The study of developing interpreting competence is an insightful perspective to explore how L2 learners develop complex language skills. It is unclear how relevant cognitive resources are mobilized in interpreting training to produce intended progress. To clarify this issue, we conducted a longitudinal experiment exploring changes in the relationship between English-to-Chinese/Chinese-to-English (L2-L1/L1-L2) consecutive interpreting (CI) competence and two related capacities (language competence and memory capacity) in interpreter trainees. Thirteen tasks were administered at the 2nd and 10th month of CI training (Time 1 and 2) to measure CI performance, language skills and working memory spans. There were two main results. First, Pearson correlation analyses revealed a general pattern in which more training resulted in more correlations between cognitive abilities and CI performance. As there was little L1-L2 CI training for these participants, CI performance was only correlated with general English proficiency at Time 1; at Time 2, it correlated with three indices (English listening span, source language listening comprehension and summary writing). For the L2-L1 direction on which participants were mostly trained, CI performance was correlated with the same six indices at both timepoints (general English proficiency, source language listening comprehension and summary writing, English listening and speaking span, Chinese speaking span). Second, structural equation modelling showed that a valid model of CI competence could only be established at Time 2, with working memory influencing CI performance via language competence (which directly influences CI performance). We conclude that the development of interpreting competence results from the self-organization of relevant systems, whereby a better-coordinated structure of mobilised components emerges. Implications for the development of complex language skills and future research directions will be discussed.