all received intensive perceptual training that focused their attention on subtle phonetic contrasts between the speech sounds of the target language and those of their L1. We suggest that this may have helped them to rely less on the categorical mode and more on the continuous mode of perception, as they did when they acquired their L1, and thus to gradually work out what the relevant sound cues in the L2 are ( Martohardjono & Flynn, 1995; see also Hammond, 1995) and to establish correct perceptual targets (Flege, 1995) for the L2 speech sounds. In addition, the very advanced learners had all received intensive training in the production of L2 speech sounds aimed at developing the finely tuned motor control required for accurate pronunciation. In sum, what we suggest is that the success of the exceptional adult learners we identified may have been at least partly due to the combination of three factors: high motivation, continued access to massive L2 input, and intensive training in the perception and production of L2 speech sounds. Clearly, much more work in this area is called for, and subsequent studies of ultimate attainment should put more effort into identifying the psychological and contextual correlates of exceptionally successful L2 learning.
So far, our studies have focused on the pronunciation of British English and French by adult learners with a Dutch Ll background. It is an empirical question whether the findings we reported in this chapter can be generalized to pairings of L1s and L1s that are typologically more distant than the L1-L2 pairings in our experiments. We intend to explore this issue in future studies with very advanced learners of Dutch who have Turkish, Moroccan Arabic, or Berber L1 backgrounds.
To conclude, although the speech of adult L2 learners is typically accented, it seems that we have identified at least some individuals who have beaten the predictions of the critical period hypothesis for accent by attaining a native-like pronunciation of an L2. A major challenge for the future would be to identify which (combinations of) learner, context, and language variables (L1-L2 pairings) are instrumental in making nativelike attainment possible.
I gratefully acknowledge the contributions by Marie-José Palmen, Brigitte Planken, Chantal van Summeren, and Erik Schils to the research reported on in this chapter. I also thank David Birdsong, Kees de Bot, James Emil Flege, Daan Hermans, Margriet Jagtman, Annemieke Jansen-van Dieten, and Eric Kellerman for their valuable discussions and comments.