隨著網路日益發達，使消費者在做購買決策時所需要參考的商品資訊來源更多樣。同時為了降低對於商品的品質不確定性，人們傾向在網路上尋找其他已購買此商品的消費者所撰寫之線上商品評論。根據文獻探討，消費者在瀏覽線上商品評論時，其評論的影響力會因為其評論方向而有所不同而產生偏誤。本研究主張線上商品評論的負面偏誤並非皆存在於各種情況，其負面偏誤關係會因為眾多因素而產生變化。 本研究根據Nelson(1970; 1974)對商品分類的方式，將商品分為搜尋性商品與經驗性商品，探討在不同商品類別之下，評論方向對消費者商品態度影響的負面偏誤關係之差異；並且主張線上商品評論之負面偏誤關係會由於消費者之商品涉入程度與主觀商品知識程度不同而有差異。 透過實驗設計各種情境並發放實體問卷，有效問卷共計296份。模型驗證部分，本研究使用SPSS 22、Amos 22與SmartPLS(v3.2.4)統計軟體分析各路徑之影響關係，結果顯示：(1)整體而言，線上商品評論對消費者商品態度的影響之負面偏誤確實存在。(2)在同樣程度的商品涉入之下，搜尋性商品的負面偏誤比經驗性商品更為顯著。(3)主觀商品知識程度並無顯著的干擾作用存在。(4)在經驗性商品之下，商品涉入程度對負面偏誤有顯著的負向干擾效果，即商品涉入越高，負面偏誤越小；搜尋性商品下，商品涉入程度對負面偏誤的干擾作用並不顯著。 ;As internet and technology advance rapidly, the diversity of product information source that consumers require when making purchase decision is much more than be-fore. For reducing the uncertainty of the product’s quality, people tend to seek online product reviews that written by consumers who had purchased the product. From pre-vious studies, the effect of reviews would vary due to the review valence. This article argues that the negativity bias of online product reviews cannot be universally applied, the relationship would vary due to lots of factors instead. According to Nelson (1970; 1974), products can be classified into search goods and experience goods, this paper examines and compares the negativity bias of online product reviews of different product types. We also suggest that the negativity bias would vary because of different product involvement and subjective knowledge. By designing different senarios of our experiment, data was then collected through entity questionnaire and 296 effective responses were collected. Statistical softwares such as SPSS 22, Amos 22 and SmartPLS(v3.2.4) were used for analyzing purposes. Results from our experiments imply that (1) the negativity bias of online product re-views on consumers’ product attitude does exist significantly overall. (2) Specifically, given the degree of product involvement, the negativity bias of search goods is signifi-cantly higher than that of experience goods. (3) Subjective knowledge has no signifi-cant moderating effect on negativity bias. (4) For experience goods, product involve-ment has a significant negative moderating effect on negativity bias, but there’s no moderating effect on negativity bias for search goods.