This thesis investigates the structure and properties of end-linked polydimethylsiloxane (PDMS) elastomers prepared with multiple, discrete molar mass distributions. Even in unimodal networks, which contain elastic chains of a single molar mass distribution, the length of the precursors used to synthesize an end-linked polymer network will have a great influence on the network's mechanical properties. When the average molar masses of the short and long chain components in bimodal networks are widely separated, improved mechanical properties can result. For instance, these elastomers can be stretched to high elongation ratios and display an upturn in stress before fracture. Therefore, the best bimodal networks can absorb more energy before failure than unimodal networks with similar low strain properties. Experimental determination of mechanical properties for a wide variety of bimodal networks systematically identifies which compositions demonstrate optimal performance. These results are combined with theoretical calculations and comparisons to models of rubber elasticity to provide evidence of the microstructure connected with mechanical reinforcement. The methods are then extended to a series of PDMS elastomers with three precursor molar mass distributions. Some of these trimodal networks have mechanical properties which surpass even the bimodal networks. The relationship between the structure of these end-linked networks and their mechanical properties under applied stress are examined at the molecular scale using 2 H-NMR spectroscopy. Unimodal networks show decreasing spectral wings with increasing precursor molar mass, indicating that the longer chains are less perturbed from their initial states when changing from the melt state to the elastic state. These general lineshapes persist for the selectively labeled short and long chains in bimodal networks, and the short chain networks with the best mechanical properties show the most prominent spectral wings. 2H-NMR experiments on stretched samples reveal that the short chains sustain more of the applied load than the long chains in optimal networks. Longer chains in unimodal or bimodal samples can display inner and outer doublets at high strain that are related to the excluded volume interactions between segments and the effect of the chemical cross-links, respectively.