Previous studies found that caffeine consumers acquired a liking for the flavour of novel caffeinated drinks when these drinks were consumed repeatedly in a caffeine-deprived, but not nondeprived, state. Expression of this acquired liking appeared acutely sensitive to current caffeine deprivation state, but the use of between-subjects designs confounded interpretation of those studies. The present study evaluated these findings further using a within-subject design, with one flavour paired with caffeine (CS + ) and the second with the absence of caffeine (CS-). During four CS + and four CS- training days, 32 moderate caffeine consumers alternatively consumed a novel flavoured drink a CS + paired with caffeine and a CS- flavour paired with placebo. Participants evaluated both drinks before and after training in two motivational states: caffeine deprived and nondeprived. As predicted, pleasantness ratings for the caffeine-paired flavour increased overall. However, this acquired liking was only significant when tested in a caffeine-deprived state. These data are consistent with a conditioned-flavour preference model and imply that expression of acquired liking for a novel caffeinated flavour depends on the need for the effects of caffeine at the time when the drink is evaluated.