This dissertation explores the effect of visuals on test-takers’ L2 listeningprocesses. It uses a mixed methods design, following a concurrent transformative model (Creswell & Poth, 2018). There are three components to the study. The first is a comparison of group performance, aiming to shed light on how test-takers’ test performance may be affected by the inclusion of visuals. A total of 190 undergraduate students in China took part in this component. After being randomly and evenly placed in two groups according to their estimated proficiency level, their listening ability was assessed by a pre-test. They then took one of the two versions of a post-test, one where the participants could both see and hear the speakers, and the other one where they could only hear the speakers. A comparison of the two groups’ post-test scores using ANCOVA showed that the audio-visual participants scored significantly higher than the audio-only participants on the post-test. This indicates that the inclusion of visuals significantly enhanced the participants’ test performance. The second component of the study is a retrospective cued recall that investigatestest-takers’ specific listening processes. Ten participants in each group were randomly selected to join this part to the study. Their post-test test-taking processes were recorded and served as the stimuli for them to recall their viewing patterns, note-taking practice, identification of authentic features, and question-answering approaches. The comparisons between the two groups showed individual and group similarities, along with differences in every aspect of the listening processes. The last component is an analysis of the notes taken by the test-takers. Thequalitative participants’ notes taken during the post-test were rated regarding the correct levels of information noted, overall organization, and irrelevant and incorrect information noted. Using Spearman’s correlations, the note quality was correlated with the post-test scores in each condition. No significant correlations were found. This indicates that there was no substantive relationship between the participants’ note quality and test scores. Based on the results of the study, I advocate for the practice of including visualsin L2 listening tests, and an expansion of the construct definition of academic lecture listening to include the assessment of test-takers’ ability to understand visual cues. I also propose a five-dimension authenticity scale to measure the authenticity level of L2 listening input used on L2 listening tests. The study also has several other practical implications that are useful for L2 listening research, test development, and teaching and learning. Key limitations are acknowledged and discussed, and future research directions are suggested. / Teaching & Learning