本研究的主要目的是探討利用兒童期自閉症評量表（Childhood Autism RatingScale, CARS）診斷3歲以下自閉症類幼兒的效度。受試者為48名生理年齡在24～36個月大的幼兒，包括：自閉症類幼兒（平均生理年齡30.42個月，平均心理年齡19.59個月）及非自閉症類幼兒（平均生理年齡28.96個月，平均心理年齡20.43個月）各24名。經由訊號偵測理論決定CARS總分的最佳切截分數，發現28.5/29是相對較好的選擇。以28.5/29做為切截分數，敏感度與精確度分別是 .92與 .88。藉由比較兩組受試者在CARS全部15題的差異，選出9題關鍵題目，再利用訊號偵測理論找出這9題關鍵題目總分的最佳切截分數，結果發現，18是相對較好的選擇，敏感度與精確度依舊相當理想。以18做為切截分數，敏感度與精確度分別是1與 .92。本研究結果支持CARS運用在3歲以下自閉症類幼兒的診斷效度，利用9題關鍵題目的敏感度與精確度，比起全部15題，相對更為理想。未來研究需要蒐集更多的受試者，再次探究與驗證CARS在3歲以下自閉症類幼兒的篩檢效度。 The purpose of the study was to examine the validity of using the Childhood Autism Rating Scale (CARS) to diagnose young children with autism spectrum disorders (ASD) less than three years old. The participants were comprised of 48 young children at age from 24 to 36 months, including 24 young children in each of an ASD group (mean chronological age = 30.42 months, mean mental age = 19.59 months) and a non-ASD group (mean chronological age = 28.96 months, mean mental age = 20.43 months). Using signal detecting procedure, the 28.5/29 was used as cut-off to distinguish young children with ASD from young children with non-ASD. The sensitivity and specificity were .92 and .88 respectively. We selected nine items from all 15 items of CARS as critical items based on comparing between young children with ASD and with non-ASD. Using signal detecting procedure again, the 18 was used as cut-off of the nine critical items to distinguish young children with ASD from young children with non-ASD. Again, the results showed that good sensitivity and specificity. The sensitivity and specificity were 1 and .92 respectively. The current results supported that utility of using CARS to diagnose accuracy of young children with ASD. However, the nine critical items showed better sensitivity and specificity. Future studies are needed to recruit larger sample size and validate the accuracy of using CARS for diagnosing young children with ASD under three years old.