The interactions among the components of a biological system can be given a logical representation that is useful for reasoning about them. One of the relevant problems that may be raised in this context is finding what would explain a given behaviour of some component; in other terms, generating hypotheses that, when added to the logical theory modelling the system, imply that behaviour. Temporal aspects have to be taken into account, in order to model the causality relationship that may link the behaviour of a given component to that of another one. This paper presents a hypothesis generation method for linear temporal logic theories whose formulae have a restricted syntactic form, which is however sufficient to model cellular and molecular interactions, as they are often represented by biologists. The method exploits the duality between hypothesis generation and consequence finding, and can therefore be also used to infer the consequences of a given fact. It is based on a resolution system proposed by Cavalli and Fariñas del Cerro in 1984.