There’s a lot of advice on this and almost all of it collapses into one instruction: assume it’s you.
Which is unhelpful in both directions.
If it genuinely is you, “assume it’s you” doesn’t tell you which part, so you’ll do a vague bout of self-criticism and change nothing. And if it genuinely isn’t — if you’re dealing with someone evasive, or hostile, or simply not very good — then assuming it’s you means spending two years working on the wrong thing while being managed badly.
The question deserves an actual method. Here are five tests, in the order they’re worth running.
The Method
Attributing a repeated communication failure requires distinguishing between a general pattern in the speaker, a mismatch specific to a pairing, and a problem in the other party. Five checks separate them. Distribution: whether the difficulty occurs with most people or with one, since a general pattern indicates a general cause. Symmetry: whether the other person experiences comparable difficulty with others, which can usually be established informally. Adjustment: whether the speaker has attempted delivery in the other party’s register, since without that attempt no evidence exists either way. Category: whether the difficulty concerns how meaning travels or what is being done, as evasion, dishonesty and hostility are not communication problems. And response: whether a genuine adjustment produces any measurable change across several attempts. The usual outcome is shared responsibility in unequal proportions, and the practically relevant portion is the speaker’s, since it is the only part available to act on.
Five tests. The first two are cheap.
Test One: Distribution
Does this happen with most people, or with one?
The single most informative question, and it takes about a minute.
Count. Not impressions — actual instances. How many people do you have this specific difficulty with?
One or two out of many points at a pairing problem or at them. It’s difficult to hold a general deficiency responsible for a failure that occurs in one relationship out of thirty.
Most people, most of the time points at you. A pattern that travels across everyone you deal with has one common element, and it isn’t the thirty different people.
The trap here is that everyone answers “just this one person” on instinct, because the one is salient and the others aren’t. Do the count properly, and count the people who quietly find you hard work as well as the ones who’ve said so.
Test Two: Symmetry
Do they have this difficulty with other people too?
Also cheap, and it requires one conversation.
Not “is Mark difficult” — that’s asking someone to gossip and you’ll get a diplomatic answer. Something narrower: “When you’re getting something signed off by Mark, what works?”
That question is legitimate, useful, and it produces the information. If the answer is a knowing look and a technique, you’ve learned that the difficulty is general to him. If the answer is “he’s fine, just send it over,” you’ve learned that it’s specific to you and him.
Either result is worth having. The second one isn’t a verdict on you, incidentally — a pairing mismatch is still a two-person phenomenon.
Test Three: Adjustment
Have you tried leading with their register rather than yours?
Here’s where most people fail the test without realising it.
If you’ve delivered the same way every time — leading with evidence, or with outcomes, or with rapport, whichever is yours — then you don’t have any evidence about whether they’re difficult. You have evidence that one particular delivery doesn’t work on them, which is a much weaker claim.
You can’t adjudicate this until you’ve made an actual attempt in their mode. Listen for their first substantive question: how do we know that is evidence, what does this get us is outcome, who else is involved is relationship, what exactly do you mean is precision. Then lead with theirs for ninety seconds.
That’s the room where half your words disappear, and until you’ve done it once, this whole question is unanswerable.
Running this test costs one conversation and it settles more than the other four combined.
Test Four: Category
Is the difficulty about how meaning travels, or about what’s being done?
This one matters because it prevents a specific error in the opposite direction.
Register mismatch, over-explanation, and mode incompatibility are communication problems, and they respond to communication fixes.
Evasion is not. Dishonesty is not. Someone who agrees in meetings and undermines afterwards, who withholds information deliberately, who is contemptuous, or who says one thing to you and another to your manager — none of that is a delivery issue, and no amount of adjusting your register will touch it.
The check: would this be fine if we simply understood each other better?
If yes, it’s communication. If the behaviour would still be a problem with perfect mutual comprehension, it isn’t, and treating it as one means you’ll spend a year adjusting your delivery to someone acting in bad faith.
That’s the failure mode of “assume it’s you,” and it’s expensive.
Test Five: Response
When you adjust properly, does anything move?
The last test and the slowest, because it needs several attempts.
Make a genuine adjustment — real, sustained, in their register — and observe over three or four exchanges.
Something moves, even slightly: it was at least partly yours, and you now have a lever.
Nothing moves at all, across several genuine attempts: that’s real information. Not proof it’s them, but the hypothesis that a delivery change would fix it has now been tested and failed.
The word doing the work here is genuine. One half-hearted attempt that you were expecting to fail doesn’t count, and people run that version constantly so they can conclude it’s the other person.
The Honest Answer
Run all five and the result is almost never clean.
The usual finding is both, in unequal proportions. There’s a real mismatch, they’re genuinely harder work than average, and you’ve been leading with your own mode for two years and calling it their problem.
Which is unsatisfying if you wanted a verdict, and useful if you wanted a lever, because of one asymmetry:
You can only act on your half.
That’s not a moral claim and it isn’t an instruction to take the blame. It’s arithmetic. Their portion may be larger than yours and it’s still not available to you. Yours may be thirty per cent and it’s the only thirty per cent you can move.
So the practical question isn’t “whose fault is it.” It’s: what proportion is mine, and what happens if I move it?
Frequently a thirty per cent adjustment on your side produces a disproportionate change on theirs, because you’ve stopped supplying the input their difficult behaviour was responding to. And frequently it doesn’t, and then you know.
When It Is Them
Worth addressing, because most articles on this stop at the self-examination and leave people stranded.
If the tests point outward — the difficulty is specific to them, they’re like this with others, you’ve adjusted genuinely and nothing moved, and the behaviour would persist under perfect comprehension — then you have a different problem and different options.
Reduce the surface. Fewer interactions, more of them in writing, more with a third party present. Not avoidance for its own sake — reduction of a channel that reliably produces nothing.
Change the channel. Some people who are impossible live are perfectly reasonable in writing, and vice versa. This costs nothing to test.
Escalate or route around. Legitimate, and worth doing early rather than after two years of trying to fix yourself.
Accept the cost and decide it’s worth paying. Sometimes it is. Naming it as a cost you’ve chosen is considerably better than absorbing it while feeling inadequate.
What This Is Not
It is not a method for proving you’re blameless. Anyone who runs these tests with that intention will find a way to pass all five, because they’re honest tests rather than mechanical ones. If you’re hoping for a particular answer before you start, you’ll get it.
It is not a claim that most difficulties are the other person’s. In my experience most are shared and a decent proportion are predominantly the speaker’s, usually through mode rigidity. The tests exist to prevent both errors, not to license one.
It is not applicable to power imbalances. If someone has authority over you and behaves badly, the question of attribution matters much less than the question of what you can do about it. That’s a different problem and these tests aren’t the right instrument.
The Cost
Running the tests honestly usually finds more of your half than you expected.
Most people arrive at this question already convinced. They want a method that will confirm what they’ve concluded, and the method doesn’t cooperate — particularly test one, which involves counting, and test three, which involves discovering you’ve never actually tried.
The specific discomfort is test three. Finding out that you’ve spent two years describing someone as difficult without once delivering in a form they can receive is not a pleasant discovery, and it’s the most common result.
Second cost: the tests take a conversation and several exchanges, which means you can’t have the answer today. People want adjudication immediately, and the honest version of this takes a fortnight.
Common Questions
Is it my communication or is it them?
Usually both in unequal proportions. Five checks separate them: how many people you have this difficulty with, whether they have it with others, whether you’ve genuinely tried their register, whether the problem is delivery or conduct, and whether a real adjustment moves anything.
How do you tell if you’re the problem?
Distribution is the fastest indicator. If the difficulty occurs with most people you deal with, there’s one common element and it isn’t the thirty different people. If it occurs with one or two, that points at a pairing or at them.
What if I’ve never tried adjusting?
Then the question is currently unanswerable. Delivering the same way every time gives you evidence that one approach doesn’t work on them, not evidence that they’re difficult. Test three settles more than the other four combined.
How do I tell a communication problem from bad behaviour?
Ask whether the difficulty would disappear if you simply understood each other perfectly. If it would, it’s communication. If the behaviour would still be a problem, it isn’t — and evasion, dishonesty and contempt aren’t delivery issues.
What if it really is them?
Then reduce the surface, move to a channel where they function better, escalate or route around, or explicitly decide the cost is worth paying. Naming it as a chosen cost beats absorbing it while feeling inadequate.
Why does my half matter if their half is bigger?
Because it’s the only part available to act on. That’s arithmetic rather than fairness. A modest change on your side also frequently produces a disproportionate change on theirs, since you stop supplying the input their behaviour was responding to.
Where to Start
Two tests, this week. They’re the cheap ones and between them they usually resolve it.
Test one, tonight. Write down how many people you have this specific difficulty with, and how many you don’t. Actual names, actual count. Include the ones who’ve never said anything.
Test three, at the next opportunity. Listen for their first substantive question, identify their register, and lead with it for ninety seconds. Change nothing else about your material.
That’s it. Ten minutes of counting and one adjusted conversation.
Most people find that test one is more even-handed than they expected — the difficulty is genuinely specific to a couple of people — and that test three produces a visible change on the first attempt, which resolves the question in an unwelcome direction.
Both of those are good outcomes. What you’re getting rid of is the version where you’ve been running an unresolved attribution question for two years, which costs you something every time you deal with them and never produces an answer.
I’ve spent twenty-five years working on how ordinary people develop capabilities that look impossible from the outside, and one of the more useful is the ability to hold “this is partly mine” and “this is partly theirs” at the same time without needing to settle which is larger.
Most people can only manage one of the two, and whichever one they’ve picked, it’s been running for years.