Background: Understanding algorithms is crucial for programming education, yet their abstract nature often challenges students. Algorithm visualisation (AV) has been proven effective in enhancing algorithmic thinking among university students. However, its efficacy for elementary school students and the optimal forms of AV tools remain unclear. Objectives: This study aims to assess learners' performance, motivation, and behaviour under three AV forms (i.e., algorithm animation, static visualisation, and no visualisation) from both scientific and behavioural perspectives. Methods: A quasiexperimental design was employed, involving 104 sixth-grade students (aged 11-12) from a K-12 school in eastern China. A 9-week algorithm-teaching activity covering the optimal path, enumeration, and search algorithms was conducted in an in-school extension class. Two experimental groups and one control group each used a different AV form. Quantitative data were collected through questionnaires and an algorithm competency test (ACT), whereas behavioural data were analysed from computer screen recordings and classroom video recordings. Results and Conclusions: Although no significant differences were found in overall learning performance, algorithm animation was particularly beneficial for high-proficiency students. Algorithm animation and static visualisation significantly enhanced students' learning motivation compared with no visualisation. A behavioural analysis revealed that students using algorithm animation demonstrated greater autonomy and initiative, whereas those students who did not use visualisation preferred passive learning. This study on AV-based algorithm teaching concludes that introducing AV effectively improves students' initiative and motivation, providing insights for integrating visualisations in instructor-mediated classrooms.