This paper presents a new method for the verification of the correct pronunciation of spoken words. This process is based on speech recognition technology. It can be particularly useful when applied to the field of SLA (Second Language Acquisition) in learning environments or Computer-Aided Language Learning (CALL) systems, where the students can practice their pronunciation skills. This method uses an artificial neural network plus a specific grammar for each utterance to compare the text of the expected utterance with the sequence of phonemes recognized in the speech input, in order to detect the pronunciation errors.