背景和目的：由於憂鬱兒童盛行率有逐年升高的趨勢，且兒童憂鬱症狀復原時間長、影響層面廣泛，故早期發現及介入相當重要。但因學齡兒童透過語言表達自我情緒困擾的能力有限，因此本研究擬針對房樹人投射測驗發展出一個有信度並可區辨出兒童憂鬱情緒之計分方式。方法：利用房樹人測驗(HTP)、兒童青少年憂鬱量表(CDI)和瑞文氏圖形推理測驗進行個案評估。排除低IQ者，並將符合CDI19 分以上且經由專家挑選出有問題的圖畫納入臨床組；CDI19 分以下且未由專家挑選出圖畫納入控制組。由專家和研究者進行兩組圖畫分析，挑選出初步指標。並由研究者訓練評分者進行向度精緻化和評估。結果：(1)本研究歸類出九大項和二十七小項房數人投射測驗計分細項。(2)肯德爾合諧係數分析顯示評分者間信度[rho = 0.898, p < 0.00]。(3)使用獨立樣本T 檢定檢測臨床組與控制組之區辨效度，結果顯示本研究所發展之計分向度具有區辨臨床組和控制組之效度趨勢[t(1,19) = 1.34, p = 0.197]。結論：(1)本研究所發展出之房數人投射測驗計分向度具有良好之評分者間信度；(2)未來研究可增加參與人數，並將不同共病進行區分，以求區辨效度。

Background and Purpose: Epidemiological research shows that there has been an increase in childhood depression. Available evidence also suggests that depression is a recurrent condition. Therefore early detection and intervention are important. Because school-aged children are unable to adequately express their emotional difficulties verbally, this research attempted to develop a reliable scoring system that can distinguish childhood depression from non-clinical controls by employing house-tree-person (HTP) drawings. Methods: Children aged 10~12 years were invited to partake in the research.HTP drawings, the Children'sDepression Inventory (CDI), and Raven's progressive matrices (SPMs) were administered to the subjects. Children with a CDI score of > 19 were classified as the clinical group,while thosewith a CDI score of < 19were classified as the controls. The researchers and an expert studied and analyzed the HTP drawings of the clinical and control groups, and then marked the indicators that distinguished the 2 groups. The indicators were then revised and finalized by the researchers and 2 other raters. Results: The results showed that (1) 9 major indicators and 27 minor indicators were found; (2) the Kendall coefficient of concordance showed that the inter-rater reliability was consistent (rho = 0.898, p < 0.00); and (3) the t-test was employed to verify the discriminant validity between the clinical and control groups. The result showed that a nearly significant discriminant validity was found with the scoring system developed in this research (t(1,19) = 1.34, p = 0.197). Conclusions: The scoring system developed from this study has good inter-rater reliability. It is hypothesized that using more subjects would increase the significance level of the discriminant validity.