This paper discusses the problem of sensible qualities, an important, but underestimated topic in Leibniz's epistemology. In the first section, the confused character of sensible ideas is considered. Produced by the sensation alone, ideas of sensible qualities cannot be part of distinct descriptions of bodies. This is why Leibniz proposes to resolve sensible qualities by means of primary or mechanical qualities, a thesis which is analysed in the second section. Here, I discuss his conception of nominal definitions as distinct empirical (...) representations. The provisional and modifiable status of nominal definitions is then explained in the third section. Since nominal descriptions always contain sensible determinations, Leibniz claims that empirical knowledge is indefinitely changeable according to progress in sciences. In the final section, I address the criterion of coherence that enables us to approve of hypotheses that are based on both sensible and empirical properties. (shrink)