This dissertation examines three important questions with regard to third party or independent candidates in American presidential elections. First, this dissertation tests standard theories of American voting behavior when citizens have a third option to exit from the two-party system. In particular, the spatial theory of electoral competition provides a useful guidelines and hypotheses. In the context of American presidential elections, this research posits that the effects of ideology and issues are largely dependent upon perceived spatial positions of third party candidates. The empirical evidence suggests that when the ideological position of a third party candidate is extreme, e.g., Wallace's perceived placement in 1968, the effect of the ideological distance is relatively large and the vote choice model explains a significant portion of the probability of voter choice of each candidate. In contrast when a third party candidate reveals a centrist tendency, e.g., Perot's perceived placement in 1992 or 1996, the effect of the ideological distance is substantially reduced. The empirical evidence also adds a more weight to these findings in that voters are more heavily influenced by nonpolicy factors such as partisanship, candidate personal qualities, and economic evaluations in the latter type of election. The second research question is related to the electoral strength of third party candidates. The empirical evidence suggests that voters' disaffection from the current two-party system is a key element of the third party success. Nonetheless, this paper finds that the substitutability of third party candidates, e.g., policy innovations and high quality candidacies, should be regarded as a more important source of third parties' electoral strength. The third research area investigates the electoral weakness of third party candidates. Extending Duverger's psychological effect, this paper examines the wasted vote argument or strategic voting among third party supporters in multicandidate American presidential elections. The statistical analysis finds that voters behave so as to maximize expected utility. The empirical evidence confirms that the net difference in expected utility between the first and second preferences significantly influences voters' sophisticated calculus of strategic choice. Overall, this research contributes to the existing knowledge of American voting behavior by providing in-depth analysis of electoral dynamics introduced by third party candidates. The theoretical framework as well as the empirical analysis suggests that a spatial model of multicandidate elections deserves further examination.