Lyniesha
Ward
a,
Fridah
Rotich
a,
Jeffrey R.
Raker
b,
Regis
Komperda
c,
Sachin
Nedungadi
d and
Maia
Popova
*a
aUniversity of North Carolina at Greensboro, Greensboro, North Carolina, USA. E-mail: m_popova@uncg.edu
bUniversity of South Florida, Tampa, Florida, USA
cSan Diego State University, San Diego, California, USA
dUniversity of Nebraska Omaha, Omaha, Nebraska, USA
First published on 4th November 2024
This paper describes the design and evaluation of the rganic chemistry
epresentational
ompetence
ssessment (ORCA). Grounded in Kozma and Russell's representational competence framework, the ORCA measures the learner's ability to interpret, translate, and use six commonly used representations of molecular structure (condensed structures, Lewis structures, skeletal structures, wedge-dash diagrams, Newman projections, and chair conformations). Semi-structured interviews with 38 first-semester organic chemistry learners informed the development of the ORCA items. The ORCA was developed and refined through three pilot administrations involving a total of 3477 first-semester organic chemistry students from multiple institutions. The final version of the ORCA was completed by 1494 students across five institutions. Various analyses provided evidence for the validity and reliability of the data generated by the assessment. Both one-factor and three-factor correlated structures were explored via confirmatory factor analysis. The one-factor model better captured the underlying structure of the data, which suggests that representational competence is better evaluated as a unified construct rather than as distinct, separate skills. The ORCA data reveal that the representational competence skills are interconnected and should consistently be reinforced throughout the organic chemistry course.
Mastering the “language” of chemical representations is not an easy task. Novices must learn about representations (e.g., how to interpret representations) and learn with representations (e.g., use representations to make sense of chemical phenomena) (Talanquer, 2022). Research shows that learners rely on heuristics when making sense of representations (Talanquer, 2014), primarily focus on surface features of representations (Cooper et al., 2010; Popova and Bretz, 2018c), and struggle to connect the external features of representations to the conceptual information embedded in them (Popova and Bretz, 2018a, b; Ward et al., 2022; Rotich et al., 2024). Despite the important role that representational competence plays in learner success in chemistry (Seufert, 2003; Kozma and Russell, 2005; Ainsworth, 2006; Schönborn and Anderson, 2008; Stieff et al., 2016), and the considerable number of investigations into learners’ ability to reason about symbolic and submicroscopic representations (Bodner and Domin, 2000; Cooper et al., 2010; Padalkar and Hegarty, 2012; Padalkar and Hegarty, 2015; Stieff et al., 2016; Stull et al., 2016; Stull and Hegarty, 2016; Miller and Kim, 2017; Lansangan et al., 2018; Popova and Bretz, 2018a, b, c; Connor et al., 2019; Wright and Oliver-Hoyo, 2020; Farheen and Lewis, 2021; Ward et al., 2022; Farheen et al., 2024), a limited number of assessment instruments exist that measure learner competence with representations. Within this manuscript, we describe the development of an assessment to measure aspects of learner representational competence in the context of multiple commonly used representations of molecular structure.
(a) use representations to describe observable chemical phenomena in terms of underlying molecular entities and processes (use),
(b) generate or select a representation and explain why it is appropriate for a particular purpose (generate and select),
(c) use words to interpret features of a particular representation (interpret),
(d) make connections across different related representations by mapping features of one representation onto those of another (translate),
(e) use representations in social situations to support claims, draw inferences, and make predictions about chemical phenomena (use),
(f) describe the limitations and affordances of different representations (identify affordances and limitations), and
(g) take the epistemological position that representations correspond to but are distinct from the phenomena that are observed (take epistemological position).
Most skills outlined above can be developed at a lower foundational level or a higher meta-level that allows for the reflective and purposeful use of representations (diSessa and Sherin, 2000; diSessa, 2004; Gilbert, 2005; Kozma and Russell, 2005; Ward et al., 2022). For the remainder of the article, we will use the succinct labels in parentheses to reference each skill.
RC is grounded in the cognitive theory of multimedia learning and situative theory. Together, these two theories highlight the interplay between cognitive processing and the social context of learning for understanding how students master chemical representations. The cognitive theory of multimedia learning assumes that learning occurs as a result of processing and synthesizing information across the audio and visual modes of instructional information (Mayer, 2002; Kozma and Russell, 2005). In particular, the ability to effectively process visual information is critical when it comes to learning chemical representations, as these often involve intricate visual details that convey essential information about sub-microscopic particles and processes. Situative theory suggests that the physical and social characteristics of a setting shape the processes of understanding and learning within that setting (Lave, 1991; Lave and Wenger, 1991; Kozma and Russell, 2005). As individuals become integrated into a community, they progressively develop in using its representational systems to construct new knowledge and communicate information. This integration and participation in communal practices enable learners to understand and manipulate the representations used by that community, facilitating deeper learning and expertise in the subject matter.
Several assumptions underlie the development of RC. First, a developmental trajectory is assumed, implying that learners' ability to use chemistry representations increases as they progress in their learning. With continued learning, representations become a valuable tool for constructing and communicating understanding, resulting in the sophistication of RC over time (Kozma and Russell, 2005). Second, levels of RC are not assumed to be consistent for different types of representations. For example, a learner may be proficient with most of the RC skills for a particular representation (such as Lewis structures) and only be able to interpret the features of another (such as reaction coordinate diagrams) (Kohl and Finkelstein, 2006; Stieff et al., 2011; Halverson and Friedrichsen, 2013). Finally, RC and conceptual understanding are separate but related components of learner knowledge (Hinze et al., 2013; Maroo and Johnson, 2018). Proficiency in understanding chemical concepts is achieved in tandem with proficiency in utilizing the visualizations that represent and explain those concepts (diSessa, 2004; Kozma and Russell, 2005).
Studies investigating learner RC in chemistry have primarily relied on the use of interviews. Interviews provide nuanced and rich data on how learners engage and reason with representations that result in transferable conclusions. Many of these studies suggested a relationship between RC and conceptual understanding (Lansangan et al., 2007; Höst et al., 2012; Hiong and Daniel, 2015; Pande et al., 2015; Lansangan et al., 2018; Pande and Chandrasekharan, 2022). However, interview data do not allow for drawing robust, generalizable conclusions for various groups of learners, for example, those experiencing different learning environments (Herrington and Daubenmire, 2014). Studies that use assessments that can be easily administered and scored on a large scale have the potential to improve our understanding of RC as a construct and how to effectively develop RC through instruction. This need was highlighted by Kozma and Russell, who noted that “new assessments must be designed and used that measure investigation practices and related skills, such as visualization skills or representational competence” (Kozma and Russell, 2005, p. 142).
RC (Sim and Daniel, 2014; Connor et al., 2021), or frameworks related to RC (e.g., Visualization Competence of Matter (Chang and Tzeng, 2017) and Visual Model Comprehension (Dickmann et al., 2019)) have been the foundation of some assessments developed to elicit secondary or postsecondary chemistry learners’ ability to make sense of representations. These assessments have corroborated findings from interviews stating that representational abilities are greater for learners who received more chemistry instruction (Chang and Tzeng, 2017; Vlacholia et al., 2017) or who have a higher conceptual understanding (Sim and Daniel, 2014; Dickmann et al., 2019). In addition, a few studies found that competence varies by skill or topic (Chang and Tzeng, 2017; Chang, 2018). However, these assessments are either not used to investigate learner RC explicitly (Chang and Tzeng, 2017; Vlacholia et al., 2017; Wang et al., 2017; Chang, 2018; Dickmann et al., 2019), are focused on a single aspect of RC (Connor et al., 2021), or a limited number of representations (Sim and Daniel, 2014; Connor et al., 2021). Some of these assessments have useful but limited evidence for the validity of resulting data, primarily in the form of content validity using expert panels (Chang and Tzeng, 2017; Chang, 2018). This contrasts with more robust evaluations of validity and reliability, including content validity as well as a broader array of psychometric evidence. To advance our understanding of learner RC, quality assessments that demonstrate evidence of valid and reliable data are needed that comprehensively capture multiple RC skills in the context of multiple representations (Kozma and Russell, 2005).
The ORCA is designed as a multiple-choice assessment, permitting testing with a large number of participants and allowing for quick grading. The assessment captures three RC skills: the ability to (a) use words to interpret features of a particular representation – the interpret skill; (b) make connections across different related representations by mapping features of one representation onto those of another – the translate skill; and (c) use representations to draw inferences – the use skill. These skills are among the most commonly taught by organic chemistry educators (Linenberger and Holme, 2015; Popova and Jones, 2021; Jones et al., 2022) and in organic chemistry textbooks (Gurung et al., 2022). The ORCA does not assess the learner's ability to generate representations, another skill typically taught in the target course, as it is difficult to evaluate this skill properly using multiple-choice questions. Other RC skills were not considered for the ORCA as current instruction offers little to no support for those skills (Linenberger and Holme, 2015; Popova and Jones, 2021; Gurung et al., 2022; Jones et al., 2022). Two research questions guide this study:
1. What evidence exists for the validity and reliability of the data generated from the Organic chemistry Representational Competence Assessment (ORCA)?
2. What do the data collected with ORCA reveal about learner representational competence?
Method | Outcome | Application |
---|---|---|
Stage I – assessment development | ||
Semi-structured think-aloud interviews | Elicit learner ability to reason with representations | Generate items and distractors from participant responses to establish validity evidence based on test content |
Expert evaluation | Obtain feedback from experts about the appropriateness of the items for the measurement of the target construct | Establish validity evidence based on test content and revise items accordingly |
Stage II – pilot administrations | ||
Item analysis | Understand item difficulty, item discrimination, and how well distractors function | Revise items accordingly to ensure item quality |
Factor analysis | Investigate the number of factors, the variables that load onto each factor, and the level of correlation among factors | Establish internal structure validity and revise instrument accordingly |
Cognitive interviews | Obtain insight into respondents’ thought processes and feedback about items | Establish response process validity and revise items accordingly |
Expert evaluation | Obtain additional feedback from experts about the appropriateness of the items for the measurement of the target construct | Establish validity evidence based on test content and revise items accordingly |
Stage III – final administration | ||
Item analysis | Understand item difficulty, item discrimination, and how well distractors function | Ensure item quality |
Factor analysis | Confirm the factor structure of the revised ORCA instrument | Establish internal structure validity |
Reliability coefficient | Understand correlations between items | Establish internal consistency/reliability |
We iteratively developed the ORCA by using a sequential, mixed-methods exploratory design (Towns, 2008) and a bottom-up approach that originated from analyses of learner reasoning about representations of molecular structure collected using semi-structured interviews (Bowen, 1994). At the time of their participation, all participants were in the first semester of a year-long organic chemistry course, having completed the two-semester general chemistry sequence. We wrote assessment items from the interview data collected at one institution (Stage I), and we refined these items through three pilot administrations with participants at six institutions (Stage II) using the Qualtrics platform. The final ORCA version was administered to learners at five institutions (Stage III) via Qualtrics. Only learners who completed the entire assessment received extra credit for their participation, which is why all participants completed ORCA in its entirety, and we did not have any missing data.
The interviews were designed to elicit learners’ RC skills. To capture how learners interpret a representation, participants were asked to (a) describe what the representation communicates, (b) decode each diagrammatic feature of the representation (e.g., dashes, wedges, lines, symbols, etc.), (c) explain the purpose of the representation, and (d) identify the atomic composition for the given representation. To elicit how learners use a representation, participants were asked to examine a given representation and make inferences about (a) chemical and physical properties, (b) bonding, and (c) energy and stability. To capture how learners translate between various representations of molecular structure, the participants were given a specific representation and asked to choose the corresponding structure(s) from four options. They were informed that each task may have more than one corresponding structure. After selecting the corresponding structure(s), the participants were asked to explain why they believed the structure(s) corresponded and why the other options did not.
The data from the interviews were transcribed verbatim, inductively coded using ATLAS.ti software (version 9, https://atlasti.com/), and analyzed using constant comparative analysis (Glaser, 1965). The first and the second author coded the data and, depending on a code, obtained 84–100% agreement (Saldaña, 2013). Furthermore, they discussed each case of disagreement in their coding until a 100% negotiated agreement was reached. In addition, the first, second, and sixth authors met weekly to discuss the codes and analyze the data for patterns. Some of the qualitative results from these analyses have been published elsewhere (Ward et al., 2022; Rotich et al., 2024).
We then used these data to generate over 100 assessment items to capture how learners interpret, translate, and use six representations of molecular structure. Fig. 1 depicts how we used participant interview data (Fig. 1A) to write an item that asks participants to use Newman projections to make inferences about stability (Fig. 1B). Response choices, including distractors, were developed using participants’ ideas from the interview and reflected the most common patterns in participant thinking. We reviewed all the items and eliminated items if: (a) there were too many items about the same representation or skill, (b) items were complex multiple-choice questions (Albanese, 1993; Towns, 2014), or (c) items were not representative of the skills learners typically learn in the course (i.e., translating between a chair conformation and a condensed structure). This process resulted in condensing the assessment to 64 items.
An expert panel of six organic chemistry instructors and chemistry assessment developers evaluated all 64 items for content validity. We used the feedback from the expert panel members (Fig. 1C) to eliminate some items and refine other items (Fig. 1D). This process resulted in 48 items for which the expert panel established that the items adequately covered the relevant content to measure the target constructs (interpret, translate between, and use representations of molecular structure) and were appropriate for the target population (organic chemistry learners). In addition, they provided feedback about terminology and the scoring of items.
• What does this question ask you to do?
• What information do you believe is relevant for responding to this question?
• Generate a representation that corresponds to this structure (translate items only).
• How did you decide to pick 〈response option selected〉?
• Did you experience any challenges when providing your response to this question?
During the interview, participants first addressed poor-performing items from the assessment and then the six newly written items. We used responses from the interviews to refine the item stems, modify distractors, and add or delete items. For some items, participants pointed out key features of the representations that cued them to eliminate a distractor or made the correct answer obvious. Based on participants’ feedback, we made adjustments to these items, deleted two items where participants described confusing terminology, and added four of the six newly generated use items to the assessment.
Next, we identified any potential data outliers. Based on the recommendations of Aguinis and colleagues (2013), we used single (i.e., box plots and percentage analysis) and multiple construct techniques (i.e., scatterplots, Mahalanobis distance, and studentized deleted residuals). Fourteen outliers (i.e., complete response sets for a given participant) were removed through these analyses. Outlier data had interpret item scores of zero and often had extremely high (e.g., 340 minutes) or low (e.g., 94 seconds) completion times. This process resulted in a final sample of 1494 responses for analysis.
We conducted CFA using MPlus 8 (Muthén and Muthén, 2017) to investigate the internal structure of the assessment instrument. With the RC framework (specifically, the interpret, translate, and use skills) as a theoretical basis, we tested two hypothesized models to evaluate and confirm the number and nature of latent factors. The Means and Variance Adjusted Weighted Least Squares (WLSMV) method was best suited for the analysis (Finney and DiStefano, 2013) as all measured variables were categorical (binary). A good model fit for these categorical data is more stringent than the criteria suggested by Hu and Bentler (Hu and Bentler, 1999; Beauducel and Herzberg, 2006; Xia and Yang, 2019; McNeish and Wolf, 2023). Though the model fit cutoffs are not absolute and the conditions for their use are still undergoing research (McNeish et al., 2018; McNeish and Wolf, 2023), we use criteria adopted by Komperda and colleagues for good fit (i.e., Comparative Fit Index (CFI) > 0.95 and the Root Mean Square Error of Approximation (RMSEA) < 0.05). The proposed models were congeneric, which is why McDonald's omega (ω) was calculated as a reliability indicator to examine the internal consistency of the items within each latent factor (Hancock and An, 2023; Komperda et al., 2018).
Model | Fit statistics | McDonald's omega (ω) | ||||||
---|---|---|---|---|---|---|---|---|
χ 2 | df | p | RMSEA (≤0.05) | CFI (≥0.95) | Interpret | Translate | Use | |
A – 3 factor | 694 | 296 | <0.0001 | 0.030 | 0.955 | 0.807 | 0.843 | 0.554 |
B – 1 factor | 741 | 299 | <0.0001 | 0.031 | 0.950 | 0.897 |
The data meet the fit criteria for Model A. However, internal consistency for the factors, as measured through McDonalds’ omega reliability coefficient, varied by skill. The translate and interpret factors have a higher reliability, indicating that common constructs (i.e., the ability to translate between representations and the ability to interpret features of an individual representation, respectively) explain a larger amount of the observed variance to the total variance (Komperda et al., 2018). The interpret skill and translate skill items require learners to reason by attending to the diagrammatic features within single or multiple representations (e.g., interpreting dashes and wedges in wedge-dash diagrams, mapping dashes and wedges onto the appropriate axial and equatorial substituents in chair conformations). For this reason, the interpret skill and translate skill factors are highly correlated (Fig. 3A), as students reason about the features of the representations (i.e., composition, connectivity, or spatial information, if applicable) to interpret and translate between representations.
In contrast, the use skill factor has a lower McDonalds’ omega, so a common construct (i.e., the ability to use representations to make inferences) explains less of the observed variance. This is not surprising because the use skill items require participants to not only reason by attending to the features of representations but also by extracting the relevant domain-specific conceptual knowledge embedded in the representation. Moreover, this conceptual knowledge varies by representation (e.g., evaluating the arrangement of substituents in Newman projections to make an inference about stability; evaluating the functional groups in skeletal structures to make an inference about physical properties). In this case, not only do learners need to be able to make sense of the features of a representation, but they also need to have a conceptual understanding of stability or physical properties, and know which relevant features of the representation to attend to make inferences about these concepts. Therefore, a lower reliability for the use skill factor is hypothetically expected as it may not be a homogenous construct (Taber, 2018); however, this lower reliability also limits the inferences that can be made about the use skill from the items within this assessment. Moreover, the factor loadings (Fig. 3A) for the interpret and translate factors are higher than those in the use factor, further indicating that the interpret and translate skill factors explain more variance within the items (Bandalos, 2018).
Komperda and colleagues (2018) suggest that when “an instrument is known to be composed of multiple scales where scores will be reported separately, each scale should be evaluated to determine if it fits a single-factor model, and a reliability value should be provided for each scale of data.” Data for each factor of Model A were evaluated with individual latent factor models; these CFA results provide support that interpret and translate scales fit the data and have factor loadings >0.3 (see Fig. S1 and Table S1 in the Appendix, ESI†). However, while the RMSEA is appropriate, the CFI for the use skill is below the more stringent cut-off for categorical data. Given that the use scale has a weaker model fit, these analyses provide evidence that only the interpret and translate scales can function independently. In summary, although Model A has a solid theoretical basis, aligned with the stance that RC is a “set of skills and practices that allow a person to reflectively use a variety of representations” (Kozma and Russell, 2005, p. 131), and the data meet the thresholds for the fit statistics, the use scale has items that have lower factor loadings, weaker internal consistency, and, unlike the interpret and translate scales, cannot be used independently.
In comparison, Model B is a single-factor model where every item loads onto a single latent factor—RC. The data fit Model B (Fig. 3B), suggesting that this model effectively captures the underlying structure of the data. Model B accounts for the high latent correlation between the interpret and translate skill factors (evidenced as 0.934 in Fig. 3A). While it is possible that the ORCA may be limited in distinguishing these factors, it is more likely that these skills are very related and should be considered one factor, as reflected in Model B. Additionally, the internal consistency (as measured by McDonald's omega) is higher for the one-factor model B than the three-factor model A. While the increased internal consistency may be attributed to the larger number of items within a single factor (Malkewitz et al., 2023), the empirical data better align with the single-factor structure. This suggests that in the context of representations of molecular structure, these RC skills are better evaluated as a unified construct rather than as distinct, separate skills. Below, we delve deeper into the implications of these findings and share insights into measuring student RC.
Two examples from our item-level analysis of participant performance further illustrate the interconnectedness of RC skills. Specifically, we observed instances of students having the necessary conceptual understanding to use a representation to make an inference but not being able to connect this knowledge to the appropriate features of representations, which relates to the interpret skill.
Item U15, shown in Fig. 4A (the same item as in Fig. 1D), required participants to make inferences about the stability of Newman projections and justify their reasoning by relying on the concepts embedded in the diagrammatic features of the representation. With the answer choices in this item, participants could have relied on the concepts of substituent size or electronegativity. Fig. 4B demonstrates that the likelihood of selecting the correct answer C is lower for bottom ORCA performers (bottom 27%) and higher for top performers (top 27%). Almost a quarter of participants (23%, Fig. 4C) selected response choice A, in which they chose the wrong structure while reasoning about the appropriate concept of substituent size. These participants knew the productive conceptual information to make inferences about stability but could not connect that knowledge to the proper diagrammatic features of the Newman projection. Finally, almost a quarter of the participants relied on the wrong conceptual information (electronegativity) and selected responses D (13%) or B (11%). This quarter of participants comprise less than 9% of the top performers but over 45% of the bottom performers. For this item, the bottom performers struggled to identify the productive concept of interest.
Item U14 (Fig. 5A) demonstrated a similar pattern. The correct answer, B, was chosen by 41% of all participants. A slight majority of bottom performers (55%) relied on the productive concept reflected in choices B and D. These participants were able to identify the appropriate feature (charge) and connect it to the appropriate concept (octet). However, half of them could not connect that knowledge to the appropriate diagrammatic features to make accurate inferences about stability. Thus, they incorrectly selected the structure with an atom with an incomplete octet (D). Almost a third (30%) of all participants selected response choice D. Finally, nearly a third of the participants chose responses A (20%) or C (9%). These participants relied on features like the presence or absence of double bonds that were not as relevant to the concept of interest (stability) in this prompt.
These examples highlight the complex interplay of RC skills required to answer such questions. Successfully using representations to make inferences about a concept demands a combination of knowledge and ability to (1) interpret diagrammatic features of a representation, (2) conceptually understand the concept of interest, (3) distinguish the relevant from irrelevant features of the representation in relation to the concept of interest, and (4) connect the relevant features of the representation to one's conceptual understanding to make an inference about the concept of interest. Each of these steps is necessary when making inferences about representations.
Item T9 is another item that some may assume is easy but had relatively surprising responses. The item requires learners to translate between a cyclic skeletal structure and a condensed structure with two of the four answer choices in which the ring is open (Fig. 7A). Most participants selected the correct answer, but over 15% selected a non-cyclic condensed structure (Fig. 7C, options A and B). Participants who select these incorrect answer options for a translate task do not adequately understand the representations or the chemical phenomena represented, even after a whole semester of organic chemistry. As shown, developing RC should be intentional and repeatedly reinforced throughout the curricula.
Second, the ORCA assesses learner RC in the context of multiple representations of molecular structure but is not exhaustive (e.g., representations such as electrostatic potential maps and ball-and-stick models are not included even though they have unique affordances to students (Farheen et al., 2024)); we selected symbolic representations only to feature representations with similar dimensions and features (Talanquer, 2022). More assessments are needed to capture learner RC with other representations in chemistry.
Third, the use items capture learners’ ability to make inferences about stability and boiling point only, even though there are other properties that the six target representations could convey. This narrower focus connects to what is covered in instruction and ensures the assessment can be administered in a reasonable timeframe (about 20 minutes). There is potential to consider additional items that focus on other concepts (e.g., aromaticity, acidity).
Fourth, the participants who completed the final assessment also took the assessment during the pilot III administration. There were at least six weeks between assessment administrations (depending on the institution), but this means there is a possibility of practice effects (i.e., improved performance due to familiarity with test items) (American Educational Research Association et al., 2014). Currently, there is no benchmark for when practice effects diminish. Other studies have as low as one month between test administrations (Bretz and Linenberger, 2012).
Lastly, the assessment was administered only online; however, it can be easily adapted to the classroom setting for paper-and pencil administration. Administering the assessment online was a purposeful choice; this allowed for easy access to learners and a quick collection of data. Assessments given in different conditions (such as virtual versus in class), with differing constraints (e.g., not being able to sketch notes), and scored for correctness versus completion (including for bonus points) all likely have varying but overall consistent data (Harle and Towns, 2011).
Two models were proposed to evaluate the internal structure validity of the data generated: a three-factor Model A and a one-factor Model B. While Model A has a solid theoretical basis, our data showed a better fit with Model B. This single-factor model consolidates the skills into a unified latent RC construct, accounting for the high correlation between the interpret and translate skill factors in Model A. This model can be readily used to assess organic chemistry learner RC in the context of representations of molecular structure. Our study is an important step toward advancing and reconceptualizing our understanding of RC as a network of interconnected skills.
For example, ORCA can be used as a diagnostic tool early in the course to identify areas where students may need additional support, such as with more complex skills like using representations. Based on students' performance, instructors can implement progressive scaffolding, offering explicit guidance early in the semester on how to use different representations. As students build confidence, the level of support can be gradually reduced to encourage independent use of representations.
At the same time, instructors should not assume that it is sufficient to discuss a given representation once, early in the semester. Our findings show that even by the end of the semester, many students struggled with basic tasks, such as identifying implicit hydrogen atoms in wedge-dash diagrams or distinguishing between cyclic and acyclic molecules. It is, therefore, imperative that instructors periodically incorporate explicit explanations of the features of representations that they use in their instruction without assuming that students mastered these representations at the beginning of the course.
Additionally, it's crucial for students to articulate their reasoning, especially when tackling more complex tasks that require using representations to make inferences. Our item analysis revealed that for students to effectively use molecular representations to make inferences, they must be able to do four essential steps: (1) interpret the diagrammatic features, (2) understand the underlying concept, (3) identify relevant versus irrelevant features, and (4) connect those relevant features to their conceptual understanding to draw inferences. Without eliciting students’ reasoning, an instructor or researcher will not be able to understand which of these steps present the challenge for students when using representations to make inferences. This complexity should be taken into consideration when designing assessment tasks or interview prompts.
The ORCA can also be used or adapted as an assessment instrument in studies that aim to advance our understanding of RC as a construct or how to best support learners in developing RC. Prior to using the ORCA, it is imperative that researchers evaluate their data for evidence of validity and reliability (American Educational Research Association et al., 2014; Lazenby et al., 2023). For example, the ORCA can provide evidence of how instructional practices impact learner RC and can be used in longitudinal studies to show how learner RC changes throughout instruction (Kozma et al., 2000). These investigations could help elucidate a learning progression for the development of RC or identify an optimal organic chemistry curriculum for developing RC with representations of molecular structure. While there are existing interventions that support individual RC skills in organic chemistry (Stieff et al., 2016; Stull et al., 2016; Stull and Hegarty, 2016), more research needs to be done to develop interventions that support RC more broadly. In this way, the ORCA can be used as a pre-post measure to provide empirical evidence of approaches that may be useful to support organic chemistry learners.
Future research could also use the ORCA data to conduct Structural Equation Modelling (SEM) to evaluate the relationship between learner RC and other relevant constructs such as their visuospatial ability, conceptual understanding, or success in the organic chemistry course (Nitz et al., 2012; Sim and Daniel, 2014; Stieff et al., 2014, 2018; Stieff and DeSutter, 2020).
While this work focuses on symbolic representations of molecular structure, organic chemistry and other chemistry disciplines also rely on a variety of other representations (e.g., ball-and-stick models, spectra, reaction coordinate diagrams, phase diagrams, and molecular orbital diagrams) that require RC. The process we used to develop the ORCA can be adapted to assess RC skills in these other contexts. Expanding this approach across different contexts will deepen our understanding of how learners develop RC. We encourage the community to build on this work to refine our collective understanding of RC and explore the most effective strategies for supporting learners in developing these important skills.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3rp00188a |
This journal is © The Royal Society of Chemistry 2025 |