The effects of emotional connotation on emotional Stroop interference in anxiety were examined. First, a classical conditioning paradigm was used in which neutral words and nonwords were paired with either negative or neutral pictures. These conditioned stimuli were then presented in an emotional Stroop paradigm. Finally, participants rated each word and nonword for emotional connotation. The high-anxious group demonstrated significant interference for the nonwords that had been negatively conditioned, and these effects did not dissipate over time. The affective rating data supported the view that nonwords, but not the words had been successfully conditioned in the high-anxious group. This experiment provides evidence for the importance of emotional connotation rather than confounded semantic factors in the emotional Stroop effect.