Difference from Control Test

The difference from control test is classified as an overall difference test. It is similar to the degree of difference test, in that it is used to determine if there is, in fact, a difference between one or more test samples and a control. And, more importantly, if there is a difference, its size can also be measured with this test.
This test is most useful in situations such as quality assurance and control, where there is a heterogeneous quality to the products being tested. In this case, some difference is to be expected between batches, and it is the size of those differences that determines a decision. A duo-trio, triangle, or other difference test would not be as effective as they do not measure the size of the difference. The difference from control test is also useful when fatigue and carryover are an issue, as it can be used as a two-sample, rather than a multi-sample, test1.
When conducting this test, 20-50 subjects should be used to provide meaningful results. It does not matter if the panel is trained or untrained, but there should not be both. Each subject is given a labeled control sample and one or more test samples. Within the test samples, the control can also be present as a blind-control, and the subjects should be alerted to this. The blind control helps to establish a base line for the rest of the test samples, as most blind controls will get a non-zero score due to individual variability2. They are then to rate the size of the difference between each test sample and the control. This rating can be done on a line or category scale.
To interpret the results, determine the means of the observed size of the difference from control, both for each type of sample, and for the blind-control samples. Each mean can be compared using analysis of variance, or, if only using one test sample, a paired t-test can be used1.
There are some issues that can be involved with this test. One issue is not having enough subjects participating. This is seen in many quality control/assurance situations. Another issue is that statistically failing to reject the null hypothesis of there being no difference between test and control, does not necessarily mean that there is no sensory difference between the samples.


1 Meilgaard, M, Civille, CV, Carr, BT. 2007. Sensory Evaluation Techniques. 4th ed. CRC Press. 448 p.
2 Lawless, HT, Heymann, H. 2010. Sensory Evaluation of Food: Principles and Practices. 2nd ed. Springer. 850 p.