This study examined the impact of AR on student learning outcomes in an introductory linguistics course, explored cognitive load (CL) across topics, identified factors contributing to CL, and assessed students’ perceptions of AR usefulness (PU) and ease of use (PEU). A mixed-methods, single-group design integrating both quantitative and qualitative analyses was employed. Quantitative data were collected from 15 undergraduates. Pre- and post-tests and CL questionnaires were administered to these students. The pre-test assessed language evolution knowledge, regional pronunciation and grammar differences, social influences on language use, and the language and culture relationship. The post-test measured knowledge gained after the AR-based learning experience, focusing on language acquisition at different life stages, nonverbal communication as a structured linguistic system, and the development of a written communication system. The adapted CL questionnaire consisted of 7 items: 3 assessing Extraneous Cognitive Load (ECL), 2 for Intrinsic Cognitive Load (ICL), and 2 for Germane Cognitive Load (GCL). For Qualitative data, 15 students participated in semi-structured interviews. In each session, two students were interviewed by two interviewers. After transcription, the interview data were thematically analyzed using a predefined coding scheme based on CL theory (ICL, ECL, GCL) and the Technology Acceptance Model (PU, PEU), with clear category definitions and coding examples ensuring consistency by two independent raters. For example, the statement “The content was too complex to understand” was coded as “complex material” under ICL Category, while “it (AR) makes me understand the material” was coded as “facilitate students learning” under PU. Interrater reliability was high, Cohen’s kappa = 0.947.

The results indicated a significant improvement in student learning outcomes after using AR (t(14) = 2.37, p = .033, d = 0.61). The nonparametric Friedman test examined the differences in CLs across different topics. The results revealed a statistically significant difference among the CL types, χ2(2)=39.414,p<.001. The mean rank scores indicated that GCL was the highest (2.53 ), suggesting greater cognitive engagement in meaning-making, followed by ICL (2.03) and ECL (1.44), which were relatively lower. Qualitative findings showed that ECL stemmed from unfamiliar instructional formats and information overload, while real-world connections and group discussions supported learning. Technical challenges in Gesture Language affected usability, whereas Written Language benefited from instructional support. Students’ perceptions of AR’s PU evolved over time, with positive PU emerging in the last three topics as scepticism about AR faded. However, usability challenges remained as PEU perceptions varied; while students reported positive PEU in Gesture and Written Language, negative PEU emerged in FLA and SLA, indicating early adaptation difficulties.

These findings underscore AR’s potential to enhance conceptual learning by balancing CLs and bridging abstract theories with real-world experiences. Future research should explore targeted instructional strategies, such as structuring AR interactions to minimize cognitive overload, integrating AR with collaborative discussions to reinforce learning, and optimizing content presentation to enhance perceived usefulness. Additionally, technical optimisations are needed to maximize AR’s educational impact.

Keywords: Augmented Reality, Cognitive Loads, Technology Acceptance Model.