The same is true of S and N. Sure, S and N don't need to be bimodal if you don't define them that way. You can define your system to be any way you want. But that doesn't mean it is valid.
If, on the other hand, S and N were distributed bimodally, you would have some kind of evidence that S and N are labeling the effects of some kind or cluster of real physiological differences, rather than just being arbitrary logical constructs that humans are using to categorize.
There are issues where reasonable people can disagree and issues where there's a right and wrong, and this is one of the latter issues. Whether a personality dimension is capable of demonstrating psychometrically respectable "validity" and whether it's bimodal are independent issues.
As further discussed in the "Big Five is science and the MBTI is astrology" and "Real psychologists reject the MBTI" sections of
this post, the dichotomy side of the MBTI can now point to decades of respectable data support, and McCrae and Costa (the leading Big Five psychologists) long ago acknowledged that the MBTI, besides effectively tapping into four of the Big Five factors, also passed muster in the reliability and validity departments.
And the Big Five dimensions also have decades of respectable support for their validity.
And that's without either the Big Five or MBTI dimensions demonstrating a bimodal distribution.
"Validity" basically has to do with whether the theoretical constructs are found to significantly correlate with real-world stuff that goes beyond the specific things that the applicable test asked the subjects about. To the extent that a personality dimension demonstrates that, it has validity, regardless of what the shape of the distribution curve (in terms of weak and strong preferences) may turn out to be.
In the spoiler are membership stats for TC and Personality Café. For each type, the first percentage is the percentage of that type at the forum, the second percentage (in parentheses) is the estimated "general population" percentage from the official MBTI folks (from
this page), and the final number on the right is the self-selection ratio for that type — i.e., the ratio of the forum percentage to the general population percentage.
Looking at the PerC stats (the larger sample): 62% of the members are INs (as compared to 11% of the general population), and 83% of the members are N's (as compared to 27% of the general population).
Every S type has a self-selection ratio of 0.6 or lower, and no N type has a self-selection ratio below 1.0. And the
lowest self-selection ratio for the IN types is
13 times higher than the
highest self-selection ratio for the ES types.
The stats suggest than an average MBTI IN is something like
40 times mores likely than an average MBTI ES to join a personality-related internet forum.
And the stats for the second forum are strikingly similar to the ones for the first.
Those typings aren't "official," but that's a non-issue for purposes of this discussion (and also really a non-issue given the
magnitude of the N and I skews).
Assume for the sake of discussion that the relevant typings were all "official MBTI" typings. In that case, those would unquestionably be a strongly supportive set of results from the standpoint of the validity of the MBTI E/I and S/N dichotomies. The MBTI doesn't ask people if they use internet forums (or the internet), and it doesn't ask them if they're interested in personality (or psychology). But the type frequency pattern at both forums (involving relatively large samples) is almost
perfectly in line with a type-related explanation that says that (1) an N preference has a very large impact on the likelihood that someone will participate in personality-related internet forums, and (2) introversion also has a substantial impact (but not as large as an N preference).
And as far as that validity support is concerned, the
distribution curve of mild and strong preferences is irrelevant.