Comprehensive Notes on Chapter 1 (Laws & Theories) and Singer’s Level-of-Analysis Problem

Laws: Basic Characteristics

  • Defined as invariant or highly constant relations between variables (independent aa ➔ dependent bb)
    • Absolute law: a⇒ba \Rightarrow b always holds
    • Probabilistic law: a⇒b with probability xa \Rightarrow b \text{ with probability } x
  • Based on repeated observation; repetition creates expectation of recurrence
  • In natural sciences, even probabilistic laws carry strong necessity; in social sciences they are only “law-like”
    • Example: income category ⇒ voting Democratic with probability xx
    • Must pass counter-factual tests (e.g., reduce voter bb’s income and expect change in probability)

Two Rival Definitions of “Theory”

  • Quantitative (collection-of-laws) view
    • Theory = set of inter-connected, empirically verified laws
    • Supports step-by-step “building” of theory via accumulating correlations
    • Example narrative: Schliemann’s Troy wall hypothesis; Karl Deutsch’s shift from Yes/No to How-Much questions
  • Explanatory (qualitative) view adopted by the author
    • Theory explains why laws hold; qualitatively different from laws
    • Contains theoretical notions (assumptions, ideal constructs) that are invented, not induced
    • Evaluation criterion = explanatory power, not truth/falsity of individual statements

Limits of Inductivism & Correlational Thinking

  • "Inductivist illusion" (Claude Lévi-Strauss): belief that more data ⇒ truth
  • Correlation ≠ causation; numbers describe but do not explain
    • Correlations are mere mathematical artifacts; may be spurious or genuinely connected—statistics can’t decide
  • Aristotelian vs. Galilean/Newtonian physics used as cautionary tale
  • Without theory, researchers face infinite data & infinite possible combinations; risk of triviality and overwhelm (Ashby’s 20 000-star cluster example)
  • Knowledge must precede theory and theory must guide knowledge – apparent dilemma akin to Plato’s “know everything before anything”

Laws vs. Theories: Key Distinctions

  • Laws: "facts of observation"; empirical; potentially permanent
  • Theories: “speculative processes” explaining laws; may be supplanted by better theories (Conant, Platt, Poincaré)
  • Theoretical notions
    • Concepts (force, gene, national interest)
    • Assumptions (mass at a point, unitary state actor)
    • Neither true nor false; judged by success of encompassing theory

Models, Reality & Simplification

  • Two meanings of model
    1. Representation of theory (organismic, mechanical, mathematical)
    2. Simplified picture of reality (e.g., scale model airplane)
  • Explanatory gain usually comes from departing cleverly from reality, not mirroring it
    • Science = "lowering degree of empiricism in solving problems" (James Conant)
  • Over-realistic models (state-centric political models) risk becoming identical with reality → lose explanatory leverage
Four Main Simplification Devices
  1. Isolation – treat a few factors while holding "other things equal"
  2. Abstraction – leave some elements aside to focus on others
  3. Aggregation – lump disparate items per theoretical criteria
  4. Idealization – assume limit/perfection not found in practice

Theory Construction & Creative Leap

  • Neither deduction nor induction alone yields theory; both needed plus creative intuition
  • Theorist must envision hidden organization of domain; pattern “not visible to naked eye”
  • Simplification highlights central tendencies, propelling principles, essential factors

Meaning of Terms & Theory Dependence

  • Even descriptive terms shift meaning across theories (Pepper, Kuhn)
    • Sun–Earth–Mars sets pre/post-Copernicus example
  • Operational definitions insufficient; must be rooted in theoretical context
  • Weak/conflicting IR theories breed ambiguous concepts: power, pole, stability, structure, system

Due Process of Inquiry & Methodological Choice

  • Must ask at outset:
    1. Is analytic (two-variable) method appropriate?
    2. Are statistical (many-variable) tools suitable?
    3. Does organized complexity require a systemic approach?
  • Choice determines permissible operations and validity of findings (Landau’s “due process of inquiry”)

Seven-Step Protocol for Testing Theories

  1. State the theory explicitly
  2. Infer hypotheses from it
  3. Test hypotheses via observation/experiment
  4. Use theory’s own definitions in steps 2-3
  5. Control perturbing variables outside the theory
  6. Devise multiple, demanding, distinct tests
  7. Interpret failures: total rejection, repair, or narrowing of scope?
Cautions
  • Hasty rejection on first failure often unwarranted; likewise, passing tests ≠ proof of truth (theory can never be proved)
  • "Rigorous" testing of vague theory = methodological fetishism
  • Tests must match precision/generalization level of expectations

Illustration of Bad Practice: Singer, Bremer & Stuckey (1972)

  • Attempt to test "parity-fluidity" vs. "preponderance-stability" views using concentration-of-capability variables
  • Problems: undefined/contradictory theories; arbitrary variables (power concentration ≠ polarity); vague outcome concept (war, conflict, instability conflated); no control for perturbing variables; inconclusive tests but no follow-up redesign

The Level-of-Analysis Problem (J. David Singer)

Conceptual Ladder
  • Micro vs. macro focus: parts (flowers) vs. wholes (garden), individuals vs. system
  • IR scholarship historically drifts vertically without stable focus; often defaults to nation-state level but with “vertical drift”
Requirements of an Analytical Model
  1. Accurate description of phenomena (mapping reality)
  2. Valid, parsimonious explanation (causal understanding prioritized over exhaustive detail)
  3. Capacity for prediction (reliable anticipations; demands less than explanation)
Systemic Level of Analysis (International System as Whole)

Pros

  • Most comprehensive; captures patterns of coalition, power configuration, norms, systemic stability
  • Manageable: avoids complex intra-state details; useful for broad prediction
    Cons
  • Risk of deterministic bias: over-emphasizes system → underplays state autonomy
  • Tends toward homogenizing states ("black-box" or "billiard-ball" images); over-assumes uniform interest defined as power (Morgenthau)
  • Explanation limited to correlations; internal causal mechanisms obscured
National-State Level of Analysis

Pros

  • Allows rich differentiation among actors; opens comparative study of foreign policy
  • Captures motives, ideology, domestic variables absent in systemic focus
    Cons
  • Danger of overdifferentiation; may exaggerate uniqueness and foster ethnocentric "Ptolemaic" parochialism
  • Comparative work requires balanced attention to similarities; often lacking in practice
Implication for Theory Building
  • Choice of level shapes variables deemed significant, causal narratives, and methodological tools
  • Neither level suffices alone; scholars must be explicit about choice, aware of its biases, and possibly integrate multiple levels

Practical & Philosophical Takeaways

  • Empirical work needs theoretical scaffolding; otherwise risks infinite data with no explanation
  • Theories are provisional “artistic creations” (Platt) judged by explanatory reach; always replaceable by better ones (Poincaré)
  • Ethical/policy relevance: Without explanatory theory, control or informed intervention in international affairs remains elusive—prediction alone cannot guide action
  • Scholars must resist both inductivist data-mining and premature methodological rigor; start with clear conceptualization, simplification, and properly matched testing strategy