Estimation bias away from zero in mixed sequences: Mental representation, memory recall, and integration of negative numbers in mean estimates of number sequences.

Processing numerical information, including negative numbers, is crucial for many everyday decisions. Recent debates question whether deviations from expected value maximization arise solely from subjective preferences or from regularities in number processing. Here, we examine how the mental representation, memory recall, and integration of sequentially experienced numbers affect mean estimates, with a focus on negative numbers. In three experiments, participants estimated the mean of number sequences containing negative, positive, or mixed types of numbers. Experiment 1 found an estimation bias toward zero for positive-only and negative-only sequences. This bias is consistent with a symmetrically compressed mental number line for positive and negative numbers and offers an alternative explanation for the reflection effect that is independent of subjective preferences. However, in mixed sequences with nonzero means, the pattern reversed, and participants showed an estimation bias away from zero. Experiments 2 and 3 demonstrate that this bias is not due to differences in number representation or memory recall between positive and negative numbers in pure and mixed settings. Instead, higher response times and estimation noise for mixed compared to pure sequences hint at qualitative differences in how numbers are integrated. In sum, our findings suggest that the processing of numeric information differs between contexts varying in number ranges. These findings emphasize the need to distinguish deviations from expected value maximization that reflect limitations in information processing and those that are shaped by subjective preferences toward risks and losses. (PsycInfo Database Record (c) 2026 APA, all rights reserved).