Lecture 8 - The Speaking Brain Flashcards
(34 cards)
Non-human language. Chimp study & sign language - Gardner et al 1989
Washoe (chimp) - learnt 200 manual signs.
However, evidence of overgeneralisation (signed hurt when meaning tattoo) & combined words for unfamiliar objects (waterbird for duck)
Bonobo chimp Kanzi study - Savage-Rumbaugh et al
Learned how to use arbitrary written symbols to communicate. However, mainly food requests (learnt through reward)
What is The Chinese Room study? - Searle 1980
Syntax doesn’t suffice for semantics - to use language you have to understand the meanings of the thought contents
- English speaker who doesn’t understand Chinese receives written questions in Chinese.
- He is given a set of instructions in English that guide him on how to manipulate the Chinese symbols based on their shapes and patterns.
- He can follow the instructions, understand the symbols, and produce appropriate responses in Chinese.
- He does not genuinely understand Chinese.
This challenges artificial intelligence, which posits that a computer program, running on the appropriate hardware, can possess genuine understanding and consciousness.
The Cohort Model
- Process of word recognition, the brain initially activates a set of potential word candidates based on the input’s phonetic features.
- Competition between similar sounding words.
- As more phonetic information is processed, the cohort model proposes that the set of potential word candidates narrows down.
- Brain eliminates words from the cohort that do not match the subsequent phonetic information. In our example, as the brain receives the next sound, “a,” the word “boy” would be eliminated from the cohort, as it does not match the current input.
What is the uniqueness point in relation to the cohort model?
Where the acoustic input unambiguously corresponds to only one known word - process of elimination. Model explains recognition of single spoken words.
What ERP component is involved in recognising the form of a spoken word?
N400 - retrieves its meaning and links to wider context of the utterance
How does N400 measure semantics? - Kutus & Hilliard 1980
- ERP reflects negative peak at 400 ms after the onset of a word.
- Amplitude of N400 depends on whether the word is appropriate. ‘He took a sip from the glass’ = semantically appropriate. ‘He took a sip from the transmitter’ = semantically inappropriate.
- Large N400 peak to contextually anomalous word (inappropriate) AND factually wrong sentences ‘dutch trains are white’ (they are yellow - PPT that know this will have a large N400 peak)
ERP N400 component and the development of semantic representation - infant study (Parish & Csibra 2012)
‘Look at the duck’ - move wall down and reveal duck or other object. 9 month old infants detect the mismatch between object appearing from behind occluder and the label given . N400 peaked higher when incongruent image was revealed . N400 modulated by semantic connection between expected and presented word
What does amodal representation of semantic memory mean?
- Semantic memory is not tied to one or more perceptual systems.
- Features are represented as abstract knowledge
- independent of input or output modality
Collins & Quinlan 1969 Hierarchical model - amodal model of speech recognition
SUPERORDINATE -> ORDINATE -> SUBORDINATE
Superordinate (animal / transport) to ordinate (bird / car) to subordinate (ostrich / Ferrari)
- Goes up the hierarchy (faster at classifying a robin as a bird than an animal)
- Not all concepts are hierarchical (truth or law)
What brain area is involved in semantic memory?
Lateral temporal lobes - the ventral ‘what’ route of speech processing
What specific brain area is involved in superordinate category?
Posterior temporal regions (less specific - animal)
What specific brain area is involved in subordinate category?
Anterior temporal regions - ordinate (bird) to subordinate (robin) = a gradient from posterior to anterior.
What impairments do patients with anterior temporal lobe damage have?
Lack of ability to make subordinate classifications.
Grounded models of semantic memory - argues the opposite of amodal
The grounded model suggests that when we encounter a word, we activate mental representations that involve sensory and motor information associated with the word’s referents.
Who formed the fully grounded model of semantic memory?
Allport, 1985
What is the fully grounded model?
Different features of a concept is represented in different information channels (modalities) - the same channels the info is acquired through.
What modalities are connected to the word ‘telephone’?
Auditory regions - how it sounds
Visual regions - how it looks
Action related regions - how to use it
What neural substrates are activated when processing action words? - Hauk et al 2004
Activates body based neural representations . ‘Kick’ ‘lick’ ‘pick’ all activate corresponding part of motor cortex (leg, mouth, finger). Evidence for grounded cognition
What do Mahon & Caramazza (2008) suggest?
Core system within semantic memory is amodal but modality-specific representations are evoked more as a downstream by-product. Core system of word classification / semantics is amodal but processing leads to activation of sensory-specific info.
What is the sensory functional distinction? - Warrington & Shallice, 1984
Certain categories may depend on certain types of knowledge - animals and fruit and veg may be defined more by their sensory properties whereas inanimate objects defined by their functions
What brain activity occurs when naming pictures of animals? - Chao et al 2002
Category related activity in the ventral occipito-temporal cortex- occurs when Ppt read names of or questions about animals and tools (reading names and answering questions, naming images and viewing images)
Different categories rely on different characteristics (animals vs tools)
Animal - sensory characteristics more important (features such as eyes, mouth)
Tools - man made - functional characteristics (shape and function)
IS category specificity innate?
Some categories are hardwired - animals, plants, tools
Prenatal face processing
Newborns prefer biological motion
