The purpose of this thesis was to study secondary school students' motivation to learn a second foreign language in addition to English. In addition to the empirical investigation of L3 motivation over a program of study and the testing of the widely-held assumption that L2 English impacts negatively on L3 motivation, the aim was also to contribute to the conceptual development of self-based motivation theory by examining the evolution and development of language-speaking/using selves, and by addressing the issue of interference between different self-guides.  In Studies I and II the L3 motivational trajectories of two samples of secondary school students (n=532, n=169) were mapped across grades 4 – 6 (Study I) and grades 6 – 9 (Study II), with a particular focus on differences in the trajectories of girls' and boys' ideal language-speaking/using selves. The results of Studies I and II revealed a pattern where initial gender differences, although remaining stable after a year of learning, thereafter follow different developmental paths. While boys' ideal L3 selves declined by the end of grade 9, girls' ideal L3 selves became stronger. Although a similar pattern was found for L2 English selves, the gender gap here was not as marked.  In Study III the hypothesis that, as a result of negative cross-referencing between ideal L2 and ideal L3 selves, L2 English would have a negative effect on L3 motivation was tested in a sample of 9th grade students (n= 101). Analysis of the data indicates that students are aware of the ideal L2 English self in L3 learning situations and support was found for the hypothesised negative effect on L3 motivation, with the impact being stronger among boys. In Study IV the hypothesised processes of negative cross-referencing were examined in a series of in-depth interviews with four participants selected using a maximum variation sampling strategy. Analysis of the data revealed that when cross-referencing takes place, some students seem to invoke counteracting resources. In the discussion of the findings it is suggested that, rather than interference, competition may provide a conceptually more coherent descriptor of the processes of cognition that take place in the working self-concept when more than one possible language self is active. The implications of the findings for theoretical development are discussed in relation to both qualitative applications of the L2 Motivational Self System (Dörnyei, 2005), and the proposed ID component in the Dynamic Model of Multilingualism (Herdina & Jessner, 2002). Finally, the educational implications of the findings are discussed and a series of proposals for classroom interventions are put forward