Threat Rigidity Effects in Organizational Behavior: A Multilevel Analysis

Staw, Sandelands & Dutton · Team Learning, Safety & Error, Power & Equity · Administrative Science Quarterly · 1981 · Paywalled

Open this neighbourhood on the live map →

The mechanism paper for why groups stop learning at the moment learning matters most. Staw, Sandelands and Dutton propose one effect operating at three levels: a threat to the vital interests of an individual, a group or an organisation produces a restriction in information processing, meaning a narrowed field of attention, simplified information codes and fewer channels in use, and a constriction of control, meaning influence concentrated in fewer hands and moved to higher levels of the hierarchy. Together these make behaviour less varied, pushing the entity toward its well-learned or dominant response. The argument is assembled by review rather than from new data, taking individual-level evidence largely from mid-century experimental work on stress, perception and problem solving, group-level evidence from research on cohesion and conformity, and organisation-level evidence from studies of crisis, disaster and fiscal retrenchment. For a map about psychological safety the relevant reading is at the group level, and it is uncomfortable. Where a threat is attributed to an external source and the group believes it can prevail, the paper predicts increased cohesiveness, support for existing leadership and pressure for uniformity, which express themselves as restriction of information and constriction of control: divergent solutions are ignored, deviant positions downplayed, dominant members prevail and influence centralises. In other words the conditions that suppress speaking up are not a failure of a team under threat, they are the ordinary response of a team under threat, and they arrive attached to things that feel like virtues at the time. There is also a specific and still-underused move on groupthink here: Janis attributed pressure for uniformity to cohesiveness, but every case Janis used featured an external threat, so Staw and colleagues suggest threat rather than cohesion may be the initiating variable, noting an experiment by Flowers in which raising cohesiveness did not degrade decision making as Janis predicted. The second half of the group model is the part usually forgotten. Where the threat is attributed instead to internal deficiencies, or where the group thinks it will not succeed, the prediction reverses: cohesiveness falls, leadership becomes unstable, dissension appears, and the effect is an input of new information and a loosening of control. Threat does not simply close a group down. What it does depends on where the threat is thought to come from and whether the group thinks it can win. The paper is also more careful about its own thesis than later citation has been. Rigidity is not presented as always maladaptive. Under incremental change, where the underlying causal relationships are unaltered, falling back on a previously successful response may be exactly right and can reduce the threat. It is under radical change, where prior learning no longer applies, that rigidity intensifies the threat and produces the maladaptive cycle. The authors put it as doing better at what one already knows being at best a mixed blessing, and describe threat rigidity as two-edged, amplifying both the survival and the extinction potential of an organisation. Anyone reaching for this paper as evidence that tightening up under pressure is simply wrong is using it against its own argument. Two limits the authors state themselves, and one a reader should add. To gather enough literature to span three levels they admitted studies whose notion of threat only overlaps their own, and the operationalisations of information processing and control are not equivalent across levels; they say in the introduction that this slippage makes their conclusions more speculative than they would like, and close by calling the article a starting point rather than a summary of work on the topic. They add that the dysfunctional cases may show most clearly in laboratories precisely because an experimenter can change the causal rules abruptly, which is rarer in the field. The individual-level evidence is also old, much of it from the 1940s and 1950s, and the cross-level parallel is best treated as a well-argued proposition rather than a demonstrated fact. Neither psychological safety nor Edmondson appears anywhere in the paper, which is chronology rather than oversight, and the bridge is that this supplies the mechanism the construct needs and rarely states: psychological safety is not a stable property of a team but a casualty of the conditions that make it most valuable, and any intervention assuming a team under threat will behave like a team at rest has assumed away the effect.

Staw, B.M., Sandelands, L.E. and Dutton, J.E. (1981) 'Threat rigidity effects in organizational behavior: a multilevel analysis', Administrative Science Quarterly, 26(4), pp. 501-524.

Neighbours (13)

View the source paper →