Augmented reality (AR) can represent a contextualised scientific inquiry environment in which students may explore the real world and develop science process skills via interacting with rich information from virtual systems. However, it remains a challenge for most students to complete scientific inquiry tasks without proper support. Research evidence has suggested the potential of reflective scaffolding when applying scientific inquiry. Accordingly, we designed a contextualised reflective mechanism-based AR learning model to assist students in completing scientific inquiry tasks. Guided by the proposed model, we designed four stages of scientific inquiry learning: conceptual understanding, reflective cognition, in-depth inquiry, and knowledge building. A quasi-experiment and lag sequential analysis were conducted by recruiting 81 sixth-grade students to examine the effects of the proposed model on their scientific inquiry learning performances, higher order thinking, and behavioural patterns. The experimental results reveal that the proposed approach improved students' inquiry learning performances and higher order thinking tendency (problem-solving tendency and metacognitive awareness). Moreover, the evidence from this study also suggests that the students who learned with the proposed approach exhibited more observation, comparison, exploration, and reflection behavioural patterns in the field trip than those who learned without the contextualised reflective mechanism. Implications are discussed.