The main purposes of this study were (a) to compare themathematics attitude dimension underlying the construction of theChinese version of Fennema-Sherman Mathematics Attitudes ScalesInventory (FSMAS) with the American sample found by Broadbooks,Elmore, Pedersen, and Bleyer; (b) to examine the relationshipsbetween mathematics anxiety and selected student demographic,socioeconomic, academic, and mathematics attitudes scales variablesby multivariate statistical techniques. Data used in this study were collected from junior high schoolstudents in Tainan, Taiwan, the Republic of China (R.O.C.). Thesubjects were selected by a multi-stage stratified sampling method.The first stage unit is the school. The population was stratified intotwo strata: Tainan City and Tainan County. Two junior high schoolswere selected from each stratum and then each school was stratifiedby achievement level. The second stage unit is one class within eachachievement level. Three classes were selected from each school. Thetotal sample size was 526. The 96 items on the FSMAS (excluding the mathematics anxietyitems) were factor analyzed and eight factors were retained whichaccounted for about 70.24 of common variance. The eight dimensionswere similar, but not completely consistent with the FSMASdimensions as articulated by Broadbooks &t al. After the attitude dimension factor scores were constructed, factorregression analysis (FRA) method was employed and it was foundthat confidence and motivation in learning mathematics dimensionmade a significant, unique contribution in explaining mathematicsanxiety. The absolute contribution of this PC predictor was about54.17 of the total variance of mathematics anxiety. Correlationanalysis was used to investigate the relationships between the Yenstudent profile variables and mathematics anxiety. Most of thefindings of the present study were similar to those found by earlierresearchers. Hierarchy of variables analyses (HVA) were employedand 61.18 of the total variance of mathematics anxiety wasexplained by the 24 Yen student profile variables. From a uniquecontribution point of view, the confidence and motivation in learningmathematics dimension was the most potent predictor which couldaccount for approximately 46.09 of the total variance of mathematicsanxiety.