Neural patterns reflect conceptual grasp of novice students following first class learning in physics

Neural patterns reflect conceptual grasp of novice students following first class learning in physics | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Neural patterns reflect conceptual grasp of novice students following first class learning in physics Joshua S. Cetron, Megan E. Hillis, Solomon G. Diamond, Vicki V. May, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6992513/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 22 Feb, 2026 Read the published version in npj Science of Learning → Version 1 posted 9 You are reading this latest preprint version Abstract Students in STEM fields frequently learn new abstract concepts as they build knowledge for scientific innovation. Yet little work has investigated how patterns of neural activity reflect the emergence of this newly learned conceptual information. In a single lesson and lab activity, participants learned about physics concepts, then subsequently completed an fMRI session. We identified neural patterns tracking students’ newly acquired STEM concept knowledge, using a machine-learning classifier to assess the embedding of concept-relevant categories in students’ neural representations of the task stimuli. Patterns in several parietal and temporal regions reflected conceptual knowledge acquired during the lesson. Crucially, a regression analysis further demonstrated that greater concept-relevant organization of the stimuli in these brain regions was associated with better performance on behavioral concept knowledge assessments. Results suggest that after only brief exposure to new STEM topics, early evidence of comprehension can be identified in the individualized neural patterns of novice learners. Biological sciences/Neuroscience Biological sciences/Psychology Social science/Psychology cognitive neuroscience STEM learning fMRI knowledge representation Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Acquiring new abstract concepts is one of the most remarkable abilities of the human brain. Unlike most nonhuman animals, humans can quickly generalize abstract principles from brief exposure to sparse examples 1 . Correspondingly, our education system is structured in large part around teaching students to organize and apply a wide variety of abstract concepts. But how do we know when we have acquired a new concept? In traditional learning environments, concept acquisition is often probed with explicit tests of knowledge, such as a written exam testing fact-retrieval. Once some knowledge has been acquired, more practical tests of knowledge may involve transfer tasks: tasks that require novel application of the learned concepts to complete. But both explicit knowledge tests and transfer tests of knowledge still require us to infer that a student understands a concept from their behavior in a limited circumstance: their performance on the exam or on the conceptual transfer task. While useful, these sources of inference are inherently subject to noise from several sources (e.g., sampling of test items, learner performance on a given day in a given testing situation, etc.). To draw from an additional source of data regarding the knowledge a learner has acquired, it can be informative to inspect the human concept acquisition system itself: the brain. Prior research on the neural basis of concept learning—particularly STEM concept learning—has shown that patterns of neural activity can indeed reflect abstract concept knowledge 2 – 11 . These abstract representations can be identified using multivariate representational analyses of functional neuroimaging data, such as representational similarity analysis 12 , informational network analysis 2 , informational connectivity analysis 13 , and inter-subject correlation analysis 14 . Many of these studies investigating the neural representations of learned concepts compare two groups of individuals with substantially different levels of knowledge: advanced learners, who are expected to have acquired the concepts already, and novices, who are naïve to the concepts. Comparing novices to more advanced learners (or even experts) allows for the detection of larger differences in concept knowledge representations, since expertise has been shown to affect conceptual organization across a variety of domains (e.g., chess, as in seminal work by Chase and Simon 15 ; also in physics 16 ). However, such between-group comparisons of concept knowledge representations do not inform us about the trajectory of concept learning, because the two groups are typically situated at either end of the learning process. Characterizing the learning process by identifying signs of early conceptual understanding is particularly important in STEM education because of the cumulative nature of many STEM topics (e.g., understanding object mechanics requires first understanding Newtonian force). One study by Mason and Just 9 demonstrated that as participants were exposed to incremental information about how simple mechanical systems (such as a fire extinguisher) work, patterns of neural activity evoked by the individual objects shifted from primarily visual areas to parietal and frontal areas associated with functions that could support deeper understanding such as mental animation. These results provide compelling evidence that relatively short training can evoke functional changes in novices’ conceptual representations at the group level. However, it remains untested whether neural indicators of concept knowledge can be used to capture individual variation in how well novice learners grasp new STEM concepts upon their first exposure to a topic. One study by Meshulam and colleagues 17 followed a single group of STEM students longitudinally as they progressed through an introductory undergraduate computer science course, measuring neural responses to recorded lecture videos at various timepoints during the learning period. They found that students whose neural activity patterns looked more like the group’s average neural activity performed better on the course’s final exam. However, this study did not specifically examine the initial concept-acquisition phase of learning, instead focusing on learning outcomes after a full academic term of study. Moreover, as the focus of this study was on the neural alignment of students to each other and to their expert instructors—indexed by inter-subject neural correlation—the method did not allow for an analysis of specific conceptual knowledge structures encoded in students’ neural activity. In contrast, methods such as RSA and others cited above can be used to model the embedding of concept knowledge structures within neural activity patterns directly. In the present experiment, we introduced novice students (i.e., students with no advanced physics or engineering experience) to two core concepts from static mechanics, a domain of mechanical engineering, in a controlled laboratory setting. An overview of the present experiment is illustrated in Fig. 1 . These novices received one hour of instruction on the concepts. Initially, participants were split into two groups, one of which (“lab group”) received an interactive laboratory activity as part of their instruction, and the other (“textbook group”) received only written instructional material. Due to comparable performance of these two groups on all of the learning outcomes measured (see Fig. 2 ), we analyze them as a single pooled sample. Then, within one week of the learning session, participants returned for an fMRI scan session during which they completed the conceptual near-transfer task: a free-body diagram (FBD) evaluation task. The near-transfer task depicted structures that fell into expert-defined categories based on the mechanical forces at play: cantilever, truss, or vertical load. Critically, participants were never made aware that these categories existed, and were not asked to categorize the stimuli at all. Rather, participants were only exposed to the underlying concepts that give rise to the categories. Our investigation differs from prior work in two notable ways. Firstly, rather than comparing completely naïve participants to advanced learners, we assess novice students during the initial phase of their learning trajectory, after they have learned about two STEM concepts in a short (1-hour) lesson. Secondly, while previous work on the initial phase of learning has focused on either inter-subject measures which do not directly probe the structure of conceptual knowledge, or object identity which includes many features including many which are not diagnostic of the relevant concepts, here we leveraged stimuli which fell into discrete concept-relevant categories which are knowable to experts and advanced learners but not to novices. This design allowed us to use the decodability of category structure from participants’ neural patterns as an indicator of participants’ deeper conceptual understanding of the underlying concepts, as shown in previous work 2 , 3 . Indeed, we found that the decodability of these conceptual categories from participants’ multivariate neural representations of the stimuli correlated with participants’ performance on more traditional assessments of learning, including quiz-style assessments focused on fact retrieval, which were administered throughout the experiment, as well as a conceptual near-transfer task involving the evaluation of free body diagrams completed during fMRI scanning. These results demonstrate that very brief exposure to new concepts can yield identifiable neural evidence of abstract knowledge acquisition, which is meaningfully related to performance on functional knowledge assessments, even in individuals with no prior domain expertise. Results Behavioral Scores: Quiz Performance. Participants’ knowledge of material covered in the learning sessions was assessed via two written quizzes at each of four timepoints (numbered T1 through T4; see Fig. 1 ). The two quizzes were designed in collaboration with engineering instructors, including authors S.D. and V.M., to probe fact retrieval of two mechanical engineering topics – Quiz 1 tested knowledge of linear forces, while Quiz 2 addressed rotational tendencies called “moments”. Mean Quiz 1 and Quiz 2 scores for each group at each timepoint are reported in Table 1 . Table 1 note: Means and standard deviations (SD) for quiz scores at each timepoint. N indicates sample size for each group as well as combined sample size for “All participants” scores and standard deviations, which were computed directly from participant-level performance data, not pre-averaged by group. “Overall” quiz score averages are computed across participants’ Quiz 1 and Quiz 2 scores for each timepoint. T4 averages include only data from participants who completed T4 follow-up quizzes. Note that for participants who completed T4, quiz scores were not separated by topic, hence the empty cells for Quiz 1 and Quiz 2 topic scores at T4. Table 1 Average quiz scores at each timepoint Timepoint Group N Quiz 1 Quiz 2 Overall Mean SD Mean SD Mean SD T1 Lab 25 64.81% 12.37% 42.69% 13.94% 53.75% 11.71% Textbook 25 66.80% 15.67% 45.60% 11.21% 56.20% 10.90% All participants 50 65.78% 13.98% 44.12% 12.64% 54.95% 11.27% T2 Lab 25 81.73% 10.48% 60.19% 14.46% 70.96% 10.27% Textbook 25 84.60% 9.12% 67.60% 10.22% 76.10% 7.88% All participants 50 83.14% 9.85% 63.82% 12.98% 73.48% 9.45% T3 Lab 25 79.23% 13.24% 56.54% 15.08% 67.88% 11.70% Textbook 25 84.00% 10.99% 65.00% 12.08% 74.50% 9.52% All participants 50 81.57% 12.31% 60.69% 14.21% 71.13% 11.09% T4 Lab 13 65.18% 15.25% Textbook 13 67.41% 13.66% All participants 26 66.30% 14.23% For both groups, quiz scores showed a clear learning curve reflected in a mixed-effects polynomial regression model with linear and quadratic effects for experimental timepoint, an interaction effect included for instruction group, and random intercepts included by participant ID (Table 2 , Fig. 2 ). On average, there was a significant positive linear effect of experimental timepoint (𝛽 = 0.36, SE = 0.13, p = 0.006), coupled with a significant negative quadratic effect (𝛽 = -0.82, SE = 0.12, p = 1.08x10 − 10 ). These effects indicate that participants improved on the quizzes overall after the learning session between T1 and T2, and then began to exhibit a downturn in performance as the elapsed time since learning increased (i.e., into T3 and T4). There was no significant effect of instruction group (lab vs. textbook) on quiz performance, neither in the form of an interaction with the timepoint effect nor in the form of a main effect of instruction group (Table 2 , Fig. 2 ). Table 2 note: Random intercepts were included for participant ID to account for the repeated-measures design. Significant effects are indicated in bold. Table 2 Polynomial regression effects for quiz scores over time and by group Parameter Estimate SE df t p (Intercept) 0.64 0.02 50.50 33.79 Lab) 0.04 0.03 50.35 1.66 0.10 Timepoint (linear) 0.36 0.13 127.77 2.82 0.01 Timepoint (quadratic) -0.82 0.12 126.41 -7.04 1.08x10 − 10 Interaction: Group x Timepoint (linear) 0.08 0.18 127.68 0.46 0.64 Interaction: Group x Timepoint (quadratic) -0.20 0.17 126.34 -1.22 0.23 Behavioral Scores: Free Body Diagram Task. During the fMRI session at T2, participants completed a concept knowledge task in which they were presented with photographs of real-world structures followed by a diagram labeled according to Newtonian forces that must be acting upon the highlighted section of the structure, and asked to judge the correctness of each diagram. Each task element consisted of a binary forced-choice response (indicating either a correctly labeled or incorrectly labeled diagram), and participants made 12 judgments per run for 8 fMRI task runs. Every stimulus item had a correctly-labeled version and an incorrectly-labeled version, and each version was shown an equal number of times across the runs, with only one version of the stimulus item shown in a given run (i.e., the incorrect and correct versions of the same stimulus were never shown together in the same fMRI run). After averaging performance on the FBD task for each subject across all their item judgments over all 8 fMRI runs, the group average (i.e., grand mean) task accuracy for each instruction group was 63.21% (SD = 10.76%) for the lab group and 68.12% (SD = 10.82%) for the textbook group. A two-sample t-test revealed that there was no significant difference in how well the lab and textbook groups performed on the FBD task ( t = -1.57, df = 45.63, p = 0.12). Coupled with the lack of an effect of group on the quiz score results, this suggests that there was no significant difference in how well the lab and textbook groups learned the relevant material during our short experimental intervention. For this reason, all subsequent analyses are performed using all the participants pooled together, without differentiating between those who received the lab or textbook lesson. Neural Analysis: Measuring Conceptual Understanding . Our overall goal for the analysis of neural data was to assess whether participants’ understanding of the conceptual relationships between the items was reflected in neural activity patterns with respect to the “expert” categorical model of the twelve stimuli. In the conceptual transfer (FBD) task administered during fMRI scanning, participants viewed a set of images of real-world structures and considered the interacting forces on a highlighted component of each structure. Unbeknownst to the participants, each of the structures fell into one of three mechanical categories (cantilever, truss, or vertical load). Importantly, participants were never explicitly instructed about these three categories. Therefore, the ability of a classifier to distinguish patterns of neural activity belonging to these three categories reflects deeper-level conceptual understanding of the newly-learned physics concepts. Here we assess the presence of this concept knowledge reflected in neural patterns in each participant, and we confirm the relationship of these patterns to learned conceptual information by assessing the correlation between the performance of this neural classifier with more traditional tests of knowledge (both explicit knowledge tests and transfer test performance). As a first step, we conducted a whole-brain searchlight analysis for each participant, computing a dissimilarity matrix (DM) of the correlation distances between each of the individual’s item-level neural responses to the stimuli at each searchlight location. These searchlight DMs were then averaged by cortical parcel using the Schaefer (2018) cortical parcellation atlas (300-parcel version) 18 , to yield a single parcel-average DM for each of the 300 cortical parcels for each participant. Having organized the neural results by cortical parcel in this way, we proceeded with the analysis by computing the extent to which expert categorical information was present first in the group-level average data. We then used the brain regions identified in the group-level analysis to threshold the individual-level data into relevant cortical parcels, and then computed “neural scores” for each participant as an indicator of the degree of concept knowledge embedded in their neural representations (their “neural score”, see prior work by Cetron and colleagues 2 ). Finally, using a regression modeling approach, we identified the relationship between participants’ neural scores and their behavioral concept knowledge test scores. Group Average Classification. As a feature reduction step, we sought to identify brain areas which were sensitive to the categorical distinctions between the stimuli in the group average neural responses. In each cortical parcel of the Schaefer (2018) 300-parcel atlas, we ran 1,000 iterations of support vector machine classification with iterative half-sample leave-one-item-per-category-out cross-validation (described in Methods). Figure 3 shows the distribution of SVM accuracy scores for the categorical model across 1,000 iterations for each parcel. Only parcels in which the average classification accuracy across 1,000 iterations was significantly greater than a permuted null model and additionally greater than 41% (one item above chance classification) were considered to contain significant information about categorical distinctions between the items in the group average. In total, 160 parcels met these criteria, shown in green in Fig. 3 . Subsequent analyses of individual performance were performed only on these parcels. Individual Classification: Neural Scores . In each of the 160 parcels identified by the group average classification analysis, we then attempted to classify the mechanical categories in the individual neural data. This analysis followed the same 1,000 iteration leave-one-item-per-category-out procedure, but in this case only one individual was held out from the training set at a time, and their data from the held out items used at testing. Thus, the classifier performance when any given subject was held out can be treated as an “individual neural score”, a measure of the separability of the categories in their own neural data. After calculating mean neural scores over 1,000 iterations of SVM classification, a linear mixed effects analysis of the neural score as a predictor of behavioral scores (quizzes, FBD performance) was performed at each of the 160 parcels included in the mask of areas defined by category separability in the group average neural activity. The linear mixed effects model for each parcel was fitted with random intercepts for each subject and each knowledge assessment type. This use of mixed effects modeling to implement multivariate regression (i.e., regression models with multiple response variables) is a common modeling strategy when the multiple response variables are considered to represent the same underlying construct but may be measured with slight differences 19 . This is especially useful when the response variables are correlated with one another because a multivariate regression model accounts for these covariances during estimation 20 – 22 . A heatmap of observed beta values for neural score as a predictor of behavioral knowledge assessment scores in these 160 parcels is shown in Fig. 4 . We also computed a null distribution of beta values by randomizing the mapping of behavioral to neural scores 1,000 times for each parcel and recording the beta values of neural score as a predictor of behavioral scores in these permuted mappings, which simulate the beta values that could occur by chance in a given parcel, considering the distributions of that parcel’s behavioral and neural scores 23 . Six parcels were identified where observed beta values for individual neural score as a predictor of behavioral score was greater than 1.65 standard deviations above the permuted null distribution. These parcels are located in the left inferior temporal gyrus (parcel 50, 𝛽 = 0.29, Z = 1.70), the left intraparietal sulcus (parcel 93, 𝛽 = 0.28, Z = 2.08), the right supramarginal gyrus (parcel 247, 𝛽 = 0.18, Z = 1.71), the right postcentral sulcus (parcel 205, 𝛽 = 0.25, Z = 2.16), and the bilateral precuneus (parcels 116, 𝛽 = 0.27, Z = 1.67; and 268, 𝛽 = 0.24, Z = 1.98). The observed beta values, Z-scores of those values in the permuted null distribution, and anatomical labels of each significant parcel from the in the 17-network Yeo atlas 24 as well as the Destrieux atlas 25 are reported in Table 3 . This result provides evidence that individuals’ neural category separability in these areas was significantly predictive of their knowledge as measured by the various behavioral assessments. For each of the six significant parcels, scatter plots of individual neural and behavioral scores with the overall regression line and random intercepts for each knowledge assessment (FBD task accuracy, Quiz 1 score, and Quiz 2 score) are shown in Fig. 5 . Table 3 note: Regression coefficients for the six parcels in which neural scores predicted behavioral scores with observed beta values at least 1.65 standard deviations outside the permuted null distribution of beta parameters. For each parcel, we report the observed beta parameter value (* = two-sample t-test against null distribution p < 0.001), the Z-score of the beta value in the permuted null distribution, the anatomical label of the parcel from the Destrieux atlas projected to the FSAverage5 surface, and the parcel location in the 17-network Yeo parcellation atlas. Table 3 Regression results for parcels where neural scores best tracked behavioral scores Parcel 𝛽 Z Destrieux Label (SUMA) Yeo Label (17 networks) 50 0.29* 1.70 L inferior temporal gyrus LH Dorsal Attention A Temporal Occipital 93 0.28* 2.08 L intraparietal sulcus/superior parietal gyrus LH Control A Intraparietal Sulcus 116 0.27* 1.67 L precuneus LH Default A Medial Prefrontal Cortex 205 0.25* 2.16 R postcentral sulcus RH Dorsal Attention B Post Central 247 0.18* 1.71 R supramarginal gyrus RH Control A Intraparietal Sulcus 268 0.24* 1.98 R precuneus RH Control C Precuneus Discussion In the present study, we have demonstrated that concept knowledge can be detected within the neural activity patterns of novice learners even after brief exposure to new concepts. Using neural scores derived from the classification of individuals’ neural representations of stimuli into relevant conceptual categories, we identified at least six areas of the brain where neural scores significantly predicted performance on traditional behavioral measures of learning, including topic quizzes and a conceptual near-transfer task. Among the areas where individual neural scores significantly predicted behavioral performance is the left IPS, which is commonly associated with processing of visuospatial information (in particular involving the estimation of magnitudes and quantities) and visuomotor action planning, guidance, and execution 26 – 29 . This result aligns with previous work including our own which implicates dorsal-stream regions, especially the IPS, in the neural representation of physics knowledge 2 , 3 , 10 , 26 , 30 . In particular, the IPS was implicated in representation of abstract physics concepts in the early learning phase described by Mason and Just 9 , one of the few other training studies (i.e., where students were instructed about a new physics concept) in this subfield of cognitive neuroscience. Neural scores also predicted physics knowledge in the ventral ITG, which is commonly associated with processing of visual feature representations that are diagnostic of object categories 12 , 31 . The literature on category learning in the ventral stream is generally centered on perceptual categories, but there is evidence from prior studies (e.g., from work by Connolly and colleagues 5 , and Haxby and colleagues 31 ) that conceptual categories are also differentiated in ventral temporal regions. As previously discussed, an important feature distinguishing the present study from other work is the choice to study novices after a brief period of learning rather than comparing participants who were advanced learners or even experts with those who were entirely naive. Our focus here is on the very early steps along the learning trajectory. Thus, it is particularly striking that in the left IPS and ventral ITG—two areas in which neural patterns distinguished engineering students from non-engineering students in the previous study 3 —behavioral test scores were significantly correlated with neural scores for novices with only about an hour of relevant training. In our previous study, the physics concept representations of advanced learners showed a posterior-to-anterior gradient in both the dorsal and ventral streams. In our current study, briefly-trained novices exhibit physics concept representations in ventral and dorsal regions that are posterior to the key regions we observed for the advanced learners in the prior study, but still anterior to any neural results we observed for naïve novice group (who received no training). Although the methodology of these two studies differed in important ways, this suggests the possibility that the anterior shift in representing physics concepts in the brain may be a progression through the learning process, although testing this directly would require a longitudinal fMRI investigation with higher temporal resolution (i.e., more time points over a much longer learning period). Other areas which showed significant prediction of neural scores included right parietal regions such as the supramarginal gyrus and postcentral sulcus, which have also been shown to play a role in motor imagery and spatial memory 32 , 33 (among other functions), and bilateral parcels in the precuneus, an area which has been associated with visuospatial memory, episodic memory retrieval, and semantic processing 34 – 36 . In particular, conceptual representations in the precuneus have been shown to contain information about amodal semantic relationships between concrete object categories 7 . These results suggest that this property may extend to more complex and implicit categorical relationships as well. It is also crucial to note that the novice participants in the present study were never explicitly instructed about the existence of the mechanical categories used in the classification analysis to derive their neural scores. Participants were not aware of any categories for the stimuli in the FBD task, nor were they ever asked to employ a categorical strategy during any of the behavioral tasks. Thus, any conceptual category information represented in participants’ neural data reflected an emergent understanding of the deeper STEM concepts implicit in those categories, rather than the simple retrieval of memorized facts. This interpretation of the neural score results is further supported by the correlation between the neural and behavioral scores: the participants who displayed greater implicit knowledge through their neural representations also displayed greater explicit knowledge through their behavioral test scores. Due to the comparable behavioral performance that we observed between the laboratory and textbook participant groups, we limited our analysis of the neural data to the full set of participants, pooled together rather than separated by learning condition. Although we initially hypothesized an advantage of hands-on learning over text-based instruction, we did not find evidence of any such distinction in this very brief intervention. It is possible that the advantage of hands-on instruction would have become more observable over longer timescales of learning. Future research may attempt to explore how hands-on experience could shape neural representations of physics concepts over a longer learning period, and how any such neural effects of laboratory-based learning may relate to pencil-and-paper knowledge assessments. The decodability of subtle shifts in understanding on the individual level demonstrated by the present study would suggest that these methods could be well-suited to such comparisons of longer-term classroom interventions. The results of the present study show that through multivariate pattern analysis methods, it is possible to detect individual-level differences in conceptual understanding in novices with brief exposure (one hour) to new STEM concepts. By leveraging categorical distinctions between items in the stimulus set which are apparent to experts and advanced learners but not to novices, we derived neural scores using SVM classification. In six cortical parcels in the brain, these neural scores predicted individual performance on behavioral assessments of learning, providing evidence that multivariate neural activity patterns in these regions reflected conceptual understanding. Methods and insights from this line of work may help shape educational approaches in the future, such as providing an additional modality with which to examine learning for applied curriculum development research. Using multimodal data-driven approaches, future research can yield deeper insight into the early stages of learning, where novice students begin to build the conceptual scaffolds that are essential for understanding complex topics in science, math, and other domains of knowledge. Methods Participants . Fifty-three Dartmouth College students participated in this study. Three were excluded: one due to excessive motion in the scanner, one due to a stimulus presentation error, and one due to an illness that the participant self-reported to have affected their memory after the experiment had concluded. The resulting sample had an N of 50 (32 female; mean age at T1 = 19.85 years, SD = 1.17). All participants were sent a follow-up knowledge assessment at least one month after the scan session. Nineteen participants responded to this online survey. All participants were fluent English speakers who completed an eligibility survey verifying they had little-to-no prior coursework experience in physics or mechanical engineering. Specifically, we excluded any participants with more than 1 semester of introductory college-level physics experience (or AP equivalent). We also excluded any participants who were majoring in engineering or physics, even if they had not yet completed any advanced engineering or physics coursework. These were the same exclusion criteria used to recruit the previous studies’ novice group. Participants provided informed consent on each day of data collection, and were offered either curricular extra credit points or a small monetary compensation. All procedures were approved by the Dartmouth Committee for the Protection of Human Subjects (CPHS E (#5): IRB00006768; https://www.dartmouth.edu/cphs/ ). Materials. All materials and stimuli used in the present experiment that do not contain copyrighted content can be viewed and downloaded on the Open Science Framework in our project repository by following this link: https://osf.io/ub4z3/?view_only=2ef76a55f5f2405cb621c54090dedb7a . During the behavioral session of the experiment, participants learned about static mechanics, an area of physics relevant to mechanical engineering. All participants were given a primer on the basic premise of static mechanics, consisting of a short powerpoint presentation (8 slides) explaining that unmoving objects in the world are able to maintain equilibrium (i.e., remain unmoving) due to the balance of forces exerted both by and upon the objects. This primer was created by the authors and validated by authors S.D. and V.M., who have domain expertise in engineering physics. These slides are included in the OSF project repository. The static mechanics primer was shown to participants twice: first as their initial introduction to the learning portion of the behavioral session (after baseline assessments were completed), and again prior to the fMRI task during the fMRI session. During the learning portion of the behavioral session following the static mechanics primer, participants were given more detailed introductions to each of the two topics within static mechanics that they would be learning about: linear forces (Topic 1) and rotational tendencies called “moments” (Topic 2). The main learning portion for each topic would consist of interactive instructional materials that varied by experimental condition (a “textbook” condition and a “lab” condition), but prior to each of these interactive sessions, participants would review a set of slides containing text, pictures, and video clips describing the essential principles of each topic. These slides were created by the authors and utilized content from an EdX course on static mechanics taught by author V.M. Textbook condition instruction materials. After the first topic area introduction, participants in the textbook condition reviewed a second set of slides accompanied by a printed inquiry sheet. Using the slides, textbook participants reviewed a series of specific examples illustrating concepts central to the topic area. As they reviewed the examples, they occasionally encountered question prompts, which were also printed on their inquiry sheets. Participants marked their answers to the questions on the inquiry sheet, and then reviewed those answers on subsequent slides, tracking the accuracy of their responses throughout the instruction period. Participants were instructed not to be concerned about the accuracy of their responses to these inquiry questions (we did not analyze their performance in this phase of instruction). After completing the example slides and inquiry sheet for Topic 1, they repeated the topic area introduction and example slide/inquiry sheet procedure for Topic 2. Lab condition instruction materials. Lab participants completed a procedure that paralleled the textbook procedure as closely as possible, except that they reviewed the topic-related examples using a series of activities rather than a series of slides. The activities were designed based on common physics and engineering laboratory projects, such as the construction and deconstruction of a truss from plastic drinking straws to illustrate tension and compression in a truss system (Topic 1). Participants received instructions for each activity on a printed sheet, completing the activities on their own except when the experimenter was needed to assist with the execution of an activity. As in the textbook condition, participants were periodically prompted to answer inquiry questions using an accompanying inquiry sheet, and tracked their progress as they completed the examples. The lab activities and inquiry questions were paired as closely as possible with the textbook examples and inquiry questions, such the two conditions illustrated the same concepts in the same order as one another, and such that the inquiry questions could be phrased as similarly as possible across conditions. This section of the behavioral session constituted the only difference between the lab and textbook conditions. Copies of inquiry materials for both conditions (excluding those containing copyrighted content) can be found on the OSF project repository. Topic Quizzes. All participants completed knowledge assessments in the form of two topic-specific quizzes focused on fact retrieval and administered across at least three timepoints in the experiment: first as an initial baseline measurement at the very beginning of the experiment (T1); second as a test of learning after the instruction session (T2); and third as a repeat knowledge test after the fMRI scan session (T3). All participants were sent the quizzes a fourth time at least one month after the scan session (T4), but not every participant completed this final assessment. At each timepoint, the quizzes participants completed were identical (i.e., the material being tested was always the same). Participants were never given feedback regarding their performance on the quizzes. We designed the fact-retrieval quizzes for each of the two mechanical engineering topics in collaboration with engineering instructors, including authors S.D. and V.M. Assessments were multiple-choice and involved both verbal and visual questions probing concept knowledge. Performance on each topic was analyzed separately, due to disparities in difficulty by topic (topic 2 was considerably harder for students to comprehend than topic 1). Copies of the quiz materials can be found on the OSF project repository. Free body diagram task (fMRI near-transfer task). During fMRI scanning, participants completed a conceptual near-transfer task involving the evaluation of photographs of real-world structures labeled as free body diagrams (FBDs). This constituted a near-transfer task because participants had never seen this task before, and had to apply their recently-learned concept knowledge associated with linear forces and moments to complete the task successfully. Stimulus photographs for the FBD task were selected from the set of 24 images utilized by the previous study 2 . In each of these images, which included lampposts, bridges, awnings, and other similar structures, mechanical engineering experts (authors S.D. and V.M.) labeled a particular component of each structure by outlining it in red. Each of the highlighted components fell into one of three mechanical categories (cantilever, truss, or vertical load). Importantly, participants were never explicitly instructed about these three categories. Expert model. To query the degree of conceptual information present in a given representation of the 12 real-world structure stimuli, we compared representations to an expert model of the similarity between the stimulus items. Specifically, author S.D. compared a larger set of structure stimuli (originally 24 items) in a series of pairwise similarity comparisons, focusing on the mechanical similarity between the items. This pairwise similarity model yielded three categories of mechanical structures: vertical loads, trusses, and cantilevers. In the present study, we selected four representative items from each of the three categories, resulting in the 12-item stimulus set used here. (For the full 24-item stimulus set, see previous study 2 , 3 .) Neural representations for each participant were constructed based on their neural pattern responses to each of these 12 stimuli. Then, the degree of concept information present in a given representation was identified by fitting a categorical classifier to the participant’s representation, where the true category labels correspond to the mechanical structure categories identified in the expert model. Behavioral similarity probe (not analyzed). In addition to the quizzes, participants’ representations of concept-relevant stimuli were also queried via pairwise similarity ratings of the 12 stimulus items they would review in the FBD task. These ratings were made at each of the same timepoints as the quiz assessments. We assessed the correspondence between these explicit similarity ratings and the expert categorical model using the same classification procedure as was applied to the neural similarity matrices (see Methods subheading “Multivariate Analysis”). The correlations between these classifier results and participants’ quiz scores for each time point are summarized in Supplementary Fig. 1. However, since these classifier scores neither correlate with individual quiz score nor show a clear learning trajectory over time, we have insufficient evidence that this task as administered was able to measure the relevant constructs. For example, rather than considering mechanical similarity, these novice participants may have relied on more surface-level similarities that were deliberately built into the stimulus set. As such, we do not discuss the results from the behavioral similarity ratings in the present paper. Procedure. Participants completed two sessions in this experiment: a behavioral session for concept instruction, and an fMRI scan session. An overview of activities completed during each session, as well as the online follow-up which occurred one month later, is shown in Fig. 1 . The behavioral session began with participants giving their informed consent to participate in the study, and then completing baseline assessments on the topic quizzes and provided information on their prior experience in any physics and engineering courses. Participants were expected to be unfamiliar with the concepts assessed in the quizzes at this timepoint (T1). After completing baseline assessments, all participants reviewed the static mechanics primer slides, and then began the topic-specific learning sessions. First, participants learned about linear forces (topic 1). They reviewed the topic 1 introductory slides, and then completed the interactive learning condition to which they had been randomly assigned: either the textbook (computer-based) interactive materials condition, or the laboratory interactive materials condition. The textbook condition involved reviewing slides that prompted the participant to make predictions about specific examples involving linear forces, and then learning about the solutions on subsequent slides. The laboratory condition involved completing activities involving analogous linear force examples, and learning about the solutions after completing the activities. After concluding with the topic 1 materials (and taking an optional short break), participants repeated the procedure with the materials for topic 2: rotational tendencies (known as “moments”). Once participants had reviewed all learning materials, they proceeded to the post-learning quiz assessments. These quizzes were identical to the baseline quizzes, but at this timepoint (T2) participants now had learned about the topics being assessed. After the T2 assessments, the behavioral session was complete, and participants were compensated for their time. fMRI session. Within a maximum of 1 week of the behavioral session, participants returned to complete an fMRI scan session. Before scanning, participants were shown the static mechanics primer again, as well as an introduction to the FBD task they would be performing in the fMRI scanner. Upon entering the fMRI scanner, participants reviewed these primer materials once more, and then were familiarized with the 12 stimulus items they would be evaluating during the FBD task. After familiarization and structural scans, participants completed 8 runs of the FBD task during functional scanning. The stimuli and design for this task are described in detail in previous work 2 . Participants reviewed the FBD stimuli individually, first assessing each stimulus, and then responding to the prompt via button press after a jittered fixation. Analyzed functional data were drawn from the consideration period prior to button presses. Once the functional scans were complete, participants filled out the assessment quizzes again, now at the post-scan timepoint (T3). After the T3 assessments, the fMRI session was complete, and participants were compensated for their time. Long-term recall follow-up. After completing the required behavioral and fMRI sessions, participants were invited to complete a 1-month follow-up evaluation (timepoint T4) via online survey, which only some participants completed. Participants who completed this assessment were compensated for this additional time spent. fMRI Data Acquisition. A 3 Tesla Siemens PRISMA fMRI scanner with a 32-channel head coil was used to acquire the fMRI data. A single high-resolution T1-weighted anatomical scan (FOV = 240mm, Flip Angle = 8°, 192 slices) and eight functional runs were performed for each participant. Each 2D EPI sequence consisted of 186 measurements with a 240mm field of view to provide full brain coverage over 52 slices (Flip Angle = 75°; TE = 35 ms; TR = 2000 ms; 2.5mm 3 voxels). In the scanner, stimuli were presented using PsychoPy version 1.84.2 37 (using Python 2.7). Image Preprocessing and Univariate Analyses. Brain images were preprocessed using the FSL FEAT software package 38 . The FSL brain extraction tool (BET) was used to skull-strip each T1-weighted anatomical image. Skull-stripping, motion correction, slice timing correction, and highpass temporal filtering were also applied to each functional EPI volume. Finally, the functional runs were registered to the participant’s individual anatomical space using the FSL linear registration tool 39 . A first-level univariate regression model using the general linear model was then calculated at the item level, such that beta-value estimates for each stimulus were generated separately for each run for each participant. Then, a second-level GLM combined these estimates across functional runs, yielding a single contrast estimate for each of the twelve items. All beta-value estimates were then aligned to the individual’s T1 volume and resampled to 2mm 3 using the FSL mathematical manipulation tool. For each subject, cortical surface reconstructions were generated for the T1-weighted anatomical image using FreeSurfer’s recon-all toolbox 40 and fitted to standard mesh grids based on an icosahedron with 32 linear divisions, yielding 20,484 nodes for the whole-brain cortical surface in Surface Mapping (SUMA) format 41 . Sulcal alignment of each participant’s cortical surface to the FreeSurfer average brain 42 was performed to allow anatomical correspondence between surface nodes across participants. We also performed the same transformation on an atlas of 300 parcels defined according to abrupt transitions in functional connectivity patterns in resting state fMRI 18 to create a mapping between the parcellation and the 20,484-node surface space. Multivariate Analysis. We used the PyMVPA toolbox 43 to conduct a whole-brain searchlight analysis for each participant using spherical 5mm searchlights. In each searchlight, the correlation distance between the individual’s item-level betas for each of the twelve stimuli was computed to form a dissimilarity matrix (DM) for each node and its surrounding neighborhood. These DMs were then averaged for all surface nodes belonging to each cortical parcel in the Schaefer (2018) 300-parcel atlas 18 , yielding a single average DM for each parcel. As a feature reduction step, we sought to identify parcels where categorical information about the twelve items (cantilevers vs. vertical loads vs. trusses) was represented in the neural activity of the group as a whole. Each parcel-level DM was projected into two dimensions using multidimensional scaling (MDS), and a support vector machine (SVM) classifier with a radial basis function kernel was employed using half-sample leave-one-item-per-category-out cross validation for 1,000 iterations and the mean accuracy score was taken. On each iteration, the twelve items were randomly assigned to four folds such that each fold consisted of a test set with one item from each of the three categories. On each fold, the model was trained on nine rows (leaving one item per category out) of an average DM constructed from a randomly drawn half (25) of the participants, then tested on the three held-out items in the average DM of the held-out participants. The division of both participants and items into the training and test sets was randomized on each iteration. These steps were implemented in Python using the scikit-learn package 44 . At each parcel, we also performed a permutation test by shuffling the training data labels and repeating the same procedure for another 1,000 iterations. The permutation test results for relevant parcels are shown in Supplementary Fig. 2. Out of 300 parcels, we identified 160 in which the mean SVM accuracy for the categorical model was both significantly greater than the permuted null model and additionally greater than 41%, indicating that the classifier could reliably categorize the held-out items at least 8% (the value of one item) greater than chance (33%). These parcels were treated as a binary mask in subsequent analyses. In these 160 parcels which had been shown to be sensitive to categorical distinctions between the stimuli in the group average , we then probed individual parcel DMs using the same leave-one-item-per-category-out cross-validation procedure to assess the degree to which each individual’s patterns of neural activity in each parcel reflected the relationships between the items. As before, on each fold of the 1,000 iterations, the SVM classifier was trained on nine out of twelve items in the average DM of the training set. However, rather than holding out a randomly selected half of the sample, we trained the classifier on all but one participant, then tested on the three held-out items for the remaining participant. The mean classification accuracy for each individual was taken to be that individual’s “neural score” for each parcel. In each of the 160 parcels identified as significant at the group level, we performed a linear mixed effects analysis of the relationship between participants’ behavioral scores (quiz scores, FBD performance) and neural scores. These models were created in R 45 using the lme4 package 46 . We entered neural score as a fixed effect, and included behavioral score type (Quiz 1, Quiz 2, or FBD performance) and intercepts for each participant as random effects. We then computed a null distribution of beta values by shuffling the mapping of behavioral to neural scores and recording the beta values of neural score as a predictor of behavioral scores over 1,000 permutations. We define significant prediction as a parcel where the observed beta value for neural score as a predictor of behavioral scores is greater than 1.65 standard deviations above this permuted null distribution of beta values, a critical Z value which corresponds to an alpha level of 0.05. Declarations Competing Interests The authors declare no competing interests. Author Contribution Authors J.S.C. and M.E.H. contributed equally to the work and are co-first authors of the present manuscript. J.S.C. and D.J.M.K. designed and conducted the study. J.S.C., M.E.H., and D.J.M.K. performed the analysis and wrote the manuscript. S.G.D. and V.V.M. contributed to study design (procedure design, assessment design, and stimulus creation). All authors reviewed the manuscript. Acknowledgement D.J.M.K. and M.E.H. were supported by a National Science Foundation award (DRL-2201304). The authors thank J. Hayes for his assistance with data collection and the members of the Cognitive Neuroscience of Learning Lab at Dartmouth for scientific discussions and feedback on this manuscript. We thank T. Sackett and the Dartmouth Brain Imaging Center for facilitating fMRI research at Dartmouth. Data Availability The datasets generated during the current study are available from the corresponding author upon reasonable request, where participant data privacy policies permit. The code generated during the current study is available from the corresponding author upon reasonable request. References Lake, B. M., Linzen, T. & Baroni, M. Human few-shot learning of compositional instructions. arXiv.org https://arxiv.org/abs/1901.04587v2 (2019). Cetron, J. S. et al. Decoding individual differences in STEM learning from functional MRI data. Nature Communications 10, 2027 (2019). Cetron, J. S. et al. Using the force: STEM knowledge and experience construct shared neural representations of engineering concepts. npj Sci. Learn. 5, 6 (2020). Chao, L. L., Weisberg, J. & Martin, A. Experience-dependent Modulation of Category-related Cortical Activity. Cerebral Cortex 12, 545–551 (2002). Connolly, A. C. et al. The Representation of Biological Classes in the Human Brain. J. Neurosci. 32, 2608–2618 (2012). Coutanche, M. N. & Thompson-Schill, S. L. Rapid consolidation of new knowledge in adulthood via fast mapping. Trends in Cognitive Sciences 19, 486–488 (2015). Fairhall, S. L. & Caramazza, A. Brain Regions That Represent Amodal Conceptual Knowledge. J. Neurosci. 33, 10552–10558 (2013). Martin, A. The Representation of Object Concepts in the Brain. Annual Review of Psychology 58, 25–45 (2007). Mason, R. A. & Just, M. A. Physics instruction induces changes in neural knowledge representation during successive stages of learning. NeuroImage 111, 36–48 (2015). Mason, R. A. & Just, M. A. Neural Representations of Physics Concepts. Psychological Science 27, 904–913 (2016). Weisberg, J., van Turennout, M. & Martin, A. A Neural System for Learning about Object Function. Cerebral Cortex 17, 513–521 (2007). Kriegeskorte, N., Mur, M. & Bandettini, P. Representational similarity analysis - connecting the branches of systems neuroscience. Frontiers in Systems Neuroscience 2, (2008). Coutanche, M. & Thompson-Schill, S. Informational connectivity: identifying synchronized discriminability of multi-voxel patterns across the brain. Frontiers in Human Neuroscience 7, (2013). Hasson, U., Nir, Y., Levy, I., Fuhrmann, G. & Malach, R. Intersubject Synchronization of Cortical Activity During Natural Vision. Science 303, 1634–1640 (2004). Chase, W. G. & Simon, H. A. Perception in chess. Cognitive Psychology 4, 55–81 (1973). Chi, M. T. H., Feltovich, P. J. & Glaser, R. Categorization and representation of physics problems by experts and novices. Cognitive Science 5, 121–152 (1981). Meshulam, M. et al. Neural alignment predicts learning outcomes in students taking an introduction to computer science course. Nature Communications 12, (2021). Schaefer, A. et al. Local-Global Parcellation of the Human Cerebral Cortex from Intrinsic Functional Connectivity MRI. Cereb Cortex 28, 3095–3114 (2018). Butler, E. A., Gross, J. J. & Barnard, K. Testing the effects of suppression and reappraisal on emotional concordance using a multivariate multilevel model. Biological Psychology 98, 6–18 (2014). Hox, J., Moerbeek, M. & Schoot, R. van de. Multilevel Analysis: Techniques and Applications, Third Edition . (Routledge, New York, 2017). doi: 10.4324/9781315650982 . Twisk, J. W. R. Applied Mixed Model Analysis: A Practical Guide . (Cambridge University Press, Cambridge, 2019). doi: 10.1017/9781108635660 . Snijders, T. A. B. & Bosker, R. J. Chapter 16: Multivariate Multilevel Models. in Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling 282–288 (Sage Publishers, London, 2012). Giacalone, M. & Alibrandi, A. Overview and Main Advances in Permutation Tests for Linear Regression Models. JMSS 5, (2015). Thomas Yeo, B. T. et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of Neurophysiology 106, 1125–1165 (2011). Destrieux, C., Fischl, B., Dale, A. & Halgren, E. Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature. NeuroImage 53, 1–15 (2010). Ansari, D. Effects of development and enculturation on number representation in the brain. Nat Rev Neurosci 9, 278–291 (2008). Grefkes, C. & Fink, G. R. REVIEW: The functional organization of the intraparietal sulcus in humans and monkeys. Journal of Anatomy 207, 3–17 (2005). Simon, O., Mangin, J.-F., Cohen, L., Le Bihan, D. & Dehaene, S. Topographical Layout of Hand, Eye, Calculation, and Language-Related Areas in the Human Parietal Lobe. Neuron 33, 475–487 (2002). Simon, O. et al. Automatized clustering and functional geometry of human parietofrontal networks for language, space, and number. NeuroImage 23, 1192–1202 (2004). Kontra, C., Lyons, D. J., Fischer, S. M. & Beilock, S. L. Physical Experience Enhances Science Learning. Psychol Sci 26, 737–749 (2015). Haxby, J. V. et al. A Common, High-Dimensional Model of the Representational Space in Human Ventral Temporal Cortex. Neuron 72, 404–416 (2011). Kaufmann, L. et al. A developmental fMRI study of nonsymbolic numerical and spatial processing. Cortex 44, 376–385 (2008). Russ, M. O., Mack, W., Grama, C.-R., Lanfermann, H. & Knopf, M. Enactment effect in memory: evidence concerning the function of the supramarginal gyrus. Exp Brain Res 149, 497–504 (2003). Cavanna, A. E. & Trimble, M. R. The precuneus: a review of its functional anatomy and behavioural correlates. Brain 129, 564–583 (2006). Schott, B. H. et al. Gradual acquisition of visuospatial associative memory representations via the dorsal precuneus. Human Brain Mapping 40, 1554–1570 (2019). Trimble, M. R. & Cavanna, A. E. Chapter 3.7 The role of the precuneus in episodic memory. in Handbook of Behavioral Neuroscience (eds. Dere, E., Easton, A., Nadel, L. & Huston, J. P.) vol. 18 363–377 (Elsevier, 2008). Peirce, J. et al. PsychoPy2: Experiments in behavior made easy. Behav Res 51, 195–203 (2019). Jenkinson, M., Beckmann, C. F., Behrens, T. E. J., Woolrich, M. W. & Smith, S. M. FSL. NeuroImage 62, 782–790 (2012). Smith, S. M. Fast robust automated brain extraction. Human Brain Mapping 17, 143–155 (2002). Fischl, B. & Dale, A. M. Measuring the thickness of the human cerebral cortex from magnetic resonance images. Proceedings of the National Academy of Sciences 97, 11050–11055 (2000). Saad, Z. S. & Reynolds, R. C. SUMA. NeuroImage 62, 768–773 (2012). Fischl, B., Sereno, M. I., Tootell, R. B. H. & Dale, A. M. High-resolution intersubject averaging and a coordinate system for the cortical surface. Hum. Brain Mapp. 8, 272–284 (1999). Hanke, M. et al. PyMVPA: a Python Toolbox for Multivariate Pattern Analysis of fMRI Data. Neuroinformatics 7, 37–53 (2009). Pedregosa, F. et al. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research 12, 2825–2830 (2011). Team, R. C. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria (2021). Bates, D., Mächler, M., Bolker, B. & Walker, S. Fitting Linear Mixed-Effects Models using lme4. Preprint at https://doi.org/10.48550/arXiv.1406.5823 (2014). Additional Declarations No competing interests reported. Supplementary Files suppfig12025.png suppfig22025.png Statics2MSsuppfigures07082025.docx Cite Share Download PDF Status: Published Journal Publication published 22 Feb, 2026 Read the published version in npj Science of Learning → Version 1 posted Editorial decision: Revision requested 15 Sep, 2025 Reviews received at journal 05 Sep, 2025 Reviews received at journal 14 Aug, 2025 Reviewers agreed at journal 08 Aug, 2025 Reviewers agreed at journal 18 Jul, 2025 Reviewers invited by journal 17 Jul, 2025 Editor assigned by journal 16 Jul, 2025 Submission checks completed at journal 14 Jul, 2025 First submitted to journal 27 Jun, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6992513","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":488542537,"identity":"0ba6355f-5dfc-46b6-8124-d7f1f87d4d52","order_by":0,"name":"Joshua S. Cetron","email":"","orcid":"","institution":"Harvard University","correspondingAuthor":false,"prefix":"","firstName":"Joshua","middleName":"S.","lastName":"Cetron","suffix":""},{"id":488542538,"identity":"7c1881da-1472-4e5b-8684-b3e7cda935bf","order_by":1,"name":"Megan E. Hillis","email":"","orcid":"","institution":"Dartmouth College","correspondingAuthor":false,"prefix":"","firstName":"Megan","middleName":"E.","lastName":"Hillis","suffix":""},{"id":488542539,"identity":"0531fdd6-94c5-4a85-9830-cac91278a6cf","order_by":2,"name":"Solomon G. Diamond","email":"","orcid":"","institution":"Dartmouth College","correspondingAuthor":false,"prefix":"","firstName":"Solomon","middleName":"G.","lastName":"Diamond","suffix":""},{"id":488542540,"identity":"89139ec5-877f-4b38-bb05-fb537f23b039","order_by":3,"name":"Vicki V. May","email":"","orcid":"","institution":"Dartmouth College","correspondingAuthor":false,"prefix":"","firstName":"Vicki","middleName":"V.","lastName":"May","suffix":""},{"id":488542541,"identity":"aac67118-0820-4dce-a6cb-0728e5e6585b","order_by":4,"name":"David J. M. Kraemer","email":"data:image/png;base64,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","orcid":"","institution":"Dartmouth College","correspondingAuthor":true,"prefix":"","firstName":"David","middleName":"J. M.","lastName":"Kraemer","suffix":""}],"badges":[],"createdAt":"2025-06-27 14:23:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6992513/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6992513/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41539-025-00394-3","type":"published","date":"2026-02-22T15:57:52+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":87334687,"identity":"cd47b52c-2497-460b-97ed-0bc683c37f3d","added_by":"auto","created_at":"2025-07-22 20:16:55","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":721374,"visible":true,"origin":"","legend":"\u003cp\u003eExperiment and analysis procedure. Participants completed behavioral knowledge assessments at four timepoints: immediately before learning (T1), immediately after learning (T2), prior to the fMRI session a maximum of one week later (T3), and as a follow-up one month later (T4). Neural data collected at timepoint T3 was used to derive neural scores. At T1, participants were assigned to one of two instructional groups, but due to comparable quiz performance between the two groups, we analyzed them as a single pooled sample in all subsequent analyses. We used support vector machine (SVM) classification to analyze the ability of multivariate neural activity patterns to predict behavioral outcomes from timepoint T3. SVM analysis was run across the brain within each of the cortical parcels within the Schaefer (2018) cortical parcellation atlas (300-parcel version).\u003c/p\u003e","description":"","filename":"figure12025.png","url":"https://assets-eu.researchsquare.com/files/rs-6992513/v1/f6da482f33862b0ba027521f.png"},{"id":87334489,"identity":"bc4dad3d-655f-4323-a770-e8fd21230527","added_by":"auto","created_at":"2025-07-22 20:08:55","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":593101,"visible":true,"origin":"","legend":"\u003cp\u003eModel-predicted quiz score learning curves compared to raw data. Model predictions from the polynomial regression showing the quadratic learning curves (bold lines with 95% confidence intervals in shaded bands) are displayed for each instruction group. Model-estimated learning curves are superimposed over raw average quiz score data for lab-instruction participants (orange) and textbook-instruction participants (purple) at each experimental timepoint. Points and smoothed histogram violins are jittered for visualization only; all data comes from discrete timepoints T1 through T4. Results from the regression model show that the learning curve’s quadratic effect was negatively-signed and statistically significant (indicated by the upper bracket annotation and stars, \u003cem\u003ep \u003c/em\u003e\u0026lt; 0.001). Participants learned the concepts as the experiment progressed initially, and then as the time increased from the content lessons (between T1 and T2), participants showed some forgetting as indicated by the downturn of the quadratic effect around T3 and into T4. By contrast, we detected no significant interaction modifying this effect by instruction group (indicated by the lower bracket annotation; “ns” = “not significant”), nor did we detect a significant main effect of instruction group. This suggests that instruction format did not have a significant impact on learning over the timecourse of the experiment. We therefore analyzed the two groups together as a single population for the remaining analyses.\u003c/p\u003e","description":"","filename":"fig22025.png","url":"https://assets-eu.researchsquare.com/files/rs-6992513/v1/8a7ec875bf4c7eb2f2f9ce31.png"},{"id":87334491,"identity":"de7d797b-d572-4c43-9d67-0ad3204fb022","added_by":"auto","created_at":"2025-07-22 20:08:55","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1729631,"visible":true,"origin":"","legend":"\u003cp\u003eSelected parcels from group average classification.\u003cem\u003e \u003c/em\u003eGroup average classification accuracy over 1,000 iterations of leave-one-item-per-category-out SVM classification was performed in each cortical parcel of the Schaefer (2018) 300-parcel atlas. (a.) The distribution of scores in all 300 parcels is shown. The solid line indicates performance of the permuted null model, in which the training set labels were scrambled before classification (“true chance”). The dashed line indicates performance greater than 41%, indicating that the classifier could reliably categorize the held-out items at least 8% (the value of one item) greater than chance (33%). Green bars indicate scores greater than this 41% criterion. (b.) Parcels with mean classification scores greater than 41% were used as a binary mask for subsequent analyses, as shown on the semi-inflated cortical surface.\u003c/p\u003e","description":"","filename":"fig32025.png","url":"https://assets-eu.researchsquare.com/files/rs-6992513/v1/290a5e90a68f8ef9892bd103.png"},{"id":87334690,"identity":"e660563f-fba7-429c-851d-70442f812447","added_by":"auto","created_at":"2025-07-22 20:16:55","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":2658829,"visible":true,"origin":"","legend":"\u003cp\u003eLinear mixed model results in positively predictive parcels.\u003cem\u003e \u003c/em\u003eResults from linear mixed model analysis of individual neural score as a predictor of behavioral scores in each of the 160 parcels which survived group-level thresholding are displayed on a semi-inflated cortical surface. At each positively predictive parcel, the observed beta value reflects the extent to which category separability in the individual’s neural activity predicted performance on the three behavioral tasks (Quiz 1, Quiz 2, FBD task). At each parcel, we also conducted a permutation test by repeating this regression 1000 times with the mapping of behavioral to neural scores randomized, to simulate the distribution of beta values that could occur by chance given the data. Six parcels with observed beta values at least 1.65 standard deviations (significance level of 0.05) above this permuted null distribution are indicated.\u003c/p\u003e","description":"","filename":"fig42025.png","url":"https://assets-eu.researchsquare.com/files/rs-6992513/v1/68ba86440243d6d89636fe37.png"},{"id":87334517,"identity":"af11d1b2-0db4-4bfc-b4fa-c9154ee26fbe","added_by":"auto","created_at":"2025-07-22 20:08:55","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1523701,"visible":true,"origin":"","legend":"\u003cp\u003eLinear mixed effects regression in significant parcels. Neural score at each significant parcel predicted concept knowledge in a linear mixed-effects model. The overall regression line is plotted as the black line over the behavioral score data for each knowledge assessment (FBD task accuracy, Quiz 1 score, and Quiz 2 score). The colored lines represent the individualized regression lines for each outcome measure, which deviate from the overall regression line according to the random intercepts extracted from the model.\u003c/p\u003e","description":"","filename":"fig52025.png","url":"https://assets-eu.researchsquare.com/files/rs-6992513/v1/c57ffd7269700c74f23e5904.png"},{"id":103251487,"identity":"9be37067-478a-4faa-b1d0-24867e60fb51","added_by":"auto","created_at":"2026-02-23 16:09:22","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8383034,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6992513/v1/6e364d94-296c-4e37-ae90-5f086a923c3d.pdf"},{"id":87334688,"identity":"ac593790-a477-4099-8a57-9f51cf8fbd6e","added_by":"auto","created_at":"2025-07-22 20:16:55","extension":"png","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":210640,"visible":true,"origin":"","legend":"","description":"","filename":"suppfig12025.png","url":"https://assets-eu.researchsquare.com/files/rs-6992513/v1/22fd22f4415d46249f19aec0.png"},{"id":87334493,"identity":"c68be20a-6a7d-4ff1-a388-0a3f3671d30f","added_by":"auto","created_at":"2025-07-22 20:08:55","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":682088,"visible":true,"origin":"","legend":"","description":"","filename":"suppfig22025.png","url":"https://assets-eu.researchsquare.com/files/rs-6992513/v1/7891ad6ac8f24bee03802f06.png"},{"id":87335243,"identity":"e1626331-ab38-45e2-9b0c-c345c05113c9","added_by":"auto","created_at":"2025-07-22 20:24:55","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":257850,"visible":true,"origin":"","legend":"","description":"","filename":"Statics2MSsuppfigures07082025.docx","url":"https://assets-eu.researchsquare.com/files/rs-6992513/v1/b06379f7563b90afda2216fd.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Neural patterns reflect conceptual grasp of novice students following first class learning in physics","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAcquiring new abstract concepts is one of the most remarkable abilities of the human brain. Unlike most nonhuman animals, humans can quickly generalize abstract principles from brief exposure to sparse examples\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Correspondingly, our education system is structured in large part around teaching students to organize and apply a wide variety of abstract concepts.\u003c/p\u003e\u003cp\u003eBut how do we know when we have acquired a new concept? In traditional learning environments, concept acquisition is often probed with explicit tests of knowledge, such as a written exam testing fact-retrieval. Once some knowledge has been acquired, more practical tests of knowledge may involve transfer tasks: tasks that require novel application of the learned concepts to complete. But both explicit knowledge tests and transfer tests of knowledge still require us to infer that a student understands a concept from their behavior in a limited circumstance: their performance on the exam or on the conceptual transfer task. While useful, these sources of inference are inherently subject to noise from several sources (e.g., sampling of test items, learner performance on a given day in a given testing situation, etc.). To draw from an additional source of data regarding the knowledge a learner has acquired, it can be informative to inspect the human concept acquisition system itself: the brain.\u003c/p\u003e\u003cp\u003ePrior research on the neural basis of concept learning\u0026mdash;particularly STEM concept learning\u0026mdash;has shown that patterns of neural activity can indeed reflect abstract concept knowledge\u003csup\u003e\u003cspan additionalcitationids=\"CR3 CR4 CR5 CR6 CR7 CR8 CR9 CR10\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. These abstract representations can be identified using multivariate representational analyses of functional neuroimaging data, such as representational similarity analysis\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e, informational network analysis\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e, informational connectivity analysis\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e, and inter-subject correlation analysis\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Many of these studies investigating the neural representations of learned concepts compare two groups of individuals with substantially different levels of knowledge: advanced learners, who are expected to have acquired the concepts already, and novices, who are na\u0026iuml;ve to the concepts. Comparing novices to more advanced learners (or even experts) allows for the detection of larger differences in concept knowledge representations, since expertise has been shown to affect conceptual organization across a variety of domains (e.g., chess, as in seminal work by Chase and Simon\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e; also in physics\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e).\u003c/p\u003e\u003cp\u003eHowever, such between-group comparisons of concept knowledge representations do not inform us about the trajectory of concept learning, because the two groups are typically situated at either end of the learning process. Characterizing the learning process by identifying signs of early conceptual understanding is particularly important in STEM education because of the cumulative nature of many STEM topics (e.g., understanding object mechanics requires first understanding Newtonian force). One study by Mason and Just\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e demonstrated that as participants were exposed to incremental information about how simple mechanical systems (such as a fire extinguisher) work, patterns of neural activity evoked by the individual objects shifted from primarily visual areas to parietal and frontal areas associated with functions that could support deeper understanding such as mental animation. These results provide compelling evidence that relatively short training can evoke functional changes in novices\u0026rsquo; conceptual representations at the group level. However, it remains untested whether neural indicators of concept knowledge can be used to capture individual variation in \u003cem\u003ehow well\u003c/em\u003e novice learners grasp new STEM concepts upon their first exposure to a topic.\u003c/p\u003e\u003cp\u003eOne study by Meshulam and colleagues\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e followed a single group of STEM students longitudinally as they progressed through an introductory undergraduate computer science course, measuring neural responses to recorded lecture videos at various timepoints during the learning period. They found that students whose neural activity patterns looked more like the group\u0026rsquo;s average neural activity performed better on the course\u0026rsquo;s final exam. However, this study did not specifically examine the initial concept-acquisition phase of learning, instead focusing on learning outcomes after a full academic term of study. Moreover, as the focus of this study was on the neural alignment of students to each other and to their expert instructors\u0026mdash;indexed by inter-subject neural correlation\u0026mdash;the method did not allow for an analysis of specific conceptual knowledge structures encoded in students\u0026rsquo; neural activity. In contrast, methods such as RSA and others cited above can be used to model the embedding of concept knowledge structures within neural activity patterns directly.\u003c/p\u003e\u003cp\u003eIn the present experiment, we introduced novice students (i.e., students with no advanced physics or engineering experience) to two core concepts from static mechanics, a domain of mechanical engineering, in a controlled laboratory setting. An overview of the present experiment is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. These novices received one hour of instruction on the concepts. Initially, participants were split into two groups, one of which (\u0026ldquo;lab group\u0026rdquo;) received an interactive laboratory activity as part of their instruction, and the other (\u0026ldquo;textbook group\u0026rdquo;) received only written instructional material. Due to comparable performance of these two groups on all of the learning outcomes measured (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e), we analyze them as a single pooled sample. Then, within one week of the learning session, participants returned for an fMRI scan session during which they completed the conceptual near-transfer task: a free-body diagram (FBD) evaluation task. The near-transfer task depicted structures that fell into expert-defined categories based on the mechanical forces at play: cantilever, truss, or vertical load. Critically, participants were never made aware that these categories existed, and were not asked to categorize the stimuli at all. Rather, participants were only exposed to the underlying concepts that give rise to the categories.\u003c/p\u003e\u003cp\u003eOur investigation differs from prior work in two notable ways. Firstly, rather than comparing completely na\u0026iuml;ve participants to advanced learners, we assess novice students during the initial phase of their learning trajectory, after they have learned about two STEM concepts in a short (1-hour) lesson. Secondly, while previous work on the initial phase of learning has focused on either inter-subject measures which do not directly probe the structure of conceptual knowledge, or object identity which includes many features including many which are not diagnostic of the relevant concepts, here we leveraged stimuli which fell into discrete concept-relevant categories which are knowable to experts and advanced learners but not to novices. This design allowed us to use the decodability of category structure from participants\u0026rsquo; neural patterns as an indicator of participants\u0026rsquo; deeper conceptual understanding of the underlying concepts, as shown in previous work\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e,\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. Indeed, we found that the decodability of these conceptual categories from participants\u0026rsquo; multivariate neural representations of the stimuli correlated with participants\u0026rsquo; performance on more traditional assessments of learning, including quiz-style assessments focused on fact retrieval, which were administered throughout the experiment, as well as a conceptual near-transfer task involving the evaluation of free body diagrams completed during fMRI scanning. These results demonstrate that very brief exposure to new concepts can yield identifiable neural evidence of abstract knowledge acquisition, which is meaningfully related to performance on functional knowledge assessments, even in individuals with no prior domain expertise.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cb\u003eBehavioral Scores: Quiz Performance.\u003c/b\u003e Participants\u0026rsquo; knowledge of material covered in the learning sessions was assessed via two written quizzes at each of four timepoints (numbered T1 through T4; see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The two quizzes were designed in collaboration with engineering instructors, including authors S.D. and V.M., to probe fact retrieval of two mechanical engineering topics \u0026ndash; Quiz 1 tested knowledge of linear forces, while Quiz 2 addressed rotational tendencies called \u0026ldquo;moments\u0026rdquo;. Mean Quiz 1 and Quiz 2 scores for each group at each timepoint are reported in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e note: Means and standard deviations (SD) for quiz scores at each timepoint. N indicates sample size for each group as well as combined sample size for \u0026ldquo;All participants\u0026rdquo; scores and standard deviations, which were computed directly from participant-level performance data, not pre-averaged by group. \u0026ldquo;Overall\u0026rdquo; quiz score averages are computed across participants\u0026rsquo; Quiz 1 and Quiz 2 scores for each timepoint. T4 averages include only data from participants who completed T4 follow-up quizzes. Note that for participants who completed T4, quiz scores were not separated by topic, hence the empty cells for Quiz 1 and Quiz 2 topic scores at T4.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eAverage quiz scores at each timepoint\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"9\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTimepoint\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGroup\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eN\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003eQuiz 1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e\u003cp\u003eQuiz 2\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e\u003cp\u003eOverall\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMean\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSD\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eMean\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eSD\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eMean\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eSD\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eT1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLab\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e64.81%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e12.37%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e42.69%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e13.94%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e53.75%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e11.71%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTextbook\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e66.80%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e15.67%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e45.60%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e11.21%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e56.20%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e10.90%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAll participants\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e65.78%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e13.98%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e44.12%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e12.64%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e54.95%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e11.27%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eT2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLab\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e81.73%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e10.48%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e60.19%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e14.46%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e70.96%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e10.27%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTextbook\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e84.60%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e9.12%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e67.60%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e10.22%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e76.10%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e7.88%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAll participants\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e83.14%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e9.85%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e63.82%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e12.98%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e73.48%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e9.45%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eT3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLab\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e79.23%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e13.24%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e56.54%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e15.08%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e67.88%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e11.70%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTextbook\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e84.00%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e10.99%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e65.00%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e12.08%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e74.50%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e9.52%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAll participants\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e81.57%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e12.31%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e60.69%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e14.21%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e71.13%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e11.09%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eT4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLab\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e65.18%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e15.25%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTextbook\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e67.41%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e13.66%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAll participants\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e66.30%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e14.23%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eFor both groups, quiz scores showed a clear learning curve reflected in a mixed-effects polynomial regression model with linear and quadratic effects for experimental timepoint, an interaction effect included for instruction group, and random intercepts included by participant ID (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). On average, there was a significant positive linear effect of experimental timepoint (\u0026#120573; = 0.36, SE\u0026thinsp;=\u0026thinsp;0.13, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.006), coupled with a significant negative quadratic effect (\u0026#120573; = -0.82, SE\u0026thinsp;=\u0026thinsp;0.12, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1.08x10\u003csup\u003e\u0026minus;\u0026thinsp;10\u003c/sup\u003e). These effects indicate that participants improved on the quizzes overall after the learning session between T1 and T2, and then began to exhibit a downturn in performance as the elapsed time since learning increased (i.e., into T3 and T4).\u003c/p\u003e\u003cp\u003eThere was no significant effect of instruction group (lab vs. textbook) on quiz performance, neither in the form of an interaction with the timepoint effect nor in the form of a main effect of instruction group (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e note: Random intercepts were included for participant ID to account for the repeated-measures design. Significant effects are indicated in bold.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePolynomial regression effects for quiz scores over time and by group\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eParameter\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEstimate\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSE\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003edf\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cem\u003et\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e(Intercept)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e0.64\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e0.02\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e50.50\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e33.79\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;2x10\u003c/b\u003e\u003csup\u003e\u003cb\u003e\u0026minus;\u0026thinsp;16\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInstruction Group (Textbook\u0026thinsp;\u0026gt;\u0026thinsp;Lab)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e50.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.10\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eTimepoint (linear)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e0.36\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e0.13\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e127.77\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e2.82\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e0.01\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eTimepoint (quadratic)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e-0.82\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e0.12\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e126.41\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e-7.04\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e1.08x10\u003c/b\u003e\u003csup\u003e\u003cb\u003e\u0026minus;\u0026thinsp;10\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInteraction: Group x Timepoint (linear)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.08\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e127.68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.64\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInteraction: Group x Timepoint (quadratic)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e126.34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e-1.22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.23\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eBehavioral Scores: Free Body Diagram Task.\u003c/b\u003e During the fMRI session at T2, participants completed a concept knowledge task in which they were presented with photographs of real-world structures followed by a diagram labeled according to Newtonian forces that must be acting upon the highlighted section of the structure, and asked to judge the correctness of each diagram. Each task element consisted of a binary forced-choice response (indicating either a correctly labeled or incorrectly labeled diagram), and participants made 12 judgments per run for 8 fMRI task runs. Every stimulus item had a correctly-labeled version and an incorrectly-labeled version, and each version was shown an equal number of times across the runs, with only one version of the stimulus item shown in a given run (i.e., the incorrect and correct versions of the same stimulus were never shown together in the same fMRI run).\u003c/p\u003e\u003cp\u003eAfter averaging performance on the FBD task for each subject across all their item judgments over all 8 fMRI runs, the group average (i.e., grand mean) task accuracy for each instruction group was 63.21% (SD\u0026thinsp;=\u0026thinsp;10.76%) for the lab group and 68.12% (SD\u0026thinsp;=\u0026thinsp;10.82%) for the textbook group. A two-sample t-test revealed that there was no significant difference in how well the lab and textbook groups performed on the FBD task (\u003cem\u003et\u003c/em\u003e = -1.57, df\u0026thinsp;=\u0026thinsp;45.63, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.12). Coupled with the lack of an effect of group on the quiz score results, this suggests that there was no significant difference in how well the lab and textbook groups learned the relevant material during our short experimental intervention. For this reason, all subsequent analyses are performed using all the participants pooled together, without differentiating between those who received the lab or textbook lesson.\u003c/p\u003e\u003cp\u003e\u003cb\u003eNeural Analysis: Measuring Conceptual Understanding\u003c/b\u003e. Our overall goal for the analysis of neural data was to assess whether participants\u0026rsquo; understanding of the conceptual relationships between the items was reflected in neural activity patterns with respect to the \u0026ldquo;expert\u0026rdquo; categorical model of the twelve stimuli. In the conceptual transfer (FBD) task administered during fMRI scanning, participants viewed a set of images of real-world structures and considered the interacting forces on a highlighted component of each structure. Unbeknownst to the participants, each of the structures fell into one of three mechanical categories (cantilever, truss, or vertical load). Importantly, participants were \u003cem\u003enever explicitly instructed\u003c/em\u003e about these three categories. Therefore, the ability of a classifier to distinguish patterns of neural activity belonging to these three categories reflects deeper-level conceptual understanding of the newly-learned physics concepts. Here we assess the presence of this concept knowledge reflected in neural patterns in each participant, and we confirm the relationship of these patterns to learned conceptual information by assessing the correlation between the performance of this neural classifier with more traditional tests of knowledge (both explicit knowledge tests and transfer test performance).\u003c/p\u003e\u003cp\u003eAs a first step, we conducted a whole-brain searchlight analysis for each participant, computing a dissimilarity matrix (DM) of the correlation distances between each of the individual\u0026rsquo;s item-level neural responses to the stimuli at each searchlight location. These searchlight DMs were then averaged by cortical parcel using the Schaefer (2018) cortical parcellation atlas (300-parcel version)\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e, to yield a single parcel-average DM for each of the 300 cortical parcels for each participant.\u003c/p\u003e\u003cp\u003eHaving organized the neural results by cortical parcel in this way, we proceeded with the analysis by computing the extent to which expert categorical information was present first in the group-level average data. We then used the brain regions identified in the group-level analysis to threshold the individual-level data into relevant cortical parcels, and then computed \u0026ldquo;neural scores\u0026rdquo; for each participant as an indicator of the degree of concept knowledge embedded in their neural representations (their \u0026ldquo;neural score\u0026rdquo;, see prior work by Cetron and colleagues\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e). Finally, using a regression modeling approach, we identified the relationship between participants\u0026rsquo; neural scores and their behavioral concept knowledge test scores.\u003c/p\u003e\u003cp\u003e\u003cb\u003eGroup Average Classification.\u003c/b\u003e As a feature reduction step, we sought to identify brain areas which were sensitive to the categorical distinctions between the stimuli in the group average neural responses. In each cortical parcel of the Schaefer (2018) 300-parcel atlas, we ran 1,000 iterations of support vector machine classification with iterative half-sample leave-one-item-per-category-out cross-validation (described in Methods). Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows the distribution of SVM accuracy scores for the categorical model across 1,000 iterations for each parcel. Only parcels in which the average classification accuracy across 1,000 iterations was significantly greater than a permuted null model and additionally greater than 41% (one item above chance classification) were considered to contain significant information about categorical distinctions between the items in the group average. In total, 160 parcels met these criteria, shown in green in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. Subsequent analyses of individual performance were performed only on these parcels.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eIndividual Classification: Neural Scores\u003c/b\u003e. In each of the 160 parcels identified by the group average classification analysis, we then attempted to classify the mechanical categories in the individual neural data. This analysis followed the same 1,000 iteration leave-one-item-per-category-out procedure, but in this case only one individual was held out from the training set at a time, and their data from the held out items used at testing. Thus, the classifier performance when any given subject was held out can be treated as an \u0026ldquo;individual neural score\u0026rdquo;, a measure of the separability of the categories in their own neural data.\u003c/p\u003e\u003cp\u003eAfter calculating mean neural scores over 1,000 iterations of SVM classification, a linear mixed effects analysis of the neural score as a predictor of behavioral scores (quizzes, FBD performance) was performed at each of the 160 parcels included in the mask of areas defined by category separability in the group average neural activity. The linear mixed effects model for each parcel was fitted with random intercepts for each subject and each knowledge assessment type. This use of mixed effects modeling to implement multivariate regression (i.e., regression models with multiple response variables) is a common modeling strategy when the multiple response variables are considered to represent the same underlying construct but may be measured with slight differences\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. This is especially useful when the response variables are correlated with one another because a multivariate regression model accounts for these covariances during estimation\u003csup\u003e\u003cspan additionalcitationids=\"CR21\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. A heatmap of observed beta values for neural score as a predictor of behavioral knowledge assessment scores in these 160 parcels is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eWe also computed a null distribution of beta values by randomizing the mapping of behavioral to neural scores 1,000 times for each parcel and recording the beta values of neural score as a predictor of behavioral scores in these permuted mappings, which simulate the beta values that could occur by chance in a given parcel, considering the distributions of that parcel\u0026rsquo;s behavioral and neural scores\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. Six parcels were identified where observed beta values for individual neural score as a predictor of behavioral score was greater than 1.65 standard deviations above the permuted null distribution. These parcels are located in the left inferior temporal gyrus (parcel 50, \u0026#120573; = 0.29, Z\u0026thinsp;=\u0026thinsp;1.70), the left intraparietal sulcus (parcel 93, \u0026#120573; = 0.28, Z\u0026thinsp;=\u0026thinsp;2.08), the right supramarginal gyrus (parcel 247, \u0026#120573; = 0.18, Z\u0026thinsp;=\u0026thinsp;1.71), the right postcentral sulcus (parcel 205, \u0026#120573; = 0.25, Z\u0026thinsp;=\u0026thinsp;2.16), and the bilateral precuneus (parcels 116, \u0026#120573; = 0.27, Z\u0026thinsp;=\u0026thinsp;1.67; and 268, \u0026#120573; = 0.24, Z\u0026thinsp;=\u0026thinsp;1.98). The observed beta values, Z-scores of those values in the permuted null distribution, and anatomical labels of each significant parcel from the in the 17-network Yeo atlas\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e as well as the Destrieux atlas\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e are reported in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eThis result provides evidence that individuals\u0026rsquo; neural category separability in these areas was significantly predictive of their knowledge as measured by the various behavioral assessments. For each of the six significant parcels, scatter plots of individual neural and behavioral scores with the overall regression line and random intercepts for each knowledge assessment (FBD task accuracy, Quiz 1 score, and Quiz 2 score) are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e note: Regression coefficients for the six parcels in which neural scores predicted behavioral scores with observed beta values at least 1.65 standard deviations outside the permuted null distribution of beta parameters. For each parcel, we report the observed beta parameter value (* = two-sample t-test against null distribution p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), the Z-score of the beta value in the permuted null distribution, the anatomical label of the parcel from the Destrieux atlas projected to the FSAverage5 surface, and the parcel location in the 17-network Yeo parcellation atlas.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eRegression results for parcels where neural scores best tracked behavioral scores\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eParcel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026#120573;\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eZ\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDestrieux Label\u003c/p\u003e\u003cp\u003e(SUMA)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eYeo Label\u003c/p\u003e\u003cp\u003e(17 networks)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.29*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.70\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eL inferior temporal gyrus\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eLH Dorsal Attention A Temporal Occipital\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e93\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.28*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2.08\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eL intraparietal sulcus/superior parietal gyrus\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eLH Control A Intraparietal Sulcus\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e116\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.27*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eL precuneus\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eLH Default A Medial Prefrontal Cortex\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e205\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.25*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eR postcentral sulcus\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eRH Dorsal Attention B Post Central\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e247\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.18*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.71\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eR supramarginal gyrus\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eRH Control A Intraparietal Sulcus\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e268\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.24*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eR precuneus\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eRH Control C Precuneus\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn the present study, we have demonstrated that concept knowledge can be detected within the neural activity patterns of novice learners even after brief exposure to new concepts. Using neural scores derived from the classification of individuals\u0026rsquo; neural representations of stimuli into relevant conceptual categories, we identified at least six areas of the brain where neural scores significantly predicted performance on traditional behavioral measures of learning, including topic quizzes and a conceptual near-transfer task.\u003c/p\u003e\u003cp\u003eAmong the areas where individual neural scores significantly predicted behavioral performance is the left IPS, which is commonly associated with processing of visuospatial information (in particular involving the estimation of magnitudes and quantities) and visuomotor action planning, guidance, and execution\u003csup\u003e\u003cspan additionalcitationids=\"CR27 CR28\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. This result aligns with previous work including our own which implicates dorsal-stream regions, especially the IPS, in the neural representation of physics knowledge\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e,\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. In particular, the IPS was implicated in representation of abstract physics concepts in the early learning phase described by Mason and Just\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e, one of the few other training studies (i.e., where students were instructed about a new physics concept) in this subfield of cognitive neuroscience.\u003c/p\u003e\u003cp\u003eNeural scores also predicted physics knowledge in the ventral ITG, which is commonly associated with processing of visual feature representations that are diagnostic of object categories\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. The literature on category learning in the ventral stream is generally centered on perceptual categories, but there is evidence from prior studies (e.g., from work by Connolly and colleagues\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e, and Haxby and colleagues\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e) that conceptual categories are also differentiated in ventral temporal regions.\u003c/p\u003e\u003cp\u003eAs previously discussed, an important feature distinguishing the present study from other work is the choice to study novices after a brief period of learning rather than comparing participants who were advanced learners or even experts with those who were entirely naive. Our focus here is on the very early steps along the learning trajectory. Thus, it is particularly striking that in the left IPS and ventral ITG\u0026mdash;two areas in which neural patterns distinguished engineering students from non-engineering students in the previous study\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e\u0026mdash;behavioral test scores were significantly correlated with neural scores for novices with only about an hour of relevant training. In our previous study, the physics concept representations of advanced learners showed a posterior-to-anterior gradient in both the dorsal and ventral streams. In our current study, briefly-trained novices exhibit physics concept representations in ventral and dorsal regions that are posterior to the key regions we observed for the advanced learners in the prior study, but still anterior to any neural results we observed for na\u0026iuml;ve novice group (who received no training). Although the methodology of these two studies differed in important ways, this suggests the possibility that the anterior shift in representing physics concepts in the brain may be a progression through the learning process, although testing this directly would require a longitudinal fMRI investigation with higher temporal resolution (i.e., more time points over a much longer learning period).\u003c/p\u003e\u003cp\u003eOther areas which showed significant prediction of neural scores included right parietal regions such as the supramarginal gyrus and postcentral sulcus, which have also been shown to play a role in motor imagery and spatial memory\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e (among other functions), and bilateral parcels in the precuneus, an area which has been associated with visuospatial memory, episodic memory retrieval, and semantic processing\u003csup\u003e\u003cspan additionalcitationids=\"CR35\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. In particular, conceptual representations in the precuneus have been shown to contain information about amodal semantic relationships between concrete object categories\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. These results suggest that this property may extend to more complex and implicit categorical relationships as well.\u003c/p\u003e\u003cp\u003eIt is also crucial to note that the novice participants in the present study were never explicitly instructed about the existence of the mechanical categories used in the classification analysis to derive their neural scores. Participants were not aware of any categories for the stimuli in the FBD task, nor were they ever asked to employ a categorical strategy during any of the behavioral tasks. \u003cem\u003eThus, any conceptual category information represented in participants\u0026rsquo; neural data reflected an emergent understanding of the deeper STEM concepts implicit in those categories, rather than the simple retrieval of memorized facts.\u003c/em\u003e This interpretation of the neural score results is further supported by the correlation between the neural and behavioral scores: the participants who displayed greater implicit knowledge through their neural representations also displayed greater explicit knowledge through their behavioral test scores.\u003c/p\u003e\u003cp\u003eDue to the comparable behavioral performance that we observed between the laboratory and textbook participant groups, we limited our analysis of the neural data to the full set of participants, pooled together rather than separated by learning condition. Although we initially hypothesized an advantage of hands-on learning over text-based instruction, we did not find evidence of any such distinction in this very brief intervention. It is possible that the advantage of hands-on instruction would have become more observable over longer timescales of learning. Future research may attempt to explore how hands-on experience could shape neural representations of physics concepts over a longer learning period, and how any such neural effects of laboratory-based learning may relate to pencil-and-paper knowledge assessments. The decodability of subtle shifts in understanding on the individual level demonstrated by the present study would suggest that these methods could be well-suited to such comparisons of longer-term classroom interventions.\u003c/p\u003e\u003cp\u003eThe results of the present study show that through multivariate pattern analysis methods, it is possible to detect individual-level differences in conceptual understanding in novices with brief exposure (one hour) to new STEM concepts. By leveraging categorical distinctions between items in the stimulus set which are apparent to experts and advanced learners but not to novices, we derived neural scores using SVM classification. In six cortical parcels in the brain, these neural scores predicted individual performance on behavioral assessments of learning, providing evidence that multivariate neural activity patterns in these regions reflected conceptual understanding. Methods and insights from this line of work may help shape educational approaches in the future, such as providing an additional modality with which to examine learning for applied curriculum development research. Using multimodal data-driven approaches, future research can yield deeper insight into the early stages of learning, where novice students begin to build the conceptual scaffolds that are essential for understanding complex topics in science, math, and other domains of knowledge.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cb\u003eParticipants\u003c/b\u003e. Fifty-three Dartmouth College students participated in this study. Three were excluded: one due to excessive motion in the scanner, one due to a stimulus presentation error, and one due to an illness that the participant self-reported to have affected their memory after the experiment had concluded. The resulting sample had an N of 50 (32 female; mean age at T1\u0026thinsp;=\u0026thinsp;19.85 years, SD\u0026thinsp;=\u0026thinsp;1.17). All participants were sent a follow-up knowledge assessment at least one month after the scan session. Nineteen participants responded to this online survey.\u003c/p\u003e\u003cp\u003eAll participants were fluent English speakers who completed an eligibility survey verifying they had little-to-no prior coursework experience in physics or mechanical engineering. Specifically, we excluded any participants with more than 1 semester of introductory college-level physics experience (or AP equivalent). We also excluded any participants who were majoring in engineering or physics, even if they had not yet completed any advanced engineering or physics coursework. These were the same exclusion criteria used to recruit the previous studies\u0026rsquo; novice group.\u003c/p\u003e\u003cp\u003eParticipants provided informed consent on each day of data collection, and were offered either curricular extra credit points or a small monetary compensation. All procedures were approved by the Dartmouth Committee for the Protection of Human Subjects (CPHS E (#5): IRB00006768; \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.dartmouth.edu/cphs/\u003c/span\u003e\u003cspan address=\"https://www.dartmouth.edu/cphs/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eMaterials.\u003c/b\u003e All materials and stimuli used in the present experiment that do not contain copyrighted content can be viewed and downloaded on the Open Science Framework in our project repository by following this link: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://osf.io/ub4z3/?view_only=2ef76a55f5f2405cb621c54090dedb7a\u003c/span\u003e\u003cspan address=\"https://osf.io/ub4z3/?view_only=2ef76a55f5f2405cb621c54090dedb7a\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e .\u003c/p\u003e\u003cp\u003eDuring the behavioral session of the experiment, participants learned about static mechanics, an area of physics relevant to mechanical engineering. All participants were given a primer on the basic premise of static mechanics, consisting of a short powerpoint presentation (8 slides) explaining that unmoving objects in the world are able to maintain equilibrium (i.e., remain unmoving) due to the balance of forces exerted both by and upon the objects. This primer was created by the authors and validated by authors S.D. and V.M., who have domain expertise in engineering physics. These slides are included in the OSF project repository.\u003c/p\u003e\u003cp\u003eThe static mechanics primer was shown to participants twice: first as their initial introduction to the learning portion of the behavioral session (after baseline assessments were completed), and again prior to the fMRI task during the fMRI session.\u003c/p\u003e\u003cp\u003eDuring the learning portion of the behavioral session following the static mechanics primer, participants were given more detailed introductions to each of the two topics within static mechanics that they would be learning about: linear forces (Topic 1) and rotational tendencies called \u0026ldquo;moments\u0026rdquo; (Topic 2). The main learning portion for each topic would consist of interactive instructional materials that varied by experimental condition (a \u0026ldquo;textbook\u0026rdquo; condition and a \u0026ldquo;lab\u0026rdquo; condition), but prior to each of these interactive sessions, participants would review a set of slides containing text, pictures, and video clips describing the essential principles of each topic. These slides were created by the authors and utilized content from an EdX course on static mechanics taught by author V.M.\u003c/p\u003e\u003cp\u003e\u003cb\u003eTextbook condition instruction materials.\u003c/b\u003e After the first topic area introduction, participants in the textbook condition reviewed a second set of slides accompanied by a printed inquiry sheet. Using the slides, textbook participants reviewed a series of specific examples illustrating concepts central to the topic area. As they reviewed the examples, they occasionally encountered question prompts, which were also printed on their inquiry sheets. Participants marked their answers to the questions on the inquiry sheet, and then reviewed those answers on subsequent slides, tracking the accuracy of their responses throughout the instruction period. Participants were instructed not to be concerned about the accuracy of their responses to these inquiry questions (we did not analyze their performance in this phase of instruction). After completing the example slides and inquiry sheet for Topic 1, they repeated the topic area introduction and example slide/inquiry sheet procedure for Topic 2.\u003c/p\u003e\u003cp\u003e\u003cb\u003eLab condition instruction materials.\u003c/b\u003e Lab participants completed a procedure that paralleled the textbook procedure as closely as possible, except that they reviewed the topic-related examples using a series of activities rather than a series of slides. The activities were designed based on common physics and engineering laboratory projects, such as the construction and deconstruction of a truss from plastic drinking straws to illustrate tension and compression in a truss system (Topic 1). Participants received instructions for each activity on a printed sheet, completing the activities on their own except when the experimenter was needed to assist with the execution of an activity. As in the textbook condition, participants were periodically prompted to answer inquiry questions using an accompanying inquiry sheet, and tracked their progress as they completed the examples.\u003c/p\u003e\u003cp\u003eThe lab activities and inquiry questions were paired as closely as possible with the textbook examples and inquiry questions, such the two conditions illustrated the same concepts in the same order as one another, and such that the inquiry questions could be phrased as similarly as possible across conditions. This section of the behavioral session constituted the only difference between the lab and textbook conditions. Copies of inquiry materials for both conditions (excluding those containing copyrighted content) can be found on the OSF project repository.\u003c/p\u003e\u003cp\u003e\u003cb\u003eTopic Quizzes.\u003c/b\u003e All participants completed knowledge assessments in the form of two topic-specific quizzes focused on fact retrieval and administered across at least three timepoints in the experiment: first as an initial baseline measurement at the very beginning of the experiment (T1); second as a test of learning after the instruction session (T2); and third as a repeat knowledge test after the fMRI scan session (T3). All participants were sent the quizzes a fourth time at least one month after the scan session (T4), but not every participant completed this final assessment. At each timepoint, the quizzes participants completed were identical (i.e., the material being tested was always the same). Participants were never given feedback regarding their performance on the quizzes.\u003c/p\u003e\u003cp\u003eWe designed the fact-retrieval quizzes for each of the two mechanical engineering topics in collaboration with engineering instructors, including authors S.D. and V.M. Assessments were multiple-choice and involved both verbal and visual questions probing concept knowledge. Performance on each topic was analyzed separately, due to disparities in difficulty by topic (topic 2 was considerably harder for students to comprehend than topic 1). Copies of the quiz materials can be found on the OSF project repository.\u003c/p\u003e\u003cp\u003e\u003cb\u003eFree body diagram task (fMRI near-transfer task).\u003c/b\u003e During fMRI scanning, participants completed a conceptual near-transfer task involving the evaluation of photographs of real-world structures labeled as free body diagrams (FBDs). This constituted a near-transfer task because participants had never seen this task before, and had to apply their recently-learned concept knowledge associated with linear forces and moments to complete the task successfully. Stimulus photographs for the FBD task were selected from the set of 24 images utilized by the previous study\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. In each of these images, which included lampposts, bridges, awnings, and other similar structures, mechanical engineering experts (authors S.D. and V.M.) labeled a particular component of each structure by outlining it in red. Each of the highlighted components fell into one of three mechanical categories (cantilever, truss, or vertical load). Importantly, participants were \u003cem\u003enever explicitly instructed\u003c/em\u003e about these three categories.\u003c/p\u003e\u003cp\u003e\u003cb\u003eExpert model.\u003c/b\u003e To query the degree of conceptual information present in a given representation of the 12 real-world structure stimuli, we compared representations to an expert model of the similarity between the stimulus items. Specifically, author S.D. compared a larger set of structure stimuli (originally 24 items) in a series of pairwise similarity comparisons, focusing on the mechanical similarity between the items. This pairwise similarity model yielded three categories of mechanical structures: vertical loads, trusses, and cantilevers. In the present study, we selected four representative items from each of the three categories, resulting in the 12-item stimulus set used here. (For the full 24-item stimulus set, see previous study\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e,\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e.) Neural representations for each participant were constructed based on their neural pattern responses to each of these 12 stimuli. Then, the degree of concept information present in a given representation was identified by fitting a categorical classifier to the participant\u0026rsquo;s representation, where the true category labels correspond to the mechanical structure categories identified in the expert model.\u003c/p\u003e\u003cp\u003e\u003cb\u003eBehavioral similarity probe (not analyzed).\u003c/b\u003e In addition to the quizzes, participants\u0026rsquo; representations of concept-relevant stimuli were also queried via pairwise similarity ratings of the 12 stimulus items they would review in the FBD task. These ratings were made at each of the same timepoints as the quiz assessments. We assessed the correspondence between these explicit similarity ratings and the expert categorical model using the same classification procedure as was applied to the neural similarity matrices (see Methods subheading \u0026ldquo;Multivariate Analysis\u0026rdquo;). The correlations between these classifier results and participants\u0026rsquo; quiz scores for each time point are summarized in Supplementary Fig.\u0026nbsp;1. However, since these classifier scores neither correlate with individual quiz score nor show a clear learning trajectory over time, we have insufficient evidence that this task as administered was able to measure the relevant constructs. For example, rather than considering mechanical similarity, these novice participants may have relied on more surface-level similarities that were deliberately built into the stimulus set. As such, we do not discuss the results from the behavioral similarity ratings in the present paper.\u003c/p\u003e\u003cp\u003e\u003cb\u003eProcedure.\u003c/b\u003e Participants completed two sessions in this experiment: a behavioral session for concept instruction, and an fMRI scan session. An overview of activities completed during each session, as well as the online follow-up which occurred one month later, is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eThe behavioral session began with participants giving their informed consent to participate in the study, and then completing baseline assessments on the topic quizzes and provided information on their prior experience in any physics and engineering courses. Participants were expected to be unfamiliar with the concepts assessed in the quizzes at this timepoint (T1). After completing baseline assessments, all participants reviewed the static mechanics primer slides, and then began the topic-specific learning sessions.\u003c/p\u003e\u003cp\u003eFirst, participants learned about linear forces (topic 1). They reviewed the topic 1 introductory slides, and then completed the interactive learning condition to which they had been randomly assigned: either the textbook (computer-based) interactive materials condition, or the laboratory interactive materials condition. The textbook condition involved reviewing slides that prompted the participant to make predictions about specific examples involving linear forces, and then learning about the solutions on subsequent slides. The laboratory condition involved completing activities involving analogous linear force examples, and learning about the solutions after completing the activities. After concluding with the topic 1 materials (and taking an optional short break), participants repeated the procedure with the materials for topic 2: rotational tendencies (known as \u0026ldquo;moments\u0026rdquo;).\u003c/p\u003e\u003cp\u003eOnce participants had reviewed all learning materials, they proceeded to the post-learning quiz assessments. These quizzes were identical to the baseline quizzes, but at this timepoint (T2) participants now had learned about the topics being assessed. After the T2 assessments, the behavioral session was complete, and participants were compensated for their time.\u003c/p\u003e\u003cp\u003e\u003cb\u003efMRI session.\u003c/b\u003e Within a maximum of 1 week of the behavioral session, participants returned to complete an fMRI scan session. Before scanning, participants were shown the static mechanics primer again, as well as an introduction to the FBD task they would be performing in the fMRI scanner. Upon entering the fMRI scanner, participants reviewed these primer materials once more, and then were familiarized with the 12 stimulus items they would be evaluating during the FBD task.\u003c/p\u003e\u003cp\u003eAfter familiarization and structural scans, participants completed 8 runs of the FBD task during functional scanning. The stimuli and design for this task are described in detail in previous work\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Participants reviewed the FBD stimuli individually, first assessing each stimulus, and then responding to the prompt via button press after a jittered fixation. Analyzed functional data were drawn from the consideration period prior to button presses.\u003c/p\u003e\u003cp\u003eOnce the functional scans were complete, participants filled out the assessment quizzes again, now at the post-scan timepoint (T3). After the T3 assessments, the fMRI session was complete, and participants were compensated for their time.\u003c/p\u003e\u003cp\u003e\u003cb\u003eLong-term recall follow-up.\u003c/b\u003e After completing the required behavioral and fMRI sessions, participants were invited to complete a 1-month follow-up evaluation (timepoint T4) via online survey, which only some participants completed. Participants who completed this assessment were compensated for this additional time spent.\u003c/p\u003e\u003cp\u003e\u003cb\u003efMRI Data Acquisition.\u003c/b\u003e A 3 Tesla Siemens PRISMA fMRI scanner with a 32-channel head coil was used to acquire the fMRI data. A single high-resolution T1-weighted anatomical scan (FOV\u0026thinsp;=\u0026thinsp;240mm, Flip Angle\u0026thinsp;=\u0026thinsp;8\u0026deg;, 192 slices) and eight functional runs were performed for each participant. Each 2D EPI sequence consisted of 186 measurements with a 240mm field of view to provide full brain coverage over 52 slices (Flip Angle\u0026thinsp;=\u0026thinsp;75\u0026deg;; TE\u0026thinsp;=\u0026thinsp;35 ms; TR\u0026thinsp;=\u0026thinsp;2000 ms; 2.5mm\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e voxels). In the scanner, stimuli were presented using PsychoPy version 1.84.2\u003csup\u003e37\u003c/sup\u003e (using Python 2.7).\u003c/p\u003e\u003cp\u003e\u003cb\u003eImage Preprocessing and Univariate Analyses.\u003c/b\u003e Brain images were preprocessed using the FSL FEAT software package\u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. The FSL brain extraction tool (BET) was used to skull-strip each T1-weighted anatomical image. Skull-stripping, motion correction, slice timing correction, and highpass temporal filtering were also applied to each functional EPI volume. Finally, the functional runs were registered to the participant\u0026rsquo;s individual anatomical space using the FSL linear registration tool\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eA first-level univariate regression model using the general linear model was then calculated at the item level, such that beta-value estimates for each stimulus were generated separately for each run for each participant. Then, a second-level GLM combined these estimates across functional runs, yielding a single contrast estimate for each of the twelve items. All beta-value estimates were then aligned to the individual\u0026rsquo;s T1 volume and resampled to 2mm\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e using the FSL mathematical manipulation tool.\u003c/p\u003e\u003cp\u003eFor each subject, cortical surface reconstructions were generated for the T1-weighted anatomical image using FreeSurfer\u0026rsquo;s recon-all toolbox\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e and fitted to standard mesh grids based on an icosahedron with 32 linear divisions, yielding 20,484 nodes for the whole-brain cortical surface in Surface Mapping (SUMA) format\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. Sulcal alignment of each participant\u0026rsquo;s cortical surface to the FreeSurfer average brain\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e was performed to allow anatomical correspondence between surface nodes across participants. We also performed the same transformation on an atlas of 300 parcels defined according to abrupt transitions in functional connectivity patterns in resting state fMRI\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e to create a mapping between the parcellation and the 20,484-node surface space.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMultivariate Analysis.\u003c/b\u003e We used the PyMVPA toolbox\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e to conduct a whole-brain searchlight analysis for each participant using spherical 5mm searchlights. In each searchlight, the correlation distance between the individual\u0026rsquo;s item-level betas for each of the twelve stimuli was computed to form a dissimilarity matrix (DM) for each node and its surrounding neighborhood. These DMs were then averaged for all surface nodes belonging to each cortical parcel in the Schaefer (2018) 300-parcel atlas\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e, yielding a single average DM for each parcel.\u003c/p\u003e\u003cp\u003eAs a feature reduction step, we sought to identify parcels where categorical information about the twelve items (cantilevers vs. vertical loads vs. trusses) was represented in the neural activity of the group as a whole. Each parcel-level DM was projected into two dimensions using multidimensional scaling (MDS), and a support vector machine (SVM) classifier with a radial basis function kernel was employed using half-sample leave-one-item-per-category-out cross validation for 1,000 iterations and the mean accuracy score was taken. On each iteration, the twelve items were randomly assigned to four folds such that each fold consisted of a test set with one item from each of the three categories. On each fold, the model was trained on nine rows (leaving one item per category out) of an average DM constructed from a randomly drawn half (25) of the participants, then tested on the three held-out items in the average DM of the held-out participants. The division of both participants and items into the training and test sets was randomized on each iteration. These steps were implemented in Python using the scikit-learn package\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. At each parcel, we also performed a permutation test by shuffling the training data labels and repeating the same procedure for another 1,000 iterations. The permutation test results for relevant parcels are shown in Supplementary Fig.\u0026nbsp;2. Out of 300 parcels, we identified 160 in which the mean SVM accuracy for the categorical model was both significantly greater than the permuted null model and additionally greater than 41%, indicating that the classifier could reliably categorize the held-out items at least 8% (the value of one item) greater than chance (33%). These parcels were treated as a binary mask in subsequent analyses.\u003c/p\u003e\u003cp\u003eIn these 160 parcels which had been shown to be sensitive to categorical distinctions between the stimuli in the \u003cem\u003egroup average\u003c/em\u003e, we then probed individual parcel DMs using the same leave-one-item-per-category-out cross-validation procedure to assess the degree to which each \u003cem\u003eindividual\u0026rsquo;s\u003c/em\u003e patterns of neural activity in each parcel reflected the relationships between the items. As before, on each fold of the 1,000 iterations, the SVM classifier was trained on nine out of twelve items in the average DM of the training set. However, rather than holding out a randomly selected half of the sample, we trained the classifier on all but one participant, then tested on the three held-out items for the remaining participant. The mean classification accuracy for each individual was taken to be that individual\u0026rsquo;s \u0026ldquo;neural score\u0026rdquo; for each parcel.\u003c/p\u003e\u003cp\u003eIn each of the 160 parcels identified as significant at the group level, we performed a linear mixed effects analysis of the relationship between participants\u0026rsquo; behavioral scores (quiz scores, FBD performance) and neural scores. These models were created in R\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e using the lme4 package\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. We entered neural score as a fixed effect, and included behavioral score type (Quiz 1, Quiz 2, or FBD performance) and intercepts for each participant as random effects. We then computed a null distribution of beta values by shuffling the mapping of behavioral to neural scores and recording the beta values of neural score as a predictor of behavioral scores over 1,000 permutations. We define significant prediction as a parcel where the observed beta value for neural score as a predictor of behavioral scores is greater than 1.65 standard deviations above this permuted null distribution of beta values, a critical Z value which corresponds to an alpha level of 0.05.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003ch2\u003eCompeting Interests\u003c/h2\u003e\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAuthors J.S.C. and M.E.H. contributed equally to the work and are co-first authors of the present manuscript. J.S.C. and D.J.M.K. designed and conducted the study. J.S.C., M.E.H., and D.J.M.K. performed the analysis and wrote the manuscript. S.G.D. and V.V.M. contributed to study design (procedure design, assessment design, and stimulus creation). All authors reviewed the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eD.J.M.K. and M.E.H. were supported by a National Science Foundation award (DRL-2201304). The authors thank J. Hayes for his assistance with data collection and the members of the Cognitive Neuroscience of Learning Lab at Dartmouth for scientific discussions and feedback on this manuscript. We thank T. Sackett and the Dartmouth Brain Imaging Center for facilitating fMRI research at Dartmouth.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets generated during the current study are available from the corresponding author upon reasonable request, where participant data privacy policies permit. The code generated during the current study is available from the corresponding author upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eLake, B. M., Linzen, T. \u0026amp; Baroni, M. Human few-shot learning of compositional instructions. \u003cem\u003earXiv.org\u003c/em\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://arxiv.org/abs/1901.04587v2\u003c/span\u003e\u003cspan address=\"https://arxiv.org/abs/1901.04587v2\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCetron, J. S. \u003cem\u003eet al.\u003c/em\u003e Decoding individual differences in STEM learning from functional MRI data. \u003cem\u003eNature Communications\u003c/em\u003e 10, 2027 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCetron, J. S. \u003cem\u003eet al.\u003c/em\u003e Using the force: STEM knowledge and experience construct shared neural representations of engineering concepts. \u003cem\u003enpj Sci. Learn.\u003c/em\u003e 5, 6 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChao, L. L., Weisberg, J. \u0026amp; Martin, A. Experience-dependent Modulation of Category-related Cortical Activity. \u003cem\u003eCerebral Cortex\u003c/em\u003e 12, 545\u0026ndash;551 (2002).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eConnolly, A. C. \u003cem\u003eet al.\u003c/em\u003e The Representation of Biological Classes in the Human Brain. \u003cem\u003eJ. Neurosci.\u003c/em\u003e 32, 2608\u0026ndash;2618 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCoutanche, M. N. \u0026amp; Thompson-Schill, S. L. Rapid consolidation of new knowledge in adulthood via fast mapping. \u003cem\u003eTrends in Cognitive Sciences\u003c/em\u003e 19, 486\u0026ndash;488 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFairhall, S. L. \u0026amp; Caramazza, A. Brain Regions That Represent Amodal Conceptual Knowledge. \u003cem\u003eJ. Neurosci.\u003c/em\u003e 33, 10552\u0026ndash;10558 (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMartin, A. The Representation of Object Concepts in the Brain. \u003cem\u003eAnnual Review of Psychology\u003c/em\u003e 58, 25\u0026ndash;45 (2007).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMason, R. A. \u0026amp; Just, M. A. Physics instruction induces changes in neural knowledge representation during successive stages of learning. \u003cem\u003eNeuroImage\u003c/em\u003e 111, 36\u0026ndash;48 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMason, R. A. \u0026amp; Just, M. A. Neural Representations of Physics Concepts. \u003cem\u003ePsychological Science\u003c/em\u003e 27, 904\u0026ndash;913 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWeisberg, J., van Turennout, M. \u0026amp; Martin, A. A Neural System for Learning about Object Function. \u003cem\u003eCerebral Cortex\u003c/em\u003e 17, 513\u0026ndash;521 (2007).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKriegeskorte, N., Mur, M. \u0026amp; Bandettini, P. Representational similarity analysis - connecting the branches of systems neuroscience. \u003cem\u003eFrontiers in Systems Neuroscience\u003c/em\u003e 2, (2008).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCoutanche, M. \u0026amp; Thompson-Schill, S. Informational connectivity: identifying synchronized discriminability of multi-voxel patterns across the brain. \u003cem\u003eFrontiers in Human Neuroscience\u003c/em\u003e 7, (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHasson, U., Nir, Y., Levy, I., Fuhrmann, G. \u0026amp; Malach, R. Intersubject Synchronization of Cortical Activity During Natural Vision. \u003cem\u003eScience\u003c/em\u003e 303, 1634\u0026ndash;1640 (2004).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChase, W. G. \u0026amp; Simon, H. A. Perception in chess. \u003cem\u003eCognitive Psychology\u003c/em\u003e 4, 55\u0026ndash;81 (1973).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChi, M. T. H., Feltovich, P. J. \u0026amp; Glaser, R. Categorization and representation of physics problems by experts and novices. \u003cem\u003eCognitive Science\u003c/em\u003e 5, 121\u0026ndash;152 (1981).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMeshulam, M. \u003cem\u003eet al.\u003c/em\u003e Neural alignment predicts learning outcomes in students taking an introduction to computer science course. \u003cem\u003eNature Communications\u003c/em\u003e 12, (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchaefer, A. \u003cem\u003eet al.\u003c/em\u003e Local-Global Parcellation of the Human Cerebral Cortex from Intrinsic Functional Connectivity MRI. \u003cem\u003eCereb Cortex\u003c/em\u003e 28, 3095\u0026ndash;3114 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eButler, E. A., Gross, J. J. \u0026amp; Barnard, K. Testing the effects of suppression and reappraisal on emotional concordance using a multivariate multilevel model. \u003cem\u003eBiological Psychology\u003c/em\u003e 98, 6\u0026ndash;18 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHox, J., Moerbeek, M. \u0026amp; Schoot, R. van de. \u003cem\u003eMultilevel Analysis: Techniques and Applications, Third Edition\u003c/em\u003e. (Routledge, New York, 2017). doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4324/9781315650982\u003c/span\u003e\u003cspan address=\"10.4324/9781315650982\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTwisk, J. W. R. \u003cem\u003eApplied Mixed Model Analysis: A Practical Guide\u003c/em\u003e. (Cambridge University Press, Cambridge, 2019). doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1017/9781108635660\u003c/span\u003e\u003cspan address=\"10.1017/9781108635660\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSnijders, T. A. B. \u0026amp; Bosker, R. J. Chapter 16: Multivariate Multilevel Models. in \u003cem\u003eMultilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling\u003c/em\u003e 282\u0026ndash;288 (Sage Publishers, London, 2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGiacalone, M. \u0026amp; Alibrandi, A. Overview and Main Advances in Permutation Tests for Linear Regression Models. \u003cem\u003eJMSS\u003c/em\u003e 5, (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eThomas Yeo, B. T. \u003cem\u003eet al.\u003c/em\u003e The organization of the human cerebral cortex estimated by intrinsic functional connectivity. \u003cem\u003eJournal of Neurophysiology\u003c/em\u003e 106, 1125\u0026ndash;1165 (2011).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDestrieux, C., Fischl, B., Dale, A. \u0026amp; Halgren, E. Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature. \u003cem\u003eNeuroImage\u003c/em\u003e 53, 1\u0026ndash;15 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAnsari, D. Effects of development and enculturation on number representation in the brain. \u003cem\u003eNat Rev Neurosci\u003c/em\u003e 9, 278\u0026ndash;291 (2008).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGrefkes, C. \u0026amp; Fink, G. R. REVIEW: The functional organization of the intraparietal sulcus in humans and monkeys. \u003cem\u003eJournal of Anatomy\u003c/em\u003e 207, 3\u0026ndash;17 (2005).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSimon, O., Mangin, J.-F., Cohen, L., Le Bihan, D. \u0026amp; Dehaene, S. Topographical Layout of Hand, Eye, Calculation, and Language-Related Areas in the Human Parietal Lobe. \u003cem\u003eNeuron\u003c/em\u003e 33, 475\u0026ndash;487 (2002).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSimon, O. \u003cem\u003eet al.\u003c/em\u003e Automatized clustering and functional geometry of human parietofrontal networks for language, space, and number. \u003cem\u003eNeuroImage\u003c/em\u003e 23, 1192\u0026ndash;1202 (2004).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKontra, C., Lyons, D. J., Fischer, S. M. \u0026amp; Beilock, S. L. Physical Experience Enhances Science Learning. \u003cem\u003ePsychol Sci\u003c/em\u003e 26, 737\u0026ndash;749 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHaxby, J. V. \u003cem\u003eet al.\u003c/em\u003e A Common, High-Dimensional Model of the Representational Space in Human Ventral Temporal Cortex. \u003cem\u003eNeuron\u003c/em\u003e 72, 404\u0026ndash;416 (2011).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKaufmann, L. \u003cem\u003eet al.\u003c/em\u003e A developmental fMRI study of nonsymbolic numerical and spatial processing. \u003cem\u003eCortex\u003c/em\u003e 44, 376\u0026ndash;385 (2008).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRuss, M. O., Mack, W., Grama, C.-R., Lanfermann, H. \u0026amp; Knopf, M. Enactment effect in memory: evidence concerning the function of the supramarginal gyrus. \u003cem\u003eExp Brain Res\u003c/em\u003e 149, 497\u0026ndash;504 (2003).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCavanna, A. E. \u0026amp; Trimble, M. R. The precuneus: a review of its functional anatomy and behavioural correlates. \u003cem\u003eBrain\u003c/em\u003e 129, 564\u0026ndash;583 (2006).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchott, B. H. \u003cem\u003eet al.\u003c/em\u003e Gradual acquisition of visuospatial associative memory representations via the dorsal precuneus. \u003cem\u003eHuman Brain Mapping\u003c/em\u003e 40, 1554\u0026ndash;1570 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTrimble, M. R. \u0026amp; Cavanna, A. E. Chapter 3.7 The role of the precuneus in episodic memory. in \u003cem\u003eHandbook of Behavioral Neuroscience\u003c/em\u003e (eds. Dere, E., Easton, A., Nadel, L. \u0026amp; Huston, J. P.) vol. 18 363\u0026ndash;377 (Elsevier, 2008).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePeirce, J. \u003cem\u003eet al.\u003c/em\u003e PsychoPy2: Experiments in behavior made easy. \u003cem\u003eBehav Res\u003c/em\u003e 51, 195\u0026ndash;203 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJenkinson, M., Beckmann, C. F., Behrens, T. E. J., Woolrich, M. W. \u0026amp; Smith, S. M. FSL. \u003cem\u003eNeuroImage\u003c/em\u003e 62, 782\u0026ndash;790 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSmith, S. M. Fast robust automated brain extraction. \u003cem\u003eHuman Brain Mapping\u003c/em\u003e 17, 143\u0026ndash;155 (2002).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFischl, B. \u0026amp; Dale, A. M. Measuring the thickness of the human cerebral cortex from magnetic resonance images. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 97, 11050\u0026ndash;11055 (2000).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSaad, Z. S. \u0026amp; Reynolds, R. C. SUMA. \u003cem\u003eNeuroImage\u003c/em\u003e 62, 768\u0026ndash;773 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFischl, B., Sereno, M. I., Tootell, R. B. H. \u0026amp; Dale, A. M. High-resolution intersubject averaging and a coordinate system for the cortical surface. \u003cem\u003eHum. Brain Mapp.\u003c/em\u003e 8, 272\u0026ndash;284 (1999).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHanke, M. \u003cem\u003eet al.\u003c/em\u003e PyMVPA: a Python Toolbox for Multivariate Pattern Analysis of fMRI Data. \u003cem\u003eNeuroinformatics\u003c/em\u003e 7, 37\u0026ndash;53 (2009).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePedregosa, F. \u003cem\u003eet al.\u003c/em\u003e Scikit-learn: Machine Learning in Python. \u003cem\u003eJournal of Machine Learning Research\u003c/em\u003e 12, 2825\u0026ndash;2830 (2011).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTeam, R. C. R: A language and environment for statistical computing. \u003cem\u003eR Foundation for Statistical Computing, Vienna, Austria\u003c/em\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBates, D., M\u0026auml;chler, M., Bolker, B. \u0026amp; Walker, S. Fitting Linear Mixed-Effects Models using lme4. Preprint at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.1406.5823\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.1406.5823\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-science-of-learning","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"npjscilearn","sideBox":"Learn more about [npj Science of Learning](http://www.nature.com/npjscilearn/)","snPcode":"41539","submissionUrl":"https://mts-npjscilearn.nature.com/cgi-bin/main.plex","title":"npj Science of Learning","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"cognitive neuroscience, STEM learning, fMRI, knowledge representation","lastPublishedDoi":"10.21203/rs.3.rs-6992513/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6992513/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eStudents in STEM fields frequently learn new abstract concepts as they build knowledge for scientific innovation. Yet little work has investigated how patterns of neural activity reflect the emergence of this newly learned conceptual information. In a single lesson and lab activity, participants learned about physics concepts, then subsequently completed an fMRI session. We identified neural patterns tracking students\u0026rsquo; newly acquired STEM concept knowledge, using a machine-learning classifier to assess the embedding of concept-relevant categories in students\u0026rsquo; neural representations of the task stimuli. Patterns in several parietal and temporal regions reflected conceptual knowledge acquired during the lesson. Crucially, a regression analysis further demonstrated that greater concept-relevant organization of the stimuli in these brain regions was associated with better performance on behavioral concept knowledge assessments. Results suggest that after only brief exposure to new STEM topics, early evidence of comprehension can be identified in the individualized neural patterns of novice learners.\u003c/p\u003e","manuscriptTitle":"Neural patterns reflect conceptual grasp of novice students following first class learning in physics","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-22 20:08:50","doi":"10.21203/rs.3.rs-6992513/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-09-15T04:24:27+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-05T15:02:16+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-15T02:37:13+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"25313792579217983395535536318696463600","date":"2025-08-08T05:25:48+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"260007605877712447297858123101448485659","date":"2025-07-18T17:43:17+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-07-17T08:43:54+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-07-16T23:32:06+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-07-14T06:41:21+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Science of Learning","date":"2025-06-27T14:14:46+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-science-of-learning","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"npjscilearn","sideBox":"Learn more about [npj Science of Learning](http://www.nature.com/npjscilearn/)","snPcode":"41539","submissionUrl":"https://mts-npjscilearn.nature.com/cgi-bin/main.plex","title":"npj Science of Learning","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1be9486c-e607-4aa6-924e-5b3573d25134","owner":[],"postedDate":"July 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":51860183,"name":"Biological sciences/Neuroscience"},{"id":51860184,"name":"Biological sciences/Psychology"},{"id":51860185,"name":"Social science/Psychology"}],"tags":[],"updatedAt":"2026-02-23T16:06:00+00:00","versionOfRecord":{"articleIdentity":"rs-6992513","link":"https://doi.org/10.1038/s41539-025-00394-3","journal":{"identity":"npj-science-of-learning","isVorOnly":false,"title":"npj Science of Learning"},"publishedOn":"2026-02-22 15:57:52","publishedOnDateReadable":"February 22nd, 2026"},"versionCreatedAt":"2025-07-22 20:08:50","video":"","vorDoi":"10.1038/s41539-025-00394-3","vorDoiUrl":"https://doi.org/10.1038/s41539-025-00394-3","workflowStages":[]},"version":"v1","identity":"rs-6992513","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6992513","identity":"rs-6992513","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}