This study investigated the hypothesis that discrepant results regarding the recency effect in Alzheimer's disease (AD) patients are due to the different scoring procedures used by various authors and/or to the different number of terminal items attributed to the recency part of the curve. Our results indicate that the last two processed words are available to AD patients for recall, just as they are to controls. Words processed slightly earlier are less available to AD patients than to controls, presumably because of the accelerated forgetting rate in demented patients.