Abstract: At national level, using gross expenditure on R&D (GERD) and researchers in full-time equivalents (RFTE) as S&T (scientific and technical) input, with high cited papers and triadic patents as S&T output respectively, it is revealed that C-D production function, CES production function and VES production function show strong correlations between S&T input and paper output, but show relatively weak explanatory ability to the data variation of S&T input and patent output. Based on the analysis of output elasticity and returns to scale, it is pointed out that neither excess input of GERD nor that of RFTE is conducive to increasing paper output or patent output and that model select also plays an important role. For improving the performance of scale, it is suggested to adjust dynamically S&T input referring to the equation φEGERD+ψERFTE=1 on the premise of grasping the specific relationship between paper output and patent output.