Alternatively, some language learning tasks are suitable for software that is supported by databases, hence being able to deliver materials recorded from multiple native speakers. I have a paper on the effect of using an adaptive training system to teach second language learners to recognize Chinese tones. Perhaps students got (simulated) one-on-one training, thus the time is used more effectively, on average their recognition scores improved by 8% after 2.5 hours of training.