1、Designation: E2164 16Standard Test Method forDirectional Difference Test1This standard is issued under the fixed designation E2164; the number immediately following the designation indicates the year oforiginal adoption or, in the case of revision, the year of last revision. A number in parentheses
2、indicates the year of last reapproval. Asuperscript epsilon () indicates an editorial change since the last revision or reapproval.1. Scope1.1 This test method covers a procedure for comparing twoproducts using a two-alternative forced-choice task.1.2 This method is sometimes referred to as a paired
3、comparison test or as a 2-AFC (alternative forced choice) test.1.3 A directional difference test determines whether a dif-ference exists in the perceived intensity of a specified sensoryattribute between two samples.1.4 Directional difference testing is limited in its applicationto a specified senso
4、ry attribute and does not directly determinethe magnitude of the difference for that specific attribute.Assessors must be able to recognize and understand thespecified attribute.Alack of difference in the specified attributedoes not imply that no overall difference exists.1.5 This test method does n
5、ot address preference.1.6 A directional difference test is a simple task forassessors, and is used when sensory fatigue or carryover is aconcern. The directional difference test does not exhibit thesame level of fatigue, carryover, or adaptation as multiplesample tests such as triangle or duo-trio t
6、ests. For detail oncomparisons among the various difference tests, see Ennis (1),MacRae (2), and OMahony and Odbert (3).21.7 The procedure of the test described in this documentconsists of presenting a single pair of samples to the assessors.1.8 This standard does not purport to address all of thesa
7、fety concerns, if any, associated with its use. It is theresponsibility of the user of this standard to establish appro-priate safety and health practices and determine the applica-bility of regulatory limitations prior to use.2. Referenced Documents2.1 ASTM Standards:3E253 Terminology Relating to S
8、ensory Evaluation of Mate-rials and ProductsE456 Terminology Relating to Quality and StatisticsE1871 Guide for Serving Protocol for Sensory Evaluation ofFoods and Beverages2.2 ISO Standard:ISO 5495 Sensory AnalysisMethodologyPaired Com-parison3. Terminology3.1 For definition of terms relating to sen
9、sory analysis, seeTerminology E253, and for terms relating to statistics, seeTerminology E456.3.2 Definitions of Terms Specific to This Standard:3.2.1 (alpha) riskthe probability of concluding that aperceptible difference exists when, in reality, one does not (alsoknown as type I error or significan
10、ce level).3.2.2 (beta) riskthe probability of concluding that noperceptible difference exists when, in reality, one does (alsoknown as type II error).3.2.3 one-sided testa test in which the researcher has an apriori expectation concerning the direction of the difference. Inthis case, the alternative
11、 hypothesis will express that theperceived intensity of the specified sensory attribute is greater(that is, AB) (or lower (that is, A 65 % represents “large” values.8.3 Having defined the required sensitivity for the test using8.2, use Table 1 or Table 2 to determine the number ofassessors necessary
12、. Enter the table in the section correspond-ing to the selected value of Pmaxand the column correspondingto the selected value of . The minimum required number ofassessors is found in the row corresponding to the selectedvalue of . Alternatively, Table 1 or Table 2 can be used todevelop a set of val
13、ues for Pmax, , and that provideacceptable sensitivity while maintaining the number of asses-sors within practical limits.8.4 Often in practice, the number of assessors is determinedby material conditions (e.g., duration of the experiment,number of available assessors, quantity of sample). However,i
14、ncreasing the number of assessors increases the likelihood ofdetecting small differences. Thus, one should expect to uselarger numbers of assessors when trying to demonstrate thatsamples are similar compared to when one is trying to showthey are different.9. Procedure9.1 Prepare serving order worksh
15、eet and ballot in advanceof the test to ensure a balanced order of sample presentation ofthe two samples, A and B. Balance the serving sequences ABand BA across all assessors. Serving order worksheets shouldalso include complete sample identification information. SeeAppendix X1.9.2 It is critical to
16、 the validity of the test that assessorscannot identify the samples from the way in which they arepresented. For example, in a test evaluating flavor differences,one should avoid any subtle differences in temperature orappearance caused by factors such as the time sequence ofpreparation. It may be p
17、ossible to mask color differences usinglight filters, subdued illumination or colored vessels. Code thevessels containing the samples in a uniform manner using3-digit numbers chosen at random for each test. Preparesamples out of sight and in an identical manner: sameapparatus, same vessels, same qua
18、ntities of sample (see GuideE1871-91).9.3 Present each pair of samples simultaneously wheneverpossible, following the same spatial arrangement for eachassessor (on a line to be sampled always from left to right, orfrom front to back, etc.). Within the pair, assessors are typicallyallowed to make rep
19、eated evaluations of each sample asdesired. If the conditions of the test require the prevention ofrepeat evaluations, for example, if samples are bulky, leave anaftertaste, or show slight differences in appearance that cannotbe masked, present the samples monadically (or sequentialmonadic) and do n
20、ot allow repeated evaluations.9.4 Ask only one question about the samples. The selectionthe assessor has made on the initial question may bias the replyto subsequent questions about the samples. Responses toadditional questions may be obtained through separate tests forpreference, acceptance, degree
21、 of difference, etc. See Cham-bers and Baker Wolf (9).Asection soliciting comments may beincluded following the initial forced-choice question.9.5 The directional difference test is a forced-choice proce-dure; assessors are not allowed the option of reporting “nodifference.” An assessor who detects
22、no difference between thesamples should be instructed to make a guess and select one ofthe samples, and can indicate in the comments section that theselection was only a guess.10. Analysis and Interpretation of Results10.1 The procedure used to analyze the results of a direc-tional difference test d
23、epends on the number of assessors.10.1.1 If the number of assessors is equal to or greater thanthe value given in Table 1 (for a one-sided alternative) or Table2 (for a two-sided alternative) for the chosen values of , , andPmax, then use Table 3 to analyze the data obtained from aone-sided test and
24、 Table 4 to analyze the data from a two-sidedE2164 163test. If the number of common responses is equal to or greaterthan the number given in the table, conclude that a perceptibleattribute difference exists between the samples. If the numberof common responses is less than the number given in thetab
25、le, conclude that the samples are similar in attribute inten-sity and that no more than Pmaxof the population wouldperceive the difference at a confidence level equal to 1-.Again, the conclusions are based on the risks accepted whenthe sensitivity (that is, Pmax, , and ) was selected indetermining t
26、he number of assessors.10.1.2 If the number of assessors is less than the value givenin Table 1 or Table 2 for the chosen values of , , and Pmaxandthe researcher is primarily interested in testing for a difference,then use Table 3 to analyze the data obtained from a one-sidedtest or Table 4 to analy
27、ze the data obtained from a two-sidedtest. If the number of common responses is equal to or greaterthan the number given in the table, conclude that a perceptibleattribute difference exists between the samples at the -level ofsignificance.10.1.3 If the number of assessors is less than the value give
28、nin Table 1 or Table 2 for the chosen values of , , and Pmaxandthe researcher is primarily interested in testing for similarity,then a one-sided confidence interval is used to analyze the dataobtained from the test. The calculations are as follows:Pc5 c/nScstandard error of Pc! 5 =Pc1 2 Pc!/nconfide
29、nce limit 5 Pc1zScTABLE 1 Number of Assessors Needed for a Directional Difference Test One-Sided AlternativeNOTE 1The values recorded in this table have been rounded to the nearest whole number evenly divisible by two to allow for equal presentation ofboth pair combinations (AB and BA).NOTE 2Adapted
30、 from Meilgaard et al (8). 0.50 0.40 0.30 0.20 0.10 0.05 0.01 0.0010.50 Pmax=75 % 2 4 4 4 8 12 20 340.40 2 4 4 6 10 14 28 420.3 6 810142030 480.20 6 6 10 12 20 26 40 580.10 10 10 14 20 26 34 48 700.5 14 161824344258 820.01 22 28 34 40 50 60 80 1080.001 38 44 52 62 72 84 108 1400.50 Pmax=70% 4 4 4 81
31、21832 600.40 4 4 6 8 14 26 42 700.30 6 8 10 14 22 28 50 780.2 101220304060 940.10 14 20 22 28 40 54 80 1140.05 18 24 30 38 54 68 94 1320.01 36 42 52 64 80 96 130 1740.001 62 72 82 96 118 136 176 2280.50 Pmax=65 % 4 4 4 8 18 32 62 1020.40 4 6 8 14 30 42 76 1200.3 8 101424405488 140.20 10 18 22 32 50
32、68 110 1660.10 22 28 38 54 72 96 146 2080.05 30 42 54 70 94 120 174 2440.01 64 78 90 112 144 174 236 3200.001 108 126 144 172 210 246 318 4120.50 Pmax=60 % 4 4 8 18 42 68 134 2380.40 6 10 24 36 60 94 172 2820.30 12 22 30 50 84 120 206 3280.20 22 32 50 78 112 158 254 3840.10 46 66 86 116 168 214 322
33、4720.05 72 94 120 158 214 268 392 5540.01 142 168 208 252 326 392 536 7260.001 242 282 328 386 480 556 732 9440.50 Pmax=55 % 4 8 28 74 164 272 542 9520.40 10 36 62 124 238 362 672 11240.30 30 72 118 200 334 480 810 13020.20 82 130 194 294 452 618 1006 15560.10 170 240 338 462 658 862 1310 19060.05 2
34、82 370 476 620 866 1092 1584 22380.01 550 666 820 1008 1302 1582 2170 29280.001 962 1126 1310 1552 1908 2248 2938 3812E2164 164where:z= the one-sided critical value of the standard normaldistribution, andc = the number of common responses.Values of zfor some commonly used values of -risk are:-risk z
35、0.50 0.0000.40 0.2530.30 0.5240.20 0.8420.10 1.2820.05 1.6450.01 2.3260.001 3.090If the confidence limit is less than Pmax, then conclude thatthe samples are similar in attribute intensity (that is, no morethan Pmaxof the population would perceive a difference at the-level of significance). If the c
36、onfidence limit is greater thanPmax, then similarity has not been demonstrated.10.2 If desired, calculate a two-sided confidence interval onthe proportion of common responses. The method is describedin Appendix X4.11. Report11.1 Report the test objective, the results, and the conclu-sions. The follo
37、wing additional information is highly recom-mended:11.1.1 The purpose of the test and the nature of thetreatment studied;11.1.2 Full identification of the samples: origin, method ofpreparation, quantity, shape, storage prior to testing, servingsize, and temperature. (Sample information should commun
38、i-cate that all storage, handling, and preparation was done insuch a way as to yield samples that differed only in the variableof interest, if at all);TABLE 2 Number of Assessors Needed for a Directional Difference Test Two-Sided AlternativeNOTE 1The values recorded in this table have been rounded t
39、o the nearest whole number evenly divisible by two to allow for equal presentation ofboth pair combinations (AB and BA).NOTE 2Adapted from Meilgaard et al (8). 0.50 0.40 0.30 0.20 0.10 0.05 0.01 0.0010.50 Pmax=75 % 2 6 8 12 16 24 34 520.40 6 6 10 12 20 26 40 580.30 6 8 12 16 22 30 42 640.20 10 10 14
40、 20 26 34 48 700.10 14 16 18 24 34 42 58 820.05 18 20 26 30 42 50 68 920.01 26 34 40 44 58 66 88 1180.001 42 50 58 66 78 90 118 1500.50 Pmax=70% 6 81216263454 860.4 6101220304060 943 8141822344468 1020.20 14 20 22 28 40 54 80 1140.10 18 24 30 38 54 68 94 1320.05 26 36 40 50 66 80 110 1500.01 44 50 6
41、0 74 92 108 144 1920.001 68 78 90 102 126 148 188 2400.50 Pmax=65 % 8 14 18 30 44 64 98 1560.40 10 18 22 32 50 68 110 1660.30 14 20 30 42 60 82 126 1880.20 22 28 38 54 72 96 146 2080.10 30 42 54 70 94 120 174 2440.05 44 56 68 90 114 146 200 2760.01 74 92 108 132 164 196 262 3460.001 122 140 162 188
42、230 268 342 4400.50 Pmax=60 % 16 28 36 64 98 136 230 3520.40 22 32 50 78 112 158 254 3840.30 32 44 66 90 134 180 284 4260.20 46 66 86 116 168 214 322 4720.10 72 120 158 214 268 392 5540.05 102 126 158 200 264 328 456 6360.01 172 204 242 292 374 446 596 7960.001 276 318 364 426 520 604 782 10100.50 P
43、max=55 % 50 96 156 240 394 544 910 14240.40 82 130 194 294 452 618 1006 15560.30 110 174 254 360 550 722 1130 17020.20 170 240 338 462 658 862 1310 19060.10 282 370 476 620 866 1092 1584 22380.05 390 498 620 786 1056 1302 1834 25440.01 670 802 964 1168 1494 1782 2408 32040.001 1090 1260 1462 1708 20
44、94 2440 3152 4064E2164 16511.1.3 The number of assessors, the number of selections ofeach sample, and the result of the statistical analysis;11.1.4 Assessors: age, gender, experience in sensory testingwith the product, and experience with the samples in the test;TABLE 3 Number of Selected Responses
45、Needed ForSignificance in a Directional Difference Test, One-SidedAlternativeNOTE 1Entries are the minimum number of common responsesrequired for significance at the stated significance level (column) for thecorresponding number of assessors n (row). Reject the assumption of “nodifference” if the nu
46、mber of correct responses is greater than or equal tothe tabled value.NOTE 2For values of n not in the table, compute the missing entry asfollows: Minimum number of responses (x) = nearest whole numbergreater thanx=(n/2)+zn/4 , where z varies with the significance levelas follows: 0.84 for =0.20; 1.
47、28 for = 0.10; 1.64 for = 0.05; 2.33 for = 0.01; 3.10 for = 0.001. This calculation is an approximation. Thevalue obtained may differ from the exact value as presented in the table,but the difference never exceeds one response. Exact values can beobtained from binomial distribution functions widely
48、available in statis-tical computer packages.NOTE 3Adapted from Meilgaard et al (8).Significance level (%)n .50 .20 .10 .05 .01 .0014 3 4 4 . . .5 4 4 5 5 . .6 4 5 6 6 . .7 4 6 6 7 7 .8 5 6 7 7 8 .9 6 7 7 8 9 .10 6 7 8 9 10 1011 6 8 9 9 10 1112 7 8 9 10 11 1213 7 9 10 10 12 1314 8 10 10 11 12 1315 9
49、10 11 12 13 1416 9 11 12 12 14 1517 9 11 12 13 14 1618 10 12 13 13 15 1619 10 12 13 14 15 1720 11 13 14 15 16 1821 12 13 14 15 17 1822 12 14 15 16 17 1923 12 15 16 16 18 2024 13 15 16 17 19 2025 13 16 17 18 19 2126 14 16 17 18 20 2227 14 17 18 19 20 2228 15 17 18 19 21 2329 16 18 19 20 22 2430 16 18 20 20 22 2431 16 19 20 21 23 2532 17 19 21 22 24 2633 17 20 21 22 24 2634 18 20 22 23 25 2735 19 21 22 23 25 2736 19 22 23 24 26 2840 21 24 25 26 28 3144 23 26 27 28 31 3348 25 28 29 31 33 3652 27 30 32 33 35 3856 29 32 34 35 38 4060 31
