Fiol & Lyles (1985) established the foundational distinction: "Learning involves the development of insights, knowledge, and associations between past actions, the effectiveness of those actions, and future actions." Adaptation is behavioral change that may not involve cognition. Organizations can adapt without learning (superstitious behavior change) and learn without adapting (organizational cynicism — understanding without action). Hedberg (1981) confirmed: "organisations do not have brains, but they have cognitive systems and memories" — learning is an organizational property, not just aggregated individual learning.

Huber (1991) identified that "an entity learns if, through its processing of information, the range of its potential behaviors is changed." Critically, learning need not produce improved performance — organizations can learn things that are wrong. The four processes map to distinct infrastructure requirements: acquisition (how knowledge enters), distribution (how it spreads), interpretation (how it gains shared meaning), and memory (how it persists). Easterby-Smith (1997) documented six distinct disciplinary perspectives on organizational learning (psychology, management science, sociology, strategy, production management, cultural anthropology) — each defining "organizational learning" differently, producing frameworks that rarely interoperate.

Single-loop learning detects and corrects errors without questioning governing variables (values, norms, assumptions). Double-loop learning questions the governing variables themselves. The critical structural insight: governing variables are not stated policies but the actual values driving behavior — the "theory-in-use" that often diverges from "espoused theory." Argyris found that "nearly all study participants espouse Model II values when asked how they would behave, but virtually all operate from Model I in actual problematic situations." Model I governing values (achieve purposes, maximize winning, minimize negative feelings, be rational) create defensive routines that make governing variables undiscussable — "and make the undiscussability undiscussable." The paradox, articulated across Argyris's body of work (1990; see also Argyris & Schön 1996): "The actions we take to promote productive organizational learning actually inhibit deeper learning."

Argyris shows that Model I behavior optimizes locally (avoiding embarrassment → maintaining social cohesion) while systematically preventing global improvement (learning → governing variable change). Chlon (S11) proves that AI training under log-loss optimizes locally (next-token prediction) while systematically breaking the symmetries governance requires. Both exhibit the same pattern: a system that cannot fix itself through more of the same process, because the optimization objective is misaligned with the desired outcome. Both require architectural intervention — changing the system's structure, not scaling its existing process. Original synthesis. No published work makes this parallel explicit.

Intuiting: "the preconscious recognition of the pattern and/or possibilities inherent in a personal stream of experience" (individual). Interpreting: "the explaining, through words and/or actions, of an insight or idea to one's self and to others" (individual → group). Integrating: "the process of developing shared understanding among individuals and of taking coordinated action through mutual adjustment" (group → organization). Institutionalizing: "the process of ensuring that routinized actions occur" (organization). Feed-forward dynamics carry new learning from individual to organization (exploration). Feedback dynamics carry institutionalized knowledge back through perception (exploitation). The tension between feed-forward and feedback is the learning-level expression of March's exploration/exploitation tension.

Crossan, Maurer & White (2011) found that "although some subsequent research has added to the original work, the challenge to develop an accepted theory remains unrealized." Empirical research tested processes in isolation, missing the crucial interlinkages. Lawrence et al. (2005) addressed the power gap: four forms of power (force, manipulation, domination, discipline) affect each I process. Most critically, power suppresses the feed-forward path — people with novel insights self-censor when articulation threatens powerful stakeholders.

March (1991): "Exploration includes things captured by terms such as search, variation, risk taking, experimentation, play, flexibility, discovery, innovation." "Exploitation includes such things as refinement, choice, production, efficiency, selection, implementation, execution." The tension is zero-sum for resources. The returns asymmetry biases toward exploitation: exploitation yields "positive, proximate, and predictable" returns; exploration yields "uncertain, distant, and often negative" returns. Self-reinforcing dynamics: "Each increase in competence at an activity increases the likelihood of rewards for engaging in that activity, thereby further increasing the competence and the likelihood." The competency trap: "It is quite possible for competence in an inferior activity to become great enough to exclude superior activities with which an organization has little experience."

March's most striking finding: "adaptive processes, by refining exploitation more rapidly than exploration, are likely to become effective in the short run but self-destructive in the long run." And: slow learners and personnel turnover improve long-run organizational knowledge because they maintain belief diversity.

Levinthal & March (1993): Temporal myopia — learning overweights recent experience and underweights historical patterns. Spatial myopia — learning from local successes/failures misses system-level patterns. Failure myopia — successful people underestimate risks, and organizations promote successful people, so success biases accumulate in decision-making authority. The core mechanism: "By simplifying experience and specializing adaptive responses, learning improves organizational performance on average. However, the same mechanisms of learning that lead to improvements also lead to limits to those improvements."

Structural ambidexterity (Tushman & O'Reilly, 1996): separate units for exploration and exploitation, integrated through senior leadership. Contextual ambidexterity (Gibson & Birkinshaw, 2004): individual employees manage the tension within their own work, enabled by organizational context (discipline + stretch + support + trust). Sequential ambidexterity (Brown & Eisenhardt, 1997): temporal cycling between exploration and exploitation periods. O'Reilly & Tushman (2013) confirmed: structurally ambidextrous organizations outperform those attempting to resolve the tension within a single unit. Gupta et al. (2006) showed: within a single domain the tension IS zero-sum; across domains or over time, simultaneous pursuit is possible.

Cohen & Levinthal (1990): absorptive capacity is "largely a function of the firm's level of prior related knowledge." Path dependence creates lockout: "once a firm ceases investing in its absorptive capacity in a quickly moving field, it may never assimilate and exploit new information in that field." Two mechanisms: expectation formation (cannot see what you have no framework for seeing) and cumulative difficulty (harder to catch up the longer you wait). R&D serves a dual function: generating innovation directly AND developing absorptive capacity to exploit external knowledge. Zahra & George (2002) split this into potential absorptive capacity (acquisition + assimilation) and realized absorptive capacity (transformation + exploitation), with the efficiency factor as the ratio of realized to potential.

Teece et al. (1997) defined dynamic capabilities as "the firm's ability to integrate, build, and reconfigure internal and external competences to address rapidly changing environments." Teece (2007) refined into sensing, seizing, and transforming. Zollo & Winter (2002) explained how dynamic capabilities are built: experience accumulation (tacit, semi-automatic), knowledge articulation (making implicit explicit through collective discussion), and knowledge codification (encoding in procedures and tools). The counterintuitive finding: codification is most valuable for rare, causally ambiguous tasks where experience alone cannot build reliable routines — not for frequent routine tasks where managers typically invest in documentation.

Cannon & Edmondson (2005): cognitive barriers (causal ambiguity, attribution errors), emotional barriers (failure as identity threat, positive illusions), political barriers (status hierarchies, career protection). The structural innovation: "decoupling the discovery of failure from its evaluation and consequences." Organizations that learn from failure have infrastructure where separate roles identify problems (detection) and later determine accountability (judgment), with explicit norms protecting early reporters. "A natural consequence of punishing failures is that employees learn not to identify them, let alone analyze them." Sitkin (1992): small failures are more valuable for learning than large successes because they stimulate search without triggering crisis responses.

Intuiting stage: superstitious learning (false causal attribution from small samples; Levitt & March 1988) and competency trap (success blinds to alternatives; March 1991). Interpreting stage: defensive routines (undiscussable topics prevent shared interpretation; Argyris 1990) and groupthink (premature consensus suppresses dissent; Janis 1982). Integrating stage: political resistance (power structures block integration; Lawrence et al. 2005). Institutionalizing stage: premature institutionalization (embedding before sufficient evidence; Crossan 1999) and institutional inertia (embedded patterns resist evidence; Hedberg 1981). Cross-level: learning myopias ×3 — temporal, spatial, and failure biases (Levinthal & March 1993). Transfer: individual→org failure — learning not captured or shared (Kim 1993).

Hedberg (1981): "Knowledge grows, and simultaneously it becomes obsolete as reality changes. Understanding involves both learning new knowledge and discarding obsolete and misleading knowledge." Tsang & Zahra (2008) distinguished routine unlearning (changing procedures) from belief unlearning (changing assumptions) — the second is vastly harder because beliefs are implicit, embedded in culture rather than documented. Becker (2010) identified three conditions for unlearning: awareness of obsolescence, willingness to abandon, and availability of replacement. Argote (2013) added that deliberate forgetting of obsolete knowledge is necessary for adaptation. This report infers that distinguishing deliberate from accidental forgetting requires lineage — knowing why a pattern existed and whether the original justification still holds.

Gavetti & Levinthal (2000): forward-looking search uses mental models to evaluate alternatives without trying them — more efficient but prone to model errors. Backward-looking search learns from actual outcomes — more reliable but slow and costly. The dual-process finding is structurally parallel to LeCun's System 1/2 (S8): reactive/compiled (backward-looking experience) vs. deliberative/model-based (forward-looking projection). Starbuck (2017) added a meta-critique: organizational learning research itself suffers from the pathologies it studies — temporal myopia, success bias, retrospective accounts subject to defensive routines.