Clinical experience suggests that many symptoms occur together. In this paper, we examine the rationale and evidence base for symptom clusters in different medical fields, particularly the cluster phenomenon in cancer. Cancer symptom clusters are a reality. Various symptoms that cluster clinically have also been verified statistically. Specific clusters such as nausea-vomiting, anxiety-depression, and cough-dyspnea are evident on both clinical observation and in research investigation. Fatigue-pain and fatigue-insomnia-pain have also been demonstrated statistically as clusters. Another proposed cluster 'depression-fatigue-pain' seems relevant to clinical practice. Other clusters may serve only as theoretical models that illustrate possible common biological etiologies in cancer; they need to be validated in future research. Analysis of the literature is complicated by considerable inconsistencies across studies. Discrepancies between clinically defined and statistically obtained clusters raise important questions. We must consider the analytical techniques used, and how methodology might influence cluster occurrence and composition. Further research is warranted to establish universally accepted statistical methods and assessment tools for symptom cluster research.