Any theory of a phenomenon should include a precise definition and description of its target phenomenon (or explanandum), as well as clear demarcation of conceptually and empirically different phenomena. Theories of dreaming should clearly state i) in what way dreaming is a different type of phenomenon from sleep (or any particular stage of sleep), and ii) in what way dreaming is a special form of mental activity occurring during sleep. In our approach the starting points are that while sleep and its different stages can be defined by objective behavioural and neurophysiological criteria, dreaming is a subjective phenomenon; a special, complex altered state of consciousness that can be differentiated from simple sleep mentation. Quite independently from any functional considerations, the general, universal form of dreaming, as most dream researchers currently agree, is a complex, multi-modal simulation of the sensory perceptual world, inhabited by a simulated self or a self-model (Hobson 2009; Metzinger 2003, 2013; Nielsen 2010; Windt 2010). A fruitful idea in biology is that form suggests function; thus the form that dreaming takes, a world-simulation, most likely suggests that the major functions of dreaming have something to do with world-simulation. The most frequent dream contents are therefore the most likely candidates for reflecting the specific function(s) of dreaming: how, when, under what circumstances, and what contents to simulate. Thus, to state that dreaming is an internal world-simulation is to describe the general form that this phenomenon universally takes, but not necessarily its function. The function(s) of the simulation, according to our view, are mainly related to the specific contents selected for simulation.[1]
Furthermore, a proper theory of dreaming should be simple yet covering, so that the same general principles apply to many types of dreams, including the pathologies of dreaming, animal dreaming, and other special cases; the theory should be fruitful, so that it leads to new ideas, hypotheses, and new directions for active research; it should be empirically testable, so that it leads to risky predictions whose accuracy can be objectively checked. It should have both predictive and explanatory power.
Of course, these virtues are desirable in any scientific theory of any phenomenon. When there are rival theories of the same phenomenon, they should be compared with regard to their overall strengths and weaknesses as scientific theories. If they are consistent with each other, perhaps they can be combined into a single, more covering theory. If they are inconsistent with each other, their differing predictions should be empirically tested. After their relative strengths and weaknesses are compared, it should be possible to say which ones are stronger than others.