Concepts & Categories

Concepts & Categories

  • Mental representations of classes of things (Murphy, 2002)

  • ‘Glue’ that hold out world together

What are Concepts for?

  • Concepts allow us to make sense of the world, and to make accurate predictions about novel items of a category

The Classical View of Concepts

  • Grounded in philosophical tradition of formal logic

  • Concepts can be defined by the presence or or absence of features (e.g. if conditions A,B,C and D are satisfied, then object X belongs to category Y)

  • Necessity and Sufficiency: Properties A, B, C, and D are both necessary and sufficient for object X to be categorized as an instance of Y

  • Concepts are merely lists of necessary and sufficient conditions

Problems with Classical View

  • Wittgenstein: What are the necessary and sufficient conditions for the concepts of ‘game’?

  • Even fairly simple concepts do not have a definition making it difficult to create a concept list

  • Wittgenstein: What are the the necessary and sufficient conditions for concept of ‘game’

    • Inconsistency with an individual-

      • McCloskey and Glucksberg (1978): when asked whether an olive is a fruit, people changed their minds when asked more than once

    • A dog is a dog is a dog is a dog?

      • Classical view predicts that all members of a category are ‘equal’. In other words, a Chinese crested dog is equally as good of an example of ‘dog’ as a golden retriever

Typicality and The Fall of the Classical View

Eleanor Rosch and Typicality

  • Rosch: members of a category differ in terms of how well they represent the category as a whole. A sparrow is a typical bird, whereas a chicken is a atypical bird.

  • Typical it’s are categorized more quickly and more consistently

  • What makes an item “typical”?

  • Family Resemblance- the extent to which exemplar X shares attributes with other exemplars in category A, but not exemplars from carry B, determines ‘typicality’

Typicality

  • Exemplars of a category can vary widely in terms of how ‘typical’ they are for that category

  • Categorization behaviour is sensitive to exemplar typicality

  • Rips, Shoben, & Smith (1973): participants were faster to verify the robin is a bird than ostrich because it is a more typical

  • Induction: the ability to generalize or extend properties of some category members to others

    • Premise: Robins have sesamoid bones?

      • Do sparrows have sesamoid bones?

  • Rips (1975): People are more likely to believe sparrows have  sesamoid bones when the premise involves a typical category member (e.g. robin) than when we premise involves an atypical category member (e.g. penguin)

Typicality and Fear Generalization

  • Dunsmoor and Murphy (2014): Does stimulus typicality determine how broadly conditioned fear is generalized?

    • Typical group: Typical mammals paired with aversive shock during conditioning phase

    • Atypical group: Atypical mammals paired with aversive shock during conditioning phase

  • Does stimulus typicality determine how broadly conditioned fear is generalized?

    • Typical mammals paired with aversive shock during conditioning phase

    • Atypical mammals paired with aversive shock during conditioning phase

  • Does stimulus typicality determine how broadly conditioned fear is generalized

    • Broad conditioning makes generalized fear easier

Alternatives to the Classical View

Prototype Theory

  • Concepts are represented by a categorical ‘prototype’

  • Prototypes capture the central tendency of a category- ‘summary’ representation

  • Categorization is based on the similarity between an exemplar and the prototype for that concept

Posner and Keele (1968)

  • Training Phase

    • Participants see multiple exemplars from ‘different’ categories

    • Exemplars consist of low and high distortion items from each ‘category’

  • Test Phase

    • Participants are shown new exemplars and the prototypes themselves from each ‘category’ and asked to classify them

  • Results: Better accuracy for low vs. high distortion exemplars

  • Accuracy highest for prototypes despite never seeing them

  • Exposure to exemplars created a ‘prototype’ representation for each category

  • Categorization of new exemplars was based on comparison to prototype representation

Exemplar Theory

  • Subjects do not form an abstract ‘prototype’ for each category

  • Individual exemplars are stored in memory

  • Categorization is based on the similarity between a new test item and stored exemplars

Whittlesea (1983)

  • Created categories based in CVCV strings (i.e.FRUIG, NOBAL) as prototypes

  • Only children of the words are shown and people were told to guess which parent the word came from

  • Similarity of exemplars to prototype varied by changing one, two, or three letters from category prototype

  • Do people categorize based on similarity to stored exemplars?

  • Exemplar theory suggest as we move away from the children of the exemplar it should become more difficult to recognize as belonging to that category

  • Prototype theory suggests as we move away it should not matter as we are comparing the example to a prototype we’ve made up

Brooks, Norman, Allen (1991)

  • Seasoned dermatologist and inexperienced medical residents shown labelled pictures of different skin diseases- asked to judge how ‘typical’ each one is of that diagnostic category

  • The asked to categorize new pictures (unlabeled) that were from the same diagnostic categories, but were either superficially similar or dissimilar from the original images

  • Results showed that doctors were more likely to diagnose images that looked typical to examples they had just been shown

Alternatives to Classical View

  • A general consensus regarding the correct theory of concepts (prototypes vs. Exemplars) has yet to be reached

  • There is evidence for both theories

  • Some have suggested that when leaning a new category, people initially use individual exemplars to guide their categorization decision , but switch to using a prototype once sufficient experience accrues

    • Not whoever eve experts rely on similarity between novel exemplars and previously encountered examples (possibly even more than novices)

The Role of Prior Knowledge

  • Majority of research on concepts has focuses on category leaning with artificial materials

    • This is done to ensure that prior knowledge can’t influence category learning

    • Allows researchers to examine ‘pure’ category representations

  • However, most new concepts that we learn are in some way associated with related  prior knowledge

Lin & Murphy (1997)

  • When groups were told different uses for an item then showed a collection of images of FINSIHHHHHH

  • Prior knowledge has a large impact on how people categorized objects

How are Concepts Organized?

  • Categorical knowledge is structured in a nested hierarchy- taxonomic organization

  • Transitivity- All dogs are mammals, and all mammals are animals. Therefore all dogs are animals

  • Property Inheritance- all lower categories inherit the properties associated with higher taxonomic

Importance of Basic Categorization

  • Preferred Level of categorization

    • Most often used when spontaneously naming an object

  • Balance the trade between informativeness and distinctiveness

  • Are often the first categories that children learn based on sorting and naming

Rosch et Al. (1976)

  • Faster Res to verify category image pairs when category label is at basic category level

  • Demonstrates that there is seemingly something special about the basic level (default level)

  • Basic category identification is the last to degenerate in patients with dementia

Categorization and Expertise

  • Does expertise in a particular domain affect preferred level of categorization?

  • Expertise on some level represents categorization at its highest level

  • It takes a dog expert the same amount if time to classify a cocker spaniel as a

    • Expertise eliminates the advantage FINSIH THIS

Chi Feltovich, Glaser (1981)

  • Asked novices and experts to classify different types if physics problems

  • Novices tended to sort problems based on superficial properties, whereas experts tended to sort based on underlying commonalities, forming few overall groups

  • Experts know when to lump and when to split when appropriate and are not distracted by false commonalities