ABSTRACT
The synthetic creation of speech by computer has been a goal which has given considerable impetus to the testing and refinement of phonetic models. Developments in the specification of acoustic segments, and of their temporal integration, have led to increases in the realism and intelligibility of synthetic speech; but no less important have been advances in the description of intonation. Perfecting the segmental aspects of synthetic speech is of little value if its intonation is such as to misdirect listeners away from the informationally important parts of the utterance, or to antagonise them through undesirable attitudinal cues, such as those to hostility or impatience.
