brycewang-stanford Theory Development (jms-theory-development)
When to trigger
- Hypotheses read as bald predictions ("A is positively related to B") with no mechanism
- An inductive study produces rich description but no abstracted theoretical model
- The argument leans on one borrowed citation rather than a developed logic
- A reviewer says "the theory is thin," "this is description, not theory," or "what is the contribution to theory?"
- You are unsure whether to write a priori hypotheses or emergent propositions
The JMS theory bar
JMS exists to advance management and organization theory, and — unlike a US empirical journal — it treats the deductive and inductive paths as equally legitimate. The bar is the same for both: a reader must finish understanding why the relationship or pattern holds and under what conditions it changes. Pick the path the design demands; do not force a qualitative study into a hypothetico-deductive template or write post-hoc hypotheses around quantitative results.
Path A — Deductive: build a hypothesis as a mechanism chain
For each hypothesis, write the explicit chain and skip no step:
- Antecedent — the predictor and why it matters in this organizational context.
- Mechanism — the theoretical process linking antecedent to outcome (e.g., institutional logics, attention allocation, identity work, resource dependence, signalling). Name the theory and the process verb (legitimates, depletes, redirects attention, signals).
- Outcome — the dependent construct, defined at a stated level of analysis (individual / team / firm / field).
- Direction & form — sign, and whether linear, curvilinear, or threshold (justify a turning point with two opposing forces, not a significant squared term).
- Boundary — the moderator(s) and the theoretical reason the mechanism strengthens or reverses. "It depends" is not a theory until you say why.
Keep hypotheses to a defensible set (a focal effect plus a mechanism and one or two boundaries); a "kitchen sink" of ten thin predictions signals weak theory. A boxes-and-arrows model figure must map one-to-one to the hypotheses.
Path B — Inductive: abstract data into a grounded theoretical model
For qualitative theory-building, JMS does not want a priori hypotheses — it wants a credible journey from data to a model. The Gioia-style discipline is the field reference:
- First-order codes — informant-centric terms, staying close to the data.
- Second-order themes — researcher-centric theoretical concepts that aggregate the codes.
- Aggregate dimensions — the small set of constructs that organise the themes.
- Process / theoretical model — a figure showing how the dimensions relate dynamically, with propositions (not hypotheses) abstracting the relationships.
Make the data-to-theory chain visible (a data-structure figure plus representative quotes table is the convention) so a reviewer can audit how a quote became a construct. For process/longitudinal work, use temporal bracketing or a visual-mapping strategy and theorise the mechanism of change over time, not just a stage sequence.
Engage rival explanations
Whichever path you take, name at least one alternative theoretical account and show why yours explains the phenomenon better or differently. JMS reviewers reward a paper that takes the strongest counter-theory seriously, not one that ignores it.
Checklist
- Path chosen (deductive mechanism / inductive grounded model) and consistent with the design
- Deductive: every hypothesis has an explicit mechanism, not just a predicted sign; level of analysis stated
- Inductive: first-order → second-order → aggregate-dimensions chain is explicit and auditable; propositions (not hypotheses)
- A theoretical/process model figure maps one-to-one to hypotheses or propositions
- Boundaries/moderators (or scope conditions) are theorised with substantive reasons
- At least one rival theoretical explanation is named and engaged
- No HARKing: deductive hypotheses fixed before analysis; inductive constructs traceable to data
Anti-patterns
- Description masquerading as theory: a vivid case with no abstracted constructs or model
- Bald predictions: "we hypothesise A relates to B" with no why
- Borrowed theory by name only: invoking dynamic capabilities or institutional logics without using its logic
- HARKing: hypothesising after results are known; every hypothesis supported, none surprising
- Stage models without mechanism: a phase sequence that never explains what drives the transitions
- Boundary without reason: a moderator with no theoretical account of why the effect differs
Output format
【Path】deductive (hypotheses) / inductive (grounded model)
【Focal theory / conversation】... (process verb: ...)
【Deductive】H1 antecedent→mechanism→outcome (level); H2 mechanism; H3+ boundary+reason
【Inductive】first-order → second-order → aggregate dimensions → propositions
【Model figure】matches hypotheses/propositions one-to-one? yes/no
【Rival explanation engaged】...
【Next step】jms-literature-positioning, then jms-methods