In four experiments we investigated whether interpretative biases found in
anxious patients and high-trait anxious individuals can be induced by train
ing in unselected volunteers. Repeated exposure to emotionally valenced (th
reatening) meanings of homographs during training was followed by relativel
y faster resolution of word fragments and faster lexical decisions for targ
ets that matched the trained valence. Similar effects were found whether pa
rticipants generated the meanings themselves, or verified a particular mean
ing of the homograph. Finally, comparison with a baseline condition confirm
ed that systematic exposure to threatening-but not non-threatening-resoluti
ons of ambiguous words led to generally faster access times for congruent m
eanings, thus resembling the interpretative bias seen in anxiety states.