An experiment was carried out to investigate errors in translation exercises done by French students studying English as a second language. A code was devised to rate errors as being: (1) lexical or grammatical, and (2) related to the signifier or the signified, with further subdivisions within these groups. While this method has the advantage that it is rigorous, practical, and informative, some classifications are admittedly ambiguous and some errors difficult to classify. The errors were compiled from two series of translation examinations, and from a test especially compiled to facilitate certain errors. Although this research must be considered non-definitive, some conclusions may be drawn from it. Among these are the importance of: (1) the grammar test over the essay as a reliable measure of grammatical competence; (2) gradation in the presentation of grammatical and syntactic structure; and (3) grammar courses and composition courses to fill the gaps demonstrated by analyses such as this. (Author/AM)