This paper provides an introduction to the mathematical models that describe the shape of the term structure of interest rates across time. In essence, all these so-called term structure models are driven by the assumption that arbitrage opportunities are absent. The intuitive concept of absence of arbitrage can be linked directly to the existence of a pricing kernel and a risk neutral probability measure. The latter concepts are at the heart of the finance literature and play a unifying role in it. Moreover, by assuming that the state of the economy is well-described by factors that follow diffusion dynamics, factor-dependent expressions for prices and yields can be derived. Typically and for reasons of tractability, additional model assumptions are imposed on the factor dynamics, giving rise to the so-called affine class of term structure models. We discuss the fundamental trade-off between empirical flexibility and theoretical rigor that applies to all models within the affine class of term structure models. Recently, the class of quadratic term structure models has been proposed and seems to outperform the affine class in terms of matching the economic moments of the yield curve. However, given the lack of uniform data samples and the widely differing estimation methods, much robustness work remains to be done.