Lecture 6: Semantic Memory Flashcards
What is a knowledge category?
- group of objects that belong together and have something in common
What is an exemplar?
- item in the category
Why are categories important for prediction?
- categories allow us to predict what is likely in new situations because they treat all exemplars as similar
What is the classic view of categorization?
- categories are defined by a list of necessary and sufficient features
What is a necessary feature?
- item must have all defining features to be included in the category
- i.e. humans are bipedal mammals
What is a sufficient feature?
- attribute not required for category membership
What is the problem with the classic view?
- items can be categorized as whether they are more or less typical examples of a category than others (graded membership)
- leads to typicality effects
What is prototype theory?
- categories have a graded structure
- rather than a set of defining features, exemplars have characteristic features
How is category membership determined according to prototype theory?
- determined by matching the item with prototype stored in memory
- categories have a central tendency where exemplars with most characteristic features are found
- all category members share a family resemblance even if they are not typical
What is exemplar theory?
- items are divided into categories and sorted based on defining features + exceptions
- categories are based on mental accounts of each experience, instance, or example of the encounters a person has had with members of that category
- generalize from many members to form a prototype
What is the problem with similarity theories like prototype and exemplar theory?
- people can give typicality ratings to clearly defined categories
(rank these numbers based on their “odd-ness”) - people can make up reasons that two items are similar
(how are a television and a banana similar?) - category members may not be typical, and typical cases may not be category members
(is a tomato a fruit or a vegetable?)
What is the theory-based or explanation-based view of meaning?
- knowledge and beliefs are important for categorizing
- we have complex IMPLICIT ideas about categories and concepts
→ includes idea of the CAUSE of category membership, not just features
→ “psychological essentialism” i.e. things just are that way because they are
How did Rips’s 1989 experiment demonstrate psychological essentialism?
- cats can’t be anything else but cats
- toasters can become something other than a toaster
- something about them is essentially this or that
What are semantic network models?
- concerned with how different items are related to each other
- all network models are based on idea that nodes contain information that are connected to each other by directional pathways
How are nodes in semantic network models activated?
- activated via spreading activation to connected pathways
→ primes related concepts
What is Collins & Quillian’s hierarchical model?
- nodes are in a network where a set of concepts that are related to each other are connected
- concepts can activate some or several nodes
What is Collins & Loftus’ Semantic Relatedness model?
- nodes are organised based on the strength of their relationship
- stronger associations are represented by shorter pathways
→ thus typical exemplars have shorter pathways
What are feature comparison models of semantic memory?
- believe categories exist as a list of features (both defining and characteristic)
→ similar to classic view
How is category retrieval accomplished by feature comparison models?
- accomplished by comparing feature lists in 2 stages:
→ Stage I is a fast search for large feature overlap.
→ Stage II is a slower search for defining feature overlap.
What are connectionist models of semantic memory?
- based on how brain is structured
- consists of highly connected neuron-like units
What are the features of the units in connectionist models?
- layer of output, hidden, and input units
- each unit can be inactive, excitatory, or inhibitory
- each unit has a connection with different weights (+1 to -1)
How is information stored in connections models?
- stored as a pattern of activation
- new information changes the weights so the pattern
Why is remembering reconstructive?
- use general knowledge and expectations based on past experiences to organize memories
What was Bartlett’s work with memory?
- repeated reproduction technique
→ used stories, poems and had people memorize them
→ then asked to retell the story several times over several intervals - reported the constructive nature of memory
What is the advantage of schemas?
- help us organize our memories
- focus our limited resources
- reduce memory load
- make inferences
What was Bransford & Johnson’s work on memory?
- strong evidence that schemas help organize information and improve recall
- passage provided with/without context
- can link details to well-established knowledge in LTM
What was Brewer & Treyens’s work on memory?
- office memory study
- showed how people use schemas to reduce memory load
→ people showed good memory for “office schema” items
→ people “remembered” items that were consistent wiith “office schema” but not present
What was Bower, Black & Turner’s work on memory?
- participants remember script-like passages
→ irrelevant details (32%)
→ script actions (38%)
→ obstacles/distractions (53%) - participants able to direct limited attention and memory to info that is important for the situation based on past experience
What is a “schema-copy-plus-tag” approach?
- memory load is reduced by storing only one general copy of a schema/script
- resources can “tag” information that doesn’t match the schema
- explains why memory is “good” for schema consistent information and why we often include false memories to match our schemas
What is semantic memory?
- permanent memory store of general world knowledge
- conceptual knowledge
- similar between individuals in terms of structure and processes
What brain regions are dependent on semantic memory?
- lateral temporal lobe
- anterior prefrontal cortex
What is a semantic network?
- interrelated set of concepts that provide the structure of semantic memory
What is a node in a semantic network?
- a point or location in the semantic space
- represents a concept
What is a pathway in a semantic network?
- links concept nodes
- are labeled directional associations between concepts
What is spreading activation?
- retrieval of information from the semantic network
What is a proposition in a semantic network?
- a relation between two concepts
What are semantic features?
- simple, one-element properties of the concept
- defining (essential) or characteristic (common but not essential)
What is sentence verification task?
- simple sentences are presented for yes/no decisions
- response time measures are used
- theoretically two concepts closer together should have lower response times
What is the semantic relatedness effect?
- concepts that are more highly related are retrieved faster
What are the advantages of connectionist models?
- similar to the network of neurons in the brain
- units are similar to neurons in the brain that fire/don’t fire
- positive and negative weights between units parallel excitatory/inhibitory neural synapses
- activity of a connectionist model is massively parallel i.e. multiple processes are co-occurring in a model at various levels
What is a category-specific deficit?
- disruption in which person loses access to one semantic category of words or concepts while not losing others
What did Warrington & Shalice suggest was the cause of category-specific deficit?
- selective disruption of sensory knowledge in semantic memory
→ know animals and other living things by sight, smell, sounds
→ know objects by function
What is lexical memory?
- mental lexicon or dictionary where word knowledge (as distinct from conceptual knowledge) is stored
What is anomia/anomic aphasia?
- inability to name semantic concept, a deficit in word finding
What is semantic priming?
- broad activation of concepts in semantic memory
- fundamental consequence of retrieval from semantic memory
What are the basic principles of semantic priming?
→ process takes time
→ activation of primed concepts is smaller the more removed concepts are from the origin
→ effect decays across time
What is a lexical decision task?
- priming task
→ people judge whether a string of letters is a word
What is reconstructive memory?
- construct a memory by combining elements from the original together with existing knowledge
What aspects of Bartlett’s research lead him to believe that memory is reconstructive?
1 - Omissions
→ good recall of main events
→ poor recall of details like specific names or minor events
2 - Normalization/Rationalization
→ successive recalls had people add material that was not in the original story to make them more “normal”
→ removed weirder aspects of the story
What is a schema?
- mental framework or body of knowledge about some topic
What was Sulin & Dooling’s work that shows the distortion of recall?
- participants presented with a passage describing a political figure who was either fictitious or Adolf Hitler
→ when presented with statements they had to decide if they were the same or not in the passage
→ Hitler group agreed more with statements matching previous knowledge about him (“hates Jews”, “obsessed with world domination”)
→ thematic distortion increases over time
What are scripts?
- semantic knowledge that guides understanding of ordered events
What is the theory behind scripts?
- people have generalized memory of experienced events which are invoked/retrieved when a new experience matches an old script
What is a frame in a script?
- details about events within the script
What is a default value in a script?
- common, typical value or concept that occurs on the script
What was Abelson’s work on scripts?
- restaurant scripts were used to show that subsequent events in a story are interpreted with reference to a script which is activated by elements in the story
→ “John went to the restaurant and realized he forgot his reading glasses.” = “He was trying to read a menu.”
What is a prime?
- any item presented first to see whether it influences a later processing
What is the target in a priming task?
- item follows the prime
What are the two results of priming?
- facilitation/benefit: positive influence where the target is easier or faster to process
- inhibition/cost: negative influence on processing where the prime impedes performance to the target
What are two studies that show evidence of the use of scripts?
- Smith & Graesser
→ given passsages and correcting or typical scripts, showed higher recall/recognition for atypical events - Hannigan & Reinitz
→ shown either cause or effect scenes
→ when later shown new “effect” scenes people recalled them even when they weren’t initially present
What is the neuroimaging evidence of probabilistic/exemplar theories?
- rule-based categorization appears to involve brain regions implicated in cognitive control
- similarity-based categorization appears to involve areas implicated in the configural processing of perceptual images
What are ad hoc categories?
- categories that are created based on situational circumstances
→ have characteristics of regular categories
What is psychological essentialism?
- people treat members of a category as if they have the same underlying, perhaps invisible, property or essence
