Identification in the limit, originally due to Gold [Gold, Information and Control, 1967], is a widely used computation model for inductive inference and human language acquisition. We consider a nonconstructive extension to Gold's model. Our current topic is the problem of applying the notions of reliability and refutability to nonconstructive identification. Four general identification situations are defined and two of them are studied. Thus some questions left open in [Kucevalovs, 2010] are now closed.