Purpose: Validated methods to assess diet of non-European infants are sparse. We assessed the validity and reliability of a semi-quantitative food-frequency questionnaire (FFQ) for South Asian infants in Canada.Methods: We developed an 80-item FFQ to assess infant nutrient intake in the South Asian Birth Cohort study (START). Caregivers completed the FFQ twice along with two 24-hour diet recalls. We measured infant plasma ferritin to cross-validate reported iron intake. We evaluated validity using Spearman's rho (ρ), and reliability using the intraclass correlation coefficient.Results: Seventy-six caregivers provided 2 FFQs and 2 24-hour diet recalls. Energy-adjusted, de-attenuated correlations between the FFQs and 24-hour diet recalls ranged from -0.29 (monounsaturated fat) through 1.00 (cholesterol). The FFQ overestimated energy intake by 128%. Iron intake by 24-hour diet recalls correlated with plasma ferritin (r = 0.41; P = 0.01; n = 37), but iron intake by FFQ did not. The average reproducibility coefficient of the FFQ ranged from 0.24 (macronutrients) to 0.65 (minerals).Conclusions: Among South Asian infants living in Canada, at least 2 days of diet recall completed with the primary caregiver yields more valid and reproducible estimates of nutrient intakes than a semi-quantitative FFQ, and it highlights that careful selection of FFQ portion sizes is important for assessing dietary intake with an FFQ.