The Field Guide › Paper
A three-wave longitudinal field study, conducted across 55 clinical units (224 members) in a large hospital, that traces a chain from leader behaviour to unit performance and positions psychological safety as its pivot. The model is sequential: leader inclusiveness — leaders being open to input and making themselves accessible and available, the Nembhard and Edmondson (2006) conception — raises members' psychological safety; unit psychological safety climate enables the unit to learn from its failures (reflecting on what went wrong and addressing root causes rather than hiding or repeating them); and that learning in turn predicts subsequent unit performance, judged independently by senior managers two waves later. Measuring the constructs at separate times and from separate sources, the study finds that learning from failures fully mediates the link between psychological safety climate and later performance: once learning from failures is in the model, psychological safety no longer predicts performance directly, which is strong evidence for the long-standing claim that psychological safety is the substrate for learning rather than a performance driver in its own right. The paper's distinctive contribution is a moderation: leader inclusiveness matters most where units are already performing poorly. In low-performing units, whose members are more disoriented and more sensitive to their leaders' cues, inclusive behaviour does the most to lift psychological safety, narrowing the gap with high-performing units (whose members tend to feel safe regardless). The authors read this as a way to break a vicious cycle in which poor performance brings repercussions, repercussions erode safety, and low safety suppresses the voice and failure-learning that might have improved performance. A supplementary analysis adds that climate strength matters alongside climate level: psychological safety climate predicts performance most strongly when it is also shared, meaning there is low within-unit variance in members' perceptions. For this corpus the paper does two jobs at once: it thickens the thin leadership-antecedent strand by carrying Nembhard and Edmondson's inclusiveness construct through to a hard, independently-rated outcome, and it supplies a clean empirical instance of the psychological-safety-enables-learning mechanism. The design is longitudinal and multi-source but remains correlational, so the causal ordering is inferred from timing and measurement separation rather than established by intervention.