Semantics of Words Lecture

305 lecture on semantics of words. This is in preparation for the problem set that is due Monday, March 31st, when we get back from spring break. So you're probably on your spring break right now, in fact, and I hope you're enjoying it. I'm actually in San Francisco. I'm in the process of moving to San Francisco, and so I'm sitting here in a nice little park in the Levi Strauss Plaza, and so if you hear some traffic noises and things in the background, that is why.

hope you're enjoying it I'm actually in San Francisco I'm in the process of moving to San Francisco and uh so I'm sitting here and a nice little Park in the Levi Strauss uh Plaza and uh so if you hear some traffic

So let's see. Why don't we go back to this picture that we looked at last week, when we looked at the lexicon? Back then I referred to this idea that language is, at its core, a relationship between sound and meaning. This is actually not my idea. This is the idea of Ferdinand de Saussure, a Swiss linguist who is kind of the father of modern linguistics. So we had this picture where language is at the interface between the world of sound and the world of meaning, and I illustrated the world of meaning with a little thought bubble down there at the bottom. We've covered all these areas of theoretical linguistics: phonology, morphology, syntax, and all that. What's left is semantics. So we're going to look at how this interface between language and thought, or language and concepts, works. I had a lecture called "Language and Thought," and I decided to cancel that. It's not really essential to core linguistics. So we're just going to plunge ahead into the semantics of words today.

that what language is at its core is a relationship between sound and meaning this is actually not my idea this is the idea of Ferdinand sosur a Swiss linguist who's kind of the father of modern Linguistics so we had this kind of

This is an interesting topic because it gets us out of linguistics to some extent, and it gets us into related areas, in particular philosophy and psychology. This area of semantics, the study of meaning, is associated with those two fields just as much as with linguistics. So let's start with some philosophical issues having to do with word meaning. For example, what is word meaning really, at the basic abstract level? How do we get this mapping that Saussure talked about between the vague, amorphous world of sound and the equally vague and amorphous world of meaning? How does this mapping come about? In particular, how do certain sets of phones that make up words come to be associated with particular concepts? How does the word "tree," that sound, come to be associated with trees? Philosophers have different ideas about this.

of theoretical Linguistics now phology morphology syntax and all that's left is semantics so we're going to look at how uh this kind of this interface between language and and thought or language and Concepts I had a lecture called language

Here I've put up a quote from Humpty Dumpty. This is from Lewis Carroll's *Through the Looking-Glass*, and Humpty Dumpty says, "When I use a word, it means just what I choose it to mean, neither more nor less." So that's the idea that he can have any association between sound and meaning that he chooses. We might call that an idiosyncratic theory of meaning, but that's just a non-starter. It would never work, because everyone would have their own separate mapping from sound to meaning, and that's not how language works. Language works by social convention. Everyone has to agree that "tree" means tree, "dog" means dog, and "Google" means Google, and so on. That's the only way language could work socially. We can't have our own private kinds of meanings. So the idiosyncratic approach we can rule out right away. But there are two other approaches that were discussed by Plato. This is the oldest recorded linguistic discussion, at least in the European tradition. It's a Platonic dialogue called the *Cratylus*, written around 400 BC, and it's a debate, as these dialogues always are, between Hermogenes and Cratylus about this very question: how do words acquire their meaning? You can find it online and read it yourself if you want. Hermogenes takes what's called the conventionalist position, which is sort of what I just said: society as a whole decides on this relationship. We all agree that "dog" means dog and "cat" means cat. That's an approach we could call conventionalist, and it posits an arbitrary relationship between sound and meaning. It's arbitrary in the sense that it could have gone either way. We could have decided that "cat" means dog and "dog" means cat. We could have switched the sounds around. There's nothing in this view that's particularly doglike about the word "dog," and so on. I think this is the most sensible position to take regarding language, and it was also the position of Saussure. Think about it: if this weren't the case, then all the languages of the world would be the same. If "dog" naturally meant dog, then all the languages, Swahili and Japanese and Greek and French, would all have a word that sounds like "dog."

interesting topic because it gets us out of linguistics to some extent and it gets us into some related areas in particular philosophy and psychology so this is uh this area of semantics which is the study of meaning is really associated

The alternative position is called the naturalist position, and it was advocated by Cratylus in this dialogue. He said no, there is actually a natural relationship between words and concepts. In other words, certain words are better than others, and certain words for dog are more doglike than other words. So there's an ideal word for dogs, an ideal word for horses. That's not a surprise, coming from a Greek philosopher. The ancient Greeks, the Platonists, believed in essentialism: there's some essential horseness to a horse and essential dogness to a dog, and so on. But these days that position seems pretty untenable. As I said, if there were an ideal word for dog, then every language would use that word, but of course they're all different. Spanish uses *perro*, Japanese uses *inu*, French uses *chien*, German uses *Hund*, and in English we use "dog." None of them seem to be related. That fact about the world's languages seems to support the conventionalist position: there's basically just an arbitrary relationship between sound and meaning that constitutes words. Arbitrary, that is, but still consistent with the phonotactics of the language, the syllable structure, the phonological structure, and so on.

abstract level how how do we get this mapping that soir talked about between this vague amorphous world of sound and the equally vague and amorphous world of meaning how does this mapping come about and in particular how do certain

All right. Now that we've examined the basic philosophical relation between sound and meaning, we can look at some other ideas about meaning that come from the philosophical tradition. In particular, we can distinguish two kinds of meaning. This is a distinction made in chapter 7 between reference and other types of meaning. So let's look at reference first. Reference is a kind of meaning where a word refers directly to some object in the real world. Imagine a proper name like "Barack Obama." What does "Barack Obama" mean? It really refers to the person, the guy who sat in the White House. You could imagine a piece of fishing line attached to the word on one end and to the guy Barack Obama on the other. That's roughly what reference is: a direct connection between a word and some entity in the world. It's a really good theory of meaning, at least for proper names like "Barack Obama," "Microsoft," "San Francisco," and so on. These proper names really do refer directly to objects in the world, and that kind of meaning is called reference. But meaning can't be that simple, because we have words that don't seem to refer to objects in the world. Verbs and adjectives, for instance, don't refer to entities the way "Barack Obama" does. Even common nouns like "chair" are tricky. You could have a reference theory that says the word "chair" refers to the set of all chairs in the world, or something like that, but it's probably better to just say th

ideas about this uh here I I put up um a quote from Humpty Dumpty this is from leis Carrol's Through the Looking Glass and Humpty Dumpty says when I use a word it means just what I choose it to mean

neither more nor less so that's that's the idea that he can just have any association between sound and meaning that he chooses we might call that an idiosyncratic theory of meaning but that's you know that's just a non-starter that would never work

because then everyone would have their own separate mapping from sound to meaning and that's not how language Works language works by social convention everyone has to agree that tree means tree and dog means dog and Google means Google and so on that's the

only way that language could work socially so we can't have our own private kinds of meanings so the idiosyncratic approach we can rule out right away but there are two other approaches that uh were discussed by Plato this is the oldest recorded

linguistic discussion in in at least the the European tradition this is a dialogue uh of platon dialogue called the catalus was uh written down in 400 BC and it's a debate as these uh these dialogues always are between hermogenes

and cratus and about this very question how do words acquire their meaning and in this discussion uh you can find it online and read it yourself if you want but hermogenes takes What's called the conventionalist position which is sort of what I just said that

you know society as a whole decides on this relationship we all agree that dog means dog and cat means cat so that that's an approach that uh we could say is conventionalist and and it it posits an arbitrary relationship

between sound and meaning so you know it's arbitrary in the sense that it could have gone either way you know we could have decided that cat means dog and dog means cat you know we could have switched around the sounds there's

nothing in this view that's particularly doglike about the word dog and so on and that I think is the most sensible position to take regarding language and this was also the position of soer Because if you think about it

you know if if um if this weren't the case then all the languages of the world would be the same you know every if dog naturally meant dog then all the languages Swahili and Japanese and Greek and French they would all have have a

word that sounds like dog and the alternative position is is called the naturalist position and this was advocated by cratus in this dialogue and he said no there's there's actually a natural meaning between words and Concepts so in other words certain words

are better than others certain certain words for dog are more doglike than other words and so there's this kind of uh ideal word for dogs an ideal word for horse and that's that's not a surprise that uh a Greek philosopher would

Advocate a position like that remember the ancient Greeks they were the platonists believed in essential essentialism you know that there's some essential horseness to a horse and essential dogness to a dog and so on but these days that that position

does seems pretty untenable right because like I said uh if there was a ual word for dog then every language would use that word but of course they're all different you know Spanish uses perro and and Japanese uses Inu and

French uses Shan and German uses H and in English we use dog and so on they're they're and none of and they don't seem to be related so that that fact about the world's languages seems to support the

conventionalist position that you know there's basically just an arbitrary relationship between sound and meaning that constitutes words arbitrary that is you know but still consistent with the phonotactics of the language you know the syllable structure and phonological structure and

so on all right so now that we've examined that issue about you know the basic philosophical relation between sound and meaning we can look at uh some other ideas about meaning that come from the philosophical tradition and in

particular we can we can distinguish two kinds of meaning this is a distinction they make in chapter 7 between reference and other types of meaning so let's look at reference first reference is a is a kind of meaning where a word

refers directly to some object in the real world you could imagine a I don't know take a take a take a proper name like Barack Obama right and then imagine what Barack Obama means it really refers to the person the guy who's sitting there

in the White House you could imagine a sort of a this you know I don't know a piece of fishing line or something attached to the word on one hand and then to the guy Barack Obama on the

other hand and that's that's sort of what reference is like it's a direct connection between a word and some entity in the world and that's a really good theory of me meaning it at least for proper names like Barack Obama

Microsoft San Francisco and so on these proper names really refer directly to objects in the world and that that kind of meaning is called reference but meaning can't be that simple uh because you know we have words that don't seem to refer to objects in

the world you know especi you know things like uh verbs and adjectives and things like that they don't refer to entities in the way that Barack Obama does even common nouns like chair uh you know you could have a

reference theory that says that the word chair refers to the set of all chairs in the world or something like that but uh it's probably better to just say that well there are other kinds of meaning besides reference and that gets us into

the second approach conceptualism and that says in addition to just words referring to entities in the world uh we have this idea of Concepts which are not directly entities in the world but refer to abstractions in the mind uh these are things in the mind so

maybe the chair uh from a conceptualist point of view refers to the concept the general concept of a chair and this seems more plausible for some types of words like alive and near uh that are very kind of complex in

their meaning and they they seem to denote Concepts rather than entities in the world so we still we need both of these kinds of meaning let me tell you a little bit about uh frga there's a philosopher named uh gab fra who considered this

kind of issue and he he he posed a a a paradox that explains why we can't get by with just a theory of reference where words refer to entities in the world and he he cited this uh example about the

ancient Greeks so the ancient Greeks um had uh names for these entities up in the sky and one they called Hesperus this was the evening star and when you walked out after dinner you would see Hesperus up in the up in the sky a very bright star

that was the first star to appear in the evening and then they would go to bed and they would wake up in the morning and they'd go out outside at dawn and they would see phosphorus phosphorus was a a very bright Morning Star they called

it the morning star and it was in a different area of the sky so for a long time the ancient Greeks believed that Hesperus and phosphorus were different Stars until the uh great mathematician Pythagoras he figured out that actually

they are the same star and in fact they're the planet Venus what we today know of as the planet Venus and so of course it's uh moving and that's why it's in two different positions and so anyway look at sentence

one here this is what Pythagoras discovered Hesperus is phosphorus and what frga noticed is that you know this was a great discovery but if if our theory of meaning is just a theory that of reference you know where hesperis just

refers to the actual entity in the sky and phosphorus the same sentence one would be equivalent to sentence two hesperis is hesperis because again if if the theory is just reference then you can just substitute the entity for the word this

this under the theory of reference in other words one and two have exactly the same informational content and frga pointed out that that's kind of a ridiculous result I mean sentence one is a great discovery and sentence two is

just a tautology and it has no informational value whatsoever so that led uh fragga to propose that we need two kinds of meaning in other words reference is not enough because it leads to these kinds of paradoxes and so he

proposed two kinds of meaning reference and sense and you can read in chapter 7 they have a nice discussion about sense and reference and the difference between them but essentially as I mentioned reference is the relationship between between words and entities in the world

and sense is more about Concepts it's about our mental images of words and so on you might think of Senses as kind of like dictionary descriptions so the sense of a chair is something like you know something you sit on that has four

legs or three legs and a back to support you and the sense of the word ant is uh the sister of one of your parents and so on it's sort of like dictionary definitions and that's a that's a second kind of

meaning so reference is a direct relationship between a proper name and some entity in the world and then sense is a more abstract kind of meaning that has to do with mental Notions or concepts of words and so fr's point was

once you have this kind of Distinction you can make sense out of these paradoxes like Hesperus as phosphorus you could say something like even though Hesperus and phosphorus have different senses you know one has the sense of being the evening star and the

other is the Morning Star they happen to have the same reference or if you don't like that Hesperus phosphorous stuff think of more contemporary examples think of uh say Clark Kent is Superman that's really an equivalent sentence so uh you know we have two

senses for those two proper names Clark Kent is the mild-mannered reporter for the Daily Planet and Superman is the Man of Steel and they're very different in their senses and then at some point Lois Lane figures out that even though these

terms have different senses they actually have the same reference they refer to the same entity in the world so I think fr's distinction here is very useful because it helps explain sentences like that like Clark ktis Superman so even though they have different

senses they have the same reference or take a sentence like there is no Santa Claus and if you analyze it from this point of view you say what that means is that even though Santa Claus has a sense

you know this guy who lives in the North Pole and delivers presents to children on Christmas uh nonetheless he has no reference there is no actual entity in the world that that Santa Claus refers to so having both sense and reference is

really useful for your uh for your theory of meaning there all right so let's let's move on from the these philosophical distinctions like sense and reference and look at some of the uh ideas from psychology and uh one thing that

psychologists like to point out is that uh common nouns in particular seem to correspond to C categories in the human mind so these words like mountain and apple and tree and so on those really denote categories and objects can be a

member of that category an individual object is an apple and there are many apples and we when we see an object we kind of put it into these categories we say oh that's an apple and that's a pair

and these distinctions are based on perceptual cues you know texture size color uh function these kinds of things and these categories furthermore are not ABS they're they're they're organized they're not just floating around by themselves they form these

kinds of taxonomies I put an example of a taxonomy here on the screen uh just a very simple toy taxonomy here but it shows fish and bird and so on and and uh these hyper nms and hyponyms the uh entities that are above

and below in generality so a shark is a kind of fish and a fish is a kind of animal and a bird is a kind of animal and and so on and these have different properties and it kind of makes sense

that humans would organize their categories this way right because it it supports inference that's the whole point of this you know we've evolved to you know make these kinds of inferences like oh if if this is a kind of fish

then that means I can eat it and it swims and it has all these properties that that fish have so it's very useful to have these kind of inferential properties associated with our categories let's see here's another example from wordnet this big thesaurus

you can download from the Linguistics department at Princeton University uh this shows a just a small fraction of their big taxonomy that they constructed including natural versus artificial objects and things like that uh I'm going to cite here some work

by a psychologist named Elanor Ro who's done a lot of work on semantics of words and and these ideas about categories she points out that you know within a taxonomy there's a particular layer of that taxonomy in other words a

particular level that seems to be kind of optimal and uh optimal in a couple senses one is that this this is the level that children learn so if you look at the objects that I've pictured here on the slide you know if you ask a child

or an adult for that matter to identify those objects they would say well that's a rabbit a car and a chair and that's exactly the level of generality that she's talking about so we don't get more specific or or more General right we

wouldn't just refer ref to it as a vehicle or furniture or mammal because those just aren't as useful and when children are learning language they seem to be hardwired to kind of go for this uh basic level category of rabbit car and

so on she's her idea was that this is kind of useful because this is the level at which you can maximize your inferences you know so if I tell you something is a rabbit uh you really it's you that's the the best kind of

information you can have for making decisions about that if I told you it was a mammal that's just too vague that's not enough uh information for you another contribution of Elanor Ro was um what's called prototype Theory and this was her observation

that the entities that make up these categories are not at an equal level some are better than others in other words some entities are better Birds than others they're they are more birdlike and uh Furniture not all furniture is

created equal right some types of furniture are better examples like chair and sofa those are excellent examples of furniture and then other things you know are maybe sort of like Furniture but not quite you know like a I don't know like

you have a computer a monitor stand or something is is that furniture eh not really not in the sense that that a chair is and same with birds you know if I ask you to think about a bird you'd

probably think about a robin or a Jay or a dove or one of these birds that you know the so-called passerines that perch and sing and uh feed worms to their young and and things like that and you're probably less likely to think

about a penguin or an ostrich which are technically Birds but they're not very good birds are they they're not they're not prototypical Birds they don't you know Penguins can't fly in the air like most birds can and and so on so uh this

is the idea that some exemplars are better than others in these different uh categories this is she was building on philosophical work by viken Stein this guy and he he pointed out the same thing about how some categories like he had

the example of a game are pretty fuzzy you know it's it's hard to pin down a really good definition of what a game is and but there are other kinds of situations where we do have very hard and fast rules for what counts

as a bird or what counts as a game and so on so this is I I I put it on this slide here as the the two kinds of mental categories that we use so the first one is rule-based categories

and sometimes when we use a word we have a very strict definition in mind in in certain contexts like say a legal context or mathematics when we're doing mathematics we can't be kind of vague and fuzzy we have to be exact so these

mathematical Notions like an odd number or a prime number they have very specific definitions and every number is going to either be prime or not or odd or even and so on and you know sometimes in a legalistic setting we'll have strict

definitions of what counts as say a a grandmother you know you when in a in a legal situation when a grandmother is trying to get custody of her grandchild you know the court may want her to prove that she's a grandmother which means

that she's the mother of the of the parent of the child or something and then for most situations we kind of get by with fuzzy categories this is what Ellen R Ro was talking about you know the prototype categories

of bird and Tool and furniture and so on vegetable you know usually we just kind of think of a vegetable as something we throw in a salad and we don't really use a strict rule-based definition like a botanist would use or something like

that and both of these kinds of ways of thinking about meaning are useful you know it just depends on the situation you know there's a fuzzy notion of a grandmother which is a kindly woman who has gray hair and

bakes cookies for you or something like that and then there's some some situations we think of a grandmother in a you know more rule-based context where you have to prove that the that it's the mother of the parent of the child and so

on so another issue that comes up in uh chapter 7 they spent a lot of time in chapter 7 talking about relations between words and uh four in particular they talk about are we already mentioned hyper nms and hyponyms these are terms

that are more or less General than other terms in a taxonomy uh and others that uh you've I'm sure heard of are uh lexical relations are synonyms and antonyms uh here I put an example of some synonyms from wordnet this uh

thesaurus that I mentioned that comes from Princeton and they uh they essentially list the senses of each term in their database by linking it to other words that have similar senses and you can see some of those links here the the words

themselves are in red and then the links between them show some of the similar senses so these are words that are at least Le in part uh synonymous with each other there's a question about whether any two words are truly synonymous in

English or any other language but you might say that you know you can think of a Continuum of synonymy you know uh words are kind of more or less synonymous they have shared meaning let's let's say I'll mention I won't mention anonyms

but uh you should read uh some of the discussion of antonyms in Chapter 7 the different kinds of antonyms they talk about complimentary and uh uh what were the other terms uh Converse was another term I put some

questions on the problem set about those kinds of antonyms gradable was another one uh so some of these adjectives like tall and short we think of as antonyms uh but they're gradable because uh just because you're not tall doesn't mean

you're not short you might be sort of somewhere in between whereas other other types of uh antonyms are complimentary you know if you're not male then you have to be female and if you're not dead then you have to be alive and so on and

all this uh I'll let you read about that in uh chapter 7 so uh another final topic to look at here is this notion of dxis this is very important this will come come up uh later in the course when we talk about

child language acquisition this is a special kind of meaning it's it's a kind of reference remember reference was meaning that Associates a word with some entity in the world and um dxis is a kind of reference it's called dxis by linguists

and it's called indexical by philosophers but these are words that are interpreted in con context they include pronouns these personal pronouns like I and you and he and it and so on and notice the meaning of those pronouns

is not fixed right it depends who's making the utterance so when I say I it's very different from when you say I and when I say it it depends on the situation what it refers to and so words

that behave this way are called dtic terms or indexical this uh comes from the Greek word uh for index you know like your index finger for pointing these are words that point uh in other words and they can

point to different things in different contexts every language in the world seems to have these kinds of dtic forms that is forms that don't have an absolute reference but a changeable reference that depends on the context and people have studied you know

what they call dtic systems in different languages um English and Spanish for example tend to be very me centered you know we have words like dtic terms like here and there in English and Spanish and it and it really refers to me the speaker when

something is near me I say it's here and when something is away from me then I say it's there and other languages like gave examples here from Fijian they uh take the speaker into account they have different dtic terms

for near you so there might be a there that means there by you and then there's yet another there that means they're away from both of us for example so that's a that's an example of a dtic system uh in these different

languages all right well let's um let's review some of these ideas that we looked at then so first was this notion of reference the reference was this direct relation between a word and some entity in the world and uh we saw this

is a pretty good theory of meaning at least when it's restricted to proper names like Microsoft and Barack Obama so we say that the word Barack Obama refers to the actual entity in the world pronouns are like this as well you know

pronouns refer to entities in the world uh but that's not enough to have a a theory of reference you also have to have a theory of sense so you need two kinds of meaning and words like you know tradition and

alive and near uh human these uh may or may not have a reference but they certainly also have a sense that is a uh sort of a cognitive content to them that can be captured in these dictionary definitions and what frga pointed out is

that we need both kinds of meaning we need to talk about sense and reference as two distinct and equally important kinds of meaning and then we looked at uh these categories we saw that common nouns like chair and dog and tree they corresponds

to these categories and we saw some of the ideas about categories how you have hyps and hyponyms and uh prototypical categories you know these they're they're kind of membership in a category is kind of fuzzy fuzzy right um

you know tomato is kind of a vegetable in some sense it I don't know a botanist might argue that it's really a fruit or something and so some of these membership in these in these categories can be um a little

problematic sometimes and we saw that certain exemplars are better you know a robin is kind of a in a sense a better example of a bird as more prototypical bird than say an ostrich or a penguin and then finally we looked at these um

dtic terms so this is a special small class of words personal pronouns like I and me and you and him and uh other expressions like here and there now yesterday tomorrow and notice that their meaning is not fixed it changes every

time you use it so the word tomorrow changes meaning uh depending on the day that it's used and so on and uh so this this is a this is an important distinction that'll come up later in some future uh problem sets these these uh

dtic dtic uh terms well that should be enough to uh get you going on your problem set there on the semantics of words and we'll get together on Monday March 31st and we'll look at semantics of larger units we'll move up to phrases

and sentences and look at some of the studies from philosophers and psychologists about meaning that's associated with propositions and sentences and and larger units of meaning all right hope you had a great spring break and I will see you Monday

morning