Cognitive, Understanding Meaning, WEEK 9 Flashcards
How do we know what words mean?
- e.g. what does the word “cat” mean?
- Approach 1 would be to define the word > e.g. a carnivorous mammal long domesticated and kept by humans as a pet or for catching rats and mice (probably both)
- BUT, does approach 1 actually cover the meaning of cat? > w/ this definition, how do you know what mammal means? Or rat or pet?
- Approach 2 for explaining the meaning of something is by pointing at it and saying that is it or it is one of those > e.g. pointing to a cat
- BUT, for approach 2, you cannot always point to things > they may be inanimate e.g. cannot explain what trust means by pointing at it.
- Pointing at things also does not relate what it means to other things > e.g. by pointing at a cat we won’t know it is related to being a pet or collecting mice. Part of the meaning of cat is that it can breathe, has hair and eyes but pointing at it does not cover those meanings
Approaches to meaning: Semantic Networks
• One approach to explaining meaning is to put forward semantic networks
• Semantic networks can be shown by a hierarchical network of unitary nodes > there is no internal structure within nodes + there is labelled links between nodes
• E.G. if you look at the network diagram, we see a hierarchy with “animal” at the top so that means EVERYTHING BELOW is also an animal alongside the labelled link which is “breathes” meaning every node below is an animal and has the property of breathing.
• Others may differ, e.g. one type of animal is a reptile which has scales + is cold blooded, below this in the hierarchy is snake which is a type of animal and reptile, having scales and is cold-blooded but is different from others as it has no legs.
• This works well for proposing hierarchies and assigning properties to each node + properties that are inherited > everything below the node the animal breeds so is inherited (e.g everything below animal breathes)
Predicts sentence verification times (more links to cross, more time)
Evidence for semantic networks
• Sentence verification data supports semantic networks > asking ppts to verify sentences meaning they have to say whether it is true or false + measure the time needed to verify the sentence
• E.G. A robin is a robin (fastest – no links), A robin is a bird (faster – 1 link), A robin is an animal (slower – 2 links)
• For robin is a bird, you only need to cross one link so we are quicker to verify it (robin is on the level just below bird) for robin is an animal you need to cross 2 links, this is why people are slower here because 2 links need to be crossed
• This is true for properties too e.g. A robin has wings (faster – wings at bird level) > only birds mainly have wings, A robin has lungs (slower – lungs at animal level) > slower because you store the robin having lungs at animal level > all animals have lungs + is not explicit to robins > this is how we store meaningful information in our semantic memory
• HOWEVER, A cow is a mammal (slower than) A cow is an animal > this happens even though you cross less links for a cow is a mammal than animal > this is because there is also an effect of familiarity (more familiar with a cow being an animal than mammal)
• A robin is a bird (faster than) A penguin is a bird > this is because robins are a more typical bird than penguins > there is a typicality effect (what fits the typicality or typical characteristic > more common)
• means semantic networks as an explanation for explaining meaning is not enough alone + does not get the full scope > there may be different ways in which people organise semantic memory e.g. in terms of semantic relatedness or semantic distance instead of the hierarchical models
• Models focusing on semantic relatedness + distance deal with typicality and familiarity better
We don’t know the best way to understand how our semantic memory is organised yet
Problems with networks
• The problem lies with the definition as the network implies we have sharp definitions but is it true that our definitions are actually this precise?
• clear definitions need necessary + sufficient conditions > well-defined set of attributes where if all of them are present we have X (object, person etc)
• E.g. Bachelor > never married, but old enough to be > Male. These are the necessary AND sufficient conditions to be a bachelor
• Problem: almost impossible to find a set of necessary + sufficient conditions + be able to adequately define something in a way that captures it’s the true meaning
• E.g. define “game” using necessary and sufficient conditions > very complex to do so because there are so many different types of games (board games, Olympic games, card games) > they have some things in common like it is something you play/participate in but there are many differences > cannot even say something where you win or lose as not all games are like that > cannot capture a true/nuanced definition > Wittgenstein famous example
• We have a complicated network with some things overlapping + others not > it is hard to truly define something precisely using necessary + sufficient conditions > Wittgenstein shows a problem for learning and meaning
○ Problem for learning: how do children figure out what a game is? If you cannot tell a child what a game is then how does the child understand what a game is? The child eventually understands this but how? The set is not arbitary aka not determined by chance as not everything is a game + child knows/learns this
○ Problem for meaning: people’s definition of game may differ so how do we understand each other’s meaning?
Word meaning is different from world knowledge
- How does this distinction influence how we comprehend language?
- Do we first compute the meaning of word then later relate it to our world knowledge?
- Or do we integrate the meaning of the word and world knowledge at the same time when we listen to language?
Word meaning & world knowledge: Hagoort, Hald, Bastiaansen and Petersson (2004)
- suggests we integrate meaning + world knowledge when we are comprehending language
• Ppts from the Netherlands are shown three statements + see if they detect a mismatch using brain activity (EEG)
1. Dutch trains are yellow and very crowded (True)
2. Dutch trains are sour and very crowded (False due to meaning of word sour > sour is a taste)
3. Dutch trains are white and very crowded (False because doesn’t fit world knowledge, Dutch trains are always yellow)
• The researchers wanted to see if they detect a mismatch in 3 as fast as 2 > used EEG because it is time sensitive (commonly used in language research as it unfolds + processes very quickly)
• In the exp, researchers looked at the N400 which is an ERP (a potential in the signal which comes about due to an event > EEG measurements during sentence reading > N400 amplitude is an index of brain detecting a mismatch
• N400 is a negative ERP > you get a negative change in the voltage signal in ERPs + happens around 250ms after the event + peaks at around 400ms (why it is called N400)
• When your brain detects a mismatch between word meaning + sentence context, you get an N400 ERP
• Point 0 in the figure shows when the word yellow, sour or white is said > result shows response to yellow is not very high, does not go negative because a mismatch has not been detected
• But response to white (sentence incorrect based on world knowledge) there is a negative ERP, N400 (not as clear as sour)
• Similarly, response to sour (incorrect based on word meaning) shows a negative ERP + N400 > very clearly shows N400 which is expected as “sour” does not fit sentence context
• people detect the mismatch in 3 as quick as in 2 > difference between incorrect sentence based on world knowledge + word meaning is minimal
• In 3, the sentence context is not the only present context, we have our world knowledge too providing context
• people can detect a sentence mismatch (word meaning) as fast as they can for world knowledge > supporting that when reading a sentence, brain retrieves and integrates world knowledge + word meaning at the same time
Broca’s area
- Broca’s area is considered to make up the inferior frontal gyrus which is within the frontal lobe > for most people the Broca’s area resides in the left cerebral hemisphere
- Broca’s area was named after Paul Broca who suggested this area is important in speech production
- Based this from case studies of patients who had damage to this area + had a speech deficit > famous case where patient could only say the word 10 > didn’t have MRI scanners back then + look at brain after death + found he had damage to an area in the frontal lobe > Broca’s area
- The particular condition Broca observed is called Broca’s aphasia
- Broca’s aphasia: deficit in ability to produce language > patients with Broca’s aphasia, reading and writing are also often impaired but language comprehension is relatively preserved
What is the precise role of Broca’s area in language?
• Still being debated > damage to Broca’s area can disrupt language production but we don’t know exactly what language-related function is lost to cause that disruption
• Some suggest that damage to Broca’s area is associated to muscle movement which produce speech (e.g. tongue and mouth)
• Others argue it is associated to grammar, syntax, verbal working memory or all
• Broca’s area is thought to also have non-linguistic functions such as role in language comprehension, movement + understanding movement of others
Although Broca’s area seems to play a role in language, the overall function is more complex > too simplified
Wernicke’s area
• Wernicke’s area is suggested to reside in the left cerebral cortex, near the junction between the temporal and parietal lobes > posterior temporal lobe > serves language comprehension
• Found by Carl Wernicke who suggests that damage to this area results in a deficit where patients are able to produce speech which resembles fluent language but is actually meaningless, comprehension was impaired > called Wernicke’s aphasia
• Wernicke’s aphasia: Patients with this use made-up words or words substituted for another to produce speech making little sense. They also show a deficit in their ability to comprehend language
• Wernicke proposed a model for language that involved both the region he discovered and Broca’s area
○ Wernicke-Geschwind model: suggests Wernicke’s area is responsible for making plans for meaningful speech
○ Alongside this, the Broca’s area is then responsible for taking those plans and determining movements + send to motor cortex to vocalise them (e.g. moving tongue + mouth to speak)
Is the Wernicke-Geschwind model too simplistic?
- Research suggests the model is too simplistic as studies indicate language likely requires widespread networks + cannot be boiled down to a connection between two areas > too simple
- Evidence also suggests that the Wernicke’s area may be involved in speech production not just comprehension > thus researchers are still trying to understand the precise contribution of Wernicke’s area to language.
Outdated view of how the brain supports language
- Old views suggest only in the left side of the brain we have two crucial regions > Broca’s area and Wernicke’s area
- Based on language of aphasia patients
- Due to technological advances we can better understand brain behaviour > even though this is dated, Broca’s and Wernicke’s area are still influential and important today due to the impact they had before
The challenge of processing language
• to do with integrating information > when you process language you don’t speak or listen to single words > the challenge and uniqueness comes from the fact we produce multiple words which is more difficult + requires integration of different words at the level of meaning
• Meaning integration: integrating different words can change the meaning
○ E.g., “flat” > we know the meaning of flat on it’s own however it can mean different things when integrated with other words
○ E.g. flat beer, flat tire, flat note > when you pair “flat” with a second word, we compute and integrate the words to understand the meaning > integration of words can alter the meaning of the first word
• Don’t integrate info just at the level of meaning but also at level of phonology and syntax etc..
• Syntactic (grammatical) integration: grammar and syntactic roles can change how a sentence is seen when the information is integrated
○ E.g., “dog, chase, cat” > suggests dog is the subject + is chasing the cat
○ But, switching to “cat, chase, dog” changes the meaning so now the cat is the subject + chasing the dog
• processing language is difficult because you are not just retrieving info, you are integrating information too
How to find which brain regions underlie syntactic comprehension?
• studied with priming paradigms where the ppt is first presented with a priming stimulus + when they see the next stimulus which is related to the prime, their RT or brain activity is lower = repression suppression effect
• E.g. Priming effect in the brain is called a repetition suppression effect. E.g. in the case of reading words, the brain areas responsible for reading words need fewer resources the second time the word is presented, so there is a suppression effect in these brain areas. > the brain areas will light up less so the second time (less active) > W8
• E.g. first time you listen to a sentence with a certain syntactic structure like a passive sentence, your brain will show a big change because it is the first time you are seeing it so you are more responsive > the second time you hear a similar syntactic structure such as another passive sentence, there is less of a change in brain activity, less of a response
• This feature can be used to find which brain regions support syntactic comprehension
• Ppts are in the scanner + choose to repeat or not repeat the syntactic structure of the sentence + those regions with the repetition-suppression response (areas giving less of a response second time round) can be seen as areas in the brain supporting syntactic comprehension
• The brain image comes from a meta-analysis which looked at different studies conducting the same kind of manipulations where there is a repetition-suppression response
• We see the repetition-suppression response in the left inferior frontal gyrus which is also known as the Broca’s region > repetition-suppression is also in the posterior temporal lobe which is referred to as Wernicke’s area > somewhat supports the more dated views
• However, we can see that other regions come into play too, this effect can be seen in the parietal regions + in a bigger part of the frontal cortex (extends from frontal cortex to pre-motor cortex) + is not only in the left but also in the right side of the brain in the right inferior frontal cortex
This shows a widespread network underlies syntactic comprehension + importantly, not just Wernicke’s area which is meant to be for comprehension or Broca’s area for production.
Which brain regions underlie syntactic comprehension?
• Ppts are in the scanner + choose to repeat or not repeat the syntactic structure of the sentence + those regions with the repetition-suppression response (areas giving less of a response second time round) can be seen as areas in the brain supporting syntactic comprehension
• The brain image comes from a meta-analysis which looked at different studies conducting the same kind of manipulations where there is a repetition-suppression response
• repetition-suppression response in the left inferior frontal gyrus which is also known as the Broca’s region > repetition-suppression is also in the posterior temporal lobe which is referred to as Wernicke’s area > somewhat supports the more dated views
• However, we can see that other regions come into play too, this effect can be seen in the parietal regions + in a bigger part of the frontal cortex (extends from frontal cortex to pre-motor cortex) + is not only in the left but also in the right side of the brain in the right inferior frontal cortex
This shows a widespread network underlies syntactic comprehension + importantly, not just Wernicke’s area which is meant to be for comprehension or Broca’s area for production.
Which brain regions underlie language + syntactic production?
- If we look at language production in a similar way as with syntactic comprehension, the same areas seem to play a role > includes posterior temporal, middle and superior posterior temporal gyrus > exterior temporal lobe (Wernicke’s area)
- Also involves Broca’s area extending into the frontal cortex + premotor cortex
- All of these areas are important in syntactic production + comprehension > not one area is responsible for production or comprehension, a network of areas are important and respond for both > shouldn’t make a distinction between language production and language comprehension
