Two experiments are reported that explore the influence of strength of the prime–target relationship on the observed priming effects in young, healthy old, and individuals diagnosed with dementia of the Alzheimer type (DAT). In Experiment 1, participants were auditorily presented primes (FURNITURE) and after varying delays presented visual targets that were (1) high-strength related (e.g., SOFA), (2) low-strength related (e.g., RUG), or (3) unrelated control words (e.g., COW or DEER). The results indicated that the DAT individuals produced relatively larger priming effects than both the young and the healthy old, but these data could be accommodated by increases in effect size due to general slowing of response latencies. In Experiment 2, the same cross-modal priming paradigm was used with ambiguous words presented as primes (e.g., BANK) and either high-dominant (e.g., MONEY) or low-dominant (e.g., RIVER) words as targets. The results of Experiment 2 produced a qualitatively distinct pattern of priming that indicated DAT individuals only produced priming for high-dominant targets and not for low-dominant targets, whereas, the healthy control groups produced equivalent priming for both high- and low-dominant targets. The discussion focuses on the implication that these results have for the interpretation of semantic priming effects, in general, along with implications for the apparent semantic memory loss in DAT individuals. (JINS, 1999, 5, 626–640.)