formal models of memory

Threshold Model

Memories are retrieved when their activation exceeds a certain threshold. This general activation level helps in judging whether something has been encountered before, but uncertainty and guessing can occur.

Generate-Recognize Model of Recall

Recall happens in two stages: first, potential memory traces are generated, and then these are checked for recognition. This integrates recognition into the recall process. The recognition failure effect suggests the need for variations.

Network Theories of Memory

Memory is a large number of interconnected units (nodes) linked by associations. Often used for semantic memory, can be hierarchical or based on relatedness.

Spreading Activation

A principle where the activation of one memory node spreads to related nodes through associative links, influencing the availability of other concepts. Relevant to understanding priming.

Adaptive Control of Thought (ACT) Model

A comprehensive semantic memory network with a propositional network for LTM and a production memory for altering memory content (mental steps). Distinguishes between declarative memory (propositions) and production memory (mental steps), using type nodes and token nodes with limited activation.

Latent Semantic Analysis (LSA)

A natural language model that derives contextualized word meanings from their co-occurrences in text. Represents knowledge as relationships between concepts in a high-dimensional semantic space.

Global Matching Models

Models that do not assume a strong pre-defined organization of memory. Organization emerges from the retrieval process.

Search of Associative Memory (SAM)

A global matching model where retrieval is a probabilistic access of traces based on the match between the retrieval cue and stored memories.

MINERVA 2

A global matching model that accounts for how remembering changes the memory itself.

REM and SARKAE

Memory models that combine elements from different existing models.

CHARM and TODAM

Models exploring how multiple memory traces can be superimposed on a limited representational structure.

Parallel Distributed Processing (PDP) Models

Models where information is encoded in a complex pattern of weights between interconnected units. Processing is distributed, and symbolic or "localist" representations are often needed for meaning.

Dual Process Models of Memory

Memory retrieval involves simpler, automatic, familiarity-based processes and more complex, deliberative, conscious recollection-based processes. Evidence comes from double dissociations.