Maintenance
Part of MBTI explorers: three habits that read action as type
5 honest notes on MBTI explorers compatibility
MBTI explorers compatibility charts fill 136 pairings nobody tested: the arithmetic, the reliability ceiling, and the verdict inside the ratings.
What to take away
- The chart is a set of impressions arranged in a grid, and the grid is what makes it read as data.
- Sixteen codes give 136 unordered pairs. No chart that rates MBTI explorers compatibility reports how a single cell was determined.
- A rating cannot be steadier than the codes it is built from, and four-letter codes move on retest.
- Explorer types are the four SP codesISFP, ESFP, ISTP, and ESTP — and they get a warm short-term rating and a doubtful long-term one. That split is a moral judgment wearing a preference label.
- The usable version is two people, two self-typings, and one named difference. It predicts nothing.
The arithmetic behind a 136-cell chart
Sixteen codes produce 136 unordered pairs, or 256 if a chart treats each side of a pairing as distinct. Every cell carries a rating.
The charts in circulation are not the instrument itself. The MBTI is the Myers-Briggs Type Indicator, published by The Myers-Briggs Company, and the MBTI name is a trademark of the Myers & Briggs Foundation. The rated tables shared on sites like 16Personalities and Truity are third-party products built on the same sixteen-code idea.
For one cell to mean anything, its author needs a defined outcome, a way to measure it, a set of couples of that exact combination, and a comparison group drawn the same way.
Multiply by 136. That is the missing work. What happens instead is simpler. Someone reads two type descriptions, forms an impression, and writes a rating. The chart is a set of impressions arranged in a grid.
[figure 1]
Chart Appearance vs Reality
What the chart looks like
- Findings
- 136 findings
- Rating per pair
- A rating per pair
- Coverage
- Comprehensive coverage
What it is
- Findings
- 136 impressions
- Rating per pair
- Summary of two descriptions
- Coverage
- Written, not checked
| What the chart looks like | What it actually is |
|---|---|
| A rated cell | One person's impression of two descriptions |
| A tested pairing | No sample and no outcome measure |
| A finding about types | A statement about two labels |
The reliability ceiling
A rating built from two inputs inherits the wobble in both. A four-letter code is settled by four cuts through four sets of answers. A code sitting near a cut can come back different on a second sitting.
Retest data for the Myers-Briggs Type Indicator commonly report four-letter agreement in a rough range near half of respondents over a few weeks, with single-letter agreement higher. That is what repeatability asks of any measurement.
Put two such codes in a cell. If either one moves, the pair belongs in a different cell with a different rating. The arithmetic of four cuts works through the same problem from the scoring side.
So the rating reads as a fact about two people. It is a fact about two labels that might have been different labels a month ago.
What explorer codes get rated as
The four explorer codes — ISFP, ESFP, ISTP, and ESTP — are described in terms of responsiveness, comfort with improvisation, and attention to what is present. On a pairing chart that becomes a familiar verdict. Enjoyable, energetic, unreliable over time.
Nothing in the model turns a preference for acting on the current situation into a forecast about commitment. The step is editorial. It is ordinary moralizing about spontaneity, attached to a code so it reads as an assessment. It is the same habit behind celebrity type profiles and sun-sign readings: a loose impression hardened into a label.
The base rate problem finishes the job. A pairing that fits the verdict gets remembered and retold. Pairings that do not are unremarkable, so nobody counts them. The testable-claim standard treats the same gap from the other direction.
Four questions to ask a chart
- What outcome is being rated, and how would anyone measure it?
- How many couples of this exact pairing were observed, and against what comparison group?
- Would both codes come back the same on a second sitting?
- What is the rate of that outcome among all couples, with type set aside?
The first two questions stop most charts immediately.
[figure 2]
Four Questions Before Using a Rating
- What outcome does the rating predict?
- What would count against it?
- Where did the two codes come from?
- Would you accept the same reasoning elsewhere?
The sentinel compatibility charts fail on a related ground, since the pairs they rate never asked what the pairing was supposed to produce.
Example: naming a planning-horizon difference
Two people who both type as explorer codes come to the same decision from opposite directions. One wants it made now and adjusted later. The other wants it settled and left alone.
Called a preference, that difference is workable. Called a rating, it turns into a verdict about which of them is the problem. Said out loud, it is a sentence both can finish.
That naming is the whole of the useful version. It predicts nothing and rates nothing. General accounts of personality testing hold the same boundary: a self-report instrument describes how a person sees themselves, not how a relationship will run. The explorer overview draws that limit from the observer's side.
Common questions
Do any two codes genuinely clash?
The model names none. Charts do, and different charts name different ones using the same model.
Is there research on personality similarity in couples?
There is work on personality and relationships in general, using continuous trait measures such as the Big Five rather than four-letter codes. That literature does not produce a 136-cell grid.
My partner and I are rated poorly and we are fine. What does that mean?
That the rating was never a prediction. Nothing about the relationship needs explaining.
Could a good chart be built?
In principle, with defined outcomes and real comparison groups, for a few combinations. It would be a large piece of work and it would not look like the charts in circulation.
