Gliomas are the most common intracranial tumors. In the US, approximately 15,000 patients die with glioblastoma per year (CBTRUS 2002). Despite modern diagnostics and treatments the median survival time does not exceed 15 months. However, it has long been observed that after surgical removal, tumors recur predominantly within 1 cm of the resection cavity. This is mainly due to the fact that at the time of surgery, cells from the bulk tumor have already invaded normal brain tissue. Decades ago Matsukado showed that more than 50% of untreated brain tumors had already reached the contralateral hemisphere (J Neurosurg 18: 636-644, 1961). Therefore one of the most important hallmarks of malignant gliomas is their invasive behavior. Dandy already recognized the highly invasive characteristics of this tumor type and performed hemispherectomy in patients with preoperative hemiplegia (J Am Med Assoc 90: 823-825, 1928). Despite his and others' heroic efforts, recurrence was detected as early as 3 months after surgery (Bell, LJ: J Neurosurg 6: 285-293, 1949), leading to the discontinuation of this radical approach. Diffuse gliomas remain a particularly challenging clinical management problem. Over the last 20 years no significant increase in survival of patients suffering from this disease has been achieved. Even drugs directed against newly identified targets like MMPs or angiogenesis-related targets fail to increase survival duration (Tonn, Goldbrunner: Acta Neurochir Suppl 88: 163-167, 2003) Furthermore, anti-angiogenic drugs have been shown to increase glioma invasiveness, finally leading to gliomatosis cerebri. (Lamszus et al.: Acta Neurochir Suppl 88: 169-177, 2003). In this review we focus on the main features which may underlie the invasive phenotype of human gliomas, and offer a biological basis for optimism towards therapeutic advances to come.