The cosmological holographic principle, which states that the total observable entropy of de Sitter space (including gravitational quanta) is bounded by S < π/H 2 , where H is the expansion rate, is used to estimate the magnitude of quantum-gravitational effects on inflationary perturbations. The constraint is shown to imply that the initial states of the vacua for perturbation modes are not independent, and that field theory is not a reliable tool to study transplanckian effects on mode amplitudes and phases. It is argued that holographic discreteness alters the continuous, random-phase gaussian distribution predicted by standard field theory for fluctuations in the inflaton field that lead to cosmic background anisotropy. A toy model, applied in the context of scalar inflaton perturbations produced during standard slow-roll inflation, and assuming that horizonscale perturbations "freeze out" in discrete steps separated by one bit of observable information, predicts discrete steps in anisotropy separated by ∆T ≈ 10 −10 K. It is conjectured that the Hilbert space of a typical observable perturbation is equivalent to that of no more than about 10 5 binary spins (approximately the inverse of the final scalar metric perturbation amplitude, independent of H and other parameters), and that some manifestations of this discreteness may be observable.