OBJECTIVE: To assess the validity of a semi-quantitative food-frequency questionnaire (SFFQ) used in a large nation-wide dietary survey among 12-month-old Norwegian infants. METHODS: The SFFQ was administered to the parents about 1 week before the infants turned 12-month-old. The participants filled in the SFFQ and 1-2 weeks later they kept weighed food records for a total of 7 days. Both methods were completed for 64 infants. RESULTS: The SFFQ overestimated energy intake with 25% and gave significantly higher estimates of all nutrients compared with the records, except for calcium. Much of the difference between the methods disappeared when nutrient density was compared. Spearman correlation coefficients between pairwise measurements of nutrient intakes from the food records and the SFFQ ranged from 0.18 for vitamin D to 0.72 for polyunsaturated fatty acid intake (median r=0.50). On average 38% of the infants were classified in the same quartile with the two methods, and 3% in the opposite quartile. The correlations for food items varied from 0.28 for sweetened drinks to 0.83 for commercial porridge (median r=0.62). CONCLUSION: This study indicated that the SFFQ overestimates average absolute nutrient intakes. However, the questionnaire gave better estimates for average nutrient densities than for absolute nutrient intakes. The capability of the questionnaire to rank infants according to intake of nutrients and food items was moderate, but at the same level as others have observed with food-frequency questionnaires.