Objective: To assess the errors between estimates of amounts based on photographs and actual quantities of food presented to adults or eaten on the previous day. Design: A photographic atlas was constructed by taking digital photographs of three pre-weighed portion sizes of 212 traditional Lebanese dishes. In a first approach, ten portions of real pre-weighed foods were assessed using the photographic atlas. In a second approach, the participants weighed all foods consumed at one meal and recalled the amount of food that was eaten the previous day using the photographic atlas. Differences between actual quantities and estimation were assessed using a signed rank test (P , 0?05). Spearman's correlation coefficients and bias (Bland-Altman plot) between the methods were calculated. Setting: Lebanese university canteen and Lebanese homes. Subjects: Forty adults (twenty males, twenty females) completed the first protocol and fifty adults (twenty-five males, twenty-five females) completed the second protocol; all were volunteers aged 21 to 62 years. Results: Mean differences between actual and estimated portion sizes by photographs were between 213?1 % and 124?5 % when pre-weighed foods were presented, and between 210?4 % and 13?8 % when foods were consumed the day before. No significant differences were found between actual and estimated portion sizes except for three dishes (loubieh bil zeit, tabbouleh and yachnet bazella). Bland-Altman plots showed good agreement for all dishes with a negligible bias of 20?2 g between estimated and presented portions, and 26?3 g between estimated and eaten portions. Conclusions: In the Lebanese diet, food photographs seem a reliable tool for quantification of food portion sizes.