It is quite often that a common passage is used to measure more than one latent trait. For instance, we might ask subjects to judge how important is a property (e.g., plentiful experiences, abundant knowledge, open-minded) to creativity development and how much do they posses this property. In order to save space, the common passages are listed in the first column, the ratings on "importance" in the second, and the rating on degrees of "possession" in the third. With this editing, the subjects read the common passage first, judge its importance, and then judge the degrees they have. After responding to the first common passage, they move on to the second common passage, judge the importance and then the degrees, etc. It is suspicious that a response procedure like this would result in interference between the two latent traits because people might overestimate (or underestimate) their degrees of possess if they consider the property (un)important. In this paper, we propose a procedure to detect local dependence between traits when common passages are used. To do so, a virtual item (may be called item bundle) should be formed by crossing the response categories in the first trait and those to the second trait. Some parameters are added to model the local dependence. Results of simulation studies show the parameters could be recovered very well. A real data set was analyzed to show implications and applications of the proposed detection procedures.