Membrane separation of gases has evolved into an important separation technology for Various gas mixtures (specifically O-2/N-2). Aromatic engineering polymers such as polysulfones, polycarbonates, and polyimides comprise commercially utilized membranes for these applications. The ability to predict permeability and permselectivity from polymeric structural units is highly desired in order to streamline synthetic approaches to optimum membrane candidates. A group contribution methodology is outlined in this paper which demonstrates excellent predictability of permeability (for O-2, N-2 and He) and good prediction of permselectivity for the O-2/N-2 and He/N-2 gas pairs. This procedure utilizes the basic equation : lnP = Sigma(i=1)(n) empty set(i) ln P-i where empty set(i)=volume fraction of a structural unit i and P-i=the permeability contribution of the structural unit Experimental permeability data are employed to set up an array of equations (of the above equation) solved by least squares fit. The values of empty set(i) are calculated using computer software programs to predict molar volume contributions. The structural units are chosen around the chemical bond. This procedure shows promising results when applied to aromatic polymers chosen from the classes of polysulfones, polycarbonates, polyarylates, poly(aryl ketones) and poly(aryl ethers). This procedure has been utilized to determine the contributions of 24 structural units employing 65 polymers which comprise the database. Excellent agreement within the database is observed and good agreement outside the database is also demonstrated. This procedure allows for a quantitative assessment of the structure/permeability (permselectivity) relationships for polymers of interest for membrane separation, and thus demonstrates group contribution methodology can be applied to both polymer permeability and permselectivity. Further refinements by addition of other polymeric classes (e.g. polyimides and polyamides) as well as additional expansion of the database should prove to be a valuable technique to predict the membrane separation potential of a wide variety of polymeric materials.