本研究的主要目的在探討正反向試題敘述對人格測驗產生的影響。研究對象為102位國中生及68位大學生。研究工具為賴氏人格測驗原版及作者依題意轉換之反向題本。結果發現，正反向題目敘述在賴氏人格測驗上的答題反應一致性頗高。若針對轉換的方式而言，加“不”所形成的反向題目其受試者對正反向題目作答反應一致性較高。當以相近反義詞的方式轉換為反向題目時，受試者對正反向題目作答一致性偏低。此外，對測量不同人格特質的量尺而言，受試者對正反向題本之答題一致性也不同。易受情境影響的分量尺如變異性，其Kappa平均值較低。本研究結果還發現國中生對正反向題本之作答反應一致性顯著比大學生高，女生對正反向題本之作答反應一致性又顯著比男生高。針對測驗總分而言，本研究結果顯示，正反向題本個分量尺的總分平均數大多達顯著差異。測量正向人何特質分量尺的總分平均數顯著高於測量負向人格特質的分量尺。換言之，受試者對測量正向特質的題目傾向回答“是”，而對測量負向人格特質的題目傾向回答“否”。針對各分量尺之α系數的比較部分則是正反向題本之α係數無顯著差異。無論正反向題本，各分量尺間之α係數也相差不大。最後針對賴氏人格測驗正反向題本之效標關聯效度的探討耜分則發現反向題本對人格他評的預測性比正向題本好。而人格自評與他評的相關則是測量正向人格特質的量尺比測量負向人格特質的量尺高。因此本研究建議(1)對情意量尺而言，撰寫題目的過程及題目敘述的推敲可能比認知測驗更為重要。(2)題目敘述對人格測驗而言可能比非人格測驗複雜，因為受試者作答時會符合社會期許，因此對於測量正向或負向人格特質的題目，其作答反應偏差的情況也會有差異。(3)目前李克量表以反向計皆來處理反向敘述題目的作法應不會造成太大的測驗誤差。

　　　　　The purpose of this study is to investigate the impact of positive and negative item phrasing on affective measure. The subjects were 102 junior high school students and 68 university students. The total 152 subjects were available finally. The Lai's Personality Inventory was used for this study. The kappa consistency was applied for the statistic analysis. The findings were: (1)In general, no matter positive or negative item phrasing, the consistency of most item responses between original (positive) and reversed (negative) versions of the inventory were significantly high. (2)The consistency of item response between positive and negative test versions for reversing method by using "not" to create a negative item phrasing was significantly higher that that for the other reversing method by using negative worded (parallel) item phrasing. (3)The consistency of item response between positive and negative versions of the test was significantly varied among the 13 sub-scales (characteristics). (4)Junior high school students yielded significantly higher kappa consistency than university students. (5)Female students yielded significantly higher kappa consistency than male students. (6)The sub-scales' total scores and the test total scores between positive and negative versions were mostly significantly different. The total scores of the sub-scales measured positive characteristics were significantly higher than those measured negative characteristics. That is, the subjects tended to response "no" to the items which measured negative characteristics, and tended to response "yes" to the items which measured positive characteristics. (7)Positive and negative versions of the test did not result in differentα coefficients for all sub-scales. (8)According to investigating criterion-related validation of the inventory, the negative version of the test had more potential to predict the criterion (personality rating by people) than the positive version. Therefore, the study suggests: (1)Item phrasing in affective measures may be more important and critical than cognitive measures. (2)Personality inventories may be more influenced by the social desirability than attitude or interest scales. (3)The way of reversing scoring usually applied to evaluate the negative worded items in Likert Scales should not result in any serious test bias.