Poor visual acuity (VA) is a risk factor for falls, and a common impediment to rehabilitation, but conventional VA testing is difficult in dysphasic, deaf or confused patients. In the Cardiff acuity test (CAT) observation of the subject's eye movements (preferential looking) indicates if they can see a vanishing optotype on a card. The test is quick, and requires no speech or understanding on the part of the subject. We consider its usefulness in frail, elderly patients. Seventy-three patients aged 47-99 (mean 78) years, were tested in good lighting, wearing their usual spectacles. Cards ordered 'A' to 'K' with increasingly faint targets were sequentially presented at 1 m, until the subject's eye movements indicated the target to be invisible. We then performed conventional Snellen acuity chart testing. Twenty-three were retested by a second observer, and 24 subjects were retested by the same observer on another day. We used correlation coefficients to confirm inter-observer (r=0.95, P<0.01), and test-retest (r=0.97, P<0.01) reliability. Snellen chart measurements of VA were possible in all but six subjects. The results of CAT and Snellen acuity tests showed statistically significant correlation (r=0.35, P<0.01). The World Health Organisation (WHO) define significant impairment as a Snellen test acuity below 6/18, and using the 'G' card as a threshold were able to detect this degree of impairment with a sensitivity of 91.7%, and a specificity of 90.9%. The CAT is reliable, and highly acceptable to elderly patients, and may be useful as a screening tool in clinical practice, and for epidemiological purposes.