Professor Hobson (Reference Hobson2003) argues admirably for the continued relevance of psychoanalysis in a mainstream psychiatric journal. But is his suggested rapprochement between psychoanalysis and contemporary neuroscience really desirable?

Contemporary neuroscience as illustrated by his example of ‘mirror neurons’ typically assumes an ‘empiricist’ worldview. In brief, imitation is assumed to be an acquired process in which information is abstracted from experience using associative learning. The current focus is on the anatomical location of the associative learning responsible for imitation (Reference Rizzolatti, Fogassi and GalleseRizzolatti et al, 2001).

In contrast, psychoanalysis derives from an older, rationalist philosophical tradition. It assumes the existence of both innate beliefs, such as persecutory anxiety, and distinct mental mechanisms, such as introjection or Klein's paranoid–schizoid position, that do not rely on associative learning.

These two philosophies have been in tension for centuries. One option is to make psychoanalysis more empiricist by down-playing the innateness and divergent mental mechanisms of classical theory. This is seen in attempts to incorporate ‘theory of mind’ deficits into a psychodynamic understanding of mental states (Reference FonagyFonagy, 1991).

But will associative learning form the secure basis for understanding the mind that empiricism proposed? Practical attempts to develop this philosophy, such as neural networks, are not actually very good at accounting for how we manage to form and fix true beliefs. Rationalist accounts of higher brain functions such as language do exist (Reference ChomskyChomsky, 1959). Given such uncertainty rationalism continues to merit the foothold within psychiatric practice that psychoanalysis provides. Too much mutual understanding between psychoanalysis and contemporary neuroscience might not be good for psychiatry.