Affective priming studies have demonstrated that subliminally presented prime words can exert an influence on responses towards positive or negative target stimuli. In the present series of experiments, it was investigated whether these findings can be extended to pictorial stimuli. Ideographically selected positive, neutral, and negative picture primes that were sandwich-masked immediately preceded positive or negative target pictures (Experiment 1) or words (Experiments 2 & 3). Evaluative categorization responses to these target stimuli were significantly influenced by the valence of the prime. First, it was demonstrated that high anxious participants were selectively slowed when the subliminally presented prime was negative (Experiments 1 & 2). Second, the affective congruence between primes and targets also exerted an influence on the responses, but in a direction that is opposite to what is typically observed in affective priming research. These reverse priming effects are situated within a series of recent similar findings, and implications for theories of affective priming are discussed.