Studies investigating developmental synaesthesia have sought to describe a number of qualities that might capture in behavioural terms the defining characteristics of this unusual phenomenon. The task of generating a definition is made more difficult by the fact that any description of synaesthesia must be broad enough to capture the 61 different variants of the condition already reported to date. Given these difficulties, the current literature now contains a number of conflicting assumptions about the nature of this condition. Here, I attempt to address several of these divisive areas from a set of contemporary definitions. I present evidence that might argue against previous claims that synaesthesia is (a) a 'merging of the senses', which (b) gives rise to consistent synaesthetic associations over time, with (c) synaesthetic associations that are spatially extended. I then investigate the possible benefits of moving from a behavioural definition to a neurobiological one and explore the ways in which this might force a rethink about the potential outermost boundaries of this fascinating condition.