Fifty-one male and female volunteers aged 18-90 years from a wide variety of social and occupational backgrounds completed 7284 assessments of portion size in relation to food photographs. Subjects were shown six portion sizes (two small, two medium and two large) for each of six foods, and asked to compare the amount on the plate in front of them to (a) a series of eight photographs showing weights of portions from the 5th to the 95th centile of portion size (British Adult Dietary Survey), or (b) a single photograph of the average (median) portion size. Photographs were prepared either in colour or in black and white, and in two different sizes. The order of presentation of foods; use of black and white or colour; the size of photographs; and presentation of eight or average photographs were each randomized independently. On average, the mean differences between the portion size presented and the estimate of portion size using the photographs varied from -8 to +6 g (-4 to +5%) for the series of eight photographs, and from -34 to -1 g (-23 to +9%) for the single average photograph. Large portion sizes tended to be underestimated more than medium or small portion sizes, especially when using the average photograph (from -79 to -14 g, -37 to -13%). Being female, 65 years and over, or retired, or seeing photographs in colour, were all associated with small but statistically significant overestimations of portion size. Having a body mass index > or = 30 kg/m2 was associated with an 8% underestimate of portion size. We conclude that use of a series of eight photographs is associated with relatively small errors in portion size perception, whereas use of an average photograph is consistently associated with substantial underestimation across a variety of foods.