Forty-nine children aged between 8 and 10 years old from primary schools in two socially disparate areas of Edinburgh participated in this study. Repertory-grid method obtained information from the children regarding their perceptions of eight vegetables commonly consumed in the UK (baked beans, carrots, tomatoes, cauliflower, turnip (swede), cabbage, sweetcorn and peas). Between 6 and 23 personal constructs were elicited from each child (mean = 15). Individual score-sheets contained each child's unique set of constructs represented by linear scales. Children scored all eight vegetables for every construct, and for preference. Data were analysed using Generalised Procrustes Analysis. The first five dimensions of the product space (explaining a total of 88% of the variance) separated the vegetables into distinct groups according to physical and sensory properties, requisite preparation, and appropriateness for eating with particular meals. Socio-economic background was found to influence children's perceptions of the vegetables. Internal preference mapping examined the variance in the children's hedonic responses. Three clusters were identified, with baked beans being particularly liked. Friedman's two-way ANOVA confirmed that the children's preferences for the vegetables were not equal (p < 0.01), with carrots ranked the most liked, and cauliflower ranked least liked. Mann-Whitney tests showed that the socio-economic status did significantly influence preferences for three of the eight vegetables (p < 0.05). There were no significant correlations between preferences and either gender, or frequency of vegetable consumption. (C) 1999 Elsevier Science Ltd.. All rights reserved.