The entire context must be considered during clustering, as there is no target word around which to adjust the test or training scope, for example. (These are options in discriminate.pl) Typical examples of headless contexts include email or other short messages or documents, where the goal is to cluster them based on topic.