In three experiments, we examined the effects of prime-target stimulus onset asynchrony (SOA) and the proportion of related primes and targets (relatedness proportion, or RP) on semantic priming when the prime was either named or was searched for a specific letter. In Experiment 1, with an RP of .50, priming occurred at SOAs of 240 and 840 msec when the prime was named, but no priming was found at either SOA when the prime was searched for a letter. In Experiment 2 the RP was either .20 or .80, and the SOA was set at 1,700 msec; priming again was found in both conditions when the prime was named, but only in the RP .80 condition when a letter search task was performed on the prime. In Experiment 3, both the proportion of related trials and SOA were varied; as in the previous experiments, no priming effects were found with the letter search task for either SOA in the RP .20 condition, but the priming effect was reinstated in the RP .80 condition. These results are discussed with respect to how limited capacity resources are allocated and how they influence semantic priming effects.