Abstract(#br)A new paradigm of the duo‐trio difference‐preference (DT‐Preference) test was developed as alternatives to the paired‐preference test with a no preference option, to correct the overestimation of preference responses. Triadic sample presentation in a preference test elicits odd sample bias, i.e., preferring the sample selected as the same as the reference. In the two replicated DT‐Preference test applied for fermented milk product testing using 208 consumers, this psychological bias in preference responses was apparent in the first test. This bias referred to as the tried first reference bias was eliminated by treating those responses that changed depending on which reference was presented as though no preference had been expressed, leading to more reliable sample... preferences, meaningful to consumers. The perceptual discriminability in DT‐Preference tests was not decreased in comparison to the sole difference tests, while when the preferred sample was used as a reference, the discrimination increased.(#br)Practical Applications(#br)Consumer acceptance mainly determines the sensory quality of products. Paired‐preference test with a no preference option is a broadly used sensory acceptance test method. Yet this preference test method suffers from preference bias caused by the order of sample presentation and/or hidden demand characteristic of the preference test. Such a psychological bias can generate preference responses toward either product presented in a pair for slight or no differences in sample preference, which is not meaningful for consumer behavior. The new paradigm of the duo‐trio difference‐preference test with a perceptual reference introduced in this article uses another type of psychological bias – odd sample bias – as a means to regulate such biased preferences expressed in a predictable manner. Using this approach, the validity and reliability of preference data analysis can be improved by finding more accurate response proportions of two‐way preference responses versus no preference responses.