SemPrag 1: Intro, Lexical Meaning, Theta-Roles

Question 1

What does semantics study?

Answer 1

Meaning: mental representations, as opposed to form.

Question 2

Difference between sentence & utterance?

Answer 2

A sentence is an abstract linguistic form with certain invariant properties: constituent structure & compositionality. On the other hand, an utterance is a situated (real) instance of language use, which may or may not take the form of a sentence.

Question 3

Difference between sentence meaning and utterance meaning?

Answer 3

Sentence meaning is meaning derived from the linguistic forms without any context. Utterance meaning studies Speaker Meaning: what the speaker is actually conveying in context by using the sentence/phrase.

Question 4

What are context-dependent expressions?

Answer 4

Expressions have invariable meaning but also refer to the speech situation: ‘I’ (who?), ‘now’ (when?), ‘local’ (to where?), ‘tall’ (compared to what?), ‘arrive’ (where?). For words like ‘after’ & ‘local’, the relation is specified but the extent (how long after / how close) is not.

Question 5

Based on the existence of context-dependent expressions, what middle ground has been proposed between sentence meaning and speaker meaning?

Answer 5

‘What is said’: the linguistic meaning incorporating the context for context dependent expressions, however not including speaker intention. Grice 1978: lexical meaning and sentence structure AFTER disambiguation and reference assignment.

Question 6

How to test semantic theories?

Answer 6

Using native speakers’ judgements about truth & falsity.

Question 7

How do semanticists define sentence meaning?

Answer 7

As the ‘truth conditions’ of the linguistic expression. These are the conditions which would have to hold for the expression to be true. This is very different to the sentence’s truth VALUE (whether or not it is actually true).

Question 8

What is a proposition?

Answer 8

The aspects of a sentence that are relevant when we evaluate their truth-conditions (at ‘what is said’ level). (essentially truth conditional meaning).

A function mapping possible worlds to truth values.

Question 9

What is entailment?

Answer 9

A special kind of implication which is not context-dependent, defined as: A semantic relation by which the truth of S1 guarantees the truth of S2.

Question 10

What is contradiction?

Answer 10

The truth of one sentence guarantees the falsity of the other.

Question 11

What is the difference between pragmatic inferences and entailment?

Answer 11

Entailment is context-independent, whereas pragmatic inferences are context-dependent.

Question 12

Grice: Natural vs Non-natural meaning?

Answer 12

In the 1948 article “Meaning”, Grice describes “natural meaning” using the example of “Those spots mean (meant) measles.”

And describes “non-natural meaning” using the example of “John means that he’ll be late” or “‘Schnee’ means ‘snow’”.

Grice does not define these two senses of the verb ‘to mean’, and does not offer an explicit theory that separates the ideas they’re used to express. Instead, he relies on five differences in ordinary language usage to show that we use the word in (at least) two different ways.

Question 13

Why are some utterances which seem not to be complete sentences considered to be anyway?

Answer 13

Ellipsis: ‘what notebook did she choose?’ ‘the one with the yellow cover’. Is ellipsis of ‘she chose’ rather than a phrasal utterance such as ‘ouch’.

Question 14

How to check for contradiction?

Answer 14

Look for superordinate and subordinate nouns, antonyms & negatives.

Question 15

What is ambiguity vs anomaly? (Larson, R. & G. Segal 1995)

Answer 15

Ambiguity is where a sentence has multiple meanings. (Lexical & syntactic ambiguity). Whereas anomaly is where a sentence has an aberrant (departing from accepted standard), e.g ‘colourless green ideas sleep furiously’.

Question 16

Words vs Lexemes? (Lyons, J. 1995).

Answer 16

Words: composite units with both form & meaning.
Lexemes: a unit of the lexicon.
Not all words are lexemes and not all lexemes are words.
Lexemes can be transformed by rules into lexically composite expressions (e.g polite > politeness).
A phrasal lexeme is not transformed by rules but is rather simply two lexemes combining to form a lexically distinct lexical unit (lexeme) e.g pass muster or red herring.
Lexical meaning is distinct from grammatical meaning so girl vs girls is one lexeme distinguished by grammatical number. (language dependent).

Question 17

Difference between homonymy and polysemy? (Lyons, J. 1995).

Answer 17

Homonymy is where two lexemes with different meanings have the same form.
Polysemy is where one lexeme has multiple etymologically related senses (sometimes due to metaphorical extension). However should senses have distinct lexical entries? And what constitutes a distinct sense (e.g the various meanings of fast).
(1) sense enumeration lexicon: create a distinct entry for each kind of ‘fast’, ‘red’, ‘kiss’, ‘window’, ‘Dickens’ etc.
(2) generative lexicon: specify some rules for deriving the distinct senses from a ‘basic’ sense + contextual information (Pustejovsky, 1995)

Question 18

C.S. Peirce Types vs Tokens?

Answer 18

Types are distinct words, whereas tokens are instances of word use (whether distinct or not).

Question 19

Synonymy? (Lyons, J. 1995)

Answer 19

Refers to lexemes & lexically complex expressions.
For ABSOLUTE synonymy:
i) all meanings are identical
ii) synonymous in all contexts
iii) semantically equivalent on descriptive and non-descriptive dimensions of meaning. (non-descriptive is expressive meaning e.g the difference between stingy and economical).

Question 20

What is the goal of semantic theory?

Answer 20

to explain the external and internal significance of language compositionality.

External: information bearing potential in the real world.
Internal: inference from necessity (entailment) or abduction (likelihood).
Compositionality: meaning of a complex expression is derived from the meaning of its parts and the way they are put together. (we believe in compositionality due to productivity, systematicity and finiteness of means).

Question 21

What are the approaches to semantic theory?

Answer 21

Referential Meaning or Mentalistic Approaches (classical theory, prototype theory, theory-theory, causal theory).

Question 22

Gottlob Frege: The Referential Theory?

Answer 22

The meaning of a word is what it denotes (its extension) (direct relation with the world).
Terms are proper names to denote individuals.
One-place predicates denote sets of individuals who have the relevant property (table, linguist).
According to this theory there is no intermediate level between an expression and its meaning

Question 23

Problems with Referential Theory?

Answer 23

There should be a sensitivity to the difference between proper names and noun phrases.
Doesn’t account for changing meaning e.g ‘the U.S president’.
Also doesn’t account for words with no real world reference e.g ‘Santa Claus’.
Sense & Reference

Question 24

Sense vs Reference

Answer 24

If sense is considered meaning (the mental representation which corresponds to the linguistic form) and reference is considered the real world object to which the linguistic form refers.

The Morning Star vs The Evening Star
Speakers had no knowledge that they were the same star: sense was different.
However Referent was the same!

Laurence & Margolis 1999 note that senses are the indirect referents of expressions in intensional contexts so take on bizarre properties.
Can freely substitute ‘the morning / evening star is bright’ without affecting the truth value, however the same substitutions are not possible for ‘Sue believes the evening star is bright’.

Question 25

Within the 4 theories of concepts, how can they be grouped?

Answer 25

Classical, prototype and theory theory all believe in an inferential role (for classical theory, only certain inferences are informative), whereas causal theory believes in some causal/historical relation between expression and extension.

Question 26

Necessary and Sufficient Conditions?

Answer 26

A is a necessary condition for B when it cannot be the case that B is true and A is false.
A is a sufficient condition for B when it cannot be the case that A is true and B is false.

Question 27

Classical Theory (Laurence & Margolis 1999)

Answer 27

Most concepts (especially lexical concepts) are structured mental representations that encode a set of necessary and sufficient conditions for their application, if possible in sensory or perceptual terms. (Although more recent theorists have discarded with that last bit, e.g brother). 
e.g Bachelor = human AND male AND NOT married AND adult. (However it is hard to choose the primitive features - isn't this choice arbitrary?) 
This holds for verbs too: 'John almost killed Bill' did John almost cause bill to die? Or did John cause Bill to almost die?

Question 28

Problems with Classical theory:

Answer 28

The Retreat from definitions:
Arbitrary feature choice - like how would we acquire the primitive feature ‘male’ ?
Failure to give definition from any but a few obvious words cannot capture nuances of meaning (is the pope a bachelor?) (is carpet furniture?) L&M call this conceptual fuzziness.
Plato’s Problem: what are the composite concepts of knowledge, justice, truth & beauty. Not sensory for sure…

Ignorance & Error: (based on the ideas of Kripke and Putnam) do people who think that a tomato is a vegetable not know what a tomato is?

Psychological reality: Processing time for convinced (causexbelieveyz) is same for believed (believexy) which should be faster. (Kintsch 1974)

Cannot explain typicality effects. (well actually some models can. The idea that all members of a set all share every member of a group of concepts could not, but the idea (Smith & Medin 1981) that categorisation is an exhaustive feature matching process would explain it because more erroneous features would cause more errors in the matching process.

Question 29

Prototype Theory

Rosch & Mervis 1975

Answer 29

Most definitions are really approximate or partial.
‘Most concepts are complex mental representations whose structure encodes a statistical analysis of the properties their members tend to have.’
e.g BIRD: flies, sings, nests in trees, lays eggs.
A member of the category need only exhibit some of the features rather than necessarily all of them as in Classical Theory (which are weighted differently based on their frequency). Meaning is assigned using a category features list and typicality rating.

Internal Significance: Inferential relations exist among concepts by virtue of their prototypical features.
External Significance: We recognise something as e.g a bachelor by checking that it satisfies many of the typical category features for being a bachelor.

Jaszczolt says the core prototype might contain an abstract set of characteristic features, a result of our cultural experience (think bachelor) a real world example.

Question 30

Problems with Prototype Theory?

Answer 30

Compositionality: Pet Fish isn’t in anyway a combination of the prototypes.

Some concepts have no prototype: e.g 3

Cases where typicality effects affect even well-defined concepts (how is a prototype the explanation here): e.g prime number.

Ignorance & error.

Perhaps a dual model is required. It’s suggested that prototypes used for quick judgements and classical for more considered, because prototypes often can’t determine reference (need essential info?) so in a dual model would be relegated to identification where necessary conditions would hold for reference determination.

For more see L&M 1999 pgs 32-43.

Question 31

Compositionality isn’t just syntax. What are the different types of adjective and how does compositionality affect them? (Partee 1995)

Answer 31

Intersective Adjective:
‘green ball’ is at the intersection of things that are green and balls.
Subsective Adjective:
‘good pianist’ is a subset of the group ‘pianists’.
Privative Adjectives: deprive the noun of its properties.
Intersective adjectives mean the same thing when separated from their noun, however subsective adjectives would take on a more general meaning and privative adjectives would create a contradiction.

Question 32

What are basic concepts?

Answer 32

Minimal concepts to arrive decompositionally at the meaning of a lexical item.

Question 33

What is a 1 place, 2 place, 3 place predicate.

Answer 33

Has different number of arguments. Father (x, y) is two place.

Question 34

What is the backwards E?

Answer 34

The existation quantifier: There exists…

Question 35

What are the two meanings of ‘Pete opened the window again’.

Answer 35

Repetitive meaning, (widescope applies to whole action)

Restitutive meaning, (narrowscope just refers to the window becoming open again).

Question 36

Two basic models for what concepts are:
(Laurence & Margolis 1999)
They take it as given that concepts are mental rather than abstract.

Answer 36

Containment model: a concept is a structured complex of other concepts only if it has those concepts as proper parts. So ‘tony plays trumpet’ is made up of concepts ‘tony’ ‘plays’ ‘trumpet’.
The Inferential Model: one concept is a structured complex of other concepts if it stands in privileged relation to these other concepts by way of inferential disposition. So Red can implicate colour without tokening it. (red colour).

Question 37

Can senses be mental entities?

Answer 37

Frege said no because senses are objective and mental entities are subjective, however Laurence & Margolis 1999 posit that people’s subjective mental entities can share senses.

Question 38

What can Classical Theory account for?

L&M1999 calling it very explanatory

Answer 38

Compelling account for categorisation, epistemic justification (verifying something’s category), concept acquisition (assemble the theory)

Kant: analytic inferences are drawn from the meaning of the word, whereas synthetic ones are drawn from related concepts.
Analytic: smith is an unmarried man. so smith is a man.
Synthetic: smith is a bodybuilder. so smith is a man.
Analytic! smith is a bachelor, so smith is a man.
(seems to be contained in the meaning even without word man, you don’t have to think about likelihood). However some such as Quine have argued against analyticity.

Question 39

Problems with Classical Theory which are solved by Prototype Theory?

Answer 39

• lack of clear definitions is not a problem because prototype posits no definitional structure
• can wholeheartedly embrace the Quinean critique of analyticity as it does not posit necessary conditions.
• concept acquisition is slightly better due to analysing features which TEND to co-occur but still struggles with the fact that lots of concepts aren’t sensory.
• categorisation accounts perfectly for typicality effects

Question 40

What is the Theory-Theory of concepts? L&M1999

Answer 40

Concepts are representations whose structure consists in their relations to other concepts as specified by a mental theory. (embedding theories).

This is useful in explaining the tendency not to categorise based on observable properties, (psychological essentialism - which even happens in children pg 46) e.g a cat painted like a skunk is not categorised as a skunk due to biological makeup.

Question 41

Problems for Theory-Theory? L&M1999

Answer 41

Ignorance & Error - It is possible to have a concept in spite of its being tied up with a deficient (very vague essence place holders if using genetic makeup) or erroneous mental theory (small pox is caused by god?).
Stability - The content of a concept can’t remain invariant across changes in mental theory.
‘Mysteries of Science’ - how do cognitive theories change and develop.

Question 42

Neoclassical Theory? L&M1999

Answer 42

Believes that concepts encode only ‘partial definitions’ and do away with the idea that minimal concepts are perceptual/sensory.

Problems:
Partial definitions are not useful for reference determination
Ignorance and Error
Another Problem I do not understand 59.

Question 43

Atomism? L&M1999

Answer 43

Lexical concepts are primitive they have no structure.

Problems:
The position that many concepts are innate is bizarre e.g xylophone, carburetor.
Can’t explain psychological processes such as categorisation.
No explanation for analyticity (though…)
Compositionality.
5th i dont get.

Question 44

Saeed: Referring & Non-referring expressions

Answer 44

some words don’t refer but contribute meaning (all, so, very, maybe).
(Intrinsically non-referring).

Nouns are potentially referring expressions.

Nouns refer in sentences like ‘I performed an operation this morning.’ (1 specific operation).
Rather than ‘An operation can be emptionally challenging’. (no specific operation referred to).

Question 45

Saeed: Constant vs Variable Reference

Answer 45

The Eiffel Tower vs You (deictic) but also things like the US President.

Question 46

Saeed: What is the extension of an expression?

Answer 46

The set of things which could possibly be the referent of that expression.

Question 47

Saeed: what is denotation?

Answer 47

The relationship between an expression and its extension.

Question 48

Saeed: description theory and causal theory. (Names)

Answer 48

Description theory (e.g Russel 1967): understanding a name and identifying a referent are dependent on associating the name with the right description.

Causal Theory: (Devitt & Sterelny, 1987). Names come from a history of usage back to an original usage or ‘grounding’, social knowledge.

Question 49

Saeed: Lexicalised Concepts

Answer 49

A concept to which a single word is attached. He posits that some words become lexicalised due to utility (frequency of usage).

Question 50

Saeed: conceptual hierachies

Answer 50

If you know that a specific lexicalised concept belongs to the extension of a broader lexicalised concept, then the first one inherits meaning form the broader one.
Models have been proposed for conceptual networks.

A lower (more specific) concept is a hyponym of the more general one.

Question 51

Saeed: Meronymy

Answer 51

Describes the part-whole relationship between lexical items.

Question 52

Antonymy:

Answer 52

Either binary or gradable ‘opposite’ pairs.

Question 53

Hyperonymy:

Answer 53

The opposite of Hyponymy.

Question 54

How to distinguish ambiguity and polysemy?

In ‘Ambiguity Tests and How to Fail them’ by Sadock & Zwicky 1975, they refer to what I’m calling ‘polysemy’ as lack of specification. As the issue arises not from the existence of many senses but the fact that none are picked out.

Ambiguity involves a difference in semantic structure not a difference in understanding (potentially context-dependent).

Answer 54

Ellipsis and Coordination test.

John drew a gun and Bill drew a gun.
John drew a gun and Bill did too.
John and Bill each drew a gun.
For ambiguous verb ‘draw’ it cannot have a crossed reading in the ‘did too’
The lovers kissed and the politicians kissed.
The lovers kissed and the politicians did too.
The lovers kissed and so did the politicians.
For different senses of kissing (romantic vs greeting) you can still distinguish with the ‘did too’.

Question 55

Some ideas underlying theta-roles.

Answer 55

There is no one-to-one mapping between syntactic position of an NP and the role it performs. (although specific VPs determine specific patterns, it is not random). The role it performs is ‘participant/semantic/thematic > theta roles’.

Question 56

Problems with theta-roles.

Answer 56

No conclusive definition of number or specific theta roles.

What are they?! Theta-roles are not primitives but a cluster of prototypical properties. (Dowty 1991)

Sometimes hard to attribute to a given NP. Chomsky 1988 gives the ‘Theta-Criterion’ which states that there must be a one-to-one correspondence between NPs and theta-roles. Alternatively Jackendoff 1990 (and others) suggest that 1 NP could fulfil multiple roles. He elaborates into the ‘thematic tier’ describing spacial relations and the ‘action tier’ which describes ACTOR-PATIENT type relations, the idea being that 1 NP could fulfil a thematic role on each of these tiers.

Question 57

Two hypotheses for which role will surface as the syntactic subject (NOT the order of all roles in the sentence).

Answer 57

Dowty 1991:578
AGENT > INSTRUMENT/EXPERIENCER > PATIENT > SOURCE/GOAL

evidence: ‘a torpedo sinks a ship’ instrument before patient.

Givon (1984)
AGENT > RECIPIENT > PATIENT > INSTRUMENT > LOCATION

Question 58

Participant vs Non-Participant Roles

Answer 58

Participants are obligatorily specified whereas non-participants are optional arguments (as distinct from adjuncts).

Question 59

Autonomy of Syntax vs Functionalist Approaches

Answer 59

Chomsky’s principle where syntax is independent of the semantic properties of the words used.

Functionalist approaches - use of particular grammatical forms is strongly even deterministically linked to the presence of particular semantic or pragmatic functions in discourse - Tomlin 1990

Question 60

How are theta-roles thought to be incorporated in the lexicon

Answer 60

lexicon - phon, morph, lexical meaning, syntactically relevant semantic meaning in the form of a theta-grid (AGENT, PATIENT), underscore indicating syntactic subject.

Question 61

Let’s clear it up: Polysemy vs Indexicality

Answer 61

indexicals have a parameter which is filled by an object determined in the context without requiring any further sense construction. ‘I, now, today’

Other context-dependent (polysemous) words would require actual sense construction. ‘fast, red, tall’.

There are many cases where this distinction is blurred. Some items such as ‘here’ can be used without sense construction but sometimes the sense construction is required.

Question 62

Roberto’s list of theta roles:

Which seems to be Saeed’s !!!

Answer 62

Agent: perpetrates action intentionally (prototypical features - volition, sentience, movement).
Patient: undergoes the effect of some action (often change of state)
Theme: is moved by an action
Experiencer: undergoes a feeling (Saeed: is aware of the action or state described by the predicate but which is not in control of it can be sensory not just emotional feelings).
Stimulus: causes a feeling (psychological effect)
Beneficiary: a recipient (of some object) / the entity for whose benefit the action was performed (I filled in the form for my brother).
Goal: The entity towards which something moves
Source: the entity from which something moves, either literally or metaphorically
Location: place of occurrence
Instrument: the means by which an action is performed or something comes about.
ACTOR: Some writers (e.g Foley and Van Valin 1984, Jackendoff 1990) have suggested that AGENT is one type of a broader thematic role ACTOR: performs, effects, instigates or controls the action. So every agent is an actor but not the other way around.
FORCE: sometimes force is used instead of instrument for an inanimate entity which causes something e.g the wind.
RECIPIENT: sometimes identified (e.g Andrews, 1985) as a type of goal in actions describing change of possession.

Question 63

Roberto’s essay on whether syntactic position is a good predictor of semantic role.

Answer 63

1) there is a tendency for S=agent and O=patient
2) this is only a tendency. Some verbs such as ‘die’ take the patient as the subject. Also where the agent role is omitted e.g ‘the ice broke’ the patient takes the subject.
3) Also passive voice totally reverses this pattern for verbs which DO follow the tendency. (Givon 1990 thinks this is to foreground patient and background the agent, Kuno 1987 thinks it’s to show greater empathy with patient, fun example about Mary slapping on 168 of Saeed. There are sooo many other ways in which voice works in diff languages).
4) Psychological verbs which have different theta-roles have varying patterns depending on the verb.

Question 64

Justification for the positing of theta-roles.

Answer 64

The claim that in some languages theta-roles play a role in the morphology of verbal agreement. (Mithun 1991) gives examples of the pronominal verbal prefixes in Lakhota. wa-marks an agent argument and pa- a patient.

Question 65

Zwicky and Sadock 1975

Polar opposites and Privative opposites

Tests for ambiguity (in addition to those seen above - ellipsis and coordination)

Answer 65

Polar: one has feature F+ the other F-, otherwise identical.

Privative: one has feature F specified whereas other does not.

Generally these tests can detect ambiguity, but if they fail they can only SUGGEST lack of specification.

Contradiction can be used to detect ambiguity in privative opposites. For the more general one you can assert the general understanding and supply material that implies denial of it without problem, which you can’t for more specific. Also inconsistency under substitution (w/ synonyms or whatever if we believe in them). Whereas consistency under substitution would imply lack of sense specification. The idea in these cases being that a ‘lexical implication rule’ meant that the existence of one word implied the existence of the other. (That was a brilliant idea / that was a great idea).
Two distinct syntactic surface structures constitutes ambiguity.
Adding extra material (often known as co-occurrence restrictions) is not a good way to test for ambiguity, as adding stuff could be specifying sense or disambiguating by selecting one underlying structure. There are occasional circumstances where it works: special distribution. If the ability of each meaning of a phrase to occur in certain places (e.g after certain adverbs) is slightly different it’s ambiguous.
If you believe in transformational grammar, transformations (well established ones like wh) can disambiguate ambiguous cases but they will never specify sense.