Improving the performance of A-Not A sensory discrimination ratings by modifying sample presentation probability

Sensory discrimination analysis based on signal detection theory (SDT) deals with measuring small sensory differences between samples required for various business objectives, including quality control, reformulation, cost reduction, and product development. Variants of the A-Not A test designs have been used for SDT analysis using sensory panels to measure the sensory differences between a reference and multiple test products because they are economical and sensitive for such multiple product comparisons. On the other hand, the test protocols of conducting the A-Not A ratings affect the test performance and data quality. The present study examined the effects of equal sample presentation probability between reference vs test samples (0.5 reference presentation probability, 0.5 RPP) with and without informing a panel of RPP on the test performance of A-Not A and A-Not AR ratings using a reference with three test products. SDT analyses including the novel d'Rec analysis were used to examine these effects. The overall experiment consisted of three protocols: Control - 0.25 RPP without informing; Modified-1 protocol - 0.5 RPP without informing; Modified-2 protocol - 0.5 RPP with informing of the RPP. Compared to the Control, the Modified-1 protocol employing 0.5 RPP without informing improved discrimination performance significantly (higher d ' values) for the A-Not A ratings. Compared with the A-Not A ratings, the effects of the Modified-1 protocol were reduced for the A-Not AR ratings, showing the effects of reminders assisting the evaluations. The test instructions providing information of 0.5 RPP were inefficient. It shifted the panel's middle decision criteria (c0) close to zero but did not improve the discrimination performance better than the Control. Overall, the efficiency of 0.5 RPP was confirmed using sensory panels for both A-Not A and A-Not A ratings, and it was shown that the novel d'Rec analysis is an efficient way to study the reasons for changes in a panel's performance and diagnose the sensory data quality obtained from sensory panels.

Automated summary

Sensory discrimination analysis based on signal detection theory is important for measuring small sensory differences between samples for various business objectives. A-Not A test designs are used for this analysis, but the test protocols can affect the test performance and data quality. This study found that equal sample presentation probability without informing the panel is effective for improving discrimination performance in A-Not A ratings, but its effects are reduced in A-Not AR ratings.