The observation
Amy Edmondson’s research on psychological safety and learning behaviour, based on a field study of 51 manufacturing work teams published in Administrative Science Quarterly, found that psychological safety predicted learning behaviour within a team even after controlling for team efficacy — and that learning behaviour, in turn, was what drove team performance. The finding carries an implication that extends well beyond the specific teams studied: performance is substantially produced through an ongoing process of noticing problems, raising them, and adjusting — not primarily through a fixed set of individual traits or a single evaluative snapshot taken at year-end.
This has a direct bearing on why annual or even quarterly performance reviews so often feel disconnected from the actual texture of daily work. A review is, structurally, a retrospective summary compiled well after the fact, usually from whatever documentation happened to survive the year — a handful of formally logged achievements, a manager’s general impression, sometimes a self-assessment written under time pressure close to the deadline. The learning behaviour that Edmondson’s research identifies as the actual driver of performance — the small, frequent, often informal moments of noticing something and adjusting — mostly does not get documented as it happens, because documenting it was never anyone’s job in the moment. By the time the formal review arrives, most of what actually produced the year’s results has already been forgotten, smoothed over, or folded into a single generic description that captures none of its texture.
Why it happens
The structural mismatch has two separate sources, and they compound each other. The first is timing: a review conducted annually or quarterly is separated from the events it is meant to evaluate by weeks or months, and human memory for specific, undocumented events degrades quickly and unevenly over that kind of interval. What gets remembered by the time of the review is not necessarily what mattered most; it is what was most recent, most dramatic, or most easily narrated — which introduces a systematic bias entirely independent of anyone’s honesty or good intentions.
The second source is that the moments Edmondson’s research identifies as most predictive of performance — raising a concern, testing an idea, correcting course after noticing something was off — are, almost by definition, informal and low-visibility events. They do not generate a ticket, a report, or a line item anyone was asked to log. A single high-visibility achievement — a project delivered, a target hit — is easy to recall and easy to write into a review months later. The steady accumulation of smaller adjustments that made the achievement possible in the first place is not, and it is precisely this accumulation that the research identifies as the real engine of performance.
What leaders and employees alike often conclude, when a review feels disconnected from the work, is that the review process itself is poorly designed or poorly administered — too infrequent, too bureaucratic, too reliant on a form that does not capture what actually matters. This is sometimes accurate as a design critique. The evidence points to something more fundamental sitting underneath the design problem: the review is being asked to summarise a continuous, largely informal process using a retrospective, largely formal instrument, and some amount of disconnect between the two is close to structurally guaranteed regardless of how well the form itself is written.
What it looks like in practice
Consider an organisation where an annual review covers twelve months of a person’s work. In preparing for the conversation, both the manager and the employee find themselves reaching mainly for a handful of visible, easily narrated moments — a project that shipped, a target that was hit, perhaps one difficult period that stands out because it was genuinely stressful at the time. What neither of them can easily reconstruct, months later, is the accumulation of smaller adjustments across the year that actually produced those visible results: a concern raised early that avoided a larger problem later, a process quietly adjusted after noticing something was not working, a piece of feedback given and absorbed in a five-minute conversation that nobody thought to record.
None of this is anyone’s failure in the moment. Nobody was asked to log it, and logging every small adjustment as it happened would itself be a significant and arguably disproportionate burden. The result, though, is a review that ends up summarising the visible tip of the year rather than the ongoing process underneath it — which is why both participants in the conversation often come away with a nagging sense that the review captured something true but somehow missed the actual substance of how the year unfolded.
This disconnect tends to be most acute in roles and teams where the real value of the work lies precisely in the kind of continuous, low-visibility adjustment that Edmondson’s research identifies — collaborative, judgement-heavy work where the outcome depends less on any single achievement and more on an accumulated pattern of noticing and correcting. Roles with clearer, more countable outputs are somewhat more resistant to the disconnect, simply because more of what mattered happens to leave a visible trace behind it.
For how this pattern presents in other kinds of organisation, see other sectors.
Questions worth considering
- What proportion of what actually shaped this person’s year would survive being written down only from memory, months after the fact?
- Are the small, informal moments of noticing and adjusting — the ones research links most closely to performance — visible anywhere before the formal review arrives, or only the handful of large, easily narrated events?
- Is the disconnect people feel about reviews a critique of the review’s design, or a more structural mismatch between a continuous process and a retrospective instrument?
This perspective is part of Performance.
Your team is facing similar challenges?
Subscribe to The Riddle Ledger
Perspectives are sent as they are published, with no promotional email, and you can unsubscribe at any time.