How Do We Build and Test Psychological Theories?
From phenomena and assumptions to models, predictions and revision
- Level
- Intermediate
- Reading load
- 8 core readings in 5 units
- Prerequisites
- How to Read Psychological Science, or equivalent familiarity with experiments, measurement and uncertainty. Elementary algebra helps with the learning model. The guides introduce theoretical and statistical terms; no programming experience is assumed.
- Last reviewed
- September 2026
Introduction
An experiment can reveal a difference while leaving its cause obscure; a model can predict well while leaving the process uncertain. A psychological theory undertakes a further task: to connect evidence with an account of how minds, behavior, and environments work. The more explicit that connection becomes, the more clearly we can question it, compare it with alternatives, and discover where it needs revision.
This syllabus follows the work of building an explanation and bringing evidence to bear on it. Begin with the concepts, measures and assumptions needed for an informative test. Examine a practical construction framework and what is required to express a proposed process as a model that can be run and compared with observations. Then work through a learning model and an experiment that challenges its extension to human causal reasoning. The final units distinguish successful prediction and close agreement between a model and observed data from explanation, and ask what should happen when evidence disagrees with a theory.
By the end, you should be able to identify what a theory is trying to explain, trace a prediction to its assumptions, compare what different accounts expect, and propose a revision or new test that could change the argument. The readings advance positions about scientific practice; part of the task is to evaluate those positions rather than adopt a single recipe.
Unit 1 of 5
What must be clear before a theory can be tested?
A test begins before data collection. The meaning of a concept, the adequacy of a measure and the conditions of a claim affect what either a positive or negative result could establish. Begin by identifying the weakest connection between an explanation and its proposed test.
Reading
1 of 8Why Hypothesis Testers Should Spend Less Time Testing Hypotheses
Scheel, A. M., Tiokhin, L., Isager, P. M., & Lakens, D. (2021). Why hypothesis testers should spend less time testing hypotheses. Perspectives on Psychological Science, 16(4), 744–755. https://doi.org/10.1177/1745691620966795
Methodological contribution · Critical reappraisal · Moderate · Full work: 12 pages
Trace one proposed prediction through its concepts, measures and auxiliary assumptions.
Why you’re reading it. Establish why groundwork can be a scientific contribution in its own right.
A question to consider. For the claim that interruptions impair learning, what must be specified before an experiment could test the explanation?
Unit 2 of 5
How does an explanation become a model?
The construction framework asks how an account might generate a phenomenon. The modeling argument then distinguishes a theory from its specification and implementation. Compare these proposals as ways to make reasoning explicit, and retain room to question their assumptions.
Reading
2 of 8Theory Construction Methodology
Borsboom, D., van der Maas, H. L. J., Dalege, J., Kievit, R. A., & Haig, B. D. (2021). Theory construction methodology: A practical framework for building theories in psychology. Perspectives on Psychological Science, 16(4), 756–766. https://doi.org/10.1177/1745691620969647
Methodological contribution · Foundational argument · Challenging · Full work: 11 pages
Separate the phenomenon to explain from the proposed mechanism and the simulation used to explore it.
Why you’re reading it. Give theory construction a sequence of decisions that can be inspected.
A question to consider. Would reproducing the positive correlations among cognitive tests establish that abilities promote one another’s development?
Reading
3 of 8How Computational Modeling Can Force Theory Building
Guest, O., & Martin, A. E. (2021). How computational modeling can force theory building in psychological science. Perspectives on Psychological Science, 16(4), 789–802. https://doi.org/10.1177/1745691620970585
Methodological contribution · Foundational argument · Challenging · Full work: 14 pages
Ask which commitments become visible at each stage of the proposed pathway.
Why you’re reading it. Prepare to read a formal psychological model as an argument with explicit choices.
A question to consider. How could a model’s implementation reveal an ambiguity that its verbal description concealed?
Unit 3 of 5
What would distinguish competing explanations?
The Rescorla–Wagner chapter makes changes in learning depend on a shared prediction error. Translate the rule into expectations about cue competition. Waldmann and Holyoak then ask what happens when people interpret cues as causes or as effects. Applying an animal-learning model to a human task introduces assumptions that deserve their own examination.
Reading
4 of 8A Theory of Pavlovian Conditioning: Variations in the Effectiveness of Reinforcement and Nonreinforcement
Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. In A. H. Black & W. F. Prokasy (Eds.), Classical conditioning II: Current research and theory (Chapter 3, pp. 64–99). Appleton-Century-Crofts.
Foundational argument · Historical turning point · Challenging · Full work: 36 pages
Track how the combined prediction of cues determines learning about each cue.
Why you’re reading it. Work through a theory that makes related effects follow from a common rule.
A question to consider. When a well-learned cue is joined by a new cue, what changes in the outcome would alter the model’s prediction for learning about the newcomer?
Reading
5 of 8Predictive and Diagnostic Learning Within Causal Models: Asymmetries in Cue Competition
Waldmann, M. R., & Holyoak, K. J. (1992). Predictive and diagnostic learning within causal models: Asymmetries in cue competition. Journal of Experimental Psychology: General, 121(2), 222–236. https://doi.org/10.1037/0096-3445.121.2.222
Landmark empirical study · Foundational argument · Moderate · Full work: 15 pages
Compare the causal interpretations of the tasks with the symmetrical cue-competition prediction under examination.
Why you’re reading it. Evaluate an experiment designed to make competing accounts disagree.
A question to consider. Which assumption must change for the original learning rule to accommodate different judgments when cues are described as causes rather than effects?
Unit 4 of 5
Do good predictions or good fit explain a process?
Predicting new observations is an achievement that fitting old observations does not guarantee. Neither achievement automatically identifies a mechanism. Compare an argument for stronger predictive evaluation with a demonstration of how different accounts can fit similar statistical patterns.
Reading
6 of 8Choosing Prediction Over Explanation in Psychology: Lessons From Machine Learning
Yarkoni, T., & Westfall, J. (2017). Choosing prediction over explanation in psychology: Lessons from machine learning. Perspectives on Psychological Science, 12(6), 1100–1122. https://doi.org/10.1177/1745691617693393
Methodological contribution · Foundational argument · Challenging · Full work: 23 pages
Distinguish prediction on genuinely new observations from fitting the data used to develop a model.
Why you’re reading it. Add a criterion for evaluating performance while keeping explanatory claims separate.
A question to consider. If a model predicts students’ exam scores accurately, what would remain unknown about an intervention intended to improve learning?
Reading
7 of 8Theory Building in the Factor and Network Literature
Fried, E. I. (2020). Lack of theory building and testing impedes progress in the factor and network literature. Psychological Inquiry, 31(4), 271–288. https://doi.org/10.1080/1047840x.2020.1853461
Critical reappraisal · Methodological contribution · Challenging · Full work: 18 pages
Ask what the data can distinguish before interpreting the name of a fitted model.
Why you’re reading it. Expose a different gap: excellent fit can leave the generating process unresolved.
A question to consider. What intervention or repeated observation would make two otherwise similar accounts predict different outcomes?
Unit 5 of 5
When should evidence change a theory?
Return to the connection between a theory and an empirical prediction. A result can be informative only relative to assumptions about what should have happened. Use the concluding argument to distinguish revising a mechanism, checking an auxiliary assumption and adding an explanation that shields a theory from every possible result.
Reading
8 of 8Addressing the Theory Crisis in Psychology
Oberauer, K., & Lewandowsky, S. (2019). Addressing the theory crisis in psychology. Psychonomic Bulletin & Review, 26(5), 1596–1618. https://doi.org/10.3758/s13423-019-01645-2
Critical reappraisal · Methodological contribution · Challenging · Full work: 23 pages
Compare an effect merely suggested by a theory with a prediction that follows from its specified commitments.
Why you’re reading it. Decide what evidence should prompt checking, revising or abandoning an account.
A question to consider. For the learning-model comparison, name one possible auxiliary failure and one new test that would make your proposed revision more than an after-the-fact accommodation.
Scope and omissions
The eight core works span five units and 152 full-work pages: seven journal articles and the complete Rescorla–Wagner book chapter. This is an intermediate path. Formal rules, probability arguments and modeling examples require patient reading; the guidance helps identify the inference each technical choice supports. The author-hosted Guest and Martin copy uses online-first pagination.
The central worked comparison concerns associative learning and human causal reasoning. Other readings use intelligence, measurement and psychopathology to examine theoretical claims. These examples do not make one model, population or laboratory task representative of all psychology. Formal and quantitative approaches receive sustained attention; qualitative theory development, interpretive traditions, implementation skills and a full history of philosophy of science are beyond this path. The selection does not establish that all worthwhile explanations must take the same mathematical form. Optional works broaden the criteria for theory evaluation, including conceptual plausibility, history and researcher responsibility.
Synthesis
A theory gives an inquiry direction by joining a phenomenon to a proposed explanation and working out what should follow. Evidence then becomes a means of choosing among alternatives. Models make assumptions visible, while leaving us responsible for their justification. Prediction may establish usefulness without establishing mechanism, and a good fit may leave several explanations alive.
The worked learning example also shows why disagreement should be specific. Evidence against one set of representations or assumptions is not automatically a verdict on every theory in a broad tradition. Equally, a theory that can accommodate any outcome needs stronger constraints before evidence can improve it. Revisions should clarify what changes and make further consequences available for examination.
Choose an everyday psychological claim, such as interruption making it harder to remember instructions. State the phenomenon separately from two possible explanations. Give each explanation one prediction, identify a comparison on which they disagree, and name the measurement or auxiliary assumption most likely to complicate that comparison. Explain what each possible result would lead you to reconsider. Finally, state who and what your proposed account leaves outside its scope.
For the measurement decisions that connect a construct with observations, continue with How Do We Know What a Psychological Test Measures?. To pursue the worked learning example and its alternatives, explore How Do We Learn What Predicts What?.
For the comparisons and assumptions needed to identify an effect, continue with What Would Justify a Causal Claim?.
Further reading
Optional. Not part of the core sequence.
- Platt, J. R. (1964). Strong inference. Science, 146(3642), 347–353. Paper. A historical argument for developing alternatives and choosing discriminating experiments. Compare its ideal of decisive inference with the auxiliary assumptions examined in the core.
- Meehl, P. E. (1967). Theory-testing in psychology and physics: A methodological paradox. Philosophy of Science, 34(2), 103–115. Paper. The historical critique behind several core arguments: precise prediction and a statistically significant directional difference need not test a theory with equal severity.
- van Rooij, I., & Baggio, G. (2021). Theory before the test: How to build high-verisimilitude explanatory theories in psychological science. Perspectives on Psychological Science, 16(4), 682–697. Paper. An alternative emphasis on explaining capacities and examining theoretical plausibility before empirical tests. Public full text.
- Miller, R. R., Barnet, R. C., & Grahame, N. J. (1995). Assessment of the Rescorla-Wagner model. Psychological Bulletin, 117(3), 363–386. Paper. A detailed appraisal of the model’s assumptions, successes and failures. Extends the worked example without treating historical influence as evidence that every prediction is correct.
- Muthukrishna, M., & Henrich, J. (2019). A problem in theory. Nature Human Behaviour, 3(3), 221–229. Paper. Broadens the discussion to theoretical frameworks, connections across disciplines and the organization of research programs. Its argument for an evolutionary framework is a position to evaluate.
- Guest, O. (2024). What makes a good theory, and how do we make a theory good? Computational Brain & Behavior, 7(4), 508–522. Paper. Proposes criteria involving assumptions, scientific dialogue, relations with evidence and potential harms. Extends evaluation beyond fit or a single score. Publisher full text.