Three experiments using the Reicher task show that performance on low-frequ
ency words is disrupted if the incorrect alternative forms a higher frequen
cy word. This neighborhood frequency effect occurs for both energy and patt
ern masks and for different sets of items. When the upcoming word is primed
and its accessibility is enhanced, the neighborhood frequency effect is el
iminated. Experiments 4a and 4b tested the neighborhood frequency effect us
ing a same-different task and a signal-detection analysis. Neighborhood fre
quency affected the decision criterion but not the sensitivity of the perce
ptual System. Experiment 5 showed that many words with a frequency in the r
ange from 1 to 5 per million are not recognized out of context. These resul
ts are consistent with the hypothesis that frequency effects in the Reicher
task are caused by a bias in the decision system and can be simulated with
the stochastic activation-verification model.