There is one study large enough to settle the argument, and it is not about the MBTI — it is about personality in general, which is where the data live. Dyrenforth, Kashy, Donnellan and Lucas (2010) used nationally representative samples of married couples in Australia, the United Kingdom and Germany — 23,250 people, around eleven thousand six hundred couples — to separate three things that almost always arrive tangled: the effect of a person’s own traits, the effect of the partner’s traits, and the effect of the similarity between the two. The third is what pages like this one sell. With the first two accounted for, it explained less than 0.5% of the variance — in relationship satisfaction and in life satisfaction too. Less than half of one per cent. What predicted satisfaction was each person’s own traits: a person’s own accounted for around 6% of the variance in relationship satisfaction; the partner’s, for 1% to 3% — and there the heaviest were agreeableness, conscientiousness and emotional stability.
Notice what that does to the recognition axis in particular. It is, by construction, a measure of similarity — and similarity is precisely the variable that barely showed up. The complement axis fares no better: nobody has measured that either, and the notion that opposites complete each other has, in the literature, even less support than the notion that likes understand each other. Both axes on this page are in the same boat, and the boat is a theoretical construction.
The nuance comes from Montoya, Horton and Kirchner (2008), a meta-analysis of similarity and attraction. It does attract: the effect is robust when what is measured is initial attraction, especially in designs where one person rates another they have not really met. In existing relationships, the effect of actual similarity stops showing up, and what goes on predicting there is perceived similarity. Which is to say: similarity predicts well who you will find interesting in the first conversation, and badly how the two of you will be doing in year three. The table below, and every compatibility list ever handed to you as an INTJ, measures the first thing and sells it as the second.
Translated into your life, without hedging: you do not need to find an ENFP. There is no right person waiting with the correct four letters, and you owe nobody an explanation for having fallen for an ISFJ. The useful question was never which letters — it is whether both people are mature enough, and maturity here has a practical, uncomfortable definition: knowing what you feel before anyone asks, saying it before it turns into filed resentment, and changing the plan when the data change. None of those three is a function of a four-letter code. All three are a function of work done — which is, point for point, what the Journey page is about.
And to close the circle honestly: almost none of this tested type. What exists is little, small and unreplicated, and nothing in it supports a ranking by four-letter code. Anyone trying to do better would hit an earlier problem first: readministered a few weeks apart, the instrument returns at least one different letter for close to half of people in the classic studies, and type dynamics — the very basis both of this page’s rules are built on — has never assembled consistent evidence. The Myers & Briggs Foundation is explicit: the instrument was not designed to select people and does not measure ability or competence. A compatibility page is precisely that, selection and prediction. Which is why it appears here with the rules printed over the top of it, rather than as a lone number on a screen.