How specific is the learning in an auditory frequency discrimination task?

Background: There is ample evidence showing that frequency discrimination can dramatically improve with practice. Nevertheless, the ability to generalize the learning gains to conditions that were not encountered during training is still controversial. Therefore, the purpose of the present study was to evaluate the extent of generalization to the untrained ear, to an untrained frequency (2000 Hz) and to an untrained (noisy) environment by continuing to train the generalization conditions. Methods: Twenty-one participants took part in two training phases. The first training phase included nine training sessions that consisted each of six threshold measurements of difference limen for frequency (DLF) at 1000 Hz, and a 10th testing session that consisted of three DLF thresholds at 1000 Hz. The generalization phase included four training sessions that consisted each of six DLF threshold estimates, two days at each generalization condition (each participant was trained in two out of three generalization conditions, thus the total number of participants trained in each generalization condition was 14). All DLF thresholds were estimated using a three-interval, two-alternative forced choice adaptive procedure. Results: Results showed that (1) the initial thresholds in the generalization conditions were reduced following training, and (2) new learning was observed in the subsequent training sessions suggesting that the learning was partially specific to the trained ear, frequency and environment. Conclusions: These results have important theoretical and clinical implications regarding the extent of the plasticity of the auditory system and the neural processes that underlie this plasticity.