Self-consistency model implies that the confidence judgements are based on the consistency of the retrieved representations (Koriat, 2012). The aim of this study was to test the generalizability of the assumptions that follow from self-consistency model using base-rate tasks. A total of 77 psychology students of the Faculty of Humanities and Social Sciences in Rijeka participated in this research. 24 base-rate tasks were repeatedly presented in five blocks. After solving each task participants made a confidence judgment. The response times were also measured. The number of repetitions of the answer from the first block in the following blocks was correlated with confidence judgements given in the first block. The average confidence of frequent answers increased significantly with item consistency, while for rare answers it significantly decreased with item consistency. For frequent answers, participants were significantly faster when they choose the frequent choice across all five trials, compared to when they choose frequent choice three or four times. Participants were the fastest and the most confident when answering to the consistent items, slower and less confident when giving frequent answers to non-consistent items, and the slowest and the least confident when giving rare answers to non-consistent items. At the between-subject level, participants were more confident and faster for majority than for minority answers. A significant item consensualityaccuracy interaction on confidence, as well as on response time, was observed. On consensually wrong items, participants were more confident and faster for wrong than for correct answers. The obtained results imply that assumptions of the self-consistency model can be generalized to the base-rate tasks.