This structured encoding enables accurate decoding of volitional speech segments and could be applied in the development of brain–machine interfaces for restoring speech in paralysed individuals.