Halliday & Matthiessen (1999: 533-4):
We have referred all along to the primarily visual nature of human experience: how much of it is constituted as location, and especially movement, in space. Now, both gestural and vocal resources involve positioning and moving the organs of articulation in space; but the position and movement of the vocal organs, besides being largely out of sight of the listener, is very much mote constrained; and while this permits a limited degree of iconicity (association of close vowels with 'small', open with 'large', for example) this can never be more than a marginal feature of the system as a whole. Thus even allowing for the additional iconic potential of loudness and length, the expression systems of spoken languages must remain prototypically conventional. This is the familiar principle of the "arbitrariness" (i.e. conventionality) of the linguistic sign.