Firm principles which can be relied on to analyze and discuss textual and graphical code representations are still missing. We propose a framework relying on ScanVis, an extension of the Semiology of Graphics that models the perception and scanning of abstract graphics, to model and to provide plausible explanations of phenomena pertaining to the visual perception of representations of code. This framework unifies many aspects of the visual layout and appearance of programming languages and reveals similarities and substantial differences in the visual operations required by those notations. We also show how the framework may help compare and generate representations of programming languages with respect to visual perception. This work suggests that the gap between textual and graphical languages is narrow, and that all kind of programming languages should rely on the capability of the human visual system.