Theory Responsive Case Sampling Design

Select the next case because it can sharpen, challenge, extend, or saturate the emerging account—not because it statistically represents a population.

Version

v1 · 2026-08-24 · History

Solution archetype #

1074

Problem family

Uncertainty, Evidence & Inference Failure

Problem subfamily

Sampling, Selection, Missingness & Generalization

Disposition check

The queue target theoretical_sampling was checked against accepted archetypes, alias and variant indices, component and mechanism indices, reconciliation merge maps, and previous outputs through queue position 41. The closest accepted neighbors are important but do not absorb the target. representative_sampling_design governs samples that credibly stand in for a population or system; the target prime explicitly selects cases by what they teach an emerging model rather than by population representativeness. deviant_case_analysis governs focused inquiry into pattern-violating cases; it can supply a mechanism or variant here, but it does not own the full sequential sampling loop across confirming, contrasting, boundary, and rival-discriminating cases. generalization_validation tests a fitted pattern on new cases; this archetype helps build and scope the pattern before validation. The result is a full draft rather than a disposition-only record.

When This Archetype Applies

Complete catalog groundingAt least one sufficient condition set is fully represented by existing primes or domain-specific abstractions.

Diagnostic problem

An inquiry is trying to build or refine an explanatory account from cases, but the next observation is chosen either as if the goal were population representativeness or as an ad hoc interesting example. Fixed representative sampling can miss the contrast, exception, or boundary condition needed for model refinement; opportunistic case selection can create confirmation bias, scope creep, and untraceable narrative fitting. The structural problem is the absence of a disciplined link between the current state of the model, the learning question for the next case, and the stop condition for enough cases.

Applicability expression6 distinct conditions

Theory-building purposeandSequential case selectionandUnresolved model structureandVariable case learning valueandNonrepresentative sampling goalandAssessable learning saturation

Algebraic1∧2∧3∧4∧5∧6

groundedpartly groundedopen

6 conditions, all required.

6Required in every casenumbered 1–6

These hold no matter which pattern applies.

1

Theory-building purpose · grounded

Theory-building or model-refinement goal more

primeTheoretical Sampling— Select the next case by what it would teach the emerging model rather than by what it represents about a population, interleaving selection with analysis and stopping when new cases add only confirmation.

and — also required

2

Sequential case selection · grounded

Cases can be selected sequentially more

primeTheoretical Sampling— Select the next case by what it would teach the emerging model rather than by what it represents about a population, interleaving selection with analysis and stopping when new cases add only confirmation.

and — also required

3

Unresolved model structure · grounded

Current model has unresolved structure more

primeTheoretical Sampling— Select the next case by what it would teach the emerging model rather than by what it represents about a population, interleaving selection with analysis and stopping when new cases add only confirmation.

and — also required

4

Variable case learning value · grounded

Learning value differs across possible cases more

primeTheoretical Sampling— Select the next case by what it would teach the emerging model rather than by what it represents about a population, interleaving selection with analysis and stopping when new cases add only confirmation.

and — also required

5

Nonrepresentative sampling goal · grounded

Representativeness is not the primary criterion more

primeTheoretical Sampling— Select the next case by what it would teach the emerging model rather than by what it represents about a population, interleaving selection with analysis and stopping when new cases add only confirmation.

and — also required

6

Assessable learning saturation · grounded

Saturation or diminishing learning can be assessed more

primeTheoretical Sampling— Select the next case by what it would teach the emerging model rather than by what it represents about a population, interleaving selection with analysis and stopping when new cases add only confirmation.

Coverage

6 of 6 conditions grounded.

Read the methodology · Download the trigger-logic data

How to use this archetype

Use this archetype when the inquiry is still learning what the account should be. The first move is not to ask, "How many cases are enough?" It is to ask, "What does the current model fail to explain, and what case could teach us the most about that gap?" Each new case should have a learning question before collection or analysis, and each analyzed case should leave a trace of what changed: a new concept, a sharper property, a revised relation, a boundary condition, a rival explanation removed, or only confirmation.

Practical pattern

A practical implementation begins with an emerging model frame. The team writes the current categories, relationships, process story, and uncertainty limits. It then maintains an analytic gap register. Candidate cases are compared for their expected contribution: typical, deviant, extreme, contrasting, boundary-testing, confirming, or rival-discriminating. The chosen case is analyzed, the model is revised, and the next selection decision is made from the new model state. The loop continues until additional strategically chosen cases are no longer adding material conceptual, relational, process, or boundary information.

Key components

Component	Description
Emerging Model Frame ↗	The model frame is the current account under construction. It may be a grounded-theory category system, a process model, a causal explanation, an implementation account, or a domain diagnosis. It does not need to be final, but it must be explicit enough that the next case can be selected for a reason.
Analytic Gap Register ↗	The analytic gap register names what the model does not yet explain. Gaps include ambiguous categories, missing relations, rival explanations, untested contexts, mechanism uncertainty, boundary conditions, and cases that might contradict the account. This register is the main protection against collecting more cases merely because they are available.
Case Learning Question ↗	Each selected case should answer a named learning question. A question might be, "Does this condition still matter when the organizational role changes?" or "Can the model explain a case where the expected constraint is absent?" The learning question links sampling to model refinement.
Case Contrast Palette ↗	The palette keeps multiple case roles visible. Typical cases stabilize categories; deviant cases threaten the model; boundary cases locate scope; extreme cases reveal mechanisms under stress; contrast cases separate dimensions; confirming cases test saturation. A strong design rarely uses only one role.
Model Revision Register ↗	After each case, the revision register records what changed. If nothing changed, that matters too, but only after the case was strategically selected. Repeated confirmation can support saturation; unrecorded confirmation only creates narrative fluency.
Conceptual Saturation Criterion ↗	Saturation is a stop rule at a specified resolution. It should say what counts as "nothing materially new": no new category, no new property, no new relation, no new process step, no new boundary condition, or no change in decision-relevant confidence. Saturation is not a statistical claim and not proof that no possible case could ever matter.

Common mechanisms

• Grounded theory sampling memo: records the current category, property, and reason for the next case.
• Constant comparison matrix: compares new cases against existing cases and model categories.
• Theoretical gap matrix: maps model gaps to candidate cases that could clarify them.
• Negative-case sampling pass: actively seeks cases that could disconfirm or scope the account.
• Maximum-variation case round: explores range and invariants without claiming representativeness.
• Boundary-case probe: chooses edge cases to learn where the account stops applying.
• Rival-explanation discriminator: selects a case because it can distinguish between live explanations.
• Saturation review memo: documents why new cases have or have not stopped changing the model.
• Case selection audit trail: preserves case-order, rationale, and post-analysis effects.
• Transferability claim check: aligns final claims with what the case logic can support.

Parameter dimensions

Important parameters include the granularity of the emerging model, the definition of a case, access constraints, acceptable ethical burden, candidate-case diversity, tolerance for deviant cases, saturation resolution, memoing discipline, stakeholder review level, and how strongly transferability claims must be bounded. Tightening the saturation criterion increases rigor but can prolong inquiry. Increasing case variation can reveal boundaries but may reduce comparability unless the analysis framework is maintained.

Invariants to preserve

Preserve the link between model state and case choice. Preserve selection rationales before they are retrofitted to findings. Preserve a legitimate path for disconfirming and boundary cases. Preserve the distinction between conceptual saturation and population representativeness. Preserve unresolved anomalies rather than forcing them into the model. Preserve enough audit trail that later readers can reconstruct why the case sequence unfolded as it did.

Target outcomes

The archetype should produce an inquiry trail where each case has a clear job, each job is tied to a live model uncertainty, and each analysis changes the next choice or supports a documented stop decision. The final account should have sharper categories, better boundary conditions, clearer mechanisms, and more honest transferability claims than a pile of loosely accumulated cases.

Tradeoffs and failure modes

The central tradeoff is adaptive learning versus auditability. Adaptivity lets the inquiry follow what it is learning, but it also creates room for cherry-picking. The second tradeoff is depth versus breadth: a highly informative case can deepen mechanism understanding while leaving scope underexplored. The third tradeoff is saturation versus premature closure. Stopping matters, but so does checking whether countercases and boundary cases remain unexamined.

Common failures include confirmation-only case sequences, post-hoc selection rationales, overclaiming representativeness, endless sampling without a stop rule, hidden access constraints, and model immunization against counterexamples. The mitigations are explicit learning questions, negative-case triggers, scope notes, saturation reviews, and a visible model revision register.

Neighbor distinctions

Representative Sampling Design asks whether observations can stand in for a population or system. Theory-Responsive Case Sampling Design asks what case will most improve the emerging account. A study can use both, but they answer different questions.

Deviant Case Analysis explains a case that violates expectations. This archetype can choose a deviant case as one step in a longer sampling sequence, but it also chooses typical, boundary, confirming, maximum-variation, and rival-discriminating cases.

Generalization Validation tests a model after it exists. Theory-responsive sampling constructs and scopes the model before or during validation.

Hypothesis Testing Frame structures a claim test against alternatives and error costs. This archetype may generate hypotheses and rival explanations before they are formal enough for testing.

Boundary Critique Audit examines what a boundary includes and excludes. This archetype uses boundary critique to choose cases and to keep transferability claims honest.

Examples

An ethnographer adds a night-shift team after early fieldwork suggests that informal workarounds depend on supervisor absence. A product researcher selects both fast-success users and abandoned users to refine a model of onboarding friction. A policy team selects jurisdictions with similar formal rules but different implementation capacity to separate rival explanations. A clinical case inquiry adds atypical presentations to revise syndrome boundaries. A safety investigation adds near-miss cases from adjacent systems when the current causal model cannot explain workload-triggered failures.

Non-examples

A random survey sample is not this archetype unless it is part of a separate representative-sampling design. A dramatic anecdote selected to support a conclusion is not this archetype. A fixed sample collected before analysis and never revised is not theoretical sampling. An A/B test of a predeclared claim belongs primarily to hypothesis testing. A single outlier investigation may be deviant case analysis without being a full theory-responsive sampling loop.

Quality notes

The target prime is included as a direct source prime. Proposed-prime handling is controlled: this draft introduces no proposed primes. The pre-draft disposition check found close sampling, validation, and deviant-case neighbors but no accepted archetype, alias, component, mechanism, reconciliation-map entry, or previous queue output that should absorb the target prime.

Common Mechanisms

10 documented mechanisms across 4 implementation forms.

The grouping reflects forms represented among the mechanisms currently documented for this archetype; an absent form is not necessarily an impossible implementation.

Analysis, Modeling & Optimization · 2 mechanisms

• Constant Comparison Matrix — Compares each new case against prior cases and the current categories, forcing every difference into a model revision.
• Theoretical Gap Matrix — Maps the model's open gaps against candidate cases to rank which case would teach the most next.

Assessment, Review & Assurance · 2 mechanisms

• Saturation Review Memo — Documents whether newly sampled cases have stopped changing the model, and convenes the decision to stop.
• Transferability Claim Check — Audits the final claims against what the sampled cases can actually support, trimming overreach.

Experiment, Test & Rehearsal · 4 mechanisms

• Boundary Case Probe — Selects a case at the model's suspected edge to find out where the account stops applying.
• Maximum Variation Case Round — Samples deliberately across the widest range of cases to see which findings survive maximum difference.
• Negative Case Sampling Pass — Actively hunts for a case that could disconfirm or puncture the current account rather than confirm it.
• Rival Explanation Discriminator — Chooses the one case whose outcome would separate two still-live rival explanations.

Record, Log & Register · 2 mechanisms

• Case Selection Audit Trail — Preserves the versioned, time-ordered record of the sampling path — memos, access constraints, and each case's model effect — so the sequence can be reconstructed and defended.
• Grounded Theory Sampling Memo — Records the current category, the open gap, and the reason for the next case before it is collected.

Compression statement

Theory-Responsive Case Sampling Design applies when inquiry is building or refining an explanatory model through cases, episodes, sites, documents, interviews, observations, or events. Instead of fixing a representative sample up front, the process alternates analysis and selection: infer the current model, identify its unresolved concepts, relations, boundary conditions, exceptions, and weak comparisons, choose the next case for its expected learning value, analyze it, revise the model and scope, and stop when new cases mainly confirm already stable categories rather than adding material distinctions.

Canonical formula: emerging_model + unresolved_concept_or_relation + case_learning_question + traceable_selection_rationale + analysis_loop + saturation_stop_rule -> scoped_theory_refinement

Related Abstractions

Abstractions this archetype builds on — directly (a source ingredient) or as a related pattern. Links follow the typed catalog namespace.

Built directly on (6)

• Feedback: Outputs influence inputs.
• Inductive Reasoning: Specific to general inference.
• Negative Case Analysis: Deliberately hunt the cases that would overturn your account, then revise or scope it.
• Refinement: Iteratively improving a candidate solution toward adequacy through repeated cycles of evaluation and adjustment that narrow the gap to a target, rather than deriving the answer in one shot.
• Theoretical Sampling: Select the next case by what it would teach the emerging model rather than by what it represents about a population, interleaving selection with analysis and stopping when new cases add only confirmation.
• Validation: Confirming that an artifact actually solves the intended problem in its real operational context, as distinct from confirming it was merely built to specification.

Also references 17 related abstractions

• Abductive Reasoning: Infer the hypothesis that would best explain a surprising observation, accepted provisionally and held defeasibly against better candidates.
• Boundary Critique: Examines inclusion/exclusion assumptions.
• Comparative Method: Systematically juxtaposing selected cases so that their similarities and differences do the causal-inference work that controlled experiments cannot.
• Confirmation Bias: Favor confirming evidence.
• Evidence: A defeasible, provenance-bearing relation between an observable trace and a hypothesis about an unobservable state.
• Falsifiability: A claim is scientific only if it could in principle be empirically refuted.
• Hermeneutic Circle: Whole/part interpretation loop.
• Hypothesis Testing (Null vs. Alternative): Null vs alternative evaluation.
• Iteration: Repeats steps to refine outcomes.
• Overfitting: Poor generalization.

▸ Show 7 more

Variants

Narrower or domain-specific specializations that share this archetype's core structure. Recognized variants are established; candidate variants are provisional.

Negative-Case Theoretical Sampling · risk or failure variant · recognized

Selects cases likely to contradict or force scope revision in the emerging account.

• Distinct from parent: The parent includes multiple case roles; this variant makes countercase search the leading selection criterion.
• Use when: The current model may be over-confirmed; A plausible counterexample class is known but underexamined.
• Typical domains: ethnography qualitative methods, safety investigation
• Common mechanisms: negative case sampling pass, case selection audit trail

Maximum-Variation Theoretical Sampling · subtype · recognized

Selects widely different cases to discover which properties vary and which remain invariant.

• Distinct from parent: The parent can choose cases by many learning roles; this variant prioritizes range and invariance.
• Use when: The model may be too narrow; Important dimensions of variation are known but not yet explored.
• Typical domains: design research, policy learning
• Common mechanisms: maximum variation case round, constant comparison matrix

Boundary-Condition Case Sampling · subtype · recognized

Selects cases at the edge of the model to learn where it stops applying.

• Distinct from parent: The parent can refine any model element; this variant focuses on applicability boundaries.
• Use when: The model seems plausible but scope conditions are unclear; The inquiry risks overclaiming transferability.
• Typical domains: clinical case inquiry, policy learning
• Common mechanisms: boundary case probe, transferability claim check

Confirming Saturation Check Sampling · temporal variant · recognized

Adds final cases mainly to verify that new data no longer change the model at the required resolution.

• Distinct from parent: The parent includes discovery and refinement; this variant emphasizes closure and diminishing marginal learning.
• Use when: The model appears stable; Stopping would be consequential or contested.
• Typical domains: ethnography qualitative methods, organizational research
• Common mechanisms: saturation review memo, case selection audit trail

Near names: Theoretical Sampling Design, Grounded Theory Sampling Design, Emergent-Model Case Sampling, Analytic Case Selection Loop.

Editorial Notes

Problem Classification

Classification: Uncertainty, Evidence & Inference Failure → Sampling, Selection, Missingness & Generalization

Problem kernel: case selection lacks a governed link to model state and learning need

Rationale: The next case is selected either mechanically for representativeness or opportunistically for interest, without a governed link among current model state, learning contrast, boundary condition, and stopping rule. Explanatory case reasoning governs how cases support or challenge a hypothesis; the earlier structural defect here is which cases enter the evidence set and why.

Boundary considered: Uncertainty, Evidence & Inference Failure → Explanatory Hypothesis, Pattern & Case Reasoning

Why this classification prevailed: Sampling selection governs which observations enter inquiry and under what stop rule; explanatory case reasoning governs what pattern or hypothesis may be inferred from the cases once selected.

Review outcome: Adjudicated after independent review; high confidence.