In some cases, I am uncertain which is the best version. This explains some apparent duplications.
The first two are on the philosophical motivation for NF. The last two are general surveys of alternative set theories.
it does have a connection to NF3, of course, but it is not really an NF paper.
This was the original program of my thesis, and led to the Watson program.
This was an isolated foray into mainstream mathematics. The paper is based on my Masters thesis. My advisor admitted later that it was really a PhD thesis, but it would have been quite disastrous for me to get a PhD thesis in functional analysis...without knowing any functional analysis!
This has been my main line of work. Substantial time spent thinking about the model theory of NFU is manifest in the strong axioms paper, with a footnote in the paper about models preserving information. The tangled type theory paper is the start of my main line of attack on the consistency problem for NF itself. The other papers touch on other consistent fragments of NF or important techniques in NF. The acyclic comprehension paper is about a very peculiar proposal about the axiomatics of NFU...which turns out to work perfectly. The Tarski's Theorem and NFU paper is an oddment.
Also this: preserves3 seems to be the submitted version; I have no idea what preserves4 actually is in versioning terms.
This paper also appears under the general NF set theory heading, but it says something about the type theory of PM.
This also appeared under stratified lambda calculus above. The untyped lambda calculus paper which appears there also mentions Watson but seems less a Watson paper than this one.