Hypothesis Grounding Skill

Ask clarifying questions when:

• The causal mechanism linking X → Y is unclear
• The user lists theories without explaining how they connect
• Variables seem disconnected (potential coherence problem)
• The user hasn't articulated why the null hypothesis is plausible
• Boundary conditions are unspecified

Phase 2: Diagnose and Advise

Assess the user's current draft or idea against the three pillars. Check for the common pitfalls identified in Sparrowe & Mayer (2011):

Pitfall 1: Argument by citation

• Symptom: Multiple sentences begin with "Smith (2002) found..." or "According to Jones (2015)..."
• The theory section reads as a literature review rather than a logical argument
• Fix: Strip citations, check if the argument stands on its own logic, then re-add citations to support (not replace) the reasoning

Pitfall 2: Ignoring the conversation

• Symptom: Highly relevant prior work is not cited or engaged
• Readers will question the contribution if closely related papers are neglected
• Fix: Identify the theoretical conversation the paper enters and articulate how the current work builds upon, extends, or challenges prior work

Pitfall 3: Lack of specificity

• Symptom: The explanatory logic relies on a general theory (e.g., social exchange theory, trait activation theory) without explaining the specific mechanism for the specific variables
• A general principle like "favors beget reciprocation" does not automatically ground a specific claim like "civility predicts job performance"
• Fix: Bridge from the general theoretical principle to the specific operationalization through step-by-step explanatory logic

Pitfall 4: Fragmented theorizing

• Symptom: Each hypothesis in the model draws on a different, unrelated theory
• Creates the impression of post hoc theorizing or data-driven model specification
• Fix: Integrate the multiple theoretical perspectives into a coherent overarching narrative. Explain why multiple theories are needed and how they fit together (see Phase 4)

Pitfall 5: Stating the obvious

• Symptom: The hypothesis seems trivially true; a reviewer cannot imagine when the null could hold
• Fix: "Flirt with the null" — explicitly consider why the relationship might not exist or might reverse. Frame the hypothesis as an alternative to a plausible null

Phase 3: Build the Argument (The Logic of Explanatory Logic)

For each hypothesis, construct the argument using a combination of three substantiation techniques:

Technique 1: Theoretical Anchoring

• Link the hypothesis to an established theory or conceptual framework
• Do NOT merely reference the theory — provide enough verbal explication that the reader understands the mechanism without having to read the original source
• Check: Is the claim in the paper consistent with the established theory's assumptions?

Technique 2: Empirical Support

• Cite empirical evidence supporting similar relationships in analogous contexts
• The implicit argument: "If it occurs in similar circumstances, it should apply here"
• CRITICAL: Empirical evidence is persuasive ONLY when accompanied by a logical rationale explaining why the same mechanism applies in the current context

Technique 3: Process Narrative

• Explain how the hypothesized relationship occurs by describing intervening states and processes
• Identify the mediating mechanisms: What happens between X changing and Y changing?
• If mediators are proposed, they should be operationalized and tested — otherwise reviewers will see them as speculative

The recommended writing process:

1. Write the argument FIRST without any citations — pure explanatory logic
2. Check: Is the logic clear, consistent, and persuasive on its own?
3. Check: Does the argument clearly lead the reader to the hypothesis (no surprise)?
4. Check: Does the reader understand why X and Y are related?
5. Only THEN add citations to give credit, anchor in theory, and show how the work builds on prior research

Phase 4: Handle Multiple Theories

When the user draws on multiple theoretical perspectives, help them choose and execute one of these integration strategies:

Strategy A: Competing Hypotheses

• Two theories make opposing predictions; let data adjudicate
• CAUTION: Must explain when and why one theory should take precedence — merely pitting theories against each other leaves readers puzzled
• Best practice: Explain the conditions under which each theory's predictions are most applicable (cf. Vanneste & Puranam, 2010)

Strategy B: Additive Combination

• Different theories explain different independent variables' effects on the same outcome
• RISK: Avoid combining variables simply because each has been shown to affect Y in prior research — that's not a conceptual framework
• Requirement: Articulate a conceptual framework that brings the perspectives together and explains their relevant differences (cf. Agarwal et al., 2004)

Strategy C: Complementary Integration

• The assumptions of one theory implicitly require those of the other to be fully realized
• Highest bar: Requires thorough understanding of the logic underpinning each theory
• Highest potential contribution: Can generate new questions and insights for each theory's literature (cf. Silverman, 1999)

Strategy D: Division of Labor

• Different theories are most applicable for different (but related) sub-questions that combine to address the phenomenon
• Example: One theory explains when a practice gains traction; another explains which firms adopt it (cf. Sherer & Lee, 2002)

For all strategies: Clearly explain why addressing the research question requires these specific theories and how exactly the theories will be joined to create a unique contribution.

Phase 5: Ensure Coherence

The final quality check ensures the hypothesis set works as a unified whole:

Coherence checklist:

• Is there a clear, overarching research question that drives all hypotheses?
• Can you explain why these variables (and only these) were selected?
• Do the hypotheses link together to create an overall contribution to the topic?
• Would a reader wonder "why wasn't variable Z included?" — if so, address this
• Does the theoretical narrative hold together, or does each hypothesis seem to come from a different paper?
• For each hypothesis: Does the preceding text lead the reader to this specific prediction (no surprise)?
• For each hypothesis: Does the reader clearly understand why X and Y are related (even if they might disagree)?

Phase 6: Output

When generating text for the user, follow these formatting principles:

Structure each hypothesis block as:

1. Opening move (1-2 paragraphs): Position the argument in the theoretical conversation. What is the relevant prior work? What gap or extension does this hypothesis address?
2. The argument (2-4 paragraphs): The core explanatory logic. Why should X predict Y? Walk through the causal mechanism step by step. This is the heart of the section.
3. The hypothesis statement: A formal, precise prediction that follows naturally from the argument. The reader should feel "of course — that's exactly what the argument predicts."

Writing style guidelines:

• Lead with the argument, not with citations
• Use citations to support reasoning, not as the reasoning itself
• Avoid starting sentences with author names unless the specific finding/argument is the point
• Make the causal chain explicit: X → mechanism → Y
• Use transition language that signals logical progression ("Because...", "This suggests that...", "To the extent that...")
• Be specific about boundary conditions and context

Quality Rubric

Use this rubric to evaluate a theory section (yours or the user's):

Reference

This skill is based on: Sparrowe, R. T., & Mayer, K. J. (2011). Publishing in AMJ—Part 4: Grounding hypotheses. Academy of Management Journal, 54(6), 1098–1102.

See also the references directory for the full checklist and examples.