Abstract
This study was conducted to test whether three days of high variability training consisting of either word-only or sentence-only materials would most effectively improve native listener accuracy for non-native speakers. Listeners were separated into a word training (N=14), sentence training (N=13) or control group (N=14). Depending on the training, targets were presented by Spanish-accented speakers as single word tokens or combined to form semantically anomalous sentences. Results showed that listeners who heard sentence-only information were best able to utilize their training. The employment of several strategies including the utilization of prosodic elements available in the spoken sentences and an exposure to a wider range of phonetic environments likely led to the improvement. The results suggest that for native listeners and non-native speakers, a high variability paradigm can be employed successfully when paired with sentence level information as all benefits found with word-only training were met or exceeded by sentence-only training.