Where does spreading activation collide?

Retrieving one concept makes the concepts linked to it in memory easier to reach, because activation spreads outward along the associative links of a semantic network. Collins and Loftus set this out in A spreading-activation theory of semantic processing in 1975, revising an earlier hierarchical model that Collins and Quillian had built in 1969, in which concepts sat in a strict tree. In the spreading-activation version the network is a web of nodes joined by links of varying strength, and reading or hearing a word sends activation from its node to its neighbours, so a related word is recognised faster. This is the standard explanation for semantic priming, where nurse is read more quickly after doctor. The mechanism here is described in words and diagrams rather than equations, and its predictions are about relative speed, not exact numbers.[1][2]

Unit

How far and fast activation travels between linked concepts

Recovery means

Activation is thought to fade as it spreads and to die away once the cue is gone, so distant nodes receive less of it

Measured by

Reaction-time studies of semantic priming and lexical decision, comparing related with unrelated cues

The same idea of activation spreading along links, but written as equations that predict how likely and how fast a memory will be retrieved. Anderson's A spreading activation theory of memory in 1983, and the later ACT-R architecture built on it, treat the activation of a memory as a base-level term that reflects how often and how recently it has been used, plus an associative term that sums the activation spreading from whatever is currently in mind. Retrieval succeeds when total activation clears a threshold, and its speed and probability follow from the activation value. This turns a verbal network theory into a fitted, quantitative model whose parameters are estimated against data. The mechanism is the same spread of activation, but here it carries a number rather than only a direction.[3]

Unit

A numeric activation value for a memory item, in the model's own units

Recovery means

Base-level activation decays as a power function of the time since each past use, so an unrehearsed memory loses activation

Measured by

Fitted to retrieval latency and error rates in memory experiments and in cognitive-model simulations

The same mechanism applied to what people hold in memory about a brand, where the brand and its attributes, benefits, and feelings are nodes joined by associative links. Keller's Conceptualizing, Measuring, and Managing Customer-Based Brand Equity in 1993 built brand knowledge on the associative network memory model, so that meeting a brand cue activates the associations linked to it and makes them easier to reach. On this account a strong brand is one whose node is tied to many favourable, distinctive associations that come to mind readily when the brand is seen. The spread of activation explains how a jingle, a colour, or a slogan can call up a whole set of brand meanings at once. The referent is consumer memory rather than a laboratory word list, but the claim about activation travelling along links is borrowed directly from the cognitive theory. The neighbouring page on brand equity follows this associative picture in detail.[4]

Unit

How readily a brand cue brings its linked associations to a consumer's mind

Recovery means

Not modelled directly; accessibility is assumed to fade without repeated exposure to the brand

Measured by

Free-association tasks, recall and recognition of brand attributes, and response-time measures of accessibility