Chunk decomposition is an important mechanism of restructuring in insight problem solving for overcoming impasse. However, one basic issue remained unresolved that why chunk tightness causes the difficulty of chunk decomposition. For this purpose, we examined the role of element type and crossed relation on chunk decomposition difficulty in three experiments based on the Chinese character subtracting task in which participants were asked to remove one target element from a source character in order to get a new character. In Experiment 1 and 2, the two variables including element type (character level vs. non-character level) and crossed relation (uncrossed relation vs. crossed relation) were manipulated in a within-subject design. Thirty two volunteers (16 female; average age: 20.38 ± 2.08 years) participated in Experiment 1, whereas thirty volunteers (15 female; average age: 20.33 ± 1.73 years) participated in Experiment 2. All the participants were native Chinese, right-handed, with normal or corrected-to-normal vision. They got proper rewards after the experiments. The main difference between Experiment 1 and Experiment 2 was stimuli presentation style: the source character and the removed element were simultaneously presented in Experiment 1 whereas the presentation of the removed element was followed by the source character in Experiment 2. There were approximately consistent results between the two experiments, which showed that relative to uncrossed relation, crossed relation lead to lower solution rates and longer response times in chunk decomposition, regardless of the type of the to-be-removed element; chunk decomposition in stroke level required longer response times compared to that in character level and this effect was limited to the uncrossed relation condition. For the condition of crossed relation, there were no differences or even different patterns. The main aim of Experiment 3 was to examine whether the effect of crossed relation on chunk decomposition difficulty still holds in contrast to another type of non-crossed relation----left-right spatial relation. For the purpose, in Experiment 3, the two variables: element type (character level vs. stroke level) and spatial relation (left-right relation vs. top-down relation vs. crossed relation) were manipulated in a within-subject design in the Chinese character subtracting task with subsequent presentation style. Thirty volunteers (15 female; average age: 20.63 ± 1.83 years) participated in Experiment 3. All the others were the same with Experiment 1 and 2. The effects of crossed relation and element type on chunk decomposition difficulty were replicated. Additionally, the solution rate was lower in the condition of crossed relation than that in uncrossed relation (whatever in the condition of left-right relation or top-down relation). There is no difference between the two latter conditions. In conclusion, we demonstrated that chunk tightness is a basic variable that can causes the difficulty of chunk decomposition, which is mainly determined by spatially crossed relation but weakly influenced by element type. Moreover, based on crossed relation, the effect of chunk tightness on the difficulty of chunk decomposition is robust across different paradigms as well as different styles of stimuli presentation (subsequent or simultaneous).