This study aimed at finding the most consistent errors of 42 medical students before and after they attended a writing course. The students were asked to write about assigned topics both at the beginning and e end of their writing course before Surface Strategy Taxonomy was used to analyse their assignment. To see if there is a significant difference between errors made by students before and after their writing instruction, paired sample t-test and Wilcoxon test were run. The results indicated that the most consistent error, both in test and retest, was omission. Furthermore, the number of errors in the retest reduced significantly compared with the number of errors in test which could be, at the first glance, interpreted as satisfactory training of writing skills. However, when the significant difference between each type of errors (omission, misordering, addition and misformation) was examined, the results showed that except for the omission error, the reduction of all the other types of errors was not significant. This result could help writing instructors in this university to know the areas of language that their students are struggling with and devote more time and energy to overcome their weakness. Furthermore, teaching and learning writing would be more purposeful.