The paper presents a review of the techniques used to date to provide secure duplex or half-duplex conversational speech communications over HF radio channels, and gives an assessment of likely future developments. The paper begins with a brief discussion of the mechanisms of speech production and hearing, with particular reference to the factors which are important for communications over the difficult HF medium. Various methods for speech scrambling based on time and frequency domain permutations are then described, and the effects of HF channel distortions on their speech quality performance is assessed. Digital vocoding techniques based on the long established channel vocoder and the more recent linear predictive coder are described, and their performance advantages over scrambling techniques are discussed. The importance of the type of modem employed in conjunction with vocoders is emphasised, particularly with respect to the techniques used to overcome the effects of fading and cochannel interference. In the future, continuing developments in semiconductor technology will have a significant impact on the amount of processing which can be performed within moderately priced equipment. The paper is concluded with some prognostications as to how the techniques used to provide narrowband secure speech will change to yield improved performance, both in terms of the speech quality and the reliability of obtaining communications over HF channels.