515 



^ -35- 



x • 



Logleatl^, CO* would txpMt ch« p«rc«aus« prtf«r«ac« (or om 

 clt«r«CC« ov«r ^S»ir to rtfltct ratpond«nc«* eT«r«ll oplnioo of ch« 

 oa« r«l«tlv« c^-tht o(h«r. As already lodicaccd chit wa« ooc ch« e««« 



with ton* clutter^' To (•ciliVats coi>p«rl«OQ b«tw««a p«rccatag« 

 preference aod overell jinxes Figure 3 (hovt a plot In which the 

 overall opinion ratings fo^igarattc E have been subtracted froa those 



for cigarette F to give the "L Oyerall Oploion": where this Is negative 



cf - ' . 



,- respondents had a higher opinion off, where it is positive they had a 



higher opinion of F. Comparison of Flgore 2 with Figure 3 shows a 



Dufflb^ of anomalies: for iostince, S6J of'<(Ujster 3 said they preferred 



clgarettcC^f SQ<) ycc oo average this sane clus^«£ had • higher overall 



oplntoo of ci^ette E; only 532 of Cluster 11 saTtf^hey preferred F 



compared vlth ^t^T^ Cluster S, but In terns of overal(]>}piiUott 



Cluster 11 favoured '^^o a slightly greater extent than d^ Cluster S. 



There are two possible explanations for this apparent anomaly: 



Explanation 1 - A considerable proportion of respondents gave the 



non-preferred cigarette a Hu^er overall rating than the preferred 



cigarette. r$ 



Explanation 2 - The percentage prenMace figures may conceal the 



strength of the preference for one or ot^er cigarette. For lostaoee, 



those respondents in Cluster 3 who preftrr^^igarette F (S6Z) 



may have preferred it only slightly, whereas cniose who preferred 



E (44Z) may have preferred it strongly. / ^ 



^C^ c 



To decide between these two alternative explanations, a^'ltional is* 



o 

 computer analyses were undertaken in which the stated prcferenctT^ ^ 



