This study examines the effect of modality and redundancy in an Augmented Reality (AR) based language learning environment for teaching vocabulary in a foreign language. The study aims to determine the effectiveness of the redundancy and modality principles, examine the relationship between cognitive load, satisfaction, anxiety, willingness, and achievement, identify predictors of achievement, and understand how students perceive AR-based vocabulary learning in three different presentation modes (animation + narration + text, animation + narration, and animation + text). The study used a pre- and post-test control group quasi-experimental design, with 97 Turkish-speaking undergraduate students from the Faculty of Education at a state university participating. The study participants were divided into three groups and practiced using the same AR-based flashcards in three different modalities (animation + narration + text, animation + narration, and animation + text). After completing a short distractive task during the practice session, the participants were given an achievement test. The analysis of the data revealed that while the group that used animation + narration outscored the other two groups, the difference was only significant between the animation + text group and the animation + narration group, which supported the modality principle. However, there was no significant difference between the animation + narration + text group and the animation + narration group, which did not support the redundancy principle. The study also included qualitative data from a focus group interview to support the quantitative findings. The conclusion of the study is that the principles of CTML (Cognitive Theory of Multimedia Learning) should be reconsidered in relation to the use of AR technology in instructional settings.