Fragile watermarking has been proposed as a means of adding additional security or functionality to biometric systems, particularly for authentication and tamper detection. In this paper we describe an experimental study on the effect of watermarking iris images with a particular class of fragile algorithm, reversible algorithms, and the ability to correctly perform iris recognition. We investigate two scenarios, matching watermarked images to unmodiﬁed images, and matching watermarked images to watermarked images. We show that different watermarking schemes give very different results for a given capacity, highlighting the importance ofinvestigation. At high embedding rates most algorithms cause signiﬁcant reduction in recognition performance. However, in many cases, for low embedding rates, recognition accuracy is improved by the watermarking process.