本論文採用民國六十七年至九十一年人力資源調查五月份的資料，建構人力資源調查合併資料檔案，根據相異指數、以及Tam（2001b）的標準化女性比例，進行職業層次、產業層次、工作層次性別隔離的程度和樣貌的趨勢分析，並以三維立體圖形方法呈現結果。研究結果顯示，職業層次的性別隔離近乎整合且標準化女性比例的百分比座落在60％至70％之職業有逐漸增多的趨勢、產業層次的性別隔離也呈現整合而中性職業增加的趨勢、至於工作層次性別隔離程度則逐漸下降且也呈現出整合的趨勢變化，而工作層次性別隔離程度高過職業層次性別隔離程度。另採用民國六十九年與七十九年台閩地區戶口及住宅普查資料完整母體資料所建構的工作層次資料，以檢驗使用人力運用調查的樣本資料得出之工作層次資料的結果，證明後者的分析具有效度。而採用普查資料進行探究也發現，無論是在更細緻的職業層次（三分位職業代碼）和工作層次（三分位職業代碼與二分位行業代碼的工作代碼組合）、或者是在與人力資源調查相同的職業層次（二分位職業代碼）和工作層次（二分位職業代碼與二分位行業代碼的工作代碼組合）的分析層次下，採用相異指數和標準化女性比例的結果皆是性別隔離程度下降、也呈現出整合的趨勢變化，工作層次性別隔離程度也高過職業層次性別隔離程度，然其僅提供十年間的資訊，實為可惜之處。 / This study uses data from 1978-2002 Manpower Utilization Survey (MUS) for analyzing, ultimately, job-level sex segregation in Taiwanese workplace. Taking into account the limitation of large amount of small-N job cells due to the relatively smaller sample sizes, the method of mid-point of interval estimation and 5-year aggregation is applied with its strength of consecutive data series: pooling data across 1978 to 1982 as estimation for year 1980, and so on, to construct 5-year aggregated MUS data. To view the job-level sex segregation pattern, job-level datasets are established by the “combination of occupation and industry codes”, from which there are units of analysis, job cells. And discarding small-N reliable sampled jobs in MUS is essential to avoid biased results for they contain extreme sex composition after comparing with those from population data of 1980 and 1990 Taiwanese Censuses. Findings indicate that the degree of occupational, industrial, and job-level sex segregation is somewhat increasing by Duncan and Duncan’s Dissimilarity Index, by which it is unable to measure the exact segregation degree because of its margin-dependent trait. However, by using Tam’s margin-free index: Standardized Percent Women, findings indicate that: (1) for occupational-level sex segregation, there is a trend of integration, and those occupations with 60%- 70% SPWOM shows a slight increase; (2) for industrial sex segregation, the trend almost remain the same, just that gender-neutral industries increase a bit, especially in recent years; (3) as for the degree of job-level sex segregation, it is declining, and there is a long-term trend of integration in Taiwanese workplace over the past 25 years. Therefore, job-level analysis does show what the past occupational and industrial analysis cannot present. And the findings from MUS are robust, based on the contrast analysis from population data of 1980 and 1990 Taiwanese Censuses; in addition, research outcomes from Census data also show integrated pattern in occupational and job-level sex segregation in that decade.