Abstract1. Understanding how animals perceive, learn and remember stimuli is critical for understanding both how cognition is shaped by natural selection, and how ecological factors impact behaviour. However, the majority of studies on cognition involve captive animals in laboratory settings. While controlled settings are required to accurately measure aspects of cognition, they may not yield realistic estimates of learning performance in natural environments. Wild bees offer a useful system in which to study cognitive ecology and comparative cognition more broadly: they encompass around 20,000 species globally, varying in characteristics such as lifehistory strategy, degree of sociality and dietary specialization. Yet, the limited number of protocols currently available for studying insect cognition has restricted research to a few commercially available bee species, in almost exclusively laboratory settings. We present a protocol (Free-Moving Proboscis Extension Response [FMPER]) tomeasure wild bees' colour preferences, learning performance and memory.3. We first used laboratory-reared bumblebees Bombus impatiens to establish that FMPER yielded results consistent with learning theory. We then successfully tested wild honeybees Apis mellifera in the laboratory and Bombus vosnesenskii at field sites. Free-Moving Proboscis ExtensionResponse is straightforward to implement, is low cost, and may be readily adapted to other flower-visiting insects. We believe it will be useful to a broad range of evolutionary biologists, behavioural ecologists and pollination ecologists interested in measuring cognitive performance in the wild and across a broader range of species.