The ease with which a reactivated memory is updated has major implicat
ions for whether or not a prior memory is likely to be retrieved in th
e future. In three experiments, we explored this problem with nonverba
l human infants, whose newly acquired memory of training in a specific
context is readily updated by novel contextual information. In Experi
ment 1, exposing infants to a novel context immediately after a succes
sful reactivation treatment neither impaired their retention in the or
iginal context nor facilitated it in the novel exposure one. In Experi
ment 2, increasing the delay between the reactivation treatment and ex
posure to the novel context also failed to facilitate retention in the
novel test context. In Experiment 3, the reactivated memory was updat
ed when the contingency was briefly experienced in the novel context i
mmediately after the reactivation treatment. Under these circumstances
, previously trained infants exhibited retention in the novel context,
but infants who had not been trained 3 weeks earlier or whose origina
l memory had not been reactivated exhibited none. The resistance of a
reactivated memory to contextual updating unless the new context is pr
edictive apparently buffers infants' memories against revision after l
ong delays by contexts that could be inappropriate. (C) 1994 John Wile
y and Sons, Inc.