517 



-37- 



Of t^M vbo mU tti«7 pctftrrcd eigtrtcc* C oolj 6Z (16 ouc of 

 V . 



2S6 r«ipoQdc«C«) s«v« « blglMr ovcrtll rtclos Co 7, tad for Cho vtsc 



>C chctii 



■Ajorlcy of chcf^JlS omC of 16) cbo diff«r«ac« la ridns vsa ooly 1° u&ic. 

 Of cboce vbo lui} cb«y preferred P oolj 5X gavt a higher overall raclog 



CO E, and for Che ^jorlcy of efaesa (11 ouc of 14} cha dif fareoca vaa 



O 

 only 1 unit. For Cha coaparaClvely snail nuaber of retpondeaCi who 



expressed no preference all-^uC ooa raced Che cigarecces elcher equally 



or differing by one unic. "*■>%, 



These results would sees Co ?ivposa of Explanacion 1. 



'^^ To seek evidence for Explanacion {.ic la necessary co look at Che 



resuC^a by cluster. Table 16 shows, for '^l^h cluaCcr in turn, Che 



average Mwrall racings given Co che Cwo cigarit^es by chose who preferred 



E, and Che »M) for chose who preferred F, wich Lfl(i>^ach caac Che dif fcreaca 



between th« Cwo<d^E) expressed aa che "t Overall OpraLoa". Cooparison 



of the ncgatlvt L ic^ytt for those who preferred E with t^ posldve A 



scores for those wbo prfttArred F, cluster by cluster, gives some indication 



of the relative atreofCba oC^cferenca for these cwo groups of respondeat. 



For chree of cha Low Need clusc6ca (1, 3 and 9) che negative A scores '*>-'' 



outweigh the positive A acoraa, indicating that che strength of preference 



for respondents in chesa chree eluaters ^ho preferred E was greater 



Chan che scrength of preference for chose wtis preferred F. For the 



High Need clusters (11, 12 and 10) the coov«rse<>^ true, those who 



preferred F doing so nore decidedly than thost who preferred E. These 



results support Explanation 2. 



