Two experiments used the category repetition procedure (Dewhurst & Anderson, 1999) to test the hypothesis that false remember (R) responses occur because participants generate associates to items presented at study. Participants in Experiment 1 studied the categorised lists either with full attention or whilst performing one of two secondary tasks (articulatory suppression or random number generation). Both secondary tasks led to a reduction in the number of false R responses, with random number generation producing the greater effect. Experiment 2 manipulated the presentation duration of study items and the instructions given to participants. The numbers of false R responses were not influenced by presentation duration, but increased when participants were explicitly instructed to make associations to study items. The findings support the view that false R responses are caused by the activation of semantic associates at encoding.