Self-image and how others see a leader: reading self–other gaps in a 360°

What a self–other gap is – and what it is not

Every 360° report contains the same quiet provocation: a number the leader gave themselves next to a number the people around them gave. Where the two diverge, you have a self–other gap – the most discussed element of any debrief and the most frequently misread, because it gets treated as a scoreboard. It is not one. The raters saw behaviour in the situations they share with the leader; the leader knows their own intentions and constraints. Both views are partial, and the gap is where they stop overlapping.

DefinitionSelf–other agreement (SOA)

The degree to which a leader’s self-ratings match the ratings given by others – supervisors, peers and direct reports – on the same dimensions. Leadership research treats the construct as an indicator of self-awareness; in a report it appears as a gap per dimension and per rater group. An early, foundational study by Atwater and Yammarino (1992) treated a leader’s self-awareness as exactly this agreement, and a 2010 review by Fleenor and colleagues in The Leadership Quarterly traces the field that followed.

Two findings from that research belong in every debrief. First, interest in self–other agreement stems from its assumed links to self-awareness and leader effectiveness – it is a meaningful signal, not a curiosity. Second, the field has measured agreement inconsistently and produced discrepant findings, which is why no headline claim that a fixed percentage of leaders overrate themselves survives the review. What survives is a pattern worth reading carefully, one leader at a time.

Four agreement categories sort every gap

The most useful frame groups leaders by the combination of direction and level. The four categories go back to Atwater and Yammarino’s 1997 self–other agreement model, which Fleenor and colleagues identify as the seminal paper of the field. Applied to one dimension of one report:

Why over-estimation deserves the most attention

Over-estimation is the category that decides whether a 360° changes anything. An under-estimator receives good news and has to learn to trust it. A leader in agreement on a weakness already knows and has, consciously or not, decided what to do about it. An over-estimator has neither an internal signal to change nor, at first, a reason to believe the raters: the feedback contradicts a self-image they are invested in, and the reflex is to discount the source – wrong raters, wrong moment, wrong reading of the item.

The gap persists precisely because nothing inside the leader argues against it. That is why an over-estimation on a dimension that matters to the role is the first thing a good consultant looks for, and the last thing to smooth over in the conversation.

A gap on capability is not a gap on behaviour

In a behaviour-only 360°, every gap answers the same question: how often does this person do X, in my eyes versus theirs? LEADBeyond 360° asks two questions per dimension, and the gaps mean different things.

The Capability question asks what the leader could do if they chose to. In the self-view that is self-efficacy in Bandura’s sense – belief in one’s own capacity – while raters can only infer it from what they observe. That is why LEADBeyond 360° compares self-efficacy not with others’ capability rating but with observed behaviour – the calibration gap: over- or under-confidence, computed only across sources, with differences within 0.75 points on the six-point scale treated as noise. How the calibration gap between capability and behaviour works has the detail.

The Behavior question asks what the leader actually does, and both sides rate the same observable behaviour. A behaviour gap is a perception gap about frequency, and the productive hypotheses are about visibility and situation: does the leader do it, but not where these raters can see it? Does it happen in calm phases and disappear under pressure – the pattern that leadership under pressure describes? Is the leader counting intention as behaviour? Each leads to a different conversation, and a report that shows only behaviour cannot tell you which one you are in.

Direction before size – and one rater group at a time

Direction

Read the sign first. Over- and under-estimation are different situations with different debriefs. A profile in which the leader sits consistently above the raters says something about self-image before you have looked at a single dimension; a profile whose direction flips from dimension to dimension is usually more differentiated and more honest.

Size

A gap is a difference between two averages, and difference scores compound the measurement error of both; on a six-point scale, a few tenths of a point are not a finding. LEADBeyond 360° sorts a dimension as a blind spot or a hidden strength only when self and others differ by more than half a point, and treats capability calibration gaps within ±0.75 as noise. Be more cautious still with small groups: Nowack and Mashihi (2012) note that a rater group with two or fewer respondents may be too small for reliable measurement – see how many raters a 360° needs.

Rater group

Peers see one leader; direct reports see another. Supervisors mostly observe strategy, results and client work; peers observe collaboration and communication; direct reports observe coaching, delegation and how the team is led. A gap that appears only with direct reports on delegation is a different hypothesis from a gap with everybody – it says something about what that group receives from the leader, not about the leader in general. Groups are shown separately only once they reach the anonymity threshold (two completed responses by default); below that, their answers are folded into “all others”, so a group-specific gap is visible only when the group is large enough to protect.

How to talk about a gap in the debrief

Feedback is not automatically helpful. In Kluger and DeNisi’s 1996 meta-analysis of feedback interventions, performance improved on average, but more than a third of the interventions made it worse – and their explanation is that effectiveness declines as attention shifts from the task to the self. For a self–other gap that is a precise instruction: keep the conversation on situations and behaviour, never on who the leader is, and never on who is right. Questions that do exactly that:

1. In which situations would these raters have seen this – and which situations do they never see?
2. If the raters are right, what would they have observed that led them to this rating?
3. Is this something I cannot do, or something I do not do when it counts?
4. Which group sees the gap, and what does that group get from me that the others do not?
5. What changes about this when I am under pressure?
6. Which of my own beliefs about leadership would have to be true for my self-rating to be accurate?

Three things to avoid: do not try to work out who said what – it ends the conversation and breaks the trust the programme depends on; do not average a group difference away into a single “others” score when the difference is the finding; and do not let a gap on one dimension become a statement about personality. A gap is local, situational and testable.

How LEADBeyond 360° visualises self–other gaps

The “Self vs. Others – Blind Spots” view shows a horizontal bar per dimension with the self-score and the others’ score side by side, across all 19 dimensions – the 12 progressive leadership dimensions and the 7 limiting tendencies, in the same higher-is-better direction throughout, so a blind spot on a limiting tendency reads the same way: the leader sees less of the pattern than the people around them do. Filters sort the dimensions into blind spots, hidden strengths and aligned, by largest difference or by pillar, and rater groups appear separately wherever they meet the anonymity threshold.

On top of that sits the pattern-hypothesis overlay. Where a limiting tendency scores clearly high, the report names the progressive dimensions the model links to that tendency and states the expected direction of the gap – explicitly labelled as a pattern hypothesis.

Where the actual gap agrees, that is an anchor for the conversation; where it does not, the report marks it as atypical – a cue to look more closely, not a contradiction to explain away. The mapping is part of the model, not an empirically derived prediction, which is why the label matters. The calibration view and the Beliefs & Behaviour view sit next to it; what a 360° report should contain walks through all twelve views.