Wolfgang de Melo discusses the great Roman scholar.

I acquired my first edition of Virgilʼs Aeneid in a bet. A fellow undergraduate and I were debating whether the rules for counterfactual statements in indirect speech could be found in our smaller grammars, or whether one would have to consult the larger ones that we used for prose composition. The rules were set out in the smaller grammars, and the next day I was presented with a lovely edition of Romeʼs most celebrated poet. Two decades have passed, and grammar and linguistics have become my life. I still enjoy more esoteric topics like the one just mentioned, but I no longer look at grammar as a collection of random rules. What matters to me now is how the different components of grammar interact, how Latin works as a system. If we understand the fundamental driving forces in this system, the minor, esoteric rules can be deduced from the bigger principles, and we can see why certain rules exist or why they can on occasion be overridden.

If I am right in thinking of my own linguistic development as a process of maturation, then grammar in antiquity went the opposite direction. The grammarians of late antiquity are collectors: they collect odd case endings; rules of syntax; stylistic terminology; and especially exceptions to whatever regularities their predecessors came up with. A deeper understanding of what is going on is rarely sought. However, there was an earlier time when some Romans thought of grammar in a more holistic way, and not just as a means to impress others with fancy terminology or obscure knowledge. The most important scholar of that earlier era was Marcus Terentius Varro (116-27 BC), a polymath who tackled history, philosophy, and many other topics in addition to linguistics. His most significant grammatical treatise is his De lingua Latina ʻOn the Latin languageʼ. I have recently edited this work, translated it, and made it more accessible through an introduction and a commentary. My book came out with OUP earlier this year, and so I am grateful for this opportunity to present what Varroʼs work is all about.

The De lingua Latina then and now

The De lingua Latina was meant as a comprehensive treatment of the Latin language in twenty-five volumes. An introductory volume was followed by six books on etymology, six on morphology, and twelve on syntax. The hexads on etymology and morphology fell into two halves each, one theoretical and one practical.

Sadly, we have only a small portion of this work today. All our manuscripts go back to one particular manuscript, F, which was written in Monte Cassino in the eleventh century, and this manuscript transmits books 5-10, in other words, the practical etymologies and the theory of morphology. In addition, late antique authors, who still had access to the entire De lingua Latina, quote or paraphrase sections, and many of these are from the lost books. This means that the transmission is not great, but not hopeless either. Let us now turn to the individual parts of the work.

Varroʼs etymology

Varroʼs discussion of morphology is not too different from how we do things today. By contrast, if we take modern etymology as the yardstick, Varroʼs etymology seems very outdated. Modern etymology only began in the nineteenth century, with scholars such as the Brothers Grimm, who did far more than collect fairy tales, or Karl Verner. By this time, it was already obvious that languages like Latin, Greek, Sanskrit, and ultimately also English, are related to each other. However, these scholars discovered that when languages change, the sound changes are regular. This ʻNeogrammarian principleʼ allows us to take the oldest stages of languages that are related to each other and to reconstruct prehistoric proto-forms. For example, Latin pater ʻfatherʼ, Greek πατήρ, Sanskrit pitār, and Old English fæder allow us to reconstruct that in the unattested mother of all these languages, the corresponding form was *pə2tēr (the asterisk shows that the form is reconstructed rather than attested, and *ə2 is a sound whose exact nature is uncertain).

A modern diagram of the Indo-European language family tree

Modern etymologists want to know what the words we use today sounded like in the oldest attested forms of the language. They want to know whether these words came in as loans from other languages, and if so, from what languages. If the words are native, they try to find cognates in related languages and to reconstruct a proto-form. For example, kitchen looks like a thoroughly English word, but it is actually a loan from Latin coquīna and shows late Latin and early English sound changes. On the other hand, town is native, with cognates in Dutch tuin ʻgardenʼ and German Zaun ʻfenceʼ. The original meaning was probably ʻenclosed areaʼ, and in Proto-West-Germanic, the ancestor of these three languages, the form was *tūn.

The term etymologia is Greek; Cicero once renders it as ueriloquium ʻtruth-speakingʼ, but this literal translation did not catch on, probably because already in antiquity etymology was not about finding the ʻtrueʼ meanings of words. In fact, the goals of ancient etymologists are not so different from our own. If Varro knows the earlier form a word, he mentions it. He also tells us which words are loan words from which other languages. For native words, he establishes Latin cognates and relates two forms to each other.

Varro was aware that Latin and Greek are related at some level, and he knew about some sound changes, such as the one we call rhotacism, whereby s between two vowels changes to r, hence the verb carēre ʻto be withoutʼ next to the adjective castus ʻchasteʼ, which was originally its participle. However, Varro could not have known that Latin and Greek are also related to the languages spoken by the Germanic and Celtic barbarians, or that sound change is regular. This means that any proto-forms he comes up with are unreliable from a modern perspective and that he does not go beyond a prehistoric form of Latin. But it does not mean that what he does is stupid or uninteresting.

However, on occasion Varroʼs work shows signs of haste. In 7.52, Varro looks at the etymology of latrō, a word which in his own time meant ʻhighwayman, robberʼ. Varro correctly points out that the earlier meaning of the word was ʻmercenaryʼ; the change in meaning shows how common people perceived the military. For Varro, the Latin word goes back to the ablative latere ʻsideʼ because mercenaries were often hired to serve as bodyguards at the side of the king, or to Greek λάτρον ʻsoldierʼs payʼ. Such double etymologies are very common in antiquity and reveal a certain caution that is often lacking in modern scholars. Today we agree with the second etymology. But then Varroʼs haste becomes apparent. The later meaning is clearly derived from the earlier one, yet Varro feels the need to derive this secondary meaning from the verb latēre ʻto be hiddenʼ.

Perhaps it is an inappropriate question whether Varro the etymologist was any good, but I will ask it anyway. The snarky answer is that by modern standards, Varro gets it right whenever the etymology is obvious, and that he gets it wrong everywhere else. However, given the state of the art in antiquity, it would be perverse to consult Varro with the expectation that he should fulfil our modern standards. Varro makes an intelligent contribution in the context of his own times. And there are many things that we can still learn from Varro. The words he discusses show what mattered to a Roman, from what sort of public offices were considered prestigious to what sort of food was popular, from weapons used by the military to tools used in agriculture. We can also see what kind of connections an educated native speaker would make between words. And as a bonus, Varro often quotes ancient poets in support of his etymologies; many of these passages would be unknown to us were it not for Varro.

Morphology then and now

Latin is a highly inflecting language. Nouns have case, number, and gender, and verbs have person, number, voice, tense, and mood. All of these categories are discussed by Varro, who also deals with additional issues like adjective gradation or diminutive formation.

Over the last few decades, much progress has been made on all these topics. Entire textbooks are devoted to categories like tense, outlining what tense systems exist across languages and how they relate to morphology and syntax. Reading these textbooks makes it much easier to learn a new language or understand how any given language develops over the centuries. But the fact remains that every second-language learner who wants to learn Latin or any other inflecting language will still have to memorize paradigms.

Varro explains what a paradigm looks like in 10.22. However, in the extant part of the work, no detailed paradigms are presented. It is possible, in fact quite likely, that readers were given paradigms in the practical books 11-13, otherwise there would be no point in explaining what a paradigm is in book 10. But the theoretical books 8-10 deal with a bigger question: how regular is language? To what extent are inflectional patterns predictable? Varroʼs answer is complex. He is the first to draw a distinction between inflection and derivation. I give and she gives are part of the same inflectional paradigm, but givegiver, and gift stand in a derivational relationship. Modern scholarship agrees with Varro that inflection is by and large regular, while derivation is far less regular both in terms of morphology and in terms of the semantic relationship between base word and derivative.

When Varro discusses concepts like tense or mood in books 8-10, he does so in a sensible manner, but always with a view to these bigger questions. This can lead to interesting results. When we learn the indicative tenses today, we distinguish between present, simple future, imperfect, perfect, future perfect, and pluperfect. For late Latin grammarians, however, the future perfect is a subjunctive. There are two reasons for this. On the one hand, the future perfect, which indicates that a future event is over before another, more central one, is rare in main clauses because of its complex semantics, and this rarity in main clauses could lead to the interpretation that we are actually dealing with a subjunctive; subjunctives are more common in subordinate clauses than in main ones. And on the other hand, in late Latin, future perfect and perfect subjunctive, already quite similar in the classical period, are beginning to fall together completely. However, the merger is not yet complete in later Latin, and the future perfect does occasionally occur in main clauses, so our modern analysis is more sensible. But this modern analysis is already found in Varro. He starts from a purely morphological basis and notes that Latin has two verb stems, the infectum and the perfectum. The system is entirely parallel: the first three tenses are present, future, and past of the infectum, and the second set of three tenses is the perfectum equivalent. Varroʼs clarity of thought is admirable.

But like any pioneer, he does occasionally overlook things. Today, we distinguish six declension classes; grammars only count to five, but the third declension has two main subtypes, the consonant stems and the i-stems. Varro counts declension classes by looking at the ablative singular, the case he calls the sixth case peculiar to Latin because Greek lacks it. He treats our two subtypes of the third declension as two separate declension classes, but he arrives at only five declensions because he forgets what we today call the fifth declension, of the type diēs ʻdayʼ.

Varroʼs discussion of grammatical gender is particularly good. He notices that Latin gender is not sex-based; for most nouns, gender is determined by declension class. But occasionally men and women are in declension classes that are overwhelmingly feminine and masculine, respectively. In these cases, sex overrides morphology, and men and women are marked according to their biology. Varro likens this situation, where a man is called Perpenna and is placed in the overwhelmingly feminine first declension, to cross-dressing. A man may wear womenʼs shoes, but he will still be a man; likewise, he may have a name that looks as if it belonged to a woman, but he will still be a man and be modified by adjectives in their masculine form. Grammatical gender is determined by agreement patterns on adjectives and pronouns, not by the outward appearance of a noun.

In this connection, Varro also remarks that wild animals, like deer or eagles, typically only have one grammatical gender, even though the animals themselves can be male or female. We only mark sex distinctions in animals through grammatical gender if the animals are culturally important to us. Varro notes that in the early period, doves were not domesticated, so the word columba could refer to a male or a female animal. But when the Romans started to breed them, a distinction between male columbus and female columba began to be made.

I am now going to ask another inappropriate question. How good was Varroʼs morphology? Well, it depends. If we just want to learn paradigms, Varroʼs successors are much more useful. Their information is far more complete and structured in a way that makes it easy to memorize. But if we want to understand the larger concepts, we can still learn something from Varro.

How syntactic was Varroʼs syntax?

In 8.1, Varro tells us that the third part of his work, that is, the last twelve books, is about ʻhow words create a sentence when they are joined with one another in a logical wayʼ. This act of joining is expressed by the verb coniungere, which corresponds to Greek συντάττειν; the derived noun is σύνταξις, whence our syntax. When we think of Latin syntax, we think of rules of case usage, sequence of tenses, and similar phenomena. But is that what Varro means by joining words in a logical way?

Scholars are divided on this issue. Some believe that Varroʼs syntax is about case usage and similar issues; others think that it is mostly about logic. For instance, Britain has beautiful castles is a true statement; but if I coordinate it with a false one, such as and its current Prime Minister is Winston Churchill, the sentence as a whole becomes false.

We have very few fragments of the syntactic books. If I had to venture a guess, I would say that Varro moves from case usage and similar topics to issues of logic and semantics. But we need to be careful. The longest fragment we have comes from book 24 and is preserved in Gellius (16.8.1-14). It deals with issues of logic, and it draws from a variety of sources. What is unclear is how much of this chapter is Varro and how much goes back to these other sources. In fact, it is not even clear whether Gellius gives us an accurate picture of the Varronian bits. In 2.25.5-10, Gellius presents a large chunk of text as a literal quotation from Varroʼs eighth book. But we do have book 8, and a comparison with it shows that Gellius is loosely paraphrasing, modifying the text and cobbling together pieces from different sections of the book. Ultimately, then, the syntactic books have to remain a mystery.

A scholar after my own heart

By now it should be apparent that Varroʼs De lingua Latina will never make it onto any sensible school curriculum. Children do not need to learn etymologies, let alone ones that go back to a pre-scientific era. Latin morphology is set out more clearly in modern textbooks than in any ancient author; and the same goes for Latin syntax. And yet I would like to think that I have not wasted six years of my life on collecting obscurities in the style of late antique grammarians. Varro has a lot going for him: he was intelligent and industrious; he absorbed knowledge like an educational sponge; and above all, he was not afraid to ask difficult, big questions and to give his own views. All too often, modern scholarship resembles the late antique habit of accumulating and collecting material on well-trodden paths. I hope that we can learn from Varro and become more open again to new approaches and bigger questions.

Professor Wolfgang de Melo is Professor of Classical Philology and Governing Body Fellow of Wolfson College, Oxford. His new edition of Varro: de lingua Latina came out in March 2019 (Oxford University Press).  He has also published The Early Latin Verb System: Archaic Forms in Plautus, Terence, and Beyond (Oxford University Press, 2007) and five volumes of Plautus for the Loeb Classical Library.