1 Masculine, Feminine, Neuter, VegetableGrammatical Gender around the World Ruth Kramer The Christina Kakava Linguistics Speaker Series Good evening everybody! Start with gratitude. Thank you, Paul, for that incredibly kind introduction and thank you for the invitation to speak in this wonderful lecture series. Thanks to all of you as well for coming out tonight – I think grammatical gender is one of the most interesting phenomena in language, and I’m very excited to talk to you (and hopefully convince you) how interesting it is So let’s get going

2 What is Grammatical Gender?feminine fourchette ‘fork’ la fourchette ‘the fork’ couteau ‘knife’ le couteau ‘the knife’ masculine Now you might be wondering (especially if you have less background in linguistics) what exactly I mean by grammatical gender. As with many things in life, it’s easiest to show by example. So let’s take an example of grammatical gender from a language many of you may be familiar with: French. Now French as a language has many nouns, two of which you’ll see up on the screen – we’ve got fourchette ‘fork’ and couteau ‘knife.’ They’re both garden-variety every day nouns, can be used in all the same ways in the language, denote objects used for eating food, they’re quite similar – at least superficially. How many people know French? If we want to say ‘the fork’ what do we say? Now how about the knife? Aha! They’re different now. And that’s the heart of grammatical gender – a way of differentiating nouns, putting them into different classes, depending on how they affect the words around them – in this case the definite article. Grammatical gender is all about noun classification. What are these classes traditionally called? Classes traditionally called feminine for fork and masculine for knife, and we’ll come back to how these terms are connected to male-ness or female-ness… NB: will use “gender” for “grammatical gender” for most of the rest of the talk! French

3 What is Grammatical Gender?Gender is the sorting of nouns into two or more classes, as reflected in the form of words associated with the noun. La petite fourchette Le petit couteau the small fork the small knife ‘the small fork’ ‘the small knife’ Traditional definition up top [STOP AT CLASSES], gender really divides up nouns into different types, think of it if you will as very similar to this object – anybody? Yes, gender is the Sorting Hat of Linguistics…. How do we know as linguists and language learners which gender a noun is in? We can tell if two nouns have two different genders if they take different forms of associated items like definite articles , like we saw in the previous slide, and adjectives [CLICK]. So here we see both a definite article and an adjective for fork and knife in French. Like with the definite articles, the adjectives look different! Petite for fork, petit for couteau. So we see more evidence here that fourchette and couteau belong to two different genders. But now we need to turn to a question that I get a lot working as I do on grammatical gender: so far so good, but why study this?

4 Why is Gender Interesting?Gender varies widely across languages. Gender affects sound, form, & meaning. How nouns are classified gives us clues about the nature of human cognition. SO many reasons, but here are three biggies: See some of that variation today in my talk. Genders vary along many dimensions across languages. -Some languages have many genders, some languages have only a few, some have none. -languages also vary in whether gender is reflected in, say, definite articles or adjectives or verbs or other parts of speech -we’ll see the most today in how languages vary in what semantic properties they use to assign genders ====so all this variation needs to be documented precisely and analyzed (why is it this way?) --- complex problem! Gender also has wide-reaching effects for the grammar of a language. It affects how word are pronounced, it affects the form of words (how you put them together and how you combine them into sentences) and (I will argue) that it also interacts deeply with the meaning of words. So it has a big impact on grammar – not a narrow or limited phenomenon from a grammatical perspectives. Connected to: classifying nouns gives us clues about what is important in human cognition. The fact that language encodes the sorting of objects/concepts in the world surely tells us something about what we consider to be important attributes of objects/concepts and tells us about our cognitive capacities to classify objects and to retain information. So that’s why gender is a highly interesting linguistic phenomenon – today’s talk you’ll see how gender varies, how it affects lots of areas of the grammar and a little bit how it connects to congition. So let’s turn specifically to what I’ll be talking about (Inspired by Corbett 1991: Ch. 1)

5 Outline Question 1: Can we predict the grammatical gender of a noun from its meaning? Yes! Question 2: What semantic properties of a noun can be used to predict its grammatical gender? For the rest of the talk, we’ll start with a couple important empirical questions about gender and how linguistics research has answered them (hoping to build your knowledge base and expose you to major results). So the first question: can we predict the grammatical gender of a noun just from its meaning? Is there some semantic property, for at least some nouns in a given gender, that determines that the noun will be in that gender? YES! Since the first question was answered positively, the second question is well, what ARE those semantic properties?? I don’t want to give away the thrill of discovery here, so for now I will just say TBD and we will see many shortly Empirical questions! Will answer with data from various languages TBD!

6 Outline, continued Question 3 (Big Picture): Why is grammatical gender related to meaning? Question 4 (Big Picture): What is the impact of these results for human cognition? I will wrap up the talk by talking about two bigger picture questions that are tougher to answer but also extremely interesting First of all: -why is grammatical gender related to meaning? In fact, we’ll see that it’s always related to meaning for at least some nouns in every language that has it, and that’s a really striking result – a bit counterintuitive, in fact. Why would it be this way?? What forces it to be this way? And Question 4 (going back to why gender is inherently interesting): I will talk a little about how the result that gender is related to meaning impacts our understanding of categorization capacities in the brain, looking very quickly at how this relates to a research strand on certain types of brain injuries that result in the inability to recognize certain types of entities. So that’s the lay of the land, let’s start with the most basic question which is: is gender related to meaning?

7 Q1: Gender Predicted from Meaning?Knife Fork French Masculine couteau Feminine fourchette Somali mindi mudac Welsh cyllell fforc Spanish cuchillo tenedor Hypothesis: there seems to be no connection between the meaning of a noun and its grammatical gender. At first glance, those of you who know languages like French, Spanish, German and Russian… or know a bit about language in general… may think that it’s not going to get us very far to try and predict the gender of a noun from its meaning. To see that, let’s return to our example from French: knife and fork. Now knife is masculine and fork is feminine. If the meaning of a word predicts its gender, we’d expect the same words ‘fork’ and ‘knife’ to have masc and fem gender respectively in other languages. So let’s take a look. Somali: nope – opposite! Welsh: nope, both fem [elicit] Spanish: nope, both masc All possible combinations of masculine and feminine gender with fork and knife attested! It’s not unreasonable to think that maybe there’s no connection between the meaning of a noun and its grammatical gender. Somali: Orwin 1995: Welsh: online dictionary through University of Wales Trinity Saint David

8 Q1: Gender Predicted from Meaning?Father Mother French Masculine père Feminine mère Somali aabbe hooyo Welsh tad mam Spanish padre madre Revised hypothesis: the grammatical gender of a human noun can be predicted from its meaning. Male → Masculine Female → Feminine But if we look at other nouns, specifically, nouns that refer to humans, the picture looks different, even for the same languages. The word ‘father’ is masculine in all of these languages, and the word mother is feminine in all of them. So, seems like gender of a human noun can be predicted from its meaning – masculine if refers to male human and feminine if refers to female human That’s where the names of the genders came from – people noticing these correlations

9 Q1: Gender Predicted from Meaning?But some languages have human nouns with fixed grammatical gender. Spanish persona ‘person’ (feminine) individuo ‘individual’ (masculine) In the face of conflicting information, what’s a linguist to do? French sentinelle ‘sentry’ (feminine) génie ‘genius’ (masculine) But even this hypothesis has some problems – yet another twist to the facts. Many languages have human nouns with a single grammatical gender, regardless of biological sex. In the face of conflicting information – that gender is sometimes predictable from meaning and sometimes it is not, what’s a linguist to do? How are we supposed to even draw a generalization from these messy facts?

10 Defining the ObservationsGrammatical gender is assigned arbitrarily for some nouns (like ‘knife’ and persona). arbitrary gender assignment Grammatical gender is assigned according to meaning for other nouns (like ‘mother’). Let’s take a step back and define some of the facts we’ve observed semantic gender assignment

11 Defining the ObservationsIn Spanish and French, most human nouns undergo semantic gender assignment. Nouns like persona are exceptions. Is there any language where all nouns are assigned gender semantically? Important fact: in languages like Spanish and French, most human nouns undergo semantic gender assignment Nouns with fixed grammatical gender like persona are exceptions and most of the nouns have the gender you would expect from the biological sex of the entity that they refer to, like we saw with father and mother, but also true for boy/girl, many names for professions like, say, artist can be masculine when referring to a man and feminine when referring to a woman, etc. So we can ask then: if these nouns with fixed grammatical gender are just exceptions, is there maybe any….

12 Gender Assignment in Tamil is SemanticAll nouns denoting women are feminine. peɳ ‘woman’ kaaɭi ‘Kali’ Tamil is a Dravidian language, spoken in India All nouns denoting men are masculine. aaɳ ‘man’ civaɴ ‘Shiva’ All other nouns are neuter. maram ‘tree’ viiʈu ‘house’ Tamil is… All nouns denoting…. You can determine the gender of a human noun and in fact any noun from its meaning – human female feminine, human male masculine, anything else neuter Excellent research website called the World Atlas of Language Structures that documents linguistic variation – did a survey, found 112 langs with grammatical gender /112 = 47% of languages with gender are similar to Tamil in having mostly/entirely semantic gender assignment, though tend not to be European languages taught commonly in US (Tamil data from Corbett 1991:9) Other languages like Tamil: Mangarayi (Australia), Alamblak (Papua New Guinea), Zayse (Ethiopia), Barasano (Colombia), etc.

13 Semantic & Arbitrary Gender Assignment across LanguagesOnly semantic gender assignment Tamil, Mangarayi Semantic and arbitrary gender assignment French, Spanish Only arbitrary gender assignment ??? Is there any language where all nouns are assigned gender arbitrarily? We can now think through the distribution of different types of gender assignment across languages…. There are no gender systems where gender (as we’ve defined it) is assigned arbitrarily with no connection to any semantic property at all. And this is a major linguistic discovery – this gap where by sheer logic we might have expected a language to exist. There is no language that assigns gender purely arbitrarily

14 Answer to Question 1 Every language with grammatical gender assigns gender semantically to at least some of its nouns. Every language has a semantic core to its gender system (major empirical discovery!) Q1: Can we predict the grammatical gender of a noun from its meaning? Yes, for at least some nouns in every language So we found ourselves arriving at an answer to Question 1 now. [read slide] Major result in ling theory – a discovery about a way the world is. Counterintuitive And so for Q1

15 Answer so far: biological sex (female/male)What About Question 2? Q2: What semantic properties of a noun can be used to predict its grammatical gender? Answer so far: biological sex (female/male) Languages vary in which nouns are sex-differentiable: Just humans (Tamil) Humans and certain animals (Spanish, French) Humans and all animals (Amharic) Question 1 had a second part! If meaning is relevant (and we have seen that it is)… So far we have seen that it is biological sex. Male entities are masculine and female entities are masculine Now we have also seen somewhat that languages vary in which nouns they actually consider differentiable in terms of male/female -just humans for Tamil -we have seen only human examples for Spanish/French like mother/father, but certain domesticated or common animals are sex-differentiable as well like dogs and cats for instance (male cat masculine, female cat feminine) -even languages where all animals sex-differentiable But if we assume that languages just draw the line somewhere, we can see that biological sex is the determining factor. For that set of nouns, gender will be assigned semantically by whether they are interpreted as male or female.

16 It is very common to use biological sex for semantic gender assignment.This the most common type of meaning relevant for grammatical gender. Numbers here are from Word Atlas of Language Structures – survey of 112 languages with grammatical gender, 84 used bio sex, and 28 did something else (75% used bio sex) What do the languages in the righthand bar use to assign gender? Animacy!

17 Animacy Example: Lealao ChinantecLealao Chinantec has two genders: (1) -i/-y suffix on adjectives, verbs, etc. (2) no suffix on adjectives, verbs, etc. Lealao Chinantec is an Otomanguean language, spoken in Oaxaca, Mexico Nouns denoting men and women both occur with -i/-y, so these nouns have the same gender. nakɨɁ-i mɨ fallen the.woman ‘The woman was made to fall.’ Mahu-i hmii mɨliuɁ died father of.child ‘The child’s father died.’ two genders: -y, and null men and women are in the same gender – both trigger –y v different from the other languages we’ve seen thus far, where animate nouns with different biological sexes had different genders

18 Animacy Example: Lealao ChinantecNouns denoting animals occur with the -i/-y suffix. dsɨɨ dxú-y dog good ‘good dog’ mɨh dxú clothing good ‘good clothing’ Here is a completely gratuitous picture of a good dog, since as the internet has proven, all things are improved with pictures of puppies or stupid cat videos (no cat videos in this presentation, though… unfortunately!) Data from Rupp 2009 with tones omitted But nouns denoting objects do not.

19 Animacy Example: Lealao ChinantecOverall: Gender assignment in Lealao Chinantec is based on animacy, not biological sex. Animate i/-y suffix Inanimate no suffix Better Answer to Question 2: What semantic properties are relevant to gender assignment? Biological sex or animacy And biological sex or animacy are used in every languages with grammatical gender – one of these two must be present if you have a gender system at all. But we’re not quite finished with exploring what semantic properties can be used to assign grammatiacl gender But that’s not all!

20 Some languages have more than 2 or 3 genders.Data from WALS: 75 languages with 2-3 genders and about 35 languages with more than three genders So in those 33 languages, they can make more distinctions than just biological sex (at most three-way: male, female, neuter) or animacy (at most two-way) And that’s what we find – in addition to biological sex/animacy (they are always used for some nouns), So let’s look at a couple of examples of the kinds of semantic properties that are used to assign gender in these cases In addition to biological sex/animacy, some of these languages use other semantic properties for gender assignment.

21 Mayali Has Four GendersMasculine nakurrng ‘son-in-law’ Mayali is a Gunwinjguan language, spoken in Australia Feminine daluk ‘woman’ Neuter kunwarre ‘muddy ground’ Vegetable mandubang ‘ironwood tree’ Masculine used for male humans and animates Feminine used for female humans and animates Neuter for objects And then a fourth gender: vegetable – used for plants and all things related to plants including plant-based food (except separable parts which are considered objects and put into the neuter gender) Mayali assigns grammatical gender semantically in this case – but based on whether or not a noun denotes a type of plant! (Family claim from Bowern 2005 book review of Mayali grammar by Evans) Data from Evans Brown and Corbett 2002

22 Gender in Bantu Bantu languages typically have 7-10 gendersmarked by prefixes on the nouns referred to with numbers Bantu languages are spoken throughout sub-Saharan Africa. Gender 1/2 is semantically assigned to human nouns. mtu ‘person’ mwenzi ‘friend’ msichana ‘girl’ Swahili (G42) ~500 Bantu languages, Niger-Congo family, divided into zones by letter as you an see on this map of central and southern Africa – that’s where they are spoken Exact number depends how you count, but it’s clear that Semantic core of Swahili Swahili distinguishes human from non-human, a kind of animacy, requirement but different form LC in that many more genders to go (Swahili data from Moxley 1998, map from uni.edu/becker Some of the other genders are also assigned semantically. What properties are relevant?

23 Gender in Bantu Many exceptions! Gender 3/4 is for plants.miti ‘trees,’ mizizi ‘roots’ Gender 7/8 is for tools kijoko ‘spoon,’ kisu ‘knife’ Gender 9/10 is for animals. mbwa ‘dog,’ nyoka ‘snake’ Gender 6 is for mass nouns. maziwa ‘milk,’ maji ‘water’ Null prefix for gender 9/10’ sand, smoke, liquids It’s important to note in the interests of representing the facts fully that there are many exceptions to these generalizations in two directions: -nouns in that gender which don’t have the semantic property (there are some things in 7/8 that are not tools eg) -nouns that have a semantic property which aren’t in the expected class (some animals are not in 9/10) But these loose, not watertight generalizations do hold across the family, and indicate there is probably some type of semantic classification occurring or that occurred in the past in the Bantu family - Bantu linguists have struggled since not black and white, but majority show tendency and many firm So this leads us back to the Answer to Questions 1 and 2 (Data from Moxley 1998, online swahili dictionary, Maho 1999) Gender 14 is used for abstract nouns uhuru ‘freedom,’ ukweli ‘truth’ Many exceptions!

24 Answers to Questions 1 and 2Every language assigns grammatical gender to some of its nouns based on animacy or biological sex. In addition, some languages assign gender semantically to nouns based on other semantic properties (plants, tools, animals, mass nouns, etc.). To close out the talk, we will look at two implications of this -Why? How can this discovery be explained? -Larger impact, especially for human cognition And we’ll take them in this order Only the big picture questions remain: Q3: Why is gender always assigned semantically? Q4: What is the larger impact of this discovery?

25 Introduction to Inflection ClassInflection class: another type of noun classification Russian Class 1 Class 2 Class 3 Class 4 Subject zakon škola kost’ vino Object školu Possessor zakona školy kosti vina Indirect object zakonu škole vinu ‘law’ ‘school’ ‘bone’ ‘wine’ Might think that grammatical gender has to be semantic because it’s about classifying nouns and nouns naturally fall into categories like animate, inanimate But it turns out that nominal classification doesn’t have to be semantic at all – there’s nothing about the inherent fact of classification that forces semantics to be part of the picture To see this – inflection class (aka declension class or conjugation class)! -way of classifying nouns like gender Let’s take a look at how it works in Russian -can see the effects of the four different classes when nouns change form depending on their syntactic position (also known as case) -take one form when a subject, one when object, etc. (if you know Russian, a few of the cases are omitted here for simplicity) -so let’s take the first class STEM IS VIN! -but has no semantics – connected to sound if anything, minimal contrast pair with gender – both noun classification where one is semantic and one is not. is there anything else that is different between them? then that thing might be the reason why gender has to be semantic! (Russian data from Corbett 1991, omitted instrumental and locative cases) -also no agreement! = Slide 2 That’s a clue

26 Inflection Class, continuedInflection class membership in Russian is not determined by any semantic property of the noun. “Semantic features on a noun stem do not suffice to predict its inflection class, that is, none of the four inflection classes correlates unambiguously with a semantic property.” (Alexiadou and Müller 2008) -but has no semantics – connected to sound if anything, minimal contrast pair with gender – both noun classification where one is semantic and one is not. we’re looking to explain why gender is semantic – is there anything else that is different between them? then that thing might be the reason why gender has to be semantic! In fact, inflection class membership is never determined by meaning across languages.

27 Agreement We saw earlier that the grammatical gender of a noun affects the form of words associated with the noun. la fourchette ‘the fork’ le couteau ‘the knife’ The definite article agrees with the noun in grammatical gender. Noun = Controller Definite article = Target Saw earlier that gender affects the form of words around it This is called agreement, like in French definite article/noun agreement French has gender agreement on definite articles Terminology: noun is called the controller since it’s the one assigned gender and controls these features, and the definite article is the target since it receives them or reflects them back to the noun Inflection class never has this effect – while gender agreement is fairly common, inflection class agreement is impossible. This will turn out to be the key to explaining why gender is related to semantics! Inflection class never behaves like this. There is no agreement in inflection class.

28 When a target & controller agree, they agree in a certain feature.Agreement, continued When a target & controller agree, they agree in a certain feature. this dog these dogs The demonstrative (target) agrees with the noun (controller) in number (sg/pl) The number feature on the controller is meaningful but the number feature on the target is not. saw this in French for definite article agrees with noun in gender, see it in English for this dog/these dogs The demonstrative ( = the target) agrees with the noun ( = controller) in number. The number feature on the controller is meaningful holds true across many different types of agreement independent (not connected to gender) hypothesis: for agreement to happen, at least some potential controllers must have semantic features and then we have arrived at an answer to our original question of why gender always has to be assigned semantically for at least some nouns Independent Hypothesis: Agreement in a feature F requires at least some potential controllers to be meaningful for that feature.

29 Why Gender Assigned Semantically?Answer: gender involves agreement, and agreement in a feature (gender) requires that feature to be meaningful on some nouns Two types of noun classification: Gender Involves agreement Meaningful for at least some nouns Inflection Class No agreement Not meaningful on any noun Gender = noun classification involved with agreement = must be interpretable! Two types of noun classification: gender and inflection class Seems like natural language accommodates both kinds of noun classification, and since gender is the kind that agrees, that’s the kind connected to semantics Might be wondering – are there more types are all possibilities attested? Let’s take a brief digression to explore this question

30 Noun Classification in LanguageTypes of Noun Classification Involves agreement? Meaningful on some nouns? Gender Y Inflection class N Cannot exist! Classifiers Here’s another way of making this same point Classifiers are independent words used with nouns that indicate many of the same semantic properties as grammatical gender – animal classifier, a plant classifier, a human classifier and so forth Example from Jacaltec Mayan (spoken in Mexico) Nouns with different classifiers all agree in the same way – classifier is the final type of possible noun classification Asking detailed questions about the semantics of gender leads us to a nuanced understanding of the different kinds of nominal classification found around language Now let’s move on to the final big picture question (Classifier data from Kilarski monograph) xil naj xuwan no lab’a saw CL.MALE John CL.ANIMAL snake ‘John saw the snake.’ Jacaltec (Mayan)

31 Gender and Human CognitionAssumption: Language is a cognitive capacity. It is learned during a critical period Speakers are largely unconscious of their ability to use it Yet speakers use it very well It can be impaired by injury to the brain Language is cognitive, like visual system, critical period for learning it, use it unconsciously and perfectly (not talking about rules like don’t split infinitives but more basic ones like put the subject first and put the wh-word in front of a sentence) but don’t know why it works like that, can have cognitive deficits from brain damage Now that we know what these properties are – we can ask certain important questions -these not others (color, flammability, usefulness, porousness) -animacy/inanimacy for infants These are all questions to be sorted out, but there’s one connection I’m currently investigating that I’ll close with that draws on answering the last question Thus: the semantic properties that are used to assign gender are important to human cognition.

32 Gender and Human Cognition: Key QuestionsWhy are these properties important and not others (color, flammability, etc.)? Are these properties distinguished by infants and children? By non-human primates? By animals? Does this result match with other research on semantic properties in human cognition? Language is cognitive, like visual system or memory system, critical period, use it unconsciously and perfectly but don’t know why it works like that, can have cognitive deficits from brain damage Now that we know what these properties are – we can ask certain important questions -these not others (color, flammability, usefulness, porousness) -animacy/inanimacy for infants These are all questions to be sorted out, but there’s one connection I’m currently investigating that I’ll close with that draws on answering the last question

33 Gender and Human CognitionHypothesis: conceptual knowledge is organized in the brain by the properties used to assign gender. Evidence: brain injuries can selectively impair the ability to recognize entities with these properties impairment for animate entities impairment for animals impairment for plants That’s a big task, but here’s a tantalizing clue that we’re going in the right direction Shown a picture of a dog and asked to name the representation of the picture This hypothesis is yet to be confirmed, but if it works out, it suggests that our brains are organized via the very same semantic properties that are used to assign grammatical gender TBD: impairment just for humans? Just for male/female entities? Mass nouns? Abstract nouns?

34 Conclusions Answer to Question 1: Every language assigns grammatical gender to at least some of its nouns based on their semantic properties. Answer to Question 2: These properties include biological sex or animacy as a minimum, and can also include plants, tools, animals, mass nouns, etc. Q1 was: can we predict the grammatical gender of a noun from its meaning? If you take away one thing, yes!

35 Conclusions, continuedAnswer to Question 3: Gender is assigned semantically to at least some nouns because gender is defined by agreement. Answer to Question 4: The semantic properties used to assign grammatical gender may correspond to the categories of conceptual knowledge in the brain. …and agreement requires that a feature be meaningful on at least some nouns Grammatical gender is deeply connected to our ability to sort and understand the world as human beings

36 Thank you!