lecture 4 - concepts and categories - what things are Flashcards
cognition
attention
language
memory episodic memory
concepts semantic memory
memory systems
diagram in notes
tromp et al ageing research reviews 2015
selective attention = stm and working memory
long term memory
semantic memory
inference
do you know if the following things have a liver?
* It probably never occurred to you to wonder if these things have livers.
* But for most of the items, you probably had a pretty good idea. Is this “memory of a particular event”?
* Not really - you don’t “remember” that the aardvark has a liver. You “infer” it from things you know about aardvarks and other animals.
* Knowledge is inferred from what we know about other related things
Helps us to make sense of the world and make predictions
semantic memory
- Knowledge about things in the world and their inter-relationships: words & meanings, objects, places, people
- “encyclopaedic” knowledge of facts shared by members of a community
- For example “Snow is white”,
- “Margaret Thatcher” was the first female prime minister of the UK”
brief recap
- Semantic memory is knowledge about things in the world (ideas, concepts, facts)
Episodic memory (next lecture) is memory about specific events
concepts and meanings - how are they organised?
triangles - three straight lines, three angles
dogs - animals, mammals, fur, four legs, bark, pet
fruit - parts of a plant, edible, sweet
concepts for organising the world
- We don’t just experience things in the world, we organise that experience
- We see individual instances (exemplars) as belonging to categories/classes/concept
- Knowing what something is
- ≈ what category it belongs to
- ≈ what it’s similar to/belongs with
- ≈ we know lots about it even though we haven’t actually encountered it before
Meaning of a word = accessing its concept
organisation of semantic information
Semantic Networks
Defining Attributes or prototype account
semantic networks 1
organising knowledge of what things are
defining attributes - an attribute possessed by all members of a category and one that distinguishes it from members of other categories
diagram in notes
basic level categories - critical to many activities eg afford particular modes of interaction. balance between generality (economy of information storage / access) and specificity (‘practicality”). a product of experience.
Semantic networks 2
how to investigate organisation? - reaction times (how long it takes to retrieve information)
economy - each subordinate category inherits attributes of superordinate ones - only need to be stored once
accessibility - information about categories accessed by searching through ‘nodes’
prediction - predictions about specific instances can be made, even if they haven’t been directly encountered. important as every situation we encounter is novel, even if its is similar to a previous one - we need to know how to respond to it
evidence for defining -attribute view
‘Does a dog bark?’ Versus ‘Does a dog reproduce?’
‘Is a Dalmatian a dog?’ versus ‘Is a Dalmatian a mammal?’
When asked to list defining attributes, people tend to start with ones on same level as ‘probe’ concept
Speed of response depends on ‘distance’ travelled to find information
problems with defining attributes
Certain attributes seem to be more salient (i.e., ‘stand out’) than others
e.g., people mention pink as an attribute of SALMON more often than has fins
Distance explanation: “has fins” is further away in hierarchy (attribute of superordinate fish)
But could also be Salience: Differences in speed of retrieval might be about difference in salience (more obvious attribute of salmon that distinguish it from other fish) rather than ‘distance’
‘has fins’ further away in hierarchy (attribute of superordinate fish), but also less salient (‘obvious’) than ‘pink’ as an attribute of salmon. Slower reaction time might reflect greater ‘distance’, or it might just be about salience
salient features depend on context
Salient feature changes depending on context, so we can’t even necessarily say that something’s salient defining attributes are properties of the ‘thing itself’
As salience changes, decisions about what things go together change (i.e. different organisation/categories) – so it isn’t determined by the thing ‘out there’
defining attributes - context
An object’s defining attributes seem to vary depending on our perspective, on the context, etc.
Suggests categories might be more flexible than the defining attribute view might imply
i.e., possession of a particular attribute doesn’t necessarily lead to a particular instance being assigned to a particular category
What something is is not
necessarily as obvious as it
might seem
Knowing what some individual thing “is” involves assigning it to a category BUT exactly defining what makes a category is not a simple matter
Difficulty being definitive about what the defining attributes are
What are the defining attributes of a GAME?
A property that all games possess and that distinguishes games from other activities
typicality - prototype view
Not all examples are equivalent, even though they possess the same defining attributes
Some examples are seen as better/more typical instances (prototypes) of a category than others
Robin versus canary
Typicality determines how long it takes people to make judgements about things
Relates to the number of
attributes a given category
member possesses
category membership is ‘graded’ rather than ‘either/or’
Even when all members possess all necessary and sufficient attributes, membership
is graded psychologically (e.g. 3 more typical than 2)
organisation of semantic information - exemplar account
Exemplar (or instance) based accounts:Categories as clusters of exemplars
(specific instances of things we’ve encountered)
Graded membership – certain exemplars are more typical members because we
have experienced them more often
diagram in notes
categories as ‘clusters’
graph in notes
categories as ‘clusters’ of exemplars in multidimensional space. no clear boundaries. correlations amongst features still enables prediction (cf. subordinate ‘inheritance’ account of prediction. Hierarchy as ‘zooming’ in and out
Focussing on different dimensions -> different categories emerge
graded ‘ad hoc’ categories
Focus on different ‘dimension’ of thing in order to form category
Things you could throw at a cat?
Stapler
Shoe
Banana
Guitar
Mobile phone
TV remote
Graded ad hoc categories, people agree on what belongs/doesn’t belong, just like ‘natural’ categories
Defining Attributes c/w exemplar account - Similarities between Defining Attributes (DA) and Exemplar (E) accounts
Importance of similarity – stimuli are placed into categories by
DA: comparing them with defining attributes/ prototypes
E: comparing with other exemplars previously encountered
Assume same general cognitive process:
new stimulus –> activation of concept in memory –> comparison/ judgement about resemblence —> categorisation
Defining Attributes (DA) versus Exemplar (E) accounts
What something is
DA: determined by possession of defining attributes (comparison with prototypes)
E: variable – categories emerge from structure of the world and our interactions with it (comparison to multiple known exemplars)
Prediction/knowledge of novel objects and events
DA: hierarchical inheritance of attributes
E: correlated features
What do we ‘store’?
DA: only details relevant to the particular concept/level
E: everything about everything (?/most salient patterns?)
DA -
pros - economical, fast judgements
cons - less flexible ( can’t account for atypicality, variability)
E -
pros - flexible ( can account for atypicality eg penguin as bird and variability)
E seems to be used with more experience (more encounters of exemplars)
cons - less economical
prototype and exemplar category learning
image in notes
A = exemplar model. accuracy increases the closer example is to previous examples
B = prototype model. accuracy increase the closer example is to prototype.
found people use both models
Bowman, Iwashita and Zeithamova 2020
prototype and exemplar representations in the brain
Prototype-related activation in ventromedial prefrontal cortex and anterior hippocampus (representation of across category attributes -> memory integration)
Exemplar-related activity in inferior frontal gyrus and lateral parietal cortex (representation of individual exemplars -> memory separation)
Depending on task demands, individuals form representations at both levels
measured brain using MRI scanner when people doing task. if higher blood flow to area of brain this area is active in this task.
found use both exemplar and prototype way of processing
Bowman, Iwashita & Zeithamova eLife 2020
Semantic dementia
- Progressive, selective loss of semantic knowledge in any modality
- Profound loss of word meanings: evident in comprehension & production (empty speech)
- Inability to recognise objects
Other cognitive abilities (e.g., episodic memory) and other aspects of language (syntax, phonology, pragmatics) seem to be much better preserved.
Everyday effects of semantic difficulties
- DM (surgeon) presented because he couldn’t remember the names of his surgical instruments.
- AM presented with difficulties in naming people and objects.
- Ate defrosting raw salmon for pudding after his lunch
- Poured orange juice on his pasta and added sugar to his wine
- JL presented with similar difficulties as AM.
- Asked his wife what the stuff was growing on his face everyday
Frightened by finding a snail in the garden
- Asked his wife what the stuff was growing on his face everyday
key points
- How knowledge/experience is organised
- what is a category?
- what makes something what it is?
- We ‘automatically’ assign instances to categories
- Although what instance we assign to what category is affected by context, perspective, experience etc.
- Depending on circumstances we may rely on DA and/or E classification
- Why do we do it/What use is it?
predictions about novel instances and events
long term memory
Long-term memory contains more than exact records of sensory information that has been perceived. It also contains information that has been transformed – organised in terms of meaning. For example, the type of information that is personally meaningful to us (such as what we had for breakfast this morning or what we were doing last night) appears to be different from the type of information that is based on general knowledge (such as knowing the capitals of the world or the order in which Shakespeare wrote his plays). These two types of memory have been termed episodic and semantic memory and the distinction was originally made by Tulving (1972).
episodic memory
Episodic memory (or autobiographical memory) provides us with a record of our life experiences. Events stored there are autobiographical and there appears to be cross-cultural agreement on when such memories are acquired (even though cultures differ in terms of the type of memory encoded) (Conway et al, 2005). Some psychologists, such as Alan Baddley, have even written about their own autobiographical memories and how they were formed (Sotgiu, 2021).
- Tulving introduced the distinction between episodic and semantic memory in 1972, based on the discovery of memory impairments in patients with amnesia: there was a dissociation whereby one type of memory was spared and the other was impaired.
- However, McKoon et al (1986) had challenged this conclusion arguing that (1) examples of parallel deficits in both types of memory in these patients could and had been reported and (2) other studies found no evidence for this distinction. As DeBrigard et al (2022) have noted, ‘the picture that has emerged is that memories do not fall neatly or independently into one system or the other’.
They argue that while it is relatively easy to parse declarative memory into episodic and semantic in the lab with learning lists, doing so with real-life narratives is much more challenging (Strikwerda-Brown et al, 2019). A special issue of Memory & Cognition published in 2022 provides alternative views to, and analysis of, Tulving’s distinction.
semantic memory
- Semantic memory consists of conceptual information such as general knowledge; it is a long-term store of data, facts and information. Our knowledge of what psychology is, the date of a loved one’s birthday, the names of the authors of this book, the components of the human sensory systems should form part of your semantic memory. Semantic memories can, of course, interact with episodic ones.
- The distinction between episodic and semantic memory reflects the fact that we make different uses of things we have learned: we describe things that happened to us or talk about facts we have learned. Tulving (1983, 1984) revised his original views of the two systems, suggesting that episodic memory is a part of semantic memory, not a separate, independent system, so the debate is ongoing.
- Perhaps the most controversial data supporting the notion of semantic memory concerns stimulus specificity, the notion that one region of the brain is more involved than others in the perception or retrieval of certain categories of object. Well-known examples of this, are face recognition and the naming of inanimate and animate objects (Warrington, 1975; Warrington and Shallice, 1984; Warrington and McCarthy, 1987).
- Warrington’s patients showed evidence of a dissociation between knowledge for living and non-living things. They were able to name non-living things but had considerable difficulty in naming living things, whether the stimuli to be named were verbal or non-verbal.
In a later study, Warrington and Shallice (1984) interpreted their findings by suggesting that the two types of object-naming depended on different processing mechanisms. Living things would be processed primarily according to perceptual and visual features such as their size, colour, shape, and so on, whereas non-living things would be processed according to their function.
thinking
- One of the most important components of cognition is thinking: categorising, reasoning and solving problems. When we think, we perceive, classify, manipulate, and combine information. When we are finished, we know something we did not know before (although our ‘knowledge’ may be incorrect)
- The purpose of thinking is, in general, to solve problems. These problems may be simple classifications ; they may involve decisions about courses of actions; or they may require the construction, testing and evaluation of complex plans of action
Much, but not all, of our thinking involves language. We certainly think to ourselves in words, but we also think in shapes and images. And some of the mental processes that affect our decisions and plans take place without our being conscious of them. Thus, we will have to consider non-verbal processes as well as verbal ones (Reber, 1992; Holyoak and Spellman, 1993)
classifying
- When we think, each object or event is not considered as a completely independent entity. Instead, we classify things, categorising them according to their characteristics. Then, when we have to solve a problem involving a particular object or situation, we can use information that we have already learned about similar objects or situations.
- Concepts are categories of objects, actions or states of being that share some attributes: cat, comet, team, destroying, playing, forgetting, happiness, truth, justice. Most thinking deals with the relations and interactions among concepts. For example, ‘the hawk caught the sparrow’ describes an interaction between two birds; ‘studying for an examination is fun’ describes an attribute of a particular action; and ‘youth is a carefree time of life’ describes an attribute of a state of being.
Concepts exist because the characteristics of objects have consequences for us. For example, angry dogs may hurt us, whereas friendly dogs may give us pleasure. Dangerous dogs tend to growl, bare their teeth and bite, whereas friendly dogs tend to prance around, wag their tails and solicit our attention. Thus, when we see a dog that growls and bares its teeth, we avoid it because it may bite us; but if we see one prancing around and wagging its tail, we may try to pat it. We have learned to avoid or approach dogs who display different sorts of behaviour through direct experience with dogs or through the vicarious experience of watching other people interact with them. The point is, we can learn the concepts of dangerous and friendly dogs from the behaviour of one set of dogs while we are young and respond appropriately to other dogs later in life. Our experiences with particular dogs generalise to others
formal concepts
- Formal concepts are defined by listing their essential characteristics, as a dictionary definition does. For example, dogs have four legs, a tail, fur and wet noses; are carnivores; can bark, growl, whine and howl; pant when they are hot; bear live young; and so on. Thus, a formal concept is a sort of category that has rules about membership and non-membership.
Psychologists have studied the nature of formally defined concepts, such as species of animals.
formal and natural concepts - collins and quillain 1969
Collins and Quillian (1969) suggested that such concepts are organised hierarchically in semantic memory. Each concept has associated with it a set of characteristics. Consider the hierarchy of concepts relating to animals. At the top is the concept ‘animal’, with which are associated the characteristics common to all animals, such as ‘has skin’, ‘can move around’, ‘eats’, ‘breathes’ and so on. Linked to the concept ‘animal’ are groups of animals, such as birds, fish and mammals, along with their characteristics.
- Collins and Quillian assumed that the characteristics common to all members of a group of related concepts (such as all birds) were attached to the general concept (in this case bird) rather than to all the members. Such an arrangement would produce an efficient and economical organisation of memory. For example, all birds have wings. Thus, we need not remember that a canary, a jay, a robin and an ostrich all have wings; we need only remember that each of these concepts belong to the category of bird and that birds have wings
Collins and Quillian tested the validity of their model by asking people questions about the characteristics of various concepts. Consider the concept ‘canary’. The investigators asked people to say true or false to statements such as, ‘a canary eats’. When the question dealt with characteristics that were specific to the concept (such as ‘can sing’, or ‘is yellow’), the subjects responded quickly. If the question dealt with a characteristic that was common to a more general concept (such as ‘has skin’ or ‘breathes’), the subjects took a longer time in answering. Presumably, when asked a question about a characteristic that applied to all birds or to all animals, the participants had to ‘travel up the tree’ from the entry for canary until they found the level that provided the answer. The further they had to go, the longer the process took.
criticism of collins and quillian
The model is attractive, but it does not realistically reflect the way in which we classify concepts and their characteristics. For example, although people may conceive of objects in terms of a hierarchy, a particular person’s hierarchy of animals need not resemble that compiled by a zoologist. For example, Rips et al (1973) found that people said yes to ‘A collie is an animal’ faster than they did to ‘A collie is a mammal’. According to Collins and Quillian’s model, animal comes above mammal in the hierarchy, so the results should have been just the opposite.
roth and mervis 1983
Although some organisation undoubtedly exists between categories and subcategories, it appears not to be perfectly logical and systematic. For example, Roth and Mervis (1983) found that people judged Chablis to be a better example of wine than of drink, but they judged champagne to be a better example of drink than of wine. This inconsistency clearly reflects people’s experience with the concepts. Chablis is obviously a wine: it is sold in bottles that resemble those used for other wines, it looks and tastes similar to other white wines, the word ‘wine’ is found on the label, and so on. By these standards, champagne appears to stand apart. A wine expert would categorise champagne as a particular type of wine. But the average person, not being well acquainted with the fact that champagne is made of fermented grape juice, encounters champagne in the context of something to drink on a special occasion, something to launch ships with, and so on. Thus, its characteristics are perceived as being rather different from those of Chablis.
Rosch (1975; Mervis and Rosch, 1981)
- Rosch (1975; Mervis and Rosch, 1981) suggested that people do not look up the meanings of concepts in their heads in the way that they seek definitions in dictionaries. The concepts we use in everyday life are natural concepts, not formal ones discovered by experts who have examined characteristics we are not aware of. Natural concepts are based on our own perceptions and interactions with things in the world. For example, some things have wings, beaks and feathers, and they fly, build nests, lay eggs and make high-pitched noises. Other things are furry, have four legs and a tail, and run around on the ground. Formal concepts consist of carefully defined sets of rules governing membership in a particular category; natural concepts are collections of memories of examples that share some similarities. Formal concepts are used primarily by experts (and by people studying to become experts), whereas natural concepts are used by ordinary people in their daily lives.
Rosch suggests that people’s natural concepts consist of collections of memories of examples, called exemplars, that share some similarities. The boundaries between formal concepts are precise, whereas those between natural concepts are fuzzy; the distinction between a member and a non-member is not always clear. Thus, to a non-expert, not all members of a concept are equally good examples of that concept. A robin is a good example of bird; a penguin or ostrich is a poor one. We may acknowledge that a penguin is a bird because we have been taught that it is, but we often qualify the category of membership by making statements such as ‘strictly speaking, a penguin is a bird’.
Exemplars represent the important characteristics of a category
characteristics that we can easily perceive or that we encounter when we interact with its members.
natural concepts
According to Rosch et al (1976), natural concepts vary in their level of precision and detail. They are arranged in a hierarchy from very detailed to very general. When we think about concepts and talk about them, we usually deal with basic-level concepts – those that make important distinctions between different categories – but do not waste time and effort with those that do not matter. For example, chair and apple are basic-level concepts. Concepts that refer to collections of basic-level concepts, such as furniture and fruit, are called superordinate concepts. Concepts that refer to types of items within a basic-level category, such as deckchair and Granny Smith, are called subordinate concepts.
basic-level concept
The basic-level concept tends to be the one that people spontaneously name when they see a member of the category. That is, all types of chair tend to be called ‘chair’, unless there is a special reason to use a more precise label. People tend to use basic-level concepts for a very good reason: cognitive economy.
subordinate concepts
- The use of subordinate concepts wastes time and effort on meaningless distinctions, and the use of superordinate concepts loses important information.
- Rosch et al (1976) presented people with various concepts and gave them 90 seconds to list as many attributes as they could for each of them. The subjects supplied few attributes for superordinate concepts but were able to think of many for basic-level concepts.
- Subordinate concepts evoked no more responses than basic-level concepts did. Thus, because they deal with a large number of individual items and their characteristics, basic-level concepts represent the maximum information in the most efficient manner. When people think about basic-level concepts, they do not have to travel up or down a tree to find the attributes that belong to the concept. The attributes are directly attached to the exemplars that constitute each concept.
- It is important to recognise that concepts can represent something more complex than simple exemplars or collections of attributes.
Goldstone et al (1991) showed participants groups of figures and asked them to indicate which were most similar to each other. When they showed the participants two triangles, two squares and two circles, the subjects said that the squares and triangles were most similar, presumably because both contained straight lines and angles. However, when they added a square to each of the pairs, the participants said that the two most similar groups were the triangles plus square and the circles plus square. The concept this time was ‘two things and a square’. If the participants were simply counting attributes, then the addition of a square to the pairs should not have changed their decision. As this study shows very clearly, concepts can include relations among elements that cannot be described by counting attributes.
concepts
Concepts are the raw material of thinking; they are what we think about. But thinking itself involves the manipulation and combination of concepts. Such thinking can take several forms, but the most common forms are deductive reasoning and inductive reasoning
deductive reasoning
Deductive reasoning consists of inferring specific instances from general principles or rules. For example, the following two series of sentences express deductive reasoning:
John is taller than Phil
Sue is shorter than Phil
Therefore, John is taller than Sue
All mammals have fur
A bat is a mammal
Therefore, a bat has fur
Deductions consist of two or more statements from which a conclusion is drawn. The first group of sentences presented above involves the application of a simple mathematical principle. The second group presents a syllogism. The syllogism, a form of deductive logic and a reasoning problem with three parts. There are two premises (one major and one minor), then a conclusion. Aristotle introduced them in his Prior Analytics in 350 BCE and they formed the first example of formal logic. For example, in this syllogism:
All humans are mammals
Neil Martin is a human
Therefore, Neil Martin is a mammal
the first premise is major and describes something that is general or universal. The second is minor and gives an example of the major premise. The conclusion follows from both
All birds can fly
Penguins are birds
Therefore, penguins can fly
Here is a syllogism where the argument is invalid:
All mammals have fur
A zilgid has fur
Therefore, a zilgid is a mammal
The reasoning is valid if it is impossible for the conclusion to be untrue if the premises are true.
People differ widely in their ability to solve syllogisms. For example, many people would agree with the conclusion of the following syllogism:
These people would be wrong; the conclusion is not warranted. The major premise says only that all mammals have fur. It leaves open the possibility that some animals that have fur are not mammals.
categories
Once you categorise a person (as an individual or as a member of a particular group), the schema of that person or group is activated. Research suggests that schemata can be organised as prototypes (Cantor and Mischel, 1979) or as exemplars (Smith and Zárate, 1992).
- Categories vary in inclusiveness. Highly inclusive categories have many members (e.g. a nation) and thus overshadow potentially important differences between people. More exclusive categories have fewer members (e.g. a family). Although these capture differences more precisely, an exclusive category structure would produce too many categories – it is a too fine-grained segmentation of the world. In general, the most cognitively accessible social categories are basic level categories which are neither too inclusive nor too exclusive.
Basic level categories are default categories that we first use to generate context-specific schemata of people – these are often based on visible cues such as skin colour, physiognomy, sex and dress (Zebrowitz, 1996). However, many factors, including the social interactive context, our interaction goals and our own personal history, can influence basic level categories and what categorisation and associated schema comes into play in a particular context.
prototype
A prototype is an abstract fuzzy set of attributes that define the category, where no instance may actually embody the attributes.
exemplar
An exemplar is a specific instance of the category. For example, if your schema of Australian people is the actor Hugh Jackman then you have an exemplar representation, whereas if what comes to mind is a general notion of billabongs, kangaroos, boomerangs and so forth then you have a prototype representation.
Note that both types of schema are equally accurate or inaccurate as a ‘true’ description of the category as a whole.
group schemata and stereotypes
- Schemata of social groups are particularly significant since they characterise large numbers of people in terms of a small number of properties that submerges the variety of differences that exist between people.
- Schemata of social groups are almost always shared among people in one group. For example, British people often believe that Americans are ‘brash’, the French think the British are ‘cold’ and so forth. Shared schemata of social groups are best described as stereotypes. Because they are closely associated with prejudice, discrimination and intergroup relations, we will return to them in the next chapter (Leyens et al, 1994).
According to Tajfel (1981), stereotypes are learned early in childhood through normal socialisation rather than direct experience. Research suggests that children’s use of stereotypes and expression of negative attitudes towards out-groups peak at around the age of 7 and then decline by 8 or 9 years of age. This may reflect cognitive developmental changes that affect the way children understand the meaning of categories and attributes, and changes in role-taking skills (Aboud, 1988; Durkin, 1995).
prejudice
- Prejudice usually refers to a person’s expression of negative views of and behaviours towards members of an ethnic group that differs from their own (Brown, 1995).
A key component of prejudice is the belief that the ethnic or ‘out-group’ is highly dissimilar to the ‘in-group’ (the person’s own social or racial group).
language
- Language is one important factor that can enhance or magnify perceived dissimilarities between groups and this is no more evident than when comparing different nationalities or cultures (Giles and Johnson, 1987; Giles and Coupland, 1991).
Language is a communicative glue, bonding otherwise highly dissimilar individuals. Not only can it allow communication between the in-group members, it can also prevent or inhibit communication with out-group members
