AS 2542.2.5:2014 pdf – Sensory analysis Method 2.5: Methodology—’A-not A’ test.
After evaluation of reference sample ‘A’, the assessors usually no longer have access to ‘A’ as a reference while cvaluating the series of test samples (sec Clause 7.2.2).
NOTI: There are variations oldie ‘A—not A’ test which allow the following:
la Initial esaluation of samplea of both ‘A and ‘not A’. tht A’.ailabilit ofrcfcrence sample ‘A’ throughout the test.
7.2.2 Evaluation of test series
Alter familiarization with reference sample ‘A’, the assessors shall evaluate a series of samples in accordance with the following requirements:
(a) All the ‘not A’ samples in a series shall be of the same food product. i.e. different test products shall not be mixed.
NOTE: There is a sariation of the ‘A—not A’ test which permits the presentation of ‘not A’ samples of the same product type hut which arc themselves different.
(b A series may contain ditTerent numbers of samples of ‘A’ and ‘not A’, but in a test each assessor shall receive the identical series (i.e. containing the same respective number of ‘A’ and ‘not A’ samples).
(c) The order of presentation of the ‘A’ and ‘not A’ samples shall be random, and die order shall be different for each assessor.
(d According to the nature of the sample, it may he necessary, in order to avoid effects of sensory adaptation. to present successive samples in the series with a specified time interval between each sample.
(e) The assessors shall evaluate each sample in the order presented and, before testing the next sample, record on the answer form provided whether the sample is ‘A’ or ‘not A’.
NOTE: Specimen answer forms are illustrated in Appendix A.
INTERPRETATION OF RF.SITLTS
At the end of the test, in its simplest form, the responses from the assessors are tabulated in a classical 2 x 2 table containing A’ and not A’ results as follows
The results are interpreted by comparing the two distributions of the classifications of A’ (n1 — ?I) and ‘not A’ (n1 — itt) to determine whcthcr the ratios of classifications by the assessors for A’ and noi A’ are significantly different, in which case the samples are deemed to be different. If there is no difference between the samples the numbers of incorrect responses for A’ and ‘not A’ will he the same.
This intcrprctation may he carried out by means of a test.
If the supervisor has no prior expectation of the results, a bilateral test of significance is applied. If the supervisor has a prior expectation of results, a unilateral test of significance is applied.
NOTES:
I Examples of the application and interpretation of the ‘A—not A’ lest are gisen in Appcndi Ii.
2 Some alues of %‘ are gisen in Appendix C.
3 Other methods of interpretation may by used, such is:
a) Where the number of responses is small, the exact Fisher test c-an be used (see
Ref. Ill.
(hI Where the ‘A—not A’ test is used to determine a detection threshold or a recognition
threshold, signal dctectiøn thoi can be used (see Rei 121. 111 and 1411.
(c) Another commonly-applied correction is the Yates correction (see Note I to
Paragraph HI, Appendix II). but other corrections may be more appropriate
(icc Ref. 151)-
9 REFERFM’ES
Ill FISHER. RA. Statistic-ui Method.c fr,r Research Workers. Edinburgh: Oliver and Boyd. 1941.
[2] GREEN. DM and SWETS, JA. Signal Detection Theort’ and P.cvchophrsics. New York: Wiley, 1966.
[3] SWETS. JA. Signal DeteCtiOn and Recognition hi’ Human Observers, New York:
Wiley. 1964.
[4] BAIRD. JC and NOMA, E, Fundamentals of Scaling and Psychophysics. New York:
Wiley. 197g.
[5] O’MAHONY, M Sensory Evaluation of Food. Statistical Methods and Procedures. New York: Marcel Dckkcr, l9S6.