Wolfgang de Melo
Many moons ago, when I was a student at the School of Oriental and African Studies, we were analysing the sounds of Dida, a language spoken in Ivory Coast. Dida has four distinctive tones, a feature which some students struggled with. Every time they were stuck, they would immediately turn to the one student whose first language was Cantonese and ask him, “What are the tones in this word?” But he was struggling just as much. Perhaps this shouldn’t have come as a surprise to anyone, as it is a very different tone system. Now that I’m back to teaching Latin, I occasionally get similar complaints from students. My first language is German, and German nouns, just like Latin and Greek ones, belong to one of three genders, masculine, feminine or neuter. Some students assume that Latin and Greek gender must be straightforward to me; and one, who had to take a German course, turned up in my office, stared at me accusingly and said, “In your language, spoons are masculine, forks are feminine and knives are neuter; explain yourself!”
This, then, is what I will try to do for Latin, but also for other languages. If you bear with me, you will see how different languages assign gender to nouns and how this is not a random process; and hopefully you will come to think of gender as not quite as pointless as it may have seemed before. But my essay is not a magic bullet: it will help you to make sense of grammatical gender assignment, and in that sense it will help you to learn languages, but it will not eliminate the effort of learning altogether. However, on our little journey together, we will also touch on some questions of broader interest: how did ancient scholars think of gender? Can nouns have more than one gender? And does grammatical gender influence the way we think about the world? We will begin our journey with the word gender itself.

Genus, gender and genre
English has a great many loanwords from Romance, and some words even entered the English language more than once. Thus, wine and vine both come from Latin uinum; the former was borrowed when the initial sound in Latin was still pronounced as w, and the latter came in after the sound had become v, as in modern French. In the same way, Latin genus entered English twice, once through Old French gendre, and once through a more modern form genre.

Latin genus is a remarkably vague word. It means a “kind”. There can be different genera or “kinds” of animals, such as birds and mammals and reptiles; and there can be many other kinds – basically, whenever a Roman would classify the world, the entities would fall into several genera. The Greek cognate of our word, genos, was similarly vague, and eventually acquired grammatical usages as well. When the Romans started to get interested in grammar, genus came to mean “kind” of noun or “gender”, but because genus could also refer to biological sex, male and female, these concepts were occasionally conflated. I will use male and female for biological sex and masculine and feminine for grammatical gender to keep the ideas separate. Grammatical gender is about classifying nouns, and that classification can be based on sex, but also on other criteria.
Genus is a neuter noun. In later Latin, the neuter eventually disappeared, being mostly absorbed by the masculine. Genus turned into masculine generem (accusative), and then into Old French gendre, whence it was borrowed into English as gender. Modern French genre is still masculine and was reborrowed in the meaning “literary kind”. German also borrowed Genre in this meaning, but in German the word is neuter.

We don’t care!
Germans, whether or not they know any French, will know that Genre is a French word, just as English speakers recognize genre as a loan. For loanwords, many people make concessions in pronunciation; the initial sound in genre is not found in native German or English words. Such concessions in pronunciation are made more often when that language is considered prestigious and is taught in schools, as is the case with French; but even then, gender is normally adapted to the native system. The reason is one of awareness: we tend to notice unusual pronunciations and accents, but we are normally less aware of grammatical features like gender or case, which we handle subconsciously. Occasionally, we are told what gender a noun should be. For example, German children are taught at school that Virus should be neuter because it is so in Latin, and Email should be neuter because in English it is referred to with it. People sometimes stick to these rules in formal writing and on other occasions when they feel observed; but the normal situation is that Virus becomes masculine because all the other Latin loans in -us are, and Email becomes feminine because Post is.

The Romans were no different. In the second century BC, Greek was increasingly becoming a prestige language. Plautus (died 184 BC) used many Greek words, but was happy to give them Latin endings; not much later, on the other hand, it became a sign of education and refinement to pronounce and inflect Greek words as if Greek. Greek nouns in -ma are neuter, and their genitive ends in -matos. The Roman elite would treat such loans as neuter, too, with a genitive either in -matos, in the Greek style, or in -matis, with an adapted ending that still isn’t fully Latin. But the masses didn’t care. Latin nouns in -ma are feminine and have a genitive in -mae, like lacrima ‘tear’, genitive lacrimae, and so the less refined writers follow this pattern for Greek words such as glaucuma, “eye disease”, diadema “crown”, or later on, in the Christian era, baptisma “baptism”.
To the native speaker, then, gender assignment is quite intuitive. Sometimes, there is variation, but this variation is not random. For example, German Butter is feminine, but in some southern varieties it is masculine (because most other nouns in –er are masculine – although it has to be said that they are agent nouns like Schwimmer ‘swimmer’); what we never find for ‘butter’ is the neuter. But what is the assignment based on? What does gender actually reflect?

Tense and time, gender and… and what?
Linguistic categories normally mirror the real world, but very imperfectly. Tense mirrors time, but very imprecisely. I could use a Latin perfect, imperfect or pluperfect for the same event which happened yesterday. Tense cannot replace time words or phrases like yesterday at 4 pm. However, tenses can give us additional information that such phrases can’t provide: the perfect would tell us that it’s an important main event; the imperfect would tell us that it’s less important, the backdrop for something else; and the pluperfect would show that it happened before some other, significant event in the past.
But what does gender mirror? The naïve answer is ‘biological sex’, and this answer is not entirely incorrect; gender can reflect sex, but in some languages, like Ojibwe (spoken in parts of Canada and the US) it is animacy rather than sex. Not every language has grammatical gender (Finnish doesn’t), but every language which does have a gender system has a core group of nouns, for example for humans, where gender assignment is based on such meanings. But most languages also take other criteria into account, such as noun classes. Let us look at gender strategies in different languages.

Gender assignment by meaning
In some languages, gender assignment is purely semantic, that is, it is done by meaning. This may feel like the ideal situation, but it is quite rare. We find such systems in southern India and Sri Lanka: the Dravidian languages typically have three genders, masculine, feminine and neuter, and these correspond very neatly to biological sex. For instance, in Tamil, masculine gender is used for male humans and deities; feminine, for female humans and deities; and neuter, for everything else. Animals can of course be male or female, but in Tamil they are neuter rather than masculine or feminine; masculine and feminine should perhaps be defined as rational entities that are male and female.
There are very few exceptions in this system: the word for child is neuter, but can be masculine or feminine if you want to specify sex. The moon is masculine because it is also a deity. And the elephant is neuter, but if you make it masculine or feminine, it refers to a human with elephant-like qualities; to Tamil speakers, that doesn’t mean large size, but a good memory.
Other languages with semantic systems categorize the world differently. Dyirbal, an Australian Aboriginal language, has four genders: gender 1 for male humans, but also most animals regardless of sex; gender 2 for female humans and dangerous substances like fire; gender 3 for plant food; and gender 4 for everything else. Again there are exceptions: birds should be in gender 1, but end up in gender 2 for mythological reasons, since they represent the spirits of dead women; fishing instruments should be in gender 4, but are associated with fish and so are marked for gender 1; and although most fish species are in gender 1, two dangerous species are in gender 2.
What Dyirbal shows is that grammatical gender, even if based on meaning, can be more complex than a mere reflection of biological sex. Gender is a grouping into different kinds, and sex and animacy play a huge role in this, but they are not sufficient.

Gender assignment by morphology
The older Indo-European languages like Sanskrit or Old Church Slavonic – we will return to Latin in more detail below – have morphologically assigned gender. That means that once we know how a noun inflects in all cases and numbers, in other words, once we know its declension class, we can typically predict its gender. Such systems are found elsewhere as well. For example, the Bantu languages, spoken across most of sub-Saharan Africa, also assign gender by morphology; in Bantu languages, scholars typically speak of ‘noun classes’ rather than genders, but this is simply a terminological difference. Let’s look at Swahili. Kikapu means “basket” and it consists of a class marker ki- and the actual noun -kapu. If we combine this with the adjective for “large”, this adjective has to agree in gender, so ki-kapu ki-kubwa. The plural is vi-kapu, and “large baskets” translates as vi-kapu vi-kubwa. This gender is referred to as ‘noun class 7/8’ by Bantuists because singular and plural are treated as separate classes in this tradition. Similarly, we have m-tu, “person”, and wa-tu, “people”, class 1/2, and a “small person” is m-tu m-dogo, while “small people” is wa-tu wa-dogo.
Morphological gender systems are never quite as neat as semantic ones because there is normally a group of nouns where meaning matters more than morphology. Typically, this group includes human beings, and sometimes other animals. In Swahili, there is no grammatical gender distinction between male and female; human beings and animals take the agreement for class 1/2, regardless of their morphology. Now, many nouns for human beings and animals are themselves marked appropriately, as we have already seen with m-tu m-dogo and wa-tu wa-dogo. But if the noun is marked differently, it will still take the agreement for class 1/2. So “rhinoceros” is ki-faru, with a prefix for class 7, but as this is an animal, a “small rhinoceros” is ki-faru m-dogo. Meaning trumps morphology.

Gender assignment by phonology
In some languages, the sound structure of a word may determine its grammatical gender. When I first encountered such a system, I was a little taken aback, but if we think of gender as classifying nouns according to specific criteria that need not be semantic, phonology is no stranger than morphology. And, as we shall see, such systems are also found in the descendants of Latin! As before, there is always a semantic core where meaning overrides other criteria. In Yimas, a language from Papua New Guinea, there are eleven gender categories. Genders 1-4 are determined by meaning, regardless of phonological form: genders 1 and 2 are for male and female humans, respectively; gender 3 is used for higher animals like dogs and crocodiles; and gender 4 is for important plants, like the coconut palm or the sago palm, which provide food. Genders 6-11 are determined by phonology: for instance, if a noun ends in -mp, it belongs to gender 7. I have not yet mentioned the incredibly useful gender 5: anything that can’t be assigned to genders 1-4 by meaning, or to genders 6-11 by phonology, ends up in gender 5, a veritable leftover basket.

Back to Latin
Now that we have seen how different languages assign gender, let’s look at Latin again. Gender in Latin is largely assigned by morphological criteria, but again with a semantic core. There are five declension classes, some of them with subclasses. Nouns of the first declension, with a nominative in –a, are feminine. Those of the second declension end in –us or –um and are masculine and neuter, respectively. Nouns of the fourth declension end in –us or –u and are masculine and neuter, respectively. And nouns of the fifth declension end in –es and are feminine.
The third declension contains nouns of all three genders. But again, gender assignment is not random. Many nouns contain derivational suffixes, like –tio for abstract nouns. Such suffixes are associated with specific genders; nouns ending in –tio are feminine, those in –men are neuter, and so on. Those nouns which don’t have such suffixes were problematic for the Romans too, and sometimes there is gender variation here.
Let us now look at those cases where meaning overrides morphology. The first semantic group consists of humans. Human beings are masculine or feminine based on biological sex, which almost always trumps morphology. Agricola (first declension), “farmer”, should be feminine, but since the noun refers to a male occupation, it is masculine. Optio, “choice”, has the feminine suffix –tio and is feminine when referring to the act of choosing, but masculine when it refers to a military rank. In Plautine comedy we find the Greek names Astaphium and Pinacium, both diminutives in Greek; in Latin, Astaphium is feminine and Pinacium is masculine because they refer to a woman and a boy, respectively. I can think of only two nouns referring to humans that are neuter: scortum, “prostitute”, originally “piece of leather”, and mancipium, “slave”, originally “purchase”. Here the original meaning is still felt clearly enough to prevent sex-based gender assignment. Other words for prostitutes and slaves, however, are feminine and masculine. Incidentally, sometimes nouns change declension class because their human referents need a gender that is in conflict with declension class; we see this with fourth-declension socrus, “mother-in-law”, which ends up as Italian suocera.

The second group contains trees and cities (feminine) as well as winds and rivers (masculine). These semantic classes are based on mythological associations that aren’t felt acutely any more, and so there are far more exceptions than in the first group. For example, ficus, “fig tree”, is feminine, but occasionally masculine adjectives can be found modifying it. And anything ending in –um is neuter, regardless of semantic associations.
And finally, we have a few exceptions that can be explained historically, but must have felt random to Romans. In the predominantly masculine fourth declension we find feminine porticus, “portico”, and manus, “hand”; but porticus is also attested as masculine, and manus is so frequent that it is learnt in early childhood, and such words retain irregularities more easily. Even in the Romance languages, the word is still feminine. In the fifth declension, the only word inherited as masculine is dies, “day”. Historically, it referred to a male sky deity, “Father Day” or Diespiter, with cognates in Greek Zeus pater and Sanskrit Dyauṣpitṛ́. The connection of dies with the deity was no longer obvious in the classical period because the nominative Diespiter had largely been replaced with what started as the vocative, Iupiter. Since all other nouns of the fifth declension are feminine, dies began to take on feminine gender as well. Cicero and Caesar, two purists who were very conscious of their language use, saw this variation, felt somewhat uncomfortable with it, and decided that dies should be masculine when meaning “day”, but feminine when meaning “appointed time”. Virgil uses the masculine gender wherever he can, but the feminine where it is required metrically; hence masculine ille dies, “that day” (feminine illa dies would fit into a hexameter equally well), but feminine summa dies, “final day”, or longa dies, “long day” (masculine summus dies and longus dies do not fit into hexameters). Eventually, the feminine gender prevailed, and many Romance reflexes of dies are feminine.
We can see the various factors at play in one and the same word, forceps, “tongs”. This is a feminine noun which by metathesis becomes forpex, still feminine. By assimilation, it turns into forfex, still most often feminine, but since we have many masculine agent nouns in –fex, “maker”, Vitruvius turns forfex into a masculine, even though the tongs don’t make anything – a case not too different from German Butter mentioned above, which is mostly feminine, but can be masculine by association with agent nouns in –er.

From Latin to French
If you had French classes that were anything like mine, you didn’t learn any gender rules. Instead, you had to learn every noun with the article, which shows you what gender the noun is. That’s not really how French children learn their language. Studies that were already out when I was at school prove that French gender is largely predictable: human beings are masculine and feminine based on their sex, and for other nouns you can predict gender based on the last sound or sounds of a word. For instance, words ending in the sound z (like église “church”) are feminine in 90% of all cases. I wish we had learnt those rules at school; it would have been a bit of work at first, but then it would have saved us a lot of trouble.
But how could the Latin system change so dramatically? The answer is, by sound change. Already in later Latin, final consonants were lost. When the distinction between –us and –um disappeared, the latter was reanalysed as masculine. Occasionally neuter plurals in –a were also reanalysed as feminine singulars; this happened to nouns referring to things that tend to occur in groups, like Latin uelum, “sail on a ship”, plural uela, but now French feminine singular voile going back to the plural. This late two-gender system survives into French.
But as final sounds dropped of, and sound change obscured many of the suffixes in the third declension, the system as a whole was reanalysed as a phonological one. Now it is the last few sounds that tell you what gender a noun is.
The semantic core still exists. But notice how even that has undergone simplification. For humans, gender assignment is based on sex. But trees and cities, winds and rivers do not follow the ancient pattern any more because the mythological connections disappeared. As for the third group, the random exceptions, only the frequent manus remained feminine, la main.

Up to this point, everything has been nice and neat. But some complications await us. Can a noun belong to two genders at once? And what if we combine nouns of different genders?
Hybrid nouns and gender resolution
Gender has been quite unambiguous so far: assigning gender to a noun may not be a simple process, because meaning and inflection and sounds may all be involved, but at least once we know what gender a noun belongs to, we are going to be fine. Or are we? We have already seen that a noun like German Butter may be masculine for some speakers and feminine for others, and so in that sense there is variation, but at least once a speaker has decided which gender it’s going to be, there is no more wiggle room. Or is there?
Actually, for some nouns there is; not for Butter nouns, but for ones like German Mädchen, “girl”, or Greek paidion (παιδίον), “little child”. These are charmingly called ‘hybrid nouns’ because they share properties of two genders. Mädchen and paidion are diminutives, and diminutives in German and Greek are neuter. Now if a base noun referring to an inanimate object is masculine or feminine, like Greek pinaks (πίναξ), “painting” (masculine), there is no issue with a neuter diminutive. But if the base noun is masculine or feminine and also human, a neuter diminutive fits with morphology, but there is a clash with sex-based gender assignment. So what are we going to do about it? Linguists have come up with the so-called Agreement Hierarchy, in the form:
Attributive > Predicate > Relative Pronoun > Personal Pronoun.
The more to the left an element is situated, the more likely it is to have agreement in accordance with morphology; the more to the right it is, the more likely it is to follow sex-based assignment. Let’s look at paidion: a child can be male or female, and for the base noun pais we get masculine or feminine agreement. With paidion, the article, an attributive element, will always be neuter; but relative pronouns can show some variation between neuter and masculine or feminine, and personal pronouns, while showing the same variation, have a greater tendency to be masculine or feminine rather than neuter. But there are also other factors involved in the choice: the older the child, the more obvious its sex is going to be, and the greater the likelihood for semantic agreement with personal pronouns.
If this seems complicated, perhaps it may help to look at English hybrid nouns. Yes, English also has a few of those: sun and moon, ships, and also pets. Sun and moon were feminine and masculine respectively in Old English. Then grammatical gender was reduced to marking on pronouns, and sun and moon became neuter. But in poetry, we can treat the sun as a he and the moon as a she, the exact opposite from Old English; this is because of a newer mythological association. Still, even then the relative pronoun is inanimate which rather than animate who. Ships can be feminine by personification, so she, but again it is which. And pets are more complicated. If you think of your dog as a family member, it’s going to be he or she, and potentially who, but perhaps also which. But for other people’s dogs it may be it.

My favourite type of hybrid nouns in Latin is pregnant animals of declension classes which are normally associated with masculine gender. Will biology win over morphology? In other words, will they be feminine or masculine? Sadly, we don’t have enough examples to see a clear trend, perhaps in accordance with the Agreement Hierarchy. In Plautus, pregnant elephants come up once (Stichus 168-9) because they are supposedly pregnant for ten years, a belief widely held in antiquity (it’s actually two years). Plautus combines second-declension elephantus with the feminine adjective grauida, “pregnant”. Varro discusses pregnant hares (De re rustica 3.12.5) and notices, correctly, that female hares can conceive even if they are already pregnant, a phenomenon known as superfetation. Lepus is a noun of the third declension, and Varro combines it with the masculine relative pronoun qui, even though the animal is obviously female.
In my undergraduate days, professors would address their male students as Herr and their female ones as Fräulein. German still had the same distinction that we find in English, Herr for Mr, with no regard to marital status, Fräulein for Miss, and Frau for Mrs. As soon as I left the country, Fräulein was abolished and replaced by Frau, which now became the equivalent of English Ms. There were three reasons for getting rid of the Fräuleins. First and foremost, people felt that one shouldn’t have to parade one’s relationship status around, a sentiment that I find myself in full agreement with; this was also the reason for introducing Ms in English. But the other two reasons have no English equivalent: many were upset that Fräulein is, formally, a diminutive; and some got agitated because it is neuter (hybrid, actually, but most people can’t see beyond the article). These two specific points leave me a little bemused. It is of course true that -lein is a diminutive suffix, but in Fräulein it has lost its diminutive status; a Fräulein is not a little woman, but an unmarried one. If you really want a diminutive meaning, you have to say Frauchen and Herrchen, with a different suffix, and these are used affectionately for female and male dog owners. As for the neuter article, that is unavoidable because of the suffix, and no one seems to mind too much that it’s the same for Mädchen. However, one interesting thing emerges: it is easier to abolish a word like Fräulein altogether than to change its gender.

If you put ‘gender resolution’ into Google, all sorts of political things will come up. But in linguistics, the term refers to what we do when we have groups of mixed genders, for instance if we coordinate a masculine and a feminine noun with ‘and’. If we now have an adjective referring to both entities, what shape will it take? Different languages adopt different strategies. Let’s begin with German. “The man and the woman” is der Mann und die Frau. If we want them both to be “good”, we have to repeat the adjective: der gute Mann und die gute Frau. This is because the adjective stands between article and noun, just like in English, and since Mann is masculine and Frau is feminine, we have to use two different articles. But if we now turn this into a plural, we can say die guten Männer und Frauen, and the adjective refers to both nouns. Why is that? Well, as you can see, the article is used only once now, and so it modifies both men and women, just like the adjective. This is possible because of a phenomenon called syncretism. Quite often languages make fewer distinctions of form in the plural than in the singular. In German, we have distinct masculine, feminine and neuter paradigms for singular articles and adjectives, but in the plural the three genders look alike (and very similar to the feminine singular, but that is a coincidence). All German speakers are aware of the three genders, but hardly anyone who has not done German grammar realizes that this distinction is only visible in the singular.
Depending on which Latin grammar you use, you will find the rules of gender resolution more or less complex, but they are almost always presented as quite firm rules. The reality, however, is so much messier. There are two conflicting tendencies at play: on the one hand, Latin can go for genuine resolution, and in this case a man and a woman will be modified by a plural adjective, and the default gender is masculine. If we coordinate inanimate entities of different genders, the plural adjective may be neuter plural. On the other hand, there is another tendency, the tendency for the adjective to agree only with the noun that stands closest to it, even though its meaning extends to all the nouns. This second tendency makes a great deal of sense; why should a masculine and a feminine noun be combined with a neuter adjective, or why should one gender win out over the other? When we examine texts, we can see that one and the same author may use both strategies, and it is not always easy to determine the factors behind the choice.

Grammatical gender and sexism
Gender resolution brings me to the question of whether grammatical gender is sexist. Since my schooldays I have been listening patiently to claims that Latin and the Romance languages are inherently sexist because if you have a group of men and women, the adjectives will be masculine. My fellow Germans may not have realized that German has gender syncretism in the plural, but at least they learned their gender resolution rules for Latin and whichever Romance languages they had at school (in my day, it used to be French or Italian). And most Italians seem to be aware of Italian gender resolution, but what is less clear to me is whether this is simply because many had Latin at school, or perhaps because gender issues are a hot topic these days. But how sexist are such gender resolution rules in reality? Did they come about because of patriarchal oppression? Or if not, do they help to cement such an unfair system? The wider question is to what extent language shapes thought.
Purely empirically, we can try to find out whether languages with different gender resolution systems have different societal norms. One such language is Hadza, a linguistic isolate spoken by a small group of hunter-gatherers in Tanzania. Hadza has two genders, masculine and feminine, assigned on semantic criteria; men and oblong objects are masculine, women and round objects are feminine. In groups consisting of men and women, it is the feminine plural suffix that is used. The Hadza are a relatively egalitarian society, as is usual among hunter-gatherers, but even here we find sex-based work patterns, with men doing more hunting and women doing more gathering. If language shapes thought, this influence is much more subtle; the Italian gender resolution system does not make you any more patriarchal than the Hadza one makes you matriarchal.
However, there is some limited evidence that the grammatical gender of inanimate objects may influence how we see them. Italian and German assign gender to humans based on sex, but for inanimate things, gender assignment is largely driven by morphology. Some psycholinguistic studies indicate that when speakers are asked to find suitable adjectives for “bridge”, Italians are more likely to use words meaning “strong” or “sturdy” for masculine ponte, whereas Germans prefer terms meaning “elegant” or “beautiful” for feminine Brücke. This shows that however large a role morphology or phonology may play in gender assignment, there is still a small but significant core of nouns for which sex is fundamental, and this core can influence other nouns within its gender group.

The wider question, to what extent grammatical categories influence our thought patterns, is associated with two brilliant linguists, Edward Sapir and Benjamin Lee Whorf, both of whom did important work on Native American languages. I use the common term ‘Sapir-Whorf hypothesis’ with some hesitation because the two scholars never published together and did not propose a hypothesis as such, but the one that bears their name, as elaborated by later scholars, comes in two versions. The strong version claims that grammatical categories determine and limit our thoughts; the weak version argues that such categories can influence and bias them. While the strong version has largely been discredited, the weak version finds some limited empirical support. However, as far as gender is concerned, we seem to be dealing with a one-way street: inanimate entities that are masculine or feminine in gender may be associated with adjectives that typically go with male or female animate entities; but this core group of animate nouns is not influenced by gender associations with inanimate ones.
Grammatical gender in and of itself is not sexist, but of course that is not to deny that sexism exists in language. However, it tends to manifest itself lexically rather than grammatically. It happens when people use derogatory and insulting terms; or it happens more subtly when a male colleague is introduced as Professor + surname, while his equally qualified female counterpart is introduced by first name. But grammatical gender is not the enemy. Let us now move back to ancient times and see what Varro has to say about gender.

“Gender is like women’s shoes”: Varro’s thoughts
Marcus Terentius Varro (116–27 BC) was a veritable polymath. His De lingua Latina, “On the Latin Language” (40s BC), was not a grammar as such, but a discussion of specific linguistic topics. While later grammarians provide fuller paradigms and descriptions, Varro is particularly interesting because he tries to understand why Latin is the way it is. One topic he discusses is gender.
Varro does not have the terminology to distinguish sex and gender, but the difference is clear to him: in 9.41, he points out that paries, “wall”, and abies, “fir-tree”, are respectively masculine and feminine grammatically, but neuter by nature. For Varro, masculine and feminine is not the equivalent of male and female, but those words are masculine and feminine which are combined with the pronouns hic and haec, respectively. This comes remarkably close to our modern definition: we now say that gender is manifested in the behaviour of associated words. Interestingly, many other Roman scholars tried to understand grammatical gender in terms of sex, definitely a step backwards after Varro’s insights.
Varro also notices the connection between declension class and gender. In 9.40, he compares gender to shoes. Women’s shoes are normally worn by women, but occasionally they are worn by men. A name like Perpenna (of Etruscan origin) refers to a man, despite being in the overwhelmingly feminine first declension. Having a name like that is like wearing women’s shoes; it does not make Perpenna a woman.
In 9.56, we find a particularly interesting discussion. Varro notes that we can change declension class so that the accompanying change of gender reflects sex differences. Hence equus, “male horse, stallion”, and equa, “female horse, mare”. Animals come in two sexes, but often the Latin language uses the same word for both. In such cases, there is an underlying gender distinction, but it is not expressed unless the animal is culturally significant. Cultural significance can change. Varro tells us, correctly, that originally male and female doves were called columba, because sex differences didn’t matter. But when the Romans started to breed doves, they also started to distinguish between male columbus and female columba. And, on this note, let us move back to modern times and English pronouns.

How linguistically viable are special pronouns?
Over the last decade or so, many people, especially of the younger generations, have come out as non-binary. Typically, these are people who are biologically male or female and have not undergone any gender reassignment surgery or hormonal treatment, but either because of gender dysphoria or for entirely different reasons, they do not want to think of themselves as belonging to their biological sex or, for that matter, to the opposite sex. This raises complex questions of pronoun use. Sometimes, such individuals wish to be referred to as they, or they prefer newly created pronouns, such as ze and hir. How linguistically viable are such innovated usages?
The usage of pronouns can change, just as any other aspect of grammar. In Latin, the second person singular pronoun is tu, and its plural is uos. In Late and Medieval Latin, the plural begins to be used for addressing individuals; the plural is considered more polite, increasing the addressee’s influence and power. The usage survives into present-day French, with tu being used for informal address and vous for formal or plural address. Partly under the pervasive French influence that we see throughout the Middle English period, a similar phenomenon happens in English: the original distinction between singular thou and plural you is modified along French lines, but most varieties of English go one step farther and eventually eliminate thou, except in traditional contexts such as the liturgy. Similar phenomena are found across Europe. Dutch even creates a new pronoun: in the second person singular, the nominative used to be gij (preserved in Flemish), and the oblique and possessive were u and uw. For formal address, a new phrase uw edelheid, “your nobility”, was created, which was shortened to U.E. and eventually to u. This is now the new nominative pronoun for formal address, and the verb agreement fluctuates between second and third person, the latter still in agreement with the etymology. However, it should be noted that in all these cases we are dealing with politeness phenomena in address, where European languages do not make gender distinctions among pronouns.
Other persons can also undergo change, often as the result of a prolonged period of intense bilingualism. In English, we can say give them, or give ’em, and ’em is normally treated as an informal shorter version of them. But historically, these are two separate pronouns. Our short ’em goes back directly to Anglo-Saxon, while them is actually derived from Old Norse. In the Old English period, the Danes invaded, and their language was still so similar to Old English that many basic words entered the English language, for example egg, or sky (the native word is heaven). My father’s native language, Konkani, is spoken in Goa in southern India. It is an Indo-European language, but has been in contact with Kannada, a Dravidian language, for centuries. Under Kannada influence, Konkani introduced a further pronoun for the first person plural, so that now there is a distinction between one we that refers to the speaker and the addressee (and excludes others), and another we that refers to the speaker and others (but excludes the addressee).

On the whole, however, changes in pronominal systems are so rare that pronouns are often used to establish whether two languages are related. Pronouns tend to be what linguists call a ‘closed class’, a small group of words that is unlikely to increase, unlike ‘open classes’, such as nouns, where new words can enter the language freely. Pronouns are acquired in early childhood and as such are handled without much thought and are very resistant to change. This, then, means that entirely new forms like ze and hir will almost certainly fail, even among the very activists who are trying to promote them.
Let us look at these neopronouns a little longer because there has been a recent study by Dennis Baron, a study which I admire in some respects, but feel ambivalent about in others. Baron is both a linguist and an activist. As a linguist, Baron has done painstaking research and collected pretty much all English neopronouns created between 1770 and 2018. The sheer number of these neopronouns is staggering, and I am impressed by Baron’s thoroughness. I am less impressed when Baron the activist takes over. On page 13, he states: ‘Since all words are invented words, if you don’t see a pronoun here that meets your needs, by all means, make one up.’ Baron’s list of neopronouns goes well beyond a hundred, and the immediate question should not be whether we need even more of them, but why the previous inventions have all been so unsuccessful.
My answer comes in two parts. First, the search for neopronouns for non-binary people does not go back that far. Most neopronouns, at least until recently, were not invented for people who think of themselves as neither male nor female, but for more philosophical issues: if I start a sentence with everyone and refer to a group consisting of men and women, can I really continue logically with he? This is a relatively minor concern for most people, and so it is perhaps not surprising that neopronouns invented with such issues in mind were doomed to failure. And second, more recent neopronouns for non-binary people have failed because every grammatical system is a simplification. A system like English, with he for men and she for women, is a simplification that does not, for example, take medical intersex conditions into account; but because these are so rare statistically, it is impossible for the majority of people to maintain the constant awareness that the introduction of a special pronoun would warrant. People can and will change lexical expressions with ease; but closed-class items like pronouns only change under very limited circumstances.

Neopronouns evoke strong emotions, but ultimately the battle that rages about them strikes me as misguided. In some conservative circles, there is a fear that the introduction of neopronouns would change the societal status quo; and in some left-leaning circles, there is such a strong push for neopronouns precisely because it is hoped that the status quo would be challenged. Both sides have an unwarranted belief in a strong form of the Sapir-Whorf hypothesis. But new pronouns would not automatically lead to greater acceptance of non-binary people, just as the abolition of gendered pronouns would not automatically lead to gender equality. Again, it is worth looking beyond English, at languages whose pronouns are not marked for gender. Within Europe, we can see this with Finnish, Hungarian and Turkish. Finland does have a strong egalitarian ethos, but so do the other Scandinavian countries – whose languages have gendered pronouns. I leave it to readers to decide whether Hungary and Turkey are more tolerant towards non-binary people than, say, Britain and the Netherlands with their gendered pronouns.
Let us move on to they. They for a non-binary individual is a little different and also more complicated. Some well-meaning linguists, trying to be modern and supportive, have claimed that this usage is an old (and hence acceptable) one, but such claims are somewhat disingenuous (quite apart from the fact that they keep telling us that old does not mean good). The oldest example listed by the Oxford English Dictionary for ‘singular they’ does indeed go back to 1375, to a work called William and the Werewolf. Here, each (man) is picked up with they (I have modernized the orthography because it is irrelevant to the argument). While I can interpret this as a singular they, it is also undeniable that there is an implicit plurality in the word each. Similar mismatches are found in less careful Latin as well, such as in Plautine comedy. In the seventeenth century, medical texts did, it is true, use they for hermaphrodites; but we are dealing with niche usages here. What is truly old is they for people whose sex is unknown, perhaps because we are seeing them from a distance. But that is a category that is very different from people we know, people who think of themselves as non-binary, but who have not undergone any kind of hormone treatment or gender reassignment surgery. Human beings automatically, instinctively, put others into boxes and categorize them as male and female. It’s something we’re extremely good at: someone’s voice heard over the telephone is normally sufficient to make a reasonably accurate guess about not just their biological sex, but also their age, and often even their height and weight. All this information that is constantly coming at us makes it very difficult to use they.
Some of my trendier linguistics undergraduates claim that they use singular they consistently and with ease, and that their parents’ failure to do so, despite trying, is a generational thing. But when I listen to the same undergraduates speaking casually, they are anything but consistent, which goes to show that pronouns are normally handled without much thought or consideration, and that people are rarely aware of the reality of their own language use. The dark corners of YouTube are filled with clips of people berating others for not using the desired pronouns, only to fail in doing so themselves mere seconds later. I believe that a little kindness on everyone’s part goes a long way, and that we should try to oblige those who think of themselves as non-binary by using they; but I also believe that any slip should be treated with good grace. Given that polite you eventually replaced thou, it is just about conceivable that gender-neutral they used for a known person may catch on.

Some final thoughts: gendered language
Our little journey has taken us from linguistic strategies for gender assignment to hybrid nouns and gender resolution, from ancient thought on gender to English pronoun usage. A completely different question that often comes up in connection with gender is how male and female speech differ; and this is the final topic we will look at here. The main problem is that we don’t have many ancient female writers. In Greek, we have Sappho and Corinna, but fragmentary and highly elaborate poetry in specific dialects does not tell us much about the everyday speech of women who spoke Attic, the best-preserved variety of ancient Greek. Scholars of Greek and Latin often look at drama, where we do get a fair amount of female speech. For Greek tragedy, Imogen Stead has written on Antigone about this issue much more competently than I could. But I can say a few words about Roman comedy.
The only two writers of comedy of whom we have complete plays are Plautus (died 184 BC) and Terence (died 159 BC). In their comedies, there are quite a few female roles, but these female characters are never main characters, and they fall into specific stock roles: the unpleasant married woman of an advanced age, the clever prostitute, the slave-girl. We cannot expect a great deal of individuality here, and given that comedy is always in verse, we must not pretend that this is anything resembling a natural speech sample. Most importantly, perhaps, we should not forget that the writers of these roles were male, which means that certain female characteristics may be exaggerated, while others may be overlooked. The situation is not that different from what we find in Enid Blyton (1897–1968), a children’s writer who held a number of unpleasant prejudices and conveyed them in her writings. In Blyton’s Malory Towers (1946–51), there are several silly French ‘mamzelles’ and uncultured American schoolgirls. Blyton imitates their accents and grammatical peculiarities, but this imitation can hardly be called realistic. Still, the imitation is effective insofar as it relies on a small number of stereotypical features that send clear signals to the reader.

Plautus and Terence do the same with female language. One difference between men and women is absolute and thus probably reflects real speech: only men swear by Hercules and only women swear by Castor (while Pollux is invoked by both sexes). But other differences are statistical, and it is hard to say whether Plautus and Terence are exaggerating. For instance, women use far more politeness markers than men, far more words for “please”. Interestingly, the difference is not just one of quantity. Men prefer the slightly less polite sis / sultis (literally “if you want”), while women have more instances of opsecro, “I entreat you”. My favourite word for “please” is amabo, “I will love you”. Women use it towards men and women alike, but men only ever use it towards women, never towards other men; an instance, perhaps, of Ancient Roman men being afraid to express feelings and vulnerability towards other men?

Wolfgang de Melo is Professor of Classical Philology at Oxford. He has published on early Latin, especially Plautus and Roman comedy, and on Varro. He teaches linguistics and comparative philology and has a special interest in linguistic typology.
Further Reading
The classic introduction to grammatical gender is Greville Corbett’s Gender (Cambridge UP, 2000). Varro’s grammatical doctrines are now collected and discussed in a work of mine, Varro: De lingua Latina; Introduction, Text, Translation, and Commentary (Oxford UP, 2019). For those interested in Roman thought on grammatical gender more broadly, I recommend Anthony Corbeill’s Sexing the World: Grammatical Gender and Biological Sex in Ancient Rome (Princeton UP, 2015). The best discussion of female speech in Plautus and Terence is still the article by the late J.N. Adams, “Female speech in Latin comedy,” Antichthon 18 (1984) 43–77.
French gender and its base in phonology is discussed in G.R. Tucker, W.E. Lambert & A. Rigault, The French Speaker’s Skill with Grammatical Gender: An Example of Rule-Governed Behavior (Mouton, The Hague, 1977).
For a discussion of the Sapir-Whorf hypothesis see John McWhorter’s The Language Hoax: Why the World Looks the Same in Any Language (Cambridge UP, 2014), and Lila Gleitman & Anna Papafragou, “Relations between language and thought,” in D. Reisberg (ed.), The Oxford Handbook of Cognitive Psychology (Oxford UP, 2013), 504–23. The study on neopronouns I discussed is Dennis Baron’s What’s Your Pronoun? Beyond He and She (Liveright, New York/London, 2020).