Cognitive Psychology Lecture 9: Conceptual Knowledge 2023 PDF

PSYC 5140 Cognitive Psychology Lecture 9: Conceptual Knowledge 2023 Instructor: Urs Maurer Knowledge • Conceptual knowledge: knowledge that enables us to recognize objects and events and to make inferences about their properties • Concept: mental representation used for a variety of cognitive f...

PSYC 5140 Cognitive Psychology Lecture 9: Conceptual Knowledge 2023 Instructor: Urs Maurer Knowledge • Conceptual knowledge: knowledge that enables us to recognize objects and events and to make inferences about their properties • Concept: mental representation used for a variety of cognitive functions – Meaning of objects, events, and abstract ideas – Mental representation of a class or individual • Categorization is the process by which things are placed into groups called categories – A category includes all possible examples of a particular concept – Our knowledge of the world is organized in categories Why Categories Are Useful Category “cat” includes: Siamese cats, Persian cats, wild cats, … Concepts provide the rules for creating categories (A cat is a furry animal that meows, …) ➔ The mental representation for “cat” would affect what animals we place in the “cat” category What is a cat? Why Categories Are Useful • Help to understand individual cases not previously encountered • “Pointers to knowledge” – Categories provide a wealth of general information about an item – Allow us to identify the special characteristics of a particular item • If there were no such things as categories, we would have a very hard time dealing with the world Structure of Lecture • Basic properties of concepts and categories • Network models of categorization • How concepts are represented in the brain Definitional Approach to Categorization • Determine category membership based on whether the object meets the definition of the category • Does not work well – Works well for some things, such as geometric objects – But not for natural objects and human-made objects • Not all members of everyday categories have the same defining features Family resemblance • Family resemblance – Wittgenstein proposed the idea of family resemblance to deal with the problem that definitions often do not include all members of a category – Things in a category resemble one another in a number of ways – Allows for some variation within a category ➔ Categorization may be based on determining how similar an object is to some standard representation of a category The Prototype Approach • Prototype = “typical” – An average representation of the members of a category that are commonly encountered – Characteristic features that describe what members of that concept are like – An average of category members encountered in the past • Prototype approach to categorization: membership in a category is determined by comparing the object to a prototype that represents the category real birds prototype The Prototype Approach The Prototype Approach Variation in Typicality • Rate the following example as to whether how well they represent the category title Category: «Bird» Variation in Typicality Owl Variation in Typicality Sparrow Variation in Typicality Penguin Variation in Typicality Bat The Prototype Approach • Not all birds are like robins, blue jays, or sparrows. • Owls and penguins are also birds • These variations represent differences in typicality • Rosch (1975): S saw a category title (e.g. birds), and a list of about 50 members of the category • S were asked to rate the extent to which each member represented the category title The Prototype Approach • High-prototypicality: category member closely resembles category prototype – “Typical” member – For category “bird” = robin • Low-prototypicality: category member does not closely resemble category prototype – For category “bird” = penguin The Prototype Approach • How well do good and poor examples of a category compare to other items within the category? • Strong positive relationship between prototypicality and family resemblance – Measuring family resemblance: list as many characteristics and attributes as possible for category members • When items have a large amount of overlap with characteristics of other items in the category, the family resemblance of these items is high • Low overlap = low family resemblance The Prototype Approach • Typicality effect: prototypical objects are processed preferentially – Highly prototypical objects judged more rapidly • Sentence verification technique • An apple is a fruit (yes/no) • A pomegranate is a fruit (yes/no) Prototypical objects are named more rapidly The Prototype Approach • Prototypical objects are named first – When asked to list as many objects in a category as possible, S tend to list the most prototypical members first • Prototypical category members are more affected by a priming stimulus • S heard the prime (e.g. “green”) • Two seconds later: they saw a pair of colors side by side • Were asked to press a key as quickly as possible if the two were the same The Prototype Approach Hearing “green” primes a highly prototypical “green” The Prototype Approach The Exemplar Approach • Concept is represented by multiple examples – rather than a single averaged prototype • Examples are actual category members – not abstract averages • To categorize, compare the new item to stored examples • This approach can explain Rosch’s results – e.g. objects that are like more of the exemplars are classified faster • Similar to prototype view – Representing a category is not defining it • Different: representation is not abstract – Descriptions of specific examples The Exemplar Approach • Explains typicality effect • Easily takes into account atypical cases – Rather than comparing a penguin to an “average” bird, we remember that there are some birds that don’t fly • Easily deals with variable categories – e.g. games • Which approach do we use? May use both – Exemplars may work best for small categories – Prototypes may work best for larger categories A Hierarchical Organization • Hierarchical organization: organization in which large more general categories are divided into smaller, more specific categories • Is there a “basic” level that is more psychologically important than other levels? A Hierarchical Organization • Rosch’s research indicates that there are different levels of categories – From general (“furniture”) to specific (“kitchen table”) • When using categories, we use one of these levels • Three levels of categories: – Superordinate level (or global level): “furniture” – Basic level: “table” – Subordinate level (or specific level): “kitchen table” Evidence that Basic-Level Is Special • Rosch et al. (1976): S were asked to list as many features as they could that would be common to all or most of the objects in the category • Going above basic level results in a large loss of information • Going below basic level results in little gain of information • ➔ Basic level is psychologically special Evidence that Basic-Level Is Special • When S asked to name these objects, they named them by their basic level name • Guitar, rather than electric guitar (specific) or musical instrument (global) • Fish, rather than trout or animal Culture and Categorization • Rosch’s experiments, were carried out on college under-graduates – Results: there is a category level, which they called “basic”, that reflects college undergraduates’ everyday experience • Coley, Medin & Atran (1997): Asked both undergraduates & horticulturalists to walk around (campus/Guatemala) and name as specifically as possible 44 different plants – 75% of the undergrads used “trees” – Horticulturalists (people who grow plants) used “specific” categories instead, such as “oak” – Guatemalan Itzaj culture, call an oak tree an “oak” and not a tree Evidence that Basic-Level Is Special • Tanaka and Taylor (1991): asked bird experts and non-experts to name pictures of objects • Experts responded by specifying the birds’ species • Experts had learned to pay attention to features of birds that non-experts were unaware of People’s knowledge and properties of objects are both important factors in categorization Evidence that Basic-Level Is Special What are you experts in? Semantic Networks • Concepts are arranged in networks that represent the way concepts are organized in the mind • Collins and Quillian (1969) – Node = category/concept – Concepts are linked – Model for how concepts and properties are associated in the mind – Hierarchical model, more specific to more general Semantic Networks • Including “can fly” at every node is inefficient – This property is placed at a higher level node • Inheritance – Lower-level items share properties of higher-level items • Cognitive economy: shared properties are only stored at higher-level nodes • Exceptions are stored at lower nodes – e.g. “ostrich” would have the property “can’t fly” Semantic Networks • Given the network’s hierarchical organization, general concepts are at the top, and specific ones at the bottom • Testable prediction: the time it takes for a person to retrieve information about a concept should be determined by the distance that must be traveled through the network • It takes longer to answer “yes” to “A canary is an animal” compared to “A canary is a bird” Semantic Networks • Spreading activation – 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 Semantic Priming • Lexical decision task – Participants read stimuli and are asked to say as quickly as possible whether the item is a word or not • Pair 1 • Pair 2 • Pair 3 • Pair 4 – Fundt – Bleem – Chair – Bread – Glurb – Dress – Money – Wheat • Meyer and Schvaneveldt (1971) – “Yes” if both strings are words; “no” if at least one was a nonword – Some pairs were closely associated – Reaction time was faster for those pairs • Spreading activation Semantic Networks • Criticism of Collins and Quillian – Cannot explain typicality effects • Typicality effect: reaction times for statements about an object are faster for more typical members of a category • The model instead predicts equal fast reaction times to both “canary” and “ostrich”, as they are one node a way from “bird” – Cognitive economy? • Some evidence that people store specific properties of concepts right at the node for the concept – Some sentence-verification results are problematic for the model • A pig is a mammal. RT= 1,476ms • A pig is an animal. RT= 1,268ms The Connectionist Approach • New approach to networks: – Criticism of semantic networks – advances in understanding how information is represented in the brain • Connectionism: an approach to creating computer models for representing cognitive processes • Also called Parallel distributed processing – Knowledge represented in the distributed activity of many units • Weights determine at each connection how strongly an incoming signal will activate the next unit The Connectionist Approach • Output units: Receive input from hidden units • lines: connections that transfer information between the units • Units: inspired by the neurons in the brain • Roughly equivalent to axons • Connection weight: how signals sent from one unit either increase or decrease the activity in the next unit • Patterns of activity in these units represent concepts and their properties • Input units: units activated by stimuli from the environment The Connectionist Approach • High connection weights: result in a strong tendency to excite the next unit • Negative connection weights: can decrease excitation or inhibit activation of the receiving unit • Activation of units in a network depends on two things: – The signal that originates in the input units – The connection weights throughout the network • A stimulus is represented by the pattern of activity that is distributed across the other units Concepts in Connectionist Networks The Connectionist Approach • For the network to operate properly, the connection weights have to be adjusted ➔ learning process • How learning occurs – Network responds to stimulus – Provided with correct response – Modifies responding to match correct response The Connectionist Approach • Error signal – Difference between actual activity of each output unit and the correct activity • Back-propagation: error signal transmitted back through the circuit • Indicates how weights should be changed to allow the output signal to match the correct signal • The process repeats until the error signal is zero – Initially weak and undifferentiated activation of property units, with many errors – Error signals are then sent back – Changes are made in connection weights – Each learning experience causes only a small change in the connection weights The Connectionist Approach The Connectionist Approach • In connectionist networks: information about each concept is contained in the distributed pattern of activity across a number of units • The operation of connectionist networks is not totally disrupted by damage – Graceful degradation: disruption of performance occurs gradually as parts of the system are damaged • Connectionist networks can explain generalization of learning. – Slow learning process that creates a network capable of handling a wide range of inputs Google and Connectionism • At Google: https://www.youtube.com/watch?v=Zwm2dUNm4Xo • On Machine Learning: https://www.youtube.com/watch?v=f_uwKZIAeM0 Categories in the Brain • Different areas of the brain may be specialized to process information about different categories – Patients with category-specific memory impairment – Poorer performance for “living things” than“nonliving things” • Sensory-Functional hypothesis: our ability to differentiate living things and artifacts depends on a semantic memory system that distinguishes sensory attributes (living things) and a system that distinguishes function (artifacts). • Predicts: a patient who can’t identify living things should have impaired sensory abilities Sensory-Functional hypothesis • S-F hypothesis predicts: a patient who can’t identify living things should have impaired sensory abilities • Caramazza & Shelton (1998): reported a patient who couldn’t identify living things, had impaired sensory memory, but who also had impaired functional ability • S-F wouldn’t predict this • S-F hypothesis predicts: a person who can’t identify artifacts should have impaired functional knowledge • Ralphs et al. (1998): reported a patient who couldn’t recognize artifacts but who had an impaired sensory ability Multiple factors approach • Search for more factors that divide up concepts within category • (rather than identifying specific brain areas of networks for different concepts) – Mechanical devices (e.g. musical instruments) overlap with: • artifacts (involve performing actions) • animals (involve sound and motion) ➔ mechanical devices have a widely distributed semantic representation: • regions important for the representation of both living things and artifacts ➔ Patients may be able to identify mechanical devices even if they perform poorly for other types of artifacts Multiple factors approach • Another differentiating factor between animals and artifacts is crowding • Crowding: When different concepts within a category share many properties – e.g., “animals” all share “eyes,” “legs,” and “the ability to move” – In contrasts, artifacts share fewer properties • Patients with category-specific impairments, may not have a category-impairment at all – Patients may have difficulty recognizing living things because they have difficulty distinguishing between items that share similar features Semantic category approach • Specific neural circuits in the brain for specific categories – Distributed over a number of different cortical areas • There may be a limited number of categories that are innately determined because of their importance for survival – E.g., faces, body parts, places • Wilmer et al. (2010) tested this idea by measuring face recognition ability in monozygotic (identical) and dizygotic (fraternal) twins – Correlation of scores between identical twins was more than twice as high as the other group – There may be genetic basis for the mechanisms involved in face recognition The Embodied Approach • Our knowledge of concepts is based on reactivation of sensory and motor processes that occur when we interact with the object – When we use a hammer: – Different sensory areas are activated due to the hammer’s size, shape, … – Motor areas are activated that are involved in carrying out actions involved in using a hammer – When we see a hammer, or see “hammer”, the same areas get activated – ➔ interaction between action and perception Mirror Neurons • Mirror neurons: Neurons that fire when we do a task or when we observe another doing that same task • https://www.youtube.com/watch?v=Xmx1qPyo8Ks • What do mirror neurons have to do with concepts? – Embodied approach: thinking about concepts causes activation of perceptual and motor areas associated with the concepts – Semantic somatotopy: Correspondence between words related to specific body parts and the location of brain activation – e.g. kicking, and reading the word “kick” activate the same brain areas Categories in the Brain Hauk et al. (2004) Categories in the Brain Foci of significant activations across 13 studies Carota et al. (2012) Summary • The three approaches agree: information about concepts is distributed across many structures in the brain • The approaches differ in their emphasis on the type of information that is most important – Category-specific approach: specialized areas of the brain and networks connecting them – Multiple-factor approach: the role of many different features and properties – Embodied approach: activity caused by the sensory and motor properties of objects Some Questions to Consider • Why is it difficult to decide if a particular object belongs to a particular category, such as “chair,” by looking up its definition? • How are the properties of various objects “filed away” in the mind? • How is information about different categories stored in the brain?