Tobacco Documents Online

ltbceun-wm . -TecEad- rb.,tt18.-tft9.f79tt38 -TC1ab l w G-b N-AV A Paired'' Comparisons Test Procedure for the Sensory Evaluation of S'trength, of Flavor and Ease of Draw of Cigarettes J, Solnts, F. Escfter and H. R. Roth i h? J. Solms and!F. Escher: Department of Food IScience, Swiss Federallnstitute of Technology, CH-8092 Zurich (Switzerland) H. R. Roth: Group of'Biometryand Theoretical Population Genetics, Department of Animal Production, Swiss Federal Institute of'Teehnology, CH-8092 Zurich (Switzerland) .(Rrceived April219,1983; Accepted May37,1983) The application of a ipaired'compattisons test procedure for the evaluation of setuory parameters of cigarettes is described: Ae parameters perceived strength of flavor and ease of draw were measured ion a non-gradt.ated linear scale. Four typical one mg tar filter cigarettes commercially availablcon the Swiss market were investigated: T7u'roleplayed'by tltefilterand the tobacco rod was investigated in detail by.'exchanging fjllers. The panel'comprised'52'persons. For thrstatisticali evaluations the sign test'and'the one-sample rtest proced,.re were employed.' The method permits a reliable evaluation of'tlrescparameters for different brands. 1. Introduction The sensory quality of cigarette smoke is a very important parameter in smoking:. Direct evaluation of this parameterr can be done only in sensory tests. Sensory evaluations are performed for many purposes. They are'e.g, carried'out in the well known area ofi characterozing food1 quality (1, 2, 3', 4)1 but they are also applied for characterizing',tobacco prod- ucts (5;, 6, 7; 8; 9). The practice of using groups of smokers for the assessment ofovera111cigarette quality or of specific quality parameters is now widespread within the tobacco industry, However, relatively little information coneerning, methods and data generated on sensory tests of cigarettes has been published. There is atI present ai need for an appro- priate testing method for the sensory'parameters of'cigaret- tes The investigations reported in this paper will describe', the application of,a paired-comparisons test procedure. The procedure was devised to meet two criteria, namely (1)'ease, of participation by relatively untrained participants, and (2)', sensitivity to even,small differences in the sensory attributes of interest'. A restrictedl number of commercial, cigarette brands was employed in the test. From a large number of possible, parameters„relative strength of flavor and ease of draw upon, smoking cigarettes were selected as interesting ones for the investigationi As a~further possibility of'applicationlof the method, the role played by the filter and the tobaeco rod in perceiving tobacco strength and ease of draw was investigatedin detail. For this purpose severallcigarettes with interchangedlfilters and rods (hybrids) were prepared and used in the experiments. 2': Experimental Part 2.1. Test products The four leading one mg,tar filtcr cigarette brands commer- ciailh aM1•ailable on the Swiss market were employed. All four brands have different filters and tobacco rods and are abbreviated with the symbols A/a, B/b, Clc and D'ld. The capitals (A, B, C, D) stand for the filters,,the kttets in small case (a„b, c, d) for the tobacco ~rods: Thus, for example. B/b stands for the Barday ' itictron" filter (fi)' with the Barclay tobacco rod (b), and the symbols A/a, Cic and Did represent the remaining three brands with their conventional filters. For test purposes, filters were exchanged according to a scheme laid'out below. The'interchanged samples are called "tiybrids". All'.cigarettes were masked so as to appear virtu- ally identical, and' were labelled with three-digit codes. Ailn samples used! in the experiments were provided by a Swiss cigarette manufacttuer: - 2.2. Method and Panel The sensory evaluation! was conducted in the foratof paired comparisons amongst, a panel l of regular cigarette smokers broadlj+ representative of'smokers in Switzerlandl Paired comparisons were chosen in order to facilitate the task of the panelists: Panelists were,invited,to:attend a series of testing sessions in an appropriately furnished and ventilated room that permitted individual and' undisturbed smoking assess- ments. For the panel work the technical facilities of a com- mercial! organization were employed (llink Ltd., Lucerne, Switzerland)- Only two pairs of'cigarettes were tested in one day: Each pair was presented on a clean ashtray.,The order of presentationn was alternated througNouti the test series. The panelist was asked to smoke each of the two cigarettes to about half its length, and to filll out a response sheet at the point during smoking when his perceptual judgments of' draw and strength were formed. The, selected parameters were restricted toitwo descriptors in onder to avoid overloading the discriminative capacities of the panelists. Afterromple- tion of', the first, pair, the ashtras,. butts and response sheet were removed and replacedlwith a second painaftera pause of 10 minutes, The pa!nelist'. was allowed to drink water in between„but not to eati or smoke anything. For comparison purFxnses, some pairs were prastntedltwice to assess consisl 185 2029397fi81

I.vrul~. 16' (49671 So ~. • tency of the panel response. Two pairs of identical testr dgarettes were. however, never presented on the same day. 'The questions and scales covered strength first and ease of draw second. In both cases the panelist a) deciced whether or not he noted any difference between the two cigarettes; b) i if so, he identified which of the two seemed stronger/ easier to draw by noting the,appropriate three-digit code; and c) indicated the extent of difference using a non-graduated linear scale of 110 cm kngth„marked as follows: very small very large di:ference' difference Each panelist attended smoking sessions at a similar time of day. All panelists tested i all I pairs, presented individually in random order. In the interval betwee n smoking trials, panel- ists were free to talk about matters unrelated to the test. Panelists first underwent a, preparatory session where thee parameters strength, and ease of draw were 'discussed, and, the use of the scales explained. During this session, cigarettes which demonstrated marked differences in draw and' trength were available so that panelists could familiarize themselves with their task. These cigarettes were of coursee different from any involved in the test itself. The panel Iof 52 panelists comprised regufar, smokers (rchree' or more cigarettes par day of I all tar categories, and age/sex quotas as follows. Sex: 27'males; 25 females. Age groups: 11 persons (6 m + 5' f) aged 20ito 24 years,15' persons (8 m 1+ 7'f)aged 25-34 years, 14 persons (5' m+ 9 f) aged 35 to 49 years, and 12 persons ( 8' m+ 4 f) aged 50 years or more. Total'duration of freldworkk was four weeks. It was an impor, tant oondition th v t' t fi i i d i ll ' a e ery pane st part c pa e n a tes t '~series, so that the data of one panelist could be treated as one' block in'thestatistiral evaluation. This allowed caltttlation off individual differences. In one set of experiments, the com- mercial form B1b was compared with the commercial forms of' AGa„ C/c and! DYd. In a second set of experiments, com- mercial forms of'cigarettes and hybrid forms were'eompared in such i a way that the role played by filter and tobacco rod of Blb in its differentiation vs the remaining cigarettescould be' assessed. For statistical evaluations •the results were first grouped, ( -caording to the direction of the difference and evaluated in a sign test. Then, the marked distances on the scales weree measured in centimeters and,ttie results evaluated with the one-sample t test procedure using the BMDP3D program (10): 4. Results 4:1. Strength of flavor The data of the pairedlcomparisons for,commercial samples BJb vs A/a, C/c and D7d are presented in Tab.1land'indicate the differences between brands (filter +tobacoo) The t test gives differences in strength in fasonof B/b vs A/a, C/c and' D/d with 3.60 to 4'.88' scale un`tts and with P=valhes <0.0004. The sign test gives similar results; 94.1 to 98.1 K% of panelists found B1b stronger than the other samples. These comparisons were also highly significant. The pair 1" was a repetition of pair 1,,made ata,laterdate'for purposes of comparison. The data of'the paired comparisons:foroommereial samples and lhybrids are presented in Tab.2.'Rte first section (pairs 4 and 15 with the' corresponding repetitions 4' and 5')' give an indication of the differertces of frlters B' and A on tobacco b. (pairs 4/4") and tobacco a (pairs 5/5'). The results for both pairs indicate a idifference in strength in favorof'fitter B over filter A with tobacco b (2.41 and 3.13 scale units„ respec- tively) and with tobacco a(1.98'and111.43 scale units, respec- tively) at a high level of significanoe (P'<0.0001). The sign test gives similar results with over 74% of the panelists in favor of filter B, an outcome with a high level of significance.. The calculated I difference of, filter effects is small and! not significant with pairs 4L5; however,,it amounts to 1.7 scale' units, and' achieved a high level of'signifieance with pairs 4"-5'. The second and third sections in Tab.2 give an indication of'f the differences of fifters B'and C on tobaccos b andic (pairs 6 and 17) and of filters B'and D on rtobaoeos b and d(pairs 8'and I 9). The data for all series indicate a difference in strength ini favor of filter B over filters C and D on the tested tobaacos with scale units ranging from 2.25 to 4.81 and achieving',high levels' of significance (P<0.00101): Similar results aree obtained in the sign i test. The calculated difference of filter effects for pairs 6-7 amounts to 2:26' scale units at a high, leveliof significance (p<0:0001)i The caleulatedidifference' with pairs 8-9 is small and not signifrcant. The fourth section (pairs 10'and 11) gives an1 indication of' the differences of tobaccos b andia with filters B and A. Thee data indicate a difference in strength in favor of tobacco b over tobacco a with both filters (scale units 3.04 and i 2.42„ respectively), ani outcome with' a highI level of'significance. (P<0.0001). Similar results are obtainedl in the sign test. The calculated difference of pairs 110-11, however, is small I and i not significant. With pair 12 the',diff erence between the two hybrids A/b and' Blal is examined. The presented data indicate no difference between these hybrids with either statistical I test. 4.2. Ease of draw The results of' the paired comparisons for commercial sam- ples BJb vs A/a, C/c and D/d are presented in Tab.3' and indicate the diff,erences between brands (filter+tobaeco). Both t tests andlsign tests gave diffk'rences:in ease of'draw in favor of Bib over A/a, C/c and Did, although,not,aliways at a very high level of significarnce_ The t test: scale units 2.13 (P = 0:0040) to 3.04 (P < 00 .00'01); the sign testi: 71.2'% (P=0:0018) to 90.2% (P<0.0fxT1) of panelists in favor of Tab.1' Paired comparisons for,eommercial samplbs-strength Pair N Comparedlsamples N tTestoflscoreddifferences Signtest Remarks. 91 r. Code for Effect mean s.e.m. P X>Y X<Y X=Y P' [A X: } (irr) , (',1%r) (~'~) ~ ll 1' BI bvs A-a (B'-A)i+'(Ib,a) 52 52 4.76 4.88 0.406 0.367 <0 0iD0il <0.0001 96.2 98.1 1.9 0 1.91 1.91 <(1 ~:000 l <(1':0001 Differences between brands 2 B' b vs' C c (B-C) +(b-c) 51 4.24' 0.387 <0.0009 98.0 0 2_01 <0~:000'1' . filter and tobacco '* `~ 3 B''b %s' D d (B'-D) +(b-d) 51 3.60 0.355' <0.0009 94.1 3'.9 2,.0i <'~0!D00'1I I R~h~.

M:.oU 16 (19e01i?b.4 Tab.2 Paiced comparisons for commerc+ial samples and hybrids- strength Pair Comparedsamples N t Test of scored differences Signtest' - Nr: Code for Effect mean s:e.m: P' X>Y X<Y X=Y P' x y (%) (%) (%) 4 B/b vs A/b (B-A)e, 50 2.411 0:375 <0.00011 74.0 10.0 16,0 <0.0001 4` 51 3.13 0373 <0:000Ii' 90.2 5.9 3.9 <0.0001 5 B/a vs A/a (B-A),, 52 1.98 0;260 <0..O00I1i 84.6 3.9 1'1.5' <0.0001 5. 51 1.43 0:236 <0.000I1! 74.5 0 25.5 <0.0001 4-5' (B-A)b-(B-A)j 50 035 0:452' 0.4423 4'-5' 51 1.70 0:407 0.00011 6 B/b vs C/b (B-C)e, 52 4.811 0:359 <0.000I1' 98:I' 0 1.9 <0.0001 7 B/c vs C/c (B-C), 52 2.54 0:376 <0.000'1i 90:4 3.9 5.8 <0.0001 6-7 (B-C)6-(B-C)t 52' 2.26 0463 ' <0.0001' 8 B/b vs D/b (B-D)b 50 2.25 0:391 <0.000'1'. 90.0 6.0 4.0 <0.0001 9 B/d vs D/d l(B-D)a 51 2.54 0314 <0.0001 90:2 3.9 5.9 <0.0001 8-9 (B-D)b-(B'-D)g 50 -0:06 0:432' 0.8924 10 B/b vs B/a (b-a)8 52' 3.04 0:385 <0.0001 86.5 3.9 9:6 <0.0001 11' A/b vs A/a (b-a), 52' 2.42 0.352' <0.0001 86.5 1.9 11.5 <0.0001 10-11, (b-a)B-(b-a)s 52 0:62 0!458 0.1801 12 AJb vs B/a -(B-A)+(b-a) 52' -0!17 0:416' 0.6889 42.3 38.5 19:2 0.8776, Tab.3' Paired comparisons for commercial samples - easrof draw Pair N Compared samples N t Test of scored differences Sign test, t.. Code for Effect mean s.c.m. P' X>Y X<Y X=Y P x y (%) (%) (%) 1 B/b vs A/a (B-A)+(b-a) 52 2.13 0:705 0.0040 71.2 26.9 1L9 0.0018 1' 52 2.57 0587 <0.00011 76.9 19.2 3.9 <0.0001 2' B/b vs C/c (B-C)+(b-c) 51 2.44 0:625 0.0003 72:6 25.5' 2.0 0.0009 3 B/b vs Did (B-D)+(b-d) 51 3.04 0:382' <0.000'1 90.2 7.8' 2.0. <D.0001 B/b: Here again, pair, 1' is aircpetition of'pair l, made at:a later date for purposes of comparison. The results of, the paired! comparisons fori commercial sam, ples and hybrids are presented iniTab.4. The pairs 4 to 9 give an indication of the differences of filter B versus filters A, C and D with tobaccos b, a„c and d. These results for ease of Remarks Differenee of fdters B and A on tobacco b Difference of frfun B and iA on tobacco a Cale: difference of filter effects Differenee of jrlters B and ~C on tobacco b Differenceof filters B and C on tobacco c Ctlb: difference of filtereffects Diiterenae of jr[ters B and, D on tobacco b Difference of flrers B and D on tobacco d Calee differenoe of filter effects Differenae of robaacosb and la , on fulter B Differenee of tobaocos b and a on filter A C2fe: difference of tobacco effects Difference between hybrids Remarks Differences ~ bet ween brands t.d, filter, and tobacco ~ ~, 4, dravv, are also in favor of filter B. Wowever„in bothi*ests the ~ measurable differences are smaller and the levels of signifi- cance are lower, if significant at alll The calculated I differ- tM ences of filter effects are small and not signifiicant, with pairs 4-5 and 8-9' The calculated difference of filter e(fects for pairs 6-7' arnounts to 2.05 scaie units with a high level of 1R7

1.1. rol.. U6~/09&b ~~No,4 Tab:4' Paired'c+omparisons for commercial samples and hybrids-easeof draw Pair Compared samples hJ IVr. tTesrof scored differences Sign test ' Remarks Code for Effect mean s.e:m: X Y 4 B/b vs A/b (B-A)b 50 0:94 0:567' 4' 51 1.53 0:454 5 B/a vs A/a (IB-A),, 52 1.70 0.486 5' 51 1.411 0.470 4-5 (B-A)6-(B-A'),, 50 -0.77 0.498 4"-5'' 51 0.12 0.399 P X>Y X<Y X=Y P M I M ' (K)' 0.1048 67:3 28.9 3.9 0.0066' Difference of 0.0014 714.5 19.6 5.9 <0.000 f filrers B and A on tobacco b 0.0010 69:2 28'.9 1.9 0.0046 Difference of', 0!0041I 663 25.5' 7.8 0.0031 frlters B and A on tobacco a 0.1308 Calt: difference 0.7695 of filter effects 6' B/b vs C/b (B-C)b 52' 3.34! 0.611 <0.0001 75.0 23.1 1.9 0.0002' Differenceof' filters B and IC on tobacco b 7 B/c vs C/c (B-C), 52' 1L29 0.529 0.0179 69.2' 28.9 1.9 0.0046 Differenceof filars B'and IC on tobacco ic. 6••-7' . (B-C)e-(B'-C)c 52 2.05' 0.447 <0.0001 Calc.differenee of fiiter effects ~ 8' B/b vs D/b (B,D)6 50 7:95 0.419 <OA001 82.4 15'.7' 2.0 <0.0001 Difference of filters B and D on tobacco b 9 B/d vs D/d (B'-D),a 51 2'.24 0.427 <0.0001 74.5 17.7 7.8' <0.0001 Difference of ftlters B and D on tobacco d' 8'-9' (B-D)b-(B-D),a 50 0:71 0.449 0.1221 Calc. difference of filter; effects 10 B/b vs B/a (b-a)B 52 0:44 0:379 0.2568 51.9 30.8 17.3 0.1263 Difference of tob'acros b and a on filter B. 11 AJb vs A/a (b-a)A 52 1.91 0:353 <0.0001 67.3 15.4 17:3 <0:0001 Difference of' tobaccos b and a on filter A 10-111 (b-a)e-{b-a)X 52 -1.47 0.465 0.0026 CaIC: difference of tobacco effects 12' AJb vs B/a -(iB-A)+(b-a) 52' -0.32 0.590 0.5888 36.5' 53.9 9!6 0.2430 significance (P <0.0f)El0); Also in this series the pairs 4' andl 5' were repetitions of experiments 4 and 5, made at a later date for purposes of comparison. Paiis 10 and 1 l gi ve an indication of'the dif, ferences of tobac- cos b and a on filters B and A. The results indicate a differ- ence in ease of draw'in favor of tobacco b vs tobacco a,only with filter A in both tests, With filter B (pair 10) the differ- ence is small and not significant. With pair 111„ however, the difference in the t test amounts to 1.9'1i scale units (~P <0:0001): The calculated difference ofi pairs 10-1',l amounts to - 1.47 scale units in the t test (P = 0.0026). With pair 12 the differenccbetween the two hybrids A/b and B/a is examined. The data indicate no difference between these hybrids with either statistical testi 5. Discussion, The results presented in thisstudy indicate that reliable data on the sensory quality of'cigdrette smoke can be obtained with a paired-comparisons test procedure. The sensitivitM1 of Difference, between hybrids the procedure seems particularly impressive, in view ofi its ability to reveaf differences among,very low delivery, prod- ucts; whereitmaybeexpectedthatdi6ferentiatio'nsbetw~cen products is more difficult. Furthermore, even a parameter, which may be diffir:ult, to evaluate, like "ease of draw", can be judged successfuly with the method. A group of 50 to 52',persons seems tobe suitable for this ty,pee of panel work. A brief preparatory session i for the panel is d'esirable; and may help in obtaining, replicable results. A paired comparisons test with a restricted number of,onNytwof descriptors was chosen in order to avoid unwanted carry- over effects and overloading of the discriminative capacity of the panelists. There was no indication, of any of these unwanted effects, although this has not' been examined in detail. The evaluationiof the results,with the one sample t test and sign test is reliable and most practical. Paired!com- parisons with commerciali brandk permit an evaluation of overalll differences. Paired comparisons with samples with exchanged filters allow an additional evaluation of filter and tobaccrn effects. The I test permits further comparisons by calculating differences between pairs or of more than two 1 RR

1.M ~.•ot 161~19a61:No A pairs. The pairs presented twice (nrs. 1i'. 4' and 5*) gave results indicating a good replicability of' the method. With these pairs, the standard'errors of the mean were reduced. and in certain cases differences were more markedi The obtained results show marked differences in relative strength of cigarette flavor and in relative ease of draw; with sample B/b giving highest values for both parameters (Tab. li and 3). The results also show that it seems to be an easierr task for the panel to evaluate strength than iease of draw. Forr strength, most of the standard errors of mean are smaller, as compared withi the corresponding data for ease of draw.. Comparing, the data for commercial'cigarettes and hybrids, some intt:resting,effe+cts can be observed. A few will be dis, cussed in the following, starting with strength (see Tab.2). Looking,first at filter effects andlexaming the calculated dif- ferences for the pairs 4 to 9, it is interesting to note that differences of filter effects for the pairs 4 vs S' and 6 vs 7 are appreciable„bur no significant difference was obtained with 8 vs 9. Looking now at tobacco effects (pairs 10 vs a l)' it is interesting to findi out that there are only minor calculated differences between these pairs. These observations indicate that filter as well as tobacco rod in combinationi are of importance for the parameter strength of'flavor. A confirma- tion of'this observation occurs with pair 12. In this test, the commercial samples B/b and A!a are compared in the form of'the hybrids A,% vs B'/a. In this hybrid, form, there is no difference in strength. although in its original form B!b is stronger than Ala (pair 1). A simple interprctation of'this outcome is that here the grea- ter strength of'Blb derives about halfi from a tobacco effect and 1 about, half from i a filter effect: Similar observations can be made with the parameter ease of' draw (tab14). Looking first at filter effeas and examining the calculated!differences for the pairs 4 to 9, the following can be said. Significant differences in the extent of the filterr effect can, be observed I with pairs 6' vs 7; but neither with pairs 4 vs 5 nor 8 vs 9; indicating',an influence of the'tobacco rod. Looking then at tobacco effects (pairs 110 vs 1J'), it is interesting to note that:there is also a difference depending on the filters_ These observations indicate that filter as well as tobacco rod in combination are of importance for the parameter ease of draw. Aconfirmatiom of this observation can, be, obtained with pair 12: In this test the commercial samples Blb and A, a are compared in foraniof thechybrids A/b vs B a. In this hybrid form there is no difference in easeof draw; although in its original form B/b is superior in case of draw as com, pared with A/a (pair 11): In reviewing filter and'tobacco eJfiects, it can be stated that the pronounced relative strength in flavor and ease of draw of cigarette Brb is due to a combination of the properties of'the filter (B) and of'the tobacco rod (b). __ ACkOOMltdEements'. Thanks are due to Dr. W S. Cain, J. B, Pierce Foundation Laboratory and Yale University„ New Haven (USA) and to Dr. E. Kosrer; Psychological I Laboratory, University of Utrecht: Utrecht (Netherlands) for advice and discussion.. 6. References 1' AMERINE, M.A., PANGBORN. R'>M., and ROESSLER. E.B.. Principles of Sensorn• Evaluation of', Food. Academic Press. Nea. York;,1965: 2 AMERICAN SOCIETY' FOR TESTING AND MATERIALS. Manual on Sensory Testing Methods. ASLMiSpecial'Tecim. Pubtica. tion No.434, Philadelnhia, 1965. 3 VAISEY GENSER. M.,IWOSKOWI7Z, H.. SOLMS. Ji and ROTH. HIR:, Sensory Response to Food - A Sensory W'orkshop. Forster Publishing Ltd1, Zurich. 19,7: 4 JOHNSTON. M_R:. ed.. Sensory Evaluation Methods - IFT Short Course 1979/60, Inuitmrof Food Technologists. Chicago. 1'9'r9: 5 MORISHITA. 1., Statistical Quality Control.,12 (1p); 867'(,1!9bl )1 6 AKAKI. T., SATAK$ K.. YAMASAKI M: and MORISHITA. L, StatistiallQuality Control, Fl (10) 816 (n963) 7 KA'PLAN; A. R. and,GLANVILLE:,E. V., Nature< 202, 1366'(I1964) 8 ENDEMANN, W. and MERKER. J., Ber: Insn Tabakforschg. (Dresden)j ,/7: 34 (1!970) 9 CAIN, W.S., in GORI. G:B. and BOCK„F:G:, Banbun Report 3, A Safe Cigarette"Cold Spring Harbor Laboratory, Cold Spring Haborn N.Y., 1980. 10 DIXON, W.1_, ed_. BMPD Statistical Software 1981, Unin: of California Press. Berkeley, 1981_ 189