A key idea in accounts of how we choose between alternative courses of action is that there is a neural representation of the 'value' of competing alternatives during decision making. Neural correlates of value are indeed widespread in the brain, but the neural mechanisms by which value comparison is implemented are less clear. One hypothesis is that these mechanisms may be analogous to those underlying perceptual decisions - which have been more carefully examined experimentally, and are also well captured by network models employing 'competition via mutual inhibition'. In this talk, I will first present recent work that uses predictions derived from these models to test which brain regions may be involved in value comparison, employing MEG data to test the predictions. I will then present data that suggest the frame of reference (goods or actions) used for value-guided choice can be readily manipulated by subtly altering the experimental paradigm. Finally, I will examine precisely what we mean by a 'representation' of value, with particular reference to an fMRI study of decisions incorporating multiple attributes.