arxiv: 2604.11598 · v2 · submitted 2026-04-13 · ⚛️ physics.ed-ph

Recognition: unknown

Multidimensional Profiles of Critical Thinking in Physics Labs: Latent Structure, Instructional Change, and Connections to Physics Identity

Marcus Kubsch , Natasha G. Holmes , Antti Lehtinen

Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:11 UTC · model grok-4.3

classification ⚛️ physics.ed-ph

keywords latent profile analysiscritical thinkingphysics labsphysics identitybelongingself-efficacyagencycross-lagged panel model

0 comments

The pith

Latent profiles show two patterns of critical thinking in physics labs that shift with instruction and link to belonging.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper applies latent profile analysis to the three scales of the Physics Lab Inventory of Critical Thinking using over five thousand matched pre- and post-instruction student records. It identifies a two-profile structure at both time points, with substantial movement between lower- and higher-performing groups during a lab course. Cross-lagged panel models then connect these profiles to affective constructs, showing that belonging prospectively predicts recognition, self-efficacy, agency, and higher-profile membership while agency and self-efficacy reinforce each other asymmetrically and recognition operates mainly downstream.

Core claim

At both pre- and post-instruction, a two-profile solution best fits the data across the three PLIC scales. Profile composition shifts substantially over instruction, with nearly half of students moving from the lower- to the higher-performing profile. Belonging emerges as the principal upstream predictor of recognition, self-efficacy, agency, and higher-knowledge profile membership, while agency and self-efficacy form a reciprocal but asymmetric loop and recognition functions primarily as a downstream construct.

What carries the argument

Latent profile analysis applied simultaneously to the evaluating data, evaluating methods, and proposing next steps scales, combined with cross-lagged panel models that link profile membership to belonging, recognition, self-efficacy, and agency across time points.

If this is right

If belonging prospectively predicts profile membership, then interventions that strengthen belonging early in a course could increase the share of students reaching the higher critical thinking profile by the end.
The stronger path from agency to later self-efficacy implies that building agency may produce larger downstream gains in self-efficacy than the reverse direction.
Small associations between course type and profile membership suggest that profile transitions occur across a range of lab instructional formats.
Recognition operating mainly as a downstream outcome indicates that gains in critical thinking profiles are more likely to precede than to follow gains in feeling recognized as a physicist.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Lab curricula might incorporate targeted belonging-building activities in the first weeks to support both identity development and critical thinking profile gains.
The same person-centered latent profile approach could be applied to critical thinking inventories in other laboratory disciplines to test whether similar two-profile structures appear.
Demographic differences in belonging could help explain observed group differences in critical thinking gains if the predictive links hold in more diverse samples.

Load-bearing premise

The two-profile latent structure adequately captures the underlying patterns in the three PLIC scales and the cross-lagged panel models correctly identify temporal precedence without substantial unmeasured confounding.

What would settle it

A replication dataset in which a three-profile solution fits the PLIC responses better than two profiles, or longitudinal measurements showing no significant prospective path from belonging to later higher-profile membership after accounting for baseline levels.

Figures

Figures reproduced from arXiv: 2604.11598 by Antti Lehtinen, Marcus Kubsch, Natasha G. Holmes.

**Figure 1.** Figure 1: Class-specific mean scores and 95% confidence intervals on the three PLIC subscales for the two-profile solution at pre-instruction (left) and post-instruction (right). Profile 1 is shown in blue, Profile 2 in orange. Percentages indicate profile membership proportions. Pre Post P1 12.7% P2 87.3% P1 44.6% P2 55.4% 51.6% 56.4% 43.6% 48.4% [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗

**Figure 2.** Figure 2: Profile transitions from pre- to post-instruction. Bar heights represent profile proportions. Flow bands connect pre and post profiles; percentages indicate transition rates within each pre-test profile. Purple flows indicate crossprofile transitions. transitioned to Profile 1 by post-test. This degree of movement indicates that students’ critical thinking profiles are not fixed; instruction reshapes th… view at source ↗

**Figure 3.** Figure 3: Final cross-lagged panel model. Horizontal gray arrows represent autoregressive stability paths. Colored diagonal arrows represent significant cross-lagged paths (p < .05). Blue paths originate from belonging; dark blue from agency; red indicates the negative agency-to-profile path. Path coefficients are standardized. tern in which agentic participation precedes gains in efficacy beliefs. Recognition as a… view at source ↗

read the original abstract

The Physics Lab Inventory of Critical Thinking (PLIC) measures three components of students' critical thinking in physics labs: evaluating data, evaluating methods, and proposing next steps. Prior work has analyzed these components in isolation or as a composite score. In this study, we apply latent profile analysis (LPA) to the three PLIC scales using a large, multi-institutional dataset of 5,513 matched pre/post student records to identify characteristic response patterns across the three components simultaneously. At both pre- and post-instruction, a two-profile solution best fit the data. Profile composition shifted substantially over instruction, with 48.4\% of students in the lower-performing profile at pre-test transitioning to the higher-performing profile at post-test, while 43.6\% of students moved in the opposite direction. Course type was statistically associated with profile membership at both timepoints, though the effect was small (Cram\'er's $V \approx 0.10$). To examine the relationship between profile transitions and students' affective development, we estimated cross-lagged panel models (CLPMs) linking profile membership to belonging, recognition, self-efficacy, and agency. Belonging emerged as the principal upstream predictor, prospectively predicting recognition, self-efficacy, agency, and higher-knowledge profile membership. Agency and self-efficacy formed a reciprocal but asymmetric loop, with the path from agency to later self-efficacy being stronger. Recognition functioned primarily as a downstream construct over this timescale. These results provide the first person-centered, multidimensional characterization of PLIC performance and demonstrate that epistemic and identity-related constructs are interlinked in physics lab learning.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

LPA on the three PLIC scales yields two profiles with substantial pre-post shifts, and CLPMs flag belonging as the main upstream predictor of profile membership and identity constructs, but the handoff from LPA to observed CLPM variables is a methodological soft spot.

read the letter

The main takeaway is that this paper moves from variable-centered analysis of the PLIC to a simultaneous latent profile analysis on its three scales and finds a two-profile solution that holds at both pre- and post-instruction. Nearly half the students shift from the lower to the higher profile over the course, and the CLPMs then tie those shifts to identity measures, with belonging standing out as the strongest prospective predictor of both profile membership and the other constructs. Agency and self-efficacy show a reciprocal but asymmetric relation while recognition acts mostly downstream. That combination of person-centered profiles and longitudinal cross-lagged links is new for this instrument and dataset. The multi-institutional sample of 5,513 matched records is a clear strength and gives the transition percentages some credibility. Course-type effects being small is also a useful negative result for lab designers. The methods themselves are standard in education research, and the abstract states the models cleanly enough to follow the logic. One real soft spot is the treatment of LPA-derived profile membership as an error-free observed variable inside the CLPM. Without an explicit correction for classification uncertainty (BCH, MLR with posteriors, or a joint model), the cross-lagged coefficients can be biased when entropy is not near 1. The abstract gives no entropy, BLRT, or robustness numbers, so that step needs verification in the full text. Profile labeling and the decision to retain exactly two classes also carry the usual interpretive dependence. This work is for physics education researchers who already use the PLIC or study identity in labs. Readers looking for multidimensional, longitudinal views of critical thinking will get concrete numbers they can build on. The data volume and the attempt to link epistemic and affective factors are substantial enough that a serious editor should send it to referees rather than desk-reject, provided the model diagnostics check out.

Referee Report

2 major / 1 minor

Summary. The manuscript applies latent profile analysis (LPA) to the three PLIC scales (evaluating data, evaluating methods, proposing next steps) on a multi-institutional sample of 5,513 matched pre/post student records, identifying a two-profile solution at both time points. It reports substantial profile transitions during instruction (48.4% low-to-high, 43.6% high-to-low) and small associations with course type, then estimates cross-lagged panel models (CLPMs) linking profile membership to belonging, recognition, self-efficacy, and agency. The central claims are that belonging is the principal upstream predictor of higher-knowledge profile membership and the other constructs, while agency and self-efficacy form a reciprocal but asymmetric loop and recognition is primarily downstream.

Significance. If the LPA solution and CLPM paths are robust, the work supplies the first person-centered, multidimensional characterization of PLIC performance and demonstrates interconnections between epistemic and identity-related constructs in physics lab learning. The large, multi-institutional sample is a clear strength for generalizability. The combination of LPA with longitudinal CLPM modeling is a methodological advance over prior composite-score or single-scale analyses of the PLIC.

major comments (2)

[CLPM Methods and Results] The description of the CLPM estimation (in the section on affective-construct linkages) treats the LPA-derived binary profile membership as an observed variable without any indicated correction for classification error (e.g., BCH three-step, MLR with posterior probabilities, or joint latent modeling). Because the central claim that belonging prospectively predicts higher-knowledge profile membership rests on these cross-lagged paths, moderate entropy would bias the coefficients and undermine the upstream-predictor interpretation.
[LPA Results] No LPA fit indices (BIC, aBIC, BLRT, entropy), missing-data handling, or robustness checks for the two-profile solution are reported in the abstract or the LPA results section. Without these diagnostics, it is impossible to verify that the two-profile model adequately captures the latent structure of the three PLIC scales or that the subsequent CLPM temporal-precedence claims are not sensitive to profile misassignment.

minor comments (1)

[Abstract] The abstract states that course type is 'statistically associated' with profile membership but does not name the test or report the exact p-value alongside Cramér's V.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback on our manuscript. Their comments identify key areas for improving the transparency and robustness of our latent profile analysis and cross-lagged panel modeling. We respond to each major comment below and will incorporate the suggested enhancements in the revised version.

read point-by-point responses

Referee: The description of the CLPM estimation (in the section on affective-construct linkages) treats the LPA-derived binary profile membership as an observed variable without any indicated correction for classification error (e.g., BCH three-step, MLR with posterior probabilities, or joint latent modeling). Because the central claim that belonging prospectively predicts higher-knowledge profile membership rests on these cross-lagged paths, moderate entropy would bias the coefficients and undermine the upstream-predictor interpretation.

Authors: We agree that classification uncertainty in LPA-derived profiles can bias subsequent regression-based estimates such as those in the CLPM, particularly if entropy is moderate. Our original analysis assigned students to profiles using the highest posterior probability and treated membership as observed. In the revision, we will explicitly report the LPA entropy value and add a sensitivity discussion or analysis (e.g., using posterior probabilities as weights or a BCH-style correction where computationally feasible) to assess the stability of the cross-lagged paths, especially the belonging-to-profile link. revision: yes
Referee: No LPA fit indices (BIC, aBIC, BLRT, entropy), missing-data handling, or robustness checks for the two-profile solution are reported in the abstract or the LPA results section. Without these diagnostics, it is impossible to verify that the two-profile model adequately captures the latent structure of the three PLIC scales or that the subsequent CLPM temporal-precedence claims are not sensitive to profile misassignment.

Authors: We acknowledge the need for greater transparency in model selection. Although the abstract notes that a two-profile solution best fit the data, the main text does not present the full set of comparative fit indices, entropy, missing-data procedures, or robustness checks. In the revised manuscript we will expand the LPA results section to include a table of BIC, aBIC, BLRT, and entropy values for one- through four-profile models, describe the use of full-information maximum likelihood for missing data, and report split-sample or bootstrap robustness checks confirming the stability of the two-profile solution. revision: yes

Circularity Check

0 steps flagged

No significant circularity; LPA-derived profiles and CLPM paths are independent of the target constructs.

full rationale

The derivation proceeds from raw PLIC scale responses via standard LPA to obtain two-profile membership at pre- and post-test, followed by separate CLPM estimation on longitudinal data that includes identity measures (belonging, recognition, self-efficacy, agency) as distinct observed variables. Profile membership is not defined in terms of the identity constructs, nor are the CLPM coefficients obtained by fitting the outcome to itself. No self-citation chain, ansatz smuggling, or renaming of known results is load-bearing for the central claims. The analysis is self-contained against external benchmarks once the LPA and CLPM specifications are accepted; classification-error concerns are methodological rather than circular.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

Analysis rests on standard LPA assumptions and post-hoc model selection; no new entities postulated.

free parameters (1)

number of latent profiles
Selected as two based on model fit indices at both time points.

axioms (2)

domain assumption Local independence of PLIC indicators conditional on profile membership
Core assumption of latent profile analysis invoked when fitting the models.
domain assumption Stationarity and correct temporal ordering in the cross-lagged panel model
Required for interpreting prospective predictions between belonging and profile membership.

pith-pipeline@v0.9.0 · 5602 in / 1332 out tokens · 70974 ms · 2026-05-10T15:11:55.805806+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

29 extracted references

[1]

AAPT Committee on Laboratories. (2014). AAPT recommendations for the undergraduate physics laboratory curriculum. American Association of Physics Teachers

2014
[2]

Bandura, A. (1997). Self-efficacy: The exercise of control. Freeman

1997
[3]

B., & Johnson, A

Carlone, H. B., & Johnson, A. (2007). Understanding the science experiences of successful women of color: Science identity as an analytic lens. Journal of Research in Science Teaching, 44(8), 1187--1218

2007
[4]

Doucette, D., Clark, R., & Singh, C. (2020). Hermione and the secretary: How gendered task division in introductory physics labs can disrupt equitable learning. European Journal of Physics, 41(3), 035702

2020
[5]

A., Yale, M

Douglas, K. A., Yale, M. S., Bennett, D. E., Haugan, M. P., & Bryan, L. A. (2014). Evaluation of Colorado Learning Attitudes about Science Survey. Physical Review Special Topics---Physics Education Research, 10(2), 020128

2014
[6]

R., Stanley, J

Dounas-Frazer, D. R., Stanley, J. T., & Lewandowski, H. J. (2017). Student ownership of projects in an upper-division optics laboratory course: A multiple case study of successful experiences. Physical Review Physics Education Research, 13(2), 020136

2017
[7]

Gr\"un, B., & Leisch, F. (2008). FlexMix version 2: Finite mixtures with concomitant variables and varying and constant parameters. Journal of Statistical Software, 28(4), 1--35

2008
[8]

Hammer, D., & Elby, A. (2003). Tapping epistemological resources for learning physics. Journal of the Learning Sciences, 12(1), 53--90

2003
[9]

M., & Shanahan, M.-C

Hazari, Z., Sonnert, G., Sadler, P. M., & Shanahan, M.-C. (2010). Connecting high school physics experiences, outcome expectations, physics identity, and physics career choice: A gender study. Journal of Research in Science Teaching, 47(8), 978--1003

2010
[10]

G., & Wieman, C

Holmes, N. G., & Wieman, C. E. (2018). Introductory physics labs: We can do better. Physics Today, 71(1), 38--45

2018
[11]

G., Olsen, J., Thomas, J

Holmes, N. G., Olsen, J., Thomas, J. L., & Wieman, C. E. (2017). Value added or misattributed? A multi-institution study on the educational benefit of labs for reinforcing physics content. Physical Review Physics Education Research, 13(1), 010129

2017
[12]

G., Wieman, C

Holmes, N. G., Wieman, C. E., & Bonn, D. A. (2015). Teaching critical thinking. Proceedings of the National Academy of Sciences, 112(36), 11199--11204

2015
[13]

Y., Marshman, E., Schunn, C

Kalender, Z. Y., Marshman, E., Schunn, C. D., Nokes-Malach, T. J., & Singh, C. (2019). Why female science, technology, engineering, and mathematics majors do not identify with physics: They do not think others see them that way. Physical Review Physics Education Research, 15(2), 020148

2019
[14]

Y., Marshman, E., Schunn, C

Kalender, Z. Y., Marshman, E., Schunn, C. D., Nokes-Malach, T. J., & Singh, C. (2020). Damage caused by women's lower self-efficacy on physics learning. Physical Review Physics Education Research, 16(1), 010118

2020
[15]

Y., Stump, E., Hubenig, L., & Holmes, N

Kalender, Z. Y., Stump, E., Hubenig, L., & Holmes, N. G. (2021). Restructuring physics labs to cultivate sense of student agency. Physical Review Physics Education Research, 17(2), 020128

2021
[16]

T., & Rhoades, B

Lanza, S. T., & Rhoades, B. L. (2013). Latent class analysis: An alternative perspective on subgroup analysis in prevention and treatment. Prevention Science, 14(2), 157--168

2013
[17]

Lehtinen, A., & Kubsch, M. (2026). Characterizing critical thinkers: Latent profiles of Finnish students' critical thinking in the physics lab. European Journal of Physics, 47(2), 025709

2026
[18]

L., Stout, J

Lewis, K. L., Stout, J. G., Finkelstein, N. D., Pollock, S. J., Miyake, A., Cohen, G. L., & Ito, T. A. (2017). Fitting in to move forward: Belonging, gender, and persistence in the physical sciences, technology, engineering, and mathematics (pSTEM). Psychology of Women Quarterly, 41(4), 420--436

2017
[19]

Lising, L., & Elby, A. (2005). The impact of epistemology on learning: A case study from introductory physics. American Journal of Physics, 73(4), 372--382

2005
[20]

M., & Hazari, Z

Lock, R. M., & Hazari, Z. (2016). Discussing underrepresentation as a means to facilitating female students' physics identity development. Physical Review Physics Education Research, 12(2), 020101

2016
[21]

Pirinen, P., Lehtinen, A., & Holmes, N. G. (2023). Impact of traditional physics lab instruction on students' critical thinking skills in a Finnish context. European Journal of Physics, 44(3), 035702

2023
[22]

N., Wieman, C

Quinn, K. N., Wieman, C. E., & Holmes, N. G. (2018). Interview validation of the Physics Lab Inventory of Critical Thinking (PLIC). In A. Traxler, Y. Cao, & S. Wolf (Eds.), 2017 Physics Education Research Conference Proceedings (pp.\ 324--327). AAPT

2018
[23]

Rainey, K., Dancy, M., Mickelson, R., Stearns, E., & Moller, S. (2018). Race and gender differences in how sense of belonging influences decisions to major in STEM. International Journal of STEM Education, 5(1), 10

2018
[24]

Rosseel, Y. (2012). lavaan: An R package for structural equation modeling. Journal of Statistical Software, 48(2), 1--36

2012
[25]

M., Stein, M

Smith, E. M., Stein, M. M., Walsh, C., & Holmes, N. G. (2020). Direct measurement of the impact of teaching experimentation in physics labs. Physical Review X, 10(1), 011029

2020
[26]

M., Dew, M., Jeon, S., & Holmes, N

Stump, E. M., Dew, M., Jeon, S., & Holmes, N. G. (2023). Taking on a manager role can support women's physics lab identity development. Physical Review Physics Education Research, 19(1), 010107

2023
[27]

J., & Holmes, N

Walsh, C., Lewandowski, H. J., & Holmes, N. G. (2022). Skills-focused lab instruction improves critical thinking skills and experimentation views for all students. Physical Review Physics Education Research, 18(1), 010128

2022
[28]

N., Wieman, C., & Holmes, N

Walsh, C., Quinn, K. N., Wieman, C., & Holmes, N. G. (2019). Quantifying critical thinking: Development and validation of the physics lab inventory of critical thinking. Physical Review Physics Education Research, 15(1), 010135

2019
[29]

E., & Holmes, N

Wieman, C. E., & Holmes, N. G. (2015). Measuring the impact of an instructional laboratory on the learning of introductory physics. American Journal of Physics, 83(11), 972--978

2015