Three experiments assessed the long-term effectiveness of the keyword
mnemonic relative to a nonmnemonic (i.e., semantic-context) learning s
trategy. Following incidental-learning instructions, cued recall was a
ssessed either immediately or after a 2-day delay. The keyword mnemoni
c produced superior immediate performance relative to the semantic-con
text strategy. However, after 2 days, there was a marked reversal in p
erformance, with higher levels of delayed recall associated with seman
tic-context learning. This pattern of findings was obtained when obscu
re English words (Experiment 1) and second-language vocabulary (Experi
ment 2) were the learning stimuli. When practice frequencies were mani
pulated (Experiment 3), increased opportunities for study were more li
kely to boost the long-term retention of keyword learners compared wit
h semantic-context learners. The implication is that keyword-based mem
ories are especially fragile over time and will benefit from repeated
testing and rehearsal.