concepts and categorisation Flashcards
William James - “The Principles of Psychology” (1890):
- Without categories and their corresponding concepts….
- …. Infants don’t separate their sensory experience into parts, but instead experience “one great blooming, buzzing confusion”
Categorisation, concepts, language:
- We need to recognize (learn to recognize in James’s view, but maybe some things are innate) types of thing (people, objects, situations, properties) in order to act consistently and achieve our aims.
- Concepts give a handle on what those types of thing (members of the categories) have in common.
- Language gives us labels for concepts – sometimes single words, sometimes longer expressions (“American Psychologist” for William James and co.)
What’s in a concept?:
- A concept tell us what makes something a member of a category.
- For something to be a bird it has to be A LIVING THING (an ANIMAL, in fact)
- For something to be a bird it has to HAVE FEATHERS
- etc.
- “has to” – so these are called NECESSARY CONDITIONS
And when you have a set of these necessary conditions, they should be together SUFFICIENT for being a bird
The classical view and its psychological counterparts:
- That’s the classical view, developed by philosophers, of many types of concept
- There are two (equivalent) ways of turning this idea into a psychological theory of how concepts are stored and used
- FEATURE THEORIES (we store the sets of conditions as lists of features)
- E.g. Smith, Shoben, and Rips 1974
- NETWORK THEORIES (we store concepts in networks with IS and HAS links (BIRD is ANIMAL; BIRD has FEATHERS)
- E.g. Collins and Quillian, 1972
Eleanor Rosch and typicality:
· Eleanor Rosch (and others) soon noted that features or network links are not all that matters
· Typical members of categories are processed more easily than atypical members (robin vs ostrich as an example of a bird)
Prototype theory:
· These observations led to the idea that concepts are represented by PROTOTYPES (not by lists of features, or relations in a network)
· Category membership depends on which prototype a particular thing is closest to.
· So, depends on a measure of closeness
· An alternative idea is that prototypes are not themselves represented, only exemplars, and it is the clustering of exemplars that determines the centre of the space that a particular concept occupies
Problems for prototype theory:
· Conceptual combination
· We don’t just use individual concepts, we can combine them
· Conceptual combination is a complex process - compare TIN + CAN => “tine can” and TIN + MINE = “tin mine”
· Conceptual combination has been seen as a particular problem for prototype theory
More problems for prototype theory:
· AD HOC CONCEPTS (Barsalou, 1983) – concepts and their corresponding categories that are put together on the fly, and therefore not stored in memory
· E.g. things to save in a house fire – although they are not stored in memory, they do show prototype effects
· MATHEMATICAL CONCEPTS (Armstrong, Gleitman, & Gleitman, 1983.e.g. odd number) – they do show prototypicality effects (7 is a more prototypical odd number than 343239089), but they have clear analytical definitions (odd numbers are not divisible by 2) – they are not DEFINED by prototypes.
“Theory” Theory:
· An alternative type of theory (“Theory” theory, Murphy & Medin, 1985) is that just as scientific concepts are defined by the role they play in scientific theories, our everyday concepts are defined by their place in lay theories about the world and how it “works”.
· ”Theory” theory deals well with conceptual combination, because, for example, lay theories tell us that cans can be made of tin, but mines cannot (but they can be a source of tin)
Basic level categories:
· Another idea from Eleanor Rosch (Rosch et al., 1976)
· In a hierarchy, such as Braeburn – Apple – Fruit, concepts at one of the levels are easiest to deal with (on all sorts of measures)
- In this case Apple, with other concepts at this level being, e.g., Banana and Orange
· At this (basic) level, the features of one type of object have a strong “correlational structure”, which is different from that of other types of object.
- Apples are similar, but apples and bananas differ
- At the higher level, fruits are diverse
- So one piece of fruit is not necessarily very similar to another one
· At the lower level, Braeburns are similar to Galas and Granny Smiths
- So, things in different categories resemble one another
Types of concept:
· Psychological work has tended to focus on concepts associated with concrete nouns. There are two main types
- Natural kinds (people, animals, plants, natural objects)
- Artefacts (man made objects: table, building, etc.)
· Less studied are abstract concepts, including scientific concepts (gravity, evolution, etc.) and social/societal concepts (family, law, government, etc.)
· There are also concepts associated with verbs (events – “hit”, states – “admire”, processes – “decay” ) and adjectives (properties of nouns, “red”) and adverbs (properties of verbs, “suddenly”)
· And then there are concepts that help to link ideas (“and”, “because”, “before”)
Abstract concepts - Lakoff and Johnsons “Metaphors we live by”:
· One account of abstract concepts is that they are understood via networks of (metaphorical) links to concrete concepts (Lakoff & Johnson’s, 1980, “Metaphors we live by”)
· Example Metaphor: an abstract state of being is a “container”
- I’m in the room. He fell into a hole. You’ll never get out of this prison.
- I’m in love. He fell into depression. You’ll never get out of trouble.
· Not just “figures of speech” but fundamental conceptual frameworks
· Is it possible to reason or communicate about abstractions without such metaphors?
Embodiment:
· The traditional treatment of concepts is via formal analysis
· A more recent set of ideas is that to understand many concepts you have to know how people interact with the world – to know what is meant by “chair” for example, you don’t just need to know that is it something to sit on, but to understand how and why people sit down (via the experience of doing it)
· Furthermore, the encoding of such concepts may have much in common with our motor knowledge of how to interact with (“use”) chairs.
Embodiment and the motor cortex - “an arm and a leg”:
· Studies looking at brain activity support this idea (Pulvermüller et al., 2005)
· They applied TMS (Transcortical Magnetic Stimulation) to motor brain regions
· Faster reactions to leg-related words (“kick”) with leg region stimulation and faster reactions to arm-related words (“pick”) with arm region stimulation
· Language is not modular or abstract but an integrated part of experience
The action compatibility effect (ACE):
· Glenberg and Kaschak (2002)
- Pushing a lever away from you to confirm that “You closed the drawer” is an OK sentence
- Pulling a lever towards you to confirm that “You opened the drawer” is an OK sentence
· The action required to respond is compatible (in direction) with the action described
· Responses are quicker than with the opposite pairings (close/towards; open/away)