Cochlear implants (CIs) restore hearing to nearly one million individuals with severe-to-profound hearing loss. However, with limited spectral and temporal resolution, CI users may rely heavily on top-down processing using cognitive resources for speech recognition in noise, and change the weighting of different acoustic cues for pitch-related listening tasks such as Mandarin tone recognition. While auditory training is known to improve CI users’ performance in these tasks as measured by percent correct scores, the effects of training on cue weighting, listening effort, and untrained tasks need to be better understood, in order to maximize the training benefits. This dissertation addressed these questions by training normal-hearing (NH) listeners listening to CI simulation. Study 1 examined whether Mandarin tone recognition training with enhanced amplitude envelope cues may improve tone recognition scores and increase the weighting of amplitude envelope cues over fundamental frequency (F0) contours. Compared to no training or natural-amplitude-envelope training, enhanced-amplitude-envelope training increased the benefits of amplitude envelope enhancement for tone recognition but did not increase the weighting of amplitude or F0 cues. Listeners attending more to amplitude envelope cues in the pre-test improved more in tone recognition after enhanced-amplitude-envelope training. Study 2 extended Study 1 to compare the generalization effects of tone recognition training alone, vowel recognition training alone, and combined tone and vowel recognition training. The results showed that tone recognition training did not improve vowel recognition or vice versa, although tones and vowels are always produced together in Mandarin. Only combined tone and vowel recognition training improved sentence recognition, showing that both suprasegmental (i.e., tones) and segmental cues (i.e., vowels) were essential for sentence recognition in Mandarin. Study 3 investigated the impact of phoneme recognition training on listening effort of sentence recognition in noise, as measured by a dual-task paradigm, pupillometry, and subjective ratings. It was found that phoneme recognition training improved sentence recognition in noise. The dual-task paradigm and pupillometry indicated that from pre-test to post-test, listening effort reduced in the control group without training, but remained unchanged in the training group. This suggests that training may have motivated listeners to stay focused on the challenging task of sentence recognition in noise. Overall, non-clinical measures such as cue weighting and listening effort can enrich our understanding of the training-induced perceptual and cognitive effects, and allow us to better predict and assess the training outcomes.