Term generation: the relative efficiency of comparative and monadic protocols.

Giboreau A. (a), Dacremont C. (b) Guerrand S. (c), Dubois D. (d)

(a) ADRIANT 54 rue Lamartine 75009 Paris ( a.giboreauadriant.com)

(b) ENSBANA-CESG, Université de Bourgogne, 1 esplanade Erasme, 21000 Dijon

(c) SNCF French railways 45 rue de Londres 75379 Paris Cedex 08

(d) CNRS LAM-LCPE 11 rue Lourmel 75015 Paris

This work is part of a more global research project bringing linguistic and psycho-linguistic knowledge into sensory descriptive science.

It deals with the early stage of the descriptive process: term generation. This step consists of making panelists verbalize their perceptions of the studied products. The success of the term generation is generally estimated through the number of obtained terms, with the underlying hypothesis that the more diverse the generated words, the more representative the final list. This diversity has to be considered from two standpoints: the number of generated terms as well as their semantic variety, as the selection process generally include a gathering of terms into homogeneous meaningful families. Thus, not only the descriptors have to be selected through diverse expression of a given meaning, but also several meaning have to described, with the ideal objective of constituting an exhaustive list.

Sensory science often considers a comparative presentation of samples as being more productive than a monadic one and this research aims at studying the characteristics and efficiency of the two ways of generating descriptive terms. Two protocols - using two different 12-member panels- were compared for the tactile description of 18 toilet papers. One protocol corresponds to a monadic presentation and description of the tactile properties of each sample, the second protocol corresponds to a free sorting task followed by a description of the constituted groups.

A total of 1800 expressions was obtained and classified. Some expressions were used several times either by the same subject or by different subjects, other were subject-specific. About half of the descriptions included a comparison (more / less) or an intensity marker (not, a bit, rather, very…). Both lexical and semantic aspects of the descriptions (i.e. form and content) were investigated.

Results showed large discrepancies between monadic and comparative protocols, regarding quantitative issues - number of terms and qualitative ones - semantic categories (descriptive, reference to a product, usage quality, value judgment).

Key words

Sensory profiling, term generation, descriptors, psycholinguistics.