本文研究對象為台灣地區1982、1992、及2000年的全職非農業就業人口，透過經濟學與社會學相關理論的引導，建構一包含人力資本、工作結構、家庭限制、及擁擠理論變項的模型，用以解釋兩性間平均薪資差異及變遷。研究發現在上述變項未經控制及控制的情形下，男性平均薪資仍較女性顯著為高，但是這項差距在1982至2000年期間已有逐漸縮小的趨勢，例如經控制相關變項後，男性平均薪資在1982、1992、及2000年分，別較女性多出30%、31%、及27%。 經以兩性平均薪資差異進行分解，發現在四組變項中，以人力資本所能解釋兩性薪資差異的比率為最高，在三個時期中分別為15%、12%、及，14%，顯示男性較女性在人力資本上仍綜合佔有優勢，特別是在年齡與年資部分；工作結構變項所能解釋兩性薪資差異的比例則分別為6%、3%、及8%'這也反映女性在工作結構的綜合分佈上，仍較男性為不利，特別是兩性在階級身份的分佈上仍明顯有別；而家庭限制變項並未能解釋顯著的兩性薪資差異，但是擁擠理論變項則有遞增的解釋力，三個時期解釋此例分別為-1%、8%、9%，顯示女性被區隔至女性化職業仍是造成兩性薪資差異重要的因素。 本研究同時比較歷年兩性薪資差異無法為上述四組變項解釋的殘差比率，結果發現這項殘差比率有遞增的趨勢，分別為55%、67%、及74%，顯示女性仍面臨該四組變項以外的其他新增歧視因素而這些新增的，歧視因素預期不僅會影響薪資水準，且與性別間存在相當關聯，未來宜進一步探索現階段就業市場中對女性可能新增的薪資歧視因素。

Based on the samples of full-time, non-farm labor force from the 1982, 1992, and 2000 Taiwanese Labor Utilization Surveys, an earnings determination model, constructed from the perspectives of human capital, segmented labor markets, a family constraint, and crowding theory, is used to examine the changes of gender differences in earnings in Taiwan. Given that the individual and job characteristics have been controlled, the results indicate that the male labor force's average monthly earnings are 30%, 31 %,and 27% greater than those of the female labor force respectively in 1982, 1992, and 2000. Generally speaking, the gender gap in earnings has been on the decline in the past 20 years. The results of a decomposing gender gap in earnings indicate that a 15%, 12%, and 14% gender gap in earnings is due to gender differences of human capital respectively in 1982, 1992, and 2000. At the same time, the perspectives of segmented labor market theory and crowding theory can also explain the non-trivial portion of the gender gap in earnings. However, there are no significant findings revealling that family constraints have resulted in a gender gap in earnings. Finally, the 55%, 67%, and 74% gender gap in earnings is still unexplainable by the model for 1982, 1992, and 2000. respectively. The increasing portions of the model’s residual indicate that some potential factors causing the gender gap in earnings still have not ,been considered in the model. The implications of the findings are also discussed in the paper.