Session 11: Lexical Semantics

While analysis of the semantic content of an utterance is possible using differently sized chunks of language (phrases, sentences, entire texts), it is common to start on the word level and to examine words that intuitively seem to “go together”. Drawing up a map of sense relations is possible only after developing terms to describe these relations. The technical vocabulary explained below is used in lexical semantics to describe the relationship between terms. Are two terms neighbors? Opposites? Do they have a part-whole relationship? Lexical semantics has the goal of answering such questions.

Word fields

As has already been discussed, semantics is concerned with meaning. One way of defining meaning is by looking at the relationship of a group of terms in unison. Do they “go together” or not? Have a look at the following examples:

eyes, hands, nose, feet

green, red, purple, yellow

dog, log, hog, fog

While the terms in the first two sets are all related to one another (they form a word field), the words in the third set make up an arbitrary mix. This is likely to be the impression of most native speakers – dog and log simply have nothing in common in terms of meaning -, but it underscores a point we made very early in this course: the arbitrariness of the sign. The words in the third set share an identical sound pattern (save for the initial phoneme), but their meaning does not reflect this in any way.

Word fields as they are described above aren’t a purely theoretical exercise. Sets, an experimental tool developed by Google can automatically predict a word field based on very limited human input. Try it yourself here.


Synonomy is the degree of sameness (in regards to meaning) that two terms share. Natural languages afford fairly little space for complete synonyms (that would not be economical) and accordingly, small meaning differences exist. Buy and purchase are an example for two near-complete synonyms. In purely semantic terms, both words mean the same thing, but their use depends on the context they are used in. Purchase is likely to be used in slightly more high-brow language, whereas buy is the more common (in both senses of the word) variant. English has a fairly high number of (near) synonyms because of the influx of French words into the lexicon.


Antonyms are binary opposition pairs such as happy – unhappy, tall – short, young – old, war – peace. Their decisive quality is that the meaning of one term automatically excludes the other – someone who is tall is not short and someone who is unhappy is not happy. Antonyms can be gradable or non-gradable, depending on whether or not we can attach inflectional morphemes to them to indicate a comparison (happy – happier – happiest vs beautiful – *beautifuller – *beautifullest)


Hyponomy describes hierarchical relations between terms. If we can say that X is a kind of Y, a hyponymous relationship exists between X and Y. The two examples below illustrate this kind of connection.

blue red green yellow purple white black

toast boil fry grill roast bake microwave

In the examples, the terms color and cook are superordinates, while the words listed below them are their hyponyms.

Related to this is the concept of meronomy, which describes part-whole relationships. A meronymical relation is slightly different from a hyponymous one: eyes, lips and nose are part of the face – they are not a kind of face.


Homophones are terms that have a similar sound pattern, but are otherwise unrelated. Examples for this are see – sea, buy – bye, might – mite, night – knight. When two terms are spelled similarly but the sound patterns differ, we speak of homographs. An example for a pair of homographs is wind, as in we wind up in the same club every weekend vs. the wind is very cold in December. When both pronunciation and writing are identical, linguists conventionally speak of homonyms (see below).


Homonyms are terms that are superficially identical (in speech and writing) but etymologically unrelated:

match = thing that you light a cigarette with
match = thing that a soccer team loses

date = a sweet kind of fruit (ger: Dattel)
date = an appointment

Note that homonyms are characterized by the fact that they look the same superficially, but are actually unrelated. Usually the etymology of a word is key in determining whether it is a homonym.


In contrast to homonymity, which describes separate words with different meanings that only happen to look similar, polysemy describes individual word with multiple and distinct senses (polysemes). The term bank, for example, can denote either the institution or the building in which the institution resides. Both meanings are associated with the same word, making bank polysemous. By contrast, a river bank is not a different meaning of the same term, but a different word entirely.

Conceptual metaphors

While the abovementioned descriptions are use to describe sense relations, conceptual metaphor is a model that aims to explain how human cognition deals with certain aspects of meaning. Based largely on ideas put forth by George Lakoff and Mark Johnson in their book Metaphors We Live By, a conceptual metaphor is an expression from ordinary language in which the meaning associated with a target domain is drawn from a source domain that is (subconsciously) perceived as sharing certain traits of the target (TARGET is SOURCE).

These examples serve to demonstrate the idea:


You make my blood boil

Let her stew

She got all steamed up

He’s just blowing off steam


She spends her time unwisely

The diversion should buy him some time

Time is money


Sally gave the idea to Sam

Sally took the idea from Sam

Sally traded ideas with Sam

Sally has an idea

Many more examples are available on George Lakoff’s website.

Key terms

  • sense relation
  • word field
  • synonymy
  • antonomy
  • hyponomy
    • meronomy
  • homophone
  • homonyms
  • polysemy
  • conceptual metaphor
    • source domain
    • target domain
Written by Cornelius in: Semantics,Summaries |

Session 10: Essentials of Semantics

One reoccurring theme in virtually all of the units that we have covered so far has been the focus on internal structure (for example of words and sentences) in contrast to meaning (i.e. what a piece of language tells us about the world). You may be relieved to learn that with semantics, we are finally venturing into the domain of meaning.

What’s the meaning of to mean?

Meaning as a concept is initially more difficult to define than you might think. The verb mean itself serves as an example for the different meanings a single word can take on:

Sorry, I didn’t mean to offend you

Great — this means we’ll have to spend another hour in the car

Dog means ‘chien’ in French

In the first example, the meaning of mean is roughly equivalent to ‘intend’, in the second it means ‘it is the consequence of something’ and in the third it is equivalent of ‘dog translates into chien‘. The examples show that an extremely common word like mean can easily be used to describe very different things.

Conventional meaning vs. social and affective meaning

Not only can words be used with different meanings in different contexts, but the entire description of what something means depends greatly on the type of expression we are talking about. Compare the following examples:

Beagles are a breed of dogs

Hey Stan, how are you?

Linguistics is really cool!

The first sentence differs from the second and third in that it makes a statement about the world that can be verified or falsified. By contrast, if someone states that he/she likes or dislikes something (the third example) this is a subjective and unverifiable statement. Such an expression still contains important information, but it has what we call affective meaning. Similarly, a question such as the one provided in the second example has social meaning, as do words we use to address people (Mister, Misses, Sir, Your Honor, Dude – also think about Sie vs. Du in German), ways of greeting and saying goodbye (Hi, Cheers, Regards, Wassup) and many other parts of language which are essential in our everyday interactions with others. Note that what could be called a third type – grammatical meaning – has already been discussed.

Semantics, however, is concerned purely with the conventional meaning of words and sentences. Conventional (or sometimes conceptual) meaning can be described in almost mathematical terms and it can be applied to sentences that we can often evaluate in terms of their truth value. Beagles are a breed of dogs is such as example: it can be described as being either true or false. Conventional meaning also captures the kind of information that content words carry (man, New York, run, happy) but not expressions whose meaning is purely social or affective (Yeah, cool, Hi, Regards) or grammatical (and, the).


Reference (or referential meaning) is one of the most basic ways of thinking about meaning. The referent of an expression is essentially ‘the thing in the world that it points to’. Thus

Barak Obama

Berlin, Alexanderplatz

Sally’s cat

These old CDs

are linguistic expressions that have definite referents, although they might not be entirely clear without knowing the context (more than one person can be called Sally, she could have more than one cat and what CDs are meant exactly in the last example depends on the situation). What counts in terms of referential meaning is the assumption that a specific cat owned by a specific Sally is meant with Sally’s cat and that with access to the context it would be clear what CDs are signified by these old CDs.

Multiple expressions can point to the same thing in the world, as the list below demonstrates.

  1. Hillary Clinton
  2. the former first lady
  3. the wife of Bill Clinton
  4. the United States Senator from New York
  5. the winner of the Ohio primary election

All of these expressions have to the same referent, yet 3 and 4 could (at least hypothetically) change at some point in time. Referring expressions and referents have a dynamic relationship – it can change as circumstances change.

By contrast, the underlined expressions below never point to anything in the world, regardless of the context they are used in (they are not referential):

A bunch of people called

He is buying a new car

Nobody showed up

Elephants are native to Africa

All of these examples have in common that no definite referent exists for them, yet they certainly mean something.


As the examples above show, it is necessary to capture meaning beyond reference. Some expressions do not have a definite referent, while others cannot be described as referential because they point to something fictional. For example

Santa Claus

Bilbo Baggins


my shiny new spaceship

all have meaning, but none of them has a referent in the real world. The term sense is generally used to describe the conventional semantic meaning of an expression that is separate (or ‘goes beyond’) its reference. Sense is also used to describe the semantic content of expressions that describe activities, states, qualities, attributes, relations etc – as you may have noticed, reference essentially only applies to nouns. Still an adjective like heavy or a preposition like under has a conventional meaning and can be defined in terms of their relation to other terms (for example, we could describe heavy as the opposite of light and under as relatively similar to below). We’ll see in the section on lexical semantics that an expression is not limited to just once sense, but can easily take on multiple senses (remember mean?).

Denotation & Connotation

The terms denotation and connotation are useful to separate the literal, value-neutral and restricted sense of an expression from its figural, cultural or associative meanings. For example, the word pig simply denotes a specific animal, but the connotation of the term is often negative and it can be used in a figurative way, for example, to describe a person. The connotation of a term depends on the usage community’s values and beliefs, whereas denotation does not.


An expression’s extension is the sum of all senses and referents to which it applies. For example, the extension of the term man is roughly 50% of the world’s population; the extension of Mike’s friend would include whomever Mike is friends with.

Semantic anomaly

You might wonder why we even have to make a distinction between conventional, social and affective meaning. Since they are all important to language users, why be so specific about it? The reason is that, like expressions that are ungrammatical, expressions that are semantically malformed are not just subjectively problematic, or wrong in certain contexts (or go against the taste of some speakers), but simply do not work and are not used by speakers. The example below illustrates the issue:

John likes basketball

”The table likes basketball

If you find the second sentence strange (and you should) it is because it is semantically anomalous – it doesn’t ‘make sense’ (I’ve used two single quotes to mark this, a bit like the asterisk used to denote sentences which are ungrammatical). Noam Chomsky’s famous example sentence Colorless green ideas sleep furiously is another instance of an expression that violates the constraints of conventional meaning. There’s nothing wrong with the sentence in social or affective terms, nor is it ungrammatical, but any reasonably competent speaker of English could tell you that it is inherently illogical. What we know about the world simply tells us that ideas can’t sleep or be green and that the table, Madagascar or world peace are generally not acceptable subjects of the verb like while Mary, those boys and nobody are. As will be shown below, this interface between ‘what we know about the world’ and language can be described quite systematically.

Semantic roles

Among other things, semantic roles allow us to explain why an example such as the one above (”The table likes basketball) is semantically anomalous. Loosely speaking, semantic roles describe ‘who does what’ in a sentence and they are often discussed in concert with syntax because they exist in parallel to syntactic roles. A classical example for how semantic roles function is passive voice:

John hugged Sue (active)

Sue was hugged by John (passive)

The blue word in each sentence is the subject of main verb hugged, the red word is the object. You’ll notice that the passive sentence does not have an object (hugged has become intransitive) and that John has been ‘stored away’ in an optional adverbial (by John). In other words, the subject of the sentence has changed as the voice has switched from active to passive.

But what about the meaning?

Clearly it is still Sue who is hugged by John, not the other way around. Semantic roles allow us to describe this dimension of ‘who did what to whom’. Here’s the example again, but this time highlighting the semantic roles:

John hugged Sue (active)

Sue was hugged by John (passive)

The two roles marked in green and yellow are called agent and patient and they stay the same when switching from active to passive because the meaning of the sentence does not change. John is still the one doing the hugging and Sue is still being hugged – while syntactically there is a switch, semantically there is no change.

Below is an overview of some essential semantic roles. Note that different theorists have proposed different roles and labeled them differently, therefore there is no absolute agreement.

Agent: The ‘doer’ of the action
Sue pushed Steve

Patient (or Theme): The ‘undergoer’ of the action
Sue pushed Steve (but also Sue fell down)

Experiencer: The entity that experiences the action
Sue felt happy

Instrument: A medium or tool used to complete the action
Sue opened the door with the key

Goal: The location or entity towards which something moves
Sue drove to Chicago

Benefactive (or Recipient): The entity that benefits from the action
Sue gave Kim the tickets

Semantic features

While semantic roles describe ‘who does what’ in a relatively basic way (someone affects someone else, someone benefits from an action) semantic features represent the specific properties something needs to have to be semantically acceptable in a certain construction. Here’s the example from above once more:

John likes basketball

”The table likes basketball

The dog ran across the field

”The refrigerator ran across the field

What’s wrong with the second and the fourth sentence? The answer is that the experience and the action described (like, run) can’t be made and performed by non-living things such as tables and refrigerators.

  John the table the dog the refrigerator
animate + - + -

The term animacy is used to describe whether or not something is what we conventionally call ‘alive’. It is apparently a requirement for the subject X to have the feature +animate in order for a sentence like X ran across the field to be semantically well-formed.

Different or additional features may be required in other contexts and the list below serves only as an example:

  John the table the dog the refrigerator lemonade
animate + - + - -
human + - - - -
canine - - + - -
liquid - - - - +

While ‘animate’ and ‘human’ are broader (and more useful) features than ‘canine’ and ‘liquid’, all of them are distinctive in certain contexts. Humans are generally the subject of verbs such as like, adore, hate and consider, some kind of dog is generally the subject of the verb bark and some form of liquid generally the direct object of the verb drink.

”Dana’s mother has no children

”The empty bucket is full

”The meeting will take place three years ago

We can generally explain semantic anomaly via some kind of feature mismatch. In the examples above, the mismatch occurs in different places: +mother and +no children do not match, +empty and +full are not compatible and +future event +past event do not work together. Note that these are not commonly used features, but reading the examples like this makes it simple to spot the semantic problem right away.

Key terms

  • conventional vs. social and affective meaning
  • reference
  • sense
  • denotation & connotation
  • extension
  • semantic anomaly
  • semantic roles
    • agent
    • patient/theme
    • instrument
    • goal
    • benefactive/recipient
  • semantic features
Written by Cornelius in: Semantics,Summaries |

Presentation for Session 11

Here is the presentation for yesterday’s session.

Session 11 – Semantic features and roles 

The PDF is available here.

Please note: there is much discussion among linguists about how to delineate semantic roles and what roles are necessary for an accurate semantic and syntactic description. If you review the literature, you may find that theme in particular is often described much more restrictively than is the case in my slides. There is no simple solution to this – what counts is that you understand the general function of semantic roles in conjunction with syntax.

Written by Cornelius in: Presentations,Semantics |

Presentation for Session 10

Below you’ll find the link to the presentation for today’s session.

Session 10 – Lexical semantics

The PDF is available here.

Written by Cornelius in: Presentations,Semantics |

Powered by WordPress | Aeros Theme | TheBuckmaker.com WordPress Themes