This article reviews the use and design of speech warnings in terms of ergonomic considerations. Firstly, it considers the benefits of using the auditory channel and technological approaches to producing artificial speech, Secondly, the characteristics of human and machine-generated speech are reviewed: the latter focusing on naturalness, intelligibility, rate of presentation, emotional content and quality. Thirdly, non-speech and speech warnings, and their potential uses are considered and the design of speech for warning applications. Given technological developments, greater use of the auditory channel for warnings is likely: taking into account human factors considerations should lead to better designed warnings.