Background. There is still lack of evidence on the optimal timing of surgery in patients with aneurysmal subarachnoid haemorrhage. Only one randomised clinical trial has been done, which showed no difference between early and late surgery. Other studies were observational in nature and most had methodological drawbacks that preclude clinically meaningful conclusions. We performed a retrospective observational study on the timing of aneurysm surgery in The Netherlands over a two-year period. Method. In eight hospitals we identified 1500 patients with an aneurysmal subarachnoid haemorrhage. They were subjected to predefined inclusion criteria. We included all patients who were admitted and were conscious at any one time between admission and the end of the third day after the haemorrhage. We categorised the clinical condition on admission according the World Federation of Neurological Surgeons (WFNS) grading scale. Early aneurysm surgery was defined as operation performed within three days after onset of subarachnoid haemorrhage; intermediate surgery as performed on days four to seven, and late surgery as performed after day seven. Outcome was classified as the proportion of patients with poor outcome (death or dependent) two to four months after onset of subarachnoid haemorrhage. We calculated crude odds ratios with late surgery as reference. We distinguished between management results (reconstructed intention to treat analysis) and surgical results (on treatment analysis). The results were adjusted for the major prognosticators for outcome after subarachnoid haemorrhage. Findings. We included 411 patients. There were 276 patients in the early surgery group, 36 in the intermediate surgery group and 99 in the late surgery group. On admission 78% were in good neurological condition (WFNS I-III). Management results. Overall, 93 patients (34%) operated on early had a poor outcome, 13 (36%) of those with intermediate surgery and 37 (37%) in the late surgery group had a poor outcome. For patients in good clinical condition on admission and planned for early surgery the adjusted odds ratio (OR) was 1.3 (95% CI 0.5 to 3.0). The adjusted OR for patients admitted in poor neurological condition (WFNS IV-V) and planned for early surgery was 0.1 (95% CI 0.0 to 0.6). Surgical results. For patients in good clinical condition on admission who underwent early operation the adjusted OR was 1.1 (95% CI 0.4 to 3.2); it was 0.2 (95% CI 0.0 to 0.9) for patients admitted in poor clinical condition. Conclusions. In this observational study we found no significant difference in outcome between early and late operation for patients in good clinical condition on admission. For patients in poor clinical condition on admission outcome was significantly better after early surgery. The optimal timing of surgery is not yet settled. Ideally, evidence on this issue should come from a randomised clinical trial. However, such a trial or even a prospective study are unlikely to be ever performed because of the rapid development of endovascular coiling.