Cognitive Psychology – Language
5 Facts about language
1. Meaning of words/concepts are fuzzy
2. Semantic memory influences language processing
3. We make a lot of inferences when processing language
4. Be wary of distinctions between comprehension and production
5. There are advantages to being bilingual
1 – Semantic memory and meaning
How do we know the meaning of words?
“CAT”
o Approach 1: definitions: a carnivorous mammal long domesticated and kept
by humans as a pet or for catching rats and mice (probably both)
BUT: what does “mammal” mean?
What does “pet” mean?
What does “rat” mean?
o Approach 2: it’s one of those point to a cat
BUT: not everything can be pointed at, and pointing can’t relate cats
to other things (breathes, has hair etc.)
Approaches to meaning: Semantic Networks
Network of unitary nodes (no internal structure) and labelled links between them.
Good for hierarchies
Properties are inherited (e.g.
everything below animal breathes)
Predicts sentence verification times
(more links to cross, more time)
Evidence for semantic networks: sentence verification data
Task: present sentence – participants must respond true/false
o A robin is a robin (fastest – no links)
o A robin is a bird (faster – 1 link)
o A robin is an animal (slower – 2 links)
o A robin has wings (faster – wings at bird level)
o A robin has lungs (slower – lungs at animal level)
BUT:
o A cow is a mammal is processed slower than a cow is an animal –
FAMILIARITY – have encountered a cow is an animal before but not
necessarily encountered the fact that a cow is a mammal before.
o A robin is a bird is processed faster than a penguin is a bird – TYPICALITY
, Semantic memory organised on basis of semantic relatedness or semantic distance
(instead of hierarchy) deals with this better.
Problems with networks
Definition problem – are definitions really as sharp as a network implies?
Necessary and sufficient conditions are the usual approach to sharp definitions
o Well defined set of attributes and if all of them are present, we have X
o E.g. bachelor – never married, but old enough to be, male
Problem: necessary and sufficient conditions can rarely be found for an adequate
definitions
Wittgenstein’s example shows the problem:
Problem for learning: how do children figure out what a game is if there is no
definition, but the set is also not arbitrary – i.e. not everything is game?
Problem for meaning: how do I know what you mean by ‘game’? What if you pick
one way to make up the family and I pick a different way?
Concepts are fuzzy!
2 – Influence of semantic memory on sentence comprehension
Word meaning vs. world knowledge
How does meaning influence sentence comprehension?
o First word meaning, the world knowledge?
o Or, integrate word meaning and world knowledge at the same time?
o The following study shows that we integrate world meaning and world
knowledge at the same time in order to understand the question/sentence
and know whether it is true or false.
Experiment by Hagoort, Hald, Bastiaansen and Petersson (2004), published in
Science.
o Participants were asked questions about the Dutch train.
o “The Dutch trains are yellow and very crowded” – TRUE
o “The Dutch trains are sour and very crowded” – FALSE – because of the
meaning of sour
o “The Dutch trains are white and very crowded” – FALSE – because Dutch
trains are yellow not white
Do people detect a mismatch in 3 as fast as they do in 2? YES
EEG measurements during sentence reading N400 amplitude is an index of brain
detecting a mismatch
3 – Discourse processing and inferencing
World knowledge is during language processing
Remember! Schemata & scripts
Semantic memory is more than just concepts, also relations amongst concepts
5 Facts about language
1. Meaning of words/concepts are fuzzy
2. Semantic memory influences language processing
3. We make a lot of inferences when processing language
4. Be wary of distinctions between comprehension and production
5. There are advantages to being bilingual
1 – Semantic memory and meaning
How do we know the meaning of words?
“CAT”
o Approach 1: definitions: a carnivorous mammal long domesticated and kept
by humans as a pet or for catching rats and mice (probably both)
BUT: what does “mammal” mean?
What does “pet” mean?
What does “rat” mean?
o Approach 2: it’s one of those point to a cat
BUT: not everything can be pointed at, and pointing can’t relate cats
to other things (breathes, has hair etc.)
Approaches to meaning: Semantic Networks
Network of unitary nodes (no internal structure) and labelled links between them.
Good for hierarchies
Properties are inherited (e.g.
everything below animal breathes)
Predicts sentence verification times
(more links to cross, more time)
Evidence for semantic networks: sentence verification data
Task: present sentence – participants must respond true/false
o A robin is a robin (fastest – no links)
o A robin is a bird (faster – 1 link)
o A robin is an animal (slower – 2 links)
o A robin has wings (faster – wings at bird level)
o A robin has lungs (slower – lungs at animal level)
BUT:
o A cow is a mammal is processed slower than a cow is an animal –
FAMILIARITY – have encountered a cow is an animal before but not
necessarily encountered the fact that a cow is a mammal before.
o A robin is a bird is processed faster than a penguin is a bird – TYPICALITY
, Semantic memory organised on basis of semantic relatedness or semantic distance
(instead of hierarchy) deals with this better.
Problems with networks
Definition problem – are definitions really as sharp as a network implies?
Necessary and sufficient conditions are the usual approach to sharp definitions
o Well defined set of attributes and if all of them are present, we have X
o E.g. bachelor – never married, but old enough to be, male
Problem: necessary and sufficient conditions can rarely be found for an adequate
definitions
Wittgenstein’s example shows the problem:
Problem for learning: how do children figure out what a game is if there is no
definition, but the set is also not arbitrary – i.e. not everything is game?
Problem for meaning: how do I know what you mean by ‘game’? What if you pick
one way to make up the family and I pick a different way?
Concepts are fuzzy!
2 – Influence of semantic memory on sentence comprehension
Word meaning vs. world knowledge
How does meaning influence sentence comprehension?
o First word meaning, the world knowledge?
o Or, integrate word meaning and world knowledge at the same time?
o The following study shows that we integrate world meaning and world
knowledge at the same time in order to understand the question/sentence
and know whether it is true or false.
Experiment by Hagoort, Hald, Bastiaansen and Petersson (2004), published in
Science.
o Participants were asked questions about the Dutch train.
o “The Dutch trains are yellow and very crowded” – TRUE
o “The Dutch trains are sour and very crowded” – FALSE – because of the
meaning of sour
o “The Dutch trains are white and very crowded” – FALSE – because Dutch
trains are yellow not white
Do people detect a mismatch in 3 as fast as they do in 2? YES
EEG measurements during sentence reading N400 amplitude is an index of brain
detecting a mismatch
3 – Discourse processing and inferencing
World knowledge is during language processing
Remember! Schemata & scripts
Semantic memory is more than just concepts, also relations amongst concepts
