Many studies have shown that observers can accurately perceive and evaluate the statistical summary of presented objects’ attribute values, such as the average, without attending to each object. However, it remains controversial how the visual system integrates the attribute values (e.g., information on size) of multiple items and computes the average value. In this study, we tested how distributions of item sizes affect the precision and bias in judging average values. We predicted that if observers utilize all of the available size information equally, the distribution would have no effect, and vice versa. Our results showed that, with novice observers, judgement precision differed among size distributions and that the observers overestimated the size of the average value compared to the actual size under all conditions. However, this was not the case for experienced observers, who showed no effects of distribution type on average assessment performance.