Knowledge
What we know
Concept
Categories of objects, events, and abstract ideas
Categorization
The process by which things are placed into groups called categories
Categories
All possible examples of a particular concept
Conceptual knowledge
Enables us to recognize objects and events and to make inferences about their properties
Definitional approach to categorization
Behavioral approach
Determine category membership based on whether object meets the definition of the category
Does not work well
Not all members of everyday categories have the same defining features
Prototype approach to categorization
Membership is determined by comparing the object to a prototype that represents the category
Items in a category that have a large amount of overlap have high family resemblance
Prototype
An average representation of the “typical” member of a category
Rosch’s typicality experiment (1975)
Question: Are all members of a category created equal?
Method: Participants rated the extent to which each member represented the category title on a scale of 1 (good) to 7 (poor)
Results: There are various degrees of “typicality”, where high typicality means that a category member closely resembles the category prototype, while a low typicality member does not
Family resemblance
Common characteristics between items in a category
Priming effect
Prototypical objects are affected more by priming
Mervis et al. experiment (1976)
Question: Testing the priming effect
Method: Asked participants to list as many objects in a category as possible
Results: Prototypical objects are named first
Prototypical objects
Members of a category with high family resemblance that are affected more by priming and are named first when people list members of the category
Exemplar approach
Behavioral approach
Concept is represented by multiple examples rather than a single prototype
Examples are actual category members instead of abstract averages
Hierarchical organization of categories
Global (Superordinate) → Basic → Specific (Subordinate)
Going above the basic level = large loss of information
Going below the basic level = little gain of information
Tanaka and Taylor’s experiment (1991)
Question: How does knowledge affect categorization?
Method: Participants used categories to name pictures of birds
Results: Experts used more specific categories, while nonexperts used more basic categories
Semantic networks (Collins and Quillian)
Network approach
Concepts are linked and arranged in networks that represent the way concepts are organized in the mind (node = category/concept)
Not meant to mirror physiology
Cognitive economy
Shared properties are only stored at higher-level nodes, while exceptions are stored at lower nodes
Spreading activation
When a node is activated, activity spreads out along all connected links
Concepts that receive activation are primed and more easily accessed from memory
Connectionist approach
Network approach
Creating computer models for representing cognitive processes (parallel distributed processing)
Knowledge represented in the distributed activity of many units
Weights determine at each connection how strongly an incoming signal will activate next unit
Graceful degradation
Performance disruption occurs gradually as parts of the system are damaged
Network operation is not totally disrupted by damage
Sensory-functional hypothesis
Physiological approach
Different brain areas may be specialized to process information about different categories
Double dissociation for categories “living things” and “nonliving things” (artifacts)
Living things → sensory properties, artifacts → functions
Multiple-factor approach
Physiological approach
There are more factors than just sensory and functional that determine how concepts are divided within a category
Hoffman & Lambon Ralph experiment (2013)
Question: What factors determine how concepts are divided within a category?
Method: Ask participants about 160 different objects and how much they associate an object with color, motion, visual form, sound, etc.
Results: People associated animals (living things) with color and motion but not with performed action, and associated artifacts (nonliving things) with mostly performed action rather than motion and color
Semantic category approach
Physiological approach
There are specific neural circuits in the brain for some specific categories, with some being more innately determined because they are important for survival (such as the fusiform face area (FFA))
Embodied approach
Physiological approach
Knowledge of concepts is based on reactivation of sensory and motor processes that occur when we interact with the object
Thinking about a concept activates perceptual and motor areas associated with that concept
Mirror neurons
Neurons that fire when we do a task or we observe another person doing that same task
Semantic somatotopy
Correspondence between words related to specific body parts and location of brain activation
Language
System of communication using sounds or symbols that enables us to express feelings, thoughts, ideas, and experiences
Human language structure
Hierarchical: small components can be combined to form larger units (words → phrases → sentences)
Governed by rules: components can be arranged in specific ways
What did the “Verbal Behavior” experiment (1957) by B.F. Skinner show?
Language is learned through reinforcement
What did the “Syntactic Structures” experiment (1957) by Noam Chomsky show?
Human language is coded in the genes, as children produce sentences they have never heard and that have never been reinforced
Underlying basis of all language is similar
Psycholinguistics
Discover psychological process by which humans acquire and process language
Focused on 4 key areas: comprehension, representation, speech production, and acquisition
Lexicon
All words that a person understands (mental dictionary)
Semantics
The meaning of language
Lexical semantics
The meaning of words
Word frequency effect experiment 1
Question: How do we determine the meaning of words?
Method: Lexical decision task - decide as quickly as possible whether strings of letters are words or nonwords
Results: We respond faster to high-frequency words
Rayner and Duffy experiment (1986)
Question: How do we determine the meaning of words?
Method: Measured fixation and gaze times of participants reading different sentences with high/low frequency words (The slow music vs. the slow waltz)
Results: Low-frequency words got longer fixation and gaze times than high-frequency words
Speech segmentation
Perceiving individual words in normal conversation even though there are no silences between spoken words
Lexical ambiguity
Words have multiple meanings, where context usually clears up ambiguity
Tanenhaus experiment (1979)
Question: How do we understand ambiguous words?
Method: Participants read the probe word out loud as quickly as possible with a context word or without a context word, measured response times
Results: Without a context word, we briefly access multiple meanings of ambiguous words before the context takes over, leading to reaction time being slightly slower
Meaning dominance
Some words are used more frequently than others
Biased dominance
When words have two or more meanings with different dominance (tin means metal and container)
Balanced dominance
When words have two or more meanings with about the same dominance (cast means play and plaster)
Meaning dominance experiment
Question: How does meaning dominance work?
Method: Measured gaze time of participants for words with multiple meanings vs words with one meaning in various sentences
Results: For balanced dominance, people look at the word with more meanings longer than a word with one meaning. However, for biased dominance, people looked at the word with more meanings the same amount of time as the word with one meaning when the more frequent meaning of the word was used. When the less frequent meaning of the word was used, they looked at the word with more meanings longer.
Parsing
Mentally grouping words into phrases to help the listener create meaning
Syntax
Rules for combining words into sentences
Garden path sentences
Sentences that begin by appearing to mean one thing, but then end up meaning something else
Temporary ambiguity
When the initial words are ambiguous, but the meaning is made clear by the end of the sentence
Garden path model of parsing (Lynn Frazier)
Listeners use heuristics to group words into phrases
Fast, but could be incorrect
Utilizes late closure
Late closure
Parser assumes new word is part of the current phrase (closes the phrase as late as possible)
Constraint-based approach to parsing
Parsing is influenced by many factors other than syntax, such as word meaning, memory load, story context, scene context, and prior experience with language
Tanenhaus & Trueswell experiment (1995)
Question: Studying constraint-based approach to parsing
Method: Participants looked at a display and were told to carry out an ambiguous set of instructions (place the apple on the towel in the box) and unambiguous instructions (move the apple that is on the towel to the box) and their eye movements were measured
Results: When there was one apple, participants’ eyes moved toward the incorrect destination ~50% of the time with ambiguous instructions, and almost never with the unambiguous instructions. When there were two apples, participants’ eyes moved toward the incorrect destination ~20% of the time with ambiguous instructions, and ~10% of the time with unambiguous instructions.
Visual world paradigm
How scene context influences how a sentence is processed/parsed
Finding (scene context)
Eye movements change when information suggests revision of interpretation of sentence is necessary
Inference (scene context)
Linguistic and nonlinguistic information is used simultaneously to process a sentence
Coherence
Representation of the text in one’s mind creates clear relations between parts of the text and between parts of the text and the story’s main topic
Anaphoric inference
Creating information during reading not explicitly stated in the text by connecting objects/people
Instrument inference
Creating information during reading not explicitly stated in the text by connecting tools or methods
Causal inference
Creating information during reading not explicitly stated in the text by connecting how events in the previous sentence caused the events in the next clause
Situation model
Mental representation of what the text is about that represents events as if experiencing the situation with the point of view of the protagonist
Stanfield & Zwaan experiment (2001)
Question: How do people understand stories?
Method: Participants would read a sentence that describes a situation (hammering a nail into a wall vs. hammering a nail into the floor) and then indicate whether a picture shows the object that was mentioned in the sentence
Results: Participants would say yes to a nail regardless of orientation, but the reaction time for a nail which “matches” the orientation it would be hammered in (sideways for wall vs. downwards for floor) would be faster than for a nail which doesn’t match
Given-new contract
Speaker constructs sentences so that they include given information and new information
Common ground
Shared mental knowledge and beliefs
Entrainment
Synchronization between conversation partners (similar gestures, speaking rating, pronunciation, etc.)
Theory of mind
Being able to understand what others feel, think, or believe
Nonverbal communication
Being able to interpret or react to the person’s gestures, facial expressions, tones of voice, and other cues to meaning
Findings of faster reaction time to statements such as “A robin is a bird” than statements
such as “A ostrich is a bird” are examples of ____ effect
typicality
According to Collins and Quillian's semantic network model, reaction time to verify "a
canary is a bird" is ___________ than the reaction time to verify "an ostrich is a bird.”
the same
What are the pros and cons of Collins & Quillian’s model?
Pros: It is a useful model for representing the connections between concepts and properties and how other concepts and properties are linked to one another
Cons: Cannot explain typicality effects (people react more rapidly to high prototypical objects)
What is one important feature of the connectionist network model?
Concepts are represented by activity that is distributed across a network (parallel distributed processing)
What is special about “basic” categories?
Large loss of information when going from basic to global/superordinate level
Little gain of information when going from basic to specific/subordinate level
According to which physiological approach to categorization are there certain types of concepts that have specific neural circuits in the brain?
Semantic category approach
What are 3 different kinds of inferences when reading narratives?
Anaphoric, instrument, and causal
Human language make it possible to create new and unique sentences because
it is (1) _____ and (2) _______.
(1) Hierarchical and (2) governed by rules
Words can have ambiguous meanings. _________ describes the relative
frequency of the meanings of ambiguous words, i.e., biased vs. balanced
Dominance
Lexical priming paradigm involves assessing meaning of words in people’ mind.
The presence of a lexical priming effect can be used to tell us what?
It shows that words that are semantically related are more easily processed compared to unrelated words and that the activation of one word can spread to related words, suggesting that the semantic network model (Collins and Quillian) is accurate
_____ helps listeners to create meaning from groups of words, but ______ ________ still occurs
Parsing; temporary ambiguity
What are the two models to solve temporary ambiguity?
Garden path model of parsing and constraint-based approach to parsing
What is the major difference between the garden path model of parsing and the constraint-based approach to parsing?
Syntax; the garden path model of parsing argues that listeners use syntax to perform parsing whereas the constraint-based approach to parsing argues that parsing is also influenced by many other factors rather than just syntax
Based on behavioral studies, what has been the alternative approach to the prototype approach to categorization?
Exemplar approach
Which approach works better (prototype vs exemplar)?
Both approaches are used, but prototype approach is used first when taking an average, and later on exceptions are taken into account with the exemplar approach
What are the four physiological approaches to categorization?
Sensory-functional hypothesis, multiple-factor approach, semantic category approach, and embodied approach
What is the difference between the sensory-functional hypothesis and multiple-factor approach?
The multiple-factor approach has more factors than just sensory and functional to determine how concepts are divided into categories
What is the one thing that all of the physiological approaches agree on about the representation of categories in the brain?
Information about concepts is distributed across many structures in the brain
Emily is 2 years old and is just learning about the category “dog”. So far, she has experienced only two dogs, one a poodle and the other a German shepherd.
On her third encounter with a dog, she will likely categorize that object as a dog if -
a. it matches the size of the poodle but is of a different breed
b. it is a breed of dog that is hairless and teacup-sized
c. it matches an exemplar of one of the dogs she has experienced
d. it is similar to an “average” for the dogs she has encountered
a, c, and d are correct
Which sentence is easier to understand due to memory load & pior experience with language?
1: "The senator who spotted the reporter shouted."
2: "The senator who the reporter spotted shouted."
1, because it imposes lower memory load and subject-relative construction is more prevalent in English
What are the 3 main types of behavioral approaches to categorization?
Definitional, prototype, and exemplar
What are the 3 main approaches to categorization?
Behavioral, network, and physiological
Rosch’s prototype experiment (1975)
Question: Are all members of a category created equal?
Method: Participants heard the name of a color and thought of the prototype of the color, and then looked at a color and were asked to say “same” if the name of the color matched the color they were looking at
Results: Prototypical objects are affected more by priming, as the prototypical example for “green” had the higher reaction speed for saying “same” compared to a light green
Transcranial magnetic stimulation
Noninvasive form of brain stimulation in which a changing magnetic field is used to induce an electric current at a specific area of the brain through electromagnetic induction