Results from clinical samples suggest low serum albumin may be associated with cognitive impairment, though evidence from population-based studies is inconclusive. Participants were 1,752 adults (699 men and 1,053 women) aged 65 years and over from the Health Survey for England 2000, a nationally representative population-based study. Cognitive impairment was assessed using the Abbreviated Mental Test Score. The cross-sectional relation of serum albumin quartiles to cognitive impairment was modelled using logistic regression. Two hundred and twelve participants were cognitively impaired (68 men and 144 women). Odds ratios (95% confidence intervals) for cognitive impairment in the first (2.2-3.8 g/dl), second (3.9-4.0 g/dl), and third (4.1-4.3 g/dl) quartiles of serum albumin compared with the fourth (4.4-5.3 g/dl) were 2.5 (1.3-5.1), 1.7 (0.9-3.5), and 1.5 (0.7-2.9), after adjustment for age, sex, education and additional risk factors for cognitive impairment (p for linear trend = 0.002). A highly similar pattern of associations was observed for men and women. Our data provide new evidence to suggest that low serum albumin is independently associated with increased odds of cognitive impairment in the elderly population.