Academic journal article Memory & Cognition

The Next Generation: The Value of Reminding

Academic journal article Memory & Cognition

The Next Generation: The Value of Reminding

Article excerpt

Published online: 31 January 2012

© Psychonomic Society, Inc. 2012

Abstract In two experiments, we investigated the influence of repeated processing in the context of the generation effect. In both experiments, participants studied words once or twice. Once-studied words either were read or were generated from a definition. Twice-studied words were read both times, generated both times, or read once and generated once. Free recall was best (in order of decreasing performance) after generating twice, after generating plus reading, and finally after generating once; any generation was better than purely reading. Recognition showed a similar pattern, except that the benefit of generating twice was not as striking as in recall and that reading plus generating was just as effective as generating twice. The overall pattern of results is accounted for by a simple model in which a second encoding results in a reminding of the first encoding, and this additional encoding supports subsequent recollection. This reminding is, consequently, more effective in recall than in recognition, and it operates in accordance with the principles of transfer-appropriate processing.

Keywords Memory . Recall . Recognition

The generation effect, introduced by Slamecka and Graf (1978), is one of the best-known encoding manipulations. Simply put, it refers to the benefit on a later memory test of producing an item from a cue at study, without seeing the entire item, as opposed to simply seeing the complete item at study. Typically, this phenomenon has involved generating a word from a cue such as a definition or an antonym, in contrast to simply reading the word. The literature contains well over 200 studies exploring the generation effect, not taking into account the considerably greater number that have just used generation as a trustworthy encoding task. Not surprisingly, the original Slamecka and Graf article is a citation classic (Slamecka, 1992).

Generation is widely viewed as a powerful way to encode, competitive with imagery (Paivio, 1969) and deep, semantic processing (Craik & Lockhart, 1972). This perception of the power of generation is supported by a recentmeta-analysis that included 86 studies. Bertsch, Pesta, Wiscott, and McDaniel (2007) reported the typical mean difference in favor of the generated items over the read items to be about 9%, or just under one half of a standard deviation (.40), leading them to refer to the effect as "robust and consistent" (p. 203), a fair evaluation of its significance. In fact, it even has everyday value: Mulligan and Lozito (2004, pp. 177-178) reviewed some of the applied studies in which generation has improved memory.

What causes the generation effect? A variety of accounts have been put forward, as briefly described by Bertsch et al. (2007). It has been suggested that memory benefits from greater effort being devoted to generated items (McFarland, Frey, & Rhodes, 1980), but Bertsch et al. demonstrated that generation difficulty is not a good predictor of memory performance, and we agree with them that the concept of effort is too broad and ill-defined (see also Mitchell & Hunt, 1989, on this issue). It has been suggested that generated items steal rehearsals from read items, a selective-rehearsal account (Slamecka & Katsaiti, 1987), but Bertsch et al. presented evidence against this view, too, including the fact that blocking trial types so that the read and generate trials are separated does not reduce (and, in fact, actually seems to increase) the generation effect. And it has been suggested that the effect is a relatively straightforward outgrowth of transfer-appropriate processing (see Morris, Bransford, & Franks, 1977), but again Bertsch et al. presented evidence against this explanation, including that the type of test that shows the largest generation effect often is not the one that would appear best matched to the type of generation that had been performed during study. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.