This study examines the semantic relationships in consumers' health-related questions, physician-provided answers, and between questions and answers with the purpose of supporting the design of health consumer question-answering systems. The information present in the text was expressed using a "pilot" ontology that was based on the semantic relationships from the Unified Medical Language System (UMLS) Semantic Network. The extracted semantic relationships were represented as a set of instances in a frame-based system. An analysis of these instances led to an understanding of how answers are related to questions. A total of 509 semantic relationship instances were identified in twelve question-answer messages (97 in the questions, 334 in the answers, 78 implied between question and answer concepts). The frequencies of relationship instances appearing in the texts were evaluated. In only half of the question-answer messages analyzed was there a direct match linking the semantic relationships in the question to those in the answer. All answers contained expansions of one or more question concepts. These concept expansions used many relationship types such as treats and results_in.