Do we need to replicate in sensory profiling studies?

Several evaluations are necessary to test individual repeatability in sensory profiling. But do additional replicates really provide more information than the first evaluation? Statistical analysis of 339 studies with at least two evaluations showed that when using the first evaluation alone: Significance of product discrimination was identical for 6 out of 7 attributes, 85% of the significant attributes had a similar vector of mean scores ie a correlation coefficient higher than 0.9, Significance of multivariate product discrimination assessed by MANOVA was identical for 97% of the datasets, 93% of the datasets had a similar product configuration (RV > 0.9), the CVA product maps were similar (according to several criteria including product configurations, attribute positions...). Further, the second evaluation was found less relevant in studies including a small number of products (< 5). Finally, a third evaluation seemed irrelevant for the improvement of the statistical analysis.