Intensive medical research over the last fifty years has left the prognosis for patients diagnosed with malignant brain tumors nearly unchanged. This suggests that a new perspective on the problem may offer important insight. We have undertaken an interdisciplinary research program, seeking to study brain tumors as complex systems. This research aims to develop computational models coupled with experimental assays to investigate the hypothesis of selforganizing behavior in tumor systems. Preliminary assays have revealed behavior consistent with this hypothesis. A cellular-automaton model to study the growth of the tumor core has been developed. This model has proven successful in reproducing macroscopic tumor growth from a limited parameter set. Further, it has been applied to investigate the importance of heterogeneity to determination of a clinical prognosis and has demonstrated the importance of understanding clonal composition in making an accurate prognosis.