We propose a phonetically-guided diagnosis of auditory deficiency, which hinges on a carefully constructed corpus of synthetic sounds. Our aim is to complement the diagnosis of sensorineural hearing deficiencies, in order to improve the correction afforded by auditory prosthesis. We first design a vowel corpus, based upon pairs of two-formant, steady-state synthetic French vowels, where the vowels of each pair are chosen to differ only in the frequency of one of their two formants. To test our method, we simulate a frequency-selective loss of audibility, by specifying a piece-wise linear audibility curve with a minimum of -40 dB at a given centre-frequency (1.3, 1.6, and 1.9 kHz). Results of perceptual experiments with normal hearing people tend to show that our synthetic data set is amenable to the diagnosis of the frequency region where the simulated hearing problem is most acute.