So I've been thinking about the 1ps pronouns in Nostratic, and I'm not
pleased. There appear to be a dizzying number of pronouns. Bomhard
reconstructs *mi, *ma, *ka, *Ha, *na (actually there are two of
these), and *?iya. He tries to explain these variants as suppletive
forms depending on number, exclusivity, and whether the verb is active
or stative.

It's a valiant effort, but I'm not convinced. First of all the *?iya
form is found only in Dravidian and Afroasiatic and so I doubt it goes
back to Nostratic. The first *na form is likewise attested in
Dravidian and Afroasiatic, but poorly elsewhere (e.g. In IE it appears
only in Tocharian). The second *na form doesn't fare much better but
it is better represented in IE (more on this later). I'm also not
convinced that this Nostratic in origin. I'll leave out discussion of
*ka and *Ha because I want to focus on the *mi and *ma forms (most
likely ablaut variations of a single pronomial element).

*ma and *mi are usually said to have support in Afroasiatic, but the
case is weak. It occurs only in the Chadic languages. In these there
is often an interchange of *m before *u and *n elsewhere. I would take
this as the starting point and assume that the *m form sometimes
spread analogically before other vowels. The Dravidian forms in *m
also tend to exist only in suffixed forms so I'm a little undecided on
the origin of first person *m in Dravidian. Elsewhere in Nostratic it
seems to be well represented.

There are some curiosities in IE that need explaining. First is the
aberrant form *h1eg'(h)o(m). The plural seems to based on *n and the
dual on *w. It seems to me that it might be possible to unify these
forms with the help of Altaic. Altaic has it's own aberration. The
nom. forms are in *b and the oblique forms are in *m. Altaic
(possibly) also has accusatives in *be whereas the rest of Nostratic
points to *m(a). At first I thought there might be some Altaic rule
about *m > *b, but these are the only two examples where this change
apparently occurs.

This leads me to the radical possibility that the Nostratic pronoun
was actually *p'. Bomhard only reconstructs three roots with *p'. One
is a nursery word and the other two are onomatopoeia. The reflexes of
*p' are thus questionable. Nostratic *p' should become *b in PIE (in
the traditional reconstruction). This is known to be very rare. I
recently read an article that mentions in passing that PIE *b may have
become *w. This would explain the 1ps forms in *w (they probably
spread to the 2nd person as the forms in *n began to dominate the 1st
person plural paradigm).

But we need to tweak this hypothesis because *b is rare, but not
unknown. This *b > *w change could not be unconditioned. Before we can
look for conditioning environments we must first ignore the bulk of *b
occurrences since they occur in onomatopoeia. The remaining examples
appear to fall almost entirely within two categories (*bak is an
exception). One, *b occurs contiguous with resonants (which is pretty
much where we would expect *w not to exist). Two, *b occurs in roots
with dentals, excluding *s. An exception to this latter rule is when
the dental is a (non-contiguous) nasal. I propose for this case that
nasal assimilation occurs and *b >*m, explaining the first person
forms in *m. I would actually project this rule all the way back to
Nostratic. For Altaic I would then propose that *p' > *b (not *p as
Bomhard proposes). This would then explain the *b/*m alternation in
the pronoun.

Back to PIE. So far we have the *m and *w forms explained (assuming
the *m form spread analogically through most of the singular
paradigm), but what about *n? This one actually derives from *m. The
nominal plural is *nsmes. This appears to either be a reduplication of
*mes with zero grading of the initial syllable or *mes with a suffix
of *mes with zero grade in the root (we see something similar with the
*w forms such as *usmes). Whatever the origin, the other forms in *n
are back formations that spread through most of the plural (and
sometimes dual). This still leaves *h1eg'(h)o(m) to be explained. I
suppose that this too is derived from *b. I take it to be a compound
form *h1e-be-m. The *h1e is perhaps an emphatic prefix. The *-m is
from the absolutive *-m (which later becomes the accusative). The *e >
*o under Rasmussen's Law. The *b dissimilates to *g (> *g') because of
the final labial nasal. A similar dissimilation happens in the dative
when *bHi > *g'Hi. The aspiration occurs only in Indic and is a result
of distant aspiration from *h1.

Uralic, Kartvelian, Eskimo-Aleut, and Chukchi-Kamchatkan aren't much
help here because they all retain *m. Presumably this is due to it
analogically spreading throughout the paradigm.

The accusative is questionable. IE and Uralic have *-m. Mongolian and
Tungus has *ma. But Turkic points to *-nVG. Indeed an accusative in
*-na is widespread. Starostin claims that Altaic had an accusative in
*be based on Tungus and Japanese. It's actually unclear what Tungus
had because there are forms in *b, *m, and *w depending on
phonological environments. Though it's tempting to trace all these
forms back to an accusative *-p'a I'll refrain from doing so until
more evidence is available about the reflexes of *p'.

Of course there's a lot of speculation going on here! I think there
needs to be more study about how *p' is reflected in all the
languages, but especially IE because there's room for variation based
on phonological environment. An interesting question is, what is the
reflex of *sp'? Does it become *sw or perhaps *sp. If the latter then
we should expect to see s-mobile variants like *sp and *w. Another
clue might be roots with the nasal infix. We'd expect *w~*mb variants.
Externally we should expect to find some examples of PIE *w (or
perhaps even *b) corresponding with Altaic *b. Another important
question is what *p' is reflected as in the other languages,
especially Uralic.

Okay, there's a lot to rip apart here, so have at it.