The development of the otic placode is believed to depend on an inductive signal from the adjacent hindbrain. A candidate for this signal is FGF-3 (Int-2), which is expressed in the hindbrain adjacent to the future ear in rhombomeres 5 and 6 (r5 and r6). However, in vitro tests (Represa et al. (1991), Nature 353, 561-563) conflict with findings from FGF-3 knockout mice (Mansour et al. (1993), Development 117, 13-28). The former suggest that FGF-3 from the hindbrain is required to induce formation of the otocyst, while the latter imply that FGF-3 is required only in the later process of otocyst differentiation. We find that in normal embryos at early stages the gene is expressed not only in r5 and r6, but also in most of the hindbrain anterior to this and in the head ectoderm in the prospective otic placode region. In kreisler mutant embryos, however, there is no heightened expression in r5 and r6, but the early patch of expression in the prospective otic placode ectoderm is still seen and the otic vesicle still forms at nearly the normal place. Subsequent malformations of the inner ear in kreisler and in FGF-3 knockout mice are similar, involving failure of the development of the endolymphatic appendage. These findings argue that FGF-3 is not required as an inductive signal for invagination of the otic placode to form a vesicle, whose future site is already marked out independently of any localized FGF-3 signal from r5 and r6. FGF-3 does, however, appear to be required for a correct pattern of differentiation within the vesicle.