Professional Documents
Culture Documents
Tautho Date Performed: February 09, 2011
“Tongue Tasters” / Cleaners Date Submitted: February 16, 2011
Ex. 10 “A” ‐ “not A” Test
I. Introduction
Discrimination tests are some of the most common methods employed in sensory science. They
are used to determine if a difference (or similarity) exists between two or more samples. Statistical
significance testing is used to analyze the data and determine whether or not samples are deemed to be
different or similar. Discrimination tests are rapid techniques and can be performed by both naive and
experienced assessors; however, a panel should not be a combination of both. These are often used
when the samples are considered to be ‘confusable’, i.e. their differences are not obvious but need to
be investigated (Kamp et al., 2009).
“A” – “not A” test is used to determine whether test samples in a series are the same as or
different from the reference sample. It is an especially useful test where triangle and duo‐trio tests
cannot be used. This may be the case where comparisons are required between products that have a
strong or lingering flavor/aftertaste when you will need to control the time between sample
presentations or if there are differences in appearance. It is also useful to determine assessor sensitivity
to a stimulus. Initially, panelists require familiarization with the reference or “A” sample. Panelists are
then presented with a series of samples, some of which are the reference sample “A” and some “not‐A”.
Generally, the panelist does not have access to the reference “A” while evaluating the test samples. The
panelist must determine whether the sample is the same (“A”) or different (“not‐A”) so it is a forced‐
choice test. Only one type of “not‐A” sample exists per test series. Panelists may test one, two or up to
10 samples in series (depending on fatigue factors). The samples are presented randomly with 3‐digit
codes and one at a time (an assessment is made and recorded before proceeding to the next sample).
All samples are prepared in an identical way and are representative of the product (Mason and
Nottingham, 2002).
The total number of responses for ‘A’ and ‘not A’ are tallied for each sample presentation. The
chi‐squared test (χ2) is used to compare the different sample presentations and their responses. When
calculating by hand, the χ2 statistic is compared to a statistical table (see Appendix 6) that shows the
minimum value required before it can be concluded that a significant difference exists between the
samples. The significance level (typically 5%) must also be specified. Alternatively, software packages
provide not only the χ2 statistic and the critical minimum value that must be exceeded, but also the
probability of making a type I error (p‐value) should it be concluded that a significant difference exists
between the samples. This analysis is not wholly appropriate for the design involving multiple sample
presentations to each assessor; however, it is commonly used and the p‐value is considered to be a
good approximation (Kamp et al., 2009).
Working equation:
χ2 = N (AD‐ BC)²
EFGH
Where:
A = number of correct responses for ‘A’ sample F = total number of ‘not A’ samples given to all
B = number of incorrect response for ‘not A’ sample panelists
C = number of incorrect response for ‘A’ sample G = total number of A and B
D = number of correct response for ‘not A’ sample H = total number of C and D
E = total number of ‘A’ samples given to all panelists N = total number of G and H
II. Materials
‐ Trays ‐ Score sheets ‐ Spitting cups
‐ Serving plates (small) ‐ Master sheet ‐ Toothpicks
‐ Distilled water ‐ Cheese samples (Eden and Cheezy) ‐ Small containers
III. Procedure
Prepare two samples Prepare the samples to be used
labeled “A” and “not A”. using the standard procedure for
sample preparation.
Two coded samples
Allow the judges to familiarize
composed of “A” and “not A”
themselves with sample “A” and “not
will be served to each judge.
A”.
The order of serving of the two
Decode the results.
samples is determined randomly.
Analyze results statistically using chi‐square.
IV. Results
Table1. Quantitative data obtained for chi‐square calculation analysis.
"not A"
89% sample
Figure1. Percentage distribution on the responses of the panelists using the “A” – “not A” test.
V. Discussion of Results
Discrimination testing is used to find the difference characteristic(s) or overall characteristics of
the sample(s). It is used when sensory specialist wants to determine whether two samples are
perceptively different. Also, this test is used for product reformulation or process (procedure) change;
and for product improvement or product development
The null hypothesis states that the long‐run probability (Pdt) of the population making a correct
selection when there is no perceptible difference between the samples is one in two (H0: Pdt = 1/2). The
alternate hypothesis is that if there is a perceptible difference between the samples the population
would match the reference and the sample correctly more frequently than one in two times. (Heymann
and Lawless, 2010). Now to decide whether to accept or reject the null hypothesis, there is test
2 2 2
criterion. If χ cal > χ tab, reject Ho and accept Ha; thus, the samples are significantly different. But if χ cal <
χ2tab, accept Ho and reject Ha; thus, the samples are not significantly different at 0.05 level of
significance.
In this exercise, the null hypothesis (Ho) is, there is no significant difference between Eden
cheese and Cheezee cheese. While the alternative hypothesis (Ha) states that there is significant
2
difference between Eden cheese and Cheezee cheese at 0.05 level of significance. The value of the χ cal
is 23.68 which was obtained using the values tabulated in table 1 and also the computation is shown in
2 2
the sample calculation section. Based on the results, the χ cal (23.68) > χ tab (3.841). Thus, Ho is rejected
and Ha is accepted. Therefore, there is significant difference between Eden cheese and Cheezee cheese
at 0.05 level of significance.
Figure 1 shows the percentage distribution of the responses of the panelists prior to “A” – “not
A” test which is able to ask if which of the given coded samples is the same with the given reference
cheese samples (“A” and “not A”). The percentage of correct responses for both “A” and “not A” sample
is 89% while percentage of the incorrect responses for both “A” and “not A” sample is 11%. This just
means that 89% of the panelists judged that each of the coded sample cheeses is the same with the
reference or the “A” and “not A” cheese samples.
VI. Conclusion
2 2
Based on the results, the χ cal (23.68) > χ tab (3.841). Thus, Ho is rejected and Ha is accepted.
Therefore, there is significant difference between Eden cheese and Cheezee cheese at 0.05 level of
significance. Rejecting Ho means there is insufficient evidence to accept it. But this does not necessarily
mean that the Ha is true, there is still a need for more panelists, trials, replicates (batch production) to
asses that there is really significant difference between the two samples of cheese which from different
brands at 0.05 level of significance (which is Eden and cheezee). Because, one may commit either of the
two possible error in this statistical analysis used. The type 1 error is rejecting the Ho when it is true and
the type 2 error is accepting Ho when it is false. Thus, these recommendations may be used to verify the
results obtained based on this particular exercise and more modifications can be used as for the
improvement of the methods or procedures of the test.
VII. Reference
Heymann, H., & Lawless, H. T. (2010). Sensory Evaluation of Food: Principles and Practices (2nd ed.).
(D. R. Heldman, Ed.) New York, USA: Springer New York Dordrecht Heidelberg London.
Kamp, S. E. et al. (2009). Sensory Evaluation: A Practical Handbook. Wiley‐Blackwell: A John Wiley &
Sons, Ltd., Publication. John Wiley & Sons Ltd, The Atrium, Southern Gates, Chichester, West Sussex,
PO19 8SQ, United Kingdom.
Mason, R. L. and Nottingham, S. M. (2002). Sensory Evaluation of Food: Statistical Methods and
Procedures. Naresuan University, Phitsanulok, Thailand. Centre for Food Technology, DPI, Brisbane.
VIII. Sample Calculation
IX. Appendix
Master Sheet
“A” – “not A” Test
NA = “Not A” ()
A = “A” ()