This study aimed to compare retracted (due to misconduct) and non-retracted articles in biochemistry, in terms of proportion of positive terms, certainty score and different certainty aspects. The data set of this study composed of 662 retracted and non-retracted articles published in the time period of 2018-2020 and indexed in Scopus. These 662 articles accounted for 331 non-retracted and 331 retracted articles, which were matched using matching and covariate balancing analysis. The analysis in this article was done using several regression models. Regarding the use of positive terms, the findings showed that retracted articles were 16% less probable to use positive terms in abstracts, titles and findings presented in conclusion and discussion compared with non-retracted articles. In addition, the results regarding the analysis of certainty language, showed that retracted articles were 15% less probable to use certain language, measured by certainty score, in presenting their scientific findings. Finally, regarding the certainty aspects, the results of regression models showed that retracted articles had 11% less likelihood to present their research findings using certain probability aspect.