knowledge Flashcards
episodic - semantic distinction
- Episodic memory:
- Reference is to oneself
- Organised temporally
- Events recalled “consciously”
- Susceptible to forgetting
- Context dependent
· Semantic memory: - Reference to knowledge only
- Not organised temporally
- Events are “known”
- Relatively permanent
- Context independent
evidence and challenges for episodic - semantic distinction
· Interdependence of the systems makes distinction unclear.
· Does memory of an important autobiographical event remain episodic or become semantic?
- Some evidence from different pathologies e.g., Semantic Dementia.
evidence from semantic dementia
· Neuropathology:
- Non-alzheimer type degenerative pathology of the polar and inferolateral temporal cortex (relative sparing of hippocampus in early stages).
· Symptoms:
- Progressive, selective deterioration in semantic memory.
- Reasonably preserved episodic memory.
- Graham et al (2000): A comparison of SD with Alzheimer’s disease (AD)
knowledge informs many aspects of language processing
· Much more than a store of word meanings or grammar rules.
· When we read or listen to spoken language we use our knowledge to make inferences - “you have expensive tastes”, “not me, my sister. She works in the city. I can only afford plonk”
- The ability to make correct inferences underlies language comprehension… getting it wrong underlies much of humour.
feature comparison models
· Concepts (mental representations of objects/categories) are stored as lists of defining or characteristic features.
· Sentence verification task:
- Reaction time taken to verify a sentence
1. Is a carrot a vegetable?
2. Is an artichoke a vegetable?
- 1 is quicker as carrot has more characteristic features, called the “Typicality Effect”.
· Limitations:
- Few concepts can be reduced to list of defining characteristics e.g., a “game”
- Features are not independent (objects with leaves are less likely to have legs than objects that have fur).
prototype models
· Category membership is not clear cut - prototype models define the “centre” (average) of a category, rather than its boundaries.
· Prototypicality effect:
- Defining characteristics of “batchelor” - male/unmarried
1. Single 32 year old man
2. 9 year old boy
3. 64 year old catholic priest
- Faster to confirm 1 than either 2 or 3 (Lakoff, 1987). Called the “Prototypicality effect”
· Categories have a graded structure - some members are more representative (prototypical) than others.
· Key claims of prototype models:
4. Prototypes differ from non-prototypes:
- These examples of a category most often generated are also those rated as most prototypical (Mervis et al, 1976).
- Semantic priming - we are faster to decide if “Apple” is a word if it is preceded by the unconsciously presented word “Fruit”. This priming effect is greater for prototypes.
- Prototypes share most features with other category members.
5. Objects are categorised at 3 different levels - superordinate, basic, subordinate.
basic level categories have special status
- We sue them to name objects
- They show the largest semantic priming effects
- They distort memory
· Benefits: - Can explain a lot of empirical findings
- Allows “loose” concepts to be created
· Limitations: - Prototypes can change with context e.g., Novick, 2003
- Prototypes “lose” information
- Experts have different prototypes…
· An alternative view is that we use exemplars, not prototypes e.g., Nosofsky, 1991.
- The two models are related as they both involve comparisons between external reality and an internally generated “standard”.
exemplar models
· Specific exemplars rather than ‘averaged’ prototypes drive the typically effects - easy to find specific memories of common objects (e.g., apple) compared to uncommon ones (e.g., fig).
· Benefits:
- Do not require abstraction
- Preserve information about variability
- Is a 19” object more likely to be a Pizza or a Ruler?
- 12” is prototypical for both but you are more likely to have
seen larger pizzas (Rips & Collins, 1993)
· Limitations:
- Applies to complex more than simple categories e.g., even no.s.
- Infers that we store every single example of a category.
- Both Exemplar and Prototype models can coexist.
network models - hierarchal nets
· Emphasise the relationship between concepts.
· Relates to feature comparison models.
· Key concepts:
- Concepts represented by Nodes.
- Nodes connected by links.
- Semantic Relatedness.
· People faster to verify - canaries are yellow than canaries have feathers.
- Limitation - participants are very fast to verify “peacocks have feathers”.
network models - semantic nets
· Key concepts:
- Semantic Distance
- Spreading activation
- Link strength or weight (depicted by length here)
· Evidence:
- Semantic priming greater for “closer” items (McNamara, 1992)
- False memory effects - DRM paradigm (Deese-Roediger-McDermott)
network models - ACT-R
· A theory of all cognition
· Attention/ visual cognition/ memory/ action
· Declarative memory is one small part
· Declarative memory consists of propositions - the smallest unit of information can be declared true or false.
1. Susan gave a cat to Maria
2. The cat was white
3. Maria is president of the club
* Susan gave a white cat to Maria who is president of the club.
* As with Collins and Loftus, the more often links are activated, the stronger (faster) they become.
* The number of links that can be supported is limited by the strength of the concept.
* Fan effect:
- Time to make true/false decisions increases with the number of items of information associated with the location and person.
- ACT-R assumes a finite amount of activation - if it is spread thinly, then decisions are slower and acquiring new connections is harder.
network models - parallel distributed processing
· “Neuron like” nodes connected by links
· A concept is represented by a pattern of activity distributed across many modes.
· Links have different weights
· Information processing proceeds in parallel - many patterns of spreading activation may occur at the same time.
- PDP models allow for spontaneous generalisation and graceful degradation.
schemas and script
· Semantic memory stores more than just concepts
· Schemas - generic information about situations/general rules that are typically accurate.
· Scripts - structured, ordered sequence of events - can be a prototype - an abstraction.
· Research into the role of schemas has long histroy (Piaget/Bartlett etc) but fell out of fashion whilst behaviourism was dominant.
- Schema research highlights the interplay between “top down” and “bottom up” information processes.
schemas and retrieval
· Schemas can distort memory (Brewer and Treyens, 1981)
· Participants spend 35 seconds waiting in office
· Most participants remembered desk/chair/shelves/typewriter etc.
· Books and filing cabinets often recalled.
· Picnic hamper and wine bottle rarely recalled
- Note, other research suggests we remember schema - inconsistent information.
summary
· There is evidence for a distinction between episodic and semantic memory.
· Semantic memory is central in inferences and language processing.
· Different models of concept storage include: feature comparison, prototype, exemplar, network.
· Evaluating these models relies on assessing the degree to which they reflect human failings e.g., false memory, generalisation, graceful degradation
- Semantic memory stores more than concepts - schemas and scripts
