Free-Comment (FC), as a response to open-ended questions, enables a word-based sensory description and discrimination of sets of products. The stability of FC outputs has never been investigated and is the purpose of the present paper. Since Check-All-That-Apply (CATA) is the most popular method for the word-based sensory description of products with consumers, the stability of FC was compared to that of CATA performed on the same products. Four red wines and four milk chocolates were evaluated according to different sensory modalities by groups of consumers following either an FC or a CATA protocol. The stability of the product configurations and the product by descriptor associations were investigated. FC outputs were slightly more stable than CATA ones. Sixty consumers enable to guarantee medium stability, if not good, of FC and CATA outputs when the investigated product space is characterized by large differences between the products. The minimum number of con sumers to obtain stable results was strongly dependent on the size of the differences between the products, which suggests that if a priori knowledge on the size of the differences between the investigated products is available, it must drive the decision of the number of consumers to include in the study rather than relying on an absolute rule. For both FC and CATA, the product configurations were more easily stable in terms of numbers of con sumers than the product by descriptor associations. Investigating the stability of the product by descriptor associations a posteriori is recommended for future FC and CATA studies.