A total of 320 students (163 males, 157 females) were administered Marks' Vividness of Visual Imagery Questionnaire (VVIQ) in either its original (blocked) format or in a revised (random) format in which the order of the items was changed. All subjects completed the test on a second occasion, either 10 min or three weeks after the first, and did so under standard or novel memory instructions in which they were asked to recall their initial responses. Mean scores were significantly higher (less vivid imagery was reported) and split-half reliability significantly lower with the random (0–692) than blocked (0–893) grouping, although females scored lower than males on both forms. Test-retest reliability was similar at 0–710 for the two versions after three weeks under reassess instructions, but was slightly elevated on the blocked relative to the random test after 10 min, where self-reported item recall, although low, was enhanced. These results suggest that, although its internal consistency and temporal stability were rather low by psychometric standards, scores on the random form of the test were less contaminated by response bias than those on the blocked version. It is therefore recommended that the random VVIQ be preferred, but that its use be confined to research purposes.