Halliday & Matthiessen (1999: 17-8):
We need, therefore, a further guiding principle in the form of some model of the processes by which meaning, and particular meanings, are created; let us call these semogenic processes. Since these processes take place through time, we need to identify the time frames, of which there are (at least) three.
(i) First, there is the evolution of human language (and of particular languages as manifestations of this). Known histories represent a small fraction of the total time scale of this evolution, perhaps 0.1%; they become relevant only where particular aspects of this evolutionary change have taken place very recently, e.g. the evolution of scientific discourse. This is the phylogenetic time frame.
(ii) Secondly, there is the development of the individual speaker (speaking subject). A speaker's history may — like that of the biological individual — recapitulate some of the evolutionary progression along epigenetic lines. But the individual experience is one of growth, not evolution, and follows the typical cycle of growth, maturation and decay. This is the ontogenetic time frame.
(iii) Thirdly, there is the unfolding of the act of meaning itself: the instantial construction of meaning in the form of a text. This is a stochastic process in which the potential for creating meaning is continually modified in the light of what has gone before; certain options are restricted or disfavoured, while others are emprobabled or opened up. We refer to this as the logogenetic time frame, using logo(s) in its original sense of 'discourse'.