In 1984, the companies Siemens, Philips and the IPO Institute of the Technical University of Eindhoven started the SPICOS project to develop a continuous speech dialogue system with a lifetime of 1984-1990. The dialogue system - hereafter called SPICOS system - was developed for the German language and worked with a vocabulary of 1000 words. The SPICOS system consisted of 3 modules - an automatic speech recognition (ASR) module, a dialogue module and a text-to-speech (TTS) module. In 1984, the recognition of continuous speech was still in its infancy. The recognition rate of existing ASR systems for continuous speech was not sufficient to serve as input to a dialogue module. The main focus of the SPICOS project was to develop an ASR module that would produce sentences with a low word error rate, and a dialogue module that would be robust enough to handle sentences with erroneous words. For the TTS module, the TTS system developed at IPO for the Dutch language was adapted to the German language. The paper describes the state of the art in the 1980s relevant to the SPICOS system and the concepts developed for the modules. These concepts are documented by various papers listed in the references. At the end of the SPICOS lifetime, a working continuous speech-driven dialogue system was demonstrated. Finally, we conclude: SPICOS has made a significant contribution to paving the way for KI systems based on the concept of data-driven learning and the availability of large speech resources.