感官分析.方法学.“A”-“非A”试验

[中文适用范围]： 本文件规定了确定两种产品样品之间是否存在可察觉感官差异的程序。该方法适用于单个感官属性或多个感官属性是否存在差异的情况。“A”-“非A”测试可用于感官分析，具体如下：a) 作为差异测试，特别是用于评估具有差异的样品，例如外观差异（难以获得完全相同的重复样品）或余味差异（难以直接比较）；b) 作为识别测试，特别是用于确定评估员或评估员组是否识别出与已知刺激相关的新刺激（例如，识别新甜味剂的甜味品质）；c) 作为感知测试，用于确定评估员区分刺激的能力。“A”-“非A”测试不适用于评估两种产品是否足够相似以可互换使用（即用于相似性测试），因为“A”-“非A”测试本身涉及所有评估员对同一产品的重复评估。这些重复评估违反了相似性检验在统计上有效的基本假设。其应用示例见附件 B。注：Bi 和 Ennis[1] 指出，无论检验中执行的重复评估的性质如何，“A”样本和“非 A”样本之间的判别距离 d' 的估计值都是相同的，但 d' 的方差估计值取决于重复评估的执行方式。因此，本文档不对“A”-“非 A”方法的 Thurstonian 分析或检验功效进行一般性讨论。感兴趣的读者可以参阅参考文献 [1] 以了解有关该主题的详细讨论。 [外文原描述]： ISO 8588:2017 specifies a procedure for determining whether a perceptible sensory difference exists between samples of two products. The method applies whether a difference exists in a single sensory attribute or in several. The "A" ? "not A" test can be used in sensory analysis in the following ways: a) as a difference test, particularly for evaluating samples having variations, for example, in appearance (making it difficult to obtain strictly identical repeat samples) or in aftertaste (making direct comparison difficult); b) as a recognition test, particularly for determining whether an assessor or group of assessors identifies a new stimulus in relation to a known stimulus (for example, recognition of the quality of the sweet taste of a new sweetener); c) as a perception test, to determine the ability of an assessor to discriminate stimuli. The "A" ? "not A" test is not appropriate for assessing if two products are sufficiently similar to be used interchangeably (i.e. for similarity testing) because the "A" ? "not A" test inherently involves replicate evaluations of the same products by all assessors. These replicate evaluations violate the basic assumptions for similarity tests to be statistically valid. Examples of its application are given in Annex B. NOTE Bi and Ennis[1] point out that the estimate of the discriminal distance, d', between the "A" and "not A" samples is the same regardless of the nature of the replicated evaluations performed in the test but that the estimate of the variance of d' does depend on how the replicate evaluations were performed. As such, no general discussion of a Thurstonian analysis of the "A" ? "not A" method, nor of the power of the test is undertaken in this document. Interested readers are referred to Reference [1] for a detailed discussion of the topic.