To identify protective dietary predictors amongst long-lived elderly people (N= 785), the "Food Habits in Later Life "(FHILL) study was undertaken among five cohorts in Japan, Sweden, Greece and Australia. Between 1988 and 1991, baseline data on food intakes were collected. There were 785 participants aged 70 and over that were followed up to seven years. Based on an alternative Cox Proportional Hazard model adjusted to age at enrollment (in 5-year intervals), gender and smoking, the legume food group showed 7-8% reduction in mortality hazard ratio for every 20g increase in daily intake with or without controlling for ethnicity (RR 0.92; 95% CI 0.85-0.99 and RR 0.93; 95% CI 0.87-0.99, respectively). Other food groups were not found to be consistently significant in predicting survival amongst the FHILL cohorts.