Verbal probability statements such as “very certain” or “impossible” are often used to communicate information regarding uncertainty. This study aims to investigate how people estimate the “informativeness” of verbal probability statements. With regard to numerical probabilities (e.g., “30%”), Keren and Teigen (2001) proposed the “search for definitive predictions” principle; according to this principle, relatively high or low probabilities are preferred to medium ones because high or low probabilities denote the occurrence or nonoccurrence of a single outcome more strongly than they do medium ones. However, whether people adhere to this principle when they estimate the informativeness of the verbal probability statement remains an unresolved issue. Through two empirical studies, this study establishes a quantitative relation between verbal probability statements and their “informativeness,” which is similar to the principle of Keren and Teigen (2001). In addition, this study shows that the directionality of verbal probability (Teigen & Brun, 1999) also affects the informativeness judgment. Finally, this study formalizes the judgment regarding the informativeness of probability statements in terms of the information theory and argues that the “search for definitive predictions” principle can be interpreted as rational information estimation under the rarity assumption (Oaksford & Chater, 1994).