Presentation is loading. Please wait.

Presentation is loading. Please wait.

Representational Hypotheses in Cognitive Psychology

Similar presentations


Presentation on theme: "Representational Hypotheses in Cognitive Psychology"— Presentation transcript:

1 Representational Hypotheses in Cognitive Psychology
Psychology 355: Cognitive Psychology Instructor: John Miyamoto 05/16/2018: Lecture 08-3 Note: This Powerpoint presentation may contain macros that I wrote to help me create the slides. The macros aren’t needed to view the slides. You can disable or delete the macros without any change to the presentation.

2 Lecture probably ends here
Outline Representational hypotheses - what are they? The centrality of human language capacity in the theory of cognition Semantic network models - a very brief discussion Is there a basic level in human conceptual representations? Neuropsychological representations of categories and concepts Lecture probably ends here Cognitive Theory = { Representational Hypotheses + Processing Hypotheses } Psych 355, Miyamoto, Spr ‘18

3 Representational Hypotheses in Cog Psych
Assumption: Cognitive ability is based on the ability to create, store, retrieve and manipulate a variety of mental representations. Representational Hypotheses Cognitive Theory = Plus Processing Hypotheses Example: Phonological Loop Representational Hypothesis: Speech sounds Processing Hypotheses: Rehearsal, capacity limits, interference Psych 355, Miyamoto, Spr '15 Mental Representations & Their Uses

4 Levels of Representation in Language
Theories of human language capacity propose representational hypotheses at many levels. Articulatory and acoustical phonetics - the study of the muscular coordination in speech and acoustical patterns in the speech signal. Phonology - the study of the sound structures of languages. Syntax - theories of sentence structure and its relation to speech context and meaning. Semantics - theories of the relation between language structure and meaning. Psych 355,, Miyamoto, Spr '18 Language-Like Representations in Cognition

5 Symbolic Representations in Cognition
The motor theory of speech comprehension: A link between speech production and speech comprehension. Models of language syntax Cognitive, computational and/or neuropsych models of language processing with emphasis on sentence structure. Inference in language use - is it a logic? is it a computer program? The modeling of human language capacity is a major concern of cognitive science. Cognitive psychology is part of cognitive science; Other parts include philosophy, linguistics, computer science, neuroscience, anthropology. Psych 355,, Miyamoto, Spr '18 Semantic Networks & Connectionist Models

6 Semantic Networks & Connectionist Models
SKIM Semantic networks & connectionist models – Two different approaches to modeling concept structure. Semantic networks & connectionist models use different formalisms. Spreading activation – activation of some concepts causes activation in associated concepts. Connectionist models focus on learning concept structure. [This topic will be omitted from Psych 355 this quarter.] Semantic networks focus on modeling human concept structure without modeling how we learn this structure. Psych 355 will briefly discuss semantic network models and ignore connectionist models. Psych 355, Miyamoto, Spr '18 Semantic Network Models

7 Semantic Network Models
Concepts are arranged in networks that represent the way concepts are organized in the mind. Any specific network model must explain: what is the relationship between the structure of the network and human thinking or behavior; what are the rules by which representations are constructed within the model; how to derive predictions from the model to different aspects of human behavior. Psych 355, Miyamoto, Spr '18 Example of a Semantic Network

8 Example of a Semantic Network
SKIM Goldstein Figure 9.12 Collins and Quillian (1969) Model for how concepts and properties are associated in the mind Node = category/concept Concepts are linked Cognitive economy: shared properties are only stored at higher-level nodes Exceptions are stored at lower nodes Inheritance Lower-level items share properties of higher-level items Psych 355, Miyamoto, Spr '18

9 Using a Semantic Network to Predict Sentence Verification
Goldstein Figure 9.13 (top) & 9.14 (bottom) Collins & Quillian (1969) Measure response time to verify whether sentence is TRUE or FALSE ("sentence verification" or "semantic memory") Spreading Activation Theory: Activation is the arousal level of a node When a node is activated, activity spreads out along all connected links Concepts that receive activation are primed and more easily accessed from memory SKIM Summary re Semantic Networks Psych 355, Miyamoto, Spr '18

10 Summary re Semantic Network Models
Concepts are represented as nodes in a semantic network. What do semantic network models (SNM's) do? SNM's explain the relationship between the network structure and human reasoning and behavior. SNM's yield predictions for cognitive experiments like semantic memory experiments, categorization judgments and causal inferences. SNM's can be part of a learning model that predict how a learner learns new concepts and conceptual relationships. There are many different SNM's that are devised for different research problems. They tend to be complicated. Psych 355,, Miyamoto, Spr '18 Is There a Basic Level of Categorization?

11 Is There a Basic Level in Categorization?
Categories can often be arranged from higher-level superordinate categories (more inclusive) to lower-level subordinate categories (less inclusive; more specific). Eleanor Rosch asked: Is there a basic level of categorization at which we habitually categorize the objects in our experience? Psych 355, Miyamoto, Spr '18 Basic Level Versus Superordinate Level: Amount of Feature Sharing

12 What Defines the Basic Level of Categorization?
Hypothesis: The basic level is the highest level at which category members share many features within the category while also being very different from objects in other categories at the same level. Two Factors that Determine the Basic Level: Within a basic level category, objects share many features with each other. Between different categories at the basic level, objects are very different from each other. Basic Level Versus Subordinate Level: Amount of Feature Sharing Psych 355, Miyamoto, Spr '18

13 What Defines the Basic Level of Categorization?
Hypothesis: The basic level is the highest level at which category members share many features within the category while also being very different from objects in other categories at the same level. Two Factors that Determine the Basic Level: Within a basic level category, objects share many features with each other. Between different categories at the basic level, objects are very different from each other. Same Slide Without Emphasis Rectangles Psych 355, Miyamoto, Spr '18

14 What Defines the Basic Level of Categorization?
Hypothesis: The basic level is the highest level at which category members share many features within the category while also being very different from objects in other categories at the same level. Two Factors that Determine the Basic Level: Within a basic level category, objects share many features with each other. Between different categories at the basic level, objects are very different from each other. Information Gain at Different Levels of Categorization Psych 355, Miyamoto, Spr '18

15 Information Provided by Categorizations at Different Levels
If you categorize something at the superordinate level, e.g., "furniture," instead of at the basic level, e.g., "table," you lose a lot of information about the object. If you categorize something at a subordinate level, e.g., "kitchen table," instead of at the basic level, e.g., "table," you don't gain a lot of information about the object. Concrete Example with Images of a Bull Do and a Pickup Truck Psych 355, Miyamoto, Spr '18

16 What Is It? Possible Answers Possible Answers An animal
Superordinate level A vehicle Superordinate level A dog Basic level A car Basic level A bulldog Subordinate level A Toyota Camry Subordinate level Psych 355, Miyamoto, Spr '18 Return to Diagram Showing Superordinate, Basic & Subordinate Levels

17 Information Provided by Categorizations at Different Levels
The basic level is the level at which we routinely categorize objects. We are capable of categorizing objects at superordinate or subordinate levels, if required to do so by a situation or task. Psych 355, Miyamoto, Spr '18 Criteria for Basic Level: Feature Listing

18 Behavioral Criteria for Basic Level Categories
SKIM Criterion 1 (Feature-Listing): Ask subjects to list features of objects in a category Category Name = (fill in the blank, e.g., “furniture” or “chair”) Instruction: “List as many characteristics or attributes that you can think of which describe the object.” Examples Categorization: FURNITURE “List as many characteristics or attributes that you can think of which describe furniture.” Categorization: CHAIR “List as many characteristics or attributes that you can think of which describe chairs.” Categorization: ARMCHAIR Superordinate Level Basic Level Subordinate Level The instructions for the feature listing task were taken from Exp 1 of Tanaka & Taylor (1991), cited in the notes to the next slide. Psych 355, Miyamoto, Spr '18 Behavioral Criteria: Basic Level has Greatest Gain in Features

19 Behavioral Criteria for Basic Level Categories (cont.)
SKIM Criterion 1 (Feature-Listing): Ask subjects to list features of objects in a category Category Name = (fill in the blank, e.g., “furniture” or “chair”) Instruction: “List as many characteristics or attributes that you can think of which describe the object.” Characteristic of the basic level: Superordinate level – relatively few features are listed (not many features of “furniture”) Basic level – many features are listed (Many features associated with “chairs”) Subordinate level – may have more features than the basic level but the increase is relatively small. (Features listed for “armchair” are not many more than features listed for “chair”) The instructions for the feature listing task were taken from Exp 1 of Tanaka & Taylor (1991), cited in the notes to the next slide. Psych 355, Miyamoto, Spr '18 Behavioral Criteria: Free Naming

20 Behavioral Criteria for Basic Level Categories (cont.)
SKIM Criterion 2 (Free Naming): Ask subjects to name a picture of an object Instruction: “What would you call this?” [show image of an object] Characteristics of the Basic Level: The basic level is the most commonly used category label. Example Show subject a picture of an armchair. The response “chair” is more likely than the response “furniture”. The response “chair” is more likely than the response “armchair”. The instructions for the feature listing task were taken from Exp 1 of Tanaka & Taylor (1991), cited in the notes to the next slide. Psych 355, Miyamoto, Spr '18 Behavioral Criteria: Category Verification Time

21 Behavioral Criteria for Basic Level Categories
SKIM Criterion 3 (Category Verification): Measure response time for deciding whether an image is an example of a given category Examples 1st subject hears “FURNITURE” Next subject sees picture of an armchair. Correct response = TRUE 1st subject hears “CHAIR” 1st subject hears “ARMCHAIR” Superordinate Level Basic Level Subordinate Level The instructions for the feature listing task were taken from Exp 1 of Tanaka & Taylor (1991), cited in the notes to the next slide. Psych 355, Miyamoto, Spr '18 Behavioral Criteria: Category Verification Time – Basic level Fastest

22 Behavioral Criteria for Basic Level Categories
SKIM Criterion 3 (Category Verification): Measure response time for deciding whether an image is an example of a given category Categorization decisions are fastest at the basic level. 1st subject hears “FURNITURE” Next subject sees picture of an armchair. Correct response = TRUE 1st subject hears “CHAIR” 1st subject hears “ARMCHAIR” Superordinate Level Basic Level Subordinate Level The instructions for the feature listing task were taken from Exp 1 of Tanaka & Taylor (1991), cited in the notes to the next slide. Slower Categorization Response Fastest Categorization Response Slower Categorization Response Psych 355, Miyamoto, Spr '18 Behavioral Criteria: Summary

23 Summary: Behavioral Criteria for Basic Level Categories .
SKIM Criterion 1 (Feature-Listing) Criterion 2 (Free Naming) Criterion 3 (Category Verification) The instructions for the feature listing task were taken from Exp 1 of Tanaka & Taylor (1991), cited in the notes to the next slide. Psych 355, Miyamoto, Spr '18 Expert Versus Non-Experts

24 The Basic Level May Not Be the Same for Experts & Non-Experts
Tanaka and Taylor (1991): Bird experts use specific bird species, e.g., "eagle", "hawk", etc., as if they are the basic level. For the average person, "bird" is basic level, but not for bird experts. Experts Non-Experts Tanaka, J. W., & Taylor, M. (1991). Object categories and expertise: Is the basic level in the eye of the beholder? Cognitive Psychology, 23, basic specific Psych 355, Miyamoto, Spr '18 Same Slide Without Emphasis Rectangles

25 The Basic Level May Not Be the Same for Experts & Non-Experts
Tanaka and Taylor (1991): Bird experts use specific bird species, e.g., "eagle", "hawk", etc., as if they are the basic level. For the average person, "bird" is basic level, but not for bird experts. basic specific Psych 355, Miyamoto, Spr '18 There May Be Cultural Differences in Categorization

26 There can be systematic cultural differences in category knowledge
Medin et al. compared Native American (Menominee Indian) and Euro-American fish experts (fisherman). Euro-American experts tended to sort fish into goal-related categories, e.g., game fish and non-game fish. "Native American fish experts ... tended to sort ecologically and were more likely to see positive and reciprocal ecological relations," e.g., fish that live together or live in predator-prey relations. Itza Maya in Guatemala categorize birds at a lower level than do typical Americans. Basic level for Itza Maya is the subordinate level for Americans. E.g., Itza Maya would say "red hawk" where an American would say "bird." Psych 355, Miyamoto, Spr '18 Summary re Category Structure

27 Summary re Category Structure
The "basic level" of categorization is the level of category structure that is used habitually by members of a given culture. The objects that are grouped together in a basic level category .... share many properties with other objects in the category, i.e., different chairs share many important properties with each other, and .... differ in important ways from objects in other categories at the same level, e.g., chairs differ in important ways from tables, lamps, cars, etc. The basic level can differ between experts and non-experts in a domain, or between members of different cultures. Psych 355, Miyamoto, Spr '18 What Is the Neural Representation of Category Knowledge?

28 Wednesday, 16 May, 2018: The Lecture Ended Here
Psych 355, Miyamoto, Spr '18

29 What Is the Neural Representation of Category Knowledge?
Are there object-specific or person-specific neural representations? Are there specific neurons that are sensitive to specific objects or persons? Are there grandmother cells? How is knowledge of categories of objects represented at the neural level? Psych 355, Miyamoto, Spr '18 Category Specific Neurons

30 Evidence for Category-Specific Neurons
Kreiman, G., Koch, C., & Fried, I. (2000). Single-cell recordings from 11 epilepsy patients awaiting surgery. Neurons found in the temporal lobe that respond best to specific classes of objects. Category-specific neurons for: faces; famous faces; animals; cars; buildings; spatial layouts; abstract patterns Psych 355, Miyamoto, Spr '18 Multimodal Category Representations

31 Multimodal Category Representations
Category knowledge includes knowledge of .... an object's visual appearance, e.g., what a dog looks like; typical sounds, e.g., what kinds of sounds you would expect from a dog; function, e.g., what you would use a hammer for; smell, e.g., what a dog smells like; what a flower smells like; what a fire smells like; taste, especially true of foods typical body actions that relate to it, e.g., we have a representation of the body actions with respect to a hammer, a chair, a chest of drawers Multimodal representations involve multiple sensory modes, vision, audition, touch, kinesthesis, and so forth. Psych 355, Miyamoto, Spr '18 Example: The Grasping Circuit

32 Example of a Multimodal Category Representation: The Grasping Circuit for Manipulable Objects
The Grasping Circuit – a neural circuit associated with use of manipulable objects, like hammers, screwdrivers, tennis rackets, etc. Involves pathways in parietal cortex Question: Are the neural circuits involved in handling a hammer part of the category representation of "hammer"? Psych 355, Miyamoto, Spr '18 fMRI Images While Viewing Hammers, Buildings, Animals & People,

33 Neural Evidence for Multimodal Representations of Category Knowledge
Hammer use activates the left ventral premotor cortex & left posterior parietal cortex (grasping circuit). Subjects are immobile in the scanner so they could not grasp anything. Nevertheless, viewing a hammer activated the grasping circuit. Supports the hypothesis that we activate object-appropriate motor association areas when we access category knowledge. Supports the view that category knowledge is multimodal. Psych 355, Miyamoto, Spr '18 Summary: Neural Representations of Category Knowledge – END

34 Summary – Neural Representations of Category Knowledge
Some neurons may be specific to particular people or objects. (Controversial Issue) Some neural mechanisms are specific to particular kinds of objects. Not necessarily single neurons – the mechanism may involve distributed processing. Neural category representations are multimodal. As yet, we only have a preliminary understanding of neural representations of category knowledge. This is a good field for someone with a scientific pioneering spirit. Psych 355, Miyamoto, Spr '18 END


Download ppt "Representational Hypotheses in Cognitive Psychology"

Similar presentations


Ads by Google