In order to assure effective traceability, food-producing animals must be identified by a tamper-proof and durable technique. With the advance in human biometric technologies, the deployment of retinal recognition technology for cattle identification and verification has been prompted. The objective of this study was to assess the accuracy of a commercially available retina biometric technology for sheep identification (i) by determining whether light conditions during retinal image capture (indoors and outdoors with shade) and different operators exerted any significant effect on the matching score of the built-in pattern matching algorithm; and (ii) by evaluating the recognition performance of the biometric system for enrolment of one retinal image per sheep and two retinal images per sheep (bimodal biometric system). Neither the light conditions nor the operators were found to have a statistically significant effect on the matching score values of the built-in algorithm; yet it was clear that the pupillary light reflex phenomenon played a major role in obtaining lower matching score values for retinal images taken outdoors. The recognition errors of the one-retina biometric system were estimated to be 0.25% for false matches and 0.82% for false non-matches. An improved bimodal biometric system, i.e., two retinas, that applies a decision criterion based on a simple OR logical operator and a sum of matching scores, has been proposed in this study in order to reduce both probabilities of false matches and false non-matches to near zero. © 2007 Elsevier B.V. All rights reserved.