Nonaka & Takeuchi (1995) established the SECI spiral — Socialization, Externalization, Combination, Internalization — as the foundational model for organizational knowledge creation. The critical bottleneck is externalization: converting tacit knowledge (personal, context-specific, hard to formalize) into explicit knowledge (codified, articulable). Externalization is the most important mode because it creates new explicit knowledge from the tacit base. But it is also the most difficult: it is cognitively expensive (articulating what you know implicitly requires significant effort), contextually destructive (the act of articulation strips away the context that gave the knowledge its richness), and operationally costly (documentation competes with productive work for limited time and attention). Nonaka & Konno (1998) introduced ba — the shared context enabling knowledge conversion — arguing that knowledge without ba is merely information. Farnese et al. (2019) found in Frontiers in Psychology that empirical evidence for SECI remains "fragmented" and "inconclusive," with tacit knowledge being "particularly elusive and difficult to test."

The resolution is not better externalization methods but a different architecture: capturing the explicit products of tacit knowledge application (decisions) rather than attempting to externalize the tacit knowledge itself. Decisions are naturally at the boundary between tacit and explicit — they involve tacit judgment that produces observable, documentable commitments.

Walsh & Ungson (1991) identified five internal retention facilities in Academy of Management Review: individuals (personal knowledge — walks out when employees leave), culture (shared beliefs — drifts when conditions change), transformations (procedures — overwritten by process improvements), structures (hierarchies — disrupted by restructuring), and ecology (physical workspace — altered by relocations). Memory is a distributed structural property of the organization, not a database. When any facility degrades, memory degrades regardless of storage capacity. Stein & Zwass (1995) added that organizational memory requires active integration, not just passive retention. Olivera (2000) introduced formal "memory systems" — organizational arrangements for collecting, storing, and distributing knowledge. The deepest limitation: the five-facility model is passive. It describes where memory resides but not how to structurally ensure preservation. Decision lineage as a sixth virtual facility overlays the existing five with active, structured, queryable governance memory.

Szulanski (1996) analyzed 271 observations of 122 best-practice transfers in eight companies in Strategic Management Journal. Contrary to the prevailing assumption that knowledge transfer fails due to motivational barriers (hoarding, "not invented here"), the primary barriers are knowledge-related: (1) lack of absorptive capacity — recipients lack the prerequisite knowledge to understand transferred knowledge, (2) causal ambiguity — even practitioners don't fully understand why their practices work, (3) arduous relationships — poor interpersonal relationships impede nuanced explanation. Von Hippel (1994) defined "sticky information" in Management Science as information costly to acquire, transfer, and use in a new location — the cost is cognitive and contextual, not financial. The knowledge stickiness paradox: routine, codifiable knowledge transfers easily but provides little advantage; deep, contextual, practice-embedded knowledge provides significant advantage but resists extraction. KM systems optimized for easy capture systematically miss the most valuable knowledge.

Decision lineage addresses stickiness by changing the architecture of preservation: rather than extracting knowledge from its context, it preserves the decision context itself — the reasoning, authority, evidence, and expected consequences that make decisions meaningful. This does not eliminate stickiness but preserves enough context to make future retrieval meaningful rather than merely informational.

Cohen & Levinthal (1990) defined absorptive capacity in Administrative Science Quarterly as "the ability to recognize the value of new external information, assimilate it, and apply it to commercial ends" — one of the most cited concepts in management research (60,000+ citations). The critical finding: absorptive capacity depends on prior knowledge. Organizations with deep knowledge can recognize new relevant knowledge; organizations lacking the foundation cannot recognize its significance even when exposed to it. This creates path-dependent learning: what an organization can learn tomorrow depends on what it knows today. Zahra & George (2002) reconceptualized absorptive capacity in Academy of Management Review as a dynamic capability with four dimensions — acquisition, assimilation, transformation, exploitation — distinguishing potential absorptive capacity (acquisition + assimilation) from realized absorptive capacity (transformation + exploitation). Most organizations have a realized-to-potential ratio (η) significantly below 1.0: they acquire more knowledge than they can transform and exploit.

Decision lineage directly addresses the transformation and exploitation dimensions — the realized absorptive capacity gap. Structured decision records with documented context enable organizations to transform absorbed knowledge into actionable governance, and decision lineage enables pattern recognition across past decisions to exploit accumulated knowledge.

Lave & Wenger (1991) fundamentally reframed learning from individual cognitive acquisition to social participation in communities of practice. Knowledge is not a thing individuals possess but a relationship between practitioners and their social practice. Wenger (1998) identified three dimensions: mutual engagement, joint enterprise, and shared repertoire. Brown & Duguid (2000) articulated the distinction between information (codifiable, storable, transmissible) and knowledge (ability to use information effectively in practice — embodied, contextual, social). Organizations persistently confuse information management with knowledge management: building better databases does not create knowledge management because knowledge lives in the practices of communities.

This presents both a challenge and a validation for governance infrastructure. The challenge: if knowledge is situated, any capture system faces inherent limitations. The validation: governance infrastructure does not attempt to extract knowledge from practice communities. It captures the decisions communities make in the course of their practice — the outputs of practice knowledge — with sufficient context to enable future interpretation. Infrastructure for communities of practice, not replacement of them.

Argote (2013) documented significant variation in knowledge depreciation rates across manufacturing contexts, depending on knowledge type and retention mechanisms. Knowledge embedded in individuals (most vulnerable to turnover) depreciates faster than knowledge embedded in tools or routines. Darr, Argote & Epple (1995) demonstrated in Management Science that even in standardized operations (pizza franchises), knowledge depreciates rapidly when production slows or personnel change. Benkard (2000) confirmed organizational forgetting in aircraft manufacturing in the American Economic Review. De Holan & Phillips (2004) distinguished four types of organizational forgetting in Management Science: managed forgetting of established knowledge (deliberate unlearning), managed forgetting of new knowledge (selective absorption), accidental forgetting of established knowledge (institutional amnesia), and accidental forgetting of new knowledge (innovation loss). Hedberg (1981) introduced the concept that organizations must unlearn obsolete knowledge.

The critical insight: not all forgetting is bad, and not all remembering is good. But strategic memory management — choosing what to retain and discard — requires knowing what you know, which requires the organizational memory infrastructure most organizations lack. Decision lineage enables intelligent forgetting: documented decision rationale allows assessment of whether the conditions that justified decisions still hold.

Nelson & Winter (1982) established routines as fundamental knowledge repositories — regular, predictable behavioral patterns that encode coordination, process, adaptation, and cultural knowledge. Routines function as organizational "truces" — stable agreements about how work will be done. Feldman & Pentland (2003) introduced the ostensive-performative distinction in Administrative Science Quarterly: the ostensive aspect is the idealized pattern (documented procedure); the performative aspect is the actual enactment by specific people at specific times. Pentland & Feldman (2005) extended this by establishing routines as a unit of analysis in their own right, observable through their performative patterns rather than only through formal documentation. No two performances are identical. Traditional documentation captures the ostensive aspect; the performative aspect — where actual organizational knowledge resides — goes undocumented.

Governance infrastructure captures decisions made during routine performance — the performative adaptation that procedure manuals miss. When practitioners deviate from standard procedure to handle specific situations, the infrastructure records the deviation, the reasoning, and the authority. Over time, accumulated decision records reveal patterns of performative adaptation that constitute genuine organizational knowledge invisible to traditional documentation.

Weick (1995) established seven properties of organizational sensemaking: grounded in identity construction, retrospective, enactive of sensible environments, social, ongoing, focused on extracted cues, and driven by plausibility rather than accuracy. The retrospection property creates a direct link to organizational memory: sensemaking requires access to the past. Organizations that cannot remember what they decided (institutional amnesia) cannot construct meaningful narratives about why they decided it. Without decision lineage, retrospective sensemaking degenerates into reconstruction — guessing about motivations, assuming about contexts, inferring about evidence. This is the "audit archaeology" problem (S1/S6): retrospective reconstruction in the absence of contemporary records.

Decision lineage provides the raw material for sensemaking — structured decision records with documented intent, authority, evidence, and expected outcomes create the retrospective resource Weick's theory requires.

Wegner (1987) introduced transactive memory systems — collective capacity of groups to encode, store, and retrieve information through "who knows what" networks. Lewis (2003) operationalized TMS through three dimensions in Journal of Applied Psychology: specialization (differentiated expertise), credibility (confidence in others' reliability), and coordination (effective knowledge access). Ren & Argote (2011) reviewed 76 empirical studies over 25 years in Academy of Management Annals. TMS are vulnerable to: personnel changes (knowledge AND meta-knowledge depart), communication breakdown, organizational growth (exponentially harder to maintain), and geographic distribution (remote work erodes informal interaction). Moreland (1999) found that teams trained together develop stronger TMS than teams assembled from individually trained experts — organizational knowledge includes meta-knowledge of how knowledge is distributed.

Decision lineage formalizes the meta-knowledge TMS depend on. Rather than informal understanding of "who knows what," structured records capture who was consulted for specific decision types, what evidence informed decisions, and how authority was delegated. Over time, this creates an explicit transactive memory system that survives personnel changes and organizational growth.

Pollitt (2000) identified four mechanisms of institutional amnesia in Prometheus: (1) decisions not documented, (2) records lost, (3) archives inaccessible, (4) records available but unused. Stark (2019) extended to government in Governance: temporal legitimacy loss, actor turnover, political novelty obsession, reform structure breaks, service outsourcing. The paradox: more data storage does not prevent knowledge loss. Storage is necessary but insufficient — memory requires active preservation infrastructure, not just databases.

50–70% of KM initiatives fail (Alavi & Leidner, 2001; Davenport & Prusak, 1998). Root causes: the tacit knowledge paradox (the most valuable knowledge is least capturable — Cook & Brown 1999), separation from work (Davenport & Prusak 1998 — documentation competes with productive work), knowledge-information conflation (Alavi & Leidner 2001 — MIS Quarterly — KMS implementations that reduce knowledge to storable information), measurement difficulties (preventive value is real but hard to quantify), and cultural barriers. Only 7% of organizations reach APQC Level 5 KM maturity. The Level 3 plateau is architectural: standalone KM systems can reach Level 3 through technology investment, but Level 4 (integration into operations) requires knowledge management to become inseparable from work itself.

The resolution is architectural: shift from knowledge capture (impossibly broad) to decision capture (bounded, structured, governance-relevant). Decisions are naturally at the boundary between tacit and explicit — they involve tacit judgment producing explicit commitments. Capturing decisions with sufficient context (intent, authority, evidence, expected outcome) preserves the governance-relevant output of knowledge application without attempting full tacit knowledge externalization.

Star & Ruhleder (1996) identified key properties of infrastructure in Information Systems Research: embeddedness (sunk into other structures), transparency (supports without requiring attention), scope (extends beyond single events), learned as membership (part of organizational participation), links with practice conventions, embodiment of standards, built on installed base, and visible only upon breakdown. Bowker & Star (1999) demonstrated that classification systems are consequential infrastructure shaping organizational perception and action. The distinction between infrastructure and application is architecturally significant: infrastructure is invisible when working; applications require conscious use. Governance infrastructure for knowledge preservation succeeds where KM applications fail because it is embedded in work (not separate from it), transparent to practitioners (not an additional burden), and learned as organizational membership (not as a special training requirement).

Argyris & Schön (1978) distinguished single-loop learning (error correction within existing frameworks) from double-loop learning (questioning governing assumptions). Most organizations operate with defensive routines (Model I) that block double-loop learning despite espousing openness (Model II). Senge (1990) distinguished adaptive learning (reacting — single-loop) from generative learning (creating — double-loop). March (1991) demonstrated in Organization Science that organizational learning creates a fundamental exploitation-exploration tension: exploitation provides immediate returns but long-term vulnerability; exploration provides long-term adaptability but short-term costs.

Decision lineage enables both loops: for single-loop, records with documented expected outcomes enable systematic comparison between intended and actual results; for double-loop, records documenting governing assumptions enable questioning when patterns of deviation emerge. For the exploitation-exploration balance, historical decision records make the balance visible and manageable.