Mental accounting based on the S-shape value function in prospect theory has been widely accepted since it was posited. The coding rules in mental accounting (CRMA) further suggest that segregating gains in multiple-gain situations and aggregating losses (to one big loss) in multiple-loss situations are preferred. CRMA have been then successfully applied to many fields. However, in our daily life, we can find occasions when people tend to aggregate gains in multiple-gain situations (accumulate small money from colleagues for a bigger wedding gift) or segregate losses in multiple-loss situations (phase a monthly donation amount to a smaller daily amount). The only study to date showing experiment results conflicting to CRMA, though, focuses on comparing the utility losses of two losses that happen on the same day and on different days. In this study, in order to resolve the inconsistency, we replicate Thaler's experiments with different dollar amounts used in both gain and loss scenarios and propose that mental threshold could be the major reason why in some situations CRMA may not be applicable. Our results show that in multiple-gain (or multiple-loss) situations, CRMA reverse when the accumulated gain (or loss) is over people's mental threshold while the individual gains (or losses) are not. Another finding is that in prior-gain or prior-loss situations, the original reference points can serve as a natural mental threshold. When the accumulated gain (or loss) of multiple gains (or losses) is over the original reference point (anchored by prior gains or losses) while the individual gains (or losses) are not, our experiment results show that people's preference on segregating gains and aggregating losses is reduced.