We demonstrate qualitatively that frictional instability theory provides a context for understanding how earthquakes may be triggered by transient loads associated with seismic waves from near and distance earthquakes. We assume that earthquake triggering is a stick-slip process and test two hypotheses about the effect of transients on the timing of instabilities using a simple spring-slider model and a rate- and state-dependent friction constitutive law. A critical triggering threshold is implicit in such a model formulation. Our first hypothesis is that transient loads lead to clock advances; i.e., transients hasten the time of earthquakes that would have happened eventually due to constant background loading alone. Modeling results demonstrate that transient loads do lead to clock advances and that the triggered instabilities may occur after the transient has ceased (i.e., triggering may be delayed). These simple "clock-advance" models predict complex relationships between the triggering delay, the clock advance, and the transient characteristics. The triggering delay and the degree of clock advance both depend nonlinearly on when in the earthquake cycle the transient load is applied. This implies that the stress required to bring about failure does not depend linearly on loading time, even when the fault is loaded at a constant rate. The timing of instability also depends nonlinearly on the transient loading rate, faster rates more rapidly hastening instability. This implies that higher-frequency and/or longer-duration seismic waves should increase the amount of clock advance. These modeling results and simple calculations suggest that near (tens of kilometers) small/moderate earthquakes and remote (thousands of kilometers) earthquakes with magnitudes 2 to 3 units larger may be equally effective at triggering seismicity. Our second hypothesis is that some triggered seismicity represents earthquakes that would not have happened without the transient load (i.e., accumulated strain energy would have been relieved via other mechanisms). We test this using two "new-seismicity" models that (1) are inherently unstable but slide at steady-state conditions under the background load and (2) are conditionally stable such that instability occurs only for sufficiently large perturbations. For the new-seismicity models, very small-amplitude transients trigger instability relative to the clock-advance models. The unstable steady-state models predict that the triggering delay depends inversely and nonlinearly on the transient amplitude (as in the clock-advance models). We were unable to generate delayed triggering with conditionally stable models. For both new-seismicity models, the potential for triggering is independent of when the transient load is applied or, equivalently, of the prestress (unlike in the clock-advance models). In these models, a critical triggering threshold appears to be inversely proportional to frequency. Further advancement of our understanding will require more sophisticated, quantitative models and observations that distinguish between our qualitative, yet distinctly different, model predictions.