In a novel environment, we establish causal relationship by inductive inference from statistical data. Although recent studi es on causal induction focus on causal structure rather than intensity, how we induce the intensity from co-occurrence of an effect and a candidate cause remains important as far as the environment is of novelty. We propose a heuristics for causal induction, the proportion of assumed-to-be rare instances (pARIs), and test its rationale and the descriptive validity. The pARIs rule is based on the rarity assumption that is quite often used in rational analysi s approach (e.g., Oaksford & Chater, 1994) intimately associated with the frame problem. As for the descriptive power, we show that pARIs best fits the experimental results with the highest correlation and the min imum error. In regard to the rationale, we tested how we assume the rarity of the events we focus on. We confirmed that we r igorously distinguish between rare and non-rare events, in spite of their identical status as data.