In　research　on　dynamic　soil-structure　interactions　under　an　earthquake　load　and　as　a　verification　research　method,　the　shaking　table　test　is　considered　an　effective　tool.　However,　the　proper　model　box/container　for　the　test　soil　should　be　used.　Subsequently,　the　boundary　effect　of　the　model　box　used　in　the　test　becomes　a　critical　factor　that　has　a　direct　impact　on　the　accuracy　of　the　model　test　result.　In　this　study,　a　three-dimensional　laminar　shear　model　box　consisting　of　19　layers　of　spherical　rings　is　designed　and　produced.　The　height　and　inner　diameter　of　the　model　box　are　1,800　mm　and　4,500　mm,　respectively.　Inter-layer　aluminum　alloy　frames　are　connected　via　support　members　that　can　move　freely　in　all　three　directions.　A　shaking　table　test　using　the　model　box　is　performed,　and　the　earthquake　response　is　measured　via　accelerometers　buried　at　various　depths　and　positions　in　the　soil　to　study　the　box　boundary　effects.　During　the　test,　AP1000　ground　motion　used　in　nuclear　power　engineering　designs　is　applied　to　the　shaking　table　from　one　horizontal　direction　(X　or　Y),　two　horizontal　directions　(X　and　Y)　or　three　directions　(X,　Y　and　Z).　The　acceleration　time-history　diagrams　and　response　spectra　at　various　observation　points　with　identical　input　directions　and　different　depths　are　compared.　The　relative　errors　of　the　soil　response　peak　acceleration　and　acceleration　response　spectra　at　observation　points　located　from　the　box　center　to　the　edge　are　calculated.　The　test　and　analysis　results　demonstrate　that　the　proposed　three-dimensional　laminar　shear　model　box　can　accurately　simulate　an　actual　scenario　with　an　infinite　boundary　condition　in　all　three　directions,　which　overcomes　the　limitation　of　the　currently　used　laminar　shear　model　box　that　only　supports　horizontal　earthquake　excitation　input　from　one　or　two　directions.