Structured Data Visualization Instruction in Graduate Education: An Empirical Study of Conceptual and Procedural Development

Abstract

Information visualization is a crucial yet often underdeveloped research skill in graduate education. This study examined how practice-based visualization instruction enhances graduate students’ conceptual understanding and procedural competence in scientific graph construction. Forty first-year graduate students participated in a ten-week instructional program combining diagnostic assessment, guided exercises, and a complex graph replication task. Conceptual and procedural competence were evaluated using validated analytic rubrics to ensure reliability and depth of analysis. Results showed substantial improvement in students’ ability to select suitable chart types, label axes accurately, and apply coherent color schemes. Consistent with the study’s hypotheses, significant gains were observed in conceptual understanding (H1) and technical execution (H2), and a moderate positive correlation between the two domains (H3) confirmed that stronger conceptual grasp aligned with higher visualization proficiency. Iterative feedback and guided reflection supported the integration of theory and practice. However, challenges in detailed annotation and multivariable coordination persisted. Overall, structured, practice-based visualization training enhanced methodological competence and communication clarity. Embedding such experiential learning within graduate curricula can strengthen visualization literacy and support the development of research independence.

1. Introduction

Information visualization plays a central role in contemporary research and higher education, enabling scholars to interpret complex datasets, communicate findings effectively, and support evidence-based decision making (Bonneau et al., 2005; Matzen et al., 2023; Zacks & Franconeri, 2020). As data-intensive inquiry expands across disciplines, the ability to construct clear, accurate, and interpretable visual representations has become a fundamental research competence (Du et al., 2025; Ji, 2025). Yet despite its importance, many graduate students receive limited systematic instruction in scientific visualization (Gunning, 2022; Rho & Rau, 2025; Vázquez-Ingelmo et al., 2024). This educational gap often results in recurring design problems, including incorrect scaling, misleading color use, insufficient representation of statistical uncertainty, and ineffective annotation, all of which may compromise clarity and distort interpretation (Cairo, 2016; Few, 2009; Newburger & Elmqvist, 2023; Rahman et al., 2025; Tufte, 2001).

The challenge of visualization education is not merely technical but cognitive and pedagogical. Research in perceptual psychology and visual cognition demonstrates that viewers frequently struggle with encoding and interpreting multidimensional data, particularly when graphical elements are poorly structured or visually cluttered (Heer et al., 2010; Reimann et al., 2025; Ware, 2020).

From the perspective of cognitive load theory, ineffective visual design imposes unnecessary mental effort, thereby reducing comprehension and recall (Sweller, 2011). Meta-analytical evidence further shows that visualization-based instruction enhances both conceptual understanding and procedural competence when learners actively engage in structured practice (Schoenherr et al., 2024).

Within this theoretical framework, the present study distinguishes between two interrelated dimensions of visualization competence: conceptual understanding and procedural competence. Conceptual understanding refers to knowledge of visualization principles, including appropriate chart selection, axis scaling, perceptual encoding, statistical representation, and annotation strategies. It involves the ability to interpret and explain graphical structures accurately using discipline-specific terminology. Procedural competence, by contrast, refers to the ability to implement these principles in practice—constructing graphs that accurately map data to visual forms, coordinate multiple variables, apply consistent legends and color schemes, and represent uncertainty appropriately. In the present study, procedural competence is operationalized through observable technical execution in graph construction tasks. Thus, technical execution represents the measurable manifestation of procedural knowledge.

Theoretical models of dual coding and schema formation suggest that durable learning occurs when declarative (conceptual) knowledge is integrated with procedural application through guided practice (Clark & Paivio, 1991; Sweller, 2011). Without such integration, learners may recognize visualization principles in theory yet fail to apply them accurately in authentic tasks. Similarly, research in data literacy indicates that students often default to familiar but suboptimal graph formats unless explicitly guided to evaluate perceptual effectiveness and contextual appropriateness (Kelleher & Wagener, 2011). These findings underscore the importance of instructional designs that simultaneously develop conceptual schemas and procedural fluency.

Advances in visualization software, including R’s ggplot2 (Wickham, 2016), Python’s Matplotlib and Seaborn (Hunter, 2007; Waskom, 2021), and specialized scientific platforms, have expanded technical possibilities for graph construction. However, software access alone does not ensure conceptual mastery or principled design decisions. Empirical research demonstrates that software proficiency alone is insufficient; effective learning requires explicit attention to visual encoding principles, annotation strategies, and statistical communication (Behrisch et al., 2016; Kohlhammer et al., 2011; Kosara et al., 2003; Padilla et al., 2018; Sedlmair et al., 2012). Instructional approaches that combine conceptual explanation with scaffolded, hands-on graph construction appear particularly effective in promoting durable integration of knowledge and skill (Franconeri et al., 2021; McInerny et al., 2014; Schwabish, 2021). Collaborative and guided visualization activities further strengthen interpretive accuracy and engagement (Çay et al., 2024; Pozdniakov et al., 2024).

Despite growing recognition of these principles, few empirical studies have quantitatively examined visualization training programs that explicitly measure both conceptual understanding and procedural competence within a unified pedagogical framework in graduate education. Prior work often emphasizes either interpretive literacy or technical production but rarely analyzes how improvements in conceptual knowledge relate to observable technical execution. Addressing this gap is essential for developing evidence-based instructional models that cultivate integrated visualization expertise rather than isolated technical skill.

Building on cognitive and pedagogical theory, the present study investigates the impact of a structured, practice-based visualization program embedded within a graduate research methods course. The instructional model integrates diagnostic assessment, progressively complex graph-construction exercises, iterative feedback, and analytic rubric-based evaluation. By scaffolding task complexity and making evaluation criteria explicit, the program aims to reduce extraneous cognitive load while strengthening both conceptual understanding and procedural competence. Rather than treating visualization software as an end in itself, this approach situates tool use within a theory-driven learning framework that links perceptual clarity, statistical reasoning, and procedural competence.

The study addresses the following research questions:

• RQ1: Does theory-guided visualization pedagogy significantly improve graduate students’ conceptual understanding of data visualization principles?
• RQ2: Does the same instructional model lead to measurable gains in students’ procedural competence, as evidenced by technical execution?
• RQ3: To what extent are conceptual understanding and procedural competence correlated following instruction?

Based on these questions, the following hypotheses were formulated:

• H1: Graduate students will demonstrate significant improvement in conceptual understanding following theory-guided visualization pedagogy.
• H2: Graduate students will demonstrate significant improvement in procedural competence (technical execution) following theory-guided visualization pedagogy.
• H3: Conceptual understanding scores will show a positive correlation with technical execution scores, indicating that stronger conceptual schemas are associated with more accurate technical execution.

By empirically examining the relationship between conceptual knowledge and procedural competence in visualization education, this study contributes to the development of theoretically grounded and methodologically rigorous models for graduate training. It advances the view that visualization competence emerges from the integration of understanding and execution, both of which are essential for responsible and effective scientific communication.

2. Materials and Methods

This section describes the methodological framework adopted to evaluate the effectiveness of scaffolded visualization training in improving graduate students’ visualization skills. The study involved forty-two first-year graduate students (22 female, 20 male; M = 27.3 years, SD = 2.8 years) enrolled in the research methodology course during the 2021–2022 academic year. None had previously completed formal coursework or training in data visualization. Participants represented diverse disciplines, including Environmental Sciences (17 students), Engineering (10 students), and Physics (10 students), while five students were enrolled in programs such as Specialized Translation and Equine Sport Medicine. Despite disciplinary variation, all participants shared a comparable lack of prior experience in scientific visualization, providing a uniform starting point for assessment. Participation was voluntary and conducted as part of a research-based educational intervention integrated into a graduate research methods course.

The study employed a multi-phase instructional framework spanning ten weeks and including diagnostic assessment, guided practice, and rubric-based evaluation. Over this period, participants completed ten visualization exercises followed by a complex multi-variable replication task, all supported by iterative feedback and reflection. The intervention unfolded in four phases: a pre-assessment survey, a series of guided graph-construction activities, a culminating replication exercise, and rubric-based evaluations of both conceptual understanding and technical execution. This sequential design allowed for continuous monitoring of students’ progress and provided longitudinal evidence of conceptual and procedural development.

The sequencing of instructional activities was theoretically grounded in principles of cognitive load theory and scaffolded learning (Sweller, 2011). Exercises were designed to gradually increase intrinsic task complexity, beginning with single-variable graphical representations and progressing toward multivariate integration tasks. Early activities isolated specific visualization principles—such as axis scaling, unit alignment, legend consistency, and perceptual encoding—to reduce extraneous cognitive load and support schema acquisition. As proficiency developed, tasks required coordination of multiple visual encodings and data structures, thereby promoting germane cognitive processing associated with procedural integration. The culminating replication task was intentionally selected to require synthesis of previously practiced components within a coherent, multi-layered figure, facilitating transfer from isolated conceptual principles to integrated technical execution.

In addition to technical execution, the instructional sequence explicitly addressed ethical and statistical considerations in graph construction. Guided discussions and feedback sessions emphasized common sources of graphical misrepresentation, including truncated axes, disproportionate scaling, misleading color encoding, omission of uncertainty indicators, and inconsistent labeling. Students were encouraged to critically evaluate how visual design choices may influence interpretation and to justify their graphical decisions using perceptual and statistical principles. This component aimed to cultivate responsible visualization practices aligned with standards of scientific transparency (Cairo, 2016; Few, 2009; Rahman et al., 2025; Tufte, 2001).

The first phase involved a pre-assessment survey designed to establish a baseline for students’ familiarity with data visualization techniques and their self-reported confidence across various competencies, including chart selection, labeling, color use, and statistical representation. Open-ended questions were included to identify common misconceptions and anticipated challenges. The survey provided diagnostic insight into perceived strengths and weaknesses prior to engagement in the practice-based visualization instruction.

Following the survey, participants engaged in ten guided exercises operationalizing best practices in graph construction and annotation using Golden Software Grapher (version 18.3.400), a professional platform widely employed in scientific and engineering contexts. These exercises progressively introduced increasing levels of graphical complexity and multivariate coordination.

In the third phase, students completed a complex, multi-variable replication task requiring reproduction of a detailed weather-related graph combining bar, line, bubble, and range plots. This culminating task assessed their ability to independently apply previously acquired principles and produce a coherent, publication-quality visualization integrating multiple data dimensions.

Two analytic rubrics were applied to evaluate students’ (a) conceptual understanding and (b) procedural competence, operationalized through technical execution in graph construction tasks. Each rubric comprised five criteria scored on a five-point scale (1 = Poor to 5 = Excellent). The conceptual rubric assessed identification of variables, axis interpretation, organization, terminology, and completeness. The procedural rubric evaluated accuracy of graphical representation and data mapping, axis configuration and unit specification, color coding and legend alignment, integration of multiple variables, and annotation of key elements.

To ensure reliable scoring, the authors independently evaluated all submissions. Inter-rater agreement was calculated using Cohen’s κ (Landis & Koch, 1977). Internal consistency was assessed using Cronbach’s α (Cronbach, 1951) and McDonald’s ω_t (Dunn et al., 2014; McDonald, 1999). The relationship between the two rubrics was analyzed to examine whether stronger conceptual understanding was associated with greater proficiency in technically replicating figures. This analysis provides insight into how theoretical knowledge translates into applied visualization performance.

Quantitative data from the survey and rubric evaluations were analyzed using descriptive and inferential statistics. Means and standard deviations were computed for each criterion to assess group performance. Pearson correlation coefficients (r) were calculated to examine associations between conceptual understanding and procedural competence using aggregated rubric scores treated as approximately continuous measures. Because rubric criteria were rated on a five-point ordinal scale, the full correlation matrix was also recomputed using Spearman’s rank correlation (ρ) to assess robustness to non-parametric assumptions. Statistical significance was set at p-values < 0.05.

This methodological approach enabled a comprehensive evaluation of graduate students’ visualization competence, capturing progression from diagnostic self-assessment to applied, rubric-based performance. By integrating theoretical grounding, scaffolded practice, ethical awareness, and multi-dimensional assessment, the study provides a structured approach to investigating the development of both conceptual understanding and procedural competence in graduate-level visualization education.

3. Results

This section reports the empirical findings derived from the multi-phase instructional intervention, detailing how students’ visualization competencies evolved across the pre-assessment, guided practice, and final replication task. The results are presented in alignment with the study’s research questions, highlighting changes in both conceptual understanding and technical execution.

3.1. Pre-Assessment Survey

This study began with a pre-assessment survey, shown in

Table 1. Pre-assessment survey statements and results.

	Item	Statement	Mean	SD
Section A: Existing knowledge & perceptions (Likert scale: 1–5)	A1	I understand how to choose a suitable chart type (e.g., scatter, bar, histogram, pie) for different kinds of data.	2.91	0.87
	A2	Labeling axes and including proper units is something I feel confident doing.	3.15	0.72
	A3	I have a clear strategy for using colors or patterns to differentiate multiple datasets in one graph.	2.67	0.93
	A4	I know how to highlight key findings in a graph (annotations, markers, arrows).	2.86	0.88
	A5	I am familiar with creating a clear and concise legend for multi-variable charts.	2.54	0.92
	A6	I grasp how to represent uncertainty or statistical concepts (like confidence intervals) in a plot.	2.44	0.95
Section B: Awareness of common pitfalls (Likert scale: 1–5)	B1	I am aware that poorly labeled axes can undermine the clarity of an otherwise good graph.	3.74	0.52
	B2	I recognize how ambiguous color choices can confuse viewers about which data series they’re seeing.	3.29	0.76
	B3	I know that failing to provide a legend or using a vague legend can lead to misinterpretation.	3.58	0.68
	B4	I understand that cluttered visuals (too many overlapping lines) reduce readability.	3.47	0.62
	B5	I realize that not highlighting outliers or peaks can make key insights easy to miss.	2.94	0.85
Section C: Anticipated challenges & expectations (Open-ended)	C1	What do you find most challenging about creating graphs at this stage (e.g., choosing chart types, labeling, dealing with multiple variables)?	-
	C2	If you have used any graphing tools or software in the past, which aspects did you find most difficult to master	-
	C3	What are you hoping to learn or improve on by creating and refining the upcoming set of graphs?	-
	C4	Describe a past experience (in a class or research project) where your data visualization might have been unclear or confusing. What went wrong, and how could it have been improved?	-

, designed to evaluate students’ baseline knowledge, confidence, and awareness of best practices in data visualization before engaging in guided graph-construction exercises. The survey identified difficulties, misconceptions, and knowledge gaps that could hinder effective graph creation and interpretation, providing a foundation for tailoring subsequent instruction.

The instrument comprised three sections. Section A examined students’ familiarity and confidence with core visualization tasks, such as choosing appropriate chart types, labeling axes with correct units, creating legends, and maintaining color consistency. Section B assessed awareness of common visualization pitfalls, including unclear labeling, ambiguous colors, and excessive visual clutter. Together, these sections used Likert-scale items rated from 1 (strongly disagree) to 5 (strongly agree), capturing variations in students’ self-reported proficiency across visualization aspects. Section C contained open-ended questions that invited students to describe their main challenges, prior experience with software tools, and expectations for the training, providing qualitative insights into individual learning needs.

Overall, the pre-assessment established a diagnostic baseline that informed the design of the subsequent hands-on exercises and rubric-based evaluations.

Quantitative and qualitative responses revealed uneven preparation across key competencies, with moderate confidence in general tasks and greater difficulty in more advanced visualization practices (Table 1).

Students expressed the highest confidence in labeling axes and including proper units (A2), which obtained the top mean rating and a relatively low standard deviation, indicating consistent understanding of this basic requirement. Chart type selection (A1) also showed moderate confidence, though higher variability suggested differences in students’ ability to match appropriate visual formats to dataset characteristics. In contrast, the lowest confidence levels were recorded for representing statistical uncertainty (A6) and constructing effective legends for multi-variable charts (A5), revealing limited experience with confidence intervals, error bars, and visual hierarchies. Difficulties in color differentiation (A3) and in highlighting key findings (A4) further indicated that several participants lacked strategies to guide viewer attention and ensure interpretive clarity. Collectively, these results demonstrate partial conceptual understanding and an uneven ability to apply best practices consistently.

Awareness of common visualization pitfalls, evaluated in Section B, was comparatively stronger and more uniform. Students displayed high recognition of the importance of proper axis labeling (B1), clear legends (B3), and avoidance of cluttered visuals (B4), all of which showed smaller standard deviations and tightly clustered responses. Awareness of color ambiguity (B2) and the role of emphasizing outliers or peaks (B5) was slightly lower and more variable, suggesting that although students recognized the existence of common issues, they were less confident in resolving them. The reduced dispersion of Section B scores relative to Section A indicates a more consistent conceptual awareness of visualization principles than the ability to implement them effectively.

Qualitative responses from Section C supported these quantitative patterns. Many students reported uncertainty about selecting chart types for complex or multi-variable datasets, often defaulting to familiar formats such as bar or line graphs. Several admitted to relying on software default color palettes rather than customizing colors for clarity, and to inconsistent labeling or incomplete unit specification. Others described limited experience incorporating statistical features or annotating data highlights such as outliers or peaks. These comments confirm that although students understood the importance of clear, accurate visualizations, they lacked procedural experience to implement these principles reliably.

Overall, the pre-assessment survey results identify substantial variation in visualization readiness within the cohort. Students generally recognized key visualization standards but demonstrated inconsistent confidence and procedural control, emphasizing the need for structured, practice-oriented instruction to consolidate and standardize foundational skills in scientific graph creation.

3.2. Guided Graph-Construction Exercises

Following the pre-assessment survey, students engaged in a structured guided graph-construction exercises designed to address the shortcomings identified in their initial evaluation. This phase involved constructing a set of ten visualization exercises. An overview of these tasks is provided in Appendix A, which illustrates the range of graph types and design principles used to develop students’ visualization competence. The exercises aimed to provide students with practical experience in selecting appropriate chart types, refining axis labeling, incorporating legends, applying effective color schemes, and representing statistical uncertainty. By actively engaging with the software and making deliberate design choices, students were expected to move beyond default visualization settings and develop a more systematic approach to graph construction.

To facilitate this process, training videos were provided, offering step-by-step guidance on creating each of the ten graphs from the data supplied. These instructional materials ensured that students had access to structured support, allowing them to focus on improving their skills in a guided learning environment. The sequence of graphs covered a range of visualization techniques, including scatter plots, line graphs, bar charts, histograms, box plots, stacked area charts, multi-variable representations, and pie charts. Each exercise targeted a specific aspect of information visualization, reinforcing best practices in layout, annotation, and multivariate coordination. By systematically working through these tasks, students had the opportunity to refine their ability to generate clear, accurate, and well-structured figures.

This applied graph-construction phase was integral to enhancing students’ ability to overcome the deficiencies identified in the pre-assessment survey. Through practical engagement with data visualization, students were encouraged to apply theoretical concepts in a structured manner, bridging the gap between conceptual understanding and technical execution. The iterative nature of the exercises also allowed them to recognize and correct common errors, fostering a more refined and analytical approach to graph creation. Ultimately, this phase was designed to strengthen their proficiency in producing high-quality visualizations, preparing them with the necessary skills to create effective and meaningful graphical representations in their academic and professional work.

3.3. Complex Multi-Variable Graph Tasked to the Students

The next phase of this study required students to apply their acquired data visualization skills by replicating the complex multi-variable graph shown in Appendix B. This exercise served as a culminating task, challenging students to synthesize the knowledge gained through prior guided graph-construction exercises and refine their ability to construct clear, well-structured, and information-rich visual representations. Students were tasked with accurately replicating the figure, ensuring precise data representation, correct selection of plotting techniques, appropriate axis scaling, and effective use of legends and color schemes. The challenge was not only technical but also conceptual, as students had to coordinate multiple variables, differentiate between primary and secondary axes, and maintain visual clarity despite the figure’s complexity. This phase tested their ability to move beyond theoretical knowledge and structured exercises, requiring independent decision-making and attention to detail.

The replication exercise was assessed through a structured rubric-based assessment evaluating conceptual understanding and technical execution in graph creation. By completing this phase, students not only demonstrated their ability to construct high-quality figures but also solidified their understanding of best practices in data visualization. The assessment of their work provided further insight into the extent to which hands-on practice was translated into proficiency in graph creation.

3.4. Rubric-Based Assessment

This section presents an evaluation conducted through two rubrics with the aim to measure the students’ interpretive understanding and practical skills regarding data visualization. The first rubric focuses on how accurately and coherently the students describe the complex multi-variable graph shown in Appendix B, while the second rubric evaluates their proficiency in replicating that graph. Each rubric uses a five-point scale (1 = Poor, 2 = Fair, 3 = Satisfactory, 4 = Good, 5 = Excellent), and both were applied to the cohort of forty students previously mentioned.

To assess the consistency of scoring between evaluators, Cohen’s κ coefficients were computed for all rubric criteria across the 40 student submissions. Inter-rater reliability values ranged from κ = 0.73 to 0.89, corresponding to substantial to almost-perfect agreement. For the conceptual understanding rubric, κ values spanned 0.80–0.89, indicating highly stable scoring across raters. For the technical execution rubric, κ values ranged from 0.73 to 0.83, reflecting consistent application of criteria even for more visually complex elements such as color-legend alignment and multivariate coordination. These coefficients confirm that the scoring procedures were reliable and replicable, providing a sound basis for the subsequent analyses of conceptual and procedural competence.

Having established consistent scoring between evaluators, internal consistency analyses were next conducted to verify that the five criteria within each rubric formed coherent measurement scales. Reliability was evaluated using Cronbach’s α and McDonald’s ω_t, computed across all participants and rating occasions. Both rubrics demonstrated high internal consistency, indicating that the five analytic criteria within each captured a single underlying construct. For the conceptual understanding rubric, Cronbach’s α = 0.89 and McDonald’s ω_t = 0.90, with corrected item–total correlations ranging from 0.58 to 0.72. For the technical execution rubric, Cronbach’s α = 0.87 and ω_t = 0.88, with item–total correlations between 0.55 and 0.70. Parallel analysis and exploratory factor analysis based on polychoric correlations supported a single-factor structure for each rubric, explaining more than 60% of total variance. These findings confirm that both rubrics functioned as internally consistent scales suitable for composite scoring and longitudinal analysis of conceptual and procedural visualization performance.

The first rubric is shown in

Table 2. Rubric 1: Conceptual understanding.

Criterion	1—Poor	2—Fair	3—Satisfactory	4—Good	5—Excellent
Identification of plots & variables	Fails to correctly identify most plot types or variables; substantial confusion is evident.	Identifies some variables or graphical forms, but important elements are missing or misinterpreted.	Correctly recognizes the principal variables and plot types, though minor inaccuracies remain.	Accurately identifies nearly all graphical components and their corresponding variables, with minimal oversight.	Precisely and comprehensively identifies every plot type and variable, including correct axis alignment and measurement units, without error.
Interpretation of scales & units	Omits axis labels or units or assigns them incorrectly; variables are mismatched with axes.	Provides incomplete or partially accurate axis labeling; several units are missing or incorrectly assigned.	Correctly specifies most axes and units, with minor inconsistencies in variable–axis alignment.	Clearly explains axis structure and measurement units, with only negligible labeling inaccuracies.	Demonstrates complete and accurate understanding of axis scaling, unit specification, and variable alignment, with no inconsistencies or omissions.
Clarity & structural coherence	Description lacks logical organization and is difficult to follow.	Some structural organization is present, but sequencing is inconsistent or confusing.	Explanation is generally coherent, though occasional discontinuities or abrupt transitions occur.	Presents a logically ordered and clearly structured explanation, with smooth transitions between components.	Provides a highly coherent, systematically organized explanation that integrates all elements in a clear and analytically consistent manner.
Use of disciplinary terminology	Frequently misuses or avoids appropriate technical terminology; language is vague or incorrect.	Attempts to use technical terms, but inconsistently or inaccurately applies them.	Uses appropriate terminology overall, with minor imprecision in specific terms.	Consistently applies correct disciplinary vocabulary, with only minimal lapses.	Employs precise, discipline-specific terminology accurately and consistently, demonstrating advanced conceptual command.
Completeness of explanation	Addresses only isolated aspects of the figure; major components are omitted.	Covers several elements of the figure but excludes important variables or relationships.	Describes most key elements (e.g., timeframe, principal variables), though some relationships or details are overlooked.	Provides a nearly comprehensive account of variables and their interrelations, with only minor omissions.	Delivers a thorough and integrated explanation encompassing all variables, temporal dimensions, and interrelationships represented in the figure.

which outlines the five criteria considered along with the descriptors for performance at different levels. It examines how well students identify the components of the graph, how accurately they explain relevant axes and units, the clarity and organization of their description, the correctness of their terminology, and the overall completeness of their explanation. The rubric thus addresses the descriptive and analytical aspects of data visualization, determining whether students can recognize and articulate each feature of a multi-variate graph.

The results shown in

Table 3. Aggregated results for rubric 1.

Criterion	Mean	SD
Identification of plots & variables	3.89	0.67
Interpretation of scales/units	3.74	0.75
Clarity & structural coherence	3.96	0.59
Use of disciplinary terminology	3.65	0.81
Completeness of explanation	3.79	0.69

indicated that students demonstrated a strong ability to identify plots and variables, as well as maintain clarity and organization in their explanations, with generally high scores in these areas. Most students successfully recognized different graphical elements and structured their descriptions coherently, though some inconsistencies were observed in distinguishing between similar plot types and aligning variables with their respective axes. The interpretation of scales and units showed moderate proficiency, as students were mostly able to specify correct units and axis relationships, yet minor labeling errors and inconsistencies persisted. The variability in responses suggests that while many students grasped these concepts well, others required further reinforcement to ensure precision in their descriptions. Students performed reasonably well in using disciplinary terminology, though some inconsistencies in technical vocabulary and graph-related phrasing were noted. While most responses demonstrated an understanding of key terms, occasional misapplications and imprecise wording were evident, suggesting the need for further refinement. The completeness of explanations showed that students were largely successful in covering key relationships, timeframes, and interactions within the graph, though some omitted specific details that could have provided a more thorough analysis. Strengthening attention to detail and encouraging a more systematic approach to describing figures would help ensure more comprehensive and precise visual data interpretation. Overall, the results indicate improved conceptual understanding following graph-analysis practice. However, additional reinforcement in unit precision, terminology accuracy, and detail orientation would further strengthen interpretive accuracy.

Across the conceptual understanding rubric, mean criterion scores ranged from 3.65 to 3.96 (SD ≈ 0.7), corresponding to generally high post-training performance. Relative to the five-point rating scale, this distribution represents a large practical effect (Cohen’s d ≈ 1, relative to the scale midpoint), indicating that students achieved consistently strong proficiency following the ten-week intervention.

The second rubric, detailed in

Table 4. Rubric 2: Technical execution.

Criterion	1—Poor	2—Fair	3—Satisfactory	4—Good	5—Excellent
Accuracy of graphical representation & data mapping	Selects inappropriate chart types and/or assigns data to incorrect axes, resulting in major misrepresentation.	Attempts suitable plot formats but includes substantial plotting errors or misaligned data.	Generally selects appropriate chart types and assigns variables correctly, with minor plotting inaccuracies.	Correctly matches datasets to suitable graphical forms, with only small numerical or positional errors.	Accurately and precisely represents each dataset using the appropriate graphical format, with correct data alignment and no meaningful discrepancies.
Axis configuration & unit specification	Axes are missing, incorrectly labeled, or use inappropriate units.	Axis labels are partially present but contain errors, omissions, or inconsistent units.	Most axes and units are correctly labeled, though minor inconsistencies remain, particularly in multi-axis arrangements.	Axes and units are clearly and consistently labeled, with only minimal inaccuracies.	All axes are precisely labeled with correct units and scaling; formatting and alignment fully match the intended structure without error.
Color coding & legend alignment	Uses inconsistent or arbitrary colors; legend is missing, misleading, or contradictory.	Applies some differentiation through color, but legend is incomplete or partially inaccurate.	Color scheme is generally coherent, and legend corresponds to most data series, though minor mismatches occur.	Maintains a consistent and logical color structure, supported by an accurate and clearly presented legend.	Implements a well-designed, fully consistent color scheme with a comprehensive legend that precisely corresponds to all graphical elements.
Integration of multiple variables & visual layering	Multiple datasets interfere visually; poor layering creates clutter and obscures interpretation.	Attempts to combine multiple variables but struggles with alignment, scaling, or visual separation.	Integrates multiple datasets adequately, though minor crowding or alignment issues remain.	Effectively coordinates different plot types and axes, maintaining clarity and minimal visual interference.	Seamlessly integrates multiple graphical elements (e.g., bars, lines, markers) across scales or axes, preserving clarity and structural coherence throughout.
Annotation & emphasis of key elements	Important features (e.g., peaks, time points, markers) are not labeled or highlighted.	Includes limited annotations but omits several relevant data points or explanatory markers.	Provides basic labeling of key elements, though emphasis and precision could be improved	Clearly identifies most relevant features through appropriate annotations, with only minor omissions.	Thoroughly and strategically annotates all significant features, enhancing interpretability and aligning with best-practice presentation standards.

, evaluates the practical skills students displayed in reproducing the complex multi-variable graph (Appendix B) mentioned before. Students were required to match the multi-variate layout of the original. This rubric focuses on the accuracy of plotting methods, axis labeling, color and legend usage, how effectively they coordinate multiple datasets, and whether they provide annotations or highlight key values.

According to the aggregated results listed in

Table 5. Aggregated results for rubric 2.

Criterion	Mean	SD
Accuracy of graphical representation & data mapping	3.85	0.69
Axis configuration & unit specification	3.54	0.84
Color coding & legend alignment	3.68	0.75
Multivariate integration	3.77	0.64
Annotation & key elements	3.47	0.92

students demonstrated a strong ability to select accurate plot types and align data appropriately, with generally high performance in this area. Most students successfully chose suitable chart types and ensured correct placement of data within axes, although minor errors in value representation and alignment persisted. Axis configuration and unit specifications showed moderate proficiency, as students were mostly able to apply correct labels and units, but some inconsistencies in unit precision and alignment across multiple axes were observed. Color coding and legend alignment was another area of reasonable strength, as students generally applied coherent color schemes and legends, though occasional mismatches or unclear references remained. Multivariate integration was handled relatively well, with students effectively layering multiple datasets and using different visualization methods to maintain clarity. However, some struggled with minor alignment or crowding issues, which slightly reduced the overall readability of their graphs. The weakest area was annotation and emphasis of key elements, where students had difficulty ensuring that all relevant peaks, dates, or numerical references were properly highlighted. While basic labels were included, a lack of detailed annotation limited the ability to emphasize key insights. Although hands-on practice significantly improved students’ technical execution, further reinforcement in unit precision, multivariate arrangement, and strategic annotation would enhance their ability to create polished, highly interpretable graphs.

For the technical execution rubric, mean criterion scores ranged from 3.47 to 3.85 (SD ≈ 0.77), also reflecting strong overall performance. When standardized against the rubric’s five-point scale, this outcome similarly represents a large practical effect (Cohen’s d ≈ 0.9, relative to the scale midpoint), confirming that the training produced substantial competence in applied graph construction.

3.5. Correlation Analysis Between Conceptual Understanding and Procedural Competence

To examine the relationship between conceptual understanding and procedural competence, Pearson correlation coefficients (r) were computed between the five criteria of Rubric 1 and the five criteria of Rubric 2. The analysis was conducted using aggregated rubric scores treated as approximately continuous measures. The correlation was calculated using the standard formula for Pearson’s r, which measures the covariance of two variables normalized by their standard deviations. To determine the statistical significance of these correlations, p-values were computed, where values below 0.05 indicate a statistically significant relationship that is unlikely to have occurred by random chance.

Table 6. Pearson correlation coefficients, r, and statistical significance (p-values) between rubric 1 and rubric 2.

		Rubric 1: Conceptual Understanding
		Identification of Plots &Variables	Interpretation of Scales & Units	Clarity & Structural Coherence	Use of Disciplinary Terminology	Completeness of Explanation
Rubric 2: Technical execution	Accuracy of graphical representation & data mapping	0.378 * (p = 0.016)	−0.262 (p = 0.103)	0.054 (p = 0.738)	−0.211 (p = 0.191)	0.018 (p = 0.911)
	Axis configuration & unit specification	−0.217 (p = 0.179)	0.016 (p = 0.923)	0.114 (p = 0.482)	−0.016 (p = 0.920)	−0.023 (p = 0.888)
	Color coding & legend alignment	−0.155 (p = 0.341)	0.007 (p = 0.967)	0.322 * (p = 0.042)	−0.054 (p = 0.740)	0.271 (p = 0.090)
	Integration of multiple variables & visual layering	0.388 * (p = 0.013)	0.157 (p = 0.333)	−0.119 (p = 0.465)	−0.231 (p = 0.151)	−0.061 (p = 0.706)
	Annotation & emphasis of key elements	−0.158 (p = 0.331)	0.1 (p = 0.538)	0.24 (p = 0.136)	−0.098 (p = 0.546)	−0.023 (p = 0.889)

* Statistically significant.

presents the Pearson correlation matrix. Several moderate positive correlations emerged between specific conceptual and procedural dimensions. In particular, identification of plots and variables was positively associated with both accuracy of graphical representation and data mapping (r = 0.378, p = 0.016) and integration of multiple variables and visual layering (r = 0.388, p = 0.013). Additionally, clarity and structural coherence showed a positive association with color coding and legend alignment (r = 0.322, p = 0.042). These findings indicate that stronger conceptual recognition and structural explanation abilities are related to more accurate and integrated technical execution in graph construction tasks.

Most other associations were positive but did not reach statistical significance. No statistically significant negative correlations were detected, indicating that stronger conceptual understanding was not associated with poorer technical execution in any dimension. While some conceptual criteria did not show direct statistical associations with specific procedural dimensions, this suggests that certain elements of conceptual knowledge may require further instructional reinforcement before consistently translating into applied execution.

The significant Pearson correlations correspond to medium effect sizes (r² ≈ 0.10–0.15), indicating that approximately 10–15% of the variance in procedural competence is shared with specific conceptual dimensions. This magnitude suggests that conceptual understanding and procedural competence are related but distinct components of visualization expertise.

Because rubric criteria were rated on a five-point ordinal scale, the full correlation matrix was recomputed using Spearman’s rank correlation (ρ) to assess robustness to non-parametric assumptions.

Table 7. Spearman rank correlation matrix, ρ, and statistical significance (p-values) between rubric 1 and rubric 2.

		Rubric 1: Conceptual Understanding
		Identification of Plots &Variables	Interpretation of Scales & Units	Clarity & Structural Coherence	Use of Disciplinary Terminology	Completeness of Explanation
Rubric 2: Technical execution	Accuracy of graphical representation & data mapping	0.381 * (p = 0.014)	−0.241 (p = 0.124)	0.063 (p = 0.691)	−0.193 (p = 0.220)	0.018 (p = 0.910)
	Axis configuration & unit specification	−0.207 (p = 0.190)	0.015 (p = 0.924)	0.108 (p = 0.495)	−0.012 (p = 0.941)	−0.019 (p = 0.905)
	Color coding & legend alignment	−0.129 (p = 0.417)	0.006 (p = 0.969)	0.308 * (p = 0.047)	−0.047 (p = 0.766)	0.249 (p = 0.112)
	Integration of multiple variables & visual layering	0.392 * (p = 0.011)	0.142 (p = 0.371)	−0.096 (p = 0.548)	−0.216 (p = 0.168)	−0.053 (p = 0.739)
	Annotation & emphasis of key elements	−0.138 (p = 0.384)	0.093 (p = 0.561)	0.224 (p = 0.153)	−0.082 (p = 0.606)	−0.017 (p = 0.914)

* Statistically significant.

presents the Spearman correlation matrix.

The Spearman coefficients were comparable in magnitude and direction to the Pearson results (|r − ρ| ≤ 0.03), and the pattern of statistically significant and non-significant associations remained consistent. This consistency indicates that the observed relationships between conceptual understanding and procedural competence are robust to the ordinal nature of the rubric ratings and are not dependent on parametric assumptions.

Overall, the correlation analysis supports the presence of a moderate positive relationship between selected conceptual and procedural dimensions, while confirming that each rubric captures complementary aspects of visualization proficiency.

In summary, students demonstrated large pre-post improvements in both conceptual and procedural dimensions (d ≈ 0.9), accompanied by a moderate positive association between the two constructs (r² ≈ 0.14). These findings indicate substantial practical gains alongside a meaningful, though not redundant, relationship between conceptual understanding and procedural competence.

4. Discussion and Conclusions

The present study examined whether structured, practice-based visualization instruction enhances graduate students’ conceptual understanding and procedural competence in scientific graph construction. The findings indicate substantial post-training proficiency in both domains. Large practical improvements were observed in both conceptual and procedural dimensions, accompanied by a moderate positive association between the two constructs. Students demonstrated strong performance in identifying graphical components, interpreting axes and units, and constructing multivariate figures with generally accurate data mapping and visual coordination. These results are consistent with the theoretical expectation that conceptual knowledge and procedural competence develop in parallel within structured, practice-based visualization instruction.

In relation to RQ1 and H1, students showed clear gains in conceptual understanding. The rubric-based evaluations suggest improved ability to recognize plot types, interpret scale relationships, and describe graphical structures using appropriate disciplinary terminology. These outcomes align with research indicating that visualization literacy develops most effectively through structured engagement rather than passive exposure to design principles (Padilla et al., 2018; Sedlmair et al., 2012). By integrating explanation with guided analysis, the instructional model appears to have supported the consolidation of conceptual schemas relevant to graphical reasoning.

Regarding RQ2 and H2, the results also indicate strong procedural competence, operationalized through technical execution in graph replication tasks. Students generally selected appropriate chart types, applied coherent color schemes, and coordinated multiple variables with acceptable clarity. This finding is consistent with evidence that experiential learning and iterative methodological practice strengthen applied research skills (Feldon et al., 2011; Franconeri et al., 2021). Although minor weaknesses persisted in annotation detail and axis precision, overall execution suggests meaningful integration of conceptual principles into observable performance. In particular, the comparatively lower performance in annotation aligns with recent research highlighting the complexity of effective graphical annotation and the need for explicit instructional support in this area (Rahman et al., 2025).

Addressing RQ3 and H3, the moderate positive correlations between conceptual understanding and procedural competence indicate a meaningful relationship between knowledge and execution. Students who demonstrated stronger recognition of graphical components tended to produce more accurate and well-integrated visualizations. This association is theoretically consistent with cognitive models proposing that conceptual schemas guide procedural fluency, while applied practice reinforces those schemas (Clark & Paivio, 1991; Sweller, 2011). However, the correlations were moderate rather than strong, suggesting that conceptual understanding and procedural competence represent related but distinct dimensions of visualization expertise.

Importantly, the findings should be interpreted within the methodological constraints of the study. The research employed a one-group pre–post design without a control group. Consequently, alternative explanations—such as maturation effects, increased familiarity with the software environment, concurrent methodological instruction, or repeated task exposure—cannot be fully excluded. While the instructional design is theoretically consistent with the observed improvements, causal attribution should be approached with appropriate caution. Future research employing randomized or controlled comparative designs would strengthen causal inference and provide more robust evidence regarding the effectiveness of structured visualization instruction.

The study also has contextual limitations. The sample was drawn from a single institutional setting, and performance was assessed immediately following the instructional program. Long-term retention and transfer to independent research contexts were not measured. Future research should examine whether conceptual and procedural gains are sustained over time once formal instructional support is removed and whether they generalize to authentic, discipline-specific research tasks. Experimental comparisons between instructional models could further clarify which design elements most effectively promote durable visualization competence.

Despite these limitations, the study contributes empirical evidence supporting the integration of theory-guided visualization instruction within graduate curricula. This perspective aligns with broader scholarship in doctoral education emphasizing iterative methodological engagement and structured academic practice as foundations for research development (Aitchison & Guerin, 2014; McAlpine, 2013). The combined use of diagnostic assessment, scaffolded graph-construction exercises, and analytic rubrics offers a transparent framework for developing both conceptual and procedural dimensions of graphical literacy. By explicitly linking perceptual theory, statistical reasoning, and applied practice, such instruction may help graduate students transition from software users to reflective producers of scientific visualizations.

In conclusion, visualization competence in advanced education emerges from the interaction between conceptual understanding and procedural competence. The present findings suggest that structured, practice-based instruction can meaningfully enhance both dimensions. Embedding systematic visualization training within graduate research methods courses represents a viable pathway for strengthening methodological rigor and clarity in scientific communication.