Introduction. Hyperglycemia increases morbimortality in cerebral infarcts. In animal models, this relationship is only seen during the initial moments of cerebral ischaemia. The time needed in humans for cerebral damage to occur due to hyperglycemia is not known. Material and methods. We included 194 patients admitted within 24 hours of the start of their first clinical episode of cerebral hemisphere infarctation. The glucemia was determined on admission and after 24, 48 and 72 hours. The neurological defect was evaluated on the Canadian scale on the seventh day. The volume of the infarct was determined on a second CT scan done between the fourth and seventh days after the episode. Results. There was a positive association between the volume of the infarct and the glycemias on admission and after 24 hours, but this was not seen in later determinations. The difference in scoring, on the Canadian scale, on the seventh day, between patients with glycemias above or below 120 mg/dl decreased from the time of admission up to the time samples were taken 72 hours later. However, statistical significance persisted during the whole period studied. Nevertheless, in a logistic regression model, the glycemia on admission was the only determination associated with the Canadian scale on the seventh day (OR= 1.02; IC 95% = 1.01-1.02). Conclusions. There is a clear association between hyperglycemia and the worst stage of the infarct, and this is most intense in the first hours after onset of the clinical features