The essay begins by looking back at the way translation was taught at the Sorbonne in the 1930s. Focussing on literary translation, there was no translation theory, no manuals, collections of exercises, and especially nothing to draw students’ attention to the particular difficulties of translation. The best feedback one received was the infamous “marginal corrections”, made up of a number of fairly opaque codes indicating the error, and constituting a vague list of possible mistakes. The essay then focuses on the notion of “error”, and whether classifications of errors are helpful to translators. It is suggested that we should distinguish between correcting errors and preventing them, but in either case that the classifications should be clearly related to a particular textbook or grammar. Some suggestions for correctors’ codes are made, and the essay concludes with some remarks on the necessity also to distinguish between translation errors and purely grammatical errors in the target text.