研究主要探討學生評鑑教師教學因素構念在跨層級架構下的合理性，以作為學生評鑑教師教學多層次議題發展的立論依據。研究對象為東部某大學大學部180 班學生，班級人數介於13 和78人之間，總樣本數為6,568 人。經多層次模式競爭及多層次複核效度檢定發現，學生評鑑教師教學量表的因素構念在學生層次分為教學準備、教材內容、教學方法、教學評量、教學態度等五個次向度，但在班級層次卻僅以單一因素構念「學生評鑑教師教學」反映不同班級間的差異。因此，本研究建議在進行跨層級的量表建構及議題探討時，應先檢視跨層級的構念是否產生改變，透過跨層級的模式競爭及效度複核，忠實呈現因素構念在多層次下的樣貌，而非僅將組間層次及組內層次的因素結構假定為相等，以避免後續跨層級推論的謬誤。

This study examined the rationality of multilevel constructs of student ratings of instruction. The sample consisted of 180 undergraduate classes from a university on the east coast of Taiwan, with a class size ranging between 13 and 78, for a total sample of 6,568 students. The results indicate five sub-dimensions of student ratings of instruction, including Teaching Preparation, Teaching Materials, Teaching Methods, Learning Evaluation, and Teaching Attitudes. The same instrument within class level, however, indicates only one single dimension. This study suggests examining the changes in cross-level constructs through multilevel model competition and multilevel cross-validation approaches. Examining the validity and rationality of the cross-level model presents the effect of each dimension under multilevel constructs to avoid the fallacy of cross-level inference.