Donat, 591) In the A.M., Donatus explains the derivation of these formae more fully, and links some of the derived forms to the conjugations to which they belong. (A.M. II.12; Holtz, Donat, 633-634) Donatus recognizes three conjugations and classifies them by the vowels of their second person singulars. He conflates the third and fourth conjugations, with the length of the i distinguishing sub-types. When does the third conjugation have –am instead of –bo in its future? Donatus asks. When it has a short i. (Holtz, Donat, 591-92) In the A.M., Donatus admits that some authorities recognize a fourth conjugation. (A.M. II.12; Holtz, Donat, 634-635) 31 The genus of verbs is not gender, but rather encompasses notions of voice and transitivity. Donatus notes in the A.M. (II.12; Holtz, Donat, 635) that some call the genera “significationes”. These are actiuum, passiuum, neutrum, deponens, commune. The voices are characterized morphologically. Verba actiua end in o and can make passives out of the themselves by adding an r. Passives end in r and can make actives out of themselves by removal of the r. Neutra are intransitive active verbs: they end in o, but “accepta r littera latina non sunt, ut sto, curro,” “with an r added, they are not Latin, as ‘stand’, ‘run’.” (In the A.M., this list is expanded to include odi, noui, memini; sum and its compounds; and impersonals like pudet. A.M. II.12; Holtz, Donat, 635) Deponentia (verbs that are active in meaning but passive in form) are conceived of as the opposite of neutra, just as passiva are of activa: they end in r, like passives, “sed ea dempta latina non sunt, ut luctor, loquor,” “but with this (letter) taken away, they are not Latin, as ‘struggle’, ‘speak’.” Communia are those verbs that end in r, like deponents, but can be treated as either active or passive in meaning, like osculor, criminor: “dicimus enim osculor te et osculor a te, criminor te et criminor a te,” “for we say I kiss you and I am kissed by you, I accuse you and I am accused by you.” (Holtz, Donat, 592-93) Number, figure, and person of the verb are as we would expect. In the A.m., the persons of the verb are signaled only by form; in the A.M., Donatus explains what is referred to by the three “persons” and moves from there into a brief discussion of the syntax of cases: ...Prima est quae dicit lego; secunda, cui dicitur legis; tertia de qua dicitur legit. Et prima persona non eget casu, sed admittit plerumque nominatiuum, ut 32 uerberor innocens, liber seruio; secunda persona trahit casum uocatiuum, ut uerberaris innocens, liber seruis; tertia trahit nominatiuum, ut uerberatur innocens, liber seruit. Etiam uerba impersonalia, quae in tur exeunt, casui seruiunt ablatiuo, ut geritur a me a te ab illo. Quae in it exeunt, casui seruiunt datiuo, ut contigit mihi tibi illi. Quae uero in et exeunt, ea modo datiuo, modo accusatiuo casui seruiunt: datiuo, ut libet mihi tibi illi; accusatiuo, ut decet me te illum. Sunt uerba praeterea quorum alia genetiui casus formulam seruant, ut misereor, reminiscor; alia datiui, ut maledico, suadeo; alia accusatiui, ut accuso, inuoco; alia ablatiui, ut abscedo, auertor; alia septimi casus, ut fruor, potior. (A.M. II.12; Holtz, Donat, 638-639) The first person is the one that speaks, “I read”; the second, the one to whom one speaks, “you read”; the third the one about whom one speaks, “he reads”. And the first person does not lack case, but very often admits the nominative, as in, “innocent, I am beaten,” “free, I serve.” The second person takes the vocative, as in, “innocent, you are beaten,” “free, you serve.” The third person takes the nominative, as in “the innocent is beaten,” “the free man serves.” Also, impersonal verbs that end in –tur serve the ablative case, as in “it is borne by me by you by him.” Those that end in –it serve the dative, as in “it 33 happened to me to you to him.” But those that end in –et sometimes serve the dative, and sometimes the accusative: the dative, as in “it is allowed to me to you to him;” the accusative as in “it befits me you him.” There are, besides, verbs of which some preserve the rule of the genitive case, as in “pity”, “recollect”; others that of the dative, as in “curse”, “persuade”; others that of the accusative, as in “accuse”, “invoke”; others that of the ablative, as in “depart”, “turn away”; others that of the seventh case, as in “enjoy”, “possess”.60 As for the tenses, tempora, Donatus appears to make a distinction between the three real or essential tenses and the five that are marked by inflections. In the A.m., he says that there are three tenses, present, preterite (exemplified by a perfect form), and future, but five tempora in declinatione: “praesens (present), ut lego; praeteritum imperfectum (preterite imperfect), ut legebam; praeteritum perfectum (preterite perfect), ut legi; praeteritum plus quam perfectum (preterite pluperfect), ut legeram; futurum (future), ut legam.”61 (Holtz, Donat, 593) In the A.M. (II.12; Holtz, Donat, 637-638), he clarifies the distinction between tempora and tempora in declinatione. There are three differentiae of the temporis praeteriti: inperfecta, perfecta, and plusquamperfecta. 60 Fruor and potior take the ablative without preposition, which Donatus identified as a “seventh case” in A.M. II.9. 61 Ancient and medieval grammar did not distinguish the future perfect indicative from the perfect subjunctive. Priscian calls them both future subjunctive. Donatus recognizes the perfect subjunctive as such and calls the future perfect indicative a future subjunctive. 34 When Donatus gives sample declinationes of lego, the situation becomes more confusing. His verb categories seem to mix morphological and semantic criteria. He conjugates the indicative of lego in present, imperfect, perfect, pluperfect, and future. But he groups the imperative and jussive subjunctive under the heading “imperative”, giving precedence to semantics over morphology. The present of Donatus’s “imperative” is lege, legat, legamus, legite, legant; the future, legito, legito, legamus, legitote, legant and legunto. The “optative” organizes the tenses of the subjunctive according to their mainverb optative uses. The present and imperfect are “utinam legerem, legeres, legeret, legeremus, legeretis, legerent;” the perfect and pluperfect are legissem, legisses, legisset, legissemus, legissetis, legissent;” the future is “utinam legam, legas, legat, legamus, legatis, legant.” The forms of the “conjunctive” naturally overlap with those of the optative, but are organized differently, according to their use in subordinate clauses, and include the perfect subjunctive. The present is “cum legam legas legat legamus legatis legant;” the imperfect is “cum legerem legeres legeret legeremus legeretis legerent;” the perfect is “cum legerim legeris legerit legerimus legeritis legerint;” the pluperfect is “cum legissem legisses legisset legissemus legissetis legissent;” and the future (not recognized as a tense of the subjunctive in modern grammars) is “cum legero legeris legerit legerimus legeritis legerint.” 35 Infinitives and participles are more or less as we would expect to find them, including the distinction between active and passive participles (which have their own chapter later in the A.m.). “Gerendi uel participialia uerba” are the gerunds and supines. The A.M. offers no further help in clarifying the tenses of the subjunctive (“optative” and “conjunctive”) as set out in the A.m. Priscian, characteristically helpful on the morphology, explains how to form the various moods and tenses from known forms, and sets out the correspondences between the tenses of the optative and conjunctive. (GL 3:453) The Adverb The adverb is “pars orationis quae adiecta uerbo significationem eius explanat atque implet,” “a part of speech which, when added to a verb, clarifies and completes its meaning.” Its accidents are significatio, comparatio, figura. (A.m. 5; Holtz, Donat, p. 596) The bulk of the chapter on the adverb in A.m. consists of a list of its significationes. There are adverbia loci, temporis, numeri, negandi, affirmandi, demonstrandi, optandi, hortandi, ordinis, interrogandi, etc., etc., etc.. A similar and expanded list is given in A.M. II.13 (Holtz, Donat, 641-642). The significationes aduerbiorum are interesting from a syntactical point of view. On the one hand, Donatus’s category aduerbium includes words that can be clause markers or 36 connective particles, and an awareness of the semantics of these markers could help a reader navigate the logical relationships between successive clauses. Thus, for example, if we know that deinde is an aduerbium ordinis, an adverb of order, we know that it articulates a relationship of temporal sequence between the clause it appears in and what went before. On the other hand, Donatus’s grouping of a large number of morphologically disparate words into semantic categories lays the groundwork for an understanding of “adverbness” that would transcend the morphological criteria that dominate his discussion of the inflecting parts of speech. The functional emphasis of the definition, “pars orationis quae adiecta uerbo significationem eius explanat atque implet,” points in this direction. So does the inclusion among the adverbs of adverbial phrases like mecum, tecum, and of locative forms of the noun Roma (Romae, Roma, Romam, “at Rome”, “from Rome”, “to Rome”). The discussion of words whose classification as adverbs is doubtful shows Donatus confronting the problems caused by a part of speech that cannot be pigeonholed morphologically: Sunt multae dictiones dubiae inter aduerbium et nominum, ut falso; inter aduerbium et pronomen, ut qui; inter aduerbium et uerbum, ut pone; inter aduerbium et participium, ut profecto; inter aduerbium et coniunctionem, ut quando; inter aduerbium et praepositionem, ut propter; inter aduerbium et interiectionem, ut heu. Horum quaedam accentu discernimus, quaedam sensu. (A.M. II.13; Holtz, Donat, 643) 37 There are many words about which it is doubtful whether they are adverb or noun, like falso; adverb or pronoun, like qui; adverb or verb, like pone; adverb or participle, like profecto; adverb or conjunction, like quando; adverb or preposition, like propter; adverb or interjection, like heu. Some of these we distinguish by accent, some by sense. Donatus’s willingness to use sense as a criterion for classifying a part of speech is a departure from his preferred method. The Participle The participle is so called because it shares the accidents of the noun and the verb, taking gender and case from nouns, tense and meaning from verbs, and number and figure from both. Donatus’s treatment of the participle in the A.m. is largely unsurprising to the modern latinist, with the occasional peculiarity of classification following naturally from his treatment of the noun and the verb. It is worth noting that when Donatus says that the participle shares accidents of the noun and verb, he is speaking strictly within the morpho-semantic terms in which he describes those principal parts of speech. That is, he nowhere comments on the behavior of the participle as noun- (or adjective-) like or as verb-like. The failure to distinguish adjectives as a separate class probably impedes recognition of the adjectival function of participles. The fact that participles can take adverbial modifiers is not mentioned, although, given the definition of the adverb, this possibility may be implicitly recognized by Donatus’s inclusion of participial forms in 38 the chapter De verbo. Mention of participles taking adverbial modifiers larger than a single word, or having objects or complements, is, unsurprisingly, entirely absent. The Conjunction Marc Baratin has argued that it is from this humble part of speech that the germ of syntax sprouts in the Roman grammatical tradition.62 As it did with the adverb, Donatus’s ruthlessly morphological description of the parts of speech fails him when he approaches the conjunction. Conjunctions are indeclinable and, unlike the other indeclinable parts of speech, the adverb and the preposition, they do not neatly attach themselves to another part of speech. The conjunction cries out to be described in relation to larger syntactical patterns. Our hopes are not disappointed. In A.m. 7, De coniunctione, Donatus defines coniunctio as “pars orationis adnectans ordinansque sententiam,” “a part of speech joining and ordering the sentence.”63 Its accidents are potestas, figura, ordo. Figurae are the same as for other parts of speech, but potestas and ordo are new. The species of potestas, “power”, are semantic categories: copulatiuas (et, -que, at, atque, ac, ast); disiunctiuas (aut, ue, uel, ne, nec, neque); expletiuas (quidem, equidem, saltim, uidelicet, quamquam, quamuis, quoque, autem, porro, porro autem, tamen); causales (si, etsi, etiamsi, quidem, quando, quandoquidem, quin, quin etiam, quatenus, sin, seu, siue, nam, namque, ni, nisi, nisi si, si enim, etenim, ne, sed, interea, licet, 62 Marc Baratin, La Naissance de la Syntaxe à Rome . (Paris: Les Editions de Minuit, 1989), esp. at 19. 39 quamobrem, praesertim, item, itemque, ceterum, alioquin, praeterea); rationales (ita, itaque, enim, enimuero, quia, quapropter, quoniam, quoniam quidem, quippe, ergo, ideo, igitur, scilicet, propterea, idcirco). (Holtz, Donat, 599-600) It will be seen from this list that Donatus’s notion of the conjunction also includes various kinds of adverbs, sentence adverbs, and connective particles, and that he freely mixes subordinating conjunctions in with other forms in his lists. Donatus’s inventory of species potestatis of conjunctions is well suited to helping students master the kinds of logical connectors they will see within and across sentences, but there is nothing in Donatus’s schema which would help a student understand hypotaxis. Ordo coniunctionum consists in the fact that “aut praepositiuae coniunctiones sunt, ut ac, ast, aut subiunctiuae, ut que, autem, aut communes, ut et, igitur:” “conjunctions are either prepositive, as ac, ast, or subjunctive, as que, autem, or common, as et, itigur.” That is, the “order” of a conjunction is whether it is placed before or after other elements, or both. Donatus does not distinguish enclitic conjunctions that are “subjoined” to a word from postpostive particles that obligatorily come in second place in a clause. The ordo of a conjunction does not tell us anything about the order of the elements it joins. Knowing the ordo of a conjunction, though, could have the advantage of suggesting to a student where to look for a given conjunction when reading. 63 I will translate sententia as “sentence” for the time being, but it must be remembered that the modern and ancient terms are not always a perfect match. 40 The Preposition The preposition, like the adverb, is defined semantically in relation to other parts of speech: “pars orationis quae praeposita aliis partibus orationis significationem earum aut complet aut mutat aut minuit;” “a part of speech which when placed before other parts of speech either fills out or changes or restricts their meaning.” (A.m. 8; Holtz, Donat, 600) Since it is indeclinable and not subject to compounding, and its ordo is inherent in its name, its only accident is casus, case, that is, accusative or ablative. Donatus therefore devotes this chapter to listing prepositions that take the accusative, with an example of each with an appropriate noun; then he does the same for the ablative. Finally he gives prepositions that take either case (in, sub, super, subter), and gives rules of thumb for the usage of each. Donatus’s definition of the preposition also includes inseparable prefixes: “praepositiones...quae dictionibus seruiunt et separari non possunt,” “which serve words and cannot be separated. (p. 600) The Interjection The interjection is, alas, without syntactical import. Isidore on the parts of speech Isidore’s treatment of the parts of speech in the Etymologies is quite brief. But because of his desire to explain rather than simply to categorize, his brief treatments are 41 often more helpful than those of the grammarians stricto sensu for understanding how the partes are actually used. Isidore is not bound by the strict format Donatus imposed on himself, and, as an encyclopedist, he is transmitting some rhetorical and philosophical views of language along with the grammatical. Thus, although he devotes separate books to the separate disciplines, he allows his reading in other fields to influence his presentation of grammatical material in Book I.64 In his introduction to the parts of speech, for example, he signals the primacy of the noun and the verb, as signifying person and action. This comes closer than anything in the basic grammars to specifying the essential elements of the clause, even if Isidore does not explain person and action in precisely those terms: De partibus orationis. Partes orationis primus Aristoteles duas tradidit, nomen et verbum; deinde Donatus octo definivit. Sed omnes ad illa duo principalia revertuntur, id est, ad nomen et verbum, quae significant personam et actum. Reliquiae adpendices sunt et ex his originem trahunt. Nam pronomen ex nomine nascitur, cuius officium fungitur, ut “orator ille”. Adverbium de nomine nascitur, ut “doctus docte”. Participium de nomine et verbo, ut “lego, legens”. Coniunctio vero et praepositio vel interiectio in complexu istarum cadunt. Ideo et nonnulli quinque partes difinierunt, quia istae superfluae sunt. (Etym. I.vi) 64 On Isidore’s sources and methods, see Jacques Fontaine, Isidore de Séville et la culture classique dans l'espagne wisigothique . (Paris: Études Augustiniennes, 1959). 42 On the parts of speech. Aristotle first transmitted two parts of speech, the noun and the verb; afterwards Donatus defined eight. But all the parts go back to those two, that is, to the noun and the verb, which signify person and action. The rest are appendages and take their origin from these. For the pronoun is born from the noun, whose office it fulfills, as in “the orator, he”. The adverb is born from the noun, as in “learned, learnedly”. The participle is born from the noun and the verb, as in “read, reading”. But the conjunction and the preposition and the interjection are suited to the connection of these others.65 Therefore some also define five parts of speech, since these last are superfluous. The chapter on the noun (Etym. I.vii) is mainly taken up with classifying nouns semantically, but the short section on the cases comes closer than anything in Donatus to explaining what the cases are actually used for. The explanations are etymological: Nominativus casus dictus quia per eum aliquid nominamus, ut “hic magister”. Genitivus, quia per eum genus cuiuscumque quaerimus, ut “huius magistri filius”, vel quod rem significamus, ut “huius magistri liber”. Dativus, quia per eum nos dare aliquid demonstramus, ut “da huic magistro”. Accusativus, quia per eum aliquem accusamus, ut “accuso hunc magistrum.” Vocativus, quia per eum aliquem vocamus, ut “o magister.” Ablativus, quia per eum nos auferre aliquid alicui significamus, ut “aufer a magistro”. (Etym. I.vii.31-32) 65 Complexus is used by Quintilian to mean “connection in discourse”, and I take it that is Isidore’s meaning here; Lewis & Short, s.v. 43 The nominative case is so called because through it we name something, as in “the master”. The genitive, because through it we seek the origin of each person, as in “the master’s son,” or because we signify a thing, as in “the master’s book”. The dative, because through it we show that we are giving something, as in “give to the master”. The accusative, because through it we accuse someone, as in “I accuse the master”. The vocative, because through it we call someone, as in “O master!” The ablative, because through it we signify that we are taking something away from someone, as in “take from the master”. Isidore’s illustration of pronominal reference in the passage quoted earlier (“orator ille”, Etym. I.vi.2) is, likewise, an improvement on Donatus’s very abstract presentation of what it means that a pronoun “pro nomine posita tantundem paene significat.” His treatment of the pronoun in Etym. I.viii gives further helpful examples. Here Isidore treats the use of the pronoun rhetorically, suggesting that the pronoun is used for variety’s sake. Although he does not explain the mechanics of reference, this type of example would help students to recognize the recurrence across several clauses of a subject already named: Pronomen dictum, quia pro vice nominis ponitur, ne fastidium faciat nomen ipsud dum iteratur. Nam cum dicimus, “Vergilius scripsit Bucolica,” addimus pronomen, “ipse scripsit Georgica;” sicque varietas significationis et fastidium tollit et ornatum inducit. (Etym. I.viii.1) 44 The pronoun is so called because it is used in place of the noun, lest the noun itself produce weariness when it is repeated. For when we say, “Vergil wrote the Eclogues,” we add a pronoun, “he himself wrote the Georgics;” and thus the variety of signification both relieves boredom and introduces ornament. Isidore’s treatment of the verb is outstanding for relating to one another the ideas of person and voice (genus) as found in Donatus. The result is a model for the rudimentary analysis of transitivity: ...Sicut autem nomen significat personam, ita verbum factum dictumque personae. In persona verbi agentis et patientis significatio est. Nam “scribo” personae factum est. Item “scribor” personae factum indicat, sed eius a quo patitur. (Etym. I.ix.1) Moreover, just as a noun signifies person, even so a verb signifies a person’s deed or saying. In the person of the verb is the signification of one who acts and the one who undergoes. For “I write” is the deed of a person. “I am written” also indicates the deed of a person, but of the one by whom it is undergone. Later in his discussion of the verb, Isidore tries to explain the meaning of the moods, rather than just listing their forms, and links the “conjunctive” to the idea of the complete sentence. He has just finished explaining the formae (that is, the aspectual 45 suffixes), and makes the rather opaque comment, “Formae enim sensum tenent, modi declinationem. Nam nescis quid sit declinatio, nisi prius didiceris quid sit sensus.” (Etym. I.ix.3) “The forms have meaning, and the moods have inflection. For you do not know what the inflection is, unless you have first learned what the sense is.” This could be taken in a purely morphological sense: the derivational, aspectual endings (frequentative, inchoative, etc.), yield base forms which must in turn be inflected for mood (as well as person, number, and tense.) Thus you don’t know how to conjugate the verb until you know what its derivational status is. Isidore’s following comments, though, suggest that he is interested in what the moods mean: Modi dicti ab eo quemadmodum sint in suis significationibus. Indicativus enim modus dicitur quia significationem habet indicantis, ut “lego”. Imperativus, quia sonum habet imperantis, ut “lege”. Optativus, quia per ipsum aliquid agere optamus, ut “utinam legerem”. (Etym. I.ix.4) Moods are so called from the mode in which they have meaning. For the indicative mood is so called because it has the sense of someone indicating, as in “I read.” The imperative, because it has the sound of someone ordering, as in, “Read!” The optative, because through it we wish to do something, as in, “Would that I were reading!” The conjunctive mood, however, as its name suggests, takes its meaning not from the point of view of the speaker, but from its status in relation to a complete sentence: 46 Coniunctivus, quia ei coniungitur aliquid, ut locutio plena sit. Nam quando dicis “cum clamem,” pendet sensus; quod si dicam, “cum clamem, quare putas quod taceam?” plenus sit sensus. (Etym. I.ix.4) The conjunctive (is so called) because something is joined to it, so that the expression may be complete. For when you say “since I am shouting,” the sense is left hanging; but if I were to say, “since I’m shouting, why do you think that I’m being silent?” the sense is complete. The cum clause in this example is not characterized as a clause, or as subordinate (just as there was no understanding of hypotaxis in Donatus’s classification of pronouns, adverbs, and conjunctions). The dependence of the part of the expression containing the conjunctive verb is seen as a problem of incomplete sense. Incompleteness is also a way of understanding the infinitive and impersonal verbs, which lack “person” in the conventional sense. Isidore approaches the idea of the complementary infinitive when he explains how to add “person” to a non-finite form: Infinitus modus dicitur eo quod tempora definiens personam verbi non definit, ut “clamare”, “clamasse”. Cui si adiungas personam: “clamare debeo, debes, debet,” fit quasi finitum. (Etym. I.ix.5) 47 The infinitive mood is so called because, while it specifies tense, it does not specify the person of the verb, as in “to shout”, “to have shouted”. If you add person to it – “I, you, he ought to shout” – it becomes as it were finite. Similarly, in explaining how “person” must be supplied for impersonal verbs, Isidore models the agent construction used with such forms: Inpersonalis dicitur quia indiget personam nominis vel pronominis, ut “legitur”: addes personam “a me”, “a te”, “ab illo”, et plene sentitur. (Etym. I.ix.5) The impersonal is so called because it wants person in the form of a noun or pronoun, as in “it is read”: you add person, “by me”, “by you”, “by him”, and the sense is complete. Isidore clarifies the difference between the infinitive and the impersonal this way: “Sed infinitus modus personam tantum verbi eget; impersonalis vero vel pronominis personam vel nominis.” “The infinitive mood lacks person only in the verb; but the impersonal needs person either in a pronoun or in a noun.” That is, the infinitive needs a complementary verb form, whereas the impersonal construction requires an agent phrase. In both cases, but more explicitly in the latter, the lack of person is seen as a deficiency in the sense of the verb. 48 This understanding of completeness in the verb recurs in an interesting way in Isidore’s explanation of the adverb: Adverbium dictum est eo quod verbis accedat, ut puta, “bene lege.” “Bene” adverbium est, “lege” verbum. Inde ergo dictum adverbium, quod semper verbo iunctum adimpleatur. Verbum enim solum sensum implet, ut “scribo.” Adverbium autem sine verbo non habet plenam significationem, ut “hodie”. Adicis illi verbum, “hodie scribo,” et iuncto verbo implesti sensum. (Etym. I.x) The adverb is so called because it attends the verb, as for example, “Read well!” “Well” is an adverb and “read” a verb. From this, therefore, it is called an adverb, because it is always fulfilled when joined to a verb. For the verb alone completes the sense, as in “I write.” The adverb, however, does not have full significance without the verb, as in “today”. Add a verb to it, “Today I write,” and, with the addition of the verb, you have completed the sense. This passage expands the notion of completeness as presented in the discussion of the conjunctive, and also implicitly picks up on the Aristotelian doctrine of the noun and verb being essential parts of speech. The notion of completeness can describe the moment in sequential processing at which a sentence is felt to be “fulfilled” enough to stand on its own, and also relates to the question of which parts of speech can be understood only as they function with other parts. As noted above, the conjunction is the ultimate example of the pars that can only be described in terms of its operation on other partes: 49 Coniunctio dicta quod sensus sententiasque coniungat. Haec enim per se nihil valet, sed in copulatione sermonum quasi quoddam exhibet glutinum. Aut enim nomina sociat, ut “Augustinus et Hieronymus”; aut verba, ut “scribit et legit”. Una autem vis omnium, sive copulent, sive disiungant. Copulativae autem coniunctiones dictae eo quod sensum vel personas coniungant, ut “ego et tu eamus ad forum.” Disiunctivae dictae quia disiungunt res aut personas, ut “ego aut tu faciamus”. Expletivae dictae quia explent propositam rem, ut puta, “si hoc non vis, saltim illud fac.”...Causales dicuntur a causa eo quod aliquid cogitent facere, ut puta, ”occido illum, quia habet aurum;” causa est. (Etym. I.xii.1-4) The conjunction is so called because it joins senses and sentences. For it has no power in itself, but in joining discourse it acts like a kind of glue.66 For it either joins nouns, as in “Augustine and Jerome”, or verbs, as in “he writes and reads”. All of them have a single force, whether they join or separate. Copulative conjunctions are so called because they join senses or persons, as in “let’s go to the Forum, you and I.” Disjunctive conjunctions are so called because they separate things or persons, as in, “let’s do it, you or I.” ...Expletive conjunctions are so called because they fill out the proposed matter, as for example, “if you don’t want this, at least do that.”67 ...Causal conjunctions are so called from the cause by reason of which they might contemplate doing something, as for example, “I kill him because he has gold;” it’s the cause. 66 Isidore’s glue metaphor is one of his most often repeated grammatical tidbits, and it will appear in expanded form in Ælfric’s Grammar . See Part III below. 50 Whereas Donatus equated the potestas of the conjunction with its semantic subclasses, Isidore distinguishes the semantics of the specific conjunction from the general force of this part of speech, the una vis omnium, which is to bring discourse together. The nature of the binding appears to be a binding in pairs and in sequence, and the units bound may be of various sizes: individual parts of speech or whole clauses. The notion of binding through hypotaxis is still absent. Summary: Potential for a syntactical understanding of the parts of speech Two overlapping categories of information emerge from these treatments of the partes that might help a student learn to negotiate continuous Latin texts. The first consists of morphological and/or semantic features of the partes that come into play when words enter into combination with one another. These features emerge from the description of the accidents of the parts of speech, are unambiguously syntactical in their import by modern standards, and include some areas of linguistic description that are quite well developed in the ancient grammars. The second category is harder to pin down, but provides a necessary complement to the first: these are the recurring modes of description that a student would encounter at various stages of the grammatical curriculum and that would begin to form a supplementary set of tools and descriptors for conceptualizing and demarcating elements of Latin text. In the first category, the best-developed points of doctrine are: 67 Saltim is being identified as the expletive conjunction. 51 Syntax of cases Although the uses of the cases are introduced haphazardly in Donatus, who emphasizes rarer case uses, Donatus, Priscian, and, especially, Isidore provide enough material for the student to work out the basic functions of the cases. Interestingly, the grammars offer more practical information on case function in the sections on the verb than in the sections on the noun and pronoun. The terminology of case rection is fairly well developed, as Vivien Law has noted:68 for instance, the comparative “serves” (servit) the ablative; prepositions serviunt the words they accompany; certain adjectives “take” (trahunt) the genitive, the dative, and so on. Verbal person The meaning of the persons of the verb is clearly explained in the Ars maior. Verbal voice and transitivity In the account of the verbal accident genus, these late-antique texts contain the germ of the later medieval obsession with transitivity. More practically for Latin students, the well-developed account of the relationship of verbal voice to the cases provides a basis for identifying the essential elements of the clause. 68 Vivien Law, “The Terminology of Medieval Latin Grammar,” Grammar and Grammarians in the Early Middle Ages , (London and New York: Longman, 1997) 260-269, at 267. 52 Isidore fulfills this potential by relating the Aristotelian doctrine of substance and action to the noun and the verb. Verbal mood and tense The division of what we call the subjunctive into two moods, the optative and the conjunctive, highlights the distinction between independent and subordinate uses of the subjunctive, although the distinction independent/dependent is nowhere present in this grammatical system. The forms and tenses proper to each mood are exemplified in a way that demonstrates practically, if not comprehensively, several of the uses of the subjunctive. Pronominal reference The relationship of pronouns to nouns is in several places exemplified and is characterized in terms of rhetorical variation. Ordo Several parts of speech are described in terms of their ordo. This always means either the relative order of pairs of words, or the placement of the pars under discussion before or after other elements Clause markers and discourse particles Various sentence adverbs, connective particles, coordinating and subordinating conjunctions, and some pronouns are classified according to their 53 semantic function. This yields a system in which the logical connectors between successive clauses could be well understood, even if the distinction between hypotaxis and parataxis was not. Summary: Habits of thought and modes of description Several broad tendencies in the way linguistic patterns are conceived recur in the description of the parts of speech. These are: Binary relationships The descriptive system is particularly strong in its account of relationships between pairs of terms: prepositions and their objects; verbs and nouns that attract unusual oblique cases – that is, the relationships signaled by the terms servire and trahere. The restriction of ordo to mutual ordering of two elements also falls under this heading. The interest in the status of words as being compounded of two elements is a related phenomenon. Compounding This interest in compounding – in what the constituents of a word are and what happens at the point where they join – is not developed in a syntactical way in the grammars. It is, however, perhaps the best-developed part of the descriptive system, and can be associated with the interest in the mutual ordering of pairs of 54 constituents described above. The term used for compound status is figura, a word that recurs throughout the grammatical curriculum in a great variety of uses. The uses of figura have in common a sense of the patterning of constituent elements.69 Completeness The test whether an expression can stand alone semantically is repeatedly invoked, to test the status of separable and inseparable morphemes, and to test the syntactical sufficiency of certain parts of speech and the dependency of others. The idea of completeness in many ways substitutes for the idea of independence and dependence in this system. Latinum and non latinum In addition to being diagnosed as complete or incomplete, expressions can be described as latinum or non latinum, “Latin” or “not Latin”. This is not a diagnosis of grammatical “correctness”, in the sense of preferable usage, but of something more like the Chomskian notion of “grammaticality”: that is, wellformedness according to the internal rules of the language. In the grammars, this distinction is invoked to distinguish forms that cannot appear (such as active 69 See for instance Aldhelm’s use of figura ’s Greek equivalent, schema , to describe the arrangement of the constituent elements of a metrical line (Part I.1, below), and Bede’s distinction between figures ( schemata ) and tropes in the De schematibus et tropis (Part II.1, pp. 119-120, below). 55 forms of deponent forms), but the distinction is also suggestive in its potential for classifying larger expressions as idiomatic or inadmissible. 70 70 The concept of grammaticality is introduced in Noam Chomsky’s Syntactic Structures (The Hague: Mouton & Co., 1964). Chomsky’s framework, of course, is what he would come to call “generative” grammar, and his definition of grammaticality must be considered in light of his definition of syntax. The opening of Syntactic Structures sets the parameters: “Syntax is the study of the principles and processes by which sentences are constructed in various languages. Syntactic investigation of a given language has as its goal the constuction of a grammar that can be viewed as a device of some sort for producing the sentences of the language under analysis.” ( Syntacic Structures , 11; my emphasis) For our purposes, it is crucial to note 1) that Chomsky’s unit of analysis is the sentence; and 2) that for him, a grammar is a device, a machine, theoretically capable of producing the infinite number of possible, “correct” sentences that a native speaker of a language could theoretically produce, and no impossible sentences. Chomsky employs “grammaticality” as the test that separates sentences that are not part of a given language from sentences that are: “The fundamental aim in the linguistic analysis of a language L is to separate the grammatical sequences which are the sentences of L from the ungrammatical sequences which are not sentences of L...One way to test the adequacy of a grammar proposed for L is to determine whether or not the sequences that it generates are actually grammatical, i.e., acceptable to a native speaker, etc.” ( Syntactic Structures , 13; author’s emphasis) This understanding of “grammar” and of grammaticality is at once in some degree independent of the data to be derived from an accumulated corpus of examples of a language that already exists, and dependent on the availability of native speakers against whose linguistic instincts the output of the grammar may be tested. Neither circumstances is friendly to the modern or medieval grammarian working with earlier forms of the Latin language. Indeed, Chomsky notes ( Syntactic Structures , p. 14, n. 1) a tension between the desire to construct a grammar for the infinitely possible set of sentences in a language and the likelihood of having to construct the grammar from a limited sample of such sentences: “Notice that to meet the aims of a grammar, given a linguistic theory, it is sufficient to have a partial knowledge of the sentences (i.e., a corpus) of the language, since a linguistic theory will state the relation between the set of observed sentences and the set of grammatical sentences.” Nonetheless, Chomsky’s method substantially excludes working from a corpus (p. 15) and depends almost exclusively on his own intuition into English. In the early Middle Ages as now, the grammarian of Latin is dependent on a written corpus, of mixed native- and non-native-speaker provenance, and lacks access to native speakers. For a discussion of the application of Chomskyan principles to Latin, and to historical languages in general, see Robin T. Lakoff, Abstract Syntax and Latin Complementation . (Cambridge, MA: M.I.T. Press, 1968). All these reservations notwithstanding, the term “grammaticality” is an appealing rubric for a fruitful area of medieval grammatical thought. The problem of defining the difference between possible and impossible 56 What’s missing? Syntax in modern teaching grammars We can compare these two checklists to subjects treated under the heading “syntax” in representative modern teaching grammars of Latin. The textbook from which I learned Latin, Jenney’s First Year Latin,71 weaves the essential case uses and extensive discussion of agreement (subject-verb and noun-adjective) into the earliest chapters. New uses of the ablative are introduced in each unit. A series of chapters on the forms and uses of the infinitive is capped with a chapter on indirect statement. (The term “indirect” introduced in this chapter is picked up later in the chapters on indirect commands and indirect questions, and the three types of indirect construction are compared formally in a review chapter after they have all been covered individually.) When the forms of the subjunctive have been covered, clause types using the subjunctive are introduced, a chapter at a time: clauses of purpose, indirect commands, clauses after verbs of fearing, indirect questions, result clauses, cum clauses. These chapters also introduce sequence of tenses. Late in the book come the ablative absolute, certain less common case uses (dative of possession), impersonal constructions, gerunds and gerundives, the passive periphrastic, and the supine. (Conditions are introduced early in second year.) There is no lesson devoted to the concept of subordination, but subordinate clauses are so designated, and subordinating conjunctions are learned along with the clause types they introduced. These clauses are characterized by internal form and semantic function. Great emphasis utterances and between more and less correct ones naturally concerned medieval grammarians, and fine judgements on these and related points had been made by native Latin speakers in the ancient grammars. 71 Charles Jenney, Jr., Rogers V. Scudder, and Eric C. Baade, First Year Latin . (Boston: Allyn and Bacon, 1979). 57 is placed on recognizing what kind of verbs in the main clause would lead one to expect (or produce) what type of subordinate clause, and composition drills emphasize this relationship. The text from which I last taught beginning Latin, Latin For Reading,72 incorporates functional syntax and discourse approaches into a grammar aimed primarily at teaching Latin reading skills, rather than composition skills. This approach privileges syntactical concepts that have a high pay-off for the inexperienced reader: the first lessons introduce the complete Latin sentence, the implications of inflection for Latin word order, and the rules for ellipsis (“gapping”) in Latin. All clauses, independent and dependent, are described in terms of their essential elements (“kernel” elements) and are grouped in a manageable number of types according to the voice and transitivity of the verbs they contain. All clauses are further classified as noun clauses, adjective clauses, or adverb clauses, a functional approach that builds on students’ knowledge of the functions of the parts of speech. Woodcock’s New Latin Syntax,73 an historically-minded handbook of syntax for students,74 begins with surveys of the uses of the oblique cases, including a chapter on 72 Glenn M. Knudsvig, Gerda M. Seligson, and Ruth S. Craig, Latin For Reading: A Beginner's Textbook with Exercises . (Ann Arbor: University of Michigan Press, 1986). 73 E.C. Woodcock, A New Latin Syntax . (London: Methuen, 1959). 74 Despite the blurbs that now appear on the cover of Woodcock recommending his textbook for A-level or University students, the terms in which he justifies his project suggest that Woodcock had in mind in the first instance a younger school audience: “ Latin syntax must seem to the average student a collection of irrational peculiarities which can be mastered only by memorizing disconnected rules based on statistics. 58 the accusative and infinitive that follows that on the accusative. The remainder of the chapters treat the uses of the participle; independent uses of the subjunctive and semantic equivalents to these uses; subordinate clauses of all types; and finally special issues relating to subordination (reported speech, subordinate and conditional clauses in indirect discourse, and so forth). There are clear continuities between the ancient and medieval school grammars and their modern counterparts, as one would expect in an extremely conservative tradition. The parts of speech still provide the framework for the rest of Latin grammar; they are, if anything, more important in the functional-syntax approach of the selfconsciously new method of Latin For Reading. Latin morphology still stands out as the most obvious obstacle for Latin-learners, and much basic syntactical teaching is built around the morphological categories: case uses are introduced early and often, and, in the more conservative approaches, the acquisition of the forms of the subjunctive governs the learning of clause types in which those forms are employed. The identification of essential clause elements has a basis in ancient grammar, as does the interest in completeness as a conceptual tool in navigating Latin sentences. The striking difference, though, between the ancient and modern textbooks is the emphasis in the latter on the form and function of subordinate clauses. The distinction main/subordinate and the function of clause markers in articulating that distinction occupies the bulk of the textbook real estate in the modern grammars. This distinction is, as we have seen, But in most Grammar Schools nowadays Latin is not begun until the age of twelve, by which time the faculty of memorizing is beginning to wane. The faculty of reason is beginning to take its place, and therefore a thread of reason ought to be supplied in the presentation of Latin syntax.” (Woodcock xv) 59 completely lacking in the late-antique texts we examined. The modern student is taught to read and to construct elaborately-nested, multiply-subordinated clauses, and is explicitly taught to manipulate verb forms according to the level of subordination of the clause in which the verb appears. The student of Donatus, Priscian,75 and Isidore, by contrast, was apparently encouraged to track the logical connectors between clauses without thinking of them in terms of hypotaxis. Equally striking is the complete absence of any mention of the non-finite constructions most frequent in Latin, most peculiar to it, and most troubling to students: participial constructions, the ablative absolute, and the accusative and infinitive. These constructions are particularly frustrating to students learning Latin because they can contain a significant amount of further-subordinated material, but, not being marked by subordinating conjunctions and finite verbs, their boundaries are difficult to recognize without training and practice. These features – the lack of a discrete clause marker, especially, and of a finite verb – are also what make them liable to exclusion from the partes description of the early grammars. They are not susceptible in the usual way to marking by any of the conjunctions or “adverbs” whose semantic forces are catalogued by Donatus; on the contrary, they are distinguished by an absence of an explicit, oneword tag to characterize the logical relationship between the subordinate and the governing clause. They cannot be recognized by the presense of the “conjunctive” mood, which would have been learned with the tag cum to signal its function. 75 Institutio de nomine pronomine et verbo only, of course. 60 There is a further bar to inclusion of the syntax of the participle in the Donatan system. (Recall that although the participle is recognized as a separate part of speech sharing accidents of the noun and the verb, its function is never broached.) The failure to recognize the adjective as a separate part of speech precludes the possibility of describing the adjectival function of the participle.76 Indeed, the problem of the participle points up the main strength and the main weakness of the partes-governed system. When one of the classically-defined partes has clear syntactical implications, as, especially, the verb, the adverb, and the conjunction, the description of that part is liable to yield syntactical information, if only accidentally. But when a crucial syntactical category is not reified, as it were, as one of the canonical parts of speech, that category is liable to exclusion from the whole system of analysis. This is the case with the adjective. Many ancient grammars describe a subcategory of the noun, the nomen adiectivum (in Priscian) or nomen appellativum (in Donatus). But because the nomen appellativum is not presented as a separate pars that can interact in a binary relationship with the nomen, the rich mine of information in the grammars on the noun’s inflectional system is never associated with the principle of gender-, number-, and case-concord between a noun and its modifier. The limited information that does emerge on pronominal reference, if only by repeated exemplification, shows how the adjective might have been treated if it, too, had the status of a separate part of speech. 76 Priscian and others recognize the nomen adiectivum as a sub-class of nouns, but without granting it a separate status whose syntactical behavior can be described as distinct from that of other nomina . 61 It is tempting to see a relationship between the basic grammars’ silence on the subject of concord and such striking gaps as the failure to describe the relative clause. As we will see, however, the relationship of noun and modifier is a frequent topic for elucidation both in the paragrammatical treatises and in syntactical glossing. The status of adjectives and “adjectiveness” is an interesting test-case for the ways in which the hidden curriculum compensates for the weaknesses of the grammars. Larger units of language: Ars maior III and Isidore’s Etymologiae I However dominant the partes framework, the grammars do offer the students some help beyond the level of the single pars. Donatus and Isidore address themselves in several places to larger units of language than the part of speech. Ars maior III and Etym. I.xxxii-xxxvii. treat the vitia (faults) of discourse and rhetorical schemes and tropes. Isidore’s chapter De grammatica (Etym. I.v) includes a definition of oratio. Donatus has a short chapter on punctuation, Ars maior I.6, and Isidore expands on this in his chapter De posituris (Etym. I.xx). A.M. III deals with barbarisms, solecisms, “cetera vitia” (other faults), and metaplasms, and then with schemes and tropes. Isidore also treats this material in Etym. I.xxxii-xxxvii. The sections in both authors devoted to schemes and tropes contain practical applications of much of the material found in the description of the partes discussed above. Bede develops this material significantly in his De schematibus et tropis, and I discuss this doctrine from a syntactical point of view at length in Part II, 62 below. For the moment, it is worth looking at how the similarities and differences between smaller and larger linguistic units are expressed in the definitions of linguistic “faults”. A barbarism, Donatus says, is una pars orationis uitiosa in communi sermone. In poemate metaplasmus, itemque in nostra loquella barbarismus, in peregrina barbarolexis dicitur. (A.M. III.1; Holtz, Donat, 653) one faulty part of speech in common discourse. In verse it is called a metaplasm, in our tongue a barbarism, and in a foreign language a barbarolexis. A solecism, on the other hand, is an analogous fault of partes in combination: Solecismus est uitium in contextu partium orationis contra regulam artis grammaticae factum. Inter solecismum et barbarismum hoc interest, quod solecismus discrepantes aut inconsequentes in se dictiones habet, barbarismus autem in singulis uerbis fit scriptis uel pronuntiatis. Quamquam multi errant, qui putant etiam in una parte orationis fieri solecismum, si aut demonstrantes uirum hanc dicamus, aut feminam hunc; aut interrogati quo pergamus, respondeamus Romae; aut unum resalutantes saluete dicamus, cum utique praecedens 63 demonstratio uel interrogatio uel salutatio uim contextae orationis obtineat. (A.M. III.2; Holtz, Donat, 655) A solecism is a fault made against the rule of the grammatical art in the connection of the parts of speech. The difference between a solecism and a barbarism is this, that a solecism has in it parts of speech that disagree or that do not logically follow, whereas a barbarism occurs in the writing or pronunciation of individual words. Nevertheless, many err who think that a solecism can also occur in a single part of speech, if, pointing out a man, we were to say “her”, or a woman, “him”; or, when asked where we are going, we were to respond, “at Rome”; or, greeting one person, we were to say, “Hey, y’all!” although certainly the foregoing pointing out or query or greeting takes its force from connected discourse. Donatus’s insistence that a mistake of gender or case or number is a barbarism is in keeping with the priority he gives elsewhere to single words and their forms: if a problem can be treated morphologically, it is, and to call the errors with hanc and hunc, Romae, and saluete barbarisms is to insist on seeing them morphologically. Similarly, when Donatus catalogues errors that he does recognize as solecisms, he classifies them as offences involving the parts of speech and each of their accidents: “Per accidentia partibus orationis tot modis fiunt solecismi, quot sunt accidentia partibus orationis.” (A.M. III.2; Holtz, Donat, 656) “As many solecisms occur through the accidents of the parts of speech as there are accidents of the parts of speech.” That is, although solecisms 64 are by definition errors of words in combination, Donatus marshals them for description strictly along the lines laid down in A.m. and A.M. II. His desire to reduce every phenomenon to the grammatical description of the parts of speech is relentless. That said, Donatus’s acknowledgement of the reasoning of those who err and want to call single-word mistakes solecisms suggests that he is aware of the fuzzy boundaries between morphology per se and morphology as an aspect of textual coherence. The rationale of the errantes as quoted above is that this kind of form “uim contextae orationis obtineat,” “takes its force from connected discourse.” This would imply that these are errors of concord or case rection, stricto sensu: that is, a referent for the pronoun has been named and the gender of the pronoun must be chosen accordingly, or a verb has been uttered (pergamus) which demands a particular case of the noun Roma. The example of returning a greeting using the second person plural to a single person, however, suggests that by contexta oratio Donatus might equally mean “ongoing” or “coherent discourse or conversation”. The other two examples could be interpreted in this pragmatic way, too: Donatus sets up the scenes of the (alleged) solecisms as little dialogues. In other words, solecisms operate at levels larger than the single word, but the word (pars or verbum) is the only linguistic unit in which Donatus has a real investment. Larger units are not clearly defined except in strict contrast to the single word. These larger units may be pairs of words with a mutual grammatical relationship but that may or may not occur in the same sentence. The larger unit may be a continuous spoken or written text that demands logical continuity. Both ideas are contained in the expression 65 “discrepantes aut inconsequentes in se dictiones,” “words disagreeing among themselves or not following properly.” For an explanation of the kind of units in which coherence might operate, we can turn again to Isidore. His chapter De grammatica contains a definition of oratio. First he gives an etymology (“oratio dicta quasi oris ratio,” “the mouth’s reason”), and then he says, Est autem oratio contextus verborum cum sensu. Contextus autem sine sensu non est oratio, quia non est oris ratio. Oratio autem plena est sensu, voce, et littera. (Etym. I.v.3-4) Oratio is, moreover, a connection of words with sense. A connection without sense is not oratio, because it is not the mouth’s reason. Rather, oratio is full of sense, voice, and letter. That is, oratio is made up of the elements of spoken and written language, and it must be a meaningful sequence of those elements. The emphasis on sensus as necessary to oratio is reminiscent of Isidore’s account of expressions that are complete or incomplete as to 66 sense. The same notion also occurs in Isidore’s discussion (largely following Donatus) of the punctuation of the parts of a periodic sentence:77 Positura est figura ad distinguendos sensus per cola et commata et periodos, quae dum ordine suo adponitur, sensum nobis lectionis ostendit...Prima positura subdistinctio dicitur; eadem et comma. Media distinctio sequens est; ipsa et cola. Ultima distinctio, quae totam sententiam cludit, ipsa et periodus. Ubi enim initio pronuntiationis necdum plena pars sensui est, et tamen respirare oportet, fit comma, id est particula sensus...Ubi autem in sequentibus iam sententia sensum praestat, sed adhuc aliquid superest de sententiae plenitudine, fit cola...Ubi uero iam per gradus pronuntiando plenam sententiae clausulam facimus, fit periodus. (Etym. I.xx.1-5)78 Punctuation is a mark for distinguishing senses by cola and commata and periods which, when they are used in their own order, show us the sense of the reading...The first punctuation mark is called a subdistinctio, which is the same thing as a comma. The media distinctio comes next, and is the same as a colon. The ultima distinctio, which concludes the whole sentence, is the same as a period. For when, at the beginning of the pronouncement, the sense is not yet full, 77 For a more complete discussion of the relationship of punctuation to the ancient theory of the period, see Malcolm B. Parkes, Pause and Effect : An Introduction to the History of Punctuation in the West . (Aldershot: Scholar Press, 1992) 65-70. 78 Isidore adds that this is the usage “apud oratores”; among the poets, a single verse is equivalent to a period, and the comma and colon are different subdivisions of the verse. (Etym. I.xx.6) 67 and nevertheless one needs to breathe, there is a comma, that is a small element of the sense...Then again, when, in what follows, the sentence has already provided the full sense but there is something remaining of the fullness of the sentence, you have a colon...But when at last by pronouncing in stages we make a full ending to the sentence, there is a period. This description is framed in terms of reading a text aloud, but since the punctuation of the text would have been part of the preparation for such textual performance, we can take Isidore’s description as indicative of the way readers were expected to track larger and larger units of meaning across the sentence.79 It is interesting that the reader is being asked to imagine moving sequentially through the text, alert to the moment when completeness of sense has arrived, and comparing that completeness of sense to how much is left of the unit described as sententia. Clearly “sense” and “sentence” are not the same thing. Isidore does not explain what he means by sententia.80 However, if we compare his use of sensus here with his use of the term in the examples of the sufficiency of certain parts of speech, we can see the germ of a pedagogical technique. We can picture a teacher asking a student, “What do you have? Is the sense complete? All right, keep going and let me know when the sense is complete.” This kind of linear tracking would encourage the reader to process the text sequentially, thinking about it semantically and formally at the same time. This technique would be crucial for texts presented orally to the student, whether or not the teacher was working from a written 79 A.K. Gavrilov discusses evidence for the relationship of textual analysis to oral performance of a text in his article “Techniques of Reading in Classical Antiquity,” The Classical Quarterly N.S. 47 (1997): 56-73. 68 text. There is no suggestion here that the reader is encouraged to go “verb hunting”: evidence for that approach will emerge in glossed manuscripts.81 Conclusion With this brief survey, we can see ways in which a mainly word-based system can be made to yield syntactical insights, and other ways in which it cries out for supplementation. The paragrammatical texts of Aldhelm and Bede offer expansions of and alternatives to the modes of description found in the basic grammatical texts, and in the following chapters we will consider these in roughly chronological order. Aldhelm’s language treatises, the earliest that survive from Anglo-Saxon England, are concerned with Latin versification, and they, along with Bede’s metrical handbook, provide several developments of the core grammatical concepts. Aldhelm’s work is representative of one strand of grammatical thought, which sees the combination of smaller linguistic units into larger ones as a matter of abstract patterning. Bede, on the other hand, is more interested in developing the idea of completeness of sense as it interacts with the formal patterns of Latin verse. For both authors, the self-contained, finite system of metrics offers an arena in which the word-level analysis of language can be tested against other, larger-thanword-level systems. 80 He does define sententia in Etym II.ii, but there it means an impersonal truism. 81 See my discussion in Part III below.