Unexplained deletions: continuing what appears to be a common theme
Without expertise about each and every language included in the hypothesis, the author cannot evaluate the claims that they themselves are making. That is also my main point. I know Germanic, Slavic, Italic, they are all difficult by the languages included in them, and I still do not dare doing Indo-European reconstruction. What competence has he who wants to reconstruct Nostratic? Of course all goes by wordlists and statistics ignoring all laws of grammar and sounds and regular correspondences which one learns if one deals with multiple languages of a family. This way in StarLing there are all kinds of wildcards for which one could easily posit the correct phonemes if one knew the languages. And of course a user who does that statistics and wildcard thing does not have a babel box.
That is a fair point, how can anyone even think of having a good grasp of all those languages that are supposedly Nostratic, enough so to construct Proto-Nostratic? Even Indo-European studies took a long time to come to where it is today, with lots of wrong ideas and assumptions. Even if Nostratic is a language family, it is beyond modern-day linguistics's abilities to reconstruct.
It is already remarkable how little reliable the reconstructions of Semitic are. In this Article Igor Diakonoff and Leonid Kogan unfold the absurdities in Vladimir Orel’s and Olga Stolbova’s Afrasian etymological dictionary:
Lua error in Module:quote at line 2959: Parameter 1 is required.
We find broken plurals – an Arabian innovation – as Semitic etyma, late Iranian borrowings, sounds in Akkadian or Hebrew that did not exist, invented meanings …
In the work by Alexander Militarev (find his name on the sidebar) concerning specifically Semitic I even found sounds postulated for Proto-Semitic that did not exist, like he postulates labialized consonants in Semitic; quoting Nr. 15 of his word list: “West Semitic 1: *ḳʷr(r) ‘to be cold’ (#2), possibly related to Afras. *ḳVr- ‘dry”. The selectiveness in forms is worrying. Like for “new” Nr. 59 he gives, reconstructing *ḥadit-, Arabic ǯadīd, though there is also حَدِيث (ḥadīṯ, “new”) which he does not mention. Such is of course caused by using weird sources for representing languages, like the popular Penrice for Arabic – Orel/Stolbova not giving any sources. For *ṭāb- (“good”) (see the discussion on the talk page there why it is ṭāb-) he lists Nr. 34 *ṭayVb for no reason, even worse than Starostin who lists *ṭayb- in StarLing.