The acquisition of fear of heights in an undergraduate student sample was investigated. Height-fearful (n = 50) and non-fearful (n = 50) groups were formed on the basis of extreme scores to the heights item on the FSS-III (Wolpe & Lang, Behaviour Research and Therapy, 2, 27-30, 1964). Subjects were then assessed with a battery of measures including the Acrophobia Questionnaire (Cohen, Behaviour Therapy, 18, 17-23, 1977), self-rating of severity (Marks & Mathews, Behaviour Research and Therapy, 17, 263-267, 1979), global rating of severity (Michelson, Behaviour Research and Therapy, 24, 263-275, 1986), and a new comprehensive origins questionnaire constructed by the authors. Results obtained question the significance of simple associative-learning events in the acquisition of fear of heights. Only 18% of fearful Ss were classified as directly conditioned cases. Furthermore, no differences between groups were found in the proportion of Ss who knew other height-fearfuls, had experienced relevant associative-learning events, or the ages at which these events had occurred. Finally, no relationships between mode of acquisition and severity or individual response patterns were obtained. In general, the data were consistent with the non-associative, Darwinian accounts of fear acquisition that continue to attract theorists from a variety of backgrounds (e.g. Bowlby, Attachment and loss. London: Penguin, 1975; Clarke & Jackson, Hypnosis and behaviour therapy: The treatment of anxiety and phobias. New York: Springer, 1983; Marks, Fears, phobias and rituals: Panic anxiety and their disorders. New York, Oxford Univ. Press, 1987). Differences with previous studies in which classical conditioning has accounted for the majority of cases are discussed in terms of the methodological differences across studies.