To test the hypothesis that glycemic thresholds for cognitive dysfunction during hypoglycemia, like those for autonomic and symptomatic responses, shift to lower plasma glucose concentrations after recent antecedent hypoglycemia in patients with type I diabetes mellitus (T1DM), 15 patients were studied on two occasions. Cognitive functions were assessed during morning hyperinsulinemic stepped hypoglycemic clamps (85, 75, 65, 55, and 45 mg/dl steps) after, in random sequence, nocturnal (2330-0300) hypoglycemia (48 ± 2 mg/dl) on one occasion and nocturnal euglycemia (109 ± 1 mg/dl) on the other. Compared with nondiabetic control subjects (n = 12), patients with T1DM had absent glucagon (P = 0.0009) and reduced epinephrine (P = 0.0010), norepinephrine (P = 0.0001), and neurogenic symptom (P = 0.0480) responses to hypoglycemia; the epinephrine (P = 0.0460) and neurogenic symptom (P = 0.0480) responses were reduced further after nocturnal hypoglycemia. After nocturnal hypoglycemia, in contrast to nocturnal euglycemia, there was less deterioration of cognitive function overall (P = 0.0065) during hypoglycemia based on analysis of the sum of standardized scores (z-scores). There was relative preservation of measures of pattern recognition and memory (the delayed non-match to sample task, P = 0.0371) and of attention (the Stroop arrow-word task, P = 0.0395), but not of measures of information processing (the paced serial addition task) or declarative memory (the delayed paragraph recall task), after nocturnal hypoglycemia. Thus, glycemic thresholds for hypoglycemic cognitive dysfunction, like those for autonomic and symptomatic responses to hypoglycemia, shift to lower plasma glucose concentrations after recent antecedent hypoglycemia in patients with TIDM.