The have yet to construction, exemplified by sentences such as John has yet to visit his
grandmother, is widespread across dialects of English. However, recent studies have revealed that behind this
apparent unity, there is significant variation in the syntactic properties of the have yet to construction.
Speakers vary with respect to (i) the status of have as an auxiliary or main verb, (ii) the status of negation
tests, and (iii) the status of a variety of related yet to constructions. The goal of this paper is to sort out
the microsyntax of have yet to across speakers, in the face of contradictory empirical claims and mutually
incompatible proposals in the existing literature. We develop an analysis based in part on two wide-scale surveys we have
conducted. With respect to have, we show that speakers who can treat it as a main verb can also treat it as an
auxiliary, but not necessarily vice-versa. We propose that the variation in this case has to do with where the perfect
features are introduced in the clause. With respect to negation, we find that speakers do not treat all the negation tests the
same, forcing us to contend with the question of how these tests work. We propose that for most speakers, only the embedded clause
is syntactically negative. Negation tests split according to whether they must target the matrix clause, or whether they can
target an embedded clause as well. In some cases, the tests reveal the same sentence to be both affirmative and negative, as we
expect: the matrix clause is syntactically affirmative, but the embedded clause, which hosts the lexical content, is syntactically
negative.