The research deals with the issue of errors in translation, and distinguishes linguistic errors from translation errors, and the latter are many and must be known and defined due to their importance to students and translators in general. The research also aims, through the comparative analytical descriptive approach, to familiarize scholars and translators with the diversity of criteria for analyzing errors in translation and the theories and approaches that are based on them. We have seen that evaluation models are divided into two types: quantitative and non-quantitative, and that quantitative models share in their dependence on counting errors and on the difference between the smaller units (vocabularies and sentences) between the source and target texts, which makes their results limited, inaccurate, and erroneous if applied at different levels. Higher, text, discourse, and non-linguistic aspects, such as culture, society, psychology, ideology... Non-quantitative models came to avoid the shortcomings of quantitative models, which cannot be generalized and lack a theoretical background. The research concluded the importance of looking at these models and integrating error teaching in translator training programs, the importance of continuous training in order to reach a good, correct and self-contained translated text without reference to the source text, and the need to use accurate concepts and terms and standardize them, as translators and researchers do not always use terminology. The translation is always the same, which leads to some difficulties in choosing the most common, accurate and correct terms, which makes us call for a unified Arabic lexicon for translation science terms, similar to the standard Arabic dictionaries in other fields.