Organizational Learning II

Mercury could name its innovation problem only in private

The company called Mercury is anonymous, and the account comes from the consultant-researchers who intervened there. That boundary matters. Within the case, however, the contradiction is unusually concrete. The chemical company had grown from one product line in the 1920s to more than $5 billion in sales by the mid-1960s. Executives feared that research was no longer producing businesses fast enough, so they created a New-Business Division that could develop and incubate technologies before asking an established division to adopt them. ¹

Ten years and about $20 million in research and development later, the new unit had produced no business of consequence. Yet senior management had not directly criticized it. The authors' detailed case account shows why the silence was not simply timidity. Mercury had also decentralized into semi-autonomous divisions. Their presidents—known inside the company as “barons”—were expected to protect a charter, deliver short-term earnings, and compete for resources. Successful corporate development often required those same presidents to surrender people, technology, or markets to a project whose return would appear elsewhere. The structure rewarded the defense of territory while corporate strategy demanded its recombination. ²

The 1996 Organizational Learning II expands and reorganizes the authors' 1978 Organizational Learning rather than merely repaginating it. It carries Mercury into an early chapter, then moves through defensive reasoning, a CIO case, classroom research, and a more comprehensive intervention. The composition is important: Chris Argyris and Donald Schön do not present a survey showing how often organizations learn. They build a theory of action from intervention cases and use that theory to design further intervention. ³

Reconstructing failures changed the object of inquiry

Mercury's participants initially offered competing diagnoses. Research lacked commercial skill; division charters were too restrictive; top management was too cautious. The consultant asked the vice president for technology, the research director, and New-Business Division staff to reconstruct actual successes and failures. That request encountered resistance. Failures had been buried, memories were distributed among people who had never compared them, and public disagreement risked blame. ⁴

When participants pieced the episodes together, familiar stories stopped looking self-evident. The belief that good ideas naturally rose to the top had been protected by a circular test: an idea was called good after it succeeded, while failures disappeared from the usable record. Repeated stories instead revealed projects that became visible as “crying babies” because they were in trouble, and speculative “fliers” that were abandoned when unfamiliar risks became clear. The most promising developments often recombined capabilities across existing divisions—precisely the work the divisional system made hardest. ⁵

This movement carries the book's famous distinctions better than the thermostat metaphor alone. Mercury could correct production errors while leaving profit targets and territorial authority intact; Argyris and Schön call that single-loop learning. Comparing success and failure made the governing contradiction itself discussable, opening the possibility of double-loop learning. The gap between Mercury's stated commitment to new businesses and the territorial behavior its incentives produced is the gap between espoused theory and theory-in-use. ⁶

The intervention did not establish that Mercury became a durably different organization. It generated a shared map and a more testable diagnosis. For the authors, organizational learning would require the result to enter collective practice and memory, not remain one manager's insight. This is where organizational ignorance becomes more than missing information: the company had arranged not to know together what many members partly knew. ⁷

The investigator is part of the learning system

The case also turns its method back on the expert. The consultant chose the initial participants, elicited the histories, grouped recurring stories, and helped build the model. Those actions made previously private knowledge inspectable. They also gave an outsider unusual authority to decide what counted as defensiveness, valid information, or productive inquiry. The account gives much more voice to executives and technical managers than to other employees, customers, or communities affected by Mercury's investments and products. ⁸

That asymmetry complicates the book's proposed movement from Model I—unilateral control, face-saving, and untested inference—toward Model II—valid information, informed choice, and public testing. Asking a subordinate to disclose an unstated judgment is not the same act as asking a division president. The first person may bear career risk while the second retains authority over the consequences. A learning intervention can reproduce unilateral control if the expert interprets resistance as pathology without testing whether secrecy is a rational response to power. ⁹

Jan Greenwood identified inconsistencies between Argyris and Schön's theory and recommended pedagogy before the second edition appeared. Her question applies recursively: does the intervention's practice embody the values it espouses? ¹⁰

A famous vocabulary proved easier to cite than to realize

The 1996 book arrived as “organizational learning” was becoming a broad field, and single- and double-loop learning traveled far beyond the authors' action science. Raanan Lipshitz's later review of that reception found Argyris and Schön widely invoked but rarely followed or directly tested. He argued that their most attractive aspirations were also difficult to realize, while practical lessons in their cases were often overlooked. ¹¹

One later response shifted attention from an ideal conversational model to institutionalized learning mechanisms. Micha Popper and Lipshitz's structural account distinguishes individual learning from organizational arrangements that collect, analyze, store, disseminate, and use information. That is an extension and a critique, not a smooth next step in a canon. It makes the durable test at Mercury sharper: after the consultant leaves, who must compare failures, where is the comparison recorded, which incentives can be revised, and who is protected when the evidence implicates someone powerful? ¹²

Mercury leaves two useful doors open. Its reconstructed project histories show how a coalition can test a corporate myth against episodes it had excluded. Its unfinished intervention shows why candor alone is not learning. The stronger claim requires evidence that inquiry changed authority, resources, routines, and the range of people entitled to challenge the organization—and that those changes survived the expert who named them. ¹³

The links to culture, informal organization, trust, and voice, The Fifth Discipline, The Fearless Organization, organizational intelligence, and benefit for all life are editorial reading paths through learning, candor, standing, and consequence. They do not assert direct influence or equivalent definitions of learning. ¹⁴