The functionalization process is not restricted to linguistic transaction, we can also apply the concept of functionalization to other types of behavioral transactions, present in hominids. The earliest ancestors of functional vocabulary need not have had any specifically linguistic characteristics. There is no reason to suppose that for hominid species, vocal sounds - perhaps proto-musical sound streams - could not carry socially significant information. Since the character of such streams would be affected by the physiology of their authors, they would be a common individuating voice print of members of a group. Opportunities for novel exploitations of such productions arise with activities involving sudden motions or alterations of motion that have similar effects on such a stream for any member of a group.
This, of course, is not the beginning of language, nor need we suppose that it is even part of the early story of language, but it is an illustration of what a pre-linguistic precursor of functionalization might have been like: In the imagined account, non-linguistic sounds naturally produced with available initial benefits acquire a role in which they contribute to the safety of groups by providing a means to distinguishing members from non-members, and individuals from the group. Derivatively, therefore, it contributes to social cohesion and other advantageous conditions. This is already an instance of a process of the same general kind as functionalization, that is, an exploitation for a novel application, of a feature preselected for some distinct earlier role. In Stephen Jay Gould's [25] language, it is an exaptation.
Merely to have said this much hardly makes it less mysterious that such changes should occur, particularly later. In what we have so far described, we find three stages and two transitions between them. One stage is from voiced sound production as a natural expression of physiology which transitions to voiced sound production as social tool, such as the soothing sounds of a mother, and constitutes an other stage in the use of sound production. There is also a stage in which voiced sound production as identificatory device has accidental features - exertion sounds, eating sounds and so on. These activities can be distinguished by sound alone without the support of visual cues. At some point we can imagine a transition towards the exploitation of these auditory discriminations in the artificial production of effects, that is, in the absence of physical and perceptual cues that give rise to these sounds in the first place.