There is now considerable evidence that a person's cognitive processing is influenced by emotional factors. It is less clear, however, how this bias arises. Two hypotheses are distinguished and compared in the present study. The Inferred Bias hypothesis asserts that the cognitive effects of an emotion are not, in general, inherent consequences of that emotion, but instead arise from earlier experiences in which the particular patterns of emotional and cognitive activity have tended to be associated. The Integral Bias hypothesis asserts, in contrast, that the cognitive effects of an emotion are in general a fundamental characteristic of that emotion. In order to compare these two hypotheses, the cognitive effects of a phobia in children were studied. An effect on the Stroop naming of spider-related words was detected in spider-phobic children as young as 6 or 7 years old. Furthermore, the magnitude of this effect did not differ significantly over the age range to 12 or 13 years old. These results can be interpreted as suggesting that in this case the cognitive effects of an emotional disturbance emerged relatively fully fledged at a young age, and as such are consistent with the Integral Bias hypothesis of their origin.