Categorization

Assigning objects or events to psychologically relevant classes. Evidence about classification speed, typicality and generalization constrains representations and decision processes, but no single behavioral effect uniquely identifies a storage format.

Category verification is a task in which a person decides whether an example belongs to a stated category. A participant might judge “a robin is a bird” as true and “a robin is a fish” as false. Researchers measure accuracy and reaction time while controlling features such as wording, familiarity and the expected response.

The task is useful because equally correct category members need not be verified equally quickly. However, response speed concerns the categorization process, not a mathematical degree of membership. Sharon Lee Armstrong and colleagues compared typicality judgments with classification for well-defined categories such as odd integers, showing why graded judgments alone need not imply fuzzy category boundaries.^[ref]

Typicality is how representative an example seems of a category. A robin may be judged a more typical bird than a penguin, although both are birds. Such judgments can reflect familiarity, similarity to other members or the purposes for which a category is used; they are not the same measurement as a yes-or-no judgment of category membership.

Even a sharply defined mathematical category can produce graded psychological judgments. People can regard some odd integers as better examples than others, despite every odd integer satisfying the same membership rule. Sharon Lee Armstrong and colleagues used such cases to show that graded typicality does not, by itself, establish that a concept lacks a definition.^[ref]

Category membership concerns whether an object or event is assigned to a specified category. It is distinct from its judged typicality, how confident a person is about the judgment and how quickly the judgment is made. A category can have a precise membership rule while its examples differ in familiarity.

For example, 3 and 97 are both odd integers: each has the form $2k+1$ for an integer $k$. A participant might nevertheless consider one a more typical example or identify it faster. Those responses concern the categorization process, rather than making one integer mathematically more odd than the other.

Sharon Lee Armstrong and collaborators compared typicality judgments and category verification for everyday categories and well-defined categories including odd and even numbers. Their experiments show why graded responses alone cannot establish that the underlying membership conditions are graded. Distinguishing membership, typicality and reaction time avoids drawing a stronger representational conclusion than the task supports.^[ref]

Color categorization groups perceived colors into classes, such as those associated with color names. It is distinct from merely detecting that two stimuli differ. A study can separately ask what a sample is called, which comparison matches it and how long the response takes.

Jonathan Winawer and collaborators compared Russian and English speakers on a speeded discrimination task using blue stimuli that crossed the Russian light-blue/dark-blue naming boundary. Russian speakers showed a category-related advantage that was disrupted by verbal interference. The design compares naming boundaries, discrimination performance and a concurrent task, rather than inferring color perception from vocabulary alone.^[ref]

The result constrains how linguistic processing can contribute during a task. It does not show that a person cannot distinguish unnamed colors, or that all aspects of color vision are fixed by the words of a language. Categorical perception, naming and reaction time provide related but different measurements.

A tendency to treat a category as having a stable underlying nature that explains its nonobvious properties. This can support expectations beyond appearance without establishing that the assumed essence is biologically real.

Representing people as members of socially meaningful classes. Social categorization can organize expectations and identity; a category label alone does not determine whether its consequences are inclusive or prejudicial.

An arrangement in which group boundaries on different dimensions do not coincide. A former outgroup member may share another valued identity with the observer. This can soften a single division, but need not eliminate exclusion of people outside both categories.

Changing the category structure through which people represent an interaction, for example by replacing two opposed groups with one inclusive group. Recategorization should be assessed through perceived membership and behavior rather than inferred from labels alone.

An account of prejudice reduction in which former outgroup members become included in a common ingroup. The proposed mechanism extends favorable ingroup responses across a previous boundary. Whether the common identity recognizes or suppresses subgroup identity is a substantive boundary condition.

Reducing the relevance of a group boundary by individuating interaction partners. A central limitation is generalization: liking a particular person need not change prejudice toward other members of that person's outgroup.

More favorable evaluation or treatment of an ingroup than an outgroup. It should be distinguished from absolute hostility toward an outgroup and measured separately from consequential discrimination.

A group represented as excluding the self in a relevant comparison. A person can belong to a social outgroup on one dimension and share an ingroup membership on another.

A group represented as including the self in a relevant context. Which group counts as an ingroup can change with the dimension of social categorization.

Forming and interpreting a concept from other concepts. A theory of representation must explain how combined concepts acquire constraints and typical instances, including cases not captured by a simple average of their separate typical examples.

An account emphasizing defining conditions for category membership. Such criteria can coexist with graded identification or typicality effects; those effects alone do not prove the absence of a definitional representation.

An account in which classification depends on stored examples and their similarity to a new case. An unseen central case can resemble many stored examples and receive strong aggregate support, so prototype-like behavior need not imply an explicitly stored average.

An account in which category judgments rely on similarity to an abstract or representative summary. It naturally explains centrality and generalization effects, but must be distinguished from exemplar theory and from a descriptive typicality gradient.

A pattern of overlapping attributes among members without requiring one attribute set shared identically by every member. The distribution of attributes in competing categories matters as well as their occurrence within one category. Family resemblance can describe a category structure without specifying how it is learned or stored.

A central or representative summary used in a categorization account. It need not be a previously encountered individual, and its selection can depend on the modeled feature distribution, comparison classes and goals.

How representative a member seems or behaves relative to its category. Graded typicality is distinct from graded membership: well-defined mathematical categories can show unequal familiarity or identification speed despite exact membership criteria.