How do female sexual preferences for male ornamental traits arise? The developmental origins of female preferences are still an understudied area, with most explanations pointing to genetic mechanisms. One intriguing, little-explored, alternative focuses on the role of associative learning in driving this process. According to this hypothesis, a preference learned in an ecological context can be transferred into a sexual context, resulting in changes in mating preferences as a by-product. I tested this hypothesis by first training female guppies to associate either orange or black colour with food delivery; I then presented videos of males with computer-manipulated coloured spots and measured female preference towards them. I also allowed females from both treatments to mate with males differing in their ratio of orange-to-black spots and measured the males' reproductive success. After training, female sexual preferences significantly diverged among treatments in the expected direction. In addition, orange males sired a greater proportion of offspring with females food-conditioned on orange compared to those conditioned on black. These results show that mating preferences can arise as a by-product of associative learning, which, via translation into variation in male fitness, can become associated with indirect genetic benefits, potentially leading to further evolution.