A rare opinion may be more meaningful than ones supported by the majority of people. Such an opinion breaks into a popular concept if people become aware of the opinion and admire it as highly acceptable, and the prevalent support grows to be an established consensus. This paper is dedicated to aid in finding the seed of this process, i.e., an opinion with the latent popularity. The structure of the co-occurrence (occurrence in the response by the same subject) between opinions is shown to the key to identifying such an opinion. KeyGraph, a text indexing method, is applied to an accumulated questionnaire-result data for visualizing such a structure. The experiment show the mixture, of the human imagination and the output of KeyGraph, clarify the significance of opinions for forthcoming consensus.