The study was to evaluate the reliability and limitation of the CT number through determining the range of variation of CT number under a variety of different conditions using experimental models. Three cylindrical phantom models containing lesion-simulating-rods were used. After filling the different-density-solutions within the phantoms, the computed tomography was performed under a variety of different conditions in order to demonstrate the variability of CT number of the same lesion. The CT number of the rods within the phantom models was measured and analyzed. RESULTS : 1. The CT number of the lesion proportionately increased as the CT number of surround increased, and the increase of the CT number of the lesion showed linear response to the increase of the CT number for the same lesion regardless ofthe change of surround. 2. Time interval, change of mAs, or slice thickness had no effect on the CT number of the lesion. 3. The CT number of the lesions tended to vary wider n the body model tan the head model. 4. The range of the CT number variation increased as the lesion decreased in size, and it was more pronounced as the difference between the CT number of the lesion and surround increased. The above results suggest that CT number should be carefully used for diagnosis with consideration of its variability under a variety of different conditions.