Challenges in determining sensory panel proficiency in descriptive profiling
C.C. Gilbert*1, R.L. Heinio2, N. Barylko-Pikielna3, F. Sinesio4, A. Hegyi5, M. Rodbotten6, M.J. Sanchez7, S. Henneberg8 and E.M. Allchurch1 on behalf of the European Sensory Network (ESN) www.www.esn-network.com
1) Campden & Chorleywood Food Research Association (UK), c.gilbert@campden.co.uk
2) VTT Technical Research Centre of Finland (FI),
3) SGGW Warsaw Agriculture University (PL),
4) Istituto Nazionale di Ricerca per gli Alimenti e la Nutrizione (IT),
5) Campden & Chorleywood Magyarország Kht (HU),
6) Matforsk Norwegian Food Research Institute (NO),
7) ainia centro tecnologico (ES),
8) SAM ASAP Sensory and Marketing Germany (DE)
Good panel performance is a prerequisite for reliable sensory data. Proficiency testing evaluates panel performance though the use of inter-laboratory trials, by comparing a sensory panel's performance with other similar panels using the same test materials.
Conclusions from the EU-funded project 'ProfiSens' (SMT-4-CL98-2227) highlighted some unresolved weaknesses of the proficiency test for descriptive profiling. These included too few attributes in common between the panels; lack of standard definitions for common attributes, resulting in panel differences in the use of basic taste attributes; absence of known characteristics or sensory distances when using products 'off the shelf'; and lack of sensory complexity when using designed products. Consequently, research has been undertaken by the European Sensory Network (ESN) to critically review and build upon the existing criteria for proficiency testing of sensory descriptive profile panels.
A proficiency trial was administered using eight double-blind-coded samples of chocolate drink powder, which were adjusted in a controlled way using different levels of glucose, fructose, vanilla and cocoa powder. To address issues of sensory complexity the test materials were developed to achieve appropriate sensory distances between samples in five flavour/taste attributes - chocolate, vanilla, sweetness, bitterness and astringency, in relation to the Just Noticeable Difference. Descriptive profiling of the samples was undertaken by sensory panels in eight European countries (I, E, D, N, FI, H, PL, UK).
The eight panels were all highly experienced multi-product panels; however, none of the panels were especially familiar with the product being evaluated. Panels were allowed to choose their own scale and attributes; however, the five 'common attributes', described above, were also strictly defined and included in the profiling exercise. Generalised Procrustes Analysis showed that the sensory complexity of the products was enough to elicit differences in panel performance; most panels could discriminate samples along the first two dimensions, while only a few were able to discriminate samples along the third dimension.
Results showed that there are advantages to using 'engineered' products; for example, the known product characteristics can be incorporated into the panel performance criteria. Finally, despite providing training samples and definitions for the common attributes, it appears that there are still cultural differences in the use of these sensory terms. However, analysis using only the common attributes often yielded results quite similar to when all attributes were included for analysis.