Governments have become increasingly interested in the ‘explosion’ of research into taxpayer behaviour. This article briefly reviews two main theories of tax compliance (‘deterrence’ and ‘non-deterrence’), before discussing the recent rapid rise of natural field experiments (NFEs) in this area. These NFEs represent a ‘win–win’ for policy-makers and academics, since they can test theories while also evaluating the impact of a specific intervention in the real world. The evidence from NFEs shows that deterrence approaches improve compliance, while the case for non-deterrence approaches is less conclusive at the moment. NFEs could be improved by paying more attention to findings from behavioural science that apparently incidental factors, such as timing, framing, and complexity, substantially affect whether compliance occurs or not. New directions for future studies include: network effects, tax payments (in addition to declarations), the behaviour of firms, the provision of public goods, and the relationship between attitudes and behaviour.