The association between dietary patterns (DP) and prevalence of hearing loss in men enrolled in the Caerphilly Prospective Study was investigated. During 1979–1983, the study recruited 2512 men aged 45–59 years. At baseline, dietary data were collected using a semi-quantitative FFQ, and a 7-d weighed food intake (WI) in a 30 % subsample. Five years later, pure-tone unaided audiometric threshold was assessed at 0·5, 1, 2 and 4 kHz. Principal component analysis (PCA) identified three DP and multiple logistic and ordinal logistic regression models examined the association with hearing loss (defined as pure-tone average of frequencies 0·5, 1, 2 and 4 kHz >25 dB). Traditional, healthy and high-sugar/low-alcohol DP were found with both FFQ and WI data. With the FFQ data, fully adjusted models demonstrated significant inverse association between the healthy DP and hearing loss both as a dichotomous variable (OR=0·83; 95 % CI 0·77, 0·90; P<0·001) and as an ordinal variable (OR=0·87; 95 % CI 0·81, 0·94; P<0·001). With the WI data, fully adjusted models showed a significant and inverse association between the healthy DP and hearing loss (OR=0·85; 95 % CI 0·73, 0·99; P<0·03), and a significant association between the traditional DP (per fifth increase) and hearing loss both as a dichotomous variable (OR=1·18; 95 % CI 1·02, 1·35; P=0·02) and as an ordinal variable (OR=1·17; 95 % CI 1·03, 1·33; P=0·02). A healthy DP was significantly and inversely associated with hearing loss in older men. The role of diet in age-related hearing loss warrants further investigation.