Concept-formation can be explained without appealing to non-physical factors. I am not arguing that perceptions are conceptions. But what I am suggesting is that conceptions are possible representations of percepts by dint of natural language and neural networks. George Lakoff and Mark Johnson (based on the work of Eleanor Rosch) have argued that we categorize things (e.g. birds, chairs, tables, and teapots) according to fuzzy prototypical characteristics (cf. Metaphors We Live By). However, the prototypes by means of which we classify various objects emanate from our somatic experiences. For example, although I had never seen a Highland Cow until the year 2002, I knew that it was a cow upon first sight. Yet the Highland Cow was unlike any cow I had ever seen before.
Lakoff and Johnson would say that I was able to associate the Scottish cow with my prototype of such animals based on past somatic experiences. But my concept of cowness is nothing more than the result of what I've perceived with my sense organs (inter alia) and how my brain represents the usual properties of bovines which it derives from distal and proximal stimuli. There is no need to posit a soul in order to account for concept-formation, language or consciousness.