Masked repetition priming in the lexical decision task was found to be greater when prime validity, defined as the proportion of repetition versus unrelated primes, was high (0.8) rather than low (0.2), even though primes were displayed for only 45 or 60 ms. Prime validity effects did not obtain when targets varied markedly from trial to trial with respect to processing difficulty. This variation appears to cause extensive prime recruitment even when prime validity is low. Reducing variability in target processing difficulty restored the influence of prime validity. Prime validity effects are anticipated by an episodic account of masked priming in which a prime event creates a resource that can be recruited to aid word identification. These effects support the idea that resource recruitment is more likely to occur when the validity of the resource is high, which creates a context that supports prime recruitment. Implications for lexical accounts of masked repetition priming are discussed.