The distinction between maintenance and elaborative rehearsal was inve
stigated in four experiments. Experiments 1-3 showed that long periods
of maintenance rehearsal, induced in a distractor recall procedure, c
onsistently produced small increases in recall performance. The increa
ses were independent of manipulations that enhance overall performance
, such as semantic relatedness and intentional learning, but seemed to
depend mainly on the occurrence of rehearsal errors. Such errors pres
umably have a dishabituating effect that draws attention to the words.
Instead of qualitatively different maintenance and elaborative rehear
sal processes, a single form of rehearsal with variable amounts of att
entional processing is suggested. Attention results in active elaborat
ion and the creation of new memory representations. Without attention,
only passive activation and strengthening of already existing represe
ntations occurs. Maintenance rehearsal is accompanied by relatively li
ttle attentional processing, but the novelty of the initial presentati
on and later rehearsal errors evoke some attentional processing which
may lead to increases in recall. In Experiment 4, the effect of mainte
nance rehearsal on an implicit test (word completion) was determined.
Because implicit memory is supposed to depend mainly on the strengthen
ing of old representations, no effect of rehearsal errors was expected
, and none was found. The role of attention is discussed. It is argued
that novelty-dependent attentional processes may play a major role in
new learning and in ensuring stability of old representations both in
the human system and in artificial neural-network models.