Accounting for the perceiving conditions that make up the delivery environment of an application helps improve the efficiency of QoS provisioning systems. It is useless to transmit information whose absence cannot be noticed by the end-user under the actual perceiving conditions. Until now, QoS architectures mostly focused on the transport system and did not integrate the studies achieved in the psychophysics area. In this paper, we propose a QoS framework that accounts for the perceiving conditions. Moreover, our framework is both flexible - i.e., customizable at will by the manager - and modular, with a clear and fine-grained layering. New mechanisms and their supporting characteristics, such as experimental curves, may be very easily introduced and managed in the proposed architecture. A case study is carried out, which shows the applicability of the framework for a video-on-demand provisioning system.