Your Journal Needs Counterexamples

Your journal needs counterexamples because a pattern can look universal when you mainly record the moments that confirm it. “I always freeze in meetings” becomes easy to prove if you open the journal only after difficult meetings and never record the ordinary ones.

An exception log tests the boundary without forcing positivity. Where the pattern stops is often the clue that makes it useful.

A journal can become a confirmation machine

Confirmation bias can affect which evidence people seek, how they weigh it, and what they remember. A 2024 study found evidence for shared components across several confirmation-bias tasks, although the research literature also warns against treating every search for confirming evidence as irrational.

Klayman and Ha’s classic paper on confirmation, disconfirmation, and information showed that a positive test strategy can be useful under some conditions. The problem appears when your test has almost no chance of producing a disconfirming result.

“Show me every entry where work drained me” will find supporting material. It cannot tell you how often work did not drain you, or what was different on those days.

Turn the pattern into a claim that can lose

Vague pattern:

Meetings destroy my focus.

Testable version:

Meetings longer than 45 minutes leave me unable to resume detailed work for the next hour.

Now you can record:

• meeting length
• meeting type
• whether you had to present
• time until you resumed focused work
• sleep, food, and prior workload

The claim can survive, narrow, or fail. Any of those outcomes teaches you more than another entry saying meetings are exhausting.

This is the same discipline behind recording a prediction before the outcome. Decide what you expect before you know which details will be convenient later.

Keep an exception log beside the pattern

When the expected pattern does not happen, record four things:

1. What should have happened? State the expected pattern.
2. What happened instead? Describe the observable difference.
3. What condition changed? Look for timing, people, stakes, preparation, sleep, or environment.
4. Can I repeat it? Identify the smallest condition worth testing again.

Example:

I expected the client call to ruin my afternoon. It did not. The agenda arrived yesterday, I was not presenting, and I took five minutes afterward to write the next step. Test the agenda and transition separately next time.

The draining calls still count, and the entry gives you three candidate variables.

One exception changes the boundary, not the verdict

Avoid the opposite error. One good meeting rarely disproves a broader pattern. A counterexample has several possible meanings:

Exception	Possible interpretation
Random one-off	The pattern may still be strong
Repeats under one condition	You may have found a boundary
Appears more often than confirming cases	The original pattern may be overstated
Follows a deliberate change	You may have found a useful intervention

The denominator matters. “Five terrible calls” means something different if you had six calls than if you had sixty.

Record relief because pain has better timing

Pain announces itself. Relief often passes without asking for a journal entry. That asymmetry leaves the archive rich in breakdowns and poor in recoveries.

This is why recording relief, not just pain improves the evidence. You are not balancing bad with good for emotional symmetry. You are capturing the conditions under which the expected cost did not appear.

Current mood can also change which examples come to mind, so use the method in A Bad Mood Can Rewrite Your Whole Week when reviewing a longer period.

Ask the journal a harder question

When a belief feels proven, search for the line that would be hardest for it to explain:

• When was this less true?
• Who is the exception?
• What happened right before the pattern failed?
• What evidence would make me narrow the claim?

A repeated thought is not evidence, and a folder of confirming examples is not a complete test. Give the pattern a real chance to be wrong. The surviving version will be far more useful.