We present a new and novel synthesis of all existing neutrino data regarding the disappearance and appearance of e and . We assume four neutrinos: e , , , as well as a heavier singlet neutrino s of a few eV. The latter may decay into a massless Goldstone boson ͑the singlet Majoron͒ and a linear combination of the doublet antineutrinos. We comment on how this scenario may be verified or falsified in future experiments. We point out how it allows us to evade the CDHSW constraints.PACS number͑s͒: 14.60. Pq, 14.60.St, 14.80.Mz Accepting the totality of present experimental evidence for neutrino oscillations ͓1-3͔, it is not unreasonable to entertain the idea that there are four light neutrinos. Since the CERN e ϩ e ϩ collider LEP experiments ͓4͔ on the invisible decay of the Z boson tell us that there are only three light doublet neutrinos, i.e. e , , and , the fourth light neutrino s should be a singlet under the electroweak SU(2) ϫU(1) gauge group. Usually, s is assumed to mix with the other neutrinos in a 4ϫ4 mass matrix for a phenomenological understanding ͓5͔ of all the data. However, given that s is different from e,, , it may have some additional unusual property, such as decay. In fact, as shown below, this is a natural consequence of the spontaneous breakdown of the lepton number in the simplest model ͓6͔, and it has some very interesting and verifiable predictions in future neutrino experiments.If only atmospheric ͓1͔ and solar ͓2͔ neutrino data are considered, then hierarchical three-neutrino oscillations withwhere m 1 Ӷm 2 Ӷm 3 , would fit the data very well ͓7͔. Here, is the 1-2 mixing angle 12 , m 3 2 and 23 have been chosen to be 10 Ϫ3 eV 2 and 45°, respectively, to fit the atmospheric neutrino data, and 13 has been taken to be zero for consistency with the CHOOZ reactor neutrino data ͓8͔. For fitting the solar neutrino data, there are three solutions:Small angle Mikheyev-SmirnovWolfenstein ͑MSW͒ effect ͓9͔: m 2 2 ϳ10 Ϫ5 eV 2 , sin 2 2ϳ10 Ϫ3 ; Large angle MSW effect ͓9͔: m 2 2 ϳ10 Ϫ5 eV 2 , sin 2 2ϳ1; Vacuum oscillations: m 2 2 ϳ10 Ϫ10 eV 2 , sin 2 2ϳ1.We now add a fourth neutrino s and assume that it mixes a little with e and to explain the Liquid Scintillation Neutrino Detector ͑LSND͒ data ͓3͔. Since the relevant ⌬m 2 is now about 1 eV 2 , it is natural to take m 4 2 ϳ1 eV 2 , but this hierarchical solution is disfavored ͓10͔, because the observed → e probability ͓3͔ is contradicted by the → data of CDHSW ͓11͔ together with the e → e data of Bugey ͓12͔. However, there are two ways that this conclusion may be evaded. ͑1͒ Let m 4 2 ϳ25 eV 2 , then the constraint due to the CDHSW experiment is not a factor, but now there are three other accelerator → e experiments: BNL-E734 ͓13͔, BNL-E776 ͓14͔, and CCFR ͓15͔, which have bounds close to but allowed by the LSND 99% likelihood contour. This is a marginal hierarchical four-neutrino oscillation solution to all the data. ͑2͒ If 4 decays, then the parameter space for an acceptable solution should open up. For example, in the CDHSW experiment, two detectors ...