Brainscape's Knowledge GenomeTM

Browse over 1 million classes created by top students, professors, publishers, and experts.


See full index

C82COG - Language > Word meaning > Flashcards

Word meaning Flashcards

1
Q

What are the main questions we aim to answer about word meaning?

A

How it is represented and processed.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Where is information about words stored?

A

In the mental lexicon and semantic memory, which are distinct and linked in a two-way path.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What information is stored in the mental lexicon?

A

Information that enables us to recognise and produce words - orthography and phonology. Includes spelling, syntactic information (e.g. nouns, verbs) and pronunciation.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What information is stored in semantic memory?

A

Word meaning. Used for identification and labelling, includes information on facts, features (used for recognition) and functions.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What evidence is there that the mental lexicon and semantic memory are distinct?

A

Semantic dementia, in which people can recognise but not label words - no links with real world.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What challenges are there to semantic theory?

A
  • Ambiguity - Bank vs. bank - look and sound the same but meanings differ, how is this represented?
  • Synonymy - Chair vs. seat
  • Set inclusion
  • Antonymy - Long is the opposite of short, the plank is long, so the plank cannot be short.
  • Partial knowledge (Putman, 1970, 1975) - ‘Beech’ and ‘larch’ denote different kinds of trees, we understand there’s a difference even if we can’t tell them apart.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What is a concept?

A

A mental representation of a category which determines how things are related or categorised. All words have an underlying concept but not all concepts are labelled by a word, for example there’s no special word for brown dogs.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Define denotation.

A

The core, essential meaning of a word and the relation between the word and the class of objects to which it can refer.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Define connotation.

A

The secondary implications or emotional/evaluative associations of a word, e.g. connotations of dog: nice, frightening, smelly etc. Not used in defining meaning as they differ from person to person.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

What is the classical approach to word meaning?

A

Referential theories, whereby the meaning of a word is the referent that it points to.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What challenges are there for referential theories of word meaning?

A

Abstract concepts e.g. justice, truth, as well as the representation of the meaning of the evening star vs. the morning star (both Venus).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What did Gottlob Frege (1892) state?

A

Reference (or extension) is only one aspect of a word’s meaning. Sense (intension) is also important, the abstract specification that determines how a word is related in meaning to other words. It specifies the properties an object must have to be a member of the class.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What are the psychological approaches to word meaning?

A

Decompositional approaches, network theories, prototype theories, and connectionist approaches.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

What theories use a decompositional approach?

A

Feature-list theories and feature verification processes.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

How can network theories be divided?

A

Into hierarchical and associative networks.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Who came up with feature-list theories?

A

Katz & Fodor (1963); Smith, Shoben & Rips (1974).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

Describe feature-list theories.

A

They ‘decompose’ meaning in terms of bivalent (Y/N) semantic features or semantic markers which capture the presence or absence of features.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

What are the assumptions of decompositional approaches?

A

Primitives (features) are finite, universal (across languages), and innate.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

What is Feature Comparison Theory (Smith, Shoben & Rips, 1974) an example of?

A

A feature-list theory, decompositional approach.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

Describe Feature Comparison Theory (Smith, Shoben & Rips, 1974).

A

Presence/absence insufficient, need distinction between types of features as some are more useful than others - words have defining and characteristic features.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

What do connectionist approaches typically emphasise?

A

Statistical patterns (co-occurrences).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

What do McRae, de Sa & Seidenberg (1997) discriminate between?

A

Intercorrelated features, which tend to occur together and are shared across many members of the same category, and distinguishing (or distinctive) features which enable us to distinguish among things and are exclusive to single items within a category.

23
Q

What did Cree, McNorgen & McRae (2006) state?

A

Distinguishing features hold a privileged status in semantic memory (they found that participants were faster to identify words from distinguishing features), making them easier to retrieve.

24
Q

Give an example of a semantic network model.

A

Collins & Quillian (1969).

25
Q

Describe Collins and Quillian’s (1969) semantic network model.

A

Hierarchically organised, concepts represented by nodes in a network. Links represent relations between concepts (categories/attributes), and smaller categories are represented at lower levels. The meaning of a word is determined by the place of the node in the network as a whole, and the principle of cognitive economy applies.

26
Q

What features are part of Collins and Quillian’s (1969) semantic network model?

A
  • Set membership (Fido is a dog)
  • Set inclusion (dogs are animals)
  • Part-whole (a seat is part of a chair)
  • Property attribution (canaries are yellow)
27
Q

What is the principle of cognitive economy?

A

The idea that information about concepts is stored at the highest appropriate level in the hierarchy.

28
Q

What did Hollan (1975) do?

A

Showed that simple semantic network and feature theories are formally equivalent - they both define words in terms of features used to label it.

29
Q

What did Rips, Smith & Shoben (1975) state about the similarities between semantic network and feature theories?

A

They may be formally equivalent, but this doesn’t imply that they’re psychologically equivalent. Different processing assumptions may result in more or less satisfactory performance models.

30
Q

What is a challenge for feature and semantic network theories?

A

They both define words in terms of necessary/sufficient conditions, but Wittgenstein (1953) pointed out that this isn’t always possible, e.g. with the word ‘game’ - the concept of the game is held together by an overlapping set of similarities between games like the similarities between members of a family - it’s difficult to find a single common feature, some examples may be more prototypical than others.

31
Q

Describe prototype theory (Rosch, 1973, 1978).

A

Concepts include a prototype e.g. a robin is more prototypical of ‘bird’ than an ostrich. Concepts also include some specification of how far category members can differ from the prototype. Boundaries may often be fuzzy rather than well-defined, for example whether a tomato is a fruit or a vegetable.

32
Q

What do prototype theories explain?

A

Why sometimes the correct classification of an object is in doubt even when its features aren’t, and why some exemplars of a concept are more typical than others and come to mind more readily.

33
Q

How can models be tested using sentence verification times?

A

Subjects are asked to respond true or false to sentences:
- Set inclusion: ‘a robin is a bird’ ‘a whale is a fruit’ and/or
- Property attribution: ‘a robin has feathers’ ‘a whale has seeds’
Words are presented in sentential context as in normal comprehension, and simple sentence frames are used to control differences in syntactic processing. Reaction times are used to infer how knowledge about word meaning is stored/used in the processes of comprehension.

34
Q

How can the principle of cognitive economy be tested?

A

As information about concepts is stored at the highest appropriate level in the hierarchy and it takes time to move through different steps in the hierarchy, sentence verification times should increase additively with distance through the hierarchy.

35
Q

What do sentence verification times show about the principle of cognitive economy?

A
  • Properties (subordinates) have slower RTs than superordinates.
  • Reaction times are slower the further in distance a sentence is.
  • Reaction time is dramatically lower for 0 levels of superordinate statements e.g. a canary is a canary because pattern matching is used rather than memory.
    This provides support for semantic network models and the principle of cognitive economy.
36
Q

What are alternative explanations to the principle of cognitive economy for sentence verification times?

A

Set size effects, or a hierarchy of information or association strengths.

37
Q

Outline the idea of set size effects.

A

Categories vary in size according to their position in the hierarchy, for example words > nouns > living > animals > dogs. Larger sets (e.g. noun) will take longer to search, and so set inclusion verification times could be explained in terms of differences in set size.

38
Q

What did Landauer & Freedman (1968) find regarding set size effects?

A

For both positive and negative instances, recognition of category membership for large categories took longer than for small categories. This supports set size as an explanation of sentence verification times.

39
Q

How did Conrad (1972) criticise Collins and Quillian’s (1969) model, and what alternative explanation was offered for sentence verification times?

A

Evidence by Collins & Quillian (1969) may have been an artefact of the hierarchies used - stimuli weren’t controlled for association strengths. Suggested that a hierarchy of information or association strengths could explain their findings.

40
Q

What did Conrad (1972) do?

A

Constructed hierarchies using normative data for properties obtained from a large group of students (asked participants to list features of words).

41
Q

What did Conrad (1972) find?

A
  • Little variation in sentence verification times within the high frequency categories.
  • Argued that Collins and Quillian (1969) used high associates as level 1 etc.
  • Therefore their property verification times can be explained in terms of differences in associative strength, rather than hierarchical distance.
42
Q

What problems are there for word meaning theories?

A

• Prototype or within-category typicality effects (Rips et al., 1973; Rosch, 1973)
- A robin is a bird (1 level) vs. an ostrich is a bird (1 level)
- A robin is a bird is easier and faster than an ostrich is a bird even though they’re both at 1 level
• Between-category typicality effects (Rips, Shoben & Smith, 1973)
- A dog is an animal is easier and faster to verify than a dog is a mammal
- But mammal is a sub-class of animal
- Solution: uncommon or specialised concepts in the hierarchy can be by-passed
• Semantic relatedness effects
• Semantic priming effects

43
Q

What did Rips, Shoben & Smith (1973) do?

A

Converted relatedness ratings into distance (semantic distance), found that prototypical birds are closer to ‘bird’. This explains within-category typicality effects.

44
Q

What did Smith, Shoben & Rips (1974) state?

A

Concepts are decomposed into sets of features which are either defining (D) or characteristic (C). Distances are shaped by objects’ characteristic features.

45
Q

What support is there for Smith, Shoben & Rips (1974)?

A

Linguistic evidence – hedges (using modifiers to represent our certainty) (Lakoff, 1972).
- ‘A robin is a true bird’ (D+C)
- ‘Technically speaking, a chicken is a bird’ (D)
- ‘Loosely speaking a bat is a bird’ (C)
Also experimental evidence (Rips et al., 1973) - semantic relatedness ratings reflect the extent to which the C features of a superordinate are similar to the features of the instance. This is problematic for hierarchical models, as semantic relatedness appears to play a significant role.

46
Q

Describe the spreading activation model (Collins & Loftus, 1975).

A

A non-hierarchical network in which nodes represent concepts and links represent shared properties (relationships). Searching the model leads to spreading activation across different nodes, which occurs automatically and in parallel.

47
Q

What evidence is there for the spreading activation model (Collins & Loftus, 1975)?

A 
Semantic priming (e.g. Meyer & Schvaneveldt, 1971)
  - Prime (doctor)  target (nurse)  fast decision LDT
  - Prime (tree)  target (nurse)  slow decision LDT
  - Activation of nurse by doctor is greater than activation of nurse by tree, which can be explained in terms of spreading activation.
Between-category typicality effects
  - A dog is an animal vs. a dog is a mammal
  - Stronger associative links between dog and animal than dog and mammal 
Prototype effects/within-category typicality effects
  - Can be explained in terms of associative links of varying strength 
Variability in negative judgment times
  - Can be explained in terms of varying associative strength
  - Explains why it takes longer to reject closely related items e.g. a canary is an ostrich vs. a canary is a salmon.
48
Q

What problems are there with the spreading activation model (Collins & Loftus, 1975)?

A
  • Proliferation of auxiliary assumptions about the way information is accessed e.g. processing factors
  • Impossible to design an experiment to test underlying theory of representation vs. auxiliary processing assumptions - difficult to falsify
49
Q

According to connectionist approaches to semantics, what is semantic memory defined by?

A

Semantic microfeatures which mediate between perception, action and language like nodes in the SA model but with no straightforward linguistic counterparts. They encode knowledge at a very low (abstract) level of semantic representation, and meaning is represented as patterns of activation across many microfeatures.

50
Q

What techniques can be used to look at co-occurrences between words?

A

Latent semantic analysis (LSA) - Landauer & Dumais (1997) - a mathematical/statistical technique for extracting and representing the similarity of meaning of words and passages by analysis of large bodies of text.
Hyperspace Analogue to language - Lund and Burgess (1996); Burgess and Lund (1997).

51
Q

What is the difference between LSA and Hyperspace Analogue to language?

A

Size of context employed:

  • LSA: paragraphs + texts
  • HAL: sentences
52
Q

What is the grounding problem?

A

Meaning is defined as relation with other words, but how do we know the meaning of those words?
What is the relation between abstract symbols and our knowledge, experiences, actions and perceptions?
These issues are also problematic for prototype and semantic network theories

53
Q

What is a possible solution to the grounding problem?

A 
Rogers et al. (2004)'s connectionist model in which meaning is represented as patterns of activation. 
Two input/output layers:
1. Verbal descriptors (4 subunits)
  - Names (goose)
  - Perceptual (has wings)
  - Functional (can fly)
  -Encyclopaedic (migrates)
2. Visual features
Semantic layer mediates between these two layers.
54
Q

What does Rogers et al. (2004)’s approach provide a good explanation for?

A

Semantic dementia.

C82COG - Language (6 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2024 Bold Learning Solutions. Terms and Conditions