<![CDATA[Integrating visual elements into high-stakes science examinations requires balancing instructional scaffolding with cognitive load optimization. This study investigates the structural and functional topography of visual representations by conducting a total population content analysis of Singapore Primary School Leaving Examination Science papers across a three-year longitudinal window from 2023 to 2025. The examined corpus evaluated every visual representation deployed across all 120 multiple-choice and short-answer test items (NMCQ=84; NSAQ=36). Guided by Cognitive Load Theory and Mayer’s Coherence Principle, 134 extracted visual units were classified across four typologies (Representational, Abstract, Data-Driven, Relational) and three functional levels (Essential, Supporting, Decorative). A critical baseline finding for item designers was the absolute absence of decorative visuals (0.00%), with 97.01% (n=130) categorized as functionally essential for problem-solving, validating a construct-driven approach to assessment layout. A Pearson’s Chi-Square Test of Independence revealed that the deployment of visual types is significantly dependent on the specific scientific theme (chi2(12)=26.76, p=0.0084), with the Cycles thematic block capturing 100% of all relational process graphics. Longitudinal analysis revealed an intentional formatting shift. While multiple-choice visual density halved over three years (from 39 down to 19 elements), representational visuals within short-answer items nearly doubled (from 6 to 11). This trajectory reveals an evolutionary transition away from utilizing graphics as analytical data puzzles toward deploying them as generative contextual scaffolds for written explanations. These findings offer item writers concrete benchmarks for optimizing cognitive processing and construct validity in international STEM testing environments.]]>
Copyrights © 2026
