Skip to content

Independent Evidence Triangulation

Cross-check a scoped claim with multiple meaningfully independent evidence streams, using both convergence and divergence to calibrate confidence and expose hidden dependence, bias, or context.

1. Overview

Independent Evidence Triangulation is the general first-order intervention for claims that should not depend on a single way of seeing. It combines multiple evidence streams only after defining the claim they share, testing whether the streams are meaningfully independent, and recording the different errors each stream can carry. The result is not a vote, a pile of citations, or an automatic proof. It is a bounded confidence update whose warrant comes from quality, effective independence, diagnostic complementarity, and honest treatment of disagreement.

The archetype fills a specific frozen-catalog gap. recursive_triangulation_of_triangulation begins after a triangulation already exists and audits that procedure at a second order. source_provenance_triangulation is a mature specialization for source type, origin, proximity, perspective, and corroboration. Neither owns the broader first-order lifecycle across methods, observers, models, contexts, time periods, and sources. This draft names that parent-level intervention while keeping both neighbors distinct.

Triangulation is valuable because agreement has evidential force only when streams had a meaningful chance to disagree. Three reports copied from one source are one lineage. Two instruments calibrated by the same flawed model can repeat the same bias. Analysts who coordinate before forming judgments can manufacture consensus. The archetype therefore treats independence as a designed and audited property rather than a number of inputs.

Divergence is equally important. A contradiction may reveal poor data, incompatible definitions, time lag, population differences, context sensitivity, model misspecification, or genuine heterogeneity. Suppressing it destroys the diagnostic benefit of triangulation. A mature application makes disagreement traceable and uses it to narrow, revise, or reject the claim.

2. Why This Pattern Exists

Every evidence channel is selective. Instruments sense some variables and not others; interviews expose meaning but depend on recall and interpretation; administrative records reflect operational definitions; experiments trade realism for control; models inherit assumptions; experts carry training and incentive structures. A single channel can be internally careful yet systematically blind to the fact that matters most.

Multiple channels can reduce this exposure when their errors are not perfectly aligned. A field observation can test whether a model’s assumptions survive contact with practice. An interview can reveal implementation conditions absent from outcome data. An independent replication can discover local procedural artifacts. A different population can reveal that an apparently general result is context-bound. These benefits arise from complementary failure modes, not from plurality alone.

The pattern also exists because modern evidence systems create deceptive multiplicity. Reports reuse datasets, dashboards reuse pipelines, papers cite review articles that trace to one experiment, and nominally independent organizations share vendors, incentives, or analytic conventions. Without lineage and common-cause analysis, the same signal can be counted repeatedly and confidence can grow without new information.

Finally, action often cannot wait for certainty. Decision makers need a disciplined way to synthesize partial and conflicting evidence while preserving limits. Triangulation creates that middle ground: stronger than intuition from one stream, more honest than declaring proof from convergence, and more usable than leaving every contradiction unresolved.

3. Problem Signature

The archetype applies when one scoped claim matters to a decision and no single stream has adequate coverage or error resistance. Typical signals include reliance on one method, duplicated sources presented as corroboration, unresolved contradictions, large confidence claims from weak lineage, or a high-consequence decision whose evidence architecture has never been mapped.

The structural problem has four layers. First is partial observability: each stream sees only part of the target. Second is correlated error: streams may share hidden causes of failure. Third is synthesis discretion: analysts can choose definitions, weights, exclusions, and stopping points after seeing results. Fourth is confidence translation: decision makers may mistake stream count or apparent consensus for warrant.

A useful diagnostic asks: What exact claim is being tested? Which observations could falsify or narrow it? What paths connect the streams? Which errors are likely shared, and which are distinct? What would material disagreement look like? How would the conclusion change if the strongest stream were removed? If these questions cannot be answered, apparent triangulation is probably only aggregation.

Anti-signatures matter. If the task is to inspect an already-completed triangulation, use the recursive audit. If the problem is solely historical or documentary provenance, use source provenance triangulation. If the claim is causal, triangulation may support robustness but cannot replace identification. If streams address different claims or incomparable units, they should not be synthesized merely because they concern the same topic.

4. Intervention Signature

Begin by freezing a claim and decision scope: object, population, context, time, outcome, and consequence. Inventory every proposed evidence stream and state what it can observe. Then map lineage and dependence across upstream sources, samples, instruments, transformations, teams, incentives, and models. Nominally separate streams that share a decisive error path should be grouped or discounted.

Select a portfolio for diagnostic complementarity. Each stream must meet an evidence-quality floor, but streams need not be identical in form or precision. The design should explain why each addition can reveal something another stream cannot. Precommit rules for comparability, convergence, material divergence, weighting, exclusion, and stopping before the favored result is known.

Evaluate streams with enough separation to reduce anchoring where feasible. Preserve their uncertainty and scope limits. Compare them at compatible units, and route contradictions into a structured investigation of definitions, timing, populations, transformations, incentives, missingness, and context. Some divergence will invalidate a stream; some will reveal a valid boundary; some will leave residual uncertainty that must remain in the conclusion.

Synthesis follows, but it is not arithmetic by default. Weight evidence by relevance, quality, independence, and uncertainty. Run sensitivity checks for dependency assumptions, weights, and exclusions. Produce a traceable confidence statement that says what is supported, under which scope, what remains unresolved, and what evidence would trigger review. This is the intervention’s output—not a claim of certainty.

5. Components

1. Claim and Decision Scope

Defines the precise claim being tested, the decision it may inform, the population, time horizon, and acceptable scope of inference. Prevents different evidence streams from appearing to converge merely because they answer different questions or use incompatible units of analysis. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. claim_and_decision_scope is structural: a tool may record it, but the record is not the component itself.

2. Evidence Stream Inventory

Enumerates each source, method, observer, dataset, model, or context treated as an evidence stream and records what it can actually observe. Makes the evidence architecture explicit before synthesis and prevents convenient streams from being added or dropped after results are known. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. evidence_stream_inventory is structural: a tool may record it, but the record is not the component itself.

3. Independence Criterion

States what meaningful independence requires for the current claim, including separation of origin, method, incentives, transformations, and error channels. Turns independence from a vague aspiration into a testable condition appropriate to the consequence of the decision. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. independence_criterion is structural: a tool may record it, but the record is not the component itself.

4. Dependency and Common-Cause Map

Maps shared upstream data, copied claims, common instruments, overlapping samples, shared training, coordinated incentives, and other paths that can correlate errors. Prevents repeated or derivative evidence from being counted as multiple confirmations. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. dependency_and_common_cause_map is structural: a tool may record it, but the record is not the component itself.

5. Method Diversity Plan

Chooses methods whose different strengths and failure modes can expose one another rather than merely repeat the same measurement logic. Creates diagnostic complementarity while keeping the methods aligned to the same scoped claim. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. method_diversity_plan is structural: a tool may record it, but the record is not the component itself.

6. Source Diversity Plan

Selects sources with relevant differences in proximity, perspective, access, incentives, and information lineage. Broadens what can be observed without mistaking demographic or institutional variety for evidential independence. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. source_diversity_plan is structural: a tool may record it, but the record is not the component itself.

7. Evidence Quality Floor

Sets minimum provenance, validity, completeness, and integrity requirements that every stream must meet before it can influence synthesis. Stops many weak signals from overwhelming one strong signal and keeps triangulation from laundering low-quality evidence. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. evidence_quality_floor is structural: a tool may record it, but the record is not the component itself.

8. Stream-Specific Uncertainty Profile

Records the uncertainty, missingness, bias exposure, sensitivity, and scope limits of each stream separately. Preserves the asymmetry among streams and prevents synthesis from erasing the reasons each one might be wrong. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. stream_specific_uncertainty_profile is structural: a tool may record it, but the record is not the component itself.

9. Convergence and Divergence Rule

Defines in advance what counts as material agreement, compatible partial agreement, informative divergence, or unresolved contradiction. Prevents post hoc declarations of convergence and gives disagreement an explicit analytical role. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. convergence_and_divergence_rule is structural: a tool may record it, but the record is not the component itself.

10. Contradiction Investigation Path

Routes important disagreement into checks of definitions, timing, populations, transformations, provenance, incentives, and hidden context. Treats divergence as evidence about model boundaries or error structure rather than as noise to be averaged away. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. contradiction_investigation_path is structural: a tool may record it, but the record is not the component itself.

11. Weighting and Non-Dominance Rule

Specifies how relevance, quality, independence, and uncertainty affect evidential weight and when no single stream may dominate. Makes combination principled without pretending all streams are equal or allowing a familiar method to silence the rest. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. weighting_and_non_dominance_rule is structural: a tool may record it, but the record is not the component itself.

12. Stopping and Sufficiency Rule

Defines when the evidence set is sufficient for the scoped decision, when another stream would add little value, and when uncertainty still requires escalation. Limits endless corroboration while preventing premature closure after the first agreeable result. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. stopping_and_sufficiency_rule is structural: a tool may record it, but the record is not the component itself.

13. Confidence Calibration Statement

Translates the synthesis into a bounded confidence claim tied to evidence quality, independence, convergence, divergence, and unresolved alternatives. Produces a decision-usable conclusion without upgrading convergence into certainty or causal proof. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. confidence_calibration_statement is structural: a tool may record it, but the record is not the component itself.

14. Traceable Evidence Synthesis

Links every synthesized conclusion to the contributing streams, weights, dependency judgments, contradictions, and exclusions. Keeps the reasoning auditable and lets later reviewers reproduce or contest the confidence update. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. traceable_evidence_synthesis is structural: a tool may record it, but the record is not the component itself.

15. Review and Update Trigger

Specifies events that require rechecking the triangulation, such as new evidence, source drift, method failure, context change, or discovery of hidden dependence. Keeps a once-defensible convergence judgment from becoming permanently authoritative after its evidence conditions change. In practice, reviewers should ask what would change if this component were missing, whether it was defined before the result was known, and whether it remains visible in the final confidence claim. review_and_update_trigger is structural: a tool may record it, but the record is not the component itself.

6. Mechanisms

1. Evidence Stream Matrix

A table that records each stream’s claim coverage, origin, method, population, time, quality, uncertainty, and known dependencies. As a template, evidence_stream_matrix implements only part of the archetype. It must inherit the scoped claim, quality floor, dependency analysis, contradiction rules, and audit requirements rather than becoming a shortcut around them.

2. Source Dependency Graph

A directed lineage map showing copied claims, shared datasets, common instruments, overlapping samples, and common-cause error paths. As a method, source_dependency_graph implements only part of the archetype. It must inherit the scoped claim, quality floor, dependency analysis, contradiction rules, and audit requirements rather than becoming a shortcut around them.

3. Multi-Method Study Design

A protocol that assigns complementary qualitative, quantitative, observational, experimental, or model-based methods to one scoped claim. As a protocol, multi_method_study_design implements only part of the archetype. It must inherit the scoped claim, quality floor, dependency analysis, contradiction rules, and audit requirements rather than becoming a shortcut around them.

4. Independent Replication Protocol

A procedure for obtaining a separately executed test with controlled information sharing and explicit comparability conditions. As a protocol, independent_replication_protocol implements only part of the archetype. It must inherit the scoped claim, quality floor, dependency analysis, contradiction rules, and audit requirements rather than becoming a shortcut around them.

5. Cross-Source Corroboration Table

A claim-by-source table that distinguishes independent support, derivative repetition, contradiction, silence, and non-comparability. As a template, cross_source_corroboration_table implements only part of the archetype. It must inherit the scoped claim, quality floor, dependency analysis, contradiction rules, and audit requirements rather than becoming a shortcut around them.

6. Convergence–Divergence Rubric

A precommitted rubric for rating agreement, partial compatibility, material conflict, and unresolved divergence. As a test or assessment, convergence_divergence_rubric implements only part of the archetype. It must inherit the scoped claim, quality floor, dependency analysis, contradiction rules, and audit requirements rather than becoming a shortcut around them.

7. Contradiction Resolution Workshop

A facilitated review that tests whether disagreement comes from definition, timing, sampling, incentives, transformation, or real context dependence. As a ritual, contradiction_resolution_workshop implements only part of the archetype. It must inherit the scoped claim, quality floor, dependency analysis, contradiction rules, and audit requirements rather than becoming a shortcut around them.

8. Blinded Parallel Analysis

A procedure in which analysts or teams evaluate the claim independently before comparing results, reducing anchoring and coordination. As a procedure, blinded_parallel_analysis implements only part of the archetype. It must inherit the scoped claim, quality floor, dependency analysis, contradiction rules, and audit requirements rather than becoming a shortcut around them.

9. Confidence Update Worksheet

A structured record of prior confidence, stream-specific likelihood, dependency discounts, contradictions, sensitivity, and resulting bounded confidence. As a template, confidence_update_worksheet implements only part of the archetype. It must inherit the scoped claim, quality floor, dependency analysis, contradiction rules, and audit requirements rather than becoming a shortcut around them.

10. Triangulation Audit Trail

A versioned record of evidence inclusion, exclusions, weights, judgments, challenges, decisions, and later updates. As a document, triangulation_audit_trail implements only part of the archetype. It must inherit the scoped claim, quality floor, dependency analysis, contradiction rules, and audit requirements rather than becoming a shortcut around them.

7. Parameters

Decision consequence

Consequence determines the required burden of evidence. Exploratory learning can tolerate lighter documentation and wider uncertainty. Credentialing, diagnosis, public safety, or irreversible policy requires stronger independence, provenance, review, and explicit dissent. The parameter should change rigor, not permit unsupported certainty.

Effective independence

Independence ranges from derivative repetition to meaningfully separate origin, method, execution, and incentives. It is claim-specific: two streams may be independent for one error and dependent for another. Record the dimensions tested and the residual common-cause risk instead of assigning a universal binary label.

Stream quality and relevance

Quality includes provenance, construct alignment, completeness, integrity, and sensitivity to bias. Relevance asks whether the stream bears on the exact scoped claim. A high-quality stream about a neighboring claim should not dominate; a relevant but weak stream should not be laundered by convergence.

Comparability tolerance

Streams often use different units, populations, definitions, and time windows. Set how much transformation is acceptable and what residual mismatch blocks direct synthesis. Transformations must remain traceable because harmonization can manufacture agreement.

Convergence threshold

Define material agreement in substantive rather than cosmetic terms. Exact numeric agreement is rarely required, but directional similarity alone may be too weak. Thresholds should reflect decision sensitivity and expected variation.

Divergence materiality

Not every difference needs escalation. Specify which contradictions could change the decision, expose a subgroup harm, reveal a scope boundary, or invalidate a stream. Preserve smaller differences as uncertainty rather than silently discarding them.

Weighting discipline

Weights may be qualitative, ordinal, probabilistic, or model-based. Whatever the form, they should reflect quality, relevance, independence, and uncertainty—not familiarity or institutional status. Report sensitivity to plausible alternatives.

Stopping and update cadence

Stop when expected information gain is small relative to cost and the remaining uncertainty is acceptable for the decision. Reopen when new evidence, source drift, method failure, context change, or hidden dependence could materially alter the confidence claim.

8. Invariants

The claim must remain stable across streams. If analysts quietly change outcome, population, time horizon, or unit, convergence loses meaning. Any scoped revision should be versioned and prior comparisons re-evaluated.

Provenance and transformation must remain traceable. A reviewer should be able to follow a conclusion back through weights and synthesis to each stream’s observations and lineage. Privacy protection may restrict disclosure, but it should not erase accountability.

Stream-specific uncertainty must survive synthesis. A combined statement cannot imply that every stream had the same precision, validity, or scope. Dependency discounts and unresolved contradictions remain visible.

Disagreement is evidence. Material dissent cannot disappear because a majority agrees, a senior actor prefers closure, or a synthesis model cannot represent it. The outcome may be narrower scope, lower confidence, more evidence, or a decision not to act.

The evidence burden grows with consequence. Low-stakes exploration and high-stakes certification may use different mechanisms, but high consequence cannot justify a lower independence or quality threshold merely because action is urgent.

Triangulation does not create causality. Convergent observations can support a descriptive or predictive claim while leaving causal alternatives open. Causal conclusions require appropriate counterfactual, temporal, confounder, or mechanism logic.

9. Outcomes

A successful application produces calibrated confidence, not necessarily agreement. It may raise confidence when high-quality independent streams converge, lower confidence when dependence is discovered, narrow a claim when effects differ by context, or defer action when contradiction remains material.

It also improves the evidence system. Source lineage becomes visible; common-mode errors are detected earlier; analysts learn which methods add information; missing perspectives become explicit; and later reviewers can reconstruct why a decision was warranted at the time.

Operational outcomes include fewer single-source failures, better prioritization of additional evidence, clearer exception and dissent handling, more defensible public explanations, and faster revision when assumptions break. These outcomes should be measured through discoveries and decisions, not by the number of streams collected.

Poor outcomes are also informative. If every stream collapses to one upstream source, the result is a dependency finding. If contexts diverge, the output may be a boundary map. If methods disagree irreducibly, the result may be an uncertainty statement and a new study design. The archetype succeeds when it makes these limits decision-usable.

10. Tradeoffs

Triangulation costs time, access, specialist coordination, and documentation. Additional streams can delay urgent action and create privacy or burden risks. The stopping rule should compare expected information gain with those costs, while high-consequence uncertainty may still justify escalation.

Diversity improves diagnostic reach but complicates synthesis. Different methods can use incompatible constructs or scales. Harmonization enables comparison but can erase distinctive meaning. Preserve raw interpretations and document every transformation.

Independence can conflict with learning. Blinded or parallel analysis reduces anchoring but prevents teams from sharing useful context early. A staged design can preserve independent first judgments, followed by transparent reconciliation.

Explicit dissent improves calibration but may reduce the rhetorical simplicity decision makers prefer. Confidence communication should distinguish productive disagreement from process failure and should not pressure analysts to manufacture unanimity.

Precommitment reduces result-driven flexibility but cannot anticipate every evidence condition. Allow exceptions through a logged challenge path that explains why the rule changed and tests whether the change favors a preferred outcome.

11. Failure Modes

Pseudo-independence is the central failure. It appears when sources copy one origin, datasets share a pipeline, methods reuse a construct error, or analysts share incentives. Map dependency at the level of the claim and error path, not organizational labels.

Majority laundering occurs when many derivative or weak streams overwhelm a high-quality contradiction. Quality floors and non-dominance rules prevent counting from substituting for warrant.

Forced convergence occurs when analysts redefine outcomes, trim time windows, transform scales, or exclude cases after seeing results. Freeze the claim, precommit comparability rules, and version changes.

Contradiction suppression wastes triangulation’s diagnostic value. Every material disagreement needs a disposition: explained artifact, valid context boundary, stream failure, unresolved uncertainty, or claim revision.

Method tokenism adds a nominally different method that observes the same construct through the same error channel. Require a statement of expected diagnostic contribution before adding a stream.

Causal overclaim treats repeated association as proof of mechanism. Separate corroboration from causal identification and use appropriate experimental, counterfactual, confounder, or temporal controls.

Endless collection arises without sufficiency criteria; frozen synthesis arises without update triggers. Both failures confuse process volume with evidence quality. Define when to stop and when to reopen.

12. Variants

Multi-Method Triangulation

Combines methods with deliberately different observation and error structures to test the same scoped claim. Use it when one method cannot observe the whole phenomenon. or method-specific bias is a central threat. Its distinctive feature is independence comes primarily from methodological diversity rather than source separation. It remains a variant because it uses the same independence, dependency, convergence, divergence, weighting, and confidence lifecycle.

Multi-Investigator Triangulation

Uses independently formed judgments from investigators with meaningfully separated assumptions, incentives, or interpretive frames. Use it when interpretive discretion is high. or a single analyst’s framing could dominate the result. Its distinctive feature is the principal diversity lies in observers and reasoning paths. It remains a variant because the independent judgments remain evidence streams combined through the parent synthesis lifecycle.

Cross-Context Triangulation

Tests whether a claim survives across populations, settings, scales, or institutional contexts. Use it when context dependence is plausible. or a conclusion may be overfit to one site or population. Its distinctive feature is evidence streams differ mainly by context while the focal claim remains comparable. It remains a variant because it preserves scoped claims, independence checks, divergence analysis, and calibrated synthesis.

Temporal Triangulation

Combines evidence taken at different times or temporal resolutions to separate persistent structure from transient conditions. Use it when the phenomenon may be episodic, delayed, or path dependent. or single-time observations risk mistaking a transient for a stable pattern. Its distinctive feature is temporal separation supplies the principal contrast among evidence streams. It remains a variant because it uses the parent’s dependence, comparability, convergence, divergence, and confidence logic.

13. Boundaries and Neighbor Distinctions

Recursive Triangulation of Triangulation

The recursive neighbor assumes a first-order triangulation exists. It audits whether streams were independent, convergence logic was valid, and confidence survives a second-order evidence layer. Independent Evidence Triangulation is the thing being audited: it selects and synthesizes first-order streams. Merging them would leave ordinary triangulation unnamed and overload the recursive record.

Source Provenance Triangulation

Source provenance triangulation is a mature specialization for source type, origin, proximity, perspective, corroboration, and confidence. It is especially strong for historical, documentary, and investigative evidence. The present archetype is broader: streams may be sources, methods, observers, instruments, models, contexts, or times. Preserve the specialization where source criticism dominates.

Independent Verification Oversight

Independent verification creates a separated verifier with access, competence, authority, and challenge rights. Triangulation may use such a verifier, but it can also be performed within one team through independent streams. The owner here is evidence architecture and synthesis, not governance separation.

Heuristic Calibration and Confidence Judgment

Heuristic calibration compares confidence in a judgment shortcut with its track record and environment. Triangulation instead evaluates one claim through multiple streams. Calibration is an outcome and supporting logic, not the whole intervention.

Metanarrative Coherence and Internal Consistency Check

Coherence asks whether parts of a story can hold together. A narrative can be internally consistent yet externally unsupported. Triangulation seeks independent bearing on a scoped claim, while coherence remains one possible stream.

Sampling, Confounder Control, and Generalization Validation

Representative sampling designs one stream to stand for a population. Confounder control protects causal estimates from third-variable distortion. Generalization validation tests performance beyond original cases. Each can participate in triangulation, but none replaces cross-stream dependence, divergence, and synthesis logic.

14. Examples

Engineering safety case

A team evaluating a suspected battery failure combines simulation, destructive inspection, charge telemetry, environmental history, and operator reports. The dependency map reveals that simulation inputs and one dashboard share the same faulty temperature estimate. The team discounts that lineage, investigates why inspection contradicts the remaining telemetry, and narrows the claim to one operating regime. The value lies in discovering dependence and scope, not in producing unanimous evidence.

Clinical diagnosis

A clinician compares patient history, examination, imaging, laboratory evidence, and longitudinal response. A discordant laboratory result triggers a sample-handling review rather than being averaged away. The conclusion states the favored diagnosis, an alternative still open, and the evidence that would require reassessment.

Historical inquiry

Researchers compare archival records, independent contemporary accounts, material evidence, and oral histories. Citation lineage shows that several later accounts derive from one memoir. They are one lineage, not multiple confirmations. Divergence by social position becomes evidence about perspective and missing records.

Program evaluation

Administrative outcomes suggest improvement, surveys show no change, and interviews show benefits concentrated among participants with stable transportation. The evaluator does not declare overall success from the largest dataset. The synthesis narrows the claim and identifies an implementation boundary.

Intelligence assessment

Teams conduct blinded parallel analysis, map source handlers and collection channels, and compare results through a convergence rubric. Shared access paths reduce effective independence. Dissent is preserved in the final confidence statement with explicit indicators for update.

15. Non-Examples

Ten articles that all cite one original experiment are not ten streams. A dashboard averaging several metrics from one transformed database is not triangulation. A panel vote is not triangulation when members share evidence, incentives, and framing. Repeating the same measurement with the same systematic error improves precision but not error diversity.

A literature review may support triangulation but is not automatically one; it needs lineage, claim alignment, independence, quality, contradiction, and synthesis rules. A replication is one potentially valuable stream, not the entire archetype. A red-team session is a mechanism when it produces an adversarial evidence stream, not proof by itself.

An audit of whether a prior triangulation followed its rules is recursive triangulation. A source-criticism protocol focused on origin and perspective is source provenance triangulation. An internal consistency scan tests coherence, not independent corroboration. A causal study without cross-stream synthesis may be excellent causal inference but is not this archetype.

Finally, combining unrelated evidence because it concerns the same broad topic is not triangulation. Streams must bear on the same scoped claim with interpretable comparability. Otherwise the combination produces ambiguity rather than confidence.

16. Review and Open Questions

This is a high-confidence gap-fill draft but still requires human acceptance review. The key catalog judgment is the parent-child boundary: q32’s sole direct source is a narrower second-order audit, so the general first-order lifecycle should remain distinct. Review should confirm that independent_evidence_triangulation is the best stable slug and that source_triangulation is retained as an alias rather than allowed to collapse the source-provenance specialization.

Review should also examine minimum independence documentation across consequence levels. Low-stakes inquiry may use a lightweight matrix; high-stakes certification may require blinded execution, provenance controls, formal dependency analysis, and independent challenge. The invariant is not paperwork volume but inspectable warrant.

Another open question is how to represent unresolved divergence in downstream indexes and decision artifacts. A single confidence score can hide subgroup or context boundaries. Catalog integration should preserve structured dissent and scope-limited conclusions.

No proposed prime is introduced. All source and related prime references validate against the frozen canonical list. Components and mechanisms are separated, the four variant records match the extension patch exactly, and the q1–q31 progress prefix remains the continuity authority. The q30 formal collapse stop stays unchanged; q32 is processed only under the bounded q31–q40 authorization.

Common Mechanisms

  • Blinded Parallel Analysis
  • Confidence Update Worksheet
  • Contradiction Resolution Workshop
  • Convergence–Divergence Rubric
  • Cross-Source Corroboration Table
  • Evidence Stream Matrix
  • Independent Replication Protocol
  • Multi-Method Study Design
  • Source Dependency Graph
  • Triangulation Audit Trail

Compression statement

Independent Evidence Triangulation designs an evidence portfolio around one explicit claim, tests whether streams are genuinely independent and comparable, applies precommitted quality and convergence rules, investigates contradictions, discounts shared error, and produces a traceable bounded confidence judgment rather than a vote count.

Canonical formula: confidence_update = f(stream_quality, claim_alignment, effective_independence, convergence, explained_divergence, residual_uncertainty)

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

Built directly on (3)

  • Aggregation: Deliberately collapsing many items into a single summary, choosing which information to discard to gain tractability.
  • Triangulation: Cross-verifying a claim by combining multiple independent sources or methods so their convergence raises confidence and their divergence exposes hidden bias or context.
  • Verification: Check that an object conforms to its specification via a defined procedure yielding evidence and a verdict.

Also references 15 related abstractions

  • Bayesian Updating: Update beliefs with evidence.
  • Calibration: Aligning a system's output to a trusted reference by measuring deviation, adjusting to reduce it, and monitoring for drift.
  • 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.
  • Correlated-Source Attribution Failure: When combined sources share underlying variation, joint inference stays strong while attribution to any individual source becomes unstable, sign-flipping, or arbitrary.
  • Data Integrity: Accuracy and consistency preserved.
  • Epistemic Humility: Calibrating the confidence of one's claims to the actual strength of the evidence and staying open to revision when new information arrives.
  • Experimental Design: Structuring an investigation through deliberate intervention, controlled assignment, and measurement so that causation can be distinguished from mere correlation and confounding.
  • Observability: Infer internal state externally.
  • Paradox of Unanimity: Agreement beyond what independent observation could produce becomes evidence against the conclusion, signaling a broken-independence failure mode.

Variants

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

Multi-Method Triangulation · mechanism family variant · recognized

Combines methods with deliberately different observation and error structures to test the same scoped claim.

  • Distinct from parent: It emphasizes method complementarity and comparability across different measurement or inquiry logics.
  • Use when: One method cannot observe the whole phenomenon; Method-specific bias is a central threat.
  • Typical domains: mixed methods research, safety assurance, program evaluation
  • Common mechanisms: multi method study design, convergence divergence rubric

Multi-Investigator Triangulation · governance variant · recognized

Uses independently formed judgments from investigators with meaningfully separated assumptions, incentives, or interpretive frames.

  • Distinct from parent: It adds role separation, information controls, and adjudication of investigator disagreement.
  • Use when: Interpretive discretion is high; A single analyst’s framing could dominate the result.
  • Typical domains: qualitative coding, intelligence analysis, clinical diagnosis
  • Common mechanisms: blinded parallel analysis, contradiction resolution workshop

Cross-Context Triangulation · domain variant · recognized

Tests whether a claim survives across populations, settings, scales, or institutional contexts.

  • Distinct from parent: It treats heterogeneity and boundary discovery as primary outcomes rather than treating all divergence as error.
  • Use when: Context dependence is plausible; A conclusion may be overfit to one site or population.
  • Typical domains: policy transfer, multi site research, organizational learning
  • Common mechanisms: evidence stream matrix, cross source corroboration table

Temporal Triangulation · temporal variant · recognized

Combines evidence taken at different times or temporal resolutions to separate persistent structure from transient conditions.

  • Distinct from parent: It requires explicit treatment of drift, lag, seasonality, and changing measurement conditions.
  • Use when: The phenomenon may be episodic, delayed, or path dependent; Single-time observations risk mistaking a transient for a stable pattern.
  • Typical domains: longitudinal research, maintenance, risk monitoring
  • Common mechanisms: evidence stream matrix, confidence update worksheet

Near names: Cross-Verification, Convergent Evidence, Corroboration.