Since they were first described by Meyer and Land [1] the classification of concepts as 'threshold' concepts has engaged many researchers, including a number of CS researchers. A variety of approaches have been employed to identify concepts that could be classified as threshold concepts, with varying success. Our own frustrations in identifying them led us to identify shortcomings in commonly-used approaches, and to the promising possibilities offered by a new direction. We describe that new direction here, and detail the path that led us to it.