Cognitive Strategies in Interpreting UML Class Diagrams with Emphasis on Load, Order and Symbolic Confusion

Cognitive Strategies in Interpreting UML Class Diagrams with Emphasis on Load, Order and Symbolic Confusion | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Cognitive Strategies in Interpreting UML Class Diagrams with Emphasis on Load, Order and Symbolic Confusion Pakiso J. Khomokhoana, Rouxan C. Fouché This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9266264/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 12 You are reading this latest preprint version Abstract Unified Modelling Language (UML) class diagrams are critical in computing education, yet students face significant cognitive challenges in interpreting these complex visual-technical representations. Building on prior semiotic analyses of UML class diagram cognition, this narrative study (based on the integrated methodology) applies the Cognitive Load Theory (CLT) to explore how 11 first-year computing students at a South African university manage mental load during the interpretation of UML class diagrams. Through the semi-structured interviews, student-created artefacts, and systematic behavioural observations across three activities of increasing complexity ( simple element identification, relationship analysis, and complex diagram construction ), we identified four primary cognitive strategies: analytical sequencing and prior knowledge; symbolic confusion and clarification; diagram orientation and redrawing; and cognitive load management through decomposition. Participants demonstrated metacognitive awareness of working memory limits, employing adaptive strategies like spatial rearrangement, symbolic error correction, and systematic breakdown of visual complexity. Two key sources of extraneous load emerged: symbolic confusion from ambiguous notation and spatial disorientation from unconventional layouts, prompting compensatory behaviours such as diagram redrawing and self-correction. This exploratory study provides preliminary evidence for the potential application of CLT to visual-technical learning, suggesting that these participants acted as active cognitive agents. Pedagogical implications include explicit training in systematic analysis, symbol fluency, standardised visual designs, and metacognitive strategies. These findings inform cognitive load management in technical education and suggest directions for broader validation. Cognitive load theory cognitive strategies diagram interpretation computing education UML class diagrams visual-technical learning Figures Figure 1 Figure 2 Figure 4 Figure 5 1. Introduction Unified Modeling Language (UML) class diagrams are a cornerstone of software engineering and information systems education (Apostol et al., 2024 ). These diagrams require students to integrate visual, logical, and symbolic reasoning when interpreting classes, relationships, and hierarchical structures (Fu et al., 2017 ; Hafeez et al., 2019 ). Recent systematic reviews have explicitly noted a lack of in-depth research on cognitive processes in model comprehension, despite numerous studies on diagram quality and tooling (Koç et al., 2021 ; Möller et al., 2025 ). Recent work has begun to examine these cognitive processes. Our previous studies identified strategies such as attention management, memory activation, and reflective thinking when students tackled UML class diagram related problems (Khomokhoana & Nkalai, 2024 ). Applying semiotic theory, we also found that visual features strongly shape understanding, while ambiguous terminology and contextual differences often lead to confusion (Khomokhoana et al., 2025 ). However, these studies did not explicitly address how students manage mental effort, a gap that is critical given the cognitive demands of interpreting UML class diagrams. Cognitive Load Theory (CLT) often appears in computing education research, but researchers rarely engage deeply with its recent developments (Duran et al., 2022 ). This gap is clear in visual-technical learning, where most CLT studies focus on text-based programming rather than visual modelling languages. The connection between diagrammatic reasoning and software engineering education remains largely unexplored (Koç et al., 2021 ), despite research showing strong links between diagram interpretation skills and programming success (Tóth & Pogatsnik, 2022 ; Zhou et al., 2023 ). UML class diagrams create unique mental demands. For example, students must simultaneously process visual symbols, spatial relationships, text elements, and abstract concepts, such as inheritance and composition (Bera, 2012 ). This multi-modal processing places a heavy demand on working memory, which research shows has severe limits (Sweller et al., 2011 ). This study employs the CLT approach to investigate how first-year computing students interpret UML class diagrams, with a focus on their mental load management. Building on our previous work, we ask: What cognitive strategies do students use when interpreting UML class diagrams? How do students manage mental overload, and what breakdown strategies do they develop? What creates unnecessary mental burden, and how do students adapt? How do cognitive strategies change as UML class diagrams become more complex? The remainder of this paper unfolds as follows: Section 2 introduces the CLT framework as it relates to visual-technical learning. Section 3 describes the study’s methodology, and Section 4 reports the findings. Section 5 presents the interpretation of these results considering theoretical and practical implications. Section 6 summarises the study contributions, and Section 7 concludes with limitations and avenues for future research. 2. Theoretical framework This study adopts the CLT as its theoretical lens to examine students’ cognitive strategies when interpreting UML class diagrams. Initially proposed by Sweller ( 1988 ) and further developed by Sweller, Ayres, and Kalyuga ( 2011 ), CLT centres on the limitations of working memory and how instructional design can hinder or enhance learning. At its core, CLT recognises that working memory is limited and that learning is most effective when cognitive resources are optimally allocated (Paas et al., 2003 ). 2.1 Cognitive load theory in visual-technical learning CLT explains how the limitations of working memory shape learning outcomes (Sweller, 1988 ; Sweller et al., 2011 ). Effective learning occurs when instructional design supports the allocation of cognitive resources between three load types: Intrinsic load stems from the inherent complexity of the learned content or task (Van Merriënboer & Sweller, 2010 ; Young et al., 2014 ). In interpreting UML class diagrams, intrinsic load includes understanding object-oriented concepts (e.g., inheritance, composition, aggregation), recognising hierarchical relationships, and comprehending the logical structure of system designs (Boustedt, 2010; 2012 ). This load is unavoidable and directly related to the learning objectives (Sweller et al., 2011 ). Importantly, intrinsic load cannot be reduced without affecting learning objectives; hence, effective management of other load types is required. Extraneous load is generated by suboptimal instructional design, including confusing layouts, unclear symbols, or irrelevant information (Van Merriënboer & Sweller, 2010 ). Chandler and Sweller’s ( 1991 ) research on the split-attention effect reveals how poorly designed visual materials can overwhelm working memory by requiring students to mentally integrate different information sources. Extraneous load in UML diagrams’ settings can occur when inheritance arrows point horizontally rather than vertically, novel language is used without explanation, or diagram elements are placed unexpectedly (Masri, 2009 ). Germane load reflects the mental effort devoted to meaningful learning processes, such as organising knowledge and integrating new information with existing understanding (also known as cognitive schemas) (Van Merriënboer & Sweller, 2010 ). According to Paas and Van Merriënboer (1994), students who actively engage in strategies such as chunking information, drawing connections, or systematically analysing complex diagrams devote more cognitive resources to building deep, organised understanding rather than to surface-level processing. This constructive utilisation of cognitive resources results in deeper comprehension and improved knowledge transfer. In the interpretation of UML class diagrams, intrinsic load arises from simultaneously processing visual symbols, spatial relationships, text elements, and abstract object-oriented principles (Bera, 2012 ). Poor UML class diagram design or ambiguous notation increases extraneous load (Chandler & Sweller, 1991 ; Masri, 2009 ), whereas strategies such as decomposition, the use of analogies, and activating prior knowledge reflect an investment in germane cognitive load (Paas & Van Merrienboer, 1994 ). Recent developments in CLT emphasise its multidimensional nature, particularly in extraneous load (Andersen & Makransky, 2021 ). These developments also call for greater integration with multimedia learning (Mayer, 2014 ), collaborative load theory (Kirschner et al., 2018 ), and technology-enhanced education (Paas & Sweller, 2014 ). However, applications in visual-technical domains such as UML remain limited, despite evidence that spatial reasoning and diagrammatic skills predict programming performance (Osztián et al., 2022 ). It should be noted that CLT has faced criticism regarding the difficulty of empirically measuring different load types separately and the potential for circular reasoning when inferring cognitive load from performance outcomes (Kalyuga & Plass, 2025 ). Despite these measurement challenges, CLT’s conceptual framework offers valuable theoretical guidance for understanding the cognitive demands of interpreting UML class diagrams. 2.2 Diagrammatic reasoning in UML interpretation Dual Coding Theory (Paivio, 1986 ) suggests that visual and verbal information are processed in interconnected but distinct systems. UML class diagrams require both systems: visual processing to recognise shapes and relationships, and verbal processing to interpret labels and abstract concepts (Shen et al., 2018 ). Effective diagrammatic reasoning depends on recognising symbolic conventions, managing spatial complexity, and integrating these with conceptual knowledge (Cheng et al., 2001 ). Students with strong diagrammatic reasoning skills tend to perform better in tasks that involve logical problem-solving and structured thinking (e.g., programming) (Tóth & Pogatsnik, 2022 ). However, when confronted with unconventional layouts or unclear symbols, novices often experience disorientation, triggering compensatory strategies such as redrawing diagrams or reordering elements. These behaviours reflect active load management rather than passive reception. 2.3 CLT and student cognitive strategies Recent reviews have revealed that the cognitive and educational dimensions of diagrammatic reasoning in software engineering remain underexplored. Empirical and review studies tend to emphasise syntactic and technical issues rather than students’ reasoning processes (Koç et al., 2021 ; Möller et al., 2025 ). Castro-Alonso et al. ( 2021 ) and Mohammadi Zenouzagh et al. ( 2025 ) suggest that students are not passive recipients, but can develop adaptive strategies when facing cognitive challenges. However, limited research has systematically examined the specific strategies students spontaneously employ when interpreting complex visual-technical materials such as UML class diagrams (Koç et al., 2021 ; Ma, 2013 ; Nugroho, 2009 ). This study investigates how students naturally respond to cognitive load challenges by examining their think-aloud protocols for evidence of: Systematic approaches to managing intrinsic load complexity. Self-correction and environmental modification to reduce extraneous load. Active engagement strategies that may reflect germane load investment. Understanding these student-generated strategies provides valuable insight into how novice learners manage cognitive load in visual-technical learning contexts. Such insights can deepen our understanding of CLT by illustrating how learners naturally adapt under varying load conditions. They can also inform instructional design aimed at optimising cognitive load management. 3. Research design and methodology 3.1 Research design This study used a narrative inquiry approach, following Plowright’s ( 2011 ) integrated-methods research framework (FraIM). FraIM enables research perspectives to emerge naturally during the study, rather than requiring fixed theoretical positions from the outset. This flexibility proved valuable for exploring cognitive load management strategies, as it allowed for the documentation of students’ natural problem-solving behaviours without imposing rigid constraints. Data collection involved multiple sources: semi-structured interviews, student-created artefacts, and systematic behavioural observations. The research focused on first-year Bachelor of Computer Information Systems (BCIS) students at a South African university. The study employed both purposeful (Singh et al., 2021 ) and convenience sampling (Mishra & Alok, 2017 ; Saunders et al., 2015 ) methods. Purposeful selection ensured participants were studying relevant course content (including aggregation and composition concepts). Convenient factors facilitated the researcher’s access through existing teaching relationships with one research team member. From a group of 105 students, 20 volunteers were selected through an open recruitment process. Eleven students completed the full research protocol. These included five males and six females, aged above 18 years, and with varying levels of programming experience. Institutional ethical approval was secured before data collection (Reference: UFS-HSD2023/1768). 3.2 Data collection procedures Individual semi-structured interviews were conducted using the think-aloud protocol in which participants completed three activities of increasing complexity. These activities followed three levels of increasing complexity, each designed to reveal different cognitive processing patterns: basic understanding (Phase 1), intermediate analysis (Phase 2), and advanced synthesis (Phase 3). Participants completed activities using pen and paper, with their written work serving as artefacts for analysis. Think-aloud protocols captured real-time cognitive processing, providing access to participants’ reasoning strategies. All interview sessions were audio-recorded with the explicit consent of the participants, ensuring complete documentation of their verbal expressions and thought processes. Participant behaviours, emotional responses, and interaction patterns were documented through structured field notes. This contributed to data triangulation and improved interpretive validity. Phase 1 — Initial Assessment Participants examined three unlabeled UML class diagrams (Figs. 1, 2 and 3), identifying and describing what they observed based on their existing knowledge. This phase assessed baseline familiarity with association, aggregation, and composition relationships without external guidance. The names of the diagrams were intentionally omitted (e.g., name withheld ) to require participants to identify them independently. Figure 1. Name withheld? Figure 2. Name withheld? Figure 3. Name withheld? Phase 2 — Relational Analysis Participants analysed and explained the relationships shown in a moderately complex boardgame class diagram (see Figs. 4 and 5). This activity required combining multiple concepts and relationship types. Phase 3 — Construction and Interpretation This phase included two activities. First, participants created original class diagrams based on detailed written specifications covering computer hardware relationships, inheritance hierarchies, and compositional structures. Second, they provided detailed explanations of a pre-constructed class diagram (Fig. 5) representing similar concepts to their own creations. 3.3 Data analysis The audio recordings were transcribed following established narrative research protocols (Creswell & Creswell, 2017 ). Initial data preparation addressed transcription inconsistencies and response clarity challenges common in open-ended questioning. A fuzzy validation approach (Parcell & Rafferty, 2017 ) was utilised to preserve authentic participant voices while ensuring analytical clarity and consistency. Analysis began with intensive data immersion through repeated audio review and comprehensive transcript examination. This familiarisation process informed the development of a structured coding system aligned with our research objectives. The 11 validated transcripts were processed using NVivo software for systematic thematic analysis. Codes were developed to reflect emerging cognitive strategies identified in participant responses, paying careful attention to analytical detail and thematic saturation. Text segments were systematically annotated, assisting in identifying connections between codes and supporting evidence. Code names were refined throughout the analysis to ensure accurate thematic representation, with a particular focus on the frequency patterns and conceptual significance of recurring elements. 4. Findings Think-aloud protocols were analysed for 11 participants who completed the three activities of increasing complexity. The findings revealed four systematic cognitive strategies that align with recent advances in multidimensional cognitive load theory (Andersen & Makransky, 2021 ). These strategies show how participants naturally adapted to the unique cognitive challenges of visual-technical learning environments. Each strategy was associated with specific types of CLT load (intrinsic, extraneous, and germane) and linked to identifiable student actions. These strategies were observed across multiple participants and levels of diagram complexity. To provide an overview of these relationships, Fig. 7 presents a concept map illustrating how each strategy is connected to CLT load types and the observed student actions. This visual represents the interplay between intrinsic complexity, extraneous distractions, and germane load investment as participants navigated tasks involving the interpretation of UML class diagrams. Table 1 summarises these strategies, the associated CLT load types, and representative student actions. This structured synthesis highlights the range of approaches participants used to manage cognitive demands, from sequencing their analysis and activating prior knowledge structures to correcting symbol misunderstandings, redrawing diagrams, and decomposing complex diagrams into smaller, more manageable units. Table 1 Cognitive strategies in UML class diagram interpretation with associated CLT load types and student actions Strategy CLT Load Type(s) Observed Student Actions Analytical sequencing and prior knowledge Intrinsic, Germane Sequential analysis, identifying main classes first, perceptual salience and using prior knowledge schemas or mental networks of prior knowledge. Symbolic confusion and clarification Extraneous, Germane Correcting symbol misinterpretations, and practising symbol fluency. Diagram orientation and redrawing Extraneous, Germane Rearranging layout, and redrawing diagrams for clarity. Decomposition for load management Intrinsic, Germane Breaking diagrams into smaller parts, and sequential processing of components. The following subsections expand on each strategy, illustrating how they manifested in student reasoning and linking them explicitly to CLT load types. The numbers indicate how many of the 11 participants demonstrated each strategy, emphasising patterns of behaviour in this narrative study. The following findings represent patterns observed in our small sample of 11 participants from a single institution. While these patterns suggest potential cognitive strategies, they should be interpreted as preliminary insights requiring replication and validation before broader theoretical claims can be made. 4.1 Analytical sequencing and prior knowledge Sequential analysis and identifying main classes first : Participants exhibited clear patterns in their approach to UML class diagrams. All 11 participants began by identifying the main classes first before moving to subclasses and attributes. This sequential analysis represents a strategy for organizing information into chunks, a behavior consistent with what CLT describes as germane load investment (Sweller et al., 2011 ). P1 explained this deliberate approach: “ The first thing I would do is analyse the diagram ”, showing how participants planned their analysis to avoid mental overload. P7 demonstrated this pattern, starting with the central element: “ Computer is the main class... there is a mouse class, a keyboard class... ”. P10 further showed this sequential analysis: “ I check the main class, then the subclasses, then analyse the components ”. This step-by-step method breaks complex tasks into smaller parts that fit within working memory limits (Paas et al., 2003 ). Perceptual salience : Before beginning their systematic analysis, five participants explicitly mentioned that their attention was drawn to visually prominent features, a response to the principle of perceptual salience. For example, P9 noted: “ Yes, it’s [the diagram] bigger. It’s the biggest thing on the page. Alright, so it caught my attention ”. Similarly, P11 observed: “ If there was maybe a bold word, I’m sure I would first look at that bold word without even looking at something else, it catches the eye ”. This automatic focus on prominent visual elements aligns with CLT principles about effective visual design (Van Merriënboer & Sweller, 2005 ), reducing unnecessary mental effort (Chandler & Sweller, 1991 ). Using prior knowledge schemas : Eight participants actively employed prior knowledge schemas to make sense of new diagrams, hence reducing intrinsic load. In this instance, P6 showed this schema activation when she connected the UML class diagram elements to familiar concepts: “ When I look at these arrows, they remind me of a certain diagram which we did in class... I think it is a component of a vehicle. I see the oil filter, oil checking... ”. By linking new information to familiar ideas, participants attempted to manage the inherent complexity of UML class diagrams. These four actions ( identifying main classes first, using prior knowledge schemas, sequential analysis, and responding to perceptual salience ) show how participants naturally developed strategies to manage mental load. As shown in Table 1 and Fig. 7 , the “ analytical sequencing and prior knowledge ” approach helped participants handle both intrinsic and germane load. 4.2 Symbolic confusion and clarification Correcting symbol misinterpretations : Ten of the 11 participants experienced misinterpretations of UML class diagram symbols, which created an extraneous load and required extra mental resources for error correction. The first key action, correcting symbol misinterpretations, appeared in various forms. For instance, P3 initially referred to an inheritance arrow as a “ method ”, but through self-correction stated: “ Chess is a type of board game ”. This self-correction process is a core part of correcting symbol misinterpretations, demonstrating how participants actively monitored their understanding and adjusted their mental resources when they identified errors. Participants also engaged in correcting symbol misinterpretations by reframing confusing visual elements in familiar terms, such as relating them to vehicle parts or personal experiences. As an example, P1 said: “ I am not familiar with this particular game, but then I know for sure that a board game has particular elements like chess. That is why I was able to see that this one is inheritance ”. While this indicates adaptive thinking, it represents the mental effort required to correct initial misunderstandings, adding extraneous load until the correct interpretation is found. Practising symbol fluency : The second key action emerged as nine participants developed from basic visual recognition to a deeper understanding. P3 particularly demonstrated this progression, starting with visual symbol processing: “ By the look of symbols, this one here is not coloured ”. After learning proper terminology, she could distinguish between relationship types: “ The one that is not shaded shows that it is an aggregation. The one in tile is a composition since it is shaded ”. This transition from surface-level identification to conceptual understanding represents practising symbol fluency in action. It indicates germane load investment, productive mental effort that builds from simple visual recognition to semantic processing. When participants could not effectively use either correcting symbol misinterpretations or practising symbol fluency, they experienced significant extraneous load (Sweller, 2020 ). P3’s initial confusion with inheritance notation forced repeated attempts at correcting symbol misinterpretations, diverting mental resources from learning objectives to error correction. Participants varied widely in their mastery of both actions. Those who had developed strong symbol fluency could focus on understanding system relationships, investing in germane load. However, participants who were still developing this fluency spent more time correcting symbol misinterpretations, using working memory for basic decoding rather than engaging in deeper analysis. This variation appeared to influence how participants engaged with complex diagrams. As shown in Table 1 and Fig. 7 , the “ symbolic confusion and clarification ” strategy involves these two specific actions: correcting symbol misinterpretations (addressing extraneous load) and practising symbol fluency (investing in germane load). Together, these actions reveal how participants actively worked to reduce confusion while building their understanding of UML class diagram notation. 4.3 Diagram orientation and redrawing Rearranging layout and redrawing for clarity : Three participants experienced unnecessary extraneous load from poor visual design when inheritance symbols appeared horizontally rather than vertically on the UML class diagram. P2 specifically addressed this challenge through two key actions: rearranging the layout and redrawing UML class diagrams for clarity. She actively restructured the UML class diagram to match her mental model, explaining: “ Yes [it’s easier] when it’s right below the main class ”. This redrawing of UML class diagrams for clarity represents load management, thereby reducing extraneous load while investing in germane load through spatial restructuring. Seven participants showed how visual conventions affect cognitive load. For example, P2 noted: “ The diamond always faces to the main class ”, while P3 emphasised: “ It is important to know where should the arrows face ”. Participants appeared to form expectations about UML diagram layouts, which may suggest developing mental models for visual processing (Sweller, 2011 ). When these expectations are violated, participants must mentally or physically rearrange the layout, creating an additional cognitive burden. Inherently, design inconsistencies forced participants to unnecessarily rearrange layout activities. For example, P2’s confusion illustrates this: “ It was tricky for me to identify it because of the way it was arranged [drawn]. Is it an inheritance or what? ”. She expected to see vertical inheritance arrows on the UML class diagram, but encountered horizontal ones. The mental effort required to reconcile this mismatch represents extraneous load that could be avoided through consistent visual design. Participants who redrew UML class diagrams reported that they found this helpful. When asked if vertical arrangement was easier, P2 simply replied: “ Yes” . By physically restructuring the UML class diagram to align with her mental model of hierarchical relationships, she reduced the cognitive effort needed for interpretation. This spatial reorganisation indicates how rearranging layout and redrawing UML class diagrams for clarity work together to manage both extraneous and germane load. As shown in Table 1 and Fig. 1, the “ diagram orientation and redrawing ” strategy involves two specific actions: rearranging the layout and redrawing diagrams for clarity. These actions helped participants transform confusing visual UML class diagram presentations into formats that support understanding, hence managing both extraneous load (from poor design) and germane load (through active restructuring). 4.4 Decomposition for load management Breaking diagrams into parts : Four participants demonstrated a clear awareness of their cognitive limits and effectively used breaking diagrams into smaller parts to manage complexity. As an example, P5 warned: “ If they [students] don’t break it down, there is a high likelihood that they will get it wrong ”. This metacognitive awareness, knowing when mental capacity is exceeded, drives the use of decomposition strategies to manage intrinsic load. Sequential analysis : The second key action, sequential analysis, was consistently employed across the four participants who used decomposition strategies. For instance, P7 advised: “ Try to break the UML class diagram into smaller and more understandable pieces ”. This deliberate decomposition technique aligns with Van Merriënboer and Sweller’s ( 2005 ) research on managing complexity in technical learning. By processing elements one at a time rather than simultaneously, participants can maintain mental effort within the limits of working memory. This insight suggests that the participants understand that trying to grasp complex diagrams all at once exceeds working memory capacity. Instead, sequential processing of components keeps cognitive demands manageable. Participants recognised that excessive visual complexity impairs understanding. Breaking UML class diagrams into smaller parts transforms overwhelming visual information into manageable chunks. This systematic division enables the sequential processing of components, handling one element at a time rather than all elements simultaneously. This approach divides high intrinsic load tasks into subtasks that fit within working memory constraints. The consistency of these decomposition strategies across participants suggests that breaking UML class diagrams into smaller parts and sequential processing of components are natural responses to complex visual-technical information. These findings highlight the importance of explicitly teaching both actions as part of UML class diagram education. As mapped in Table 1 and Fig. 7 , “ decomposition for load management ” addresses intrinsic load through germane load investment. The two actions ( breaking diagrams into smaller parts and sequential processing of components ) show how participants actively manage the inherent complexity of UML diagrams through strategic decomposition. 5. Discussion The findings suggest that UML class diagrams carry an intrinsic load due to their structural complexity. However, many student difficulties were caused by extraneous load, specifically inadequate symbol clarity and ambiguous orientations. When participants were able to reduce extraneous factors (for example, by redrawing or utilising analogies) and use deconstruction, they improved their germane load, focusing more on comprehending rather than decoding. This pattern appears consistent with CLT concepts regarding how learners might manage cognitive demands during the interpretation of technical UML class diagrams. 5.1 Theoretical implications for cognitive load theory The systematic analysis of student think-aloud protocols reveals sophisticated cognitive load management strategies, providing insight into how novice learners process complex visual-technical information. These findings suggest potential implications for CLT theory, addressing deficiencies noted in previous systematic evaluations of computing education research. Potential implications for visual-technical learning domains The four cognitive strategies ( analytical sequencing and prior knowledge, symbolic confusion and clarification, diagram orientation and redrawing, and cognitive load management through decomposition ) described in this study suggest considerations for CLT’s traditional focus. Participants in this study employed specific cognitive strategies for processing UML class diagrams that warrant further investigation, raising questions about whether technical visual domains may involve distinct cognitive processes. Observed behavioural patterns suggest that students’ engagement with visual and technical aspects of UML diagrams could be leveraged to tailor instructional support. Cognitive Load Theory suggests a potential framework for such adaptations, recommending strategies such as visual cueing, worked examples, and progressive guidance to optimise learning (Sweller et al., 2011 ). The extensive use of spontaneous cognitive load management through decomposition, diagram orientation, and redrawing approaches demonstrated that participants actively identified and minimised extraneous load sources rather than being passive recipients of instructional design. This study builds on recent research by Andersen and Makransky ( 2021 ), who found that certain types of cognitive load can be advantageous when they foster specific forms of germane processing, notably in visual-technical circumstances where diagram orientation and redrawing improve comprehension. Schema development in technical domains This study’s shift from visual symbol recognition to conceptual understanding sheds light on how technical schemas evolve. Participants initially processed UML diagram elements as distinct cognitive units, resulting in a high intrinsic load; however, with systematic practice, these elements were gradually grouped into meaningful patterns. These observations align with Ericsson and Kintsch's (1995) research on long-term working memory, implying that expertise growth in technical fields involves specialised knowledge structures that successfully increase working memory capacity. The connection between analytical sequencing and prior knowledge, and germane load investment, indicates how participants in this study appeared to develop what Van Merriënboer and Sweller ( 2005 ) term “cognitive schemas” for managing complex learning tasks. The systematic approaches observed among all 11 participants in this study suggest that analytical sequencing and prior knowledge may be common strategies for managing UML complexity. This implies that structured analytic frameworks should be explicitly taught rather than assumed to emerge naturally. This finding extends recent work by Andersen and Makransky ( 2021 ), who demonstrated similar learner-controlled approaches to cognitive load management in technology-enhanced environments. However, our study suggests that this active agency extends beyond technology-mediated learning to fundamental visual-technical interpretation tasks. This raises questions for future research about the potential transferability of these strategies across technical learning domains. 5.2 Addressing gaps in computing education research Recent systematic reviews reveal that computing education research has ‘ only superficially engaged with CLT’s theoretical developments’ and that ‘ hypotheses phrased in terms of cognitive load components are rare’ (Duran et al., 2022 ). This study contributes to addressing these gaps by examining intrinsic, extraneous, and germane load components in visual-technical learning contexts. Evidence-based UML pedagogy Identifying specific extraneous load sources offers preliminary insights that may inform CLT theory and UML class diagram pedagogical practice. The widespread occurrence of symbolic misinterpretations illustrates how unclear notation systems create extraneous load by forcing students to allocate cognitive resources to symbol decoding rather than conceptual understanding. This aligns with Chandler and Sweller’s ( 1991 ) research on split-attention effects while extending it to visual-technical domains. Participants’ sophisticated UML class diagram orientation and redrawing responses demonstrate learner-generated strategies for reducing extraneous load through environmental modification. This supports Sweller’s ( 2005 ) work on a learner-controlled approach, while revealing specific mechanisms by which students adapt to poor visual design in technical contexts. This study addresses the multidimensional cognitive load gap identified by Andersen and Makransky ( 2021 ) in computing education contexts. While their work focused on technology-enhanced learning environments, our findings demonstrate that the multidimensional conceptualisation of cognitive load is equally critical for understanding traditional visual-technical learning, such as interpreting UML class diagrams. 5.3 Practical implications for UML class diagram education The cognitive strategies ( analytical sequencing and prior knowledge, symbolic confusion and clarification, diagram orientation and redrawing, and cognitive load management through decomposition ) were found in this study. Analytical sequencing and prior knowledge instruction Given the natural evolution of sequential analysis strategies, UML class diagram training should include explicit instruction in systematic diagram interpretation techniques. Rather than relying on students to generate these spontaneously, educational materials should include formal frameworks that leverage students’ intuitive decomposition tendencies, while also providing instruction on all UML class diagram elements. Symbolic confusion and clarification support The observed widespread symbolic confusion emphasises the critical need for focused UML class diagram notation education that extends beyond simple symbol recognition and includes fluency-building activities. Students require extensive practice with symbol recognition in varied contexts to automate notation processing and free cognitive resources for higher-order analysis. Diagram orientation and redrawing support The clear distinction between intrinsic and extraneous load sources in UML class diagram interpretation suggests a potential framework for evaluating instructional materials. Curriculum designers should thoroughly review visual presentations associated with diagrams (e.g., UML class diagrams) to avoid excessive cognitive demands while retaining the fundamental complexity required for deep understanding. Cognitive load management through decomposition training The observed complex decomposition processes suggest that students would benefit from targeted instruction in cognitive load management procedures. Based on these observations, educators might consider teaching students to recognise potential signs of cognitive overload and explore strategies for managing complexity in UML diagrams. The decomposition strategies observed in this study suggest potential approaches that could be explored in instructional design for diverse student populations. 6. Conclusion This study addressed four key questions about cognitive load management in UML class diagram interpretation: what cognitive strategies students use, how they manage mental overload, what creates unnecessary mental burden, and how strategies evolve with complexity. Through think-aloud protocols with 11 first-year students across three activities of increasing complexity, we found that the participants in our exploratory study actively managed their mental load through identifiable cognitive strategies, rather than merely decoding symbols. The analysis revealed four systematic strategies that participants utilised spontaneously: analytical sequencing and prior knowledge, symbolic confusion and clarification, diagram orientation and redrawing, and cognitive load management through decomposition . These observations suggest that our participants may have developed metacognitive awareness of the demands of visual-technical learning. Participants showed agency in reducing extraneous load sources, suggesting that novice learners may be more active than previously assumed. This study provides preliminary evidence for potential applications of CLT to visual-technical domains, addressing a gap in cognitive load analysis noted in recent computing education research (Berssanette & De Francisco, 2022 ; Duran et al., 2022 ; Skulmowski & Xu, 2022 ). Our findings suggest potential behavioural indicators of intrinsic, extraneous, and germane load in the context of UML class diagram interpretation. This exploratory work contributes to extending CLT applications beyond traditional text-based contexts to visual-technical learning environments, though broader validation is needed. The study identifies how students in this sample handled symbolic confusion (extraneous load), leveraged prior knowledge schemas (intrinsic load management), and employed decomposition strategies (germane load investment), providing initial insights into how CLT constructs may manifest in visual-technical learning contexts. Practically, the study offers guidelines for UML class diagram pedagogy that combine content delivery with cognitive process support. Two main sources of extraneous load were observed: symbolic confusion resulting from unclear notation and spatial disorientation caused by unconventional layouts . Participants compensated through self-correction and diagram redrawing, illustrating both resilience and opportunities for instructional support. Pedagogical implications include the explicit teaching of systematic analysis strategies, the development of symbol fluency, consistent visual design standards, and the training of metacognitive strategies. This research indicates how novice learners in this South African university context can act as strategic participants in their own cognitive development. The findings indicate that participants developed cognitive load management strategies during the tasks, indicating potential for enhancement through targeted pedagogical interventions. By integrating content mastery with cognitive process support, cognitive load-aware pedagogy may inform technical education and support diverse student populations. This approach offers a foundation for developing instructional designs that better align with human cognitive processes in visual-technical learning contexts. 7 Limitations and future research directions Several limitations should be considered when interpreting these results. The cross-sectional approach captures student strategies at a single point in time, hence restricting the ability to draw long-term conclusions about how cognitive load management evolves. A sample of 11 first-year students from a single South African university may not reflect varied student populations or educational contexts. This exploratory study provides preliminary insights that require replication with larger, more diverse samples before broader theoretical claims can be made. Future research should investigate whether the multidimensional cognitive load patterns identified in this study and in technology-enhanced environments (Andersen & Makransky, 2021 ) represent universal cognitive adaptations to complex learning environments. It should also explore whether these patterns are specific to visual-technical and technology-mediated contexts. Future research should also investigate instructional methods that systematically teach the identified cognitive processes and assess their impact on learning outcomes across various technical areas. Longitudinal studies that analyse how cognitive load management methods evolve during degree programs could provide insights into the evolution of expertise. Furthermore, comparative studies in other cultural contexts and technical visual domains could shed light on the universality of these cognitive adjustments. Understanding and supporting these cognitive processes, rather than focusing merely on information delivery, offers a viable approach for improving learning effectiveness in increasingly complex technological areas by integrating instructional design with human cognitive architecture. Declarations Statements and Declarations: Funding: The authors did not receive support from any organisation for the submitted work. Competing Interests: The authors have no relevant financial or non-financial interests to disclose. Data Availability: The anonymised interview transcripts and student artifacts collected during this study are available from the corresponding author upon reasonable request. Ethics approval and consent to participate: Ethical Approval ─ All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the General/Human Research Ethics Committee (GHREC) of the Faculty of Natural and Agricultural Science, University of the Free State approved this research study (Reference: UFS-HSD2023/1768). Consent to participate ─ All participants were informed about the purpose of the study and their role in it. Participation was voluntary, and all participants provided written informed consent prior to taking part in the study. Consent to publish ─ Participants were informed that anonymised data may be published as part of the research findings, and all participants provided consent for publication. Author Contributions: Pakiso J. Khomokhoana: Conceptualisation, Methodology, Data collection, Formal analysis, Writing - original draft, Writing - review & editing. Rouxan C. Fouché: Conceptualisation, Investigation, Formal analysis, Writing - review & editing. All authors read and approved the final manuscript. References Andersen MS, Makransky G. The validation and further development of the multidimensional cognitive load scale for physical and online Lectures (MCLS-POL). Front Psychol. 2021;12(1). https://doi.org/10.3389/fpsyg.2021.642084 . Apostol DC, Bogdan R, Marcu M. (2024). UML Diagrams in Teaching Software Engineering Classes. A Case Study in Computer Science Class. 2024 IEEE 22nd World Symposium on Applied Machine Intelligence and Informatics (SAMI 2024) , 327–332. https://doi.org/10.1109/SAMI60510.2024.10432905 Bera P. Analyzing the Cognitive Difficulties for Developing and Using UML Class Diagrams for Domain Understanding. J Database Manage (JDM). 2012;23(3):1–29. https://doi.org/10.4018/jdm.2012070101 . Berssanette JH, De Francisco AC. Cognitive load theory in the context of teaching and learning computer programming: A systematic literature review. IEEE Trans Educ. 2022;65(3):440–9. https://doi.org/10.1109/TE.2021.3127215 . Boustedt J. Students’ different understandings of class diagrams. Comput Sci Educ. 2012;22(1):29–62. https://doi.org/10.1080/08993408.2012.665210 . Castro-Alonso JC, de Koning BB, Fiorella L, Paas F. (2021). Five strategies for optimizing instructional materials: Instructor- and learner-managed cognitive load. In Educational Psychology Review (Vol. 33, Issue 4, pp. 1379–1407). Springer. https://doi.org/10.1007/s10648-021-09606-9 Chandler P, Sweller J. Cognitive load theory and the format of instruction. Cognition Instruction. 1991;8(4):293–332. Cheng P, Lowe R, Scaife M. Cognitive science approaches to understanding diagrammatic representations. Artif Intell Rev. 2001;15(1):79–94. Creswell JW, Creswell JD. Qualitative, quantitative, and mixed methods approaches. 5th ed. SAGE Publications Inc; 2017. https://us.sagepub.com/en-us/nam/research-design/book255675 . Duran R, Zavgorodniaia A, Sorva J. Cognitive load theory in computing education research: A review. ACM Trans Comput Educ (TOCE). 2022;22(4):1–27. https://doi.org/10.1145/3483843 . Ericsson KA, Kintsch W. Long-term working memory. Psychol Rev. 1995;102(2):211–45. https://doi.org/10.1037/0033-295X.102.2.211 . Fu C, Yang D, Zhang X, Hu H. An approach to translating OCL invariants into OWL 2 DL axioms for checking inconsistency. Automated Softw Eng. 2017;24(2):295–339. https://doi.org/10.1007/s10515-017-0210-9 . Hafeez A, Ahmed M, Furqan M, Rehaman WU, Husain I. Importance and Impact of Class Diagram in Software. Indian J Sci Technol. 2019;12(25):1–7. Kalyuga S, Plass JL. Criticisms and Desired Additional Modifications. Rethinking Cognitive Load Theory. Oxford University Press; 2025. pp. 36–46. https://doi.org/10.1093/9780190078539.003.0004 . Khomokhoana PJ, Fouché RC, Nkalai TS. (2025). Semiotic analysis of UML class diagrams in pedagogy: A case of first-year BCIS students. Discov Educ 4, 342 (2025). https://doi.org/10.1007/s44217-025-00817-8 Khomokhoana PJ, Nkalai TS. (2024). Cognitive processes used by students in understanding aggregation and composition in UML: A case of first-year BCIS students. Proceedings of the 53rd Annual Conference of the Southern African Computer Lecturers’ Association (SACLA 2024) , 116–131. https://www.researchgate.net/publication/386451886 Kirschner PA, Sweller J, Kirschner F, Zambrano JR. From Cognitive Load Theory to Collaborative Cognitive Load Theory. Int J Computer-Supported Collaborative Learn. 2018;13(2):213–33. https://doi.org/10.1007/s11412-018-9277-y . Koç H, Erdoğan AM, Barjakly Y, Peker S. (2021). UML Diagrams in Software Engineering Research: A Systematic Literature Review. The 7th International Management Information Systems Conference , 13. https://doi.org/10.3390/proceedings2021074013 Ma T. Reasoning of novice and experienced software designers on creating UML class diagrams. University of Gothenburg; 2013. Masri K. (2009). Conceptual model design for better understanding . http://ir.lib.sfu.ca/handle/1892/112 Mayer RE. The Cambridge handbook of multimedia learning. Cambridge University Press; 2014. Mishra SB, Alok S. Handbook of Research Methodology: A Compendium for Scholars & Researchers. Educreation Publishing; 2017. Mohammadi Zenouzagh Z, Admiraal W, Saab N. Empowering student engagement: the dynamics of learner traits in digital feedback environments. J Comput High Educ. 2025. https://doi.org/10.1007/s12528-025-09459-z . Möller M, Winter M, Reichert M. Cognitive Factors in Process Model Comprehension—A Systematic Literature Review. Brain Sci. 2025;15(5):505. https://doi.org/10.3390/brainsci15050505 . Nugroho A. Level of detail in UML models and its impact on model comprehension: A controlled experiment. ‎Inf Softw Technol. 2009;51(12):1670–85. https://doi.org/10.1016/j.infsof.2009.04.007 . Osztián P, Kátai Z, Osztián E. On the computational thinking and diagrammatic reasoning of first-year computer science and engineering students. Front Educ. 2022;7(933316). https://doi.org/10.3389/feduc.2022.933316 . Paas FGWC, Van Merrienboer JJG. Variability of worked examples and transfer of geometrical problem-solving skills: A cognitive-load approach. J Educ Psychol. 1994;86(1):122–33. Paas F, Sweller J. Implications of cognitive load theory for multimedia learning. In: Mayer RE, editor. The Cambridge handbook of multimedia learning. Cambridge University Press; 2014. pp. 27–42. Paas F, Tuovinen JE, Tabbers H, Van Gerven PWM. (2003). Cognitive Load Measurement as a Means to Advance Cognitive Load Theory . https://doi.org/10.1207/S15326985EP3801_8 Paivio A. Mental representations: A dual coding approach. Oxford University Press; 1986. Parcell ES, Rafferty KA. Interviews, recording and transcribing. In: Allen M, editor. The SAGE Encyclopedia of Communication Research Methods. Sage Publications, Inc.; 2017. pp. 800–3. Plowright D. Using mixed methods: Frameworks for an integrated methodology. SAGE Publications, Inc; 2011. https://doi.org/10.4135/9781526485090 . Saunders M, Lewis P, Thornhill A. (2015). Research Methods for Business Students (7th ed.). Pearson Education Limited. https://www.pearson.com/store/p/research-methods-for-business-students/P100001214517 Shen Z, Tan S, Siau K. Challenges in learning unified modeling language: From the perspective of diagrammatic representation and reasoning. Commun Association Inform Syst. 2018;43(1):545–65. https://doi.org/10.17705/1CAIS.04330 . Singh DP, Ahmed N, Gupta N. Business Research Method And Project Work. SBPD; 2021. Skulmowski A, Xu KM. Understanding cognitive load in digital and online learning: A new perspective on extraneous cognitive load. Educational Psychol Rev. 2022;34(1):171–96. https://doi.org/10.1007/s10648-021-09624-7 . Sweller J. Cognitive load during problem solving: Effects on learning. Cogn Sci. 1988;12(2):257–85. https://doi.org/10.1207/s15516709cog1202_4 . Sweller J. Implications of cognitive load theory for multimedia learning. In: Mayer R, editor. The Cambridge handbook of multimedia learning. Cambridge University Press; 2005. pp. 19–30. Sweller J. Cognitive load theory. Psychol Learn Motivation. 2011;55:37–76. Sweller J. Cognitive load theory and educational technology. Education Tech Research Dev. 2020;68(1):1–16. https://doi.org/10.1007/s11423-019-09701-3 . Sweller J, Ayres P, Kalyuga S. Altering element interactivity and intrinsic cognitive load. Cognitive load theory - Explorations in the Learning Sciences, Instructional Systems and Performance Technologies. NY: Springer New York; 2011. pp. 203–18. https://doi.org/10.1007/978-1-4419-8126-4_16 . Tóth P, Pogatsnik M. Advancement of inductive reasoning of engineering students. Hung Educational Res J. 2022;13(1):86–106. https://doi.org/10.1556/063.2022.00120 . Van Merriënboer JJG, Sweller J. Cognitive load theory and complex learning: Recent developments and future directions. Source: Educational Psychol Rev. 2005;17(2):147–77. Van Merriënboer JJG, Sweller J. Cognitive load theory in health professional education: Design principles and strategies. Med Educ. 2010;44(1):85–93. https://doi.org/10.1111/j.1365-2923.2009.03498.x . Young JQ, van Merrienboer JJ, Durning SJ, Cate T, O. Cognitive load theory: Implications for medical education: AMEE guide 86. Med Teach. 2014;36(5):371–84. https://doi.org/10.3109/0142159X.2014.889290 . Zhou Y, Chai CS, Li X, Ma C, Li B, Yu D, Liang JC. Application of Metacognitive Planning Scaffolding for the Cultivation of Computational Thinking. J Educational Comput Res. 2023;61(6):1123–42. https://doi.org/10.1177/07356331231160294 . Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 15 May, 2026 Reviews received at journal 14 May, 2026 Reviews received at journal 08 May, 2026 Reviewers agreed at journal 08 May, 2026 Reviews received at journal 07 May, 2026 Reviewers agreed at journal 06 May, 2026 Reviewers agreed at journal 06 May, 2026 Reviewers invited by journal 06 May, 2026 Editor invited by journal 16 Apr, 2026 Editor assigned by journal 30 Mar, 2026 Submission checks completed at journal 30 Mar, 2026 First submitted to journal 30 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9266264","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":640167384,"identity":"b9699750-2092-4205-858d-c748cc6c4444","order_by":0,"name":"Pakiso J. Khomokhoana","email":"data:image/png;base64,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","orcid":"","institution":"University of the Witwatersrand","correspondingAuthor":true,"prefix":"","firstName":"Pakiso","middleName":"J.","lastName":"Khomokhoana","suffix":""},{"id":640167385,"identity":"fb7b8609-9b2c-481d-90d2-0cb524943b74","order_by":1,"name":"Rouxan C. Fouché","email":"","orcid":"","institution":"University of the Free State","correspondingAuthor":false,"prefix":"","firstName":"Rouxan","middleName":"C.","lastName":"Fouché","suffix":""}],"badges":[],"createdAt":"2026-03-30 11:40:34","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9266264/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9266264/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109337763,"identity":"61fbbeb6-2612-4693-a26e-de83c376606d","added_by":"auto","created_at":"2026-05-15 17:46:51","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":6534,"visible":true,"origin":"","legend":"\u003cp\u003eName withheld?\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-9266264/v1/e9faf714eddf1c1f646007ec.png"},{"id":109337764,"identity":"ca251529-93ca-4f94-adf3-343c674a935c","added_by":"auto","created_at":"2026-05-15 17:46:51","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":54404,"visible":true,"origin":"","legend":"\u003cp\u003eName withheld?\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-9266264/v1/9debacacfee1451f81c21b85.png"},{"id":109405474,"identity":"c2ebe0b5-5bd6-452b-a1fc-5e9ad47b6b39","added_by":"auto","created_at":"2026-05-17 13:18:22","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":51259,"visible":true,"origin":"","legend":"\u003cp\u003eBoardgame class diagram\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-9266264/v1/8053b4a38d3058d164be4b44.png"},{"id":109337765,"identity":"251174a6-94c3-4fc7-8101-4e9666e30883","added_by":"auto","created_at":"2026-05-15 17:46:52","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":159818,"visible":true,"origin":"","legend":"\u003cp\u003eClass diagram for a computer and its components\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-9266264/v1/71b5fb1710696cabba7d4f04.png"}],"financialInterests":"No competing interests reported.","formattedTitle":"Cognitive Strategies in Interpreting UML Class Diagrams with Emphasis on Load, Order and Symbolic Confusion","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eUnified Modeling Language (UML) class diagrams are a cornerstone of software engineering and information systems education (Apostol et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). These diagrams require students to integrate visual, logical, and symbolic reasoning when interpreting classes, relationships, and hierarchical structures (Fu et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Hafeez et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Recent systematic reviews have explicitly noted a lack of in-depth research on cognitive processes in model comprehension, despite numerous studies on diagram quality and tooling (Ko\u0026ccedil; et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; M\u0026ouml;ller et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Recent work has begun to examine these cognitive processes. Our previous studies identified strategies such as attention management, memory activation, and reflective thinking when students tackled UML class diagram related problems (Khomokhoana \u0026amp; Nkalai, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Applying semiotic theory, we also found that visual features strongly shape understanding, while ambiguous terminology and contextual differences often lead to confusion (Khomokhoana et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). However, these studies did not explicitly address how students manage mental effort, a gap that is critical given the cognitive demands of interpreting UML class diagrams.\u003c/p\u003e \u003cp\u003eCognitive Load Theory (CLT) often appears in computing education research, but researchers rarely engage deeply with its recent developments (Duran et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This gap is clear in visual-technical learning, where most CLT studies focus on text-based programming rather than visual modelling languages. The connection between diagrammatic reasoning and software engineering education remains largely unexplored (Ko\u0026ccedil; et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), despite research showing strong links between diagram interpretation skills and programming success (T\u0026oacute;th \u0026amp; Pogatsnik, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Zhou et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). UML class diagrams create unique mental demands. For example, students must simultaneously process visual symbols, spatial relationships, text elements, and abstract concepts, such as inheritance and composition (Bera, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). This multi-modal processing places a heavy demand on working memory, which research shows has severe limits (Sweller et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis study employs the CLT approach to investigate how first-year computing students interpret UML class diagrams, with a focus on their mental load management. Building on our previous work, we ask:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eWhat cognitive strategies do students use when interpreting UML class diagrams?\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eHow do students manage mental overload, and what breakdown strategies do they develop?\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eWhat creates unnecessary mental burden, and how do students adapt?\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eHow do cognitive strategies change as UML class diagrams become more complex?\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eThe remainder of this paper unfolds as follows: Section \u003cspan refid=\"Sec2\" class=\"InternalRef\"\u003e2\u003c/span\u003e introduces the CLT framework as it relates to visual-technical learning. Section \u003cspan refid=\"Sec6\" class=\"InternalRef\"\u003e3\u003c/span\u003e describes the study\u0026rsquo;s methodology, and Section \u003cspan refid=\"Sec10\" class=\"InternalRef\"\u003e4\u003c/span\u003e reports the findings. Section \u003cspan refid=\"Sec15\" class=\"InternalRef\"\u003e5\u003c/span\u003e presents the interpretation of these results considering theoretical and practical implications. Section \u003cspan refid=\"Sec19\" class=\"InternalRef\"\u003e6\u003c/span\u003e summarises the study contributions, and Section \u003cspan refid=\"Sec20\" class=\"InternalRef\"\u003e7\u003c/span\u003e concludes with limitations and avenues for future research.\u003c/p\u003e"},{"header":"2. Theoretical framework","content":"\u003cp\u003eThis study adopts the CLT as its theoretical lens to examine students\u0026rsquo; cognitive strategies when interpreting UML class diagrams. Initially proposed by Sweller (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e1988\u003c/span\u003e) and further developed by Sweller, Ayres, and Kalyuga (\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e), CLT centres on the limitations of working memory and how instructional design can hinder or enhance learning. At its core, CLT recognises that working memory is limited and that learning is most effective when cognitive resources are optimally allocated (Paas et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2003\u003c/span\u003e).\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Cognitive load theory in visual-technical learning\u003c/h2\u003e \u003cp\u003eCLT explains how the limitations of working memory shape learning outcomes (Sweller, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e1988\u003c/span\u003e; Sweller et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). Effective learning occurs when instructional design supports the allocation of cognitive resources between three load types:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eIntrinsic load\u003c/em\u003e stems from the inherent complexity of the learned content or task (Van Merri\u0026euml;nboer \u0026amp; Sweller, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Young et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). In interpreting UML class diagrams, intrinsic load includes understanding object-oriented concepts (e.g., inheritance, composition, aggregation), recognising hierarchical relationships, and comprehending the logical structure of system designs (Boustedt, 2010; \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). This load is unavoidable and directly related to the learning objectives (Sweller et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). Importantly, intrinsic load cannot be reduced without affecting learning objectives; hence, effective management of other load types is required.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eExtraneous load\u003c/em\u003e is generated by suboptimal instructional design, including confusing layouts, unclear symbols, or irrelevant information (Van Merri\u0026euml;nboer \u0026amp; Sweller, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Chandler and Sweller\u0026rsquo;s (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e1991\u003c/span\u003e) research on the split-attention effect reveals how poorly designed visual materials can overwhelm working memory by requiring students to mentally integrate different information sources. Extraneous load in UML diagrams\u0026rsquo; settings can occur when inheritance arrows point horizontally rather than vertically, novel language is used without explanation, or diagram elements are placed unexpectedly (Masri, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eGermane load\u003c/em\u003e reflects the mental effort devoted to meaningful learning processes, such as organising knowledge and integrating new information with existing understanding (also known as cognitive schemas) (Van Merri\u0026euml;nboer \u0026amp; Sweller, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). According to Paas and Van Merri\u0026euml;nboer (1994), students who actively engage in strategies such as chunking information, drawing connections, or systematically analysing complex diagrams devote more cognitive resources to building deep, organised understanding rather than to surface-level processing. This constructive utilisation of cognitive resources results in deeper comprehension and improved knowledge transfer.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eIn the interpretation of UML class diagrams, intrinsic load arises from simultaneously processing visual symbols, spatial relationships, text elements, and abstract object-oriented principles (Bera, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Poor UML class diagram design or ambiguous notation increases extraneous load (Chandler \u0026amp; Sweller, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e1991\u003c/span\u003e; Masri, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), whereas strategies such as decomposition, the use of analogies, and activating prior knowledge reflect an investment in germane cognitive load (Paas \u0026amp; Van Merrienboer, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e1994\u003c/span\u003e). Recent developments in CLT emphasise its multidimensional nature, particularly in extraneous load (Andersen \u0026amp; Makransky, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). These developments also call for greater integration with multimedia learning (Mayer, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), collaborative load theory (Kirschner et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), and technology-enhanced education (Paas \u0026amp; Sweller, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). However, applications in visual-technical domains such as UML remain limited, despite evidence that spatial reasoning and diagrammatic skills predict programming performance (Oszti\u0026aacute;n et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIt should be noted that CLT has faced criticism regarding the difficulty of empirically measuring different load types separately and the potential for circular reasoning when inferring cognitive load from performance outcomes (Kalyuga \u0026amp; Plass, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Despite these measurement challenges, CLT\u0026rsquo;s conceptual framework offers valuable theoretical guidance for understanding the cognitive demands of interpreting UML class diagrams.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Diagrammatic reasoning in UML interpretation\u003c/h2\u003e \u003cp\u003eDual Coding Theory (Paivio, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e1986\u003c/span\u003e) suggests that visual and verbal information are processed in interconnected but distinct systems. UML class diagrams require both systems: visual processing to recognise shapes and relationships, and verbal processing to interpret labels and abstract concepts (Shen et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Effective diagrammatic reasoning depends on recognising symbolic conventions, managing spatial complexity, and integrating these with conceptual knowledge (Cheng et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2001\u003c/span\u003e). Students with strong diagrammatic reasoning skills tend to perform better in tasks that involve logical problem-solving and structured thinking (e.g., programming) (T\u0026oacute;th \u0026amp; Pogatsnik, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). However, when confronted with unconventional layouts or unclear symbols, novices often experience disorientation, triggering compensatory strategies such as redrawing diagrams or reordering elements. These behaviours reflect active load management rather than passive reception.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 CLT and student cognitive strategies\u003c/h2\u003e \u003cp\u003eRecent reviews have revealed that the cognitive and educational dimensions of diagrammatic reasoning in software engineering remain underexplored. Empirical and review studies tend to emphasise syntactic and technical issues rather than students\u0026rsquo; reasoning processes (Ko\u0026ccedil; et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; M\u0026ouml;ller et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Castro-Alonso et al. (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) and Mohammadi Zenouzagh et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) suggest that students are not passive recipients, but can develop adaptive strategies when facing cognitive challenges. However, limited research has systematically examined the specific strategies students spontaneously employ when interpreting complex visual-technical materials such as UML class diagrams (Ko\u0026ccedil; et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Ma, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Nugroho, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). This study investigates how students naturally respond to cognitive load challenges by examining their think-aloud protocols for evidence of:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eSystematic approaches to managing intrinsic load complexity.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eSelf-correction and environmental modification to reduce extraneous load.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eActive engagement strategies that may reflect germane load investment.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eUnderstanding these student-generated strategies provides valuable insight into how novice learners manage cognitive load in visual-technical learning contexts. Such insights can deepen our understanding of CLT by illustrating how learners naturally adapt under varying load conditions. They can also inform instructional design aimed at optimising cognitive load management.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Research design and methodology","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Research design\u003c/h2\u003e \u003cp\u003eThis study used a narrative inquiry approach, following Plowright\u0026rsquo;s (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) integrated-methods research framework (FraIM). FraIM enables research perspectives to emerge naturally during the study, rather than requiring fixed theoretical positions from the outset. This flexibility proved valuable for exploring cognitive load management strategies, as it allowed for the documentation of students\u0026rsquo; natural problem-solving behaviours without imposing rigid constraints. Data collection involved multiple sources: semi-structured interviews, student-created artefacts, and systematic behavioural observations. The research focused on first-year Bachelor of Computer Information Systems (BCIS) students at a South African university. The study employed both purposeful (Singh et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) and convenience sampling (Mishra \u0026amp; Alok, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Saunders et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) methods. Purposeful selection ensured participants were studying relevant course content (including aggregation and composition concepts). Convenient factors facilitated the researcher\u0026rsquo;s access through existing teaching relationships with one research team member. From a group of 105 students, 20 volunteers were selected through an open recruitment process. Eleven students completed the full research protocol. These included five males and six females, aged above 18 years, and with varying levels of programming experience. Institutional ethical approval was secured before data collection (Reference: UFS-HSD2023/1768).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Data collection procedures\u003c/h2\u003e \u003cp\u003e Individual semi-structured interviews were conducted using the think-aloud protocol in which participants completed three activities of increasing complexity. These activities followed three levels of increasing complexity, each designed to reveal different cognitive processing patterns: basic understanding (Phase 1), intermediate analysis (Phase 2), and advanced synthesis (Phase 3).\u003c/p\u003e \u003cp\u003eParticipants completed activities using pen and paper, with their written work serving as artefacts for analysis. Think-aloud protocols captured real-time cognitive processing, providing access to participants\u0026rsquo; reasoning strategies. All interview sessions were audio-recorded with the explicit consent of the participants, ensuring complete documentation of their verbal expressions and thought processes. Participant behaviours, emotional responses, and interaction patterns were documented through structured field notes. This contributed to data triangulation and improved interpretive validity.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003ePhase 1 \u0026mdash; Initial Assessment\u003c/strong\u003e \u003cp\u003eParticipants examined three unlabeled UML class diagrams (Figs.\u0026nbsp;1, 2 and 3), identifying and describing what they observed based on their existing knowledge. This phase assessed baseline familiarity with association, aggregation, and composition relationships without external guidance. The names of the diagrams were intentionally omitted (e.g., \u003cem\u003ename withheld\u003c/em\u003e) to require participants to identify them independently.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Taba\" border=\"1\"\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003eFigure\u0026nbsp;1.\u003c/b\u003e Name withheld?\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003eFigure\u0026nbsp;2.\u003c/b\u003e Name withheld?\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003eFigure\u0026nbsp;3.\u003c/b\u003e Name withheld?\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003ePhase 2 \u0026mdash; Relational Analysis\u003c/strong\u003e \u003cp\u003eParticipants analysed and explained the relationships shown in a moderately complex boardgame class diagram (see Figs.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e4\u003c/span\u003e and 5). This activity required combining multiple concepts and relationship types.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003ePhase 3 \u0026mdash; Construction and Interpretation\u003c/strong\u003e \u003cp\u003eThis phase included two activities. First, participants created original class diagrams based on detailed written specifications covering computer hardware relationships, inheritance hierarchies, and compositional structures. Second, they provided detailed explanations of a pre-constructed class diagram (Fig.\u0026nbsp;5) representing similar concepts to their own creations.\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Data analysis\u003c/h2\u003e \u003cp\u003eThe audio recordings were transcribed following established narrative research protocols (Creswell \u0026amp; Creswell, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Initial data preparation addressed transcription inconsistencies and response clarity challenges common in open-ended questioning. A fuzzy validation approach (Parcell \u0026amp; Rafferty, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) was utilised to preserve authentic participant voices while ensuring analytical clarity and consistency. Analysis began with intensive data immersion through repeated audio review and comprehensive transcript examination. This familiarisation process informed the development of a structured coding system aligned with our research objectives. The 11 validated transcripts were processed using NVivo software for systematic thematic analysis. Codes were developed to reflect emerging cognitive strategies identified in participant responses, paying careful attention to analytical detail and thematic saturation. Text segments were systematically annotated, assisting in identifying connections between codes and supporting evidence. Code names were refined throughout the analysis to ensure accurate thematic representation, with a particular focus on the frequency patterns and conceptual significance of recurring elements.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Findings","content":"\u003cp\u003eThink-aloud protocols were analysed for 11 participants who completed the three activities of increasing complexity. The findings revealed four systematic cognitive strategies that align with recent advances in multidimensional cognitive load theory (Andersen \u0026amp; Makransky, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). These strategies show how participants naturally adapted to the unique cognitive challenges of visual-technical learning environments. Each strategy was associated with specific types of CLT load (intrinsic, extraneous, and germane) and linked to identifiable student actions. These strategies were observed across multiple participants and levels of diagram complexity. To provide an overview of these relationships, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e7\u003c/span\u003e presents a concept map illustrating how each strategy is connected to CLT load types and the observed student actions. This visual represents the interplay between intrinsic complexity, extraneous distractions, and germane load investment as participants navigated tasks involving the interpretation of UML class diagrams.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e summarises these strategies, the associated CLT load types, and representative student actions. This structured synthesis highlights the range of approaches participants used to manage cognitive demands, from sequencing their analysis and activating prior knowledge structures to correcting symbol misunderstandings, redrawing diagrams, and decomposing complex diagrams into smaller, more manageable units.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCognitive strategies in UML class diagram interpretation with associated CLT load types and student actions\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStrategy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCLT Load Type(s)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eObserved Student Actions\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAnalytical sequencing and prior knowledge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIntrinsic, Germane\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSequential analysis, identifying main classes first, perceptual salience and using prior knowledge schemas or mental networks of prior knowledge.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSymbolic confusion and clarification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExtraneous, Germane\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCorrecting symbol misinterpretations, and practising symbol fluency.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiagram orientation and redrawing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExtraneous, Germane\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRearranging layout, and redrawing diagrams for clarity.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDecomposition for load management\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIntrinsic, Germane\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eBreaking diagrams into smaller parts, and sequential processing of components.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe following subsections expand on each strategy, illustrating how they manifested in student reasoning and linking them explicitly to CLT load types. The numbers indicate how many of the 11 participants demonstrated each strategy, emphasising patterns of behaviour in this narrative study. The following findings represent patterns observed in our small sample of 11 participants from a single institution. While these patterns suggest potential cognitive strategies, they should be interpreted as preliminary insights requiring replication and validation before broader theoretical claims can be made.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Analytical sequencing and prior knowledge\u003c/h2\u003e \u003cp\u003e \u003cem\u003eSequential analysis and identifying main classes first\u003c/em\u003e: Participants exhibited clear patterns in their approach to UML class diagrams. All 11 participants began by identifying the main classes first before moving to subclasses and attributes. This sequential analysis represents a strategy for organizing information into chunks, a behavior consistent with what CLT describes as germane load investment (Sweller et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). P1 explained this deliberate approach: \u0026ldquo;\u003cem\u003eThe first thing I would do is analyse the diagram\u003c/em\u003e\u0026rdquo;, showing how participants planned their analysis to avoid mental overload. P7 demonstrated this pattern, starting with the central element: \u0026ldquo;\u003cem\u003eComputer is the main class... there is a mouse class, a keyboard class...\u003c/em\u003e\u0026rdquo;. P10 further showed this sequential analysis: \u0026ldquo;\u003cem\u003eI check the main class, then the subclasses, then analyse the components\u003c/em\u003e\u0026rdquo;. This step-by-step method breaks complex tasks into smaller parts that fit within working memory limits (Paas et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2003\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cem\u003ePerceptual salience\u003c/em\u003e: Before beginning their systematic analysis, five participants explicitly mentioned that their attention was drawn to visually prominent features, a response to the principle of perceptual salience. For example, P9 noted: \u0026ldquo;\u003cem\u003eYes, it\u0026rsquo;s [the diagram] bigger. It\u0026rsquo;s the biggest thing on the page. Alright, so it caught my attention\u003c/em\u003e\u0026rdquo;. Similarly, P11 observed: \u0026ldquo;\u003cem\u003eIf there was maybe a bold word, I\u0026rsquo;m sure I would first look at that bold word without even looking at something else, it catches the eye\u003c/em\u003e\u0026rdquo;. This automatic focus on prominent visual elements aligns with CLT principles about effective visual design (Van Merri\u0026euml;nboer \u0026amp; Sweller, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2005\u003c/span\u003e), reducing unnecessary mental effort (Chandler \u0026amp; Sweller, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e1991\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cem\u003eUsing prior knowledge schemas\u003c/em\u003e: Eight participants actively employed prior knowledge schemas to make sense of new diagrams, hence reducing intrinsic load. In this instance, P6 showed this schema activation when she connected the UML class diagram elements to familiar concepts: \u0026ldquo;\u003cem\u003eWhen I look at these arrows, they remind me of a certain diagram which we did in class... I think it is a component of a vehicle. I see the oil filter, oil checking...\u003c/em\u003e\u0026rdquo;. By linking new information to familiar ideas, participants attempted to manage the inherent complexity of UML class diagrams.\u003c/p\u003e \u003cp\u003eThese four actions (\u003cem\u003eidentifying main classes first, using prior knowledge schemas, sequential analysis, and responding to perceptual salience\u003c/em\u003e) show how participants naturally developed strategies to manage mental load. As shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e7\u003c/span\u003e, the \u0026ldquo;\u003cem\u003eanalytical sequencing and prior knowledge\u003c/em\u003e\u0026rdquo; approach helped participants handle both intrinsic and germane load.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Symbolic confusion and clarification\u003c/h2\u003e \u003cp\u003e \u003cem\u003eCorrecting symbol misinterpretations\u003c/em\u003e: Ten of the 11 participants experienced misinterpretations of UML class diagram symbols, which created an extraneous load and required extra mental resources for error correction. The first key action, correcting symbol misinterpretations, appeared in various forms. For instance, P3 initially referred to an inheritance arrow as a \u0026ldquo;\u003cem\u003emethod\u003c/em\u003e\u0026rdquo;, but through self-correction stated: \u0026ldquo;\u003cem\u003eChess is a type of board game\u003c/em\u003e\u0026rdquo;. This self-correction process is a core part of correcting symbol misinterpretations, demonstrating how participants actively monitored their understanding and adjusted their mental resources when they identified errors. Participants also engaged in correcting symbol misinterpretations by reframing confusing visual elements in familiar terms, such as relating them to vehicle parts or personal experiences. As an example, P1 said: \u0026ldquo;\u003cem\u003eI am not familiar with this particular game, but then I know for sure that a board game has particular elements like chess. That is why I was able to see that this one is inheritance\u003c/em\u003e\u0026rdquo;. While this indicates adaptive thinking, it represents the mental effort required to correct initial misunderstandings, adding extraneous load until the correct interpretation is found.\u003c/p\u003e \u003cp\u003e\u003cem\u003ePractising symbol fluency\u003c/em\u003e: The second key action emerged as nine participants developed from basic visual recognition to a deeper understanding. P3 particularly demonstrated this progression, starting with visual symbol processing: \u0026ldquo;\u003cem\u003eBy the look of symbols, this one here is not coloured\u003c/em\u003e\u0026rdquo;. After learning proper terminology, she could distinguish between relationship types: \u0026ldquo;\u003cem\u003eThe one that is not shaded shows that it is an aggregation. The one in tile is a composition since it is shaded\u003c/em\u003e\u0026rdquo;. This transition from surface-level identification to conceptual understanding represents practising symbol fluency in action. It indicates germane load investment, productive mental effort that builds from simple visual recognition to semantic processing. When participants could not effectively use either correcting symbol misinterpretations or practising symbol fluency, they experienced significant extraneous load (Sweller, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). P3\u0026rsquo;s initial confusion with inheritance notation forced repeated attempts at correcting symbol misinterpretations, diverting mental resources from learning objectives to error correction. Participants varied widely in their mastery of both actions. Those who had developed strong symbol fluency could focus on understanding system relationships, investing in germane load. However, participants who were still developing this fluency spent more time correcting symbol misinterpretations, using working memory for basic decoding rather than engaging in deeper analysis. This variation appeared to influence how participants engaged with complex diagrams.\u003c/p\u003e \u003cp\u003eAs shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e7\u003c/span\u003e, the \u0026ldquo;\u003cem\u003esymbolic confusion and clarification\u003c/em\u003e\u0026rdquo; strategy involves these two specific actions: correcting symbol misinterpretations (addressing extraneous load) and practising symbol fluency (investing in germane load). Together, these actions reveal how participants actively worked to reduce confusion while building their understanding of UML class diagram notation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Diagram orientation and redrawing\u003c/h2\u003e \u003cp\u003e \u003cem\u003eRearranging layout and redrawing for clarity\u003c/em\u003e: Three participants experienced unnecessary extraneous load from poor visual design when inheritance symbols appeared horizontally rather than vertically on the UML class diagram. P2 specifically addressed this challenge through two key actions: rearranging the layout and redrawing UML class diagrams for clarity. She actively restructured the UML class diagram to match her mental model, explaining: \u0026ldquo;\u003cem\u003eYes [it\u0026rsquo;s easier] when it\u0026rsquo;s right below the main class\u003c/em\u003e\u0026rdquo;. This redrawing of UML class diagrams for clarity represents load management, thereby reducing extraneous load while investing in germane load through spatial restructuring.\u003c/p\u003e \u003cp\u003eSeven participants showed how visual conventions affect cognitive load. For example, P2 noted: \u0026ldquo;\u003cem\u003eThe diamond always faces to the main class\u003c/em\u003e\u0026rdquo;, while P3 emphasised: \u0026ldquo;\u003cem\u003eIt is important to know where should the arrows face\u003c/em\u003e\u0026rdquo;. Participants appeared to form expectations about UML diagram layouts, which may suggest developing mental models for visual processing (Sweller, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). When these expectations are violated, participants must mentally or physically rearrange the layout, creating an additional cognitive burden. Inherently, design inconsistencies forced participants to unnecessarily rearrange layout activities. For example, P2\u0026rsquo;s confusion illustrates this: \u0026ldquo;\u003cem\u003eIt was tricky for me to identify it because of the way it was arranged [drawn]. Is it an inheritance or what?\u003c/em\u003e\u0026rdquo;. She expected to see vertical inheritance arrows on the UML class diagram, but encountered horizontal ones. The mental effort required to reconcile this mismatch represents extraneous load that could be avoided through consistent visual design.\u003c/p\u003e \u003cp\u003eParticipants who redrew UML class diagrams reported that they found this helpful. When asked if vertical arrangement was easier, P2 simply replied: \u0026ldquo;\u003cem\u003eYes\u0026rdquo;\u003c/em\u003e. By physically restructuring the UML class diagram to align with her mental model of hierarchical relationships, she reduced the cognitive effort needed for interpretation. This spatial reorganisation indicates how rearranging layout and redrawing UML class diagrams for clarity work together to manage both extraneous and germane load.\u003c/p\u003e \u003cp\u003eAs shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Fig.\u0026nbsp;1, the \u0026ldquo;\u003cem\u003ediagram orientation and redrawing\u003c/em\u003e\u0026rdquo; strategy involves two specific actions: rearranging the layout and redrawing diagrams for clarity. These actions helped participants transform confusing visual UML class diagram presentations into formats that support understanding, hence managing both extraneous load (from poor design) and germane load (through active restructuring).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Decomposition for load management\u003c/h2\u003e \u003cp\u003e \u003cem\u003eBreaking diagrams into parts\u003c/em\u003e: Four participants demonstrated a clear awareness of their cognitive limits and effectively used breaking diagrams into smaller parts to manage complexity. As an example, P5 warned: \u0026ldquo;\u003cem\u003eIf they [students] don\u0026rsquo;t break it down, there is a high likelihood that they will get it wrong\u003c/em\u003e\u0026rdquo;. This metacognitive awareness, knowing when mental capacity is exceeded, drives the use of decomposition strategies to manage intrinsic load.\u003c/p\u003e \u003cp\u003e \u003cem\u003eSequential analysis\u003c/em\u003e: The second key action, sequential analysis, was consistently employed across the four participants who used decomposition strategies. For instance, P7 advised: \u0026ldquo;\u003cem\u003eTry to break the UML class diagram into smaller and more understandable pieces\u003c/em\u003e\u0026rdquo;. This deliberate decomposition technique aligns with Van Merri\u0026euml;nboer and Sweller\u0026rsquo;s (\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2005\u003c/span\u003e) research on managing complexity in technical learning. By processing elements one at a time rather than simultaneously, participants can maintain mental effort within the limits of working memory. This insight suggests that the participants understand that trying to grasp complex diagrams all at once exceeds working memory capacity. Instead, sequential processing of components keeps cognitive demands manageable. Participants recognised that excessive visual complexity impairs understanding. Breaking UML class diagrams into smaller parts transforms overwhelming visual information into manageable chunks. This systematic division enables the sequential processing of components, handling one element at a time rather than all elements simultaneously. This approach divides high intrinsic load tasks into subtasks that fit within working memory constraints. The consistency of these decomposition strategies across participants suggests that breaking UML class diagrams into smaller parts and sequential processing of components are natural responses to complex visual-technical information. These findings highlight the importance of explicitly teaching both actions as part of UML class diagram education.\u003c/p\u003e \u003cp\u003eAs mapped in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e7\u003c/span\u003e, \u0026ldquo;\u003cem\u003edecomposition for load management\u003c/em\u003e\u0026rdquo; addresses intrinsic load through germane load investment. The two actions (\u003cem\u003ebreaking diagrams into smaller parts and sequential processing of components\u003c/em\u003e) show how participants actively manage the inherent complexity of UML diagrams through strategic decomposition.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Discussion","content":"\u003cp\u003eThe findings suggest that UML class diagrams carry an intrinsic load due to their structural complexity. However, many student difficulties were caused by extraneous load, specifically inadequate symbol clarity and ambiguous orientations. When participants were able to reduce extraneous factors (for example, by redrawing or utilising analogies) and use deconstruction, they improved their germane load, focusing more on comprehending rather than decoding. This pattern appears consistent with CLT concepts regarding how learners might manage cognitive demands during the interpretation of technical UML class diagrams.\u003c/p\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Theoretical implications for cognitive load theory\u003c/h2\u003e \u003cp\u003eThe systematic analysis of student think-aloud protocols reveals sophisticated cognitive load management strategies, providing insight into how novice learners process complex visual-technical information. These findings suggest potential implications for CLT theory, addressing deficiencies noted in previous systematic evaluations of computing education research.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003ePotential implications for visual-technical learning domains\u003c/strong\u003e \u003cp\u003eThe four cognitive strategies (\u003cem\u003eanalytical sequencing and prior knowledge, symbolic confusion and clarification, diagram orientation and redrawing, and cognitive load management through decomposition\u003c/em\u003e) described in this study suggest considerations for CLT\u0026rsquo;s traditional focus. Participants in this study employed specific cognitive strategies for processing UML class diagrams that warrant further investigation, raising questions about whether technical visual domains may involve distinct cognitive processes. Observed behavioural patterns suggest that students\u0026rsquo; engagement with visual and technical aspects of UML diagrams could be leveraged to tailor instructional support. Cognitive Load Theory suggests a potential framework for such adaptations, recommending strategies such as visual cueing, worked examples, and progressive guidance to optimise learning (Sweller et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). The extensive use of spontaneous cognitive load management through decomposition, diagram orientation, and redrawing approaches demonstrated that participants actively identified and minimised extraneous load sources rather than being passive recipients of instructional design. This study builds on recent research by Andersen and Makransky (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), who found that certain types of cognitive load can be advantageous when they foster specific forms of germane processing, notably in visual-technical circumstances where diagram orientation and redrawing improve comprehension.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eSchema development in technical domains\u003c/strong\u003e \u003cp\u003eThis study\u0026rsquo;s shift from visual symbol recognition to conceptual understanding sheds light on how technical schemas evolve. Participants initially processed UML diagram elements as distinct cognitive units, resulting in a high intrinsic load; however, with systematic practice, these elements were gradually grouped into meaningful patterns. These observations align with Ericsson and Kintsch's (1995) research on long-term working memory, implying that expertise growth in technical fields involves specialised knowledge structures that successfully increase working memory capacity. The connection between analytical sequencing and prior knowledge, and germane load investment, indicates how participants in this study appeared to develop what Van Merri\u0026euml;nboer and Sweller (\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2005\u003c/span\u003e) term \u0026ldquo;cognitive schemas\u0026rdquo; for managing complex learning tasks. The systematic approaches observed among all 11 participants in this study suggest that analytical sequencing and prior knowledge may be common strategies for managing UML complexity. This implies that structured analytic frameworks should be explicitly taught rather than assumed to emerge naturally. This finding extends recent work by Andersen and Makransky (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), who demonstrated similar learner-controlled approaches to cognitive load management in technology-enhanced environments. However, our study suggests that this active agency extends beyond technology-mediated learning to fundamental visual-technical interpretation tasks. This raises questions for future research about the potential transferability of these strategies across technical learning domains.\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Addressing gaps in computing education research\u003c/h2\u003e \u003cp\u003eRecent systematic reviews reveal that computing education research has \u0026lsquo;\u003cem\u003eonly superficially engaged with CLT\u0026rsquo;s theoretical developments\u0026rsquo;\u003c/em\u003e and that \u0026lsquo;\u003cem\u003ehypotheses phrased in terms of cognitive load components are rare\u0026rsquo;\u003c/em\u003e (Duran et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This study contributes to addressing these gaps by examining intrinsic, extraneous, and germane load components in visual-technical learning contexts.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eEvidence-based UML pedagogy\u003c/strong\u003e \u003cp\u003eIdentifying specific extraneous load sources offers preliminary insights that may inform CLT theory and UML class diagram pedagogical practice. The widespread occurrence of symbolic misinterpretations illustrates how unclear notation systems create extraneous load by forcing students to allocate cognitive resources to symbol decoding rather than conceptual understanding. This aligns with Chandler and Sweller\u0026rsquo;s (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e1991\u003c/span\u003e) research on split-attention effects while extending it to visual-technical domains. Participants\u0026rsquo; sophisticated UML class diagram orientation and redrawing responses demonstrate learner-generated strategies for reducing extraneous load through environmental modification. This supports Sweller\u0026rsquo;s (\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2005\u003c/span\u003e) work on a learner-controlled approach, while revealing specific mechanisms by which students adapt to poor visual design in technical contexts.\u003c/p\u003e \u003c/p\u003e \u003cp\u003eThis study addresses the multidimensional cognitive load gap identified by Andersen and Makransky (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) in computing education contexts. While their work focused on technology-enhanced learning environments, our findings demonstrate that the multidimensional conceptualisation of cognitive load is equally critical for understanding traditional visual-technical learning, such as interpreting UML class diagrams.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e5.3 Practical implications for UML class diagram education\u003c/h2\u003e \u003cp\u003eThe cognitive strategies (\u003cem\u003eanalytical sequencing and prior knowledge, symbolic confusion and clarification, diagram orientation and redrawing, and cognitive load management through decomposition\u003c/em\u003e) were found in this study.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eAnalytical sequencing and prior knowledge instruction\u003c/strong\u003e \u003cp\u003eGiven the natural evolution of sequential analysis strategies, UML class diagram training should include explicit instruction in systematic diagram interpretation techniques. Rather than relying on students to generate these spontaneously, educational materials should include formal frameworks that leverage students\u0026rsquo; intuitive decomposition tendencies, while also providing instruction on all UML class diagram elements.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eSymbolic confusion and clarification support\u003c/strong\u003e \u003cp\u003eThe observed widespread symbolic confusion emphasises the critical need for focused UML class diagram notation education that extends beyond simple symbol recognition and includes fluency-building activities. Students require extensive practice with symbol recognition in varied contexts to automate notation processing and free cognitive resources for higher-order analysis.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eDiagram orientation and redrawing support\u003c/strong\u003e \u003cp\u003eThe clear distinction between intrinsic and extraneous load sources in UML class diagram interpretation suggests a potential framework for evaluating instructional materials. Curriculum designers should thoroughly review visual presentations associated with diagrams (e.g., UML class diagrams) to avoid excessive cognitive demands while retaining the fundamental complexity required for deep understanding.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eCognitive load management through decomposition training\u003c/strong\u003e \u003cp\u003eThe observed complex decomposition processes suggest that students would benefit from targeted instruction in cognitive load management procedures. Based on these observations, educators might consider teaching students to recognise potential signs of cognitive overload and explore strategies for managing complexity in UML diagrams. The decomposition strategies observed in this study suggest potential approaches that could be explored in instructional design for diverse student populations.\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003eThis study addressed four key questions about cognitive load management in UML class diagram interpretation: what cognitive strategies students use, how they manage mental overload, what creates unnecessary mental burden, and how strategies evolve with complexity. Through think-aloud protocols with 11 first-year students across three activities of increasing complexity, we found that the participants in our exploratory study actively managed their mental load through identifiable cognitive strategies, rather than merely decoding symbols.\u003c/p\u003e \u003cp\u003eThe analysis revealed four systematic strategies that participants utilised spontaneously: \u003cem\u003eanalytical sequencing and prior knowledge, symbolic confusion and clarification, diagram orientation and redrawing, and cognitive load management through decomposition\u003c/em\u003e. These observations suggest that our participants may have developed metacognitive awareness of the demands of visual-technical learning. Participants showed agency in reducing extraneous load sources, suggesting that novice learners may be more active than previously assumed. This study provides preliminary evidence for potential applications of CLT to visual-technical domains, addressing a gap in cognitive load analysis noted in recent computing education research (Berssanette \u0026amp; De Francisco, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Duran et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Skulmowski \u0026amp; Xu, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Our findings suggest potential behavioural indicators of intrinsic, extraneous, and germane load in the context of UML class diagram interpretation. This exploratory work contributes to extending CLT applications beyond traditional text-based contexts to visual-technical learning environments, though broader validation is needed.\u003c/p\u003e \u003cp\u003eThe study identifies how students in this sample handled symbolic confusion (extraneous load), leveraged prior knowledge schemas (intrinsic load management), and employed decomposition strategies (germane load investment), providing initial insights into how CLT constructs may manifest in visual-technical learning contexts. Practically, the study offers guidelines for UML class diagram pedagogy that combine content delivery with cognitive process support. Two main sources of extraneous load were observed: \u003cem\u003esymbolic confusion resulting from unclear notation and spatial disorientation caused by unconventional layouts\u003c/em\u003e. Participants compensated through self-correction and diagram redrawing, illustrating both resilience and opportunities for instructional support. Pedagogical implications include the explicit teaching of systematic analysis strategies, the development of symbol fluency, consistent visual design standards, and the training of metacognitive strategies.\u003c/p\u003e \u003cp\u003eThis research indicates how novice learners in this South African university context can act as strategic participants in their own cognitive development. The findings indicate that participants developed cognitive load management strategies during the tasks, indicating potential for enhancement through targeted pedagogical interventions. By integrating content mastery with cognitive process support, cognitive load-aware pedagogy may inform technical education and support diverse student populations. This approach offers a foundation for developing instructional designs that better align with human cognitive processes in visual-technical learning contexts.\u003c/p\u003e"},{"header":"7 Limitations and future research directions","content":"\u003cp\u003eSeveral limitations should be considered when interpreting these results. The cross-sectional approach captures student strategies at a single point in time, hence restricting the ability to draw long-term conclusions about how cognitive load management evolves. A sample of 11 first-year students from a single South African university may not reflect varied student populations or educational contexts. This exploratory study provides preliminary insights that require replication with larger, more diverse samples before broader theoretical claims can be made.\u003c/p\u003e \u003cp\u003eFuture research should investigate whether the multidimensional cognitive load patterns identified in this study and in technology-enhanced environments (Andersen \u0026amp; Makransky, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) represent universal cognitive adaptations to complex learning environments. It should also explore whether these patterns are specific to visual-technical and technology-mediated contexts. Future research should also investigate instructional methods that systematically teach the identified cognitive processes and assess their impact on learning outcomes across various technical areas. Longitudinal studies that analyse how cognitive load management methods evolve during degree programs could provide insights into the evolution of expertise. Furthermore, comparative studies in other cultural contexts and technical visual domains could shed light on the universality of these cognitive adjustments. Understanding and supporting these cognitive processes, rather than focusing merely on information delivery, offers a viable approach for improving learning effectiveness in increasingly complex technological areas by integrating instructional design with human cognitive architecture.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eStatements and Declarations:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding: \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors did not receive support from any organisation for the submitted work.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors have no relevant financial or non-financial interests to disclose.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe anonymised interview transcripts and student artifacts collected during this study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eEthical Approval\u003c/em\u003e ─ All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the General/Human Research Ethics Committee (GHREC) of the Faculty of Natural and Agricultural Science, University of the Free State approved this research study (Reference: UFS-HSD2023/1768).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eConsent to participate ─\u003c/em\u003eAll participants were informed about the purpose of the study and their role in it. Participation was voluntary, and all participants provided written informed consent prior to taking part in the study.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eConsent to publish\u003c/em\u003e ─ Participants were informed that anonymised data may be published as part of the research findings, and all participants provided consent for publication.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePakiso J. Khomokhoana: Conceptualisation, Methodology, Data collection, Formal analysis, Writing - original draft, Writing - review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003eRouxan C. Fouché: Conceptualisation, Investigation, Formal analysis, Writing - review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003eAll authors read and approved the final manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAndersen MS, Makransky G. The validation and further development of the multidimensional cognitive load scale for physical and online Lectures (MCLS-POL). Front Psychol. 2021;12(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fpsyg.2021.642084\u003c/span\u003e\u003cspan address=\"10.3389/fpsyg.2021.642084\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eApostol DC, Bogdan R, Marcu M. (2024). UML Diagrams in Teaching Software Engineering Classes. A Case Study in Computer Science Class. \u003cem\u003e2024 IEEE 22nd World Symposium on Applied Machine Intelligence and Informatics (SAMI 2024)\u003c/em\u003e, 327\u0026ndash;332. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/SAMI60510.2024.10432905\u003c/span\u003e\u003cspan address=\"10.1109/SAMI60510.2024.10432905\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBera P. Analyzing the Cognitive Difficulties for Developing and Using UML Class Diagrams for Domain Understanding. J Database Manage (JDM). 2012;23(3):1\u0026ndash;29. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/jdm.2012070101\u003c/span\u003e\u003cspan address=\"10.4018/jdm.2012070101\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerssanette JH, De Francisco AC. Cognitive load theory in the context of teaching and learning computer programming: A systematic literature review. IEEE Trans Educ. 2022;65(3):440\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/TE.2021.3127215\u003c/span\u003e\u003cspan address=\"10.1109/TE.2021.3127215\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoustedt J. Students\u0026rsquo; different understandings of class diagrams. Comput Sci Educ. 2012;22(1):29\u0026ndash;62. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/08993408.2012.665210\u003c/span\u003e\u003cspan address=\"10.1080/08993408.2012.665210\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCastro-Alonso JC, de Koning BB, Fiorella L, Paas F. (2021). Five strategies for optimizing instructional materials: Instructor- and learner-managed cognitive load. In \u003cem\u003eEducational Psychology Review\u003c/em\u003e (Vol. 33, Issue 4, pp. 1379\u0026ndash;1407). Springer. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10648-021-09606-9\u003c/span\u003e\u003cspan address=\"10.1007/s10648-021-09606-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChandler P, Sweller J. Cognitive load theory and the format of instruction. Cognition Instruction. 1991;8(4):293\u0026ndash;332.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng P, Lowe R, Scaife M. Cognitive science approaches to understanding diagrammatic representations. Artif Intell Rev. 2001;15(1):79\u0026ndash;94.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCreswell JW, Creswell JD. Qualitative, quantitative, and mixed methods approaches. 5th ed. SAGE Publications Inc; 2017. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://us.sagepub.com/en-us/nam/research-design/book255675\u003c/span\u003e\u003cspan address=\"https://us.sagepub.com/en-us/nam/research-design/book255675\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDuran R, Zavgorodniaia A, Sorva J. Cognitive load theory in computing education research: A review. ACM Trans Comput Educ (TOCE). 2022;22(4):1\u0026ndash;27. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/3483843\u003c/span\u003e\u003cspan address=\"10.1145/3483843\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEricsson KA, Kintsch W. Long-term working memory. Psychol Rev. 1995;102(2):211\u0026ndash;45. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/0033-295X.102.2.211\u003c/span\u003e\u003cspan address=\"10.1037/0033-295X.102.2.211\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFu C, Yang D, Zhang X, Hu H. An approach to translating OCL invariants into OWL 2 DL axioms for checking inconsistency. Automated Softw Eng. 2017;24(2):295\u0026ndash;339. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10515-017-0210-9\u003c/span\u003e\u003cspan address=\"10.1007/s10515-017-0210-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHafeez A, Ahmed M, Furqan M, Rehaman WU, Husain I. Importance and Impact of Class Diagram in Software. Indian J Sci Technol. 2019;12(25):1\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKalyuga S, Plass JL. Criticisms and Desired Additional Modifications. Rethinking Cognitive Load Theory. Oxford University Press; 2025. pp. 36\u0026ndash;46. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/9780190078539.003.0004\u003c/span\u003e\u003cspan address=\"10.1093/9780190078539.003.0004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhomokhoana PJ, Fouch\u0026eacute; RC, Nkalai TS. (2025). Semiotic analysis of UML class diagrams in pedagogy: A case of first-year BCIS students. \u003cem\u003eDiscov Educ\u003c/em\u003e 4, 342 (2025). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s44217-025-00817-8\u003c/span\u003e\u003cspan address=\"10.1007/s44217-025-00817-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhomokhoana PJ, Nkalai TS. (2024). Cognitive processes used by students in understanding aggregation and composition in UML: A case of first-year BCIS students. \u003cem\u003eProceedings of the 53rd Annual Conference of the Southern African Computer Lecturers\u0026rsquo; Association (SACLA 2024)\u003c/em\u003e, 116\u0026ndash;131. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.researchgate.net/publication/386451886\u003c/span\u003e\u003cspan address=\"https://www.researchgate.net/publication/386451886\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKirschner PA, Sweller J, Kirschner F, Zambrano JR. From Cognitive Load Theory to Collaborative Cognitive Load Theory. Int J Computer-Supported Collaborative Learn. 2018;13(2):213\u0026ndash;33. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11412-018-9277-y\u003c/span\u003e\u003cspan address=\"10.1007/s11412-018-9277-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKo\u0026ccedil; H, Erdoğan AM, Barjakly Y, Peker S. (2021). UML Diagrams in Software Engineering Research: A Systematic Literature Review. \u003cem\u003eThe 7th International Management Information Systems Conference\u003c/em\u003e, 13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/proceedings2021074013\u003c/span\u003e\u003cspan address=\"10.3390/proceedings2021074013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa T. Reasoning of novice and experienced software designers on creating UML class diagrams. University of Gothenburg; 2013.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMasri K. (2009). \u003cem\u003eConceptual model design for better understanding\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://ir.lib.sfu.ca/handle/1892/112\u003c/span\u003e\u003cspan address=\"http://ir.lib.sfu.ca/handle/1892/112\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayer RE. The Cambridge handbook of multimedia learning. Cambridge University Press; 2014.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMishra SB, Alok S. Handbook of Research Methodology: A Compendium for Scholars \u0026amp; Researchers. Educreation Publishing; 2017.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMohammadi Zenouzagh Z, Admiraal W, Saab N. Empowering student engagement: the dynamics of learner traits in digital feedback environments. J Comput High Educ. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s12528-025-09459-z\u003c/span\u003e\u003cspan address=\"10.1007/s12528-025-09459-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eM\u0026ouml;ller M, Winter M, Reichert M. Cognitive Factors in Process Model Comprehension\u0026mdash;A Systematic Literature Review. Brain Sci. 2025;15(5):505. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/brainsci15050505\u003c/span\u003e\u003cspan address=\"10.3390/brainsci15050505\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNugroho A. Level of detail in UML models and its impact on model comprehension: A controlled experiment. \u0026lrm;Inf Softw Technol. 2009;51(12):1670\u0026ndash;85. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.infsof.2009.04.007\u003c/span\u003e\u003cspan address=\"10.1016/j.infsof.2009.04.007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOszti\u0026aacute;n P, K\u0026aacute;tai Z, Oszti\u0026aacute;n E. On the computational thinking and diagrammatic reasoning of first-year computer science and engineering students. Front Educ. 2022;7(933316). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/feduc.2022.933316\u003c/span\u003e\u003cspan address=\"10.3389/feduc.2022.933316\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePaas FGWC, Van Merrienboer JJG. Variability of worked examples and transfer of geometrical problem-solving skills: A cognitive-load approach. J Educ Psychol. 1994;86(1):122\u0026ndash;33.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePaas F, Sweller J. Implications of cognitive load theory for multimedia learning. In: Mayer RE, editor. The Cambridge handbook of multimedia learning. Cambridge University Press; 2014. pp. 27\u0026ndash;42.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePaas F, Tuovinen JE, Tabbers H, Van Gerven PWM. (2003). \u003cem\u003eCognitive Load Measurement as a Means to Advance Cognitive Load Theory\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1207/S15326985EP3801_8\u003c/span\u003e\u003cspan address=\"10.1207/S15326985EP3801_8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePaivio A. Mental representations: A dual coding approach. Oxford University Press; 1986.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eParcell ES, Rafferty KA. Interviews, recording and transcribing. In: Allen M, editor. The SAGE Encyclopedia of Communication Research Methods. Sage Publications, Inc.; 2017. pp. 800\u0026ndash;3.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePlowright D. Using mixed methods: Frameworks for an integrated methodology. SAGE Publications, Inc; 2011. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4135/9781526485090\u003c/span\u003e\u003cspan address=\"10.4135/9781526485090\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaunders M, Lewis P, Thornhill A. (2015). \u003cem\u003eResearch Methods for Business Students\u003c/em\u003e (7th ed.). Pearson Education Limited. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.pearson.com/store/p/research-methods-for-business-students/P100001214517\u003c/span\u003e\u003cspan address=\"https://www.pearson.com/store/p/research-methods-for-business-students/P100001214517\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShen Z, Tan S, Siau K. Challenges in learning unified modeling language: From the perspective of diagrammatic representation and reasoning. Commun Association Inform Syst. 2018;43(1):545\u0026ndash;65. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.17705/1CAIS.04330\u003c/span\u003e\u003cspan address=\"10.17705/1CAIS.04330\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSingh DP, Ahmed N, Gupta N. Business Research Method And Project Work. SBPD; 2021.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSkulmowski A, Xu KM. Understanding cognitive load in digital and online learning: A new perspective on extraneous cognitive load. Educational Psychol Rev. 2022;34(1):171\u0026ndash;96. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10648-021-09624-7\u003c/span\u003e\u003cspan address=\"10.1007/s10648-021-09624-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSweller J. Cognitive load during problem solving: Effects on learning. Cogn Sci. 1988;12(2):257\u0026ndash;85. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1207/s15516709cog1202_4\u003c/span\u003e\u003cspan address=\"10.1207/s15516709cog1202_4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSweller J. Implications of cognitive load theory for multimedia learning. In: Mayer R, editor. The Cambridge handbook of multimedia learning. Cambridge University Press; 2005. pp. 19\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSweller J. Cognitive load theory. Psychol Learn Motivation. 2011;55:37\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSweller J. Cognitive load theory and educational technology. Education Tech Research Dev. 2020;68(1):1\u0026ndash;16. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11423-019-09701-3\u003c/span\u003e\u003cspan address=\"10.1007/s11423-019-09701-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSweller J, Ayres P, Kalyuga S. Altering element interactivity and intrinsic cognitive load. Cognitive load theory - Explorations in the Learning Sciences, Instructional Systems and Performance Technologies. NY: Springer New York; 2011. pp. 203\u0026ndash;18. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-1-4419-8126-4_16\u003c/span\u003e\u003cspan address=\"10.1007/978-1-4419-8126-4_16\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eT\u0026oacute;th P, Pogatsnik M. Advancement of inductive reasoning of engineering students. Hung Educational Res J. 2022;13(1):86\u0026ndash;106. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1556/063.2022.00120\u003c/span\u003e\u003cspan address=\"10.1556/063.2022.00120\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Merri\u0026euml;nboer JJG, Sweller J. Cognitive load theory and complex learning: Recent developments and future directions. Source: Educational Psychol Rev. 2005;17(2):147\u0026ndash;77.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Merri\u0026euml;nboer JJG, Sweller J. Cognitive load theory in health professional education: Design principles and strategies. Med Educ. 2010;44(1):85\u0026ndash;93. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.1365-2923.2009.03498.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1365-2923.2009.03498.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYoung JQ, van Merrienboer JJ, Durning SJ, Cate T, O. Cognitive load theory: Implications for medical education: AMEE guide 86. Med Teach. 2014;36(5):371\u0026ndash;84. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3109/0142159X.2014.889290\u003c/span\u003e\u003cspan address=\"10.3109/0142159X.2014.889290\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou Y, Chai CS, Li X, Ma C, Li B, Yu D, Liang JC. Application of Metacognitive Planning Scaffolding for the Cultivation of Computational Thinking. J Educational Comput Res. 2023;61(6):1123\u0026ndash;42. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/07356331231160294\u003c/span\u003e\u003cspan address=\"10.1177/07356331231160294\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":false,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"diedu","sideBox":"Learn more about [Discover Education](https://www.springer.com/journal/44217)","snPcode":"44217","submissionUrl":"https://submission.nature.com/new-submission/44217/3","title":"Discover Education","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Cognitive load theory, cognitive strategies, diagram interpretation, computing education, UML class diagrams, visual-technical learning","lastPublishedDoi":"10.21203/rs.3.rs-9266264/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9266264/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eUnified Modelling Language (UML) class diagrams are critical in computing education, yet students face significant cognitive challenges in interpreting these complex visual-technical representations. Building on prior semiotic analyses of UML class diagram cognition, this narrative study (based on the integrated methodology) applies the Cognitive Load Theory (CLT) to explore how 11 first-year computing students at a South African university manage mental load during the interpretation of UML class diagrams. Through the semi-structured interviews, student-created artefacts, and systematic behavioural observations across three activities of increasing complexity (\u003cem\u003esimple element identification, relationship analysis, and complex diagram construction\u003c/em\u003e), we identified four primary cognitive strategies: analytical sequencing and prior knowledge; symbolic confusion and clarification; diagram orientation and redrawing; and cognitive load management through decomposition. Participants demonstrated metacognitive awareness of working memory limits, employing adaptive strategies like spatial rearrangement, symbolic error correction, and systematic breakdown of visual complexity. Two key sources of extraneous load emerged: symbolic confusion from ambiguous notation and spatial disorientation from unconventional layouts, prompting compensatory behaviours such as diagram redrawing and self-correction. This exploratory study provides preliminary evidence for the potential application of CLT to visual-technical learning, suggesting that these participants acted as active cognitive agents. Pedagogical implications include explicit training in systematic analysis, symbol fluency, standardised visual designs, and metacognitive strategies. These findings inform cognitive load management in technical education and suggest directions for broader validation.\u003c/p\u003e","manuscriptTitle":"Cognitive Strategies in Interpreting UML Class Diagrams with Emphasis on Load, Order and Symbolic Confusion","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-15 17:46:47","doi":"10.21203/rs.3.rs-9266264/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-05-15T11:21:53+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-14T08:08:20+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-09T02:04:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"211045628561551790636607055116404476588","date":"2026-05-09T00:52:36+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-07T05:50:29+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"235434077501744083486427084354353416338","date":"2026-05-06T20:19:02+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"279805674625960194405908380026406396621","date":"2026-05-06T11:10:22+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-05-06T08:36:59+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-04-16T20:19:58+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-31T02:33:15+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-31T02:32:32+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Education","date":"2026-03-30T11:34:25+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"diedu","sideBox":"Learn more about [Discover Education](https://www.springer.com/journal/44217)","snPcode":"44217","submissionUrl":"https://submission.nature.com/new-submission/44217/3","title":"Discover Education","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b9afc566-26bd-41ee-a515-425638d971cf","owner":[],"postedDate":"May 15th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-05-15T11:21:53+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-14T08:08:20+00:00","index":51,"fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-09T02:04:42+00:00","index":47,"fulltext":""},{"type":"reviewerAgreed","content":"211045628561551790636607055116404476588","date":"2026-05-09T00:52:36+00:00","index":46,"fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-07T05:50:29+00:00","index":43,"fulltext":""},{"type":"reviewerAgreed","content":"235434077501744083486427084354353416338","date":"2026-05-06T20:19:02+00:00","index":41,"fulltext":""},{"type":"reviewerAgreed","content":"279805674625960194405908380026406396621","date":"2026-05-06T11:10:22+00:00","index":39,"fulltext":""},{"type":"reviewersInvited","content":"15","date":"2026-05-06T08:36:59+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-05-15T17:46:47+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-15 17:46:47","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9266264","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9266264","identity":"rs-9266264","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}