The purpose of this study was to examine the clear concept of the designs for cavity preparations. Among the several parameters in cavity designs, profound understanding of isthmus width factor would facilitate selection of the appropriate cavity preparation for a specific clinical situation. In this study, the cavities were prepared on maxillary first premolar and filled with gold inaly. A two - dimensional model was composed of 1037 - node triangle elements. In this model, isthmus was varied in width at 1/4, 1/3 and 1/2 of intercuspal width and material properties were given for four element groups, i.e., enamel, dentin, pulp and gold. The 500N occlusal load varied in direction and it was examined using three types of load : concentrated load, divided load and distributed load. The models were also examined with empty cavities using the devided load and distributed load. These models were analyzed the displacement and strees distribution by the two - dimensional Finite Element Method. The results were as follows : 1. All experimental models which filled with gold inlay after cavity preparation were similar direction of displacement with control model under same load type. But in the models with empty cavities, as isthmus width was wider, the degree of displacement was increased at same load type. 2. Among the experimental models which were filled with gold inaly after cavity preparation, the model II showed the least stress concentration under concentrated load and divided load. But in the models with empty cavities, the model III showed the largest stress concentration and tooth fracture is expected regardless isthmus width. 3. All experimental models showed similar displacement pattern beneath restorative material under a concentrated load. In the models with empty cavities, a divided load resulted in a lingual displacement of the lingual cusp, but a distributed load resulted in a buccal displacement of the lingual cusp. In regard to the above results, the restored models were stronger than empty models in respect to the bending moment and tensile stress. The empty models are expected to fracture regardless isthmus width. The safest isthmus width was 1/3 of intercuspal distance, which showed the least stress concentration in respect to the effect of stress distribution.