OBJECTIVES: A prototype hearing aid with trainable amplification parameters (compression threshold, gain below the compression threshold, compression ratio, and noise suppression strength) was evaluated to answer the following research questions: (1) In everyday listening situations, do aid users prefer amplification parameters that were a result of training the aid in such situations to a set of untrained parameters that were prescribed and customized in a clinic? (2) Does the provision of trained noise suppression significantly increase the preference for the trained settings? (3) Is the preference for the trained settings correlated with a measure of training? DESIGN: Three take-home trials were conducted. I. The training trial: Eighteen subjects with symmetric, sensorineural hearing loss were recruited. The aid was fitted with untrained settings based on the NAL-NL1 prescription, and the subjects then attempted to train the aid's amplification settings to their preference in everyday situations for a period of between 1 and 4 weeks. II. The first double-blind comparison trial: Thirteen of the training-trial subjects agreed to participate. These subjects blindly compared their trained settings to untrained settings and voted for their preferred settings in everyday situations for a period of 1 week. Noise suppression was enabled for the trained settings (with trained noise suppression strength), but not for the untrained settings. III. The second double-blind comparison trial: A repeat of the first comparison trial, but with noise suppression disabled for both the trained and untrained settings. Subjects who had high trained noise suppression strength values for the first comparison trial were recruited (eight subjects agreed to participate). This trial more clearly evaluated the effect of training the compression characteristics, and differences between the results of the comparison trials showed the effect of the trained noise suppression. RESULTS: For the first comparison trial, nine subjects voted for the trained settings significantly more often than the untrained settings in real life environments. Three subjects had nonsignificant results, and one subject had a significant preference for the untrained settings. The results of the first and second comparison trials were not significantly different for seven of the eight subjects who participated in both trials. Thus, for these subjects (and the group data) the provision of trained noise suppression did not have a significant effect on the preference for the trained settings. Data logged during the trials showed that the percentage of votes for the trained settings for the first comparison trial was most strongly correlated with the number of hours of aid use during the training trial, although this was not significant when all subjects were included. CONCLUSIONS: In everyday listening situations, aid users can train this prototype aid to provide amplification parameters that, in such situations, they prefer significantly more often than untrained parameters prescribed and adjusted in a clinic. The preference for the trained settings was not significantly affected by the training of the noise suppression strength, and was moderately but not significantly correlated with the hours of aid use during the training period. Therefore, the customization of compression parameters that is currently performed in the clinic can at least partly be performed in real life listening situations by clients who have been fitted with this trainable aid.