You are on page 1of 14

This is an open access article published under an ACS AuthorChoice License, which permits

copying and redistribution of the article or any adaptations for non-commercial purposes.

Article

Cite This: J. Chem. Educ. XXXX, XXX, XXX−XXX pubs.acs.org/jchemeduc

Arguing from Spectroscopic Evidence


Ryan L. Stowe*,† and Melanie M. Cooper
Department of Chemistry, Michigan State University, 578 South Shaw Lane, East Lansing, Michigan 48824, United States
*
S Supporting Information

ABSTRACT: Constructing and critiquing evidence-based


claims is centrally important to aspiring medical professionals
and to scientists more generally. Accordingly, The National
Academy of Science’s Framework for K−12 Science Education
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

describes engaging in argument from evidence as one of the


practices that characterize work in science. However, despite
the central role argumentation plays in construction and
refinement of evidence-based explanations and models, it is
Downloaded via 179.32.107.65 on September 4, 2019 at 02:00:28 (UTC).

very often absent from K−16 science learning environments.


Here, we frame argumentation from spectroscopic evidence in
terms of flexible application of a series of procedures to pull
information from spectroscopic traces and use this evidence to
inform claims as to the structure of an unknown molecule.
Through analysis of responses to several multipart assessment
items, we examined how students analyzed, interpreted and used spectroscopic evidence to inform their claims. We found that
students were fairly adept at analyzing and interpreting data from infrared and 13C NMR traces as well as indicating
correspondence between proton environments in their structural prediction and appropriate 1H NMR peaks. Unfortunately,
none of these tasks were significantly associated with student success in proposing a claim consistent with an entire corpus of
spectroscopic data. Further, scaffolding of the task prompt had no impact on student ability to successfully construct evidence-
based structural predictions. Our findings indicate that students will require significant support to use their procedural
knowledge flexibly in order to iteratively construct and critique claims supported by spectroscopic evidence.
KEYWORDS: Second-Year Undergraduate, Organic Chemistry, Problem Solving/Decision Making, Spectroscopy,
Chemical Education Research
FEATURE: Chemical Education Research

■ INTRODUCTION
A significant portion of students enrolled in introductory organic
supported by sound interpretation of evidence.4 The process of
pulling information from various strands of data, itself the SEP
chemistry aspire to careers in the healthcare sector. Accordingly, “analyze and interpret data”, to gather evidence that supports or
much of the rhetoric justifying organic chemistry as a medical refutes a diagnosis is heavily reliant upon the use of procedural
school and dental school prerequisite focuses on analogies knowledge (i.e., the “rules of the game” for solving a given
between competencies desired in medical professionals and problem). For example, interpretation of an electrocardiogram
those alleged to be fostered by the course.1,2 We have previously (EKG) requires awareness of the parameters of a normal rhythm
argued that these competencies can be precisely defined and (including the normal amplitude, deflection, and duration of
operationalized as the Science and Engineering Practices (SEPs) each component) and that one must make comparisons between
defined by A Framework for K−12 Science Education (the these ideal parameters and a patient’s observed EKG trace.
Framework).3 The Practices describe processes characteristic of There may be a number of procedures that might characterize
work in science and encompass both ways of thinking inherent these comparisons, with some being more or less efficient for a
in experimental design (e.g., “design and carry out inves-
given scenario. Flexible command of many such approaches
tigations” and “analyze and interpret data”) and ways of
communicating and refining scientific ideas in communities coupled with knowledge of when a given procedure is most
(e.g., “engage in argument from evidence” and “evaluate and appropriate has been characterized as “deep procedural
communicate information”). knowledge” by researchers in mathematics education.5−7 Such
Diagnosis, which is the construction of a claim as to the cause knowledge is essential to efficient and accurate interpretation
of a collection of symptoms, is centrally important to nearly all
healthcare practitioners. Diagnosis may be regarded as the SEP Received: June 12, 2019
“engaging in argument from evidence”, as it requires medical Revised: July 19, 2019
professionals to construct tentative causal accounts of symptoms
© XXXX American Chemical Society and
Division of Chemical Education, Inc. A DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

and use of evidence and therefore a vital component of the “conceptual” mean with regard to knowledge.23 At present, there
overall practice of diagnosis. is significant consensus that mathematics education should
The dialogic process of construction and critique of claims minimize rote reproduction of algorithms. In pursuit of de-
that characterizes diagnosis is analogous to student engagement emphasizing the “plug and chug” focus of traditional
in argumentation in the context of spectroscopy. That is, when mathematics courses, there have been several high profile
asked to determine the identity of an unknown from analysis of initiatives that advocate teaching the conceptual basis for a
spectroscopic data, students must pull evidence from various particular procedure before the procedure itself.24−26 We will
spectra, construct a claim informed by this evidence, and analyze focus here on what is meant by “procedural knowledge” as well
whether that tentative claim is in fact consistent with all of the as whether such knowledge can be anything other than rote
evidence assembled. Students do not begin this process by application of memorized rules.
“knowing the answer”, unless the problem is an exact Hiebert and Lefevre defined procedural knowledge as “chains
reproduction of one discussed previously, and so their relation of prescriptions for manipulating symbols” in the mid-1980s.
of evidence to claim is meant to persuade the instructor or their They painted a contrast between procedural and conceptual
peers of the reasonableness of the claim itself. Importantly, the knowledge, which was said to be “rich in relationships. It can be
structural hallmarks of argumentation (i.e., construction and thought of as a connected web of knowledge.”27 Star has taken
critique of evidence-based claims) may occur via dialogue issue with Hiebert and Lefevre’s characterization of procedural
among members of a group or internally, via an individual using and conceptual knowledge in mathematics, claiming that they
different mental “voices” for claim building and vetting.8 Here, have conflated knowledge type (procedural−conceptual) and
we are focused on tasks designed to require argumentative quality (novice−expert).6,7 He has noted that flexible, context-
discourse in the minds of individual students. We are aware of no sensitive application of procedures represents a deep (that is,
existing peer-reviewed literature that examines spectroscopic expert-like) understanding of both the procedures themselves
analysis through the lens of argumentation. Prior reports instead and what constitutes an appropriate procedure for a given
focus on strategies believed to foster “problem solving scenario. It should be noted that procedure “appropriateness” is
skills”,9−14 describing the differences between expert and novice not always defined in a clear-cut manner. Star has defined the
analyses of spectroscopic data,15−18 or characterizing invalid most appropriate use of procedures as that which is most
chemical assumptions and heuristics that are believe to constrain efficient (i.e., requires the fewest steps to reach a solution).28
student reasoning during IR and 1H NMR spectroscopic This definition of “appropriateness” does not map well onto the
interpretation.19 We can find no published strategies that use of procedural knowledge to construct and critique claims
contain strong evidence that a particular intervention improves informed by spectroscopic data, where there is no defined “most
student ability to argue from spectroscopic evidence; that is, step efficient solution”. Context sensitivity, that is, recognition of
extant literature does not compare the success of large student when information from a particular data source would be helpful
cohorts engaged in the described intervention with those not so in refining a claim, is more germane to discerning when a
engaged. procedure would be most appropriate for the data analysis that
Despite the significant role argumentation plays in refining informs argumentation from spectroscopic evidence.
knowledge claims within the scientific community,20 it is very Here, we describe application of a chain of procedures to solve
often absent from instruction in both K−12 and higher a problem as activation of a series of intellectual resources. Such
education spaces.21 Our work here is an attempt to characterize “resources” were first proposed by Hammer as an extension of
student engagement in argumentation from spectroscopic diSessa’s “knowledge in pieces” approach to conceptual
evidence in the context of scaffolded assessment items.22 In change.29,30 These resources represent small-grain knowledge
particular, we are interested in the following questions: elements of a variety of types, including ideas extrapolated from
1. How successful are students at deploying procedural experience, notions about the nature and appropriate use of
resources to analyze and interpret data from a variety of knowledge, and knowledge of algorithms and heuristics.31−34
spectroscopic sources? They are not “correct” or “incorrect” in and of themselves, but
2. What associations exist between student analysis and they may be activated in ways that are not appropriate for a
interpretation of spectroscopic data and their use of this particular scenario. For example, the resource that “more effort
evidence to inform claims as to the identity of an begets more result” has been activated in all manner of
unknown? idiosyncratic, incorrect student explanations for differences in
3. To what degree do analysis, interpretation, and use of boiling point.35 As students gain facility with patterns of resource
different strands of spectroscopic data predict student activation, “locally coherent sets of resources may... become
success at proposing a structure consistent with all established as resources in their own right.”34 Thus, the term
assembled evidence? “resource” alone does not necessarily imply a fixed grain size;
4. How does varying the structure of the prompt impact some resources are smaller grain than others.
student success in constructing a claim consistent with Analysis and Interpretation of Spectroscopic Data as
provided spectroscopic evidence? Activation of Resources

■ ON PROCEDURAL KNOWLEDGE
Most work defining and exemplifying various aspects of
To analyze and interpret data from the spectroscopic traces
commonly examined in organic chemistry (e.g., infrared, 13C
NMR, and 1H NMR), one must be familiar with a range of
procedural knowledge derives from the mathematics education algorithms as well as when those algorithms could be
literature. Indeed, there has been vigorous discussion for productively used. Our use of “algorithm” here is meant to
decades among mathematics education scholars about appro- indicate that, if one follows a set of well-behaved rules, one can
priate learning environment emphasis on “procedural” versus successfully pull relevant information from a spectrum. In other
“conceptual” knowledge, as well as what “procedural” and words, analysis and interpretation of spectroscopic data requires
B DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

procedural knowledge. In characterizing data analysis and in the context of a particular problem. For example, if one does
interpretation as “procedural”, we are assuming that the spectra not know the “n + 1” rule, then they will be unable to say
analyzed are fairly unambiguous and that more advanced anything about the number of protons on carbons adjacent to a
conceptual knowledge is not required for interpretation, as given proton environment. However, knowledge of this “rule”
would be the case if a spectrum were to show, for instance, long- does not guarantee that students will consistently and
range proton coupling. All of the spectroscopic traces examined methodically apply it when critiquing structural predictions.
in the course context for this study were unambiguous. Each Cartrette and Bodner characterized individuals that were “more
procedure is likely learned as activation of a series locally successful” in relating structural claims to spectroscopic
coherent resources that, over time, become compiled into evidence as having a methodical strategy that involved effective
resources themselves. For the purposes of the analyses that application of algorithms such as the “n + 1” rule to depict
follow, we define several algorithms elementary to spectroscopic fragments indicated by portions of a spectroscopic trace en route
data analysis. These include to proposing a complete structural prediction.15 “More
• Determining the number of carbon environments by successful” problem solvers also, without exception, carefully
counting the peaks in a 13C NMR spectrum checked consistency between their prediction and assembled
• Recognizing the presence or absence of functional groups spectroscopic evidence by again employing a variety of
(e.g., hydroxyl and carbonyl functionality) in an infrared procedural resources. Here, we will examine student success at
spectrum relating evidence pulled from spectroscopic data to structural
• Determining the number of protons on carbons adjacent predictions in the context of formative and summative
to a given environment from the splitting of a particular assessment prompts.


1
H NMR peak
• Determining the number of protons in a given environ- METHODS
ment from the integration of a particular 1H NMR peak Course Context
Each of these procedures are almost certainly composed of Our work occurred in the context of a transformed, large-
smaller-grain knowledge elements for many students. Indeed, enrollment organic chemistry course known as Organic
helping students compile the locally coherent resources that Chemistry, Life, the Universe and Everything (OCLUE).22
make up each algorithm into a larger-grain resource that can be This a two-semester course and has enrolled between 350 and
deployed with little mental effort is a central goal of our 700 students since the fall of 2016. Students meet as a large
treatment of spectroscopy. However, a finely grained analysis of group for approximately 3 h/week and attend a weekly 50 min
the character of students’ procedural knowledge is beyond the recitation section, where they work in groups on tasks that
scope of this study. We cannot know, from written responses to require the use of SEPs under the guidance of a graduate
prompts, whether students have compiled the procedures noted teaching assistant. The data presented here derives from
above into resources. It should be mentioned that the OCLUE enactments from fall 2016 to fall 2018. Like the
procedures we have described are by no means exhaustive and transformed general chemistry course Chemistry, Life, the
students almost certainly bring other resources to bear on Universe, and Everything (or CLUE),36 OCLUE is structured
argumentation problems beyond what we noted. For example, around scaffolded progressions of core ideas37 that build in
helpful heuristics such as “a 6H doublet around 1 ppm in a complexity as students predict, explain, and model ever more
proton NMR typically corresponds to two methyl groups in an complex phenomena. These core ideas are not discrete topics
isopropyl functionality” are undoubtedly cultivated among the but rather large-grain ideas that can be taught at various levels of
student populace as they work through many spectroscopy sophistication, have significant power to explain phenomena,
problems. and undergird all topics in the course.3,37 In both CLUE and
The reader may note that we have not made mention of the OCLUE these ideas are “energy”, “atomic/molecular structure
physical basis for spectroscopic techniques when describing the and properties”, “stability and change in chemical systems”, and
procedures needed to analyze and interpret spectroscopic “electrostatic and bonding interactions.” Connections between
evidence; this was intentional. One need not understand the core ideas and phenomena are explicit in instruction in order to
impact of nearby magnetic environments on proton NMR peak help students develop and organize their knowledge as they
splitting to learn that the multiplicity of a particular peak progress.
corresponds to one more than the number of protons on The physical basis for spectroscopic techniques as well as
carbons adjacent to a given proton environment (the “n + 1” analysis of spectroscopic data are introduced early in the first
rule). For the purposes of our analysis, we have only addressed semester of OCLUE and build in complexity throughout the
these four algorithms because (as the reader will see) even such course. Infrared (IR) spectroscopy is introduced first, with
relatively simple resources are difficult for students to emphasis on recognizing peaks that correspond to carbonyl and
coordinate. It is an open question whether possessing hydroxyl functional groups. Carbon nuclear magnetic resonance
conceptual knowledge improves one’s ability to argue from spectroscopy (13C NMR) is discussed following IR spectroscopy
spectroscopic evidence. and before proton NMR, as 13C NMR is less information-rich
Argumentation from Spectroscopic Evidence through a than proton NMR and should therefore impose less cognitive
Resources Lens load on students. Finally, proton NMR is introduced and
The procedures we have characterized above can be used in unpacked. Spectroscopy is introduced early in the first semester
more or less efficient ways to support the construction and to provide students with evidence about molecular structure; the
critique of a variety of claims. As with “deep procedural conceptual basis for various spectroscopic techniques precedes
knowledge” more generally, facility with argumentation from student use of these techniques to make sense of data. For
spectroscopic evidence relies upon students possessing the example, students learn why electron density affects NMR
requisite resources and also being strategic with their application signals, and why carbons with electronegative substituents
C DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

appear downfield prior to analyzing and interpreting 13C NMR. Box 1. 3D-LAP Criteria for Prompts with the Potential to
As new functionality is explored throughout both semesters of Engage Students in Argumentation from Evidence
the course, the spectroscopic signatures of that functionality are
introduced and discussed. At the same time, the emphasis shifts 1. Question gives an event, observation, or phenomenon
to the use of spectroscopic evidence to support claims about 2. Question gives or asks student to make a claim based on a
molecular structure. In this report, we focus on student given event, observation, or phenomenon
engagement in argumentation from spectroscopic evidence, 3. Question asks student to provide evidence in the form of
rather than student understanding of the physical basis for that data or observations to support the claim
evidence. 4. Question asks student to provide reasoning about why
Student Participants the evidence supports the claim
These studies were carried out at a large, midwestern research
university. All student participants were informed of their rights Each prompt type disclosed below will be annotated to denote
as research subjects, and all data was obtained and handled in the criteria they fulfill. Note that not all assessments used fulfill
accordance with the Institutional Review Board. all of the criteria given; these assessments are best categorized as
A total of 300 students participated in this study. All not having the potential to elicit evidence of student engagement
participants were enrolled in both semesters of OCLUE within in argumentation. All infrared spectra used for our prompts were
a single academic year. Our sample is made up of students from derived from the Spectral Database for Organic Compounds.44
three cohorts, with each cohort consisting of 100 randomly All 13C NMR and 1H NMR spectra were created using the NMR
selected students who were enrolled in a two-semester OCLUE prediction tools found in ChemDraw Professional 15.1 for Mac.
enactment; 100 students were enrolled in the course from 2016 Following discussion of our different prompt varieties, we will
to 2017, 100 were enrolled from 2017 to 2018, and 100 were discuss our coding of these assessment tasks with focus on
enrolled from 2018 to 2019. To determine whether significant analysis and interpretation of particular strands of spectroscopic
differences existed in the demographic and academic measures data, the correlation between pulling information from spectra
for our three cohorts, we ran a series of Mann−Whitney U tests and leveraging this information in constructing claims, and
comparing, in a pairwise fashion, cohort ACT scores, grades in overall success at proposing claims consistent with provided
the first semester of OCLUE, and overall GPA prior to enrolling evidence.
in organic chemistry. In cases where students had only taken the Prompt Type 1: Low Structure
SAT, SAT scores were converted to ACT scores using Assessment items of this variety were the least structured of the
concordance tables published by the ACT.38 Apart from the prompts we administered to students. Students were shown
instance mentioned below, no significant differences in the infrared, 1H NMR, and 13C NMR spectra; given a molecular
demographic and academic measures of our cohorts were found. formula; and asked to construct a claim consistent with
The cohort enrolled in OCLUE during the 2017−2018 assembled evidence, as shown in Figure 1. “Low structure”
academic year had somewhat higher first-semester OCLUE
grades than the cohort enrolled during the 2018−2019 academic
year (mean of 3.6 vs 3.4, U = 4023.5, z = −2.655, p = 0.008, r =
0.18; small effect size). All statistical tests reported in this
contribution were conducted using version 24 of SPSS Statistics
for Mac.39 Our threshold for significance in this study was p ≤
0.01. Summaries of all statistical analyses of demographic and
academic measures are reported in Table S1, contained in the
Supporting Information.
Instruments
Our data corpus consisted of responses to four types of prompts
that were designed to engage students in analysis and
interpretation of spectroscopic data and construction of claims
consistent with assembled evidence (i.e., argumentation).
Prompt scaffolding varied in order to discern how students
could and should be supported in arguing from spectroscopic
evidence on written assessments. Assessment items were given
an summative assessments on midterm exams and as formative Figure 1. Unstructured prompt asking students to propose a structure
assessments to be completed on the online homework system consistent with several strands of spectroscopic evidence. This prompt
beSocratic. beSocratic enables students to construct explan- does not have the potential to engage students in argumentation from
ations, predictions, and models in response to open-ended evidence as defined by the 3D-LAP.
questions.40−42 Below we will discuss each prompt variety with
focus on why each was scaffolded in a particular manner. Our
scaffolding strategies were informed by the criteria Laverty et al. prompts presented students with several spectroscopic traces
put forth in the 3-Dimensional Learning Assessment Protocol that were collected by bombarding an unknown compound with
(3D-LAP) for assessments that have the potential to elicit electromagnetic radiation (EMR) of different wavelengths and
evidence of student engagement in science and engineering detecting EMR absorbed or emitted (e.g., “an observation”).
practices (SEPs).43 In particular, we focused on the criteria the Following this, respondents were asked to construct a claim
assessments had to fulfill to potentially elicit evidence of student consistent with assembled spectroscopic data. Prompts of this
engagement in argument from evidence (Box 1). type never explicitly asked students to provide evidence that
D DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

supported their claim or to link evidence to claim by way of


sound reasoning. Accordingly, they do not fulfill the 3D-LAP
criteria for having the potential to elicit evidence of student
engagement in argumentation from spectroscopic evidence
(Box 1). The most that can be concluded from responses to a
low structure prompt is whether students constructed a
consistent claim. These prompts were used as controls to
address RQ4 but could not provide information useful in
addressing RQs 1−3. These prompts were only given as part of
formative assessments.
Prompt Type 2: Moderate Structure
The second type of prompt embodies the minimum amount of
scaffolding we believe is required to meet the criteria laid forth in
the 3D-LAP for having the potential to elicit evidence of
engagement in argumentation. As shown in Figure 2, prompts of

Figure 3. Highly structured prompt designed to elicit evidence of


student engagement in argumentation from spectroscopic evidence.
The prompt spacing is condensed from that given to students. Adapted
with permission from ref 2. Copyright 2017 American Chemical
Society.

evidence they should pull from assembled traces (e.g., functional


groups indicated by the IR and number of carbon environ-
ments). Further, they were asked to indicate correspondence
between peaks in the 1H NMR and their prediction, that is,
directly relate aspects of their claim to the evidence, which we
consider as meeting the “reasoning” criteria of the 3D-LAP
(Figure 3). Prompts of this type are useful for addressing
whether students can analyze and interpret various strands of
spectroscopic data given explicit guidance to do so. “High
Figure 2. Prompt designed to engage students in argumentation from structure” prompts were given on a formative assessment
spectroscopic evidence. This represents an assessment with a moderate administered in the spring of 2018, as well as on midterm exams
amount of scaffolding. given during the springs of 2017 and 2018. The data we will use
to answer RQs 1−3 are derived from responses to these
prompts.
this type provide students with an observation analogous to that
shown in Figure 1. Additionally, students are asked to propose a Prompt Type 4: Critique Centered
claim consistent with the spectroscopic traces before them. The final prompt type we will discuss here was meant to elicit
However, unlike low structure prompts, “moderate structure” evidence of student construction and critique of evidence-based
prompts ask students to justify their prediction by citing specific claims (Figure 4). Accordingly, students were given two sources
strands of evidence; that is, they are being asked to relate of spectroscopic data obtained from an unknown (infrared and
13
evidence to their claim by way of reasoning. There existed C NMR spectra), asked to pull information from each source,
minimal cuing as to what evidence would be appropriate to and asked to construct two claims consistent with assembled
mention nor how that should be related, and so this should still evidence. Following this, students were given an 1H NMR
be considered a fairly unstructured prompt. Moderate structure spectrum corresponding to that same unknown, asked to analyze
prompts were used to examine the impact of prompt structure and interpret an aspect of this spectrum, and subsequently asked
on students’ ability to construct evidence-based claims as part of to rule out one or both of their prior claims. The final part of this
addressing RQ4. question asked for the construction of a claim consistent with all
Prompt Type 3: High Structure data sources. A critique centered prompt was given as a
formative assessment during the fall 2018 semester. A critique-
Our third prompt variety should be considered “highly
centered prompt was used to examine the impact of prompt
structured” relative to the first and second prompt types
scaffolding on students’ success in constructing evidence-based
shown in Figures 2 and 3. As with previously mentioned
claims as part of addressing RQ4.
prompts, students were called to propose a claim consistent with
spectroscopic evidence and relate specific strands of evidence to Coding Protocol
their claim. However, unlike with moderate structure prompts, The descriptors of student responses that formed the basis for
students were explicitly cued to focus on particular pieces of our analyses were binary in nature. That is, students either
E DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

one of the authors, not disagreement about the meaning of a


code. Both authors met following joint coding and reached
consensus as to the proper code each of the 30 responses
merited for each question subpart.
RQ1: How successful are students at deploying procedural
resources to analyze and interpret data from a variety of
spectroscopic sources?
Student ability to deploy the algorithms mentioned earlier to
pull information from different spectroscopic sources was
analyzed using responses to portions of high structure prompts.
Such prompts explicitly asked students to describe the
functional groups indicated by an infrared spectrum and
determine the number of chemically distinct carbons shown
by a 13C NMR spectrum. Assessments structured in this manner
also asked students to indicate which proton environment in
their prediction corresponded to which 1H NMR peak (Figure
3). In order to address RQ1, we noted whether student answers
to appropriate question subparts indicated appropriate or
inappropriate resource deployment. For example, we coded
whether students correctly indicated the number of carbon
environments when asked to do so. In analyzing the use of
algorithms relevant to 1H NMR, it was difficult to deconvolute
the two algorithms mentioned previously (i.e., use of the “n + 1”
rule and peak integration). Accordingly, they were taken
together for the purposes of RQs 1−3. That is, student ability
to indicate reasonable correspondence between a proton
environment in their prediction and a 1H NMR peak was read
to indicate facility with interpreting information from both peak
splitting and integration. Additionally, although student analysis
and interpretation of 13C NMR and IR data was decoupled from
Figure 4. Assessment prompt meant to engage students in construction claim construction in high structure items, students were
and critique of claims grounded in spectroscopic evidence. The prompt required to use their interpretation of 1H NMR data in the
spacing is condensed from that given to students. context of their structural prediction on part (e) of these
assessments. Accordingly, we describe student ability to
reasonably label the proton environments in their prediction
indicated the correct response to part of a prompt or they did as “analysis, interpretation, and use” of proton NMR data. All
not. Accordingly, there was no nuanced interpretation of coding conducted as part of this work was binary; students either
responses to negotiate among raters. As a result of this, we did correctly analyzed and interpreted particular information from
not see need for extensive testing of the reliability of our coding. spectra, or they did not. Importantly, we did not and do not
The first author coded all the data analyzed for this contribution. assume that a correct or incorrect response indicates the best a
To spot-check the first author’s description of student responses, student is capable of; responses merely indicate those resources
the second author coded a random sample of 30 responses to activated in a particular moment in response to a given prompt.
high structure exam items, and their codes were compared
against those assigned by the first author. In particular, both RQ2: What associations exist between student analysis and
authors coded (1) whether students indicated the number of interpretation of spectroscopic data and their use of this
carbon environments shown in the 13C NMR, (2) whether evidence to inform claims as to the identity of an
students noted the functionality shown in the infrared spectrum, unknown?
(3) whether students constructed a structural claim with an Engagement in argumentation from spectroscopic evidence
appropriate number of carbon environments, (4) whether requires that students use evidence from analyzed data to inform
students constructed a structural claim with the functionality the construction and critique of claims. Accordingly, we were
shown in the infrared spectrum, (5) whether students indicated interested in what associations (if any) existed between student
reasonable correlation between one or more peaks in the 1H ability to pull information from spectroscopic data (coded as
NMR spectrum and their claim, and (6) whether students described under RQ2) and incorporation of this information
constructed a claim consistent with all assembled evidence. into their structural prediction. We coded whether the claims
These codes correspond to responses to different parts of the students constructed in response to high structure prompts were
high structure prompts. For example, part A (Figure 3) asks consistent with particular strands of spectral data. For example,
students to describe the functionality indicated by the infrared does the structural prediction have six carbon environments as
spectrum. Responses to part A provided evidence as to whether the 13C NMR suggests? Does the structural prediction have a
students pulled reasonable information from the IR spectrum hydroxyl group as the infrared spectrum would indicate?
regarding functional groups present in a molecule. Agreement Following this, we analyzed association between analysis and
between the authors on the codes appropriate for student interpretation of data and the construction of claims consistent
responses to each prompt portion ranged from 83 to 100%. All with particular data sources via a series of χ2 tests. For results that
discrepancies that arose were the result of an error on the part of showed a significant association, the strength of the relationship
F DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

Table 1. Three Experiments Conducted to Determine the Impact of Prompt Structure on Students’ Construction of Claims
Consistent with Spectroscopic Evidence
Experiment Comparison Prompt Experimental Prompt Purpose
Experiment 1 Low structure High structure Determine the impact of explicitly prompting for data analysis and argument
justification
Experiment 2 Low structure Critique centered Determine the impact of explicitly prompting students to construct and
critique claims
Experiment 3 Moderate structure, two Moderate structure, three Determine the impact of varying the amount of data students had to
spectroscopic traces spectroscopic traces consider when constructing claims

Figure 5. Percentage of students who analyzed, interpreted, or used spectroscopic data from several sources as prompted by highly structured
assessment items. “Exam 1” prompts were present on summative assessments, whereas “HW” prompts represent assessments given as part of a
homework assignment. The portions of multipart assessment items germane to analysis, interpretation, or use of spectroscopic data are found in the
“Prompts” box in the bottom right of the figure.

was calculated using Cramer’s V and interpreted using because of their very explicit prompting for invocation of the
guidelines published by Cohen.45 Accordingly, a small effect algorithms mentioned previously.
would have a Cramer’s V of 0.1, a medium effect would have a RQ4: How does varying prompt structure impact student
Cramer’s V of 0.3, and a large effect would have a Cramer’s V of success in constructing a claim consistent with provided
0.5. spectroscopic evidence?
RQ3: To what degree do analysis, interpretation, and use of Three experiments were conducted to determine whether
different strands of spectroscopic data predict student different prompt structures might impact student success in
success at proposing a structure consistent with all constructing a claim consistent with spectroscopic evidence
assembled evidence? (Table 1). Past studies indicate that the manner in which a
Binary logistic regressions were performed to determine prompt is structured can powerfully affect the resources students
whether student success in proposing claims consistent with bring to bear in their responses.46 If students do not know what
spectroscopic evidence could be predicted in part by their we are asking, they might not construct the best explanation,
success analyzing and interpreting particular strands of prediction, or model they are capable of generating. Accordingly,
we hypothesized that differently structured spectroscopic
spectroscopic data. Models were constructed for high structure
argumentation prompts might impact student success in
prompts with the outcome variable being “correct prediction” in
proposing appropriate evidence-based claims.
all instances and the predictor variables being “indicate the Two of the three experiments conducted to address our first
correct number of carbon environments shown by the 13C research question focused on determining whether explicitly
NMR”, “indicate the correct functionality shown by the infrared requiring students to analyze, interpret, and use spectroscopic
spectrum”, and “indicate reasonable correspondence between data improved the percentage of our sample that constructed
some of the proton environments in your prediction and some of appropriate claims. In both of these cases, the class was
the 1H NMR peaks”. For this analysis, we will report whether any randomly assigned to either an experimental or control
predictor variable significantly increased the model’s predictive condition. The control condition was given a low structure
validity relative to the intercept-only “null model”. High prompt, whereas the experimental condition was given a high
structure prompts were chosen as the basis for our models structure or “critique focused” prompt involving use of the same
G DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

Figure 6. Percentage of students who analyzed and interpreted data from infrared and 13C NMR spectra as well as the percentage of students who
constructed a structural prediction consistent with these two types of spectroscopic data. “Exam 1” prompts were present on summative assessments,
whereas “HW” prompts represent assessments given as part of a homework assignment.

data set. To analyze whether prompt structure impacted the infrared, 13C NMR, and 1H NMR traces. Accordingly, we will
reasonableness of student claims, we coded whether each unpack student responses to high structure prompts that
student constructed a claim consistent with assembled evidence explicitly require use of relevant procedural resources (see
and compared the percentage of successful students in the Figure 3 for an example of one such prompt). The percentage of
control versus experimental conditions. Association between students from our sample who deployed these resources
condition and constructing a reasonable claim was analyzed via a appropriately is shown in Figure 5.
Pearson’s χ2 test. Recall that, for the purposes of the analyses encapsulated by
The third experiment was meant to shed light on whether Figure 5, we defined “analyze and interpret spectroscopic data”
asking students to examine fewer strands of spectroscopic as application of a set of four procedures. Thus, analysis and
evidence would impact their success in constructing a claim interpretation of data from an infrared spectrum signified that
consistent with that evidence. It has been well-established that students could correctly describe functional groups indicated by
humans have a limited capacity to process information.47−50 In a particular spectrum. As this contribution is focused on
light of this, one might guess that difficulties with engaging in productive activation of resources rather than describing
argumentation from spectroscopic evidence could emerge in difficulties, students who described a functionality shown by
part from the various procedures that make up this practice, an IR trace and also described a functionality not indicated by
overwhelming students’ working memory. Requiring fewer that same data source were counted as “analyzing and
processing tasks, by reducing the amount of data to analyze, interpreting data” here. Analysis and interpretation of 13C
interpret, and use, might therefore positively impact student NMR data focused on whether students could correctly describe
ability to construct evidence-based claims. To test this, we the number of carbon environments indicated by a spectrum.
randomly assigned students to one of two groups. Both groups It should be noted that analysis and interpretation of data
were assigned a set of two moderate structure prompts with one from 13C NMR and IR spectra were explicitly prompted for by
prompt asking for a claim consistent with two spectroscopic the high structure prompts, whereas students were not
traces (infrared and 13C NMR) and the other prompt asking for specifically prompted to extract information from 1H NMR
a structural prediction consistent with three spectroscopic traces spectra. Instead, they were asked to use evidence from a proton
(infrared, 13C NMR, and 1H NMR). Group 1 received all three NMR to justify their claims (see the portion of Figure 5 that
traces for Unknown A and two traces for Unknown B. Group 2 reads “Prompts”). For this reason, we consider indicating a
received two traces for Unknown A and three for Unknown B. reasonable correspondence between at least some of the proton
Each assessment response was coded according to whether the environments in their structural prediction and the 1H NMR
claim presented was completely consistent with the assembled spectrum as evidence students were “analyzing, interpreting, and
evidence. We examined whether there was an association
using” data from a proton NMR. As such correspondence had to
between the percentage of students who proposed a reasonable
take into account information from peak splitting and
claim and the number of strands of data students were required
multiplicity, we considered this an indicator that students
to analyze via a series of χ2 tests.


could leverage both of the proton-NMR-relevant procedures
RESULTS AND DISCUSSION previously described to interpret 1H NMR data and use that data
to justify their prediction. Interestingly, no association existed
RQ 1: Success in Analyzing and Interpreting Particular between the type of assessment (i.e., formative or summative)
Strands of Spectroscopic Data and student success in analyzing, interpreting, and using data
This research question focuses squarely on student facility with from IR and 1H NMR spectra as prompted for by the high
the four algorithms previously highlighted as integral to the structure assessments (p > 0.01 in all instances). No measures of
analysis, interpretation, and use of spectroscopic data from association could be calculated relating the type of assessment to
H DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

success analyzing and interpreting 13C NMR data, because at One could persuasively argue that analysis, interpretation, and
least one variable in each two-way table was a constant. Lack of use of data from an infrared spectrum is the simplest example of
association between type of assessment and success in prompted argumentation from spectroscopic evidence encountered in
performances indicates that students were just as successful on organic chemistry. Indeed, one need only recognize the
homework problems as they were on similarly structured exam signature of a functional group and draw some sort of molecule
prompts. incorporating that functional group to have a claim consistent
From examination of Figure 5, it is readily apparent that most with IR evidence. For example, one could interpret a strong
student respondents proved adept at deploying the procedural infrared peak around 1700 cm−1 as signifying the presence of a
resources in the context of highly structured prompts on exams carbonyl and then draw near infinite permutations of a carbonyl-
and homework. In particular, nearly everyone could pull containing compound. It is perhaps a function of the relatively
information on functional groups from the infrared spectrum simple nature of these tasks that most students were successful at
and discern the number of carbon environments from the 13C analyzing, interpreting, and using data from infrared spectra on
NMR. Both of these tasks are fairly straightforward and the formative and summative items we examined. As mentioned
practiced a great deal during lecture and on homework. Students under RQ1, most students could also correctly state the number
were similarly successful in labeling correspondence between 1H of chemically distinct carbons indicated by a carbon NMR,
NMR peaks and their claims, which required that two related which required only that they count the peaks in the spectrum.
procedures (i.e., the “n + 1” rule and interpretation of peak However, construction of a structural prediction with an
integration values) were simultaneously leveraged. appropriate number of different carbon environments proved
to be more challenging than assembling a prediction consistent
RQ 2: Association between Analysis and Interpretation of
with IR evidence. This may be due to the fact that an actionable
Spectroscopic Data and Use of that Evidence to Inform
Claims understanding of “different carbon environments” requires
students to examine their claim as a whole and be sensitive to
It is clear that most students can deploy algorithms to extract any symmetry elements that might be present in the molecule
information from spectra. What is less clear is whether this they drew.
evidence actually informs structural predictions. Here, we RQ3: Degree to Which Facility with Particular Procedures
explore whether there exist associations between analysis and Predicts Student Success in Proposing a Structure
interpretation of data and use of this evidence in claim Consistent with All Assembled Spectroscopic Evidence
construction. Our analysis leverages student responses to the
three high structure prompts previously discussed because these To determine whether analysis, interpretation, or use of
particular spectroscopic data sources predicted student success
prompts allow us to ascertain whether appropriate information
in proposing a structure consistent with all assembled evidence
is being derived from the spectra. In particular, we focused on
in the context of three highly structured prompts, we
association between the information students obtain from IR
constructed binary logistic regression models described by the
and 13C NMR data and their predictions. The percentage of
formula below:
students who analyzed and interpreted data from infrared and
carbon NMR spectra is presented alongside the percentage of logit(Y ) = B0 + B1X1 + B2 X 2 + B3X3
students whose structural predictions were consistent with this
data in Figure 6. As an example, a student who correctly stated The outcome variable Y represents student construction of a
that a 13C NMR indicated that there were six different kinds of structural prediction consistent with all assembled evidence.
carbon would be counted as “analyzing and interpreting” carbon Predictor variables X1, X2, and X3 represent analysis and
NMR data appropriately, whereas a student who constructed a interpretation of data from infrared and 13C NMR spectra (X1,
claim that had six carbon environments would be categorized as and X2, respectively) as well as indication of correspondence
using the aforementioned evidence in their argument. For the between 1H NMR peaks and some part of a structural claim
purposes of this analysis, claims were not required to be (X3). The outcome variable and all predictor variables are
consistent with all assembled data, just the data that are relevant dichotomous. B0 conveys the likelihood students will construct
to the claim. For example, if a student constructed a prediction evidence-based claims if no predictor variables improve model
with the proper number of carbon environments that was predictive validity. A model in which logit(Y) = B0 may be
inconsistent with the given 1H NMR, they would still be counted thought of as a “null model”. The regression coefficients B1, B2,
as using evidence from the 13C NMR spectrum. and B3 describe the direction and magnitude of the relationship
The majority of students who responded to our three highly between the relevant predictor variables and the logit of Y.51 In
structured prompts were able to analyze, interpret and use data all three models examined for this study, no predictor variable
from infrared and 13C NMR spectra. This may indicate that significantly improved the predictive validity of a model relative
analysis and interpretation are perceived as part of an ensemble to the analogous null model. These three logistic regressions
of activities that should culminate in construction of an suggest that, for our data set, success on tasks that required
evidence-based claim. Because almost all students could pull analysis, interpretation, or use of spectroscopic data did not
information from IR and carbon NMR spectra, the association predict student success in assembling a structural prediction
between data analysis and interpretation and structural consistent with the full corpus of spectroscopic data.
prediction could not be analyzed via a χ2 test because 25% or RQ 4: Association between Prompt Structure and
more of the cells in the contingency tables have expected counts Construction of Reasonable Claims
lower than 5, violating the assumptions of the test. That is, the Thus far, we have seen that students can very often engage in the
number of students who were unable to pull appropriate algorithms needed to pull particular information from
information from infrared and carbon NMR spectra and also spectroscopic data as well as use this information to inform
unable to construct claims incorporating this information was their predictions. We have also seen that, for high structure
too few to run a χ2 test. prompts, facility with particular procedures does not predict
I DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

Figure 7. Percentage of students who constructed claims consistent with spectroscopic evidence as part of experiments designed to show the impact of
prompt structure on student success.

success in constructing a claim consistent with all assembled goal. As this is a quantitative study, we cannot do more than
data. For our fourth research question, we sought to examine the speculate as to the cause for the results of this first experiment.
impact of different prompt structures on student construction of The results of experiment 2 mirror those from experiment 1.
reasonable claims. It is possible that assessment items that differ That is, the percentage of students who successfully constructed
from our high structure variant might be more successful in an evidence-based claim in response to a “critique-centered”
cuing students into the processes of construction and critique prompt did not differ significantly from what would be expected
essential to creating a reasonable argument. To get at the impact by chance. Accordingly, there was no significant association
of varying prompt structure on student success in constructing noted between prompt structure and percentage of students
evidence-based claims, we analyzed the results of three who proposed reasonable structures (p > 0.01). These results
experiments. Two were designed to enable inferences into the indicate that our attempt to explicitly require construction and
impact of different prompt scaffolds on the percentage of critique of structural claims was not successful in the context of
students who successfully constructed evidence-based claims, this experiment. With regard to the results of experiment 1 and
experiment 2, it may be that each subpart of highly scaffolded
and the third was meant to shed light onto whether student
questions was perceived as a separate, stand-alone entity to be
success might be improved by lowering the amount of data they
answered and forgotten rather than considered in the context of
were required to analyze. The results of these three experiments
claim construction and critique. It would be interesting to
are summarized in Figure 7. explore the impact of scaffolding from a students’ perspective
The percentage of students who constructed claims consistent through think-aloud interviews conducted in the context of
with the data set they were provided was approximately the same different prompt types.
regardless of whether they were given a high structure or low The third experiment differed from the first two in that the
structure prompt as part of experiment 1 (Figure 7). Therefore, aspect of prompt structure that changed between the
there was no significant association between prompt structure experimental and control groups was not scaffolding but the
and likelihood of constructing an evidence-based claim (p > amount of data to be analyzed. For this experiment, students
0.01). Our scaffolding strategy for high structure prompts was to were randomly placed into two groups with one group receiving
explicitly ask for focus on various facets of the assembled data set three spectroscopic traces corresponding to Unknown A (1H
with the hope that students would consider the evidence they NMR, 13C NMR, and IR) and two traces corresponding to
pulled (and wrote) when constructing and critiquing their Unknown B (13C NMR and IR) and the second group receiving
structural predictions. The results of this experiment would two spectroscopic traces corresponding to Unknown A (13C
indicate that this approach was not successful in realizing our NMR and IR) and three traces corresponding to Unknown B
J DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

(1H NMR, 13C NMR, and IR). The three-trace variant of the Deep versus Shallow Procedural Knowledge
prompts used for experiment 3 have been reproduced in the We framed our initial discussion of analysis, interpretation, and
Supporting Information for this manuscript (Figures S1 and S2). use of spectroscopic data in terms of activation of a series of
In the context of spectroscopic traces derived from Unknown A, procedural resources. Recall that facility with such resources can
we can see that students given more data to analyze were more be characterized as “deep procedural knowledge” (i.e., flexible,
successful than those given less. Indeed, in this case, there is a context-sensitive deployment of resources), or “shallow
significant association between the amount of data to be procedural knowledge” (i.e., inability to deploy resources
analyzed and successful construction of evidence-based claims: reasonably and efficiently, despite knowing a procedure).
χ2(1) = 10.4, p = 0.001, Cramer’s V = 0.33. More students given Tasks that simply ask students to analyze and interpret
three spectroscopic traces to analyze constructed successful spectroscopic data to provide straightforward information can
claims than would be expected by chance. This runs counter to be readily answered by individuals who possess shallow
our hypothesis that student success should be improved by procedural knowledge. For example, one can recognize a
reducing the cognitive load of the problem via decreasing the broad peak around 3400 cm−1 in an infrared spectrum as the
data that must be considered during claim construction. signature of a hydroxyl group without knowing anything about
However, this result was not replicated in the context of traces how to flexibly use the trove of information found within proton
derived from Unknown B. That is, no significant association was NMR spectra. Argumentation, that is, construction and critique
found between the richness of the data set presented to students of evidence-based claims, requires one to both pull evidence
and their success in constructing evidence-based claims (p > from all available data sources and iterate between claim
0.01). We cannot account for the higher percentage of students construction and evaluation of whether a certain claim is
who successfully proposed the structure of Unknown A relative consistent with assembled evidence. Facility with argumentation
to those who constructed a reasonable claim as to the structure from spectroscopic evidence, we argue, is predicated on deep
of Unknown B. Unknown A was 2-methylpropan-1-ol, whereas rather than shallow procedural knowledge. Given the relatively
Unknown B was 3-methylbutan-2-one; both produce spectro- modest success of our student sample at constructing a claim
scopic traces of similar complexity. The results of both parts of consistent with spectroscopic data, one might think that the
experiment 3 stand in opposition to our notion that students surveyed lack deep procedural knowledge. Although
argumentation from spectroscopic evidence might be over- this may be true for a subset of students, we would argue
whelming students’ processing capacity. We have no evidence focusing on what students lack is not terribly productive; it
that students are more successful in constructing evidence-based would be more fruitful to focus on what these students brought
claims when presented with less data-rich contexts. Indeed, to the table (e.g., an ability to analyze and interpret data from
given that the 1H NMR spectra were the most information-rich several sources) and how those resources might be woven
data sources our students encountered, it may be that omitting a together to better enable successful engagement in internal and
proton NMR spectrum from a prompt prevented students from external argumentative discourse.
using familiar procedural resources and forced reliance on less
Importance of Framing
familiar strategies. This may have negatively impacted student
ability to analyze, interpret, and use spectroscopic evidence to Every assessment prompt discussed here has the potential to
construct a reasonable claim as to the identity of Unknown A. engage students in an internal dialogic process of construction


and critique that bears the structural hallmarks of an
CONCLUSIONS argument.8,52 That is, to address each item, individuals should
pull evidence from spectroscopic data and iteratively construct
Analysis of the data presented here indicates that many students and critique claims until they are satisfied that said claims are
can successfully engage with the parts of homework and exam consistent with the full data corpus. Indeed, methodically
prompts that ask them to pull specific information from double-checking the consistency of claims with evidence was
spectroscopic data. Virtually everyone whose responses were done by all of the “more successful” solvers of spectroscopy
examined was quite adept at recognizing functionality in an problems in Cartrette and Bodner’s earlier work.15 However, the
infrared spectrum, discerning the number of chemically distinct fact that a task asks for students to support claims with evidence
carbons from a 13C NMR spectrum, and correctly labeling does not at all guarantee that students will perceive said task as
correspondence between proton environments in their struc- serving a persuasive function. Literature on student under-
tural prediction and peaks in an 1H NMR spectrum. It is possible standings of the type of activity they are doing, that is, their
that the significant practice OCLUE-enrolled students have had framing of that activity, suggests that students very often
with spectroscopic analysis paid dividends in terms of ready and perceive tasks or scenarios meant to provoke argumentation as a
appropriate deployment of certain procedural resources on classroom performance rather than a purposeful endeavor.53−55
homework and exams. Additionally, the majority of students Berland and Hammer characterize engaging in an argument to
were able to construct structural predictions that included please the teacher (i.e., “doing the lesson”) rather than as part of
functionality indicated by the infrared spectrum and had an figuring out a problem of interest to the classroom community as
appropriate number of carbon environments. However, logistic “pseudoargumentation”.54,55 As the activities we have created
regression models built from student responses to three ask for students to figure out the identity of an unknown without
multipart prompts found no significant association between any reason for why they might want to do so; it is possible that
successfully pulling data from a spectrum and overall model these activities were perceived as a rote performance rather than
predictive validity. Further, different prompt structures had no an opportunity to figure out something meaningful. Our tasks
impact on the success of students in constructing evidence- are very similar (though perhaps more scaffolded) than those
based claims. In summary, students struggled mightily with which appear on many organic chemistry tests across the
constructing a reasonable claim to their argument, despite being nation.2 Thus, if we ended up cuing many students into a frame
able to analyze and interpret data from several sources. at odds with the scientific practice of argumentation, many
K DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

others may do likewise. Studies in the K−12 science education We can only say what resources appeared to be used in the
space indicate that students tend to draw on resources useful for context of the prompts we have discussed. We cannot speculate
argumentation to a greater degree when they see a need for as to whether students possess or know they possess the
argumentative discourse (e.g., they want to persuade their peers intellectual resources relevant to these problems; it is possible
of the reasonableness of a claim to make sense of a students would have been more successful on prompts worded
phenomenon).55,56 Stated differently, students can very often differently or if they had more time to address summative items.
reasonably link claims to evidence and attend to alternative Second, all students surveyed as part of this study enrolled in two
arguments when they see a reason to do so, at least in the K−12 semesters of a transformed undergraduate organic chemistry
contexts studied.54 Thus, it is possible that students might be curriculum that placed substantial emphasis on argumentation
more successful at engaging in argument from spectroscopic from spectroscopic evidence. As such, they had ample practice
evidence if they perceived a need to construct and critique cultivating the procedural resources needed to analyze and
evidence-based claims. interpret spectroscopic data. Our results cannot generalize
Implications for Teaching beyond this context; we do not know how students taught using
a more traditional curriculum might have fared on the prompts
Argumentation from spectroscopic evidence, as with argumen- discussed.


tation more generally, is extremely challenging. To be successful,
students must pull information from various strands of data, ASSOCIATED CONTENT
itself a scientific practice requiring facility with a range of
resources, and then iteratively construct and critique claims until *
S Supporting Information

they arrive at one consistent with the full data corpus. Given that The Supporting Information is available on the ACS
so few students in our sample were able to arrive at reasonable Publications website at DOI: 10.1021/acs.jchemed.9b00550.
structural predictions on exams and homework, it may not be Demographic and academic measures as well as example
reasonable to ask students to predict an unknown consistent prompts (PDF, DOCX)
with spectroscopic data as one part of a larger, timed summative
exam. Ideally, we would want students (and future healthcare
professionals) to carefully reflect on consistency between
predictions and evidence, not quickly write down their “best
■ AUTHOR INFORMATION
Corresponding Author
guess” and then move on because of time constraints. This may
*E-mail: rstowe@chem.wisc.edu.
mean that argumentation from spectroscopic evidence is better
suited to laboratory environments than lecture. ORCID
The broader focus of our analytical efforts also bears Ryan L. Stowe: 0000-0002-5548-495X
mention−that is, emphasis on student resources rather than Melanie M. Cooper: 0000-0002-7050-8649
constraints, deficiencies, or misconceptions. We see little value
Present Address
in focusing substantial effort on parsing out all of the struggles †
students have with engaging in a practice such as argumentation. R.L.S.: Department of Chemistry, University of Wisconsin
It is very likely that students have a range of small-grain Madison, 1101 University Avenue, Madison, Wisconsin 53706,
knowledge elements that can be called on in more or less United States
productive ways to address an assessment item or problem Notes
posed during instruction.30,32,34,35,57−59 Characterizing resour- The authors declare no competing financial interest.


ces strung together in less-than-optimal ways only serves to tell
the community that certain students at a certain point in time ACKNOWLEDGMENTS
had particular troubles with a particular task. A more fruitful
course of action would be to examine what students know and This work was supported in part by the National Science
can do across several related contexts to get a sense of the Foundation under DUE:1502552. We would also like to thank
stability of their patterns of resource activation. From there, one Elizabeth Day for her assistance editing this manuscript and
can work toward designing learning environments that support advice on statistical analyses.
productive weaving of intuitive resources with those developed
by instruction to engage in the performances we value (such as
argumentation). Here, we have found that consistent integration
■ REFERENCES
(1) Moran, B. How to Get an A− in Organic Chemistry. The New York
of spectroscopic analysis throughout OCLUE has equipped our Times, Nov 1, 2013, p ED14.
students to analyze and interpret spectroscopic data from (2) Stowe, R. L.; Cooper, M. M. Practice What We Preach: Assessing
various sources; that is, our students can leverage several “Critical Thinking” in Organic Chemistry. J. Chem. Educ. 2017, 94 (12),
procedural resources that will be needed for engaging in 1852−1859.
(3) The National Research Council. A Framework for K−12 Science
argument from spectroscopic evidence. Our task going forward
Education: Practices, Crosscutting Concepts, and Core Ideas; National
is to consider how we might help students flexibly deploy their Academies Press: Washington, DC, 2012.
ability to use data to construct and critique evidence-based (4) Osborne, J. F.; Patterson, A. Scientific Argument and Explanation:
claims. A Necessary Distinction? Sci. Educ. 2011, 95 (4), 627−638.

■ LIMITATIONS
As with all studies, the work presented here has several
(5) Maciejewski, W.; Star, J. R. Developing Flexible Procedural
Knowledge in Undergraduate Calculus. Res. Math. Educ. 2016, 18 (3),
299−316.
(6) Star, J. R. Foregrounding Procedural Knowledge. J. Res. Math.
limitations. First and most fundamentally, this was a quantitative Educ. 2007, 38 (2), 132−135.
study in which a large number of student responses to several (7) Star, J. R. Reconceptualizing Procedural Knowledge. J. Res. Math.
multipart spectroscopic interpretation prompts were analyzed. Educ. 2005, 36 (5), 404−411.

L DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

(8) Ford, M. J. A Dialogic Account of Sense-Making in Scientific (32) Hammer, D.; Elby, A. On the Form of a Personal Epistemology.
Argumentation and Reasoning. Cognition Instruct. 2012, 30 (3), 207− In Personal Epistemology: The Psychology of Beliefs about Knowledge and
245. Knowing; Hofer, B. K., Pintrich, P. R., Eds.; Lawrence Erlbaum
(9) Pulliam, C. R.; Pfeiffer, W. F.; Thomas, A. C. Introducing NMR to Associates: Mahwah, NJ, 2002; pp 169−190.
a General Chemistry Audience: A Structural-Based Instrumental (33) Hammer, D.; Elby, A. Tapping Epistemological Resources for
Laboratory Relating Lewis Structures, Molecular Models, and 13C Learning Physics. J. Learn. Sci. 2003, 12 (1), 53−90.
NMR Data. J. Chem. Educ. 2015, 92 (8), 1378−1380. (34) Hammer, D.; Elby, A.; Scherr, R. E.; Redish, E. F. Resources,
(10) Vosegaard, T. ISpec: A Web-Based Activity for Spectroscopy Framing, and Transfer. In Transfer of Learning from a Modern
Teaching. J. Chem. Educ. 2018, 95, 97. Multidisciplinary Perspective; Mestre, J. P., Ed.; IAP: Greenwich, CT,
(11) Debska, B.; Guzowska-Swider, B. Molecular Structures from 1H 2005; pp 89−119.
NMR Spectra: Education Aided by Internet Programs. J. Chem. Educ. (35) Cooper, M. M.; Corley, L. M.; Underwood, S. M. An
2007, 84 (3), 556. Investigation of College Chemistry Students’ Understanding of
(12) Flynn, A. B. NMR Interpretation: Getting from Spectrum to Structure−Property Relationships. J. Res. Sci. Teach. 2013, 50 (6),
Structure. J. Chem. Educ. 2012, 89 (9), 1210−1212. 699−721.
(13) Winschel, G. A.; Everett, R. K.; Coppola, B. P.; Shultz, G. V.; (36) Cooper, M.; Klymkowsky, M. Chemistry, Life, the Universe, and
Lonn, S. Using Jigsaw-Style Spectroscopy Problem-Solving To Everything: A New Approach to General Chemistry, and a Model for
Elucidate Molecular Structure through Online Cooperative Learning. Curriculum Reform. J. Chem. Educ. 2013, 90 (9), 1116−1122.
J. Chem. Educ. 2015, 92 (7), 1188−1193. (37) Cooper, M. M.; Posey, L. A.; Underwood, S. M. Core Ideas and
(14) Angawi, R. F. Using a Problem Solving-Cooperative Learning Topics: Building Up or Drilling Down? J. Chem. Educ. 2017, 94 (5),
Approach To Improve Students’ Skills for Interpreting 1H NMR 541−548.
Spectra of Unknown Compounds in an Organic Spectroscopy Course. (38) ACT/SAT Concordance Tables. ACT Website. https://www.act.
J. Chem. Educ. 2014, 91 (6), 823−829. or g/con tent/ da m/a ct/u nsecur ed/documen ts/ACT- SAT-
(15) Cartrette, D. P.; Bodner, G. M. Non-Mathematical Problem Concordance-Tables.pdf (accessed June 10, 2019).
Solving in Organic Chemistry. J. Res. Sci. Teach. 2010, 47 (6), 643−660. (39) SPSS Statistics for Mac; IBM Corp.: Armonk, NY, 2017.
(16) Topczewski, J. J.; Topczewski, A. M.; Tang, H.; Kendhammer, L. (40) Bryfczynski, S.; Pargas, R. P.; Cooper, M. M.; Klymkowsky, M.;
K.; Pienta, N. J. NMR Spectra through the Eyes of a Student: Eye Dean, B. C. Teaching Data Structures with BeSocratic. In ITiCSE ’13,
Tracking Applied to NMR Items. J. Chem. Educ. 2017, 94, 29. Proceedings of the 18th ACM Conference on Innovation and
(17) Domin, D.; Bodner, G. Using Students’ Representations Technology in Computer Science Education, Canterbury, England,
Constructed during Problem Solving To Infer Conceptual Under- July 1−3, 2013; ACM: New York, NY, 2013; pp 105−110,
standing. J. Chem. Educ. 2012, 89 (7), 837−843. DOI: 10.1145/2462476.2465583.
(18) Cullipher, S.; Sevian, H. Atoms versus Bonds: How Students (41) Bryfczynski, S.; Pargas, R. P.; Cooper, M. M.; Klymkowsky, M.;
Look at Spectra. J. Chem. Educ. 2015, 92 (12), 1996−2005. Hester, J.; Grove, N. P. Classroom Uses for BeSocratic. In The impact of
(19) Connor, M. C.; Finkenstaedt-Quinn, S. A.; Shultz, G. V.
pen and touch technology on education; Hammond, T., Valintine, S.,
Constraints on Organic Chemistry Students’ Reasoning during IR and
Adler, A., Payton, M., Eds.; Springer International Publishing: New
1H NMR Spectral Interpretation. Chem. Educ. Res. Pract. 2019, 20, 522.
York, 2015; pp 127−136.
(20) Ford, M. Disciplinary Authority and Accountability in Scientific
(42) Cooper, M. M.; Underwood, S. M.; Bryfczynski, S. P.;
Practice and Learning. Sci. Educ. 2008, 92 (3), 404−423.
Klymkowsky, M. W. A Short History of the Use of Technology To
(21) Osborne, J. Arguing to Learn in Science: The Role of
Model and Analyze Student Data for Teaching and Research. In Tools of
Collaborative, Critical Discourse. Science 2010, 328 (5977), 463−466.
(22) Cooper, M. M.; Stowe, R. L.; Crandell, O. M.; Klymkowsky, M. Chemistry Education Research; ACS Symposium Series; American
W. Organic Chemistry, Life, the Universe and Everything (OCLUE): A Chemical Society: Washington, DC, 2014; Vol. 1166, pp 219−239,
Transformed Organic Chemistry Curriculum. J. Chem. Educ. 2019, DOI: 10.1021/bk-2014-1166.ch012.
under revision, DOI: 10.1021/acs.jchemed.9b00401 (43) Laverty, J. T.; Underwood, S. M.; Matz, R. L.; Posey, L. A.;
(23) Klein, D. A Quarter Century of US ‘Math Wars’ and Political Carmel, J. H.; Caballero, M. D.; Fata-Hartley, C. L.; Ebert-May, D.;
Partisanship. BSHM Bulletin: Journal of the British Society for the History Jardeleza, S. E.; Cooper, M. M. Characterizing College Science
of Mathematics 2007, 22 (1), 22−33. Assessments: The Three-Dimensional Learning Assessment Protocol.
(24) Principles and Standards for School Mathematics; National Council PLoS One 2016, 11 (9), No. e0162333.
of Teachers of Mathematics: Reston, VA, 2000. (44) Spectral Database for Organic Compounds SDBSWeb; National
(25) Principles to Actions: Ensuring Mathematical Success for All; Institute of Advanced Industrial Science and Technology, 2018.
National Council of Teachers of Mathematics: Reston, VA, 2014. https://sdbs.db.aist.go.jp (accessed May 22, 2019).
(26) Research Advisory Committee of the National Council of (45) Cohen, J. Statistical Power Analysis for the Behavioral Sciences, 2nd
Teachers of Mathematics. NCTM Curriculum and Evaluation ed.; Routledge: Hillsdale, NJ, 1988.
Standards for School Mathematics: Responses from the Research (46) Cooper, M. M.; Kouyoumdjian, H.; Underwood, S. M.
Community. J. Res. Math. Educ. 1988, 19 (4), 338−344. Investigating Students’ Reasoning about Acid−Base Reactions. J.
(27) Hiebert, J.; Lefevre, P. Conceptual and Procedural Knowledge in Chem. Educ. 2016, 93 (10), 1703−1712.
Mathematics: An Introductory Analysis. In Conceptual and procedural (47) Miller, G. A. The Magical Number Seven, plus or Minus Two:
knowledge: The case of mathematics; Hiebert, J., Ed.; Lawrence Erlbaum Some Limits on Our Capacity for Processing Information. Psychol. Rev.
Associates: Hillsdale, NJ, 1986; pp 1−27. 1956, 63 (2), 81−97.
(28) Star, J. R.; Newton, K. J. The Nature and Development of (48) De Ribaupierre, A.; Fagot, D.; Lecerf, T. Working Memory
Experts’ Strategy Flexibility for Solving Equations. ZDM Mathematics Capacity and Its Role in Cognitive Development: Are Age Differences
Education 2009, 41, 557−567. Driven by the Same Processes Across the Lifespan? In Cognitive
(29) Hammer, D. Student Resources for Learning Introductory Development and Working Memory: A Dialogue between Neo-Piagetian
Physics. Am. J. Phys. 2000, 68, S52−S59. Theories and Cognitive Approaches; Barrouillet, P., Gaillard, V., Eds.;
(30) diSessa, A. A. Knowledge in Pieces. In Constructivism in the Psychology Press: Hove, East Sussex, 2011; pp 105−133,
Computer Age; Forman, G. E., Pufall, P. B., Eds.; Lawrence Erlbaum DOI: 10.4324/9780203845837.
Associates: Hillsdale, NJ, 1988; pp 49−70. (49) Cowan, N.; Chen, Z.; Rouder, J. N. Constant Capacity in an
(31) Hammer, D. Epistemological Beliefs in Introductory Physics. Immediate Serial-Recall Task: A Logical Sequel to Miller (1956).
Cognition Instruct. 1994, 12 (2), 151−183. Psychol. Sci. 2004, 15 (9), 634−640.

M DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX
Journal of Chemical Education Article

(50) Mathy, F.; Feldman, J. What’s Magic about Magic Numbers?


Chunking and Data Compression in Short-Term Memory. Cognition
2012, 122 (3), 346−362.
(51) Peng, C.-Y. J.; Lee, K. L.; Ingersoll, G. M. An Introduction to
Logistic Regression Analysis and Reporting. J. Educ. Res. 2002, 96 (1),
3−14.
(52) Ford, M. Grasp of Practice’ as a Reasoning Resource for Inquiry
and Nature of Science Understanding. Sci. & Educ. 2008, 17 (2−3),
147−177.
(53) Enyedy, N.; Goldberg, J. Inquiry in Interaction: How Local
Adaptations of Curricula Shape Classroom Communities. J. Res. Sci.
Teach. 2004, 41 (9), 905−935.
(54) Berland, L. K.; Hammer, D. Students’ Framings and Their
Participation in Scientific Argumentation. In Perspectives on Scientific
Argumentation: Theory, Practice and Research; Khine, M. S., Ed.;
Springer Netherlands: Dordrecht, 2012; pp 73−93, DOI: 10.1007/
978-94-007-2470-9_5.
(55) Berland, L. K.; Hammer, D. Framing for Scientific
Argumentation. J. Res. Sci. Teach. 2012, 49 (1), 68−94.
(56) Cavagnetto, A. R. Argument to Foster Scientific Literacy: A
Review of Argument Interventions in K−12 Science Contexts. Rev.
Educ. Res. 2010, 80 (3), 336−371.
(57) Cooper, M. M.; Stowe, R. L. Chemistry Education Research
From Personal Empiricism to Evidence, Theory, and Informed
Practice. Chem. Rev. 2018, 118 (12), 6053−6087.
(58) diSessa, A. Alternative Conceptions and P-Prims. In Encyclopedia
of Science Education; Gunstone, R., Ed.; Springer: Netherlands, 2015; pp
34−37, DOI: 10.1007/978-94-007-2150-0_87.
(59) diSessa, A. A. A History of Conceptual Change Research:
Threads and Fault Lines. In The Cambridge Handbook of the Learning
Sciences; Cambridge University Press: New York, NY, 2014; pp 88−
108.

N DOI: 10.1021/acs.jchemed.9b00550
J. Chem. Educ. XXXX, XXX, XXX−XXX

You might also like