Oliver Tomic*, Ciaran Forde+, Conor Delahunty+, Tormod Næs*

* Nofima, Norway
+ Food Science Australia, Australia


Sensory panel leaders need to know the point at which their panel is adequately trained for the descriptive analysis task it is about to undertake. Profiling when the panel has not been adequately trained will result in poor quality data, whereas training past a point of adequate proficiency adds unnecessary cost. Criteria that determine proficiency include (i) sensitivity to differences between products for a given attribute, or discrimination, and (ii) agreement of each assessor with the remainder of the panel. Expressing overall panel performance on these two criteria for a given attribute using a simple index, where 0 represents poor performance and 100 represents excellent performance, is highly desirable. Such indices would provide an easy and comprehensible way of feedback to the panel as a group. Moreover, a threshold may be established for each attribute that needs to be surpassed before one can claim the sensory panel to be proficient. These performance indices may be derived and diagnosed from Tucker-1 correlation loadings provided by the free, open source software tool PanelCheck.

To test this potentially new PanelCheck application, performance indices were computed for the sensory profiling data of a designed experiment, a cross-over study in which two groups of sensory assessors were trained with and without feedback on their performance. Panel A was trained on chocolate milk in the standard manner and did not conduct daily rating tasks or receive any feedback on their performance. Panel B was trained on same product in a separate training session and were asked to make daily attribute rating tasks across the products. Based on the results of these rating tasks the panel received feedback was used to coach their discrimination and rating reproducibility. In the second part of the study – the two sensory panels were asked to evaluate natural yoghurts but in this instance – Panel A received the feedback and Panel B received no feedback.

Sensory panel
Chocolate milk
Natural yogurt
A
Usual training
Training + daily feedback
B
Training + daily feedback
Usual training

Across both products and panels, performance feedback improved panel training efficiency. The indices developed to measured discrimination and agreement proved effective at capturing the degree to which the sensory panel were trained and will be useful in future trials as a benchmark of panel proficiency.