This study examined the theoretical concepts, and pros and cons of item parceling. While many preceding studies have focused on the merits of item parceling, very few study have researched demerits or limitations of it. To empirically test item parceling results, data were divided by two samples: uni-dimensional sample(sample1), and multi-dimensional sample(sample 2). Then, these samples were item parceled by several different item parceling methods such as random assignment, item-toconstruct balance, and approaches to multi-dimensionality. Using these item parcels, we test that how these item parcels affect dependent variables, and compare the variations of path coefficients and model fits. The results show that the factor loadings and model fits of confirmatory factor analysis with uni-dimensionality sample shows good values than these of confirmatory factor analysis with multi-dimensionality sample. It would be logical conclusion, however there are critical points between the results of sample1 and sample2. While there is no big differences among the models in sample 1, the results are very distinguishable depending on the types of item parceling in sample 2. Thus, if the researcher should use item parceling method, the verification of uni-dimensionality of data have to be conducted before item parceling. Also, about the models which item parceling directly influence on dependent variable, there are minor differences among path coefficients in sample 1, however, there are significant differences among path coefficients in sample2. Consequently we found that dimensionality of data significantly influence the results of item parceling, and item parceling methods also critically affect path coefficients.