本文透過兩個研究，檢驗臺灣版2歲期自閉症篩檢工具（The Screening Tool for Autism in Two-Year-Olds, Taiwan version, T-STAT），評估4歲自閉症類兒童（全距42-54個月）的效度。研究一，自閉症類兒童與發展遲緩兒童各13名，採用訊號偵測理論，確認T-STAT評估4歲自閉症類兒童的切截分數。研究二，自閉症類兒童19名，以及發展遲緩兒童18名，運用研究一的切截分數於效度樣本，確認適用性。結果發現：研究一，依據訊號理論，切截分數1.25分是較好的選擇。研究二，1.25分做為切截分數，敏感度.90，精確度.89。本文兩個研究結果發現，T-STAT可以協助4歲自閉症類兒童的評估、篩檢。然而，這項結果仍需要蒐集比較大的樣本數，再次檢驗正確診斷自閉症類兒童的比例。 The Screening Tool for Autism in Two-Year-Olds, Taiwan version (T-STAT) was originally developed to recognize young children with autism spectrum disorders (ASD) between 24 and 36 months. The T-STAT assesses four domains of early socialcommunicative behaviors, including play, requesting, joint attention and imitation skills. There were two studies to examine the utility of the T-STAT for assessing preschool children with ASD whose chronological age around 4 years old (range = 42 to 54 months). Study 1, the signal detection methods were used to identify a cutoff score for the T-STAT using 26 children with ASD and developmental delay (DD) whose mental age, verbal mental age and nonverbal mental age were matched. Study 2, using the cutoff score from study 1 to validate diagnosis accuracy of children with ASD and DD which not overlapped with study one. The results of current studies showed that cutoff score of 1.25 for T-STAT could get better sensitivity and specificity. Using the cutoff score to an independent validation sample (19 ASD children and 18 DD children), the sensitivity and specificity were .90 and .89 respectively. Both study 1 and study 2 revealed that using TSTAT with lower cutoff score to differentiate children with ASD around 4 years was promising. However, it needs to recruit larger sample size and divide the subtypes of children to replicate the results.