Locusts can learn associations between olfactory stimuli and food rewards, and use the acquired memories to choose between foods according to their nutrient requirements. They are a model system for both the study of olfactory coding and insect nutritional regulation. Previous studies have used operant paradigms for conditioning freely moving locusts, restricting the study of the neural mechanisms underlying the acquisition of olfactory memories, which requires restrained preparations for electrophysiological recordings. Here we present two complementary paradigms for the classical conditioning of olfactory memories in restrained desert locusts (Schistocerca gregaria). These paradigms allow precise experimental control over the parameters influencing learning. The first paradigm is based on classical (Pavlovian) appetitive conditioning. We show that opening of the maxillary palps can be used as a measure of memory acquisition. Maxillary palp opening in response to odour presentation is significantly higher in locusts trained with paired presentation of an odour and a food reward than in locusts trained either with unpaired presentation of food and odour or the odour alone. The memory formed by this conditioning paradigm lasts for at least 24 h. In the second paradigm, we show that classical conditioning of an odour memory in restrained locusts influences their decisions in a subsequent operant task. When locusts that have been trained to associate an odour with a food reward are placed in a Y-maze, they choose the arm containing that odour significantly more often than naïve locusts. A single conditioning trial is sufficient to induce a significant bias for that odour for up to 4 h. Multiple- and block-trial training induce a significant bias that lasts at least 24 h. Thus, locusts are capable of forming appetitive olfactory memories in classical conditioning paradigms and can use these memories to modify their decisions.