487 



PUCEMENT OF CIGARETTES ^' 



Th« «bev« alcocin* d*livtri«t art averages aa a rasult of aaokiaf 



40 elsarttcat •( each aaapla (8 pores, S cliaratcas par port). Mora 



recently, saaplctjr^nd P bare b««a axaaiacd la (rcatar detail in ordar 



Xa ^ *• 

 to give iofonri^oQ about the distribucion of nicotine deliveries vithia 



each sanpla. For t^s purpeae the nicotine deliveries of M individual 



cigarecces drawn at raOp> froa each laoiple were deccrvioed. Detailed 



.^ ■ 



results are given in AppeiMF^ I. The averages, which arc considered 

 nore accurate than the figures quoted la Table 1, were: 



I - l.JATac/cig 

 y F - 1.81 ag/ci^ 



Of 'difk 12 clusters identified in the surV^ (1) it was decided to 

 include all c^ept Cluster 2 in the consumer test'.J^Clustar 2 was 

 eliainated b«cau£^the cluster comparison program indicated a rather 

 low degree of aatchiif^ between the two halves of the ssnpie in respact 

 of this cluster; It was,'^ other words, a very diffuse cluster. 



With 4 products (il« t/d, Xi) to be inter-compared in the first 



consumer test, and with the stipu^tion that every respondent should 



smoke all 4 products, it vat necessary to carry out successive paired 



comparisons. Ideally, every respondent s^uld have smoVed the products 



in all possible combinations of pairs (AB. XC» AD. BC, BD, CD). As it 



was felt that three paired comparisons was the absolute maximum that 



respondents would agree to it was decided to adopt an incomplete design 



and to analyse the results, insofar as preference was co<(cerned, using ^ 



the "Round Robin" approach (see next section). This necessiUt^ed m' 



o 

 dividing the sample of RSFT sooVers, and also the sample of KSFf ^.smokers ^ 



CB 



