Huge progress has been made in developing theoretical models of choice that jointly account for both behavioural and neural data. However, certain aspects of decision making that arise in naturalistic environments have been given less attention by cognitive neuroscience to date. For example, decision making in real-world settings is rarely confined to discrete trials; it contains many intermediate ‘information sampling’ decisions about what to attend to as the decision unfolds; and it often requires the decision-maker to actively navigate and explore the environment. These features all introduce unique challenges for the design and analysis of cognitive neuroscience experiments, but once they are accounted for, they can provide a new perspective on the neural representations that support adaptive behaviour. I will discuss recent neurophysiological experiments that address some of these challenges.