CHAPTER 1

### **1.1 Introduction: What is Semantics and Linguistic Knowledge?**

Semantics is the study of meaning in language. This book is a guide to the ideas and methods used in
semantics, which is a part of modern linguistics. The book does not focus on just one theory. Instead,
it starts from one basic idea: a person's ability to use language comes from knowledge stored inside
their mind. Semantics is the field that tries to investigate and describe this knowledge.

A very important idea in modern linguistics is that speakers have different kinds of language
knowledge. They are not all the same. For example, speakers know how to pronounce words. They
know how to put words together to make correct sentences. They also know what individual words
and whole sentences mean. To show these different types of knowledge, linguistics is divided into
different levels of analysis or study.

* **Phonology** is the study of the sounds a language has and how these sounds combine to form
words.
* **Syntax** is the study of the rules for how words can be combined to form sentences that follow
the grammar of the language.
* **Semantics** is the study of the meanings of words and sentences.

This separation into different levels feels natural and logical. We can see this when we learn a new
language. You might see a written word in a book and know what it means, but you might not know
how to say it out loud. This shows your knowledge of meaning is separate from your knowledge of
pronunciation. Another situation: you might hear a word and be able to repeat its pronunciation
perfectly, but you have no idea what it means. This shows your knowledge of pronunciation is
separate from your knowledge of meaning. You could also know a word's meaning and pronunciation
but not know the rule for making it plural. This shows that "knowing a word" is actually a combination
of different kinds of knowledge. The same is true for knowing how to build phrases and sentences; it
involves different types of knowledge working together.

Since the main job of linguistics is to describe a speaker's knowledge, the job of a semanticist (a
person who studies semantics) is to describe a speaker's *semantic knowledge*. This is the
knowledge about meaning. This knowledge allows English speakers to understand certain
relationships between sentences automatically, without having to think hard about it. For instance,
they know that the following two sentences describe the exact same situation, just from different
points of view:
* **Example 1:** "In the spine, the thoracic vertebrae are above the lumbar vertebrae."
* **Example 2:** "In the spine, the lumbar vertebrae are below the thoracic vertebrae."

Speakers also know when sentences directly contradict each other and cannot both be true:
* **Example 1:** "Addis Ababa is the capital of Ethiopia."
* **Example 2:** "Addis Ababa is not the capital of Ethiopia."

They can identify ambiguity, which is when a single sentence has more than one possible meaning:
* **Example:** "She gave her the slip." (This could mean she physically handed over a slip of paper,
or it could be an idiom meaning 'she escaped from her'.)

Furthermore, they understand a relationship called **entailment**. Entailment is a connection

between sentences where if sentence A is true, then sentence B must automatically also be true.
Another way to think about it is that it should be impossible to say that sentence A is true and
sentence B is false at the same time.
* **Example:** The sentence "Henry murdered his bank manager" (Sentence A) entails the sentence
"Henry's bank manager is dead" (Sentence B). If the first sentence is true, then the second one cannot
be false. The truth of the first sentence guarantees the truth of the second.

Knowing how the word "not" creates a contradiction, or understanding the relationship between
words like "above" and "below," or "murder" and "dead," are all parts of an English speaker's
semantic knowledge. A proper description of English semantics must be able to explain this
knowledge.

As the original definition suggests, semantics is a very wide field of study. Different scholars write
about very different topics and use quite different methods, but they all share the same general aim:
to describe semantic knowledge. Because of this variety, semantics is the most diverse field within
linguistics. In addition, semanticists need to have at least a basic familiarity with other subjects like
philosophy and psychology. These disciplines also investigate how meaning is created and shared.
Some of the questions raised in these neighboring fields have important effects on the way linguists
do semantics.

### **1.2 Semantics and Semiotics: Language as One Type of Sign System**

The basic task in semantics is to show how people use pieces of language to communicate meanings.
However, it is important to note that this is only one part of a much larger human activity:
understanding meaning in general. The ability to understand linguistic meaning is a specific example
of our more general human ability to use signs. We can see this from how we use the word "mean" in
many everyday situations that do not involve language.

* **Example 1:** "Those vultures mean there's a dead animal up ahead." Here, we are making an
inference based on cause and effect (vultures are naturally attracted to dead animals).
* **Example 2:** "The red flag means it's dangerous to swim." Here, we are using knowledge about
an arbitrary public symbol (we have agreed that the color red signifies danger).

These examples show the all-pervasive human habit of identifying and creating signs. This is the
process of making one thing stand for another, which is sometimes called **signification**. The
general study of signs and symbols is called **semiotics**. Scholars like Ferdinand de Saussure
stressed that the study of linguistic meaning is just one branch of this larger study of how sign systems
are used.

Semioticians investigate the different types of relationships that can exist between a sign and the
object it represents. In Saussure's terminology, this is the relationship between the **signifier** (the
form the sign takes, like a sound or an image) and the **signified** (the concept it represents). A
useful classification from the philosopher C.S. Peirce divides signs into three main types:

1. **Icon:** A sign that physically resembles what it represents. There is a similarity between the
sign and the object.
* **Example 1:** A detailed portrait of a person looks like that person.
* **Example 2:** A diagram of an engine in a manual looks like the real engine parts.
2. **Index:** A sign that is directly connected to its object, often through a cause-and-effect
relationship. The sign is a clue or a symptom of the object.
* **Example 1:** Smoke is an index of fire.
* **Example 2:** A high fever is an index (a symptom) of a virus.
3. **Symbol:** A sign where the link to its object is purely conventional, based on social agreement
or a learned rule. There is no resemblance or direct connection.
* **Example 1:** The stripes on a military uniform symbolize rank (e.g., a sergeant).
* **Example 2:** The word "dog" is a symbol for the animal. There is nothing in the sounds d-o-g
that looks or acts like a dog.

In this classification, words are considered verbal symbols. For the rest of this book, we will focus
specifically on linguistic meaning, leaving the broader field of semiotics. How language developed in
relation to other symbolic systems is an open question. What seems clear, however, is that language
represents humanity's most sophisticated and complex use of signs.

### **1.3 Three Major Challenges in Doing Semantics**

Analyzing a speaker's semantic knowledge is an exciting but also a very challenging task. To get an
idea of just how challenging it is, we can imagine a very simple theory of semantics, which we can call
the **definitions theory**. This theory would state that to give the meaning of a linguistic expression,
we simply need to establish definitions for all the words. Then, when a speaker combines words to
form sentences according to the grammatical rules of her language, the definitions of the words
would automatically combine to give us the meaning of the whole phrase and then the whole
sentence. Let us see what happens when we try to put this simple theory into practice.

As soon as we begin the task of attaching definitions to words, we run into a number of serious
problems. Three of these problems are particularly tricky for our simple theory.

**The First Challenge: The Problem of Circularity**

This is the challenge of defining words only by using other words. How can we state the meaning of a
word without using other words, either from the same language or a different one? This is the same
problem faced by dictionary writers. For example, if you look up the word "ferret" in a monolingual
English dictionary, you might find a definition like: "a domesticated albino variety of the polecat, bred
for hunting rabbits, rats, etc." To understand this definition, you now have to understand the words in
the definition, such as "domesticated," "albino," and "polecat." According to our aims for semantics,
we have to describe the meanings of these words too. So you look up "domesticated," and its
definition might use words like "tame" and "living with human beings." You then have to look up
"tame," and so on. If the definitions of word meaning are always given in other words, the process
might never end. We are stuck in a circle of words defining other words without ever stepping outside
of language itself to find a solid foundation for meaning. The big question is: can we ever step outside
of language in order to describe it, or are we forever trapped in circular definitions?

**The Second Challenge: The Problem of Linguistic vs. Encyclopedic Knowledge**

This problem concerns where meanings exist and what kind of knowledge they are. If we ask where
the meanings of words exist, the answer must be: in the minds of native speakers of the language.
Therefore, meaning is a kind of knowledge. This raises several difficult questions. For example, is
there a difference between this kind of knowledge and other kinds of knowledge that people have? In
particular, can we make a clear distinction between **linguistic knowledge** (knowledge about the
meaning of words themselves) and **encyclopedic knowledge** (general knowledge about the way
the world is)?

For instance, if one person believes that a whale is a fish and another person knows that it is a
mammal, do they have different meanings for the word "whale" when they both use it? Probably not,
because you would still understand me if I said, "I dreamt that I was swallowed by a whale," even if
you know I am scientifically wrong. This shows that communication can happen even when our world
knowledge differs.

There is another aspect to this problem: what should we do if we find that different speakers of the
same language have different understandings of what a word means? Whose knowledge should we
pick as the "correct" meaning? One strategy is to avoid the decision by picking just one speaker and
limiting our semantic description to that individual's personal language, which is called an
**idiolect**. Another strategy is to identify experts, like scientists, and use their knowledge. But
moving away from ordinary speakers to use a strict scientific definition has its own dangers. It risks
making semantics equivalent to all of science, which is not practical. It also ignores the obvious fact
that most of us seem to understand each other perfectly well when talking about things like animals,
without any formal training in zoology.

**The Third Challenge: The Problem of Context**

The meaning of what we say can change dramatically depending on the situation, or context. The
same sentence can mean different things in different scenarios.
* **Example 1:** The sentence "Marvelous weather you have here in Ireland" would be interpreted
as a sincere compliment on a sunny day, but as sarcastic or a joke on a rainy, windy day.
* **Example 2:** "He's dying" has a literal, medical meaning when said about a terminally ill patient,
but a metaphorical meaning when said about a stand-up comedian who is failing to get laughs from
the audience.

If features of the context are a part of an utterance's meaning, then how can we possibly include
them in our definitions? The number of possible situations, and therefore of different interpretations,
is enormous, if not infinite. It does not seem likely that we could fit all the relevant contextual
information into a fixed definition for a word or sentence.

These three issues—the problem of circularity, the question of whether linguistic knowledge is
different from general knowledge, and the problem of the contribution of context to meaning—show
that our simple "definitions theory" is too simple to do the job we want. Semantic analysis must be
more complicated than just attaching dictionary-like definitions to linguistic expressions. As we shall
see in the rest of this book, semanticists have proposed a number of more sophisticated strategies for
improving on this initial position.

### **1.4 Meeting the Challenges: Basic Strategies for Semantics**

In most current linguistic theories, semantic analysis is considered as important a part of the linguist's
job as, say, phonological analysis (the study of sounds). Different theories may disagree on the details
of the relationship between semantics and other levels of analysis like syntax and morphology (word
structure), but they all seem to agree that a linguistic analysis is incomplete without a semantic
component. We need, it seems, to establish a semantic component in our theories of language.
Therefore, we have to ask: how can we meet the three challenges outlined in the last section? Clearly,
we have to replace the simple theory of definitions with a better theory that successfully solves these
problems.

One of the aims of this book is to show how various semantic theories have sought to provide
solutions to these problems, and we will return to them in detail over subsequent chapters. For now,
we will simply mention the possible strategies that we will see developed later in the book.

**Strategy for Circularity: Designing a Semantic Metalanguage**

To cope with the problem of circularity, one solution is to design a **semantic metalanguage**. We
use the term "metalanguage" here with its usual meaning in linguistics: it is the tool we use for
description. For example, in a grammar book for Arabic that is written in French, Arabic is the
**object language** (the language being studied) and French is the **metalanguage** (the tool used
for description). An ideal semantic metalanguage would be a special language, neutral and not based
on any natural languages like English or French, used solely to describe the meanings and rules of all
languages. It would satisfy scientific criteria of clarity, economy, and consistency. We will see various
proposals for such a metalanguage in later chapters. Some linguists, however, meet claims that such a
perfect metalanguage is unattainable and argue that the best policy is to use ordinary language very
carefully to describe meaning.

For some linguists, though, even translating words into a perfect metalanguage would not be a fully
satisfactory semantic description. They reason like this: if words are symbols, they have to relate to
something; otherwise, what are they symbols of? In this view, to give the true semantics of words, we
have to "ground" them in something non-linguistic, something outside of language itself. This leads to
a philosophical debate about whether the things that words signify are real objects in the world or are
thoughts and concepts in our minds.

**Strategy for the Linguistic/Encyclopedic Divide: Modeling the Mental Lexicon**

Setting up a metalanguage can also help with the problem of relating semantic and encyclopedic
knowledge. This is because designing meaning representations for words forces us to argue about
which elements of knowledge are absolutely essential to a word's meaning and which are extra, world
knowledge. Our knowledge of words is often compared to a **mental lexicon**—a kind of mental
dictionary that every speaker has. If you open a real dictionary at the entry for "whale," the definition
is likely to begin with "large marine mammal..." This brings us back to our earlier question: does it
follow that someone who doesn't know that whales are mammals fails to understand the meaning of
the word "whale"? What if a speaker only knows that it is a large animal that lives in the sea, but is
hazy about the details after that? The real issue is figuring out the amount and type of knowledge that
it is necessary to have in order to use a word correctly. We will see aspects of this debate in later
chapters, as it is really part of the general psychological debate about how our minds represent
concepts and categories.

**Strategy for Context: Literal Meaning and Pragmatics**

In tackling the third problem, of context, one traditional solution has been to assume a split in an
expression's meaning. The meaning is divided between the local, contextual effects and a context-free
element of meaning, which we might call the **conventional** or **literal meaning**. We could
perhaps try to limit our definitions to this literal part of meaning and deal with the tricky contextual
features separately.

The other side of this approach is to investigate the role of contextual information in communication
more directly. This involves trying to establish theories of how speakers and hearers combine
knowledge of context with their linguistic knowledge. As we shall see, it seems that speakers and
hearers cooperate, using various types of contextual information. Investigating this leads us to a view
of the listener's role that is quite different from the simple, but common, analogy of decoding a coded
message. We will see that listeners have a very active role. They use what has been said, together
with their background knowledge, to make guesses and inferences about what the speaker actually
meant to communicate. The study of these processes and the role of context in them is often
assigned to a special area of study called **pragmatics**.

Each of these strategies—the creation of semantic metalanguages, the modeling of conceptual

knowledge, the theory of literal language, and factoring out context into the field of pragmatics—will
be investigated in later chapters of this book.

### **1.5 Semantics in a Model of Grammar**

**1.5.1 Introduction: Semantics as a Module**

As has been suggested already, for many linguists the aim of doing semantics is to set it up as a formal
component, or "module," in a model of grammar. This component would be parallel to other
components like syntax (sentence structure) or phonology (sounds). Linguists often like to draw
flowchart-style diagrams of grammatical models, and in many of them, there is a box labeled
"semantics," as shown in a simple diagram where sound goes into phonology, then to syntax, then to
semantics, and finally to thought.

Before we go on, it is worthwhile to consider whether it is truly justified to view semantics as a

component that is equal and parallel to, say, syntax. We saw earlier that linguists identify different
levels of analysis. Another way of describing this is to say that a speaker's linguistic knowledge is
organized into distinct modules, or is "modularized." As a result, many linguistic theories are
themselves modularized, having something like the boxes in the diagram. Our question, though,
remains: what kind of module is semantics? The answer varies from theory to theory.

The real problem is that units at all linguistic levels are part of the general enterprise: to communicate
meaning. This means that, in at least one sense, meaning is a product of *all* linguistic levels.
Changing one phoneme for another (e.g., "bat" vs. "pat"), changing one verb ending for another, or
changing the word order (e.g., "The dog bit the man" vs. "The man bit the dog") will all produce
differences in meaning. This view leads some writers to believe that meaning cannot be identified as a
separate level, independent from the study of other levels of grammar. A strong version of this view is
associated with the theory known as **Cognitive Grammar**. For example, a cognitive linguist has
claimed that "the various autonomy theses and dichotomies proposed in the linguistic literature have
to be abandoned: a strict separation of syntax, morphology and lexicon is untenable; furthermore it is
impossible to separate linguistic knowledge from extra-linguistic knowledge."
As we shall see in the course of this book, however, many other linguists do see some utility in
maintaining both types of distinction referred to above: the distinction between linguistic and non-
linguistic knowledge, and within linguistic knowledge, identifying distinct modules for knowledge
about pronunciation, grammar, and meaning.

**1.5.2 Word Meaning and Sentence Meaning**

If an independent component of semantics is identified in a grammar model, one central issue is the
relationship between word meaning and sentence meaning.

Knowing a language, especially one's native language, involves knowing thousands of words. As
mentioned earlier, some linguists call this mental store of words a **lexicon**, making a direct
parallel with the lists of words and meanings published as dictionaries. In this view, the mental lexicon
is a large but finite body of knowledge. A significant part of this knowledge must be semantic—the
meanings of the words. This lexicon is not completely static because we are continually learning new
words and sometimes forgetting old ones. It is clear, though, that at any one time we hold a large
amount of semantic knowledge in our memory.

Phrases and sentences also have meaning, of course. But an important difference between word
meaning on the one hand, and phrase and sentence meaning on the other, concerns
**productivity**. It is always possible to create new words (like "selfie" or "blog"), but this is a
relatively infrequent occurrence. On the other hand, speakers regularly create sentences that they
have never used or heard before, and they are confident that their audience will understand them.
This "creativity of sentence formation" has been particularly stressed by Noam Chomsky. It is one of
generative grammar's most important insights that a relatively small number of combinatory rules
may allow speakers to use a finite set of words to create a very large, perhaps infinite, number of
sentences.

To allow this, the rules for sentence formation must be **recursive**. Recursive rules allow for
repetitive embedding or coordination of syntactic categories, meaning the rules can be applied over
and over again within their own output.
* **Example of a Recursive Rule:** A simple compositional rule like `S → [S S (and S)*]` (which
means: a Sentence can be rewritten as a Sentence, followed by another Sentence, and the "and S"
part is optional and repeatable) will allow potentially limitless expansions:
* 1.14 a. [S S and S] (e.g., "It rained and I stayed home.")
* 1.14 b. [S S and S and S] (e.g., "It rained and I stayed home and I watched a movie.")
* 1.14 c. [S S and S and S and S] etc.

The idea is that you can always add another clause to a sentence. The same can be done within a
noun phrase (NP), allowing you to keep adding nouns: "I bought [a book and a magazine and some
pens...]"

This insight has a major implication for semantic description. Clearly, if a speaker can make up novel
sentences and these sentences are understood, then they must obey the semantic rules of the
language. Therefore, the meanings of sentences cannot be listed and memorized in a lexicon like the
meanings of words. They must be *created* by rules of combination too. Semanticists often describe
this by saying that sentence meaning is **compositional**. This term means that the meaning of a
complex expression is determined by the meanings of its component parts and the way in which they
are combined grammatically.

This brings us back to our question of levels in a grammar model. We see that meaning is, so to speak,
in two places: there is a more stable, listed body of word meanings in the lexicon, and there are the
limitless, composed meanings of sentences. How can we connect the semantic information stored in
the lexicon with the compositional meaning of sentences? It seems reasonable to conclude that
semantic rules themselves have to be compositional too, and in some sense "in step" with the
grammatical rules.
The relationship between syntactic rules and semantic rules is portrayed differently in different
theories of language. In the evolving forms of Noam Chomsky's generative grammar, syntactic rules
operate independently of semantic rules, but the two types are brought together at a later level of
representation called Logical Form. In many other theories, semantic rules and grammatical rules are
inextricably bound together from the start, so that each combination of words in a language has to be
permissible under both the syntactic and semantic rules simultaneously. Such an approach is typical
of functional approaches like Halliday's Functional Grammar and Role and Reference Grammar, as
well as variants of generative grammar like Head-Driven Phrase Structure Grammar.

### **1.6 Some Important Assumptions in Semantics**

At this point, we can introduce some basic ideas that are assumed in many semantic theories. These
ideas will come in useful in our subsequent discussion. In most cases, the initial descriptions of these
ideas will be simple and a little vague; we will try to make them more precise in subsequent chapters.

**1.6.1 Reference and Sense**

One very important point made by the linguist Ferdinand de Saussure is that the meaning of linguistic
expressions derives from two different sources: the language they are part of and the world they
describe.

* **Reference** is the relationship between a word and the actual object or entity in the world that
it points to or identifies. Words allow us to "pick out" or refer to specific parts of the world and make
statements about them.
* **Example 1:** In the sentence "He saw Paul," the name "Paul" is used to refer to a specific
individual in the world.
* **Example 2:** In the sentence "She bought a dog," the phrase "a dog" is used to refer to a
particular animal.

* **Sense**, on the other hand, is not about the world directly, but about the language system
itself. It is the network of relationships that a word has with other words in the same language. A
word's sense is its place within the vocabulary system.
* **Example 1:** Saussure's famous example compares English "sheep" and French "mouton."
They can be used to refer to the same animal, but their *sense* is different because they exist in
different language systems. English has a separate word, "mutton," for the meat of the animal, while
French uses "mouton" for both the animal and the meat. Therefore, the meaning of "sheep" is partly
defined by the existence of "mutton."
* **Example 2:** The meaning of "chair" in English is partly defined by the existence of other
words like "stool," "sofa," and "bench." The meaning of "red" is defined in relation to other color
terms like "orange," "pink," and "brown."

Saussure used a diagram to show this patterning, where each word is like an oval linked to other ovals
in a network. The meaning of a word derives both from what it can be used to refer to in the world
(reference) and from the way its meaning is defined by its relationships with other words in the
language (sense).

**1.6.2 Utterances, Sentences, and Propositions**

These three terms are used to describe different levels of abstraction in language. It is important to
understand the difference between them.

* **Utterance:** An utterance is the most concrete level. It is a real-life, physical event of speaking
(or writing). It is unique to a specific time, place, and speaker.
* **Example 1:** If you say "I'm hungry" right now, that is one utterance.
* **Example 2:** If your friend says the exact same words five minutes later, that is a second,
distinct utterance. Every time someone speaks, they are producing an utterance.

* **Sentence:** A sentence is a more abstract grammatical unit that we obtain from utterances.
Sentences are abstract because we filter out the specific details of the utterance to get to the general
pattern. If a third and fourth person also say "I'm hungry" with the same intonation, we would say
there were four *utterances* of the same *sentence*. In other words, sentences are abstracted, or
generalized, from actual language use. One example of this abstraction is direct quotation. If someone
reports, 'He said "I'm hungry",' they are unlikely to mimic the original speaker's voice exactly. They
filter out information like pitch, accent, and voice quality. Speakers recognize that at the level of the
sentence, these kinds of personal information are not important. So, we can look at sentences from
the speaker's point of view (as abstract patterns to be made real by uttering them) or from the
hearer's point of view (as abstract patterns reached by filtering out certain information from
utterances).

* **Proposition:** A proposition is the most abstract of the three. It is the core, factual content or
meaning that a sentence expresses. It is a description of a state of affairs, stripped of all grammatical
details like active/passive voice or sentence structure. Logicians discovered that for establishing rules
of valid reasoning, certain grammatical information is irrelevant. For example, the difference between
active and passive sentences does not matter for logic:
* **Example 1:** "Caesar invaded Gaul." (active)
* **Example 2:** "Gaul was invaded by Caesar." (passive)
From a logician's perspective, these sentences are equivalent because they describe the same
event. Whenever one is true, the other is true. Therefore, they share the same **proposition**,
which we can write as CAESAR INVADED GAUL.

Other sentences with different structures can also share the same proposition:
* "It was Gaul that Caesar invaded."
* "It was Caesar that invaded Gaul."
* "What Caesar invaded was Gaul."
All these different sentences seem to describe the same state of affairs. They all share the proposition
CAESAR INVADED GAUL.

Logicians often use special formulae to represent propositions, like `invade(caesar, gaul)` or
`end(war)`, where the verb is treated as a function and the subject and object are its arguments.
Some semanticists have borrowed this notion of a proposition. They use it to identify a common core
meaning that might be shared by different types of sentences. For example, the statement "Joan
made the sorbet," the question "Did Joan make the sorbet?", and the command "Joan, make the
sorbet!" might all be seen to share the propositional element JOAN MAKE THE SORBET. In this view,
the different sentence types allow the speaker to do different things with the same proposition: to
assert it, to question it, or to request that it be made true.

To sum up: **Utterances** are real pieces of speech. By filtering out certain types of information
(especially phonetic), we get to abstract grammatical elements called **sentences**. By going on to
filter out certain types of grammatical information, we get to **propositions**, which are
descriptions of states of affairs and which some writers see as a basic element of sentence meaning.

**1.6.3 Literal and Non-Literal Meaning**

This is the distinction between speaking directly and factually versus using language in a creative,
imaginative way. The basic distinction seems like common sense: we distinguish between instances
where the speaker speaks in a neutral, factually accurate way, and instances where the speaker
deliberately describes something in untrue or impossible terms in order to achieve a special effect.
* **Literal Meaning:** The basic, conventional, and factual meaning of the words.
* **Example:** "I'm hungry."
* **Non-Literal (Figurative) Meaning:** Using words in a way that is not factually true to achieve a
rhetorical effect. This is described by terms like metaphor, irony, and hyperbole.
* **Example 1:** "I'm starving." (Hyperbole - you are not actually dying of hunger).
* **Example 2:** "I could eat a horse." (Hyperbole - you are very hungry, but you couldn't actually
eat an entire horse).

On closer examination, though, it proves very difficult to draw a firm line between literal and non-
literal uses of language. One reason is that language changes over time. One of the ways words gain
new meanings is through **metaphorical extension**, where a new idea is described in terms of
something more familiar. When a metaphor is new, its figurative nature is clear.
* **Example of a new metaphor:** Expressions like "to go viral" (for an idea that spreads rapidly
online) or "to photobomb" (to unexpectedly appear in a photo) are relatively new, and their
metaphorical quality is still apparent.

However, after a while, such expressions become fossilized. Their metaphorical quality is no longer
apparent to speakers, and they become part of the normal, "literal" language.
* **Example of a dead metaphor:** The word "shuttle" in "space shuttle" comes from a weaving
tool that goes back and forth. Most people today do not think of a loom when they use the word
"shuttle"; the metaphor has died. The vocabulary of a language is full of these fossilized metaphors,
and this continuing process makes it difficult to decide the exact point at which the use of a word is
literal rather than figurative.

Facts like these have led some linguists, notably George Lakoff, to claim that there is no principled
distinction between literal and metaphorical language. They see metaphor as an integral and
fundamental part of human thought and categorization—a basic way we organize our ideas about the
world. Lakoff and Johnson identify whole clusters of metaphoric uses, giving them labels like "Time is
money" to explain expressions such as:
* "You're wasting my time."
* "This gadget will save you hours."
* "I've invested a lot of time in her."
Their claim is that entire semantic fields are systematically organized around such central metaphors,
and that their use is not just a stylistic effect, but reflects how we culturally think about time as a kind
of commodity.

Clearly, if sentences like "How do you spend your time these days?" are identified as metaphorical,
then it becomes very difficult to find any uses of language that are purely literal. Many linguists,
however, would deny that this use of "spend" is still metaphorical. They adopt a position that sees
this as an example of a "faded" or "dead" metaphor. The idea is that metaphors fade over time and
become part of normal literal language. In this approach, there is a valid distinction between literal
and non-literal language.

In what we can call the **literal language theory**, metaphors and other non-literal uses require a
different processing strategy from literal language. One view is that hearers first recognize non-literal
uses as being semantically odd or factually nonsensical (like "eating a horse"). Then, motivated by an
assumption that speakers are generally trying to make sense, the hearer makes inferences to figure
out what the speaker really meant. Some figurative expressions, like "eat a horse," are quite
conventionalized and don't require much effort to interpret. Others might require more work, as
when a reader encounters a line like "She flew her kite a bit too often" in a novel and has to reject the
literal interpretation (a fear of kites) and infer that it is a metaphor for being unfaithful.

**1.6.4 Semantics and Pragmatics**

A similarly difficult but very important distinction is the one between semantics and pragmatics.
These two terms denote related and complementary fields of study, both concerning the transmission
of meaning through language. However, drawing a clear line between them is difficult and
controversial. As a preliminary, we can look at an early use of the term "pragmatics" from Charles
Morris's division of semiotics (the study of signs):
1.33 **syntax:** the formal relations of signs to each other.
1.34 **semantics:** the relations of signs to the objects they are applicable to.
1.35 **pragmatics:** the relation of signs to interpreters (the users of the signs).

If we narrow "signs" down to "linguistic signs," this gives us a view where **pragmatics** is the study
of the speaker's and hearer's interpretation of language. A crude but helpful way to interpret this is:
* **Pragmatics** = meaning described in relation to speakers and hearers.
* **Semantics** = meaning abstracted away from the users.
Let's investigate what this might mean with a simple example. A speaker can utter the same sentence,
for example, "The place is closing," and mean to use it in different ways depending on the context. It
could be a simple statement of fact. It could be a warning to a fellow shopper to hurry up and make a
purchase. It could be a suggestion or command to a friend in a bar that it's time to leave. In fact, we
can imagine a whole series of uses for this simple sentence.

Some semanticists would claim that there is some core element of meaning common to all of these
uses. This common, non-situation-specific meaning is what **semantics** is concerned with. On the
other hand, the wide range of uses a sentence can be put to, depending on the context, would be the
object of study for **pragmatics**.

One common way of talking about this is to distinguish between **sentence meaning** and
**speaker meaning**.
* **Sentence meaning** is the literal, context-free meaning of the words and sentences themselves.
It is what the sentence means in isolation.
* **Speaker meaning** is what a particular speaker intends to communicate by uttering that
sentence in a specific context. It is what the speaker means by the sentence.

This suggests that words and sentences have a meaning independently of any particular use. This
stable meaning is then taken by a speaker and incorporated into the particular meaning she wants to
convey at any one time. In this view, **semantics is concerned with sentence meaning, and
pragmatics is concerned with speaker meaning.**

We can see how this distinction works when we consider pronouns, which are highly dependent on
context. If someone says, "Is he awake?" the listener has to understand two things:
1. The semantic knowledge: In English, "he" means something like "a male entity referred to by the
speaker, who is not the speaker or the person being spoken to."
2. The pragmatic task: To work out who, in this specific situation, the speaker is referring to by "he."

Knowing the first is part of semantic knowledge. Working out the second is a task for one's pragmatic
competence.

The advantage of such a distinction is that it might free the semanticist from having to include all
kinds of real-world and contextual knowledge in semantics. It would be the role of pragmaticists to
investigate the interaction between purely linguistic knowledge and general encyclopedic knowledge.
As we shall see, in order to understand utterances, hearers seem to use both types of knowledge,
along with knowledge about the context and common-sense reasoning. A semantics/pragmatics
division allows semanticists to concentrate on just the linguistic element in utterance comprehension.
Pragmatics would then be the field that studies how hearers fill out the semantic structure with
contextual information and make inferences that go beyond the literal meaning of what was said
(e.g., inferring that "I'm tired" might mean "Let's go home").

The semantics/pragmatics distinction seems, then, to be a useful one. The problems with it emerge
when we get down to detail: precisely which phenomena are semantic and which are pragmatic? As
we will discuss later, much of meaning seems to depend on context from the very beginning. It is
often difficult to identify a meaning for a word that does not depend on the context of its use. The
strategy in this book will not be to try too hard to draw a firm line between the two. Some theorists
are skeptical of the distinction altogether, while others accept it but draw the line in different places.
What will become clear as we proceed is that it is very difficult to separate context from language.
The structure of sentences often reveals that they are designed by speakers to be uttered in specific
contexts and with desired effects.

### **1.7 Summary of the Chapter**

In this chapter, we have taken a first look at the task of establishing semantics as a branch of
linguistics. We identified three major challenges that make this task difficult: the problem of
circularity (defining words with other words), the problem of context (how situation changes
meaning), and the problem of the status of linguistic knowledge (how it differs from general world
knowledge). We will see examples of these problems and the proposed solutions to them as we
proceed through this book.

We noted that establishing a semantics component in a linguistic theory involves deciding how to
relate the meaning of words to the meaning of sentences. We saw that word meanings are stored in a
mental lexicon, while sentence meanings are compositional and built by rules.

Finally, we introduced some important background ideas that are assumed in many semantic theories
and which we will examine in more detail in subsequent chapters. These are:
* **Reference and Sense:** The two sources of a word's meaning (the world and the language
system).
* **Utterances, Sentences, and Propositions:** The three levels of abstraction in language (from real
events to core meaning).
* **Literal and Non-Literal Meaning:** The distinction between direct and figurative language use.
* **Semantics and Pragmatics:** The division between context-free sentence meaning and context-
dependent speaker meaning.

These concepts provide the foundation for the more detailed exploration of semantic theory in the
rest of the book.

CHAPTER 2

Of course. Here is a comprehensive, simplified explanation of Chapter 2 from John I. Saeed's

"Semantics," covering every concept in an easy-to-understand, paragraph-by-paragraph format with
two examples for each major topic.

### **2.1 Introduction: How Language Connects to the World**

This chapter tackles a fundamental question: how is it possible for us to use language to describe the
world around us? How can we make sounds with our mouths that convey information to a listener
about what we are seeing, hearing, or thinking? For example, how can I tell you about a scene outside
my window just by saying words? All languages give speakers this amazing ability to describe, or to
create a model of, the things they perceive. We constantly use words to pick out individual things,
people, or places.

* **Example 1:** In the sentence "That **dog** looks vicious," the phrase "that dog" is used to pick
out a specific animal.
* **Example 2:** In "We’ve just flown back from **Paris**," the word "Paris" is used to identify a
specific city.

In semantics, this action of picking out or identifying something with words is often called
**referring** or **denoting**. So, you can use the word "Paris" to refer to or denote the city. The
actual thing being referred to—in this case, the city itself—is called the **referent**.

Some linguists, like John Lyons, make a careful distinction between these two terms. They use
**denote** to describe the stable, long-term relationship between a linguistic expression (a word or
phrase) and the world. They use **refer** to describe the specific action of a speaker who, at a
particular moment, uses a word to pick out something in the world. We will follow this usage.

* **Example of Denoting vs. Referring:** If I say, "A **sparrow** flew into the **room**," I, as the
speaker, am using the phrases "a sparrow" and "the room" to **refer** to specific things in the world
at that moment. Meanwhile, the nouns "sparrow" and "room" themselves **denote** certain classes
of items. The word "sparrow" denotes the entire class of all sparrows, and "room" denotes the entire
class of all rooms.
Another important difference follows from these definitions. **Denotation** is a stable property of
the word itself within the language system. It doesn't depend on any single use. **Reference**,
however, is a momentary relationship that changes with context. What entity a speaker refers to by
using the word "sparrow" depends entirely on the situation.

As we will see, semanticists have different views on how to approach this ability to talk about the
world. Two views are particularly important in current theories: the **referential (or denotational)
approach** and the **representational approach**.

For semanticists who adopt the **referential approach**, this act of connecting words to the world
*is* the core of meaning. To provide a semantic description of a language, we need to show how its
expressions "hook onto" the world. In this view, theories of meaning are called referential when their
basic idea is that we can explain the meaning of words and sentences by showing how they relate to
real situations. Nouns are meaningful because they denote entities in the world, and sentences are
meaningful because they denote situations and events.

* **Example:** The difference in meaning between "There is a casino in Grafton Street" and "There
isn't a casino in Grafton Street" comes from the fact that they describe two different, incompatible
situations. If we assume both sentences are about the same street at the same time, then one must
be a true description of the situation and the other must be false.

For semanticists adopting the **representational approach**, our ability to talk about the world
depends on our mental models of it. In this view, a language reflects a theory about reality—about
the types of things and situations that exist in the world. A speaker can choose to view the same real-
world situation in different ways, influenced by the conventional patterns of their language.

* **Example:** In English, we can view the same situation as an activity ("Joan **is sleeping**") or
as a state ("Joan **is asleep**").

This influence of language on how we conceptualize situations becomes even clearer when we
compare different languages. Look at the different ways of saying "you have a cold":

* **English:** "You **have** a cold." (The situation is viewed as *possession*: the person
possesses the illness.)
* **Somali:** Literally, "A cold **has** you." (Again, viewed as *possession*, but this time the
illness possesses the person.)
* **Irish:** Literally, "A cold **is on** you." (The situation is viewed as *location*: the person is the
location for the illness.)

The point is that different languages conventionalize different ways of conceptualizing the same real-
world situation, and this influences how we describe it.

Theories of meaning are called **representational** when they focus on the way our descriptions of
reality are shaped by the conceptual structures built into our language.

We can see these two approaches as focusing on different parts of the same process: talking about
the world. In **referential theories**, meaning comes from language being connected to, or
"grounded in," external reality. In **representational theories**, meaning comes from language
being a reflection of our internal conceptual structures. This difference will appear throughout this
book.

These two approaches are influenced by ideas from philosophy and psychology. In this chapter, we
will review some of the most important of these ideas. We will start by looking at the different ways
linguistic expressions can be used to refer. Then we will ask whether reference is really all there is to
meaning and examine arguments that reference itself relies on conceptual knowledge. Here we will
review some basic theories about concepts from philosophy and psychology. Finally, we will discuss
how these ideas have influenced the ways semanticists view their task.
### **2.2 Reference: How Words Point to Things**

#### **2.2.1 Types of Reference**

Let's begin by looking at some major differences in how words can be used to refer. For this
introduction, we will mostly talk about names and noun phrases (which together we can call
**nominals**), as these are the linguistic units most clearly used for referring.

**Referring and Non-Referring Expressions**

We can use this distinction in two ways. First, some linguistic expressions can *never* be used to
refer. Words like *so, very, maybe, if, not,* and *all* cannot pick out an entity in the world.

* **Example 1:** In the sentence "It is **very** hot," the word "very" modifies "hot" but does not
itself point to any object.
* **Example 2:** In "**If** it rains, we will cancel," the word "if" sets up a condition but does not
refer to anything.

These words contribute meaning to the sentences they are in and help the whole sentence denote a
situation, but they are intrinsically **non-referring items**.

By contrast, when someone says the noun "cat" in a sentence like "That **cat** looks vicious," the
noun is being used to identify an entity. So, nouns are **potentially referring expressions**.

The second use of the distinction concerns these potentially referring elements like nouns. It
distinguishes between instances when speakers use them to refer and instances when they do not.

* **Example of Referring Use:** In "They performed **a cholecystectomy** this morning," the
phrase "a cholecystectomy" is used to refer to one specific operation.
* **Example of Non-Referring (Generic) Use:** In "**A cholecystectomy** is a serious procedure,"
the same phrase is not referring to any particular operation. It has a generic interpretation, talking
about the type of procedure in general.

Sometimes a sentence can be ambiguous between a referring and a non-referring reading. A classic
joke is someone saying, "I'm looking for **a woman**." This could mean they are looking for a
specific, known woman (referring), or it could mean they are looking for any woman to, for example,
date (non-referring, generic).

**Constant versus Variable Reference**

Another difference becomes clear when we look at how referring expressions are used across many
different utterances. Some expressions almost always have the same referent.

* **Example 1:** "**The Pacific Ocean**" always refers to the same body of water.
* **Example 2:** "**The Eiffel Tower**" always refers to the same structure.

These are described as having **constant reference**.

Other expressions have a referent that is totally dependent on the context of the utterance. To
identify the referent, we need to know who is speaking, to whom, when, and where.

* **Example 1:** In "**I** wrote to **you**," the words "I" and "you" refer to different people
depending on who is talking and who is listening.
* **Example 2:** In "**She** put **it** in **my** office," we need context to know who "she" is,
what "it" is, and whose office "my" refers to.
Expressions like *I, you, she, it, this,* and *that* are said to have **variable reference**.

In reality, our examples are extremes. As we will see later, most acts of referring rely on some
contextual information. For example, to identify the referent of "**the President of the United
States**," you need to know the year the sentence was uttered.

**Referents and Extensions**

We can also make useful distinctions among the *things* that are referred to. We use the term
**referent** for the specific thing picked out by uttering an expression in a particular context.

* **Example 1:** The referent of "**the capital of Nigeria**" is, since 1991, the city of Abuja.
* **Example 2:** The referent of "**a toad**" in "I've just stepped on a toad" is that one
unfortunate animal.

The term **extension** of an expression is the *set* of all things that could possibly be the referent
of that expression.

* **Example 1:** The extension of the word "**toad**" is the set of all toads that exist, have
existed, or could exist.
* **Example 2:** The extension of the word "**chair**" is the set of all chairs.

As mentioned earlier, the relationship between an expression and its extension is called
**denotation**.

#### **2.2.2 Names**

Names might seem like the simplest case of referring expressions. They are labels for people, places,
and so on, and often seem to have little other meaning. It seems odd to ask for the "meaning" of "Karl
Marx" beyond its use to talk about a specific historical figure.

Context is still important. Names are **definite**, meaning the speaker assumes the listener can
identify the referent.

* **Example:** If someone says, "He looks just like **Brad Pitt**," they assume you know which
famous actor they are talking about.

But how do names actually work? Philosophers have proposed different theories.

One important approach is the **description theory**, associated with philosophers like Russell,
Frege, and Searle. In this theory, a name is seen as a shorthand label for a bundle of knowledge about
the referent—for one or more **definite descriptions**.

* **Example:** The name "**Christopher Marlowe**" might be associated with descriptions like
"the writer of the play *Dr. Faustus*" or "the Elizabethan playwright who was murdered in a tavern in
Deptford." Understanding the name and identifying its referent depend on linking it to the right
description.

Another, very interesting explanation is the **causal theory**, based on the ideas of Kripke and
Donnellan. According to this theory, names are socially inherited. At some original point, a name is
given to a person or place in a "grounding" event. People present at this event begin using the name.
The name is then passed on like a chain from person to person, even to people who have never met
the referent and know very little about them.

* **Example:** Millions of people use the name "**Cleopatra**." Very few of them could provide
an accurate description of her. They use the name successfully because they are part of a long social
chain of use that goes back to the original naming of the historical figure.
The great advantage of the causal theory is that it recognizes that speakers can use names correctly
with very little knowledge of the referent. The description theory emphasizes the role of identifying
knowledge, while the causal theory stresses the role of social knowledge. This debate is important
because the theory chosen for names can also be extended to other nominals, like words for natural
categories (e.g., *giraffe, gold*), which we will look at later.

#### **2.2.3 Nouns and Noun Phrases**

Nouns and noun phrases (NPs) can also be used to refer. Indefinite and definite NPs can operate like
names to pick out an individual.

* **Example:** "I spoke to **a woman** about the noise." (indefinite NP, introducing a new
referent)
* **Example:** "I spoke to **the woman** about the noise." (definite NP, assuming the listener
knows which woman)

Definite noun phrases can also form **definite descriptions**, where the referent is whoever or
whatever fits the description.

* **Example:** "She has a crush on **the captain of the hockey team**." The referent is whoever
holds that position.

An account of reference has to deal with tricky cases where there is no real-world referent, like the
famous philosophical example, "**The King of France** is bald." (There is no King of France.) Or
references to fictional entities, like "**the wizard of Oz**."

NPs can refer to groups of individuals in two ways:

* **Distributively:** Focusing on the individual members of the group. ("**The people in the lift**
avoided each other's eyes." - Each person individually avoided eye contact.)
* **Collectively:** Focusing on the group as a single unit. ("**The people in the lift** proved too
heavy for the lift motor." - The group's combined weight was the problem.)

Nominals can also denote substances, actions, and abstract ideas:

* **Substance:** "Who can afford **coffee**?"
* **Action:** "**Sleeping** is his hobby."
* **Abstract Idea:** "She has a passion for **justice**."

One important referential distinction is between **mass nouns** and **count nouns**, which is
marked grammatically.

* **Count nouns** like "hat" can be counted. We say "a hat," "three hats," "many hats."
* **Mass nouns** like "furniture" typically cannot be counted in the same way. We don't say "∗a
furniture" or "∗three furnitures." We use measures instead: "a piece of furniture," "how much
furniture?"

This distinction is not just about the world, but about how we choose to conceptualize things for the
purpose of language. This is conventionalized differently in different languages; for example, "advice"
is a mass noun in English but a count noun in Spanish ("un consejo" - an advice).

Some nominals have complex denotational behavior, like "**no student**" in "No student enjoyed
the lecture." This doesn't refer to an individual. Its meaning is more like "Of all the students, not a
single one enjoyed the lecture." This is characteristic of **quantifiers** like *each, all, every, some,
none, no*. These words give speakers great flexibility to talk about whole classes or parts of classes.

* **Example 1:** "**Every Frenchman** would recognize his face." (The whole class)
* **Example 2:** "**Some Frenchmen** voted for him twice." (A sub-part of the class)
### **2.3 Reference as a Theory of Meaning**

As we observed earlier, perhaps the simplest theory of meaning is to claim that semantics *is*
reference. In this view, to give the meaning of a word, you just show what it denotes. A simple version
of this theory, as described by Ruth Kempson, might claim:
* Proper names denote individuals.
* Common nouns denote sets of individuals.
* Verbs denote actions.
* Adjectives denote properties of individuals.
* Adverbs denote properties of actions.

However, there are several major problems with this simple theory as a complete explanation of
semantics.

**Problem 1: Words Without Real-World Referents**

Firstly, it seems to predict that many words are meaningless because it's very hard to find a real-world
referent for them. As we saw, words like *so, not, very, but, of* don't point to anything in the world.
If meaning is reference, these words would be meaningless, which is clearly not the case.

**Problem 2: Referring to Non-Existent Things**

Secondly, speakers often use expressions that do not have a referent that exists or has ever existed.

* **Example 1:** "In the painting, a **unicorn** is ignoring a maiden." (Unicorns don't exist.)
* **Example 2:** "**World War III** might be about to start." (This refers to a potential, not actual,
event.)
* **Example 3:** "**Father Christmas** might not visit you this year." (A fictional entity.)

If reference to real-world things is meaning, then we would have to say that words like *unicorn* and
*World War III* are meaningless. But the sentences that contain them clearly *do* have meaning.
Therefore, we need a more sophisticated theory.

**Problem 3: Different Expressions for the Same Referent**

A further problem is that we can use different expressions to refer to the very same individual, and
these expressions have different meanings.

* **Example:** "Then in 1981, **Anwar El Sadat** was assassinated." and "Then in 1981, **the
President of Egypt** was assassinated." Both expressions refer to the same person, but they feel
different in meaning. "The President of Egypt" conveys the idea of a political office, while "Anwar El
Sadat" is a personal label.

You might refer to your next-door neighbor as "**my neighbor**," "**Pat's mother**," or "**the
Head of Science at St. Helen's School**." These all refer to the same person but have different
meanings. It's even possible to understand two expressions without knowing they share the same
referent.

The philosopher Gottlob Frege gave a famous example: a person might understand "**the morning
star**" and "**the evening star**" as two different celestial bodies without knowing that both are
actually the planet Venus. For this person, the sentence "The morning star is the evening star" would
be informative, not a simple tautology like "Venus is Venus."

If we can understand expressions without real-world referents, use different expressions for the same
referent, and even use two expressions without knowing they corefer, then meaning cannot be
exactly the same thing as reference. There must be more to it.

One solution is to follow Frege in distinguishing two aspects of our semantic knowledge: **sense**
(Frege's *Sinn*) and **reference** (Frege's *Bedeutung*). In this division, **sense is primary
because it allows reference**. It is because we understand the *sense* of the expression "the
President of Ireland" that we can use it to refer to a particular individual at any given time. Other
descriptions of the same person will have a different *sense* but the same *reference*.

If we follow this line of argument, our semantic theory becomes more complex. The meaning of an
expression will involve both its sense and its reference. In the next section, we explore what this
"sense" element might be.

### **2.4 Mental Representations: The Idea of "Sense"**

#### **2.4.1 Introduction**

In the last section, we concluded that although reference is a crucial function of language, there must
be more to meaning than just denotation. We called this extra dimension **sense**. For the rest of
this chapter, we will explore the view that "sense" is a level of **mental representation** that stands
between words and the world. In this view, a noun gains its ability to denote because it is associated
with something in the speaker's or hearer's mind.

This solves the problem of talking about non-existent things (like unicorns), as we can have mental
representations for them. But it raises a new question: what are these mental representations like?

One simple and old idea is that these mental entities are **images**. The relationship between the
mental image and the real-world entity would be one of resemblance.

* **Example:** The word "Paris" might call to mind an image of the Eiffel Tower.
* **Example:** The word "your mother" might evoke a mental picture of her face.

This might work for some concrete words and even for imaginary entities like Batman. However, this
theory runs into serious problems with common nouns. Different speakers will have different mental
images for a word like "car" or "house," based on their personal experience. A major problem is with
words like "triangle."

* **Example:** One person's mental image of a "triangle" might be equilateral, another's might be a
right-angled triangle. It's impossible to form a single image that captures all possible triangles. This is
even more difficult for words like "animal," "food," "love," or "justice." What would an image for
"democracy" look like?

So, even if images are associated with some words, they cannot be the whole story.

The most common modification is to hypothesize that the sense of a word is not a visual image but a
more abstract element: a **concept**. This has advantages. A concept could contain the non-visual
features that make a dog a dog, or democracy, democracy. We could also create a definition for
"triangle" (e.g., "a three-sided polygon").

Another advantage for linguists is that they could share the work of describing concepts with
psychologists. Some concepts might be simple (SUN, WATER), while others are complex and tied to
cultural knowledge (MARRIAGE, RETIREMENT).

This seems reasonable, but the problem is that psychologists are still actively debating what concepts
are. Without a clear idea, we are left with vague definitions like "the sense of the word *dog* is the
concept DOG."

At this point, linguists divide. Some, like Ruth Kempson, have been skeptical of psychologists' progress
and don't see the point of basing a theory of meaning on unproven psychological models. They argue
that this just replaces one vague term ("meaning") with another ("concept"). These linguists often
favor a denotational approach and prefer to model sense in a formal, non-psychological way.
Linguists who favor a **representational approach**, however, have gone on to build models of
concepts to form the basis of semantics. We will look at some of these models later. For now, let's
follow this line of inquiry and examine some basic approaches to concepts from psychology.

#### **2.4.2 Concepts**

If we adopt the hypothesis that the meaning of a noun involves a conceptual element, then from a
linguist's perspective, two basic questions are:
1. What form do concepts take? (What do they look like in the mind?)
2. How do children acquire them, along with the words that label them?

In our discussion, we'll focus on concepts that are **lexicalized**—that is, they correspond to a single
word. Not all concepts are like this; some are described by phrases ("a tool for compacting dead
leaves into garden statuary"). Concepts become lexicalized when they are useful enough to need a
quick, single-word label. The word "microwave" (from "microwave oven") is a good example of a
concept that became lexicalized through common use.

When we talk about children acquiring concepts, we have to remember that a child's concepts may
be different from an adult's. Children might **underextend** a concept (using "dog" only for their
own pet) or **overextend** it (using "daddy" for all men). Their concepts reflect what is salient in
their world.

#### **2.4.3 Necessary and Sufficient Conditions**

One traditional approach to describing concepts is to define them using sets of **necessary and
sufficient conditions**. This approach views a concept as a list of features or attributes. If something
has all the features on the list, it is an example of the concept; if it lacks any, it is not.

* **Example for WOMAN:** x is a woman *if and only if*:

* x is human. (necessary)
* x is adult. (necessary)
* x is female. (necessary)
* ...and so on.

The attributes are **necessary** (something must have them to be a woman) and, if the list is
perfect, **sufficient** (the list is enough to define a woman).

The major problem with this approach is that it's very difficult to establish a set of conditions that all
speakers agree on, even for concrete nouns. Let's take "zebra."

* **Potential Conditions:** is an animal, has four legs, is striped, is a herbivore.

* **Problems:** What if we find a zebra that is pure white (no stripes)? It's still a zebra. What about
a zebra born with three legs? Still a zebra. What if a zebra ate some insects? Still a zebra. Which of
these features are truly *necessary*?

These might seem like silly questions, but they have serious consequences. If we can't agree on a
mutual definition for a concept, how can we all use its word-label successfully?

Another argument against this "definitional theory of concepts" comes from the philosopher Hilary
Putnam, who pointed out the role of **ignorance**. Speakers often use words correctly while
knowing very little about the identifying characteristics of the referent.

* **Example:** Many English speakers cannot tell the difference between a beech tree and an elm
tree. Yet, they can understand and truthfully say, "In the 1970s, Dutch elm disease killed a huge
number of British **elms**." They use the word successfully without knowing the necessary and
sufficient conditions for being an elm.
Putnam suggests we rely on a "division of linguistic labor": we assume that there are experts (like
botanists) who *do* know the precise definitions, and we borrow the word from them. This is similar
to the **causal theory** for names, and philosophers like Putnam and Kripke have proposed
extending the causal theory to **natural kind terms** (words for natural categories like *gold, tiger,
water*). We can use the word "silver" without being able to define it scientifically, because we are
part of a social chain that links back to an original grounding of the term with the actual metal.

#### **2.4.4 Prototypes**

Because of the problems with necessary and sufficient conditions, more sophisticated theories of
concepts have been proposed. One very influential proposal is **prototype theory**, developed by
Eleanor Rosch and her colleagues. This model views concepts as having an internal structure where
there are central, typical members and less typical, peripheral members.

* **Example for BIRD:** A **sparrow** or **robin** is a very typical (prototypical) bird. A

**penguin** or **ostrich** is a less typical member of the category.
* **Example for FURNITURE:** A **chair** is a very typical piece of furniture. A **lamp** or
**rug** is less typical.

Rosch's experimental evidence showed that speakers agree more quickly on typical members, they
come to mind faster, and are learned by children earlier. Another finding is that the boundaries
between concepts can be "fuzzy" rather than clear-cut.

This approach allows for the "whale" problem from Chapter 1. A speaker might be unsure if a whale is
a mammal or a fish because whales are not *typical* mammals (they don't live on land), but they
*do* resemble typical fish (they live in the ocean, have fins).

In psychology, there are different interpretations of prototypes. Some researchers think the prototype
is an **abstraction**—a set of characteristic features describing an average bird (small, flies, has
feathers, etc.). Others think we use **exemplars**—memories of actual typical birds we've
encountered (sparrows, pigeons)—and we judge new things by comparing them to these memories.

From within linguistics, Charles Fillmore and George Lakoff have suggested that typicality effects come
from our "folk theories" about the world, which they call **frames** (Fillmore) or **idealized
Cognitive Models (ICMs)** (Lakoff). These are not scientific theories but collections of cultural
knowledge.

* **Example for BACHELOR:** The dictionary definition might be "an unmarried man," but our
cultural ICM of marriage includes ideas about romantic love, eligible partners, and monogamy. This
ICM prevents us from applying the word "bachelor" in a straightforward way to the Pope, a man living
alone on a desert island, or Tarzan. These are "bachelors" by definition, but they are not prototypical.
In this view, using a word involves combining strict semantic knowledge with broader, encyclopedic
cultural knowledge.

#### **2.4.5 Relations Between Concepts**

An important point we've skipped so far is that conceptual knowledge is **relational**. Concepts are
linked to other concepts in a network. Your knowledge of words is also a network, as we'll see in the
next chapter.

* **Example:** If you learn that a "peccary" is a kind of wild pig, your knowledge of the PECCARY
concept "inherits" all your general knowledge about PIGS. You immediately know that peccaries are
probably animals, eat, breathe, and have legs, without being told.

These relations between concepts motivate models of **conceptual hierarchies**. In a hierarchy for
living things, the concept BIRD has attributes like *has wings, can fly*. It doesn't need to specify *is a
living organism* or *has senses* because it inherits those from the higher-level concept ANIMAL. This
structure is efficient and allows us to make predictions.

Proponents of prototype theory have also investigated hierarchies and proposed that they contain
three important levels:
* **Superordinate Level:** Very general (e.g., FURNITURE). Has few common features.
* **Basic Level:** The most useful, common level (e.g., CHAIR, TABLE). Has many common features.
* **Subordinate Level:** Very specific (e.g., ARMCHAIR, KITCHEN CHAIR). Has even more specific
features.

The **basic level** is cognitively primary: it's the level used most in everyday speech, learned first by
children, and at which objects are recognized most quickly.

#### **2.4.6 Acquiring Concepts**

Our second basic question was: how do we acquire concepts? A simple, intuitive theory is **ostensive
definition**—pointing to examples. You point to a dog and say, "That's a dog!" and the child starts to
form the concept DOG.

However, the philosopher W.V.O. Quine pointed out a problem. Ostension usually happens within a
language. His famous example is of a linguist hearing a native speaker say "Gavagai!" when a rabbit
runs by. The linguist doesn't know if "Gavagai" means "rabbit," "Look!," "It's running," "food," or even
"its tail." To understand that a word is a name for a thing, you need some prior linguistic knowledge.
This leads to a deep question: where do our first concepts and words come from? Are we born with
some basic concepts? The acquisition of concepts is clearly more complex than simple pointing.

### **2.5 Words, Concepts, and Thinking**

So far, we've assumed a link between words and concepts. But what is the relationship between these
lexicalized concepts and our general ability to think and reason? There are two famous, opposing
views on this.

#### **2.5.1 Linguistic Relativity**

The idea of **linguistic relativity** is associated with Edward Sapir and Benjamin Lee Whorf. It's the
idea that the language we speak determines or heavily influences the way we think. This idea is
appealing because when we learn different languages, we often find that words don't match up
perfectly.

* **Example 1:** Color words. The range of the English word "purple" might not exactly match the
French word "pourpre."
* **Example 2:** Verbs for getting dressed. English has general verbs like "put on." Japanese and
Korean have different verbs for putting clothes on your head, your feet, your torso, etc.

This observation—that language reflects culture—was important for anthropologists like Franz Boas.
But Sapir and Whorf took it further, suggesting that language doesn't just mirror culture; it
*determines* thought.

* **Sapir:** "Human beings are very much at the mercy of the particular language which has
become the medium of expression for their society... The 'real world' is to a large extent
unconsciously built up on the language habits of the group."
* **Whorf (stronger):** "We cut nature up, organize it into concepts, and ascribe significances as we
do, largely because we are parties to an agreement... codified in the patterns of our language. The
agreement is... absolutely obligatory; we cannot talk at all except by subscribing to the organization
and classification of data which the agreement decrees."
If this strong view is correct, then speakers of different languages literally think in different ways. This
would make creating a universal semantic metalanguage very difficult, as any metalanguage would be
biased by the linguist's native language.

#### **2.5.2 The Language of Thought Hypothesis**

The idea of linguistic relativity is rejected by many linguists and cognitive scientists. They argue that it
is a fallacy to equate thought with language. They present two main types of counter-arguments.

**Argument 1: Evidence of Thinking Without Language**

* **Example 1:** Babies and higher primates show evidence of reasoning and memory, but they
don't have language.
* **Example 2:** Artists and scientists often report having "non-linguistic" insights or thinking in
images.
* **Example 3:** Psychological experiments show that people can mentally manipulate visual
images (rotating them, scanning them) in ways that don't seem to involve language.

This evidence is used to argue that we think in a separate, non-verbal computational system: a
**Language of Thought** or **Mentalese**. When we speak, we translate our thoughts from
Mentalese into English, Arabic, or whatever spoken language we use.

**Argument 2: Language Underspecifies Meaning**

Research in semantics and pragmatics shows that what we say is often a compressed version of what
we mean. Listeners have to fill in the gaps using context and common sense.

* **Example:** If someone says, "I'm tired," they might mean "Let's go home," "I can't work
anymore," or "Can you make dinner?" The language underspecifies the full intended meaning.

This fits naturally with the idea that we are translating from a richer "language of thought" into a
sparser spoken language.

A natural extension of this view is that everybody's Mentalese is essentially the same. This leads to a
position directly opposed to linguistic relativity: all human beings share the same basic cognitive
architecture and thought processes, despite speaking different languages.

#### **2.5.3 Thought and Reality**

This leads to a final, deep layer of questions that semanticists sometimes consider: the relationship
between thought and reality. These are questions of **ontology** (what exists?) and
**epistemology** (how can we know?).

* **Idealism:** Does reality exist independently of our minds?

* **Objectivism:** Can we perceive the world as it truly is?
* **Mental Constructivism:** Is our view of reality filtered through our biological and cultural
conceptual systems?

Different semantic theories often assume different answers to these basic questions. For a linguist
who just wants to describe the meaning of words in Swahili or English, these can feel like heavy
philosophical problems. One understandable response is for linguists to focus only on language itself
—to study the meaning relations between words *within* a language—and leave the big questions
about mind and reality to philosophers and psychologists. This inward focus, which we could call
**linguistic solipsism**, leads to an interest in describing semantic relations like synonymy,
antonymy, and ambiguity, which we will explore in the next chapter.

### **2.6 Summary of the Chapter**

In this chapter, we have seen that while reference is a key part of how language works, it is difficult to
build a complete theory of meaning on reference alone. Our semantic knowledge seems to include
both **reference** and **sense**.

We identified two major approaches to meaning:

1. **Denotational/Referential Approach:** Meaning comes from language being linked to the
external world.
2. **Representational Approach:** Meaning comes from language being a reflection of our internal
conceptual structures.

Each approach has its challenges. How do denotational approaches handle imaginary entities? For
representational approaches, do we need a fully developed theory of conceptual structure to do
semantics?

We also saw how these issues are influenced by philosophy and psychology, leading linguists to adopt
one of three main positions:
1. **Focus on Language-Internal Relations:** Leave issues of mind and reality to other fields and
concentrate on describing sense relations (synonymy, antonymy, etc.) within and between languages.
2. **Strengthen Denotational Theory:** Develop a sophisticated denotational theory that can handle
all the different types of reference, including reference to non-existent things.
3. **Embrace Conceptual Structure:** Decide that meaning does rely on conceptual structure and
work to build models of linguistic concepts.

CHAPTER 3

Of course. Here is a comprehensive, simplified explanation of Chapter 2 from John I. Saeed's

"Semantics." This guide breaks down every concept into easy-to-understand paragraphs, with clear
headings and two examples for each major topic, using extremely simple vocabulary and sentence
structure.

### **Introduction: How Words Connect to the World**

This chapter asks a very basic but important question: How do words work? How can we make sounds
with our mouths that tell another person about something we see, hear, or think? For example, how
can you describe your favorite movie or what you ate for lunch just by using words? All human
languages have this amazing power. They let us build a model of the world using sounds and symbols.
We use words to point to specific things, people, or places.

* **Example 1:** If you say, "Look at **that car**!" you are using the words "that car" to point out
one specific vehicle.
* **Example 2:** If you say, "I live in **Toronto**," you are using the word "Toronto" to identify a
specific city.

In the study of meaning, which is called semantics, this act of pointing to something with words is
called **referring** or **denoting**. So, you can use the word "Toronto" to refer to, or denote, the
city. The actual thing in the world that you are pointing to—the city itself—is called the **referent**.

Some language experts make a careful difference between "refer" and "denote." They use
**denote** to talk about the general, long-term relationship between a word and the world. They
use **refer** to talk about the specific action of a speaker using a word to point to something at a
particular moment.

* **Example of Denoting vs. Referring:** If you say, "A **bird** is on the **fence**," you, as the
speaker, are using the words "a bird" and "the fence" to **refer** to specific things right now. At the
same time, the words "bird" and "fence" themselves **denote** entire categories. The word "bird"
denotes all birds everywhere, and "fence" denotes all fences.

Another key point is that **denotation** is a stable part of the language. The word "bird" always
denotes the category of birds. **Reference**, however, changes with the situation. What a speaker
refers to with the word "bird" depends on what they are looking at or talking about at that moment.

Experts who study meaning have different ideas about how this connection between words and the
world works. Two main views are very important today.

The first view is called the **referential** or **denotational approach**. For experts who follow this
view, the act of connecting words to the world *is* the most important part of meaning. To explain
what words mean, you need to show how they link to real things and situations. In this view, nouns
have meaning because they name objects, and sentences have meaning because they describe real
events.

* **Example:** The sentence "There is a pizza in the oven" has a different meaning from "There is
no pizza in the oven" because they describe two different situations in the world. One is true if the
pizza is there, and the other is true if it is not.

The second view is called the **representational approach**. Experts who follow this view believe
our ability to talk about the world comes from our mental models of it. Our language reflects our own
personal theory about what the world is like and what kinds of things exist in it. A speaker can
describe the same real event in different ways, depending on the patterns of their language.

* **Example:** In English, you can describe the same situation in two ways: "Sarah **is running**"
(an activity) or "Sarah **is a runner**" (a state). The language gives you different ways to
conceptualize the same thing.

This becomes even clearer when we look at different languages. Look at how different languages say
"you have a cold":

* **English:** "You **have** a cold." (This makes it sound like you *possess* the illness.)
* **Somali:** Literally, "A cold **has** you." (This makes it sound like the illness *possesses you*.)
* **Irish:** Literally, "A cold **is on** you." (This makes it sound like the illness is *located on
you*.)

The important idea is that different languages have built-in, conventional ways of thinking about the
same situation, and this shapes how we talk about it.

Theories of meaning are called **representational** when they focus on how our descriptions of
reality are shaped by the thinking patterns built into our language.

We can think of these two approaches as looking at two different parts of the same process. In
**referential theories**, meaning comes from language being connected to the outside world. In
**representational theories**, meaning comes from language being a picture of our internal
thoughts and ideas. This difference will come up again and again in the study of semantics.

These two approaches are influenced by ideas from philosophy and psychology. This chapter will look
at the most important of these ideas. We will start by looking at the different ways words can be used
to refer. Then we will ask if reference is really all there is to meaning. We will see arguments that
reference itself depends on what we know in our minds. Here, we will look at some basic theories
about concepts. Finally, we will discuss how these ideas have changed the way semanticists do their
work.

### **Reference: How Words Point to Things**

This section looks at the different ways we use words to point to things in the world.

#### **Types of Reference**

Let's start by looking at the big differences in how words can refer. We will mostly talk about names
and noun phrases (like "the big dog"), which are the parts of language most clearly used for pointing
at things.

**Referring and Non-Referring Expressions**

We can look at this difference in two ways. First, some words can *never* be used to point to
anything. Words like *so, very, maybe, if,* and *not* do not pick out any object, person, or place in
the world.

* **Example 1:** In the sentence "It is **very** hot," the word "very" makes "hot" stronger, but it
does not point to any thing.
* **Example 2:** In "**If** it rains, we will stay inside," the word "if" creates a condition, but it does
not refer to an object.

These words add meaning to the sentences they are in, but they are inherently **non-referring
items**.

On the other hand, when you use a noun like "cat" in a sentence like "That **cat** is sleeping," the
noun is being used to identify a thing. So, nouns are **potentially referring expressions**.

The second way we use the distinction is for these potentially referring words. It helps us see the
difference between when a speaker *is* using them to refer and when they are *not*.

* **Example of Referring Use:** In "The doctor performed **an operation** this morning," the
phrase "an operation" refers to one specific medical procedure that happened.
* **Example of Non-Referring (Generic) Use:** In "**An operation** requires a skilled surgeon," the
same phrase is not talking about any specific operation. It is talking about the general idea of
operations. This is called a generic interpretation.

Sometimes, a sentence can have two possible meanings: one where a phrase is referring and one
where it is not. A classic example is someone saying, "I'm looking for **a partner**." This could mean
they are looking for a specific person they know (referring), or it could mean they are looking for any
person to be their partner in general (non-referring).

**Constant versus Variable Reference**

Another big difference is seen when we look at how referring expressions are used in many different
sentences. Some expressions almost always point to the same thing.

* **Example 1:** "**The Moon**" always refers to the same celestial body that orbits the Earth.
* **Example 2:** "**Mount Everest**" always refers to the same mountain.

These are described as having **constant reference**.

Other expressions point to different things depending completely on the context. To know what they
are pointing to, you need to know who is speaking, who they are speaking to, and when and where
they are speaking.

* **Example 1:** In "**I** like **your** shirt," the words "I" and "your" refer to different people in
every conversation.
* **Example 2:** In "**He** left **it** over **there**," you need the context to know who "he" is,
what "it" is, and where "there" is.
Words like *I, you, he, she, it, this,* and *that* are said to have **variable reference**.

In real life, most referring is somewhere in between. For example, to know what "**the President of
France**" refers to, you need to know the year the sentence was spoken, because the person in that
job changes.

**Referents and Extensions**

We can also make useful distinctions about the *things* that are being pointed to. We use the term
**referent** for the specific thing a speaker points to when they use an expression in a specific
situation.

* **Example 1:** The referent of "**the current King of England**" is King Charles III.
* **Example 2:** The referent of "**my coffee mug**" in "Pass me my coffee mug" is that one
specific mug.

The term **extension** of an expression is the entire *set* or *collection* of all things that could
ever be the referent of that expression.

* **Example 1:** The extension of the word "**cat**" is the set of all cats that exist, have existed,
or will exist.
* **Example 2:** The extension of the word "**book**" is the set of all books.

As we learned before, the relationship between a word and its extension is called **denotation**.

#### **Names**

Names might seem like the simplest kind of referring expression. They are labels for people, places,
and companies. It feels strange to ask for the "meaning" of the name "Taylor Swift" beyond its use to
talk about that specific famous singer.

Context is still important. Names are **definite**, which means the speaker assumes the listener
knows who or what they are talking about.

* **Example:** If you say, "She reminds me of **Taylor Swift**," you assume the listener knows
which pop star you mean.

But how do names actually work? Philosophers have different theories.

One important theory is the **description theory**, linked to philosophers like Bertrand Russell and
John Searle. In this theory, a name is like a shortcut for a collection of knowledge about the person or
thing—for one or more **definite descriptions**.

* **Example:** The name "**William Shakespeare**" might be linked to descriptions like "the man
who wrote the play *Hamlet*" or "the most famous English playwright." Understanding the name
depends on knowing these descriptions.

Another, very different explanation is the **causal theory**, based on the ideas of Saul Kripke.
According to this theory, names are passed along socially. At some starting point, a name is given to a
person or place in a "baptism" or "grounding" event. People who were there start using the name.
The name is then passed from person to person, like a chain, even to people who have never met the
person and know almost nothing about them.

* **Example:** Millions of people use the name "**Julius Caesar**." Most of them could not list
many facts about him. They use the name correctly because they are part of a long social chain of use
that goes all the way back to the real Julius Caesar.
The big advantage of the causal theory is that it explains how we can use names correctly even when
we are ignorant. The description theory focuses on the knowledge we have, while the causal theory
focuses on our social connections. This debate is important because the theory we choose for names
can also be used for other words, like words for natural categories (e.g., *tiger, water*), which we will
look at later.

#### **Nouns and Noun Phrases**

Nouns and noun phrases (NPs) can also be used to refer. Indefinite NPs (like "a dog") and definite NPs
(like "the dog") can work like names to pick out an individual.

* **Example:** "I saw **a movie** last night." (indefinite NP, introducing a new thing into the
conversation)
* **Example:** "I saw **the movie** you recommended." (definite NP, assuming the listener knows
which movie)

Definite noun phrases can also be **definite descriptions**, where the referent is whatever fits the
description.

* **Example:** "She is dating **the tallest player on the team**." The referent is whoever fits that
description.

Any theory of reference has to explain tricky cases where there is no real thing being referred to, like
the famous sentence, "**The King of France** is bald." (There is no King of France.) Or when we talk
about fictional things, like "**Harry Potter**."

Noun phrases can refer to groups of people in two different ways:

* **Distributively:** This means we are thinking about the individual members of the group. ("**The
students** in the class each raised their hand." - Each student individually raised a hand.)
* **Collectively:** This means we are thinking about the group as a single unit. ("**The students**
in the class elected a representative." - The group, as a whole, did one action.)

Nouns can also point to substances, actions, and abstract ideas:

* **Substance:** "I need to buy **milk**."
* **Action:** "**Running** is good for your health."
* **Abstract Idea:** "They fought for **freedom**."

One important distinction is between **mass nouns** and **count nouns**, which is shown by
grammar.

* **Count nouns** like "chair" can be counted. We say "a chair," "two chairs," "many chairs."
* **Mass nouns** like "water" usually cannot be counted directly. We don't say "∗a water" or
"∗two waters." We use measures: "a glass of water," "two liters of water."

This distinction is not just about the world, but about how we choose to think about things for
language. This is different in different languages. For example, the word "furniture" is a mass noun in
English, but in Spanish, the equivalent word ("muebles") is a count noun.

Some noun phrases have complex behavior, like "**no teacher**" in "No teacher was happy with the
new rules." This doesn't point to one person. Its meaning is more like "Of all the teachers, not a single
one was happy." This is a feature of **quantifiers** like *each, all, every, some, none, no*. These
words let us talk easily about whole groups or parts of groups.

* **Example 1:** "**Every student** passed the test." (The whole group)
* **Example 2:** "**Some students** failed the test." (A part of the group)
### **Reference as a Theory of Meaning**

As we saw, the simplest theory of meaning is to say that semantics *is* reference. In this view, to give
the meaning of a word, you just show what it points to in the world. A simple version of this theory
might say:
* Names denote individuals.
* Common nouns denote sets of individuals.
* Verbs denote actions.
* Adjectives denote properties of individuals.
* Adverbs denote properties of actions.

However, there are several major problems with this simple theory as a full explanation of meaning.

**Problem 1: Words Without Real-World Referents**

First, this theory would say that many words are meaningless because they don't point to anything in
the real world. As we saw, words like *so, not, very,* and *but* don't point to any object. If meaning
is reference, these words would have no meaning, which is clearly wrong. They are very important for
making sentences work.

**Problem 2: Referring to Non-Existent Things**

Second, people often use words for things that do not exist or have never existed.

* **Example 1:** "The story had **a dragon** in it." (Dragons don't exist.)
* **Example 2:** "I dreamt about **a purple elephant**." (This refers to something from a dream,
not reality.)
* **Example 3:** "**Santa Claus** brings presents at Christmas." (A fictional character.)

If meaning required a real-world referent, then words like *dragon* and *Santa Claus* would be
meaningless. But the sentences that contain them are perfectly understandable. So, we need a better
theory.

**Problem 3: Different Expressions for the Same Referent**

A third problem is that we can use different expressions to talk about the exact same person or thing,
and these expressions have different meanings.

* **Example:** "**Mark Zuckerberg** founded Facebook." and "**The CEO of Meta** founded
Facebook." Both expressions refer to the same person, but they feel different. "The CEO of Meta" tells
you about his job, while "Mark Zuckerberg" is his personal name.

You might talk about your friend as "**my friend**," "**Sarah**," or "**the best soccer player on
our team**." These all refer to the same person but give you different information. It is even possible
to understand two expressions without knowing they point to the same thing.

The philosopher Gottlob Frege gave a famous example: a person might know "**the Morning Star**"
as the bright star in the morning sky and "**the Evening Star**" as the bright star in the evening sky,
without knowing they are both the planet Venus. For this person, the sentence "The Morning Star is
the Evening Star" is new and surprising information, not a boring statement like "Venus is Venus."

If we can understand words for non-existent things, use different words for the same thing, and use
two words without knowing they are the same, then meaning cannot be just reference. There must
be something more.

One solution is to follow Frege and split our knowledge of meaning into two parts: **sense** (Frege's
*Sinn*) and **reference** (Frege's *Bedeutung*). In this division, **sense is more important
because it allows us to refer**. It is because we understand the *sense* of the phrase "the Prime
Minister of Canada" that we can use it to find the right person at any given time. Another description
of the same person, like "Justin Trudeau," would have a different *sense* but the same *reference*.
If we agree with this argument, our theory of meaning becomes more complex. The meaning of an
expression involves both its sense and its reference. Next, we will explore what this "sense" might be.

### **Mental Representations: The Idea of "Sense"**

#### **Introduction**

In the last section, we decided that even though reference is a key job of language, there must be
more to meaning than just pointing. We called this extra part **sense**. For the rest of this chapter,
we will explore the idea that "sense" is a kind of **mental representation** that sits between words
and the world. In this view, a word can point to something because it is linked to something in the
speaker's or listener's mind.

This solves the problem of talking about non-existent things (like unicorns), because we can have
mental representations for them. But it creates a new question: what are these mental
representations like?

One simple, old idea is that these mental things are **pictures** or **images**. The link between
the mental picture and the real thing would be that they look alike.

* **Example:** The word "apple" might make you think of a picture of a red apple.
* **Example:** The word "your home" might bring an image of your house to mind.

This might work for some concrete words and even for imaginary things like Superman. But this
theory has big problems with common nouns. Different people will have different mental pictures for
a word like "car" or "tree," based on their own life. A major problem is with words like "triangle."

* **Example:** One person's mental image of a "triangle" might be skinny and tall, while another
person's might be short and wide. It's impossible to have one picture that includes all possible
triangles. This is even harder for words like "love," "justice," or "democracy." What would a picture of
"democracy" look like?

So, even if pictures are part of how we think of some words, they can't be the whole story.

The most common solution is to suggest that the sense of a word is not a picture but a more abstract
thing: a **concept**. This has benefits. A concept could include the non-visual features that make a
dog a dog, or that make democracy what it is. We could also write a definition for "triangle" (e.g., "a
shape with three straight sides").

Another benefit for linguists is that they could share the work of describing concepts with
psychologists. Some concepts might be simple (like WATER), while others are complex and tied to
culture (like MARRIAGE).

This seems reasonable, but the problem is that psychologists are still arguing about what concepts
are. Without a clear answer, we are stuck with vague statements like "the sense of the word *dog* is
the concept DOG."

At this point, linguists split into two groups. Some, like Ruth Kempson, are skeptical of psychology and
don't want to build a theory of meaning on unproven models. They think this just replaces one fuzzy
word ("meaning") with another ("concept"). These linguists often prefer a denotational approach and
like to model sense in a formal, logical way that isn't about the mind.

Linguists who prefer a **representational approach**, however, have gone ahead and built models of
concepts to use as the foundation for semantics. We will look at some of these models later. For now,
let's follow this path and look at some basic ideas about concepts from psychology.
#### **Concepts**

If we assume that the meaning of a noun involves a concept, then from a linguist's point of view, two
basic questions are:
1. What are concepts like? (What is their form in the mind?)
2. How do children learn them, and learn the words that go with them?

We will focus on concepts that are **lexicalized**—this means they have their own single word. Not
all concepts are like this; some need a whole phrase to describe them ("a person who collects rare
stamps"). Concepts become lexicalized when they are useful enough to need a quick label. The word
"internet" is a good example of a concept that became lexicalized because it became so common.

When we talk about children learning concepts, we have to remember that a child's concepts are not
the same as an adult's. Children might **underextend** a concept (using "dog" only for their own
family pet) or **overextend** it (using "daddy" for all men). Their concepts are based on what is
important and noticeable in their own small world.

#### **Necessary and Sufficient Conditions**

One traditional way to describe a concept is to define it using a list of **necessary and sufficient
conditions**. This approach sees a concept as a checklist of features. If something has all the features
on the list, it is an example of the concept. If it is missing even one feature, it is not.

* **Example for BACHELOR (in a simple view):** x is a bachelor *if and only if*:
* x is human. (necessary)
* x is an adult. (necessary)
* x is male. (necessary)
* x has never been married. (necessary)

The features are **necessary** (you must have them to be a bachelor) and, if the list is perfect,
**sufficient** (the list is all you need to define a bachelor).

The biggest problem with this approach is that it is very hard to make a list that everyone agrees on,
even for simple words. Let's try "game."

* **Potential Conditions:** has rules, is fun, has winners and losers.

* **Problems:** Is a game of catch with a baby a "game"? It might not have strict rules. Is a difficult
and frustrating video game still "fun"? Is a game of solitaire (played alone) a "game" even if there is
no other player to lose? Which features are truly *necessary*?

These might seem like small points, but they matter a lot. If we can't agree on a shared definition for a
concept, how can we all use the word correctly?

Another strong argument against this "checklist" theory of concepts comes from the philosopher
Hilary Putnam, who pointed out the role of **ignorance**. People often use words correctly even
when they know very little about the thing they are naming.

* **Example:** Many people cannot tell the difference between an ash tree and an oak tree. Yet,
they can understand and say, "The **ash** tree in my backyard is very old." They use the word
correctly without knowing the scientific checklist for what makes an ash tree an ash tree.

Putnam suggests we rely on a "division of linguistic labor." This means we trust that there are experts
(like botanists) who *do* know the precise definitions, and we borrow the word from them. This is
very similar to the **causal theory** for names. Philosophers like Putnam and Kripke have proposed
using the causal theory for **natural kind terms** (words for natural categories like *gold, tiger,
water*). We can use the word "tiger" without knowing its DNA, because we are part of a social chain
that goes back to the original naming of the animal.
#### **Prototypes**

Because of the problems with necessary and sufficient conditions, smarter theories of concepts have
been developed. One very influential theory is **prototype theory**, created by Eleanor Rosch. This
model says that concepts have an internal structure with central, typical examples and less typical,
peripheral examples.

* **Example for FRUIT:** An **apple** or **orange** is a very typical (prototypical) fruit. An

**avocado** or **olive** is a less typical member of the category.
* **Example for VEHICLE:** A **car** is a very typical vehicle. A **hot air balloon** or **sled** is
less typical.

Rosch's experiments showed that people agree more quickly that a sparrow is a bird than that a
penguin is a bird. Typical members come to mind faster and are learned by children earlier. She also
found that the lines between categories can be fuzzy, not sharp.

This approach explains the "whale" problem from Chapter 1. A person might be unsure if a whale is a
mammal or a fish because whales are not *typical* mammals (they don't live on land), but they *do*
look like typical fish (they live in the ocean and have fins).

In psychology, there are different ideas about what a prototype is. Some researchers think the
prototype is an **abstraction**—a set of average features that describes a typical bird (small, flies,
sings, has feathers). Others think we use **exemplars**—memories of actual, typical birds we have
seen (like robins or sparrows)—and we compare new things to these memories.

Linguists like Charles Fillmore and George Lakoff have suggested that these typicality effects come
from our everyday "folk theories" about the world, which they call **frames** (Fillmore) or
**idealized Cognitive Models (ICMs)** (Lakoff). These are not scientific theories, but packages of
cultural knowledge.

* **Example for BACHELOR (revisited):** The dictionary definition is "an unmarried man," but our
cultural ICM of marriage includes ideas about dating, being of a certain age, and being in a society
where marriage is common. This ICM is why it feels strange to call the Pope, a five-year-old boy, or a
man who has lived alone on a desert island his whole life a "bachelor." They are technically bachelors,
but they are not prototypical. In this view, using a word mixes strict meaning with broader, cultural
knowledge.

#### **Relations Between Concepts**

An important point we have missed so far is that conceptual knowledge is **relational**. Concepts
are connected to other concepts in a big network. Your knowledge of words is also a network, as we
will see in the next chapter.

* **Example:** If you learn that a "lemur" is a kind of primate, your knowledge of the LEMUR
concept automatically gets all your general knowledge about PRIMATES. You instantly know that
lemurs are probably intelligent, have forward-facing eyes, and can grasp things, without being told.

These relationships between concepts lead to models of **conceptual hierarchies**. In a hierarchy

for animals, the concept BIRD has features like *has wings* and *can fly*. It doesn't need to list *eats
food* or *breathes air* because it gets those from the higher-level concept ANIMAL. This structure is
very efficient and helps us make good guesses.

Researchers in prototype theory have also studied hierarchies and found three important levels:
* **Superordinate Level:** Very general (e.g., FURNITURE). Has few common features. (What do a
table, a lamp, and a rug all have in common? Not much.)
* **Basic Level:** The most useful, common level (e.g., CHAIR, BED). Has many common features.
(You can easily picture a chair and say what it's for.)
* **Subordinate Level:** Very specific (e.g., ROCKING CHAIR, DINING CHAIR). Has even more specific
features.

The **basic level** is the most important for our minds: it's the level we use most in daily talk, it's the
first level learned by children, and it's the level at which we recognize objects the fastest.

#### **Acquiring Concepts**

Our second basic question was: how do we get concepts? A simple, common-sense theory is
**ostensive definition**—which means pointing at examples. You point to a dog and say, "Dog!" and
the child starts to learn the concept DOG.

However, the philosopher W.V.O. Quine showed a big problem with this. Pointing usually happens
when you already speak a language. His famous example is of a linguist who hears a native speaker
say "Gavagai!" when a rabbit runs by. The linguist doesn't know if "Gavagai" means "rabbit," "Look!,"
"It's running," "food," or even "furry tail." To understand that a word is a name for a whole object,
you need some language knowledge already. This leads to a deep puzzle: where do our very first
concepts and words come from? Are we born with some basic concepts? Learning concepts is clearly
much more complicated than just pointing.

### **Words, Concepts, and Thinking**

So far, we have assumed a link between words and concepts. But what is the relationship between
these word-linked concepts and our general ability to think? There are two famous, opposite views on
this.

#### **Linguistic Relativity**

The idea of **linguistic relativity** is connected to Edward Sapir and Benjamin Lee Whorf. It is the
idea that the language you speak decides or strongly shapes the way you think. This idea is attractive
because when we learn other languages, we often find that words don't line up perfectly.

* **Example 1:** Color words. Different languages divide the color spectrum differently. Some
languages have a single word for what English calls "blue" and "green."
* **Example 2:** Words for family. English has one word for "aunt," but some languages have
different words for your mother's sister and your father's sister.

This observation—that language reflects culture—was important for anthropologists. But Sapir and
Whorf went further. They suggested that language doesn't just mirror thought; it *determines* it.

* **Sapir said:** "Human beings are very much at the mercy of the particular language which has
become the medium of expression for their society... The 'real world' is to a large extent
unconsciously built up on the language habits of the group."
* **Whorf said (more strongly):** "We cut nature up, organize it into concepts, and ascribe
significances as we do, largely because we are parties to an agreement... codified in the patterns of
our language. The agreement is... absolutely obligatory; we cannot talk at all except by subscribing to
the organization and classification of data which the agreement decrees."

If this strong view is right, then people who speak different languages actually think in different ways.
This would make it very hard to create a universal language for describing meaning, because any such
language would be biased by the linguist's own native language.

#### **The Language of Thought Hypothesis**

The idea of linguistic relativity is rejected by many linguists and cognitive scientists. They argue that it
is a mistake to confuse thought with language. They give two main kinds of counter-arguments.

**Argument 1: Evidence of Thinking Without Language**

* **Example 1:** Babies and intelligent animals like chimpanzees can solve problems and remember
things, but they do not have human language.
* **Example 2:** People often report having "aha!" moments or thinking in pictures, not words.
* **Example 3:** Experiments show that people can rotate shapes in their minds or scan mental
maps without using language.

This evidence is used to argue that we think in a separate, non-verbal system: a **Language of
Thought** or **Mentalese**. When we speak, we translate our thoughts from Mentalese into
English, Japanese, or whatever spoken language we know.

**Argument 2: Language Underspecifies Meaning**

Research shows that what we say is often a short, simplified version of what we mean. Listeners have
to use context and common sense to fill in the full meaning.

* **Example:** If someone says, "The coffee is hot," they might mean "Be careful not to spill it," "I
need a coaster," or "This is just how I like it." The words themselves don't specify the full message.

This fits perfectly with the idea that we are translating from a rich, detailed "language of thought" into
a more limited spoken language.

A natural conclusion from this view is that everyone's Mentalese is basically the same. This leads to a
position that is the direct opposite of linguistic relativity: all human beings share the same basic
thinking machinery, even though they speak different languages.

#### **Thought and Reality**

This brings us to a final, very deep layer of questions that semanticists sometimes think about: the
relationship between thought and reality. These are questions of **ontology** (what really exists?)
and **epistemology** (how can we know what exists?).

* **Idealism:** Does the world exist outside of our minds, or is it all in our heads?
* **Objectivism:** Can we see the world exactly as it is?
* **Mental Constructivism:** Is our view of the world filtered through our biological senses and our
cultural ideas?

Different semantic theories often start with different answers to these basic questions. For a linguist
who just wants to describe the meanings of words in a specific language, these can feel like heavy
philosophical problems that are not their job to solve. One understandable reaction is for linguists to
focus only on language itself—to study the meaning relations between words *inside* a language—
and leave the big questions about mind and reality to philosophers and psychologists. This inward
focus, which we could call **linguistic solipsism**, leads to an interest in describing relations like
synonymy (sameness of meaning), antonymy (oppositeness), and ambiguity, which we will explore in
the next chapter.

### **Summary of the Chapter**

In this chapter, we have seen that even though reference is a key part of how language works, it is
very hard to make a complete theory of meaning based only on reference. Our knowledge of meaning
seems to include both **reference** and **sense**.

We identified two major approaches to meaning:

1. **Denotational/Referential Approach:** Meaning comes from language being linked to the
external world.
2. **Representational Approach:** Meaning comes from language being a picture of our internal
concepts.

Each approach has its own difficulties. How can denotational approaches explain imaginary things?
For representational approaches, do we need a perfect theory of concepts before we can do
semantics?

We also saw how these issues are influenced by philosophy and psychology, which leads linguists to
choose one of three main paths:
1. **Focus on Language-Internal Relations:** Ignore the problems of mind and reality and just
describe how words relate to each other inside a language (like synonyms and antonyms).
2. **Strengthen Denotational Theory:** Build a very smart and powerful denotational theory that
can explain all types of reference, even to non-existent things.
3. **Embrace Conceptual Structure:** Decide that meaning really does depend on conceptual
structure and work hard to build good models of the concepts behind words.

CHAPTER 6

### **1. The Basic Question of Chapter 6: How Do Words Relate to the World?**

This chapter tries to answer a very simple but deep question. How can we use sounds or marks on
paper to talk about the world around us? When you say, "There's a cat on the sofa," how do those
sounds tell your listener about a real situation? All human languages have this incredible power. They
let us build a model of reality using words. We use language to pick out, or identify, specific things,
people, and places.

* **Example 1:** You point and say, "Look at **that beautiful flower**." You are using words to pick
out one specific flower in the garden.
* **Example 2:** You tell a friend, "I'm going to **the library**." You are using the word "library" to
identify a specific place.

This act of using words to pick out something in the world is central to meaning. In semantics, the
study of meaning, this act has a special name.

### **2. The Core Concepts: Referring, Denoting, and the Referent**

The act of picking out something with words is called **referring** or **denoting**. If you use the
word "London," you are referring to, or denoting, the city of London. The actual thing in the world
that you are talking about is called the **referent**. So, for the word "London," the city itself is the
referent.

Some experts make a helpful distinction between "refer" and "denote." They use **denote** to
describe the general, permanent relationship between a word and the world. They use **refer** to
describe the specific action of a person using a word to point to something at a particular time.

* **Example of Denoting vs. Referring:** Imagine you say, "A **dog** chased a **car**." In this
moment, you are using the words "a dog" and "a car" to **refer** to one specific dog and one
specific car in a real event. However, the words "dog" and "car" themselves **denote** entire
categories. The word "dog" denotes all dogs, and "car" denotes all cars. Denotation is the word's
potential; reference is the speaker's action.

This leads to another key point. **Denotation** is a stable part of the language system. The word
"tree" always denotes the category of trees. **Reference**, however, changes with every use. What
a speaker refers to with the word "tree" depends entirely on the situation—it could be the oak tree in
their yard or a pine tree in a forest.

### **3. Two Main Approaches to Meaning: Referential vs. Representational**

Linguists have different ideas about how this word-world connection works. Two main views are very
important.

The first view is called the **referential** or **denotational approach**. For linguists who follow this
view, the most important part of meaning is this direct link between words and the world. To explain
what a word means, you must show what thing or set of things it points to in reality. In this view, a
noun means something because it names an object, and a sentence means something because it
describes a real situation.

* **Example:** The sentence "The window is open" has a different meaning from "The window is
closed" because they describe two opposite situations in the world. One is true if the window is open,
and the other is true if it is closed.

The second view is called the **representational approach**. Linguists who follow this view believe
that our ability to talk about the world comes from our mental models of it. Language is a reflection of
how we think about the world. A speaker can describe the same event in different ways, influenced
by the patterns of their language.

* **Example:** In English, you can describe a person in two ways: "John **is sick**" (a state) or
"John **has a sickness**" (a possession). The language provides different ways to think about the
same condition.

This becomes even clearer when we compare different languages. Look at how different languages
express the idea of "having a cold":

* **English:** "You **have** a cold." (This frames the illness as something you *possess*.)
* **Somali:** Literally, "A cold **has** you." (This frames the illness as something that *possesses
you*.)
* **Irish:** Literally, "A cold **is on** you." (This frames the illness as something *located on you*.)

The key idea here is that different languages have built-in, conventional ways of thinking about
situations, and this shapes how we talk about them. Theories of meaning are called
**representational** when they focus on how our language is shaped by our internal conceptual
structures.

We can see these two approaches as two sides of the same coin. **Referential theories** look
outward, connecting meaning to the external world. **Representational theories** look inward,
connecting meaning to our minds. This basic difference is found throughout the study of semantics.

### **4. Starting with Reference: How Words Point to Things**

Now, let's look more closely at the different ways we use words to point to things. We will focus on
names (like "Anna") and noun phrases (like "the tall woman"), as these are the parts of language most
directly used for referring.

#### **4.1. Types of Referring Expressions**

**Referring and Non-Referring Expressions**

We can understand this distinction in two ways. First, some words can *never* be used to point to
anything in the world. Words like *very, if, not,* and *all* do not pick out any object, person, or
place.
* **Example 1:** In "The soup is **very** hot," the word "very" intensifies "hot" but does not itself
refer to any thing.
* **Example 2:** In "**If** you go, I will go," the word "if" sets up a condition but does not point to
an object.

These words are essential for building meaningful sentences, but they are inherently **non-referring
items**.

In contrast, when you use a noun like "book" in a sentence like "That **book** is interesting," the
noun is being used to identify a thing. So, nouns are **potentially referring expressions**.

The second way we use this distinction is for these potentially referring words. It helps us see when a
speaker *is* using them to refer and when they are *not*.

* **Example of Referring Use:** In "I need to buy **a new coat**," the phrase "a new coat" refers
to a specific, though not yet identified, coat that the speaker intends to purchase.
* **Example of Non-Referring (Generic) Use:** In "**A coat** keeps you warm in winter," the same
phrase "a coat" does not refer to any particular coat. It is talking about the general class of coats. This
is called a generic interpretation.

Sometimes, a sentence can be ambiguous. A person might say, "I'm looking for **a taxi**." This could
mean they are looking for one specific taxi they called (referring), or it could mean they are looking
for any available taxi (non-referring, generic).

**Constant versus Variable Reference**

Another important difference is seen in how consistent the reference is across different uses. Some
expressions almost always point to the same thing.

* **Example 1:** "**The Sun**" always refers to the same star at the center of our solar system.
* **Example 2:** "**The Pacific Ocean**" always refers to the same body of water.

These are described as having **constant reference**.

Other expressions point to different things every time they are used. To know what they are pointing
to, you need to know the context: who is speaking, who they are speaking to, and when and where
they are speaking.

* **Example 1:** In "**I** am hungry," the word "I" refers to a different person in every
conversation.
* **Example 2:** In "Please put **that** down," the word "that" refers to whatever object the
speaker is looking at or pointing to.

Words like *I, you, he, this, that,* and *now* are said to have **variable reference**. They are called
"deictic" expressions or "indexicals."

In reality, most referring falls somewhere on a spectrum. For example, to know what "**the President
of the United States**" refers to, you need to know the time period, because the person in that office
changes.

**Referents and Extensions**

We can also make useful distinctions about the *things* that are being pointed to. We use the term
**referent** for the specific thing a speaker points to in a specific situation.

* **Example 1:** The referent of "**my mother**" is different for every person who says it.
* **Example 2:** The referent of "**this page**" is the page you are reading right now.

The term **extension** of an expression is the entire *set* or *collection* of all things that could
ever be the referent of that expression. It is the whole class of things the word can point to.

* **Example 1:** The extension of the word "**planet**" is the set of all planets in our solar system
and beyond.
* **Example 2:** The extension of the word "**shoe**" is the set of all shoes that exist, have
existed, or will exist.

As we learned before, the stable relationship between a word and its extension is called
**denotation**.

#### **4.2. How Names Work**

Names (like "Maria" or "Rome") seem like the simplest kind of referring expression. They are labels
for specific individuals, places, or organizations. It feels odd to ask for the "meaning" of a name
beyond its use to identify its referent.

Context is still crucial. Names are **definite**, meaning the speaker assumes the listener can identify
the intended referent.

* **Example:** If you say, "I saw **David** yesterday," you assume your listener knows which
David you are talking about.

But how do names function? Philosophers have proposed different theories.

One important theory is the **description theory**, associated with philosophers like Bertrand
Russell and John Searle. In this theory, a name is a shorthand for a bundle of knowledge or
descriptions about its referent.

* **Example:** The name "**Albert Einstein**" might be linked to descriptions like "the physicist
who developed the theory of relativity" or "the man with the famous E=mc² equation."
Understanding the name involves knowing these descriptions.

Another, very different theory is the **causal theory**, based on the work of Saul Kripke. This theory
says that names are passed down through a social chain. There is an original event where a name is
given to a person or place—a "baptism." People who were there start using the name. The name is
then passed from person to person, like a chain, even to people who have never met the referent and
know very little about them.

* **Example:** Millions of people use the name "**Confucius**." Most people know only that he
was an ancient Chinese philosopher. They use the name correctly because they are part of a long
historical chain of use that goes back to the original naming of the real person.

The major advantage of the causal theory is that it explains how we can use names correctly even
when we are ignorant. The description theory focuses on the knowledge in our heads, while the
causal theory focuses on our social connections. This debate is important because the theory chosen
for names can also be applied to other words, like words for natural categories (e.g., *tiger, gold*).

#### **4.3. How Nouns and Noun Phrases Work**

Nouns and noun phrases (NPs) are also used to refer. An indefinite NP (like "a dog") can introduce a
new referent into a conversation, while a definite NP (like "the dog") assumes the listener already
knows which one is being discussed.

* **Example:** "I heard **a strange noise**." (indefinite NP, introducing a new thing)
* **Example:** "I heard **the strange noise** again." (definite NP, referring back to the noise
already mentioned)

Definite noun phrases can also be **definite descriptions**, where the referent is whoever or
whatever uniquely fits the description.

* **Example:** "She is **the winner of the competition**." The referent is the one person who fits
that description.

Any theory of reference must also account for difficult cases where there is no real-world referent,
like the famous sentence, "**The present King of France** is bald." (There is no King of France.) Or
when we talk about fictional entities, like "**Sherlock Holmes**."

Noun phrases can refer to groups in two different ways:

* **Distributively:** This means the statement applies to the individual members of the group.
("**The students** in the room each have a textbook." - Each student has their own book.)
* **Collectively:** This means the statement applies to the group as a single unit. ("**The
students** in the room form a committee." - The group, as a whole, is a committee.)

Nouns can also denote things that aren't single, countable objects:
* **Substances:** "We need more **water**."
* **Actions:** "**Swimming** is great exercise."
* **Abstract Ideas:** "They value **honesty** above all."

One important grammatical distinction is between **mass nouns** and **count nouns**.
* **Count nouns** like "chair" can be counted. We say "a chair," "three chairs," "many chairs."
* **Mass nouns** like "furniture" usually cannot be counted directly. We don't say "∗a furniture"
or "∗three furnitures." We use measures: "a piece of furniture," "three items of furniture."

This distinction is not just about the physical world; it's about how we choose to conceptualize things
through language. This varies across languages. For example, "information" is a mass noun in English,
but in French, the equivalent word ("des informations") is often treated as a count noun.

Some noun phrases, like those with **quantifiers**, don't refer to a single individual but allow us to
talk about quantities of a group.
* **Example 1:** "**Every cookie** was eaten." (Refers to the entire set of cookies.)
* **Example 2:** "**Some cookies** were left." (Refers to an indefinite part of the set.)

### **5. Can Reference Be the Whole Story of Meaning?**

We have seen that reference is a powerful function of language. The simplest theory of meaning
would be to say that semantics *is* reference. In this view, to give the meaning of a word, you simply
point to what it denotes in the world. A simple version of this theory might claim:
* Names denote individuals.
* Common nouns denote sets of individuals.
* Verbs denote actions.
* Adjectives denote properties.

However, this simple theory faces several major problems that show it cannot be the whole story.

**Problem 1: Words Without Real-World Referents**

First, this theory would imply that many words are meaningless because they have no real-world
referent. As we saw, words like *very, not, if,* and *but* do not point to any object. If meaning were
only reference, these words would be meaningless, which is clearly false. They are crucial for
constructing meaningful sentences.

**Problem 2: Referring to Non-Existent Things**

Second, we constantly use words to talk about things that do not exist.

* **Example 1:** "The child has an **imaginary friend**." (Imaginary friends are not real.)
* **Example 2:** "**The perfect vacation** would be on a tropical island." (This refers to a
hypothetical, not actual, vacation.)
* **Example 3:** "**Voldemort** is the villain in the Harry Potter books." (A fictional character.)

If meaning required a real-world referent, then words like *imaginary friend* and *Voldemort* would
be meaningless. But the sentences containing them are perfectly understandable and meaningful.
Therefore, meaning must involve more than just reference.

**Problem 3: Different Expressions for the Same Referent**

A third problem is that we can use different expressions to talk about the exact same person or thing,
and these expressions clearly have different meanings.

* **Example:** "**Samuel Clemens** wrote many books." and "**Mark Twain** wrote many
books." Both names refer to the same person, but they feel different. "Mark Twain" is his pen name,
and using it evokes his identity as a author, while "Samuel Clemens" is his legal name.

You might refer to the same person as "**my biology teacher**," "**Mrs. Jones**," or "**the
woman who lives next door**." These all have the same reference but different meanings. It is even
possible to understand two expressions without knowing they point to the same thing.

The philosopher Gottlob Frege gave a classic example: a person might know "**the Morning Star**"
as the bright star seen in the morning and "**the Evening Star**" as the bright star seen in the
evening, without knowing they are both the planet Venus. For this person, the sentence "The
Morning Star is the Evening Star" is informative news, not a boring, obvious statement like "Venus is
Venus."

If we can understand words for non-existent things, use different meaningful expressions for the
same thing, and use two expressions without knowing they corefer, then meaning cannot be identical
to reference. There must be an additional component.

### **6. Introducing "Sense": The Mental Dimension of Meaning**

Following Frege's argument, many linguists split meaning into two parts: **sense** (Frege's *Sinn*)
and **reference** (Frege's *Bedeutung*). In this division, **sense is the mode of presentation**—it
is the way we think about the referent. It is the conceptual knowledge that allows us to identify the
reference.

**Sense is primary because it enables reference.** It is because we understand the *sense* of the
phrase "the first person to walk on the moon" that we can use it to identify Neil Armstrong. Another
expression with the same reference, like "Neil Armstrong," has a different *sense*.

If we accept this argument, our theory of meaning becomes more complex. The meaning of an
expression involves both its sense (the conceptual information) and its reference (the thing in the
world).

### **7. What is "Sense"? The Idea of Mental Representations**

For the rest of the chapter, we explore the idea that "sense" is a level of **mental representation**
that stands between words and the world. In this view, a word is connected to a concept in our
minds, and that concept gives the word its ability to refer.

This neatly solves the problem of non-existent things. We can have mental representations for
unicorns, square circles, and fictional worlds, even though they have no real-world reference.
But this raises a new, difficult question: what is the form of these mental representations?

#### **7.1. Are Mental Representations Like Pictures?**

One simple, old idea is that these mental representations are **images**. The word "apple" calls to
mind a picture of an apple, and this mental picture resembles real apples.

* **Example:** The word "your car" might evoke a mental image of your specific car.
* **Example:** The word "dragon" might bring to mind a picture of a large, fire-breathing lizard,
even though dragons don't exist.

This might work for some concrete words, but it fails for many others. Different people have different
mental images for a word like "house" or "dog," based on their personal experience. The theory
completely breaks down for abstract words.

* **Example:** What is the mental image for "**triangle**"? One person might picture an
equilateral triangle, another a right-angled one. No single image can capture all triangles. What would
the mental image for "**justice**," "**love**," or "**democracy**" be?

So, while mental imagery might be part of our experience, it cannot be the full story of "sense."

#### **7.2. Concepts as the Basis for Sense**

The most common solution is to propose that the sense of a word is a **concept**. A concept is a
more abstract unit of mental knowledge. It could include defining features, typical characteristics, and
relationships to other concepts. A concept for DOG would contain the knowledge that allows us to
identify dogs, not just a picture of one.

This has clear advantages. We can have concepts for abstract ideas like DEMOCRACY. We can also
create definitions (e.g., "a triangle is a three-sided polygon").

For linguists, this approach has another benefit: they can collaborate with psychologists who study
concepts. Some concepts might be simple (like WATER), while others are complex and culture-specific
(like JUSTICE).

However, a major problem is that psychologists themselves do not agree on what concepts are or
how they work. Without a clear psychological model, saying "the sense of *dog* is the concept DOG"
can feel like replacing one vague term ("meaning") with another ("concept").

This dilemma divides linguists. Some, skeptical of psychology, prefer to stick to a denotational
approach or to model sense using formal logic, avoiding the mind altogether. Others, who favor a
**representational approach**, actively build models of conceptual structure to serve as the
foundation for semantics.

### **8. Exploring Concepts: Theories from Psychology**

Let's follow the representational path and look at some basic theories about what concepts are and
how they work. We will focus on concepts that are **lexicalized**—that is, they have their own
single word (like "key") rather than needing a phrase (like "a device for opening locks").

When studying how children acquire concepts, it's important to remember that a child's concepts are
not the same as an adult's. Children might **underextend** a word (using "ball" only for their
favorite red ball) or **overextend** it (using "daddy" for all men). Their concepts are shaped by what
is most salient in their immediate environment.

#### **8.1. The Definitional Theory: Necessary and Sufficient Conditions**

One traditional theory tries to define concepts using a checklist of **necessary and sufficient
conditions**. In this view, a concept is a set of features. If something has all the features on the list, it
is a member of the category. If it lacks any one feature, it is not.

* **Example for BACHELOR (in a simple, dictionary view):** For x to be a bachelor, the following
must be true:
* x is human. (necessary)
* x is an adult. (necessary)
* x is male. (necessary)
* x has never been married. (necessary)

The features are **necessary** (all are required) and, if the list is perfect, **sufficient** (the list
alone is enough to define a bachelor).

The major problem with this approach is that it is incredibly difficult to find a set of defining features
that all speakers agree on, even for common words. Let's try "game."

* **Potential Conditions:** has rules, is done for fun, has winners and losers.
* **Problems:** Does a child randomly throwing a ball against a wall have rules? Is a professional
poker player who is stressed and losing money having "fun"? Does a game of solitaire have a "loser"?
It's very hard to pin down the necessary and sufficient conditions for being a game.

If we can't agree on a definition, how do we all manage to use the word "game" successfully?

Another powerful argument comes from philosopher Hilary Putnam, who highlighted the role of
**ignorance**. People often use words correctly while knowing very little about the defining features
of the referent.

* **Example:** Many people cannot tell an **elm** tree from a **beech** tree. Yet, they can
understand and truthfully say, "There is a large **elm** tree in the park." They use the word correctly
without knowing the necessary and sufficient conditions for being an elm.

Putnam suggested we rely on a "division of linguistic labor." We defer to experts in our community
(like botanists) who *do* know the precise definitions. This is very similar to the **causal theory**
for names. Philosophers like Putnam and Kripke have argued that this causal theory can be extended
to **natural kind terms**—words for natural categories like *water, tiger,* and *gold*. We use the
word "gold" correctly not because we know its atomic number, but because we are part of a social
chain that links back to the original samples of the metal.

#### **8.2. Prototype Theory: Concepts with a Family Resemblance Structure**

Because of the problems with definitions, Eleanor Rosch and her colleagues developed **prototype
theory**. This model says that concepts are not strict checklists but have an internal structure. There
are central, "best" examples of a concept (the **prototypes**) and less typical, peripheral examples.
Category membership is a matter of degree, not a simple yes/no.

* **Example for BIRD:** A **robin** or **sparrow** is a very typical (prototypical) bird. A

**penguin**, **ostrich**, or **chicken** is a less typical member.
* **Example for FURNITURE:** A **chair** or **table** is very typical. A **lamp** or **rug** is
less typical.

Rosch's experiments showed that people are faster to agree that a robin is a bird than an ostrich is a
bird. Typical members are learned first by children and come to mind most easily. She also found that
the boundaries between categories can be fuzzy.
This approach explains the "whale" problem. A person might be unsure if a whale is a mammal or a
fish because whales are not *typical* mammals (they live in water), but they share many features
with *typical* fish.

Within psychology, there are different interpretations of what a prototype is. Some think it is an
**abstraction**—a set of averaged features that describes a typical bird (e.g., *size: small, can: fly,
has: feathers*). Others think we use **exemplars**—we store memories of many actual, typical birds
we've seen (robins, pigeons) and judge new things by their similarity to these stored examples.

Linguists like Charles Fillmore and George Lakoff have suggested that these typicality effects come
from our "folk theories" of the world, which they call **frames** (Fillmore) or **idealized Cognitive
Models (ICMs)** (Lakoff). These are not scientific theories, but culturally shared packages of
knowledge.

* **Example for BACHELOR (revisited):** The dictionary definition is "an unmarried man," but our
ICM of marriage includes ideas about being of marriageable age, being in a society where marriage is
the norm, and being eligible. This is why it feels strange to call the Pope, a five-year-old boy, or a man
in a long-term committed but unmarried relationship a "bachelor." They fit the definition but violate
our idealized model. Here, word meaning blends with cultural knowledge.

#### **8.3. How Concepts Are Related to Each Other**

Conceptual knowledge is not a set of isolated units. Concepts are linked to other concepts in a vast
network. Your knowledge of words is also a network, which we will explore in the next chapter on
lexical relations.

* **Example:** If you learn that a "wombat" is a type of **marsupial**, your concept of WOMBAT
automatically inherits general properties from your concept of MARSUPIAL (e.g., has a pouch, gives
birth to underdeveloped young).

These relationships are the basis for **conceptual hierarchies**. In a hierarchy for animals, the
concept BIRD has specific features like *has wings* and *can fly*. It doesn't need to list *eats* or
*breathes* because it inherits those properties from the higher-level concepts ANIMAL and LIVING
THING. This makes our conceptual system highly efficient.

Prototype researchers have also found that conceptual hierarchies have three basic levels that are
psychologically distinct:
* **Superordinate Level:** Very general (e.g., FURNITURE). Has few common features.
* **Basic Level:** The most useful, common level (e.g., CHAIR, TABLE). Has many common features,
is used most in speech, and is the level at which we are fastest to recognize objects.
* **Subordinate Level:** Very specific (e.g., ARMCHAIR, KITCHEN TABLE). Has even more specific
features.

The **basic level** is cognitively primary. It is the first level named by children and the most natural
level for everyday conversation.

#### **8.4. How We Acquire Concepts**

Our second basic question about concepts was: how do we get them? A seemingly straightforward
theory is **ostensive definition**—pointing to examples and naming them. A parent points to a dog
and says "dog!" and the child learns the concept.

The philosopher W.V.O. Quine famously showed a deep problem with this. Ostensive learning is
ambiguous. His famous thought experiment involves a linguist and a native speaker. A rabbit runs by,
and the native says "Gavagai!" The linguist does not know if "Gavagai" means "rabbit," "Look!," "It's
running," "animal," "dinner," or even "undetached rabbit parts." To understand that a word refers to
a whole, enduring object, you need some prior conceptual and linguistic understanding. This leads to
the puzzle of how our very first concepts and words are acquired.

### **9. The Big Debate: Do Words Shape Our Thoughts?**

We have assumed a link between words and concepts. But what is the relationship between these
language-linked concepts and our general ability to think? There are two famous, opposing views on
this.

#### **9.1. Linguistic Relativity**

The theory of **linguistic relativity** is associated with Edward Sapir and Benjamin Lee Whorf. It is
the idea that the structure of a language influences how its speakers think and perceive the world.
This idea is appealing because languages categorize the world differently.

* **Example 1:** Color categories. Some languages have a single word for blue and green.
* **Example 2:** Spatial relations. Some languages use absolute directions (north, south) instead of
relative ones (left, right) for everyday descriptions.

This observation—that language reflects culture—was important. But Sapir and Whorf took it further,
suggesting that language doesn't just reflect thought; it *determines* it.

* **Sapir wrote:** "Human beings are very much at the mercy of the particular language which has
become the medium of expression for their society... The 'real world' is to a large extent
unconsciously built up on the language habits of the group."
* **Whorf wrote more strongly:** "We cut nature up, organize it into concepts, and ascribe
significances as we do, largely because we are parties to an agreement... codified in the patterns of
our language."

If this strong view (sometimes called "linguistic determinism") is correct, then speakers of different
languages literally inhabit different conceptual worlds. This would pose a major challenge for creating
a universal semantic theory.

#### **9.2. The Language of Thought Hypothesis**

The idea of linguistic determinism is largely rejected by modern cognitive scientists. They argue that it
confuses thought with language. They offer two main counter-arguments.

**Argument 1: Evidence of Thinking Without Language**

* **Example 1:** Pre-linguistic babies and many animals can solve problems, make plans, and have
memories, demonstrating complex thought without human language.
* **Example 2:** People often have sudden insights or "think" in images or feelings, not words.
* **Example 3:** Experiments in mental imagery show people can rotate objects in their mind in
ways that don't seem language-based.

This evidence supports the idea that we think in a separate, non-linguistic system often called the
**Language of Thought** or **Mentalese**. Speaking, then, is the act of translating our thoughts
from Mentalese into a spoken language.

**Argument 2: Language Underspecifies Meaning**

What we say is almost always a stripped-down version of what we mean. Listeners use context and
world knowledge to infer the full message.

* **Example:** The statement "It's dark in here" could be a simple observation, a request for
someone to turn on the light, or a complaint about a gloomy room. The words alone do not specify
the speaker's intention.
This fits perfectly with the view that a rich, pre-linguistic thought is being compressed into a linear
stream of language.

A natural conclusion from this view is that Mentalese is universal. All humans share the same
fundamental cognitive machinery, and different languages are just different ways of expressing the
same basic thoughts. This is the direct opposite of linguistic relativity.

### **10. The Deepest Layer: Thought and Reality**

This leads to philosophical questions that underpin all of semantics: what is the relationship between
our thoughts and reality itself? These are questions of **ontology** (what exists?) and
**epistemology** (how do we know?).

* **Realism/Objectivism:** The world exists independently of our minds, and we can perceive it
more or less correctly.
* **Idealism/Mental Constructivism:** Our perception of reality is largely constructed by our minds
and our conceptual systems.

Different semantic theories are often built on different assumptions about these fundamental issues.
For a working linguist, these can feel like unanswerable philosophical problems that are beyond their
scope. One practical response is **linguistic solipsism**—to focus only on the language system itself,
studying how words relate to each other, and setting aside the big questions of mind and world. This
inward focus leads to the study of sense relations, which is the topic of the next chapter.

### **11. Summary: Pulling All the Ideas Together**

In this chapter, we have seen that while reference is a crucial function of language, it is insufficient as
a complete theory of meaning. Our semantic knowledge involves both **reference** (the word-
world connection) and **sense** (the conceptual knowledge that enables reference).

We identified two major approaches to building a semantic theory:

1. **Denotational/Referential Approach:** Grounds meaning in the external world.
2. **Representational Approach:** Grounds meaning in internal mental structures (concepts).

Each approach has its challenges. Denotational theories must explain reference to non-existent
things. Representational theories require a plausible model of conceptual structure.

Influenced by philosophy and psychology, linguists have generally taken one of three paths forward:
1. **Focus on Language-Internal Relations:** Avoid the mind-world problem and concentrate on
describing the network of meaning relations within a language (e.g., synonymy, antonymy).
2. **Strengthen Denotational Theory:** Develop a formal, logical denotational theory powerful
enough to handle all the complexities of reference.
3. **Embrace Conceptual Structure:** Commit to a representational approach and build detailed
models of the concepts that underlie word meanings.

These three paths map onto different subfields of semantics that we will encounter later in the book.
The first is the domain of traditional **lexical semantics** (Chapter 3). The second is the goal of
**formal semantics** (Chapter 10). The third is the aim of **conceptual semantics** (Chapter 9) and
**cognitive semantics** (Chapter 11).