Publisher Summary This chapter reviews the nature of redintegrative memory and sketches a theory to explain this kind of memory. Redintegrative memory is said to occur when the stimulus is part of the larger response. Redintegrative probabilities are high when the material consists of single words, adjective-noun phrases, two-digit numbers, or complete sentences. The chapter primarily focuses on the redintegrative memory of adjective-noun phrases, where the noun is better remembered in free recall, and it is the better cue in cued recall. The theory presented in the chapter describes each word's meaning through features which vary in their salience. Nouns are assumed to have more features than adjectives. The theory further assumes that words can get bonded in two different ways. I-bonding occurs when the two words have features in common; the more they overlap, the stronger the bonding. J-bonding (or fusion) occurs when the phrase as a whole acquires features that did not characterize the ingredient words. When the S notices a phrase, a network of interconnected features is assumed to register. As time passes, these features fade, and only certain ones persist. The persisting features then reinstate the entire network, which is translated into the phrase.