Our aim was to investigate whether or not children's drawings of a potentially threatening topic (a Hallowe'en witch) were made significantly smaller than drawings of a non-threatening topic (a woman). The first study confirmed an earlier claim by Craddick (1963) that drawings of a witch were made significantly smaller on the day before Hallowe'en compared to drawings made one week before or after, but we found that drawings of a non-threatening topic (a woman) changed in size in exactly the same way. Furthermore, we found no evidence that children became significantly more frightened of witches as Hallowe'en approached. In the second study, questionnaire responses were used to allocate children into two groups, those who were scared of witches and those who were not. Scared children drew both smaller witches and larger women than did non-scared children, with the result that relative heights of witches compared to women differed significantly between the two groups. However, only the difference in the height of the drawings of a woman was statistically significant. The implications of these results for the clinical assessment of children through their drawings are discussed.