In the laboratory, people classify the color of emotion-laden words slower than they do that of neutral words, the emotional Stroop effect. Outside the laboratory, people react to features of emotion-laden stimuli or threatening stimuli faster than they do to those of neutral stimuli. A possible resolution to the conundrum implicates the counternatural response demands imposed in the laboratory that do not, as a rule, provide for avoidance in the face of threat. In 2 experiments we show that when such an option is provided in the laboratory, the response latencies follow those observed in real life. These results challenge the dominant attention theory offered for the emotional Stroop effect because this theory is indifferent to the vital approach-avoidance distinction. (Contains 3 figures and 1 footnote.)