Knowledge Pt 2 Chapter 10 Knowledge Pt 2

Organizing Our Concepts Semantic memory is organized into a series of hierarchical networks.

Chapter 7 General Knowledge 3 Hierarchical Network Model (Collins & Quillian, 1969) Major concepts are represented as nodes. Properties/features are associated with each concept. Examples: Nodes: Animal; Bird; Fish; Canary Features: Has wings; Is dangerous Chapter 7 General Knowledge

Cognitive economy rules: Property information is stored as high up the hierarchy as possible to minimize redundancy. If we learn something new about a category (e.g. Fish lay eggs) we add that information to the highest level to which it applies (fish) and this information generalizes to all levels below (e.g., Salmon lay eggs). If we lean about a new type of fish (e.g., a pike) we could guess that they too lay eggs.

Chapter 7 General Knowledge 5 Superordinate Level Highest level. - all properties stored at this level apply to all linked levels below. Transitive Inheritance– All properties apply to the basic level and in turn to the subordinate level Chapter 7 General Knowledge

Chapter 7 General Knowledge Name this Item Chapter 7 General Knowledge

Typical level of communication. 7 Basic Level Typical level of communication. When asked to name pictures, people use the basic level names (e.g., hammer not ball peen hammer) Share the most common features Most general level that people use similar motions. Parents use these terms when speaking to children First words acquired by children Chapter 10 Pt 2

Basic Level Continued Allow for faster level of categorization More informative and more distinctive from other categories at the basic level. Superordinate are informative but not very distinctive Subordinate categories are distinctive but not as informative.

Chapter 7 General Knowledge 9 Subordinate Level Very similar attributes named for different categories at this level (e.g., color, distinctive features, unique characteristics). Faces are generally specified at this level (Cori) rather than at the superordinate (human) or basic level (woman/prof). Chapter 7 General Knowledge

Expertise (pg. 259) With expertise comes a restructuring of categories. Basic level shifts to the subordinate level (e.g., would call the hammer on slide 7 a ball peen hammer).

Links Are labeled as either “is a” links to define that it is a member of a higher level; “has a” or “can” links to define properties of the concept. Exceptions to higher level properties are stored at the subordinate level (e.g., can’t fly).

Hierarchical Network Model (Sentence Verification Task) 12 Hierarchical Network Model (Sentence Verification Task) Hypothesis: Sentences should take more time to process as the number of levels between the tested concept and the features increases. Task: Participants had to quickly verify the veracity (truth) of statements. Examples: Fast: “A canary is yellow” Slow: “A canary can fly” Chapter 7 General Knowledge

Chapter 7 General Knowledge 13 Hierarchical Network Model (Sentence Verification Task) Results: Responses to true statements become slower as the separation between the subject and property becomes greater. Chapter 7 General Knowledge

Chapter 7 General Knowledge 14 Problems with the hierarchical network model Typicality effect: Verification is faster for more representative (Typical) member categories, independent of hierarchical distance (Rips, Shoben, & Smith, 1973). e.g. “A PENGUIN is a bird” is slower to confirm than “A CANARY is a bird”. Typical items have more commonalities with other items in their category than atypical items. Categories are fuzzier than Collins and Quillian believed. Chapter 7 General Knowledge

Transitivity does not always work (Hampton, 1982) People agree than chairs are furniture, and that car seats are a type of chair, but don’t verify that a car seat is a type of furniture.

Where are conceptual networks (semantic memory) in our nervous system? Two Views Distributed only view: In the connections (synapses) between neurons. Features are distributed across multiple regions of the cortex (color, shape, motion, language, how to interact with). Distributed-plus-hub view: Agrees with 1) but adds a convergence zone or hub that binds the features together.

Evidence from Semantic Dementia Symptoms: Progressive anomia (impaired: naming, word comprehension, object recognition and understanding of concepts. Patients with other causes of anomia, and occasionally even healthy people, sometimes fail to name something because, although they know what it is, they cannot find the word for it at that moment. Anomia in SD does not result from this kind of word-finding difficulty but instead reflects degraded knowledge of the object or concept. Characterized by preserved fluency (phonologically and syntactically correct) and impaired language comprehension. Game Anomia

Example "The Case of Anna H." "When I showed her some kitchen matches, she recognized them at once, visually, but could not say the word 'match,' saying instead, 'That is to make fire.'“ (Oliver Sacks)

Progression of Symptoms Begins with semantic-type naming errors: initially within- category (subordinate level), “elephant” for hippopotamus, Progresses to basic level “dog” for everything, and then to superordinate level “animal”… Progressing to profound and complete anomia

Degree of success in naming (and, indeed, in any measure of conceptual knowledge) is largely independent of the modalities of input and output. Naming is most commonly assessed by presenting pictures and asking for a verbal naming response; however, SD patients are just as unsuccessful, if the stimulus is a description of the item to be named (for example, “What do we call the African animal with black and white stripes?”), or is the item’s characteristic sound (for example, a telephone ringing or a dog barking), or if the patient is asked to write the name rather than say it. (Patterson, Nestor & Rogers, 2007)

Evidence from Semantic Dementia Associated with deterioration of left temporal lobe without damage to the hippocampus. Gives support for the Distributed- plus-hub view.

Using Concepts: How does understanding the structure of concepts help us understand human Thinking? Rips (1975): Participants imagine an island with a small set of bird or mammal species. Told that a particular species had a contagious disease. Asked to estimate the percentage of each other species that would share the disease (Inductive task, e.g., what other animals do I include in this category). Estimates were significantly higher when the given instance was a typical category member.

For example, when told that sparrows (i. e For example, when told that sparrows (i.e., typical bird) had the disease, people estimated that 32% of geese also had the disease. In contrast, when told that eagles (i.e., atypical bird) had the disease they judged that only 17% of geese would have the disease. Sparrows and eagles are approximately equal in their similarity to geese, but sparrows had greater inductive strength because they are considered better examples of birds overall.

Stereotypes ~ the good side! Stereotypes are often correct! and cognitively efficient. What would you buy a 5 year old boy for his birthday? How would you speak to a State Trooper if pulled over on the highway?