A version of the Stroop test was developed which requires colour naming of spider words. Spider phobics were severely retarded on this task, but not on the conflicting colour‐word Stroop or a Stroop with more general threat words. Desensitization of phobics significantly reduced interference on the Spider Stroop. A third experiment examined the retest properties of the Stroop and the value of equivalent forms. The explanation of the Spider Stroop effect and its usefulness as an outcome measure are discussed.