In many sciences, including cognitive science and biology, it is assumed that certain physical systems process information and effectively realize computation. For example, it is being claimed that DNA is being decoded in a manner that is best described as computational or that brains are analog computers. The skeptics, however, propose that the notion of computation is purely in the eye of the observer and computational properties cannot be hold to be objective. In this paper, the author discussws the criteria for realistic ascription of computational properties to physical systems. Computational ascriptions are treated as a kind of abstract mathematical ascriptions, and he shows in what sense these ascriptions are not merely conventional but refer to natural kinds. Along with general criteria that apply to other abstract properties being ascribed in sciences, such as explanatory and predictive value and implementation of functional properties vs. instantiation, he discusses specific problems of computational descriptions. The proposed criteria cover both analog and digital computation as kinds of information processing. As a result, the claims in biology about the nature of DNA information decoding turn out to be empirical and falsifiable, and not decidable a priori in a philosopher's armchair.