An application of serially balanced designs for the study of known taste samples with the α-ASTREE electronic tongue

The α-ASTREE e-Tongue instrument uses seven sensors to characterize taste signals associated with a liquid sample. The instrument was used to study eight test preparations (comprised of a blank, four preparations corresponding to four known tastes and Sodium Topiramate in three concentrations known to have a bitter taste) and eight washes. Serially balanced residual effects designs were used to order the samples to estimate residual and main effects. The design provided for eight repeated measurements per test preparation. The experimental results suggested the following: (1) The seven sensors can be separated into three groups according to the ability to discriminate test preparations, and three of the sensors contributed little or no information. Further investigation suggested the lack of differentiability might be due to the age of the sensors. (2) The sensors discriminated known tastes from blank. The residual effect due to test preparations might appear after repeated usage. (3) Exploratory principal component analysis of the data indicated that nearly 90% of the total variability across the seven sensors could be explained by a single principal component. (4) The four standard taste preparations did not correspond to orthogonal dimensions in the principal component axes. (5) The three Sodium Topiramate test preparations could neither be associated with the corresponding known bitter taste sample nor could the three doses be shown to follow a quantitative dose-response relationship on the e-Tongue measurement scale. The practical interpretation of the results of the statistical analysis indicates only poor discriminative ability of the e-Tongue to distinguish clearly between increasing concentrations of a known bitter compound such as Sodium Topiramate. No apparent linear relationship could be discerned over increasing concentrations that would allow the quantification of bitterness.