The origins of fear of spiders in an undergraduate student sample was investigated. Spider-fearful (n=19) and non-fearful (n=19) groups were formed on the basis of extreme scores to the spider item on the FSS-III (Wolpe&Lang, 1969). Spider-fearful subjects completed Menzies and Clarke's (1993a) recently constructed Origins Questionnaire (OQ), while non-fearful subjects completed a modified version of this measure. No cases of spider fear were found to originate with direct conditioning, vicarious conditioning, or information/instruction. In contrast, 63% of the sample claimed to have always been afraid of spiders, i.e., from their first encounter. These data question the significance of simple associative-learning events in the origin of fear of spiders, and are taken to support the recently proposed non-associative account of Menzies and Clarke (1993a, 1993b). Inconsistencies with some earlier studies in which high frequencies of classical conditioning have been found are explained in terms of the differences between the various origins questionnaires used. Finally, problems associated with research relying on retrospective data are examined.