High-performing teams report more problems than weaker ones. Amy Edmondson found this studying 51 work teams in a manufacturing company, and it initially surprised her. The intuitive expectation runs in the opposite direction. Good teams work cleanly, catch issues early, and deliver without drama. What Edmondson found was that high-performing teams surfaced more issues, raised more uncertainties, and acknowledged more errors than their lower-performing counterparts. The distinguishing feature was not fewer problems; it was more visible ones.
The finding demands an explanation. Most managers encountering it assume the mechanism is cultural. Better leaders create safety, so people feel comfortable raising concerns. That reading is partly correct. But it is not the full mechanism, and the part it omits is what determines whether communication practices actually improve performance or simply produce a warmer working environment.
What safety explains
Psychological safety is the right construct for explaining why people surface problems at all. Teams where members can raise concerns without fear of embarrassment or retaliation do surface more concerns. Edmondson showed this directly. Psychological safety predicts learning behavior, and learning behavior mediates team performance. The channel runs from safety to surfacing to performance, with surfacing as the active ingredient.
The standard reading of this result treats it as evidence about organizational climate. Psychological safety is a cultural property of the team, and communication is what a healthy culture produces. Under this view, communication norms exist to keep people feeling heard and safe, to maintain the relational conditions under which honest exchange is possible. This reading is not wrong. It explains why problems surface. It does not explain why surfacing them drives performance.
The missing mechanism
Knowledge work is invisible by default. The output of a knowledge worker is not an artifact that can be examined directly. It is a reasoning process that produces a conclusion. A decision, a design choice, an architectural judgment are all conclusions that follow from reasoning, and the reasoning is not visible in the conclusion. When a team member works from a flawed premise or misunderstood constraint, the flaw does not appear in the output they produce; it appears only if the reasoning behind the output is exposed.
This is why surfacing problems drives performance. The mechanism does not run through morale or relationships. Psychological safety enables surfacing, which creates visibility, which creates the opportunity for correction before an error compounds. The performance benefit comes from the correction opportunity. Communication is what creates that opportunity by making reasoning visible before it ships.
Amazon's shift to written documents before meetings, which Bezos introduced in 2004, was built on the same premise. His argument was that prose exposes reasoning in a way slides do not. A claim that does not follow from the one before it is visible on the page; the same logical gap disappears inside a bullet point. The discipline of writing a coherent six-page document is not separable from the discipline of thinking coherently about the subject. Writing is not a communication act layered on top of thinking. It is a test of whether the thinking is actually there.
Google's Project Aristotle, studying over 180 internal teams, reached the same conclusion from a different direction. Psychological safety was the single strongest predictor of team performance, more significant than talent composition, role clarity, or technical skill. What those teams shared was not consensus or comfort. It was that members could surface uncertain ideas and challenge assumptions, which meant collective reasoning was inspectable rather than private. The performance advantage came from having more corrections available, not from having more harmonious relationships.
What changes
The epistemic framing changes what communication practices are actually for. Writing a brief before a task begins is not an administrative step for keeping managers informed. It is a test of whether the reasoning behind the task is coherent enough to be executed. The assignee who restates the objective in their own terms, describes their intended approach, and surfaces what they do not understand is putting their reasoning on the table where it can be checked. When that restatement reveals a mismatch, the misalignment is caught before it costs two weeks of work.
Pre-reads and demos shift function in the same way. Under the social framing, they are reporting mechanisms for keeping stakeholders informed and relationships functional. As epistemic infrastructure, they are peer review on live reasoning. A written document submitted before a decision meeting puts the thinking on the table where others can find the gap. A demo that shows progress and surfaces an open risk gives the team a chance to change course before the cost of doing so makes it impractical. The discipline is not in producing the document; it is in what happens when others read it.
Commitment after disagreement follows the same logic. A decision that has been debated, stated clearly, and examined by the people who will execute it can be executed without ongoing argument. Relitigating the decision during execution is a failure to distinguish the reasoning process from the execution process. Disagreement belongs in the review of the brief and the pre-read discussion. Once those have run, full execution is not deference; it is the natural consequence of having already subjected the reasoning to inspection.
When this holds
This mechanism is strongest when the work is judgment-intensive and the reasoning is genuinely uncertain. When the task involves designing systems without established precedents, evaluating approaches whose outcomes cannot be predicted, or deciding which problem to solve in the first place, the cost of invisible reasoning is high. A flawed premise that runs unchecked for two weeks is expensive to correct. The benefit of communication as peer review grows with the cost of late correction, which rises as the work becomes more novel and the criteria for success more open. Those are also the conditions under which individual reasoning is most likely to be wrong in ways the reasoner alone cannot catch.
The mechanism weakens when the work is repeatable and well-specified. A team processing high-volume, clearly defined tasks does not benefit from writing pre-reads before each one. The reasoning is established; execution is the variable. In those contexts, communication overhead designed for uncertain reasoning can reduce throughput without improving quality. The question is whether there is genuine uncertainty about what to do, how to do it, or why it matters. When that uncertainty is absent, the infrastructure built for it is more burden than benefit.
The real bar
The uncomfortable implication of the epistemic mechanism is that refusing to make reasoning visible is a quality problem, not a personality trait. The engineer who skips the pre-read, the manager who assigns without a brief, and the team that treats demos as administrative formalities are each producing work whose reasoning cannot be inspected before it matters. That is not a communication style preference. It is the same failure as delivering untested code. The reasoning is complete inside one person's head, and accessible nowhere else.
The standard is not to communicate more. It is to make the reasoning behind the work visible enough that a peer can find the gap. A contribution whose impact depends on others trusting the author's judgment, rather than being able to inspect the reasoning, is not finished regardless of how technically sound it appears. In knowledge work, finished means the thinking is accessible, not just complete. The reasoning is part of the output.