Download presentation
Presentation is loading. Please wait.
Published byAshlie Reed Modified over 9 years ago
1
Cognition – 2/e Dr. Daniel B. Willingham Chapter 7: Memory Storage ©2004 Prentice Hall PowerPoint by Glenn E. Meyer, Trinity University
2
2 What is in the Storehouse? Important Definitions: Important Definitions: Categories: A group of objects that have something in common Exemplar: An instance of a category Generalize: Usually applied to categories, it means to use information gathered from one exemplar to a different exemplar of the same category. For example, if you learn that a specific dog likes to have its stomach rubbed, you may generalize that knowledge to other dogs and assume that they too like to have their stomachs rubbed The Classic View of Categorization The Classic View of Categorization The Classic View of Categorization The Classic View of Categorization The Probabilistic View of Categorization The Probabilistic View of Categorization The Probabilistic View of Categorization The Probabilistic View of Categorization ©2004 Prentice Hall
3
3 The Classic View of Categorization The view that concepts are represented as lists of necessary and sufficient properties - Aristotle : The view that concepts are represented as lists of necessary and sufficient properties - Aristotle : Concept: The mental representation that allows one to generalize about objects in a category Important Studies o Clark (1920) as seen in Fig. 7.1 o Bruner, et al. (1956) Strategies for determining category membership Focus gambling: When trying to learn a new category, using the strategy of generating a narrow hypothesis about the necessary and sufficient properties that define a category Focus gambling: When trying to learn a new category, using the strategy of generating a narrow hypothesis about the necessary and sufficient properties that define a category Successive Scanning: When trying to learn a new category, using the strategy of formulating a hypothesis and making selections based on that hypothesis until it is disconfirmed Successive Scanning: When trying to learn a new category, using the strategy of formulating a hypothesis and making selections based on that hypothesis until it is disconfirmed o Wittgenstein (1953) - objects to classical view - for many concepts hard to list necessary and sufficient properties ©2004 Prentice Hall
4
4 The Classic View of Categorization - Continued Objections to the Classic View Objections to the Classic View Typicality Effects: Some members of category are viewed as better exemplars than others (e.g., a golden retriever typical dog, Chihuahua not) o Important Studies Rosch (1971) as seen in Table 7.1 - gradations of category membership argue against classical view Smith, et al. (1974) - More efficient categorizing typical vs.. atypical examples Most frequently generated category examples are rated as most typical - Battig and Montague (1969), Nelson (1974), Rosner and Hayes (1977) as seen in Table 7.2 Ripps (1975) - atypical features not extended to typical category members. Not true for experts - several studies Category Hierarchies - Categories nested with one another o Rosch, et al. (1976) - Three types of category structure as seen in Table 7.3 Basic level: Category that is the most inclusive of which members still share most of their features Superordinate level: Defined as a category that is one level more inclusive than the basic level category Subordinate categories Defined as a category that is one level less inclusive than the basic level category o Definitions of basic categories depend on expertise - Tanaka and Taylor (1991). Johnson and Mervyns (1997) as seen in Table 7.4 - Novices use basic level, intermediate experts - basic and subordinate, advanced experts - basic, subordinate, sub-subordinate levels. Basic level has a privileged status ©2004 Prentice Hall
5
5 The Probabilistic View of Categorization Category membership is proposed to be a matter of probability. Prototype & exemplar models fall within the probabilistic view: Category membership is proposed to be a matter of probability. Prototype & exemplar models fall within the probabilistic view: Prototypes: A prototype has all of the features that are characteristic of a category o Posner and Keele (1968) as seen in Fig. 7.2 - subjects correctly picked never seen concept prototype after viewing many exemplars o Suggests people abstract central features and store prototype Exemplar Models: Model of categorization that maintains that all exemplars are stored in memory, and categorization judgments are made by judging the similarity of the new exemplar to all the old exemplars of a category o Medin and Schaffer (1978) - contrary to Posner and Keele - abstraction could take place at retrieval and only if prototype needed as seen in Fig. 7.4 o Research supports Exemplar over prototype models (Estes, 1994; Smith& Mimda, 2000) ©2004 Prentice Hall
6
6 The Probabilistic View - Continued Problems with Similarity Models Problems with Similarity Models How to determine which features used in similarity judgements ? Similarity depends on context - Tversky (1977) as seen in Fig. 7.5 Categorization can also use rules o Imaging Studies: Smith, et al (1998) - categorization task using rules vs.. memory task. Activation measured as seen in Box 7.1 Rules - secondary visual areas, superior parietal lobe, premotor cortex, dorsolateral frontal cortex Memorization - occipital cortex and right cerebellum Suggests different loci for different categorization strategies o Ripps (1989) - demonstrated rule based categorization in seminal study o Allen and Brooks (1991) - found rules or similarity strategies can be used dependent on instruction ©2004 Prentice Hall
7
7 How Is Memory Organized? Addressing Systems Addressing Systems Addressing Systems Addressing Systems Content-Addressable Storage Content-Addressable Storage Content-Addressable Storage Content-Addressable Storage Hierarchical Theory Hierarchical Theory Hierarchical Theory Hierarchical Theory Spreading Activation Theories Spreading Activation Theories Spreading Activation Theories Spreading Activation Theories Spreading Activation Models: An Example Spreading Activation Models: An Example Spreading Activation Models: An Example Spreading Activation Models: An Example Evidence of Activation Evidence of Activation Evidence of Activation Evidence of Activation Criticisms of Spreading Activation Criticisms of Spreading Activation Criticisms of Spreading Activation Criticisms of Spreading Activation Distributed Representations (Parallel Distributed Processing) Distributed Representations (Parallel Distributed Processing) Distributed Representations (Parallel Distributed Processing) Distributed Representations (Parallel Distributed Processing) Criticisms of Parallel Distributed Processing Models Criticisms of Parallel Distributed Processing Models Criticisms of Parallel Distributed Processing Models Criticisms of Parallel Distributed Processing Models ©2004 Prentice Hall
8
8 Addressing Systems Scheme to organize memories in which each memory is given a unique address that can be used to look it up Scheme to organize memories in which each memory is given a unique address that can be used to look it up ©2004 Prentice Hall Content- Addressable Storage Scheme by which to organize memories in which the content of the memory itself serves as the storage address Scheme by which to organize memories in which the content of the memory itself serves as the storage address
9
9 Hierarchical Theory Theory of memory organization in which concepts are organized in a taxonomic hierarchy (e.g., animal is above bird, which is above canary) and characteristic properties are stored at each level - Collins and Quillian (1969, 1972) Theory of memory organization in which concepts are organized in a taxonomic hierarchy (e.g., animal is above bird, which is above canary) and characteristic properties are stored at each level - Collins and Quillian (1969, 1972) Important Terms as seen in Fig 7.6: o Nodes and Activation: Nodes are the epresentation of concepts in hierarchical and spreading activation theories. Activation is the level of energy or excitement of a node, indicating that the concept the node represents is more accessible for use by the cognitive system o Links: Representation of the relationship between concepts. In the hierarchical model the links are labeled (e.g., “has this property”), whereas in spreading activation models the links simply pass activation from one node to another o Property Inheritance: A characteristic of some models of categorization; concepts inherit properties from the concepts that are higher in the hierarchy ©2004 Prentice Hall
10
10 Hierarchical Theory - Continued Experimental Evidence: o Collins and Quillian (1969) - response time for categorization seems to support movement along the hierarchy o Problems Hierarchy didn’t always hold - as “ chicken is a animal” is faster to verify than “chicken is bird” Cognitive economy The principle of designing a cognitive system in a way that conserves resources (e.g., memory storage space - doesn’t seem to be true in subject produced hierarchies - Conrad (1972) ©2004 Prentice Hall
11
11 Spreading Activation Theories Spreading activation model - Collins and Loftus (1975) Spreading activation model - Collins and Loftus (1975) Memory is conceived of as a network of nodes connected by links, & activation spreads from node to node via links as seen in Fig. 7.7 Semantic Network - composes memory and name given to all the nodes and links in a spreading activation mode. Important properties are (Rumelhart, Hinton and McClelland, 1986) o Set of Units o State of Activation o Output Function o Pattern of Connectivity o Activation Rule o Learning Rule to Change Weights Example as seen in Fig. 7.8 based on Table 7.5 by McClelland (1981) has useful characteristics: o Retrieval of Properties o Mode is content addressable o Typicality grows naturally out of the model o Model creates default values o Resistant to faulty inputs ©2004 Prentice Hall
12
12 Spreading Activation Models - Continued Examples: Examples: Repetition priming as seen in Table 7.6: Effect in which performance of a task is biased by ones having seen the same words or pictures sometime earlier. Indicates activation of nodes is long lasting and measurable Semantic priming: Effect in which performance of a task is biased by having seen semantically related words or pictures viewed earlier. Indicates activation passes between nodes. Criticisms of Spreading Activation Criticisms of Spreading Activation Nonspecificity of how to determine location and strength of links leading to circularity How to determine how far activation spreads o Mediation priming (McNamara and Healy, 1988): Priming that goes through two links, not just one. For example, if lion primes stripes it is probably because the priming was mediated through tiger (i.e., lion primes tiger, which primes stripes. o This effect predicts too much activation through many branches and makes the spreading activation concept pointless ( Ratliff and McCoon, 1994) ©2004 Prentice Hall
13
13 Distributed Representation (Parallel Distributed Processing) Local representation :A representational scheme in which a concept has a single location (e.g., it is represented in one node in a semantic network) Local representation :A representational scheme in which a concept has a single location (e.g., it is represented in one node in a semantic network) Distributed representation :A representational scheme in which a concept is distributed across multiple units. Distributed representation :A representational scheme in which a concept is distributed across multiple units. Parallel Distributed Processing: A model using a distributed representation with nodes and links as seen in Fig. 7.10. The model learns as weights are modified. The model has several advantages Parallel Distributed Processing: A model using a distributed representation with nodes and links as seen in Fig. 7.10. The model learns as weights are modified. The model has several advantages Graceful degradation A property of a model (of memory or of another cognitive process) whereby if the model is partially damaged it is able to continue functioning, although not as accurately. The human brain often shows graceful degradation; if it is damaged, cognitive processes often are compromised, but can still partially function Learning by model is convincing. Uses modification of weights of the links Automatically finds both prototypes and exceptions to prototypes as seen in Fig. 7.7 Generalization is a natural outgrowth of this model ©2004 Prentice Hall
14
14 Criticisms of Parallel Distributed Processing Many models suffer from catastrophic interference. Learning sequences for models to avoid interference (interleaved presentation of items) is not characteristic of human learning (McCloskey and Cohen, 1989) Many models suffer from catastrophic interference. Learning sequences for models to avoid interference (interleaved presentation of items) is not characteristic of human learning (McCloskey and Cohen, 1989) Pinker and Prince (1988) suggest that rule based processes in children’s language learning, such as overregularization errors, are not handled by PDP models Pinker and Prince (1988) suggest that rule based processes in children’s language learning, such as overregularization errors, are not handled by PDP models ©2004 Prentice Hall How are Concepts Organized in the Brain? Lesions can produce specific category loss, e..g. naming living things Caramazza and Shelton (1998) Lesions can produce specific category loss, e..g. naming living things Caramazza and Shelton (1998) Suggest specific loci for various categories Suggest specific loci for various categories Neuroimaging data do not agree Neuroimaging data do not agree Different aspects of an object are represented in different loci (Martin, et al., 1995) Different aspects of an object are represented in different loci (Martin, et al., 1995) May be some localization of faces, places and parts of the body but there is overlap May be some localization of faces, places and parts of the body but there is overlap Distributed representations are probably the best bet with current knowledge Distributed representations are probably the best bet with current knowledge
15
15 What Else Is In Memory? What Are Separate Memory Systems? What Are Separate Memory Systems? What Are Separate Memory Systems? What Are Separate Memory Systems? Procedural and Declarative Memory Procedural and Declarative Memory Procedural and Declarative Memory Procedural and Declarative Memory Cognitive Differences Among Memory Systems Cognitive Differences Among Memory Systems Cognitive Differences Among Memory Systems Cognitive Differences Among Memory Systems ©2004 Prentice Hall
16
16 What Are Separate Memory Systems? Experimental research and study of amensics suggest five forms of memory that have different anatomical bases. Experimental research and study of amensics suggest five forms of memory that have different anatomical bases. Analogous to the different data formats used by computer programs Analogous to the different data formats used by computer programs Types suggested (as seen in Table 7.8) and Neural Substrate ( ) Types suggested (as seen in Table 7.8) and Neural Substrate ( ) Declarative Memory o Conscious memory of facts and events (Hippocampus and other structures) Priming o Brief activation of existing representation (Occipital, temporal, frontal cortex) Motor Skill Learning o Acquires new motor skills (Striatum, motor cortical areas) Classical Conditioning o Relationships between perceptual stimuli and motor responses (Cerebellum) Emotional Conditioning o Learns relationship between perceptual stimuli and emotional responses (Amygdala) ©2004 Prentice Hall
17
17 Procedural and Declarative Memory Definitions: Definitions: Declarative memory: Memory for facts and events, often called “knowing that” memory Procedural memory :Memory for skills, often called “knowing how” memory Anatomic dissociation: Evidence that two different tasks are supported by different parts of the brain Origin Origin HM - profound anterograde amnesia with difficulty storing new memories. Damage in medial temporal lobe Still could learn motor tasks (Corkin, 1968) arguing for anatomic dissociation Absent Processes in Amensics Absent Processes in Amensics Warrington and Weiskrantz (1968) - problems with retrieval as showed repetitions on obscured words in recognition task did help Studies indicated encoding not a problem - Mayes, et al. (1980) showed deep coding helped amensics Amensics don’t show faster forgetting (Huppert and Pierccy, 1978) ©2004 Prentice Hall
18
18 Cognitive Differences Among Memory Systems Most reliable difference is awareness – Declarative is associated with awareness, other forms of learning not necessarily aware Most reliable difference is awareness – Declarative is associated with awareness, other forms of learning not necessarily aware Declarative learning may be faster and require only one trial, unlike other types that may require many trials Declarative learning may be faster and require only one trial, unlike other types that may require many trials Episodic vs. Semantic Memories suggested by Tulving (1972) Episodic vs. Semantic Memories suggested by Tulving (1972) Episodic : Memory that is associated with a particular time and place, with a this happened-to-me feeling. More prone to forgetting. Semantic: Memories that are not associated with a particular time and place or with a feeling that the memory happened to you. Semantic memories cover world knowledge (e.g., “frogs are green”). Recalled more quickly Some supporting evidence for the distinction from neuroscience o Cases found with seemingly intact semantic memories but damaged episodic memory – KC (Tulving, et al.,1968) o Vargha-Khadem, et al., (1997, 2002) – hippocampal damage causes episodic problems with little semantic memory difficulties o Neuroimaging data not consistent as to loci of separate semantic or episodic processes Problems with Episodic vs. Semantic Distinction o Line is not clear when a memory stops being episodic and becomes semantic o Most episodic memories have semantic elements embedded in them so the separation is not clear with a particular memory ©2004 Prentice Hall
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.