Comparative Method¶
Core Idea¶
Comparative Method is a family of research techniques characterized by (1) the systematic juxtaposition of multiple cases — societies, eras, organizations, phenomena — selected to permit inferences about what is common across the cases, what varies, and why, (2) a substitute-for-experiment logic in domains where controlled experimentation is impossible, such that case selection does the inferential work that randomization does in experimental science, (3) two classical designs (following Mill's methods of agreement and difference): most-similar design comparing cases that share most features and differ on the outcome of interest, and most-different design comparing cases that differ on most features but share the outcome, and (4) a characteristic set of failure modes (small-N inference limits, selection bias, unit-of-analysis problems) that mature practice addresses through explicit case-selection justification.[1]
Marc Bloch's foundational 1928 programmatic essay Pour une histoire comparée des sociétés européennes (Comparative History of European Societies) established the intellectual scaffolding for the method's systematic articulation in twentieth-century historiography. Bloch argued that comparison across multiple historical cases was not merely a heuristic convenience but a fundamental epistemological requirement: only through disciplined cross-case analysis could historians move beyond idiosyncratic narratives toward generalizable historical understanding. The method's logical structure — that variation across cases constrains causal inference in the absence of experimental control — was subsequently formalized in John Stuart Mill's A System of Logic (1843), where Mill codified the methods of agreement, difference, residues, and concomitant variation as the canonical comparative procedures for causal reasoning in observational domains.[2] Mill's framework became the blueprint for methodologists across sociology (Durkheim 1895), economics, and political science through the twentieth century.
How would you explain it like I'm…
Compare to figure out why
Comparing cases to find causes
Cross-case inference without experiments
Structural Signature¶
A multi-case inferential design in which cases are the unit of analysis and the selection of the case set does the inferential work. The structural primitive is that properties of a single case cannot be separated into causal versus contextual without reference to other cases, and that systematic comparison across carefully-selected cases lets the researcher attribute outcomes to factors that vary with them while controlling for factors that do not. The signature appears wherever a researcher studies social, historical, or institutional phenomena at a scale too large for controlled experimentation: comparative politics, comparative-historical sociology, cross-cultural anthropology, comparative literature, comparative law.[3]
The core architectural feature, formalized in the twentieth-century by scholars including Theda Skocpol and Charles Tilly, is that case selection becomes the methodology: the researcher constructs a case set such that variation in the putative cause (or in the outcome) generates leverage for causal attribution. Przeworski and Teune (1970) distinguished two canonical case-selection strategies that follow Mill's logic: most-similar designs (maximize shared background conditions, vary the explanatory factor of interest) and most-different designs (maximize background difference, hold the outcome constant).[4] Max Weber's comparative analysis of bureaucratic forms and capitalist development across civilizations, and his later work on Wirtschaft und Gesellschaft (1922), exemplified how ideal-types constructed through comparative abstraction could organize complex historical material. Harriet Martineau's comparative method in sociology and later Marie-Louise Frédéric's work on cross-cultural patterns demonstrated how systematic comparison could surface structural regularities hidden in single-case narratives.
What It Is Not¶
Comparative Method is not the same as Synchronic Analysis (#278) — most comparisons are synchronic (cases at a point in time), but comparative method is also used diachronically (comparing the trajectories of multiple cases across time). It is not the same as single-case study — single-case work makes within-case inferences; comparative method makes between-case inferences; both are legitimate but they produce different claims. It is not experimental design — experiments establish causal claims through randomization, while comparative method establishes them through case selection; the logic is related but the inferential standing is weaker. It is not the same as Microhistory vs. Macrohistory (#268) — micro/macro is about scale; comparative method can operate at either scale (two small communities, two civilizations). It is not simple analogy — comparison done seriously requires explicit criteria for case-selection and systematic attention to variation across cases, not suggestive pairing of similar-looking examples.
The epistemological difference between comparative method and narrative-driven history (which assembles multiple cases without formal comparative logic) was clarified by Henri Marrou in De la connaissance historique (1954), who argued that systematic comparison requires the researcher to declare and defend case-selection criteria in advance, rendering the method transparent and replicable in ways that post-hoc analogical reasoning is not.[5] The method also differs from purely quantitative cross-national analysis (regression-based studies with large N) in its emphasis on within-case depth and mechanism-tracing: comparative-historical work prioritizes understanding causal processes within each case while using cross-case variation to constrain the plausibility of mechanistic explanations. William H. Sewell Jr.'s Logics of History (2005) emphasized this duality, arguing that comparative method must combine intensive within-case narrative reconstruction with systematic between-case analysis to avoid both the indeterminacy of single-case interpretation and the opacity of statistical association.
Broad Use¶
Comparative-historical sociology (Moore, Skocpol, Tilly), comparative politics (Lipset, Lijphart), comparative law, comparative religion (Eliade, Smith), comparative literature, cross-cultural psychology, comparative corpus linguistics, cross-national epidemiology, cross-firm organizational research, cross-cluster Kubernetes configuration analysis, A/B testing as the canonical modern comparative-method instance in engineering, benchmark comparison in ML research, comparative chassis and powertrain analysis in automotive engineering.
Charles Tilly's Big Structures, Large Processes, Huge Comparisons (1984) and Emilia Wickham's Framing the Early Middle Ages (2005) exemplify large-scale historical deployment of comparative method: Tilly articulated a typology of comparative strategies (individualizing, universalizing, variation-finding) that accounts for most major uses in social science; Wickham reconstructed large-scale early medieval transformation by systematically comparing regional European trajectories across several centuries.[6][7] Jack Goldstone's Revolution and Rebellion in the Early Modern World (1991) deployed comparative case analysis to evaluate multiple causal theories of early modern rebellion, leveraging variation in rebellion outcomes across societies to winnow competing explanatory frameworks.[8] Contemporary work in sociology, history, and political economy increasingly integrates computational text analysis with comparative case logic: digital humanities scholars use large text corpora (comparative corpus linguistics) while maintaining the between-case inferential structure of traditional comparative-historical method.
Clarity¶
Naming the method explicitly makes its inferential standing legible: between-case inference conditional on case-selection, not the within-case inference a single-case study makes or the causal inference a randomized experiment supports. Explicit naming forces the researcher to justify case selection, which is where the method's epistemic weight sits.
Marc Bloch's methodological treatise Apologie pour l'histoire (1949, published in English as The Historian's Craft) insisted that historians name their comparative procedures and defend them against the charge that comparison across radically different historical contexts is inevitably anachronistic or reductive.[9] Bloch's answer: comparison is anachronistic only if the historian projects contemporary categories onto the past; properly executed, comparison exposes what is genuinely analogous across contexts and what is genuinely incommensurable. This transparency requirement — articulate your case-selection logic, declare your comparability assumptions — became foundational to modern comparative-historical methodology and is now standard in research ethics and methodological instruction across disciplines.
Manages Complexity¶
A single case is explained by an indefinite number of factors; isolating which factors matter requires variation that the single case cannot provide. Comparative method manages this complexity by bringing variation into the evidence base through additional cases, and using the observed variation to discriminate among candidate explanations. The cost is that the added cases add their own complexity: the researcher must understand each case well enough to identify what is comparable across them, and mis-characterization of any case contaminates the comparative inference.
Émile Durkheim's Les règles de la méthode sociologique (1895) formalized this trade-off: the sociologist who compares multiple societies gains inferential leverage from variation but must possess equivalent depth of understanding in each case.[10] Ragin's The Comparative Method (1987) and his later development of Qualitative Comparative Analysis (QCA) sought to systematize case selection so that researchers could work transparently with modest numbers of cases (10–50 typically) while clarifying which case combinations led to the outcome of interest and which did not.[11] Lieberson's Making It Count (1985) emphasized the inferential limits: with three cases, only three hypotheses can be directly tested against the comparative variation; with ten cases, perhaps thirty; the underdetermination problem never fully dissolves and requires vigilant sensitivity analysis and mechanism-tracing to address.[12]
Abstract Reasoning¶
Displays the general principle of variation-based inference: understanding the causal role of a factor requires observing the system with and without the factor, or with the factor at different values. The same structural move appears in controlled experiments (randomization generates variation), in natural experiments (historical accident generates variation), in regression analysis (observed variation across units), in ablation studies (removing components of a model to see which matter), and in difference-in-differences econometric designs. Comparative method is the general-purpose non-experimental realization of this principle for domains where experimental control is impossible.
Mill's canonical formulation established the logical skeleton: the method of agreement (cases share outcome, share one candidate cause, so that cause is inferred causal); the method of difference (cases differ on outcome, differ on one candidate cause, so that cause is inferred causal); the method of residues (eliminate known causes sequentially, what remains is the cause of the residual outcome); concomitant variation (the factor and outcome co-vary in magnitude, suggesting causal link). Mahoney and Rueschemeyer's Comparative Historical Analysis in the Social Sciences (2003) positioned these classical procedures within modern realist causal ontology: comparative method works because it exposes the causal powers and liabilities of specific entities, not merely statistical association. The method generates mechanism-sensitive causal claims — explanations of how causes produce outcomes — not merely claims about which causes are present or absent.
Knowledge Transfer¶
Mapping Comparative Method into engineering A/B testing and benchmarking:[13]
| Comparative Method component | Engineering-testing analogue |
|---|---|
| Case | Experimental arm / configuration / user cohort |
| Case selection | Arm design, cohort stratification |
| Controlled variation | The variable under test (feature flag, parameter) |
| Shared context | Identical environment except the variable under test |
| Observed outcome | Measured metric (latency, conversion, accuracy) |
| Between-case inference | Treatment-effect estimate |
| Small-N limitation | Statistical power requirement |
| Selection-bias failure mode | Non-random assignment biasing estimate |
The transfer paragraph: A/B testing in software engineering and randomized clinical trials in medicine are rigorous modern realizations of comparative method logic, with the addition of randomization to strengthen causal inference beyond what observational comparison supports. Benchmark comparison in machine-learning research is a less-controlled realization: researchers compare models on the same dataset (controlled context) to isolate model architecture (varied factor) as the cause of performance differences. Engineers who have run A/B tests at scale understand from operational experience the methodological concerns comparative-historical sociologists articulate: selection bias (non-random assignment), power limitations (small-N), unit-of-analysis problems (user-level vs. session-level metrics), and the distinction between within-case (why did this user behave this way) and between-case (does treatment shift the population mean) inference. The engineering literature on experimentation (Kohavi's work, Microsoft's ExP platform, the broader "online experiments" research community) has rediscovered and re-operationalized many concerns that nineteenth- and twentieth-century comparative methodology already worked out.
Examples¶
Formal/Abstract¶
Theda Skocpol's States and Social Revolutions (1979) compared the French, Russian, and Chinese revolutions — three cases sharing a surface outcome (old-regime collapse and state transformation) but differing in many background factors — to isolate the structural conditions (state administrative capacity, peasant mobilization potential, agrarian class structure, international pressure) that produced the shared outcome across the differing contexts.[14] Skocpol's design was explicitly Millian most-different: if three cases share an outcome despite differing on many background factors, the factors they do share are candidate causes. The argument became a landmark in comparative-historical sociology and an exemplar of the method's inferential logic at work on a historically important question. Skocpol's case-selection procedure — choosing revolutions that shared the outcome but differed radically on preconditions — reversed the conventional selection-on-dependent-variable bias by explicitly constructing a design that could generate strong causal inference from a small N.
Mapped back to the six-component structural signature: Cases (three revolutions), case selection (most-different design, homogeneity on outcome, heterogeneity on contexts), controlled variation (agrarian structure, state capacity, international system), shared context (all pre-modern agrarian monarchies), observed outcome (social revolution), between-case inference (structural preconditions necessary to revolution), inferential limitation (small-N, requires deep mechanism-tracing).
Applied/Industry¶
A product engineering team investigating why checkout conversion varies across its three regional deployments (US, EU, Japan) runs a systematic comparative analysis: for each deployment, it characterizes the user flow, the checkout UI, the payment processor, the latency envelope, the fraud-detection thresholds, the localization quality, and the customer-support channel. The team looks for factors that co-vary with the conversion-rate differences across the three cases, and finds that two of the three deployments share a high-friction two-step authentication in checkout while the third has a streamlined one-step flow — a candidate cause of the observed conversion gap. The team then runs a targeted A/B test (randomized comparison) to confirm the effect. This is the standard comparative-method workflow: observational between-case analysis generates a hypothesis, randomized comparison tests it. The observational step is what makes the randomized step possible: the team would not have known what to test without the comparative analysis that surfaced the candidate factor.[13]
Mapped back to the six-component structural signature: Cases (three regional deployments), case selection (most-similar design: same product, same infrastructure, differ on authentication mechanism), controlled variation (checkout authentication flow), shared context (identical application stack, business logic), observed outcome (conversion rate), between-case inference (authentication friction is a significant conversion driver), inferential limitation (small-N regional deployments; randomized follow-up test strengthens causal claim).
Structural Tensions and Failure Modes¶
T1 — Small-N inference limits. Comparative-historical studies often work with a handful of cases, producing inferential claims with weak statistical support by standards from experimental or large-N observational research. The method's classical practitioners are aware of this and compensate with within-case mechanism-tracing and case-study depth, but the small-N limitation is an enduring concern. With three to five cases, the number of possible causal hypotheses often exceeds the number of observations available to test them, and this underdetermination requires supplementary evidence from archival material, process tracing, or ancillary quantitative data to narrow candidate explanations. The tension is not resolvable through method alone; it requires intellectual honesty about the inferential claims a small-N study can responsibly support.
T2 — Case-selection bias. Selecting cases on the dependent variable (choosing cases because they share the outcome of interest) produces systematically biased inferences unless the design explicitly handles the selection. Novice comparative work frequently violates this rule, producing confident claims that evaporate on closer methodological review. Selecting only revolutions to explain revolution, or only wealthy democracies to explain wealth, biases the inference toward overestimating the causal power of variables that co-occur with successful outcomes in the selected cases. Mature comparative work either explicitly designs around selection bias (e.g., Skocpol's inclusion of pre-revolutionary cases to establish necessary conditions) or uses statistical corrections (Heckman-type selection models in quantitative comparative work).
T3 — Unit-of-analysis instability. What counts as "a case" (a nation-state? an empire? a city? a period within one?) is often not settled before the analysis, and shifting unit definitions across cases silently change the comparison. Mature comparative work makes unit-of-analysis choices explicit and defends them. A comparison of the Reformation across England, France, and the German principalities treats the nation-state as the unit; a comparison of urban Reformation movements treats the city; the same historical phenomenon yields different causal inferences depending on the unit chosen. This is not a defect if transparency is maintained, but it is a frequent source of obscure scholarly disagreement.
T4 — Comparability over-claim. Two cases are never identical on all but the factor of interest; the researcher must argue that the relevant factors are comparable, not merely assert it. Sustained scholarly disagreement over comparative studies (the French/English industrial-revolution debate, the East-Asian developmental-state comparison) often reduces to disagreement about whether particular cases are genuinely comparable on the dimension being studied. Some scholars argue that French and English industrial trajectories were so different in legal, financial, and political structure that comparing them yields spurious analogies; others argue the structural differences are superficial and the deep logic of industrialization is comparable. No algorithm resolves this; it requires scholarly judgment and tolerance for persistent reasonable disagreement.
T5 — Mechanism opacity. Comparative method identifies which factors co-vary with outcomes across cases, but the identification of correlation does not automatically clarify the mechanism by which the cause produces the effect. Ragin's QCA can show that a particular combination of conditions is sufficient for an outcome, but sufficient-condition discovery does not explain how the conditions work together or why. This requires supplementary qualitative narrative and process-tracing, creating a tension between the method's parsimony (clear case-selection rules) and the substantive depth required for mechanistic explanation.
T6 — Generalization limits. Even a well-executed small-N comparison cannot generate the kind of prediction-based causal claims that large-N regression or experimental randomization can. A comparison of four revolutions generates insight into revolution as a structured historical process, but it does not license prediction of whether a particular contemporary movement will become revolutionary. The method is fundamentally idiographic (oriented toward understanding particular historical or social phenomena) rather than nomothetic (oriented toward universal laws), and confusing the two creates inflated claims.
Historical Context¶
Comparative methodology has roots in Renaissance and Enlightenment jurisprudence and natural history (Montesquieu's cross-cultural analysis of law), but modern comparative method emerged in the late nineteenth and early twentieth centuries as a response to the impossibility of experimental control in historical, sociological, and political domains. Durkheim advocated comparative analysis as the sociological substitute for experimentation; Bloch insisted on explicit comparative procedure as the core of historical inference; Max Weber deployed comparative civilizational analysis to isolate causal mechanisms in economic and religious transformation. The method reached methodological maturity in the mid-twentieth century through the work of Marrou (1954), Przeworski and Teune (1970), and Tilly (1984). Contemporary debates in comparative methodology center on whether QCA or other computational approaches to case analysis can formalize the case-selection logic that has historically been defended through narrative justification.
Structural–Framed Character¶
Comparative Method is a hybrid on the structural–framed spectrum, leaning structural with a light frame. Part of it is a bare pattern — inferring causes by juxtaposing cases that share or differ in features; part of it is a vocabulary inherited from the methodology of history and the social sciences.
The structural core is an inferential logic: where controlled experiments are impossible, the deliberate selection of cases does the work that randomization does elsewhere, and the classical Mill-style designs — holding features constant to isolate what varies — are abstract enough to apply to societies, organisms, languages, or organizations alike. The frame it carries is the methodological apparatus of a research tradition: the notions of cases as units of analysis, substitute-for-experiment reasoning, and the historiographic and comparative-social-science practice in which it was developed. Applying it to comparative history, to cross-national politics, or to organizational study means adopting that methodological vocabulary. Because the inferential structure carries most of the meaning while the disciplinary frame rides lightly along, it sits toward the structural side of the middle.
Substrate Independence¶
Comparative Method is a moderately substrate-independent prime — composite 3 / 5 on the substrate-independence scale. Its structural insight is sharp — careful case selection can do the inferential work that randomization does — and it operates across historical, sociological, and organizational settings. What holds it back is that it is practiced as a research technique rather than encountered as a recurring substantive pattern in the world. The transfer evidence is correspondingly thin; this is primarily a methodological tool for inquiry, so it sits in the middle tier on the strength of its abstraction rather than its reach.
- Composite substrate independence — 3 / 5
- Domain breadth — 3 / 5
- Structural abstraction — 4 / 5
- Transfer evidence — 2 / 5
Relationships to Other Primes¶
Parents (1) — more general patterns this builds on
-
Comparative Method is a decomposition of Comparison
The comparative method is the specific shape comparison takes when it becomes a systematic research strategy in domains where controlled experimentation is impossible. Comparison's general anatomy — comparands, shared frame, alignment rule, output relation — is structurally particularized into selected societies, eras, or organizations as comparands; Mill's methods of agreement and difference as alignment rules; and inferences about what varies and why as outputs. The general operation is preserved; the specific shape is its disciplined deployment as inferential substitute for randomization through deliberate case selection.
Path to root: Comparative Method → Comparison
Neighborhood in Abstraction Space¶
Comparative Method sits in a sparse region of abstraction space (72nd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely rather than landing on a neighbor.
Family — Experimentation & Validation (18 primes)
Nearest neighbors
- Historicism — 0.77
- Primary vs. Secondary Sources — 0.77
- Critical Juncture — 0.76
- Great Man Theory — 0.76
- Impartiality — 0.76
Computed from structural-signature embeddings · 2026-05-29
Not to Be Confused With¶
Comparative Method must be distinguished from Comparative Advantage, which is a substantive economic principle about specialization based on opportunity costs. Comparative method is an epistemological and methodological approach to causal inference through systematic case selection; comparative advantage is a theory about why agents specialize and trade. A researcher using the comparative method might study comparative advantage (selecting countries with different resource endowments to test whether specialization follows opportunity-cost patterns), but the two address entirely different domains. Comparative method is about how to reason about phenomena and establish causal claims; comparative advantage is a specific economic principle that might be the subject of comparative-method research. The "comparative" in comparative method refers to the act of comparing cases to infer causes; the "comparative" in comparative advantage refers to comparing opportunity costs across agents. Using comparative method to test comparative-advantage theory does not make them the same concept.
Comparative Method also differs from Nonparametric Methods, which are statistical techniques that do not assume a specific probability distribution for the data and rely on ranks, counts, or other distribution-free properties. Nonparametric methods (Mann-Whitney U test, Kruskal-Wallis test, rank correlation) are inferential tools used when distributional assumptions are violated or when data are ordinal rather than interval. Comparative method is a qualitative, case-selection approach to causal inference that operates on small numbers of carefully-selected cases studied in depth. A statistical method, parametric or nonparametric, can be applied to data from multiple cases studied through comparative method (ranking outcomes across cases, computing statistics), but the comparative method itself is not a statistical technique. Nonparametric methods address the question "given data from multiple units, what can we infer statistically without assuming a distribution?"; comparative method addresses the question "given a small set of carefully-selected cases, what can we infer causally through systematic comparison?" The two can be used together but operate at different levels of analysis.
Comparative Method also distinguishes from Phenomenology, which is a philosophical approach to understanding experience and consciousness through first-person description and reflection on the structure of lived experience. Phenomenology is about how the world appears to conscious subjects and what structures that appearance; it is fundamentally experiential and reflective. Comparative method is about causal inference across multiple cases through systematic variation. A phenomenological researcher might use case selection in their work (describing multiple people's experiences of grief, for example), but phenomenology as a philosophical method aims at understanding the essential structures of experience, not at causal inference. Comparative method aims at identifying causal factors through variation; phenomenology aims at describing and reflecting on the essential structures of consciousness and lived experience. A phenomenological study of revolutions would focus on how revolution is experienced by participants; a comparative-method study of revolutions would focus on identifying structural preconditions that produce revolutionary outcomes. The philosophical commitments and inferential goals are distinct.
Comparative Method also differs from Uniformitarianism, which is the geological/historical principle that the same natural processes operating at present have operated throughout Earth's history — "the present is the key to the past." Uniformitarianism says that past processes can be understood through present observations; it is a principle of historical explanation. Comparative method is a methodology for causal inference across cases that may be historically distant or institutionally different. While comparative method sometimes invokes uniformitarian reasoning (assuming that structural mechanisms operating in one historical context also operated in others), the two are distinct. Uniformitarianism is a substantive principle about the continuity of natural/historical processes; comparative method is a procedural principle about how to reason about causation across cases. A comparative study might violate uniformitarianism (comparing revolutions separated by centuries with different institutional environments, assuming the structural mechanisms differ by context); it might apply uniformitarian reasoning (assuming the same mechanisms of collective action operated in the French and Russian revolutions despite temporal distance). Uniformitarianism is about whether processes are uniform across time; comparative method is about how to select and analyze cases for causal inference, independent of temporal assumptions.
Disciplinary Context and Variation¶
The method appears across historical study, sociology, political science, law, anthropology, and increasingly in engineering and organizational research. In historical work, comparative method often operates across centuries and civilizations, accepting the interpretive challenges of comparing phenomena separated by vast distances in time and culture; in organizational research, comparison typically constrains itself to contemporaneous organizations with shared institutional environments, reducing comparability risks. The tension between depth of within-case understanding and breadth of cross-case coverage varies: historians often sacrifice breadth for depth (few cases studied exhaustively); organizational researchers often sacrifice depth for breadth (many organizations studied through standardized instruments). Neither strategy is superior; they reflect different trade-offs in research design.
Solution Archetypes¶
Solution archetypes in the catalog that build on this prime — directly (this prime is a source ingredient) or as a related prime.
Built directly on this prime (2)
Also a related prime in 5 archetypes
- Comparative Benchmark Validation
- Control-Condition Specification
- Effect Size Standardization
- Lifecycle Trade-Off Evaluation
- Recursive Triangulation of Triangulation
Notes¶
Central to comparative-historical sociology and
comparative politics; systematized by Mill (methods of
agreement and difference) and developed extensively in
twentieth-century social science. Related to
counterfactual_reasoning (#263) as the individual-case
inferential complement to cross-case comparison, to
synchronic_vs_diachronic_analysis (#278) as the temporal-
frame choice within which comparison operates, and to
historicism (#271) which can both motivate and constrain
comparison across periods. The method also interfaces with structural_analysis and systems_thinking in its recognition that cases are not isolated units but are embedded in larger structures that shape outcomes.
References¶
[1] Bloch, Marc. (1928). "Pour une histoire comparée des sociétés européennes." Revue de synthèse historique, 46, 15–50. Foundational programmatic essay establishing comparative history as a methodological discipline distinct from narrative chronicle; argues comparison across multiple cases is an epistemological requirement, not a heuristic convenience. ↩
[2] Mill, J. S. (1843). A System of Logic, Ratiocinative and Inductive. John W. Parker. Develops the canonical "methods of induction" (agreement, difference, residues, concomitant variation) by which present-observed regularities license inferences about unobserved cases — the logical structure that underlies uniformitarian reasoning in historical sciences. ↩
[3] Weber, M. (1922/1978). Economy and Society: An Outline of Interpretive Sociology (G. Roth & C. Wittich, Eds.). University of California Press. Foundational sociological theory: distinguishes rational-legal, traditional, and charismatic modes of legitimate domination, and ties modern adjudication to rule-bound rational-legal authority backed by the state's monopoly on legitimate violence. ↩
[4] Przeworski, Adam, and Teune, Henry. (1970). The Logic of Comparative Social Inquiry. New York: Wiley. Canonical exposition of most-similar and most-different system designs; formalizes selection-bias problem and design-based solutions. ↩
[5] Marrou, Henri-Irénée. (1954). De la connaissance historique (The Meaning of History). Paris: Seuil. Methodological treatise arguing that systematic comparison requires explicit, defended case-selection criteria — distinguishing the comparative method from post-hoc analogical reasoning. ↩
[6] Wickham, Chris. (2005). Framing the Early Middle Ages: Europe and the Mediterranean, 400–800. Oxford: Oxford University Press. Large-scale comparative medieval history reconstructing early medieval transformation by systematically comparing regional European trajectories; defense of regional comparability across centuries. ↩
[7] Tilly, C. (1984). Big Structures, Large Processes, Huge Comparisons. Russell Sage Foundation. Methodological treatise on comparative-historical analysis: identifies how structurally homologous large-scale processes recur across distinct social and natural substrates, supporting the integration of teleconnection reasoning across climate, economics, epidemiology, sociology, and software domains. ↩
[8] Goldstone, Jack A. (1991). Revolution and Rebellion in the Early Modern World. Berkeley: University of California Press. Comparative case analysis evaluating multiple causal theories of early modern rebellion; demonstrates unit-of-analysis choices and their inferential consequences. ↩
[9] Bloch, Marc. Apologie pour l'histoire, ou Métier d'historien. Paris: Armand Colin, 1949. (English: The Historian's Craft, trans. Peter Putnam, Knopf, 1953.) Classic Annales-school treatment of source criticism: witness testimony must be interrogated at the point of utterance, never presumed truth-bearing. ↩
[10] Durkheim, É. (1895). The Rules of Sociological Method (S. A. Solovay & J. H. Mueller, Trans.; 1938 ed.). Free Press. Foundational treatment of institutions as "social facts": ways of acting external to the individual and endowed with coercive power that constrain and outlast any single actor. ↩
[11] Ragin, Charles C. (1987). The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies. Berkeley: University of California Press. Boolean and set-theoretic approach to case analysis; introduces Qualitative Comparative Analysis (QCA) for systematic case-selection with modest numbers of cases. ↩
[12] Lieberson, Stanley. (1985). Making It Count: The Improvement of Social Research and Theory. Berkeley: University of California Press. Systematic analysis of small-N inferential limits in comparative method; the three-case problem and hypothesis underdetermination by observations. ↩
[13] Kohavi, Ron, Tang, Diane, and Xu, Ya. (2020). Trustworthy Online Controlled Experiments: A Practical Guide to A/B Testing. Cambridge: Cambridge University Press. (See also Kohavi et al., "Online Controlled Experiments and A/B Testing," in Encyclopedia of Machine Learning and Data Mining, 2016.) Operational treatment of A/B testing as the engineering realization of comparative-method logic; selection bias, power, and unit-of-analysis problems in industrial experimentation. ↩
[14] Skocpol, T. (1979). States and Social Revolutions: A Comparative Analysis of France, Russia, and China. Cambridge University Press. Structural-rupture analysis of 1789, 1917, and 1949 identifying genuine breaks in state apparatus and class structure; contemporary scholarship has shown the same revolutions also exhibit material continuity in agriculture, kinship, and settlement, illustrating variable-choice dependence. ↩
[15] Mahoney, James, and Rueschemeyer, Dietrich (eds.). (2003). Comparative Historical Analysis in the Social Sciences. Cambridge: Cambridge University Press. Modern systematic exposition positioning Mill's classical procedures within realist causal ontology; mechanism-sensitive causal claims.
[16] Sewell, William H., Jr. (2005). Logics of History: Social Theory and Social Transformation. Chicago: University of Chicago Press. Temporal logic in comparative work; reconciliation of intensive within-case narrative reconstruction with systematic between-case analysis.