The Field Guide › Paper
A two-study empirical paper that gives one of the clearest accounts of what builds psychological safety, filling a common gap in a literature that studies what safety produces more often than what produces it. Carmeli and Gittell ask how organisations can support employees to learn from failures, and locate the answer in the quality of the relationships between members. They take Gittell's relational coordination as a specific manifestation of high-quality relationships, arguing that its three dimensions, shared goals, shared knowledge, and mutual respect, foster psychological safety, which in turn enables people to surface, discuss, and learn from failures rather than conceal them. The mechanism runs through the division of labour: where members hold role-specific goals, fragmented knowledge, and little mutual respect, they fear speaking out of role and default to blame rather than learning (the paper invokes Edmondson, Bohmer and Pisano's surgical teams, in which surgeons, nurses, and anaesthesiologists stay within role even when patient safety is at stake); relationships that transcend those role divisions reduce the fear and make failure safe to examine. Two studies test the mediation model. Study 1 (100 employees across three Israeli firms) finds partial mediation, high-quality relationships affecting learning both directly and through psychological safety; Study 2 (128 employees, measured at two time points a fortnight apart to blunt common-method bias) finds full mediation, the direct path falling to non-significance once psychological safety is entered, so that relationships support failure-learning entirely through the safety they create. Safety was measured with Edmondson's (1999) team scale re-worded to the organisational level, and learning from failures with items from Tucker and Edmondson (in this corpus) that capture second-order problem-solving: addressing a problem's underlying causes rather than merely fixing its immediate symptom. Two things make the paper especially useful here. It carefully distinguishes psychological safety from trust and from perceived organisational support, following Edmondson (2004): trust concerns one's beliefs about others across a long horizon, whereas psychological safety concerns beliefs about how others will treat the self over the short, specific horizon of an act such as admitting a mistake. And it converts the finding into a practical antecedent claim: to get learning from failures, establish psychological safety, but to do that, first build shared goals, shared knowledge, and mutual respect, through relational work practices (cross-functional performance metrics, relationally-attentive hiring, boundary-spanning roles, and leaders who model respect). It connects the relationship and trust literature in this corpus to the learning-from-failure strand, with psychological safety as the mediating mechanism.