• • VlC' W . I I I sout ern rtca .'e ore, ur, in -, ' - an a ter t e .-. ast .·.· acia - ' . - • ax1mum JOHN PARKINGTON • • Introduction - Ho· ocene prehi tory in · outh rn · · frica is r1 asonably well researched, in p, art becaus , e a u , ·tantia nu ber of excavat , ed rock e ter sequences preserve bone and plant mat rials in informa · tive contexts. Regional research des . igns h . ave h,elped us go b , eyond sequenc· construction and stone tool typolo, gi es . Within these reg · ion , al ram · wo , r k s a small series of sites preserve materials fro , m e , arlier time-periods, , notably the millennia since the Last Glacial Maximum (LGM). Thus it is now p, ossible to, try to interpret earlier very partial patterns against the more detaile , d and ext , ensive info , r· mation in Holocene contexts. We · m . ay be able to make only modest statements abou . t the Terminal Pleistoc. ene but we can s.e , e how dierent it was from later mil· lennia and thus sugg , es · t alternative patterns of behaviour. In such inter . pretations, c , ov . ering as they do the transformations of local clima . t·es from the extreme of about 20 · 000 years ago, to the present, the role of -n v ironm ntal changes . in stimulating b , eh , av ioural changes is un,derstandably high lighted by many researchers. Thus unde : rstand ing th end of the P · Ieistoc · ene necessarily involves b ,oth p ala eo, e n viro nment , a l an 1 d so, cial recons , tructions. In this survey I adopt a three · -part strateg · y , . First, I present the · patterns of site u : s 1 e fro , m th Elan i ds Bay Cave, whic · h has an episodic history of occupation since th · beginning of the Late Pleist · O · cen , e (Parkington 1 ' 980, Volman 1984). After that I generalize to caves and rocksh It rs elsewhere in southern Africa by considering more brie · ly sites with similar 1 dep 1 osi tional histories. And, thirdly I move beyond the conines of cav·es an 1 d discuss patt,erns o , · f settlement in th , e subcontinent again : st what is kn , own of p , al a 1 eoenvironme nts since abo , ut 20 000 years ago. I precede thi , s with some ge · neral remarks on the kinds of , data available , and my a · ttitud· to them . •