本文首先提出了Linda問題的表象-命題雙表征這一新的解釋視角。該視角認為，Linda問題基于表象表征和命題表征可以有兩種不同的解讀與表征方式;而不同的被試在Linda問題上可能分別采取了上述表征方式之一;但由于Linda問題的特殊性，大多數被試采用了表象表征;大多數被試的這一表征取向則可能是所謂謬誤判斷出現的原因。本文通過4項研究，讓被試在基于表象表征設計的轉述版本與基于命題表征設計的轉述版間選擇接近自身理解的版本;并考察了將Linda問題修改成更符合命題表征的數學化表達形式能否降低所謂謬誤水平;還考察了增加促使被試運用命題表征的排序項"Linda是全人類中的一員"能否降低所謂謬誤水平。結果顯示，在轉述版本選擇上，大多數被試選擇了基于表象表征設計的版本;而上文所指的兩個修正版Linda問題則都降低了被試的所謂謬誤水平。這些結果支持了本文所提的視角。

A dual representation that is the mental image representation and the proposition representation of the Linda problem was proposed in this paper. We hold that people have two different but reasonable representations for the Linda problem. When people used the mental image representation, they think their task in the Linda problem was to judge to what extent of the typical images of the accountant, the feminist or the "accountantfeminist" match the image of Linda respectively according to the conversational rule. However, when people used the propositional representation, the Linda problem is just a math problem described by Tversky & Kahneman(1983). Although both of these two presentations were reasonable, the description of Linda in the Linda problem make people more often use the mental image representation. This is also the reason why most people more likely judge the Lady as accountant-feminist. To test the above assumption, four studies which contained 607 participants were conducted. In study 1, two translated versions, that are the translated versions of the mental image representation, and the translated version of frequency format representation of the Linda problem were used to investigate which version is more closed to participants’ representations. In study 2, another two translated versions, that is the translated version of mental image representation and the translated version of the ratio format representation was used to investigate which one is more closed to participants’ representations. In study 3, two new versions of the Linda problem and the original Linda problem was used to investigate the difference between of them. In one new version, the mathematics set was used to replace the professions in the original Linda problem, but the other expressions were still in nature language. In another new version, the professions were also replaced by mathematics set, and the other expressions were in math language simultaneously. In study 4, another newversion of Linda problem and the original version of the Linda problem were sent to participants randomly. The only difference between the new version of Linda problem and the original one is that the new version had a new ranking item which is "Linda is a person in human being". The results of the study 1 and 2 showed that most participants who committed the fallacy chose the mental image representation of the Linda problem as the one which matched their own representations; but the participants who didn’t commit the fallacy did not have this response bias;(2) The results of study 3 showed that the proportion of the fallacy on both of the new versions of the Linda problem were lower than the original version. However, there was no significant difference between the proportions of the fallacy on both of the new versions.(3) The results of study 4 showed that the proportion level of the fallacy on the new version of the Linda problem was lower than the original one. All the results supported the dual representations perspective of the Linda problem. This research suggested that more studies were needed to explore the mechanism of the Linda problem.