Nooteboom (1975, 1983), Tatham (1970) and several other researchers introduced perception as one of the speaker’s principal goals: speakers aim to be perceived; almost all speech production models have adopted this basic idea to account for perceiver-oriented variability in the acoustic signal. An example of a wholly speaker-oriented model can be found in MacNeilage (1970) where the goal or target is the articulation, with gamma loop control of contextual variants. TM has a different take on this: the goal is to provide a signal optimised for perception, where what is optimal may be a variable, since the listener’s ability to perceive that is, assign an appropriate labelling — is itself a variable depending on context and environment.