The rules on which a text-to-speech system is based are often hard to be checked, since the programming language used is far away from phonetic and linguistic conventions. Moreover, the program is usually written in a user dependent style and, as a consequence, it is not very readable. Some of these problems could be solved by a speech synthesis oriented programming language, as close as possible to linguistic terminology. The high-level programming language described in this paper has exactly these features, since it is expressly designed for the development of text-to-speech conversion rules and, in addition, the final output of the system is not an object but a source program. One of the aims of the project is to provide a so flexible software tool as to be able to deal with rules for linguistic processing at various levels (e.g. phonetic transcription, word stress assignment, phonological phenomena, prosodic control, etc.) up to the acoustic level. Nevertheless, though the language is text-to-speech oriented, it could be used in all kinds of text processing involving contextual rules, since input and output data structures are completely user-definable.