A variety of reference curves are used to derive predicted values for adult lung function, even within ethnically similar populations. Alternatives to percentage predicted value are sometimes used to allow for height in research. Strength of association with total mortality can be used to choose the optimal expression, between forced expiratory volume in 1s (FEV(1)) divided by height(2), FEV(1)/height(3), FEV(1)% predicted and difference from predicted. Data from the Reykjavik Study cohort, 1976-2002, included 5544 men and 8062 women randomly selected from the population. Total mortality was analysed by Cox proportional hazards regression in relation to each height-adjusted measure, allowing for age group, period of recruitment and body mass index, and smoking before or at baseline. FEV(1)/height(2) and FEV(1)/height(3) had stronger associations with mortality than FEV(1)% predicted and difference from predicted in men and in women. There were similar findings for forced vital capacity (FVC) in non-smokers and in women. FEV(1)/height(2) was slightly better predictive than FEV(1)/height(3) in men, but distributions of FEV(1)/height(3) in men and women were closer than those of FEV(1)/height(2). Clinical practise and epidemiological research would benefit from agreement on how to adjust lung function for height. Replication of these analyses in other cohort studies would inform the choice between FEV(1)/height(2) and FEV(1)/height(3).