Glossary Updates16 new terms added to the glossaries · November 12, 2026, 21:23 CET
AI TechDocKnowledge

Concepts, frames and cascades

Frames

Also known as: Barsalou frames · Frames as concept representations · Attribute–value structures

A frame, in cognitive science and semantics, is a structured representation of a concept: a central node that stands for the thing described, connected by attributes to values, which can carry attributes of their own. The frame hypothesis of Lawrence W. Barsalou holds that frames are the general format in which human cognition represents concepts. The Düsseldorf research center on the structure of representations applied the hypothesis to the meanings of words and sentences and gave frames a formal definition.

By knowledge.aitechdoc.world · Last reviewed

Definition

A frame represents a concept as a network. One node is distinguished as the central node; it stands for the individual the frame describes — an object, an event, a person. Attributes lead from a node to a value: size returns the size of a thing, agent the person who performs an action, mother the mother of an animal. Values may carry attributes of their own, so frames are recursive. Nodes can be restricted to a type, a condition stating to which kind of entity the value must belong.

In the formal version developed by Sebastian Löbner, frames are first-order: the attributes are functions from individuals to individuals, a node has at most one value per attribute, and a frame can be translated into a formula of predicate logic. Frames are drawn as diagrams or written as attribute–value matrices.

Historical development

Charles J. Fillmore introduced frames into lexical semantics in the 1970s: a word evokes a scene with its participants, such as the buyer, seller, goods and price of a commercial transaction. Marvin Minsky (1974) proposed frames as data structures for knowledge in artificial intelligence. Lawrence W. Barsalou (1992) defined frames as attribute–value sets with structural invariants and constraints and proposed them as the general format of cognitive representation. The Collaborative Research Centre 991 “The Structure of Representations in Language, Cognition, and Science” at Heinrich Heine University Düsseldorf (funded by the Deutsche Forschungsgemeinschaft) applied this frame hypothesis to semantics, syntax, morphology and phonology and to concept formation in science.

Components

  1. Central node. Represents the entity the whole frame describes; the frame as a whole is a type description of it.
  2. Attributes. Functional relations that assign exactly one value to their argument, such as color, weight, agent or time.
  3. Values. Further nodes, which may be specified, typed or left open.
  4. Type constraints. Conditions on the kind of entity a node stands for.
  5. Constraints between attributes. Dependencies such as a value of one attribute restricting the value of another.

Function and purpose

Frames make the content of a concept explicit and comparable. Because every attribute is a function, frames can be merged by unification: the meanings of a verb and its arguments combine by identifying nodes. Frames therefore serve both as a model of concepts in the mind and as a tool to describe how word meanings combine in a sentence.

Applications

The volume Concepts, Frames and Cascades shows frames at work in several fields. Thomas Gamerschlag and Wiebke Petersen analyze the readings of the German verb steigen (“climb, rise”), including its figurative use for a road that “rises”, as different frames of one verb. Kata Balogh and Rainer Osswald use decompositional frames to describe how Hungarian verbal particles set a boundary to a change. Marios Andreou and Simon Petitjean model why derived words such as English nouns in -al can denote an act or a thing, and implement the constraints in the eXtensible MetaGrammar (XMG). Leda Berio uses frames to explain why color words influence some non-linguistic tasks and not others. Samuel D. Taylor and Peter R. Sutton weight observed features in category learning by their position in a frame.

Comparison with ontologies and data models

Frames, ontologies and RDF graphs share the idea of entities connected by named relations. They differ in purpose. A frame models how a concept is represented and how meanings combine; it is a hypothesis about cognition and language. An ontology in information science fixes classes and properties for a domain so that systems can exchange and check data. Frame attributes are functional (one value each), while RDF properties may take several values. Frame theory is therefore a source of ideas for information models — central node, functional attributes, constraints — rather than a modeling language of its own.

Conclusion

Frames describe concepts as recursive attribute–value structures with a central node. The frame hypothesis treats them as the general format of concept representation; the work collected in Concepts, Frames and Cascades tests it on word meaning, category learning and perception.

Source work and licence

This article draws on:

In:

Sebastian Löbner, Thomas Gamerschlag, Tobias Kalenscher, Markus Schrenk, Henk Zeevat (eds.), Concepts, Frames and Cascades in Semantics, Cognition and Ontology. Language, Cognition, and Mind, vol. 7. Springer (Springer Nature Switzerland AG), 2021. ISBN 978-3-030-50200-3. doi.org/10.1007/978-3-030-50200-3

Open access under the CC BY 4.0 licence. The articles of this subject area summarize, compare and recontextualize the volume’s ideas in our own words and structure; no text, figures or tables are reproduced. Positions are attributed to the chapter authors; the comparisons and the relation to ontologies and documentation are ours.

Further reading

Cite this article

knowledge.aitechdoc.world. “Frames.” Encyclopedia, AI TechDoc Knowledge. Last reviewed October 13, 2026. https://knowledge.aitechdoc.world/encyclopedia/concepts-frames-and-cascades/frames