This paper generalizes the Debreu/Gorman characterization of additively dec
omposable functionals and separable preferences to infinite dimensions. The
first novelty concerns the very definition of additively decomposable func
tion as for infinite dimensions. For decision under uncertainty, our result
provides a state-dependent extension of Savage's expected utility. A chara
cterization in terms of preference conditions idensitifies the empirical co
ntent of the models; it amounts to Savage's axiom system with P4 (likelihoo
d ordering) dropped. Our approach does not require that a (probability) mea
sure on the state space be given a priori, or can be derived from extraneou
s conditions outside the realm of decision theory, Bayesian updating of new
information is still possible, even though no prior probabilities are give
n. The finding suggests that the sure- thing principle, rather than prior p
robability, is at the heart of Bayesian updating.