This empirically based study aims at establishing the frequency, distribution, tenacity and the nature of errors produced by more than 700 advanced Swedish learners of French, at different university levels. The errors are classified in grammatical and lexical categories. Four grammatical categories (articles, nouns, pronouns, verbs) have been selected for a thorough examination, and examples of different types of errors are presented and commented upon. The error analysis follows, principally, the steps suggested by Corder (1974) and implies the discussion of such concepts as error, norm and usage.The verbs present the most frequent and tenacious errors constituting at each level about a third of all errors. The proportions of the different types of errors change according to task and level. The errors concerning tenses predominate in the translation tests, those concerning conjugations in the diagnostic test. As for pronouns, they cause 40 % of all errors in the diagnostic test (fill-in/transformation tests), 7 %-8 % in the translation tests. This reduction is partly due to the fact that many pronouns are necessarily regarded as part of the construction of a verb and classed as such. Put together, the errors concerning gender, orthography and vocabulary constitute 40 % at the advanced levels. The study includes a correlational analysis aiming at confirming or confuting some hypotheses regarding the interrelation, on the one hand between the results of the different parts of the two tests given during the first term of the university studies (vocabulary, grammar, metalingual knowledge), on the other hand beween the results and some external factors such as marks in French (secondary school), stay in French-speaking country, length of previous studies of French, etc. Among other things, it is shown that the students’ results in grammar are systematically better than those in vocabulary and that six years of previous studies of French do not generally lead to better results than three years’ studies.