Halliday & Matthiessen (1999: 534):
Gestural systems, by contrast, have a far greater potential for construing experience iconically. Thus in Johnston (1989: 16) "signs are roughly graded into four classes of transparent, translucent, obscure and opaque signs, depending on how iconic a sign is"; and while most signs fall in between the two extremes — Johnston grades them as obscure or translucent, rather than opaque or transparent — many of those he labels "obscure" have a popular explanation in iconic terms (e.g. 2388: n. CAMERA, v. TAKE A PICTURE, PHOTOGRAPH. Obscure action. Popular explanation: 'holding a camera and depressing the shutter button' [p. 301]). This suggests that even if particular explanations are "nothing more than deaf folklore" (p. 16), the system as a whole is perceived as prototypically iconic; and this feature is borne out in two important respects.
One is that many of the signs construing basic categories of experience that would be learnt very early in childhood, in the transition from protolanguage to mother tongue — examples are 1291 GET; 1479 HOLD; 1824 RUN; 2473 BIRD; 2759 DRINK, CUP; 36 BED; 163 UP — are clearly iconic, and so would tend to establish iconicity as the norm.
The other is that individual signs may be modified in a distinctively iconic fashion; e.g. 1471 "v LARGE, BIG, (with amplification) great, (with amplification and stress) enormous, huge, immense"; see in particular the section on "sign modification" in Johnston (1989: 494-9). As Johnston comments (p. 513), "A language which is itself visual and spatial has far more opportunities than an auditory one to map onto itself those very visual and spatial qualities of the world it wishes to represent".