Students in English Studies departments in France are expected to target a restricted set of native accents (either Received Pronunciation – RP – or General American – GA). We recorded 307 students who read phonetically-rich sentences specially designed to elicit accent-specific pronunciation features. We listened to these specific features, ran automatic speech recognition, and performed automatic accent classification. We examined to what extent we, the listeners, agreed with one another, what insight could be gained from automatic speech recognition and accent classification, and whether our perception concurred with automatic methods. Among other things, our results show that only 7% of our students use one of the two native models consistently, with two thirds favouring the American model.