Please enable JavaScript.
Coggle requires JavaScript to display documents.
Concepts and categories (Y1) - Coggle Diagram
Concepts and categories (Y1)
Understanding objects and function of conceptualising and categorising
If we treated every object as fully unique, we could not make predictions about behaviour based on past experience
We see stimuli we recognise and we have representation of different objects in the environment, which we associate with actions and meanings - general enough to allow for fluctuation but specific enough to identify actions from this
Representations - mental structures or symbols that stand for objects or events
Concepts - mental representations of categories that group together objects, events or ideas with similar characteristics
-> Building blocks of semantic knowledge
Conceptualising how the world works - concepts are manipulated in our mind in a way that corresponds to the link between external states
Basis of learning, survival, decision making, language and self-awareness - being able to predict behaviour and situations allows us to survive make decisions effectively, which we can only do by categorising behaviour and stimuli into associations and cognitive scripts
Apply our representations to guide behaviour - based on categories and concepts
Infant categorisation - simple categories are innate, but complex details are based on cultural and social learning
Categorisation and the brain - neural representations of a category:
Category can be represented in specific brain regions e.g. fusiform face area and the parahippocampal place area
Category can be represented with distributed networks of activity
Category information in single cells - monkey's are trained to categorise cats and dogs - press a button when a cat or dog is shown, with some images a mix of the two to identify ability to categorise vaguely fits a category
Inferotemporal cortex (IT) - represents shape and features, and so neurons respond more to dogs for example
Prefrontal cortex represents category information
Summary - concepts and categories help us to interact with the world
Classical definition approach, categorisation is based on rules about necessary and sufficient features / attributes
Prototype approach - store a single prototype in memory and categorise new items according to their match to a stored stereotype
Exemplar approach - we store all exemplars in memory and categorise a new item according to their match to all stored exemplars
Summary -
concepts are important for interacting with the world and thus for survival - enable inferences and predictions to be made about the nature of objects and organisms and are essential in language
Classical theory, or defining attribute theory, proposes that we categorise objects or organisms using definitions - reasonable for how we learn artificial categories, not natural categories
Prototype approach - items to categorise are matched, feature by feature, to the best representative of each category, known as prototypes - accounts for natural categories such as context, but does not explain artificial categories
People use prototype hierarchies at super ordinate, intermediary and subordinate - with many categories, one level has special status known as the basic level, which tends to be the intermediary level
Exemplar theories propose that we categorise items by comparing them, feature by feature, with all the memory traces of what we have perceived in our life; just like prototype theories, these explain natural categories but weakly explain hierarchial organisation of concepts and how they related to other types of knowledge
Explanation based theories propose that concepts are interwoven with other types of knowledge and that people use common sense explanations to categorise objects
People use different strategies for categorising objects - the strategies they use are partly determined by cognitive limits in attention and short term memory
Definitional approach
Based on rules of necessary and sufficient features - each feature is necessary and together sufficiently describe a category
Membership of a category is binary
Research in this approach - Shepard, Hovland and Jenkins (1961) - studied if we can form categories / concepts with necessary and sufficient conditions
A category rom 8 objects comprised of size (large/small), shape (round or square) and colour (black or white)
Type I task - able to categorise by one feature
Type II - able to categorise by two features
Type IV - able to follow a rule with an exception - two categories become 3 that fit and one that does not
Type VI - random and unable to categorise
Paricipants asked to categorise the objects in different ways
Item is presented, participant guesses category, feedback and repeat
Different participants learned different category structures (between-subjects design)
Observed which category was easier to learn and why (took less time when only one trial is needed to learn a category) - when we have to use more rules and features, it takes more time
Issues with this approach -
Works for some terms, such as kinship and legal terms - however, most categories in our memory tend to be loose and fuzzy
Graded membership - some are more typical members of a category than others, and so membership may not be binary - do not need all features to sufficiently fit in a category
Family resemblance - category members often share a set of common features, but not all common features are present in all members - they therefore may not be necessary
Prototype approach - based on prototypes
Defined as average in all members - every member of that category you meet adds to the average image
A typical member of a category
Does not necessarily look like any exact member
Membership of the category is determined by how the object is compared with the prototype of the category
Lower memory demands - only single prototypes and features stored -
Memory representation of characteristic features - some are sufficient, some are necessary, others are not
Some are more typical examples of category than others due to their higher similarity to the protoype (prototypicality) - not all or none
Support for prototype approach - sentence verification task (Smith et al, 1974) - are the folowing statements true (say yes) or not (say no)
-> more prototypical objects are judged faster (prototypicality effect)
Object naming - can you name as many objects as possible in the category of fruits (apples named before tomatoes)
More protoypical objects named first (prototypicality effect)
Issues with this approach -
Sometimes categorisation may depend more on contexts than inherent prototypes
Why can atypical examples be categorised
Exemplar approach: based on examples stored in memory
Actual members of the category encountered before - store each exemplar when it is seen
Accounts for the fact that you can recognise specific dogs, not just the prototypical dog
Do we start with definitional, then prototypical, then exemplar?
If this approach does not assume that prototype exists, how does it explain the typicality effect
Objects are compared to all exemplars in memory
More typical objects are similar to more exemplars and thus are classified faster
E.g. sparrow is similar to many other birds while penguin is not
In different contexts, different exemplars can be given more weight and become more influential - some examples are more notable in some contexts than others
Support for this approach -
Keeps variability information that prototype models do not maintain (Rips and Collins, 1993) - not just the mean of prototypical approach
Atypical cases can still be categorised based on their similarity to some (need not be all or many) exemplars
Level of similarity to exemplars corresponds to shorter reaction times
Issues with exemplar approach -
Working memory cannot compute millions of comparsions involving that many exemplars simultaneously
-> People can acquire good category structure from a couple of items, and some categories with strict criteria do not need as many exemplars
A single perspective cannot explain anything well - we have to use a conjunction of experiences
Defending the exemplar approach - categories have a hierarchial organisation from more general to more specific
Global - superordinate, such as vehicle or furniture
Basic - car or motorbike
Specific - subordinate such as car brand
Things can be categorise at many levels of abstraction - more categories -> more prototypes needed, still a high memory load to storing exemplars
-> Is that a good defence?
Comparing the 3 approaches -
What is stored in semantic memory:
-> Classic view - a set of rules, necessary and sufficient features
-> Prototype theories - idealised average - the prototype
-> Exemplar theories - particular instance - the exemplar
Can it explain graded membership, typicality effect, etc -
Classic view - no
Prototype theories - Yes
Exemplar theories - Yes
Definitions may be more useful early in the learning of simple, well-defined categories
Prototypes may be more useful for categories with a clear central tendency or common features
Exemplars may be more useful later in learning when lots of examples have been experienced and when variability is high within a category
Concepts
Allow us to classify objects and language, identity and out external world as well as internal world
Concept - building blocks of semantic knowledge
A category is a class of concepts that share some common properties
An artefact category - group of man made objects designed with a specific goal or function in mind e.g. computer
A nominal category - grouped based on arbitrary characteristics e.g. odd numbers
A natural category - group of entities that exist in the natural world
A definition is a rule containing features that are together individually necessary and jointly sufficient for category membership - features part of this are known as defining features
Concepts can be organised using a hierarchy of inclusion relations
Empirical data supporting classic theory - Bruner et al (1956); how people attain concepts
Simple concepts - simple attribute
Conjunctive concepts - several attributes have to be met together
Disjunctive concepts - at least one attribute has to be resent - hardest to learn as they are rarely used in real life
Also difficult to identify the categories when information about previous trials had to be kept in short term memory; external memory aids made the task easier
Leads to strategy development such as successive scanning - keep one hypothesis in mind to focus on this
Conservative focusing - participants start with a positive instance of the concept, and alter it one feature at a time, testing validity with each positive instance
Focus gambling - more features can be changed but similar to above - used in time constraint scenarios
Explanation based theories - weaknesses of prototype theory - using central features rather than superficial ones
Exemplar approach fails to explain organisation and links between concepts
Explanation-based theories were developed to remedy these - central theme of using common-sense explanations top categorise objects, motivated by intuitive theories people have about the world
Specify what attributes should be used for categorisation e.g. a distinction between diagnostic and surface attributes
Weaknesses -
Lack specificity
Does not relate to other parts of semantic memory
Circularity of definition - concepts are based on explanations but explanations are based on concepts
High amount of cognitive processing required - has to be a stopping point for the explanation
Role of the strategies, attention and short term memory
Many mechanisms are used at the same time, and all approaches may contribute to how we look at the world -
-> Bruner et al for example explains that artificial domains use classical approaches
-> Use prototypes in beginning and exemplars later on
Exemplar models account for data whereas prototype models have a high number of instances in the category meaning learning all stimuli is not possible
Human cognition is very flexible and participants often follow the strategy implicitly or explicitly suggested (Medin and Smith, 1981)
Theories of categoristion should pay more attention to the strategies used, such as computer models incorporating the possibility of using different strategies rather than being limited to one mechanism
Limited short term memory affects strategies used
Focusing attention differently also impacts the stimuli we lean towards when defining categories
Emotions and words - dominance, arousal and valence (emotional weighting)