An evaluation of the Nahuatl data in Brian Stubbs' work on Afro-Asiatic/Uto-Aztecan

I have previously written about how, in the 16^th century Franciscan friars believed that St. Thomas the apostle had visited Mexico and preached Christianity to the natives 1500 years before the arrival of the Spaniards, and that the Indians themselves were descendants of one of the lost tribes of Israel. The Franciscans saw similarities between Indigenous religion and Christianity that they explained to themselves in this way. Today such a belief of visits from the Ancient Near East to Mesoamerica is not common among Christian denominations, but it is found today among the Latter Day Saints (also known as Mormons), whose sacred book, the Book of Mormon tells that four Hebrew tribes made it to the Americas. Where their prophets wrote the original Book of Mormon on metal plates in a language named "Reformed Egyptian". LDS scholars have, over the years, invested much time and energy in trying to find external evidence in support of the account given in the Book of Mormon, both through archaeology and linguistics. 

Front page of the work under review.

Usually, I would follow Jay S. Gould in considering scientific inquiry and religious confession to be non-overlapping magisteria, and that as long as scholars keep their religious beliefs out of their scientific inquiries then they can believe whatever they want. But sometimes this is not so easy, and this blogpost is about one of those times.

In this blogpost, I analyze the use of Nahuatl data, in Brian D. Stubbs self-published manuscript “Exploring the Explanatory Power of Semitic and Egyptian in Uto-Aztecan”. This work compares Proto-Uto-Aztecan with Semitic and Egyptian and seeks to find signs of ancient contact between early near-eastern and Egyptian peoples and Uto-Aztecan speaking Native Americans. It finds a lot of such signs, in fact more than 1500 Uto-Aztecan forms that Stubbs claims are cognate with Semitic or Egyptian forms.

Now, far-fetched proposals of relations between languages that are never known to have been in contact and which defy the conventional view of world history are a dime a dozen. I have already described Denison’s attempt to show that Nahuatl was an“Aryan language”, and Turkic nationalists frequently look to Nahuatl when seeking to explain their belief that all languages descend from the Turkic “Sun Language”. Usually, I would say that it is better not to give too much attention to far fetched claims of long distance contact across the Atlantic, and better not to waste one’s energy trying to debunk them since they are usually not playing a game in which the scientific rules and methods of historical linguistics even apply. But one thing makes this proposal different: Namely the fact that Brian D. Stubbs is a well-esteemed expert in Uto-Aztecan historical linguistics with many important publications to his name, not least his monumental catalogue of Uto-Aztecan cognate sets (Stubbs 2011). This he is, in addition to being also, apparently, a member of the LDS church. This makes for an interesting conundrum, because it both gives us an apriori reason to believe in the validity of his claims because of his expertise, and an apriori reason for doubting it, because of the fact that his claim is clearly meant to validate the narrative of the Book of Mormon, and his own religious belief. (it is certainly taken as such validation by other LDS members and publicized as such) 

I have a great deal of respect and esteem for Stubbs’ previous Uto-Aztecan comparative work, but the claim of ancient trans-Atlantic contact, even disregarding any potential religious motivation behind it, is a sufficiently extraordinary claim, for it to require inordinately extraordinary evidence in its favor, before it can be accepted.

C.S. Lewis coined the term ‘bulverism’ to describe the kind of argument in which one simply assumes that one’s opponent in a discussion is wrong, and then proceeds to explain to him the psychological mechanism that must have led them astray. If we want to avoid making ourselves guilty of bulverism, we cannot simply say “oh this is just religious crack-pottery” and leave it at that. We have to actually show, that Stubbs’ work in this case does not follow what is expected from rigorous linguistic scholarship, and that its conclusions therefore cannot be accepted as valid.

The manuscript is 444 pages and includes some, 1500 proposed cognates, so it would require quite a lot of effort to analyze all of it, so instead I will limit my analysis here to Stubbs' treatment of Nahuatl. Nahuatl being of course my field of specialization, and the area where I will be most likely to see weak spots in his argument and catch any methodological blunders.

Previous Reviews: Roberts, Elzinga and Rogers

I am not the first to analyze or review Stubbs' Semitic/Uto-Aztecan work. It has been previously reviewed by three Brigham Young University linguists all specializing in Native American languages: Dirk Elzinga, John S. Robertson and Chris Rogers. It does appear, though, that my review here is the first written by a linguist who is not affiliated with the LDS-owned BYU, and not published in a LDS related journal. 

Published in the "Interpreter: A Journal of Latter-day Saints faith and scholarship", the review by John Robertson, a specialist in the Maya language, is enthusiastic and overall accepting. Robertson concludes that he "cannot find an easy way to challenge the breadth and depth of the data". 

Published in "BYU Studies Quarterly", the review by Dirk Elzinga (a Uto-Aztecanist) is what I would call lukewarm, concluding that though the proposal looks like normal crackpottery at first glance, the authority and expertise of the author means that it merits further attention. 

Published  in the "Journal of Book of Mormon Studies" in 2019, the review by Chris Rogers, an expert in historical linguistics and the Xinka language of El Salvador and also a professor of linguistics at Brigham Young University, is much more critical. 

Rogers points out some serious flaws in the work: Primary of these is that it does not stick to the established bilateral method of comparing languages with languages and proto-languages with proto-languages, but that it frequently cherrypicks so that a form in any Semitic or Egyptian variety can be compared with a form in any Uto-Aztecan language. As has been demonstrated time and time again this multilateral method hugely increases the risk of mistaking chance resemblances for cognates, and makes it possible to prove virtually anything. As Rogers’ points out since any two languages can be expected to have between 1% and 3% chance cognates, if we add additional languages to the comparison the risk rises incrementally as well. This means that once Stubbs is comparing 30 Uto-Aztecan languages with at least three Semitic and Egyptian varieties (actually more, including at least Egyptian, Coptic, Syriac, Hebrew, Aramaic) the number of expected chance similarities far exceeds the 1528 proposed cognates. He also notes that Stubbs does not adequately explain non-matching segments (in fact no explanation at all is given in most cases), and that he does not adequately account for other potential explanations of the similarity (such as onomatopoeia). These general criticisms seem absolutely valid to me, now on having reviewed Stubbs' work myself.

In a response to Rogers’ review published as a pdf on the academia.edu site, Stubbs defends himself by stating first of all that Rogers has misunderstood his intent: He is not trying to prove that Uto-Aztecan and Semitic/Egyptian are genetically related, but that his claim is rather that proto-Uto-Aztecan was a mixed language that acquired a major portion of its vocabulary and grammar from speakers of these Afroasiatic languages. Therefore, it seems to follow, he does not need to follow the strict methods for demonstrating long distance relationships. As for the charge of cherry picking, he claims that it is only natural that some forms borrowed into the proto-langage survive only in some of the daughter languages. This is perhaps true, but he apparently does not recognize, or address, the fact that this practice leads to a much higher risk of chance resemblance being mistaken for cognates, that is, random noise being mistaken for a signal. Furthermore, it cannot simply be assumed that a given form in a daughter language is a retention just because it looks like something in Hebrew, it has to be demonstrated that it is not an innovation in the individual language. 

Though I can see why Rogers thought that Stubbs was arguing for a genetic relation given Stubbs presentation of the evidence, I wish Rogers had realized that Stubbs’ claim was in fact a proposal of language contact. Because it really is a more problematic claim.

It is problematic because there is no accepted method for demonstrating borrowing or contact induced changes, and consequently no method for falsifying them. Systematic sound laws do not apply in the transfer of elements of one language to another, a language may borrow many words or few and change them in fairly random ways as they are adapted to the borrowing language's phonology, and there really is no good way to disprove a claim about a form in one prehistoric language being borrowed from another. This is why linguists normally would never even entertain the idea of a scenario of borrowing, unless there is independent evidence suggesting probable historical contact between the two languages. In this, case there is exactly zero independent evidence of contact between Ancient Semites or Egyptians and Uto-Aztecans…except for the Book of Mormon. 

So by presenting his hypothesis as a claim of ancient contact and language mixture, Stubbs is in fact making a claim that cannot be methodically falsified. When working outside of historical disciplines, such a claim is usually called pseudoscientific, but in a historical discipline such as this, we can only strive to classify it either as convincing or unconvincing given the presented evidence. 

Analysis

So, to the evidence: In the following, I will assess as many examples as I could find of Stubbs using Nahuatl data to support a claimed cognate set. I am not looking at those parts where he uses Nahuatl as one language among many to support a PUA reconstruction, but only those where Nahuatl is the only support for a claimed relation across the families. Grantedly, most of his comparisons are to his own reconstructed proto-Uto-Aztecan forms, but he does on many occasions reach into Nahuan and compare Nahuan forms directly with forms in individual semitic varieties. Looking closely at these examples will give us feel for how Stubbs handles linguistic data - and specifically whether Stubbs' claim that dipping in to the individual languages is justified because these forms are retentions, is itself justified.  

Shadow:

šwt ‘shade, shadow’ > Nahuatl seewal-li ‘shade’

This example illustrates several problems. First is the fact that semitic and Egyptian roots are triconsonantal, whereas Uto-Aztecan roots almost always are CV or CVC roots. This means that from the outset when comparing a form across the languages you have to ignore all the vowels and look only at the consonants. This of course raises the number of potential cognates on the Uto-Aztecan side quite a lot since you can ignore basically half of every word. Secondly, Stubbs argues that all Semitic sibilants (three different ones) are reflected as *s in Uto-Aztecan. So again, this kind of merger, raises the number of potential cognates on the Semitic side. So now we have quite a large pool of potentially cognate words in each of the language families, the only thing needed for a chance resemblance to appear is if two of them have the same meaning. The greater the number of potential cognates, the greater also the chance for a semantic match or near match. Usually when demonstrating a genetic relation only words with the exact same meaning are accepted, since if we now also relax the degree of semantic fit we require, then the risk of chance resemblances increase even more. Here, the fit seems to be quite good, superficially. Because though Classical Nahuatl seewalli refers to 'shade', it does not do so etymologically. The etymology points to an original meaning of "a cool place", or "something that has cooled down". We know this, because it is composed of four different morphemes. The morpheme see ‘cold’ and the verbal formant –wa form the verb seewa ‘to be cold’, the –l suffix is an old passive form, and the –li ending is the absolutive suffix. So the word in Nahuatl is analyzable into mono-consonantal roots each with a distinct meaning that comes together to become ‘cold place’ which is then extended to also mean ‘shade’, and the analysis is not at all compatible with the proposed Semitic cognate which is a triconsonantal root.

Bow/Rainbow

šmrt ‘large bow’, pl šmrwt > -samaaloo-t of Nahuatl koo-samaaloo-tl ‘rainbow’

This example also shows the problems mentioned above, of neglecting to analyze the Nahuatl word into its roots. What is basically being compared here is šmrt with smlt, this can only be done because Stubbs decides that Nahuatl ko- is a prefix that can be removed, and because he neglects to remove the final –t which *is* a suffix and *should* be removed. But ko is not likely to be a prefix, indeed the likely historical analysis of the Nahuatl word is kosa-ma-l-o-tl, where the kosa- root is found in the word for yellow kos-tik and the word for becoming yellow kosawi, and the word for necklace koskatl. So he should compare šmrt with ksml, but that doesn’t look very much a like at all. Then there is the problem of semantic stretch, he is comparing two words that are related in English “bow” and “rainbow”, but why assume that this is a natural semantic connection? Nahuatl has other words for the weapon, and no word for the architectural feature, the Nahuatl word does  not appear related to the shape at all, but rather to the color, and Nahua myth compares it to a snake not a bow.    

Reed

 twr ‘reed’ > Nahuatl tool-in ‘cattails, reeds’;

This actually seems reasonable to me since I would probably reconstruct Nahuatl toolin as coming from an earlier form along the lines of *tawri. But still we are only matching three segments out of five, which means the risk of chance resemblance is high.

Son/Child

Hebrew bεn ‘son’; pl: bəneey3‘children (of)’ > Nahuatl *konee 'child, offspring’:

This one is a far stretch phonologically with only the –n being a direct match between the two forms. Stubbs argues that PUA *kw may be correspond to Semitic *b, but konee doesn't come from a proto-form with kw (As Stubbs notes Nahuatl is one of the only UA languages to keep PUA *kw as kw). In fact I think it probably comes from the PUA root *kumCa reconstructed by Stubbs (2011) as meaning "husband" and as "male".  

Serpent/Twin

Egyptian qrђt ‘serpent’, Egyptian qrђ ‘friend, partner’ > UA/Nahuatl koŋwa ‘snake, twin’

Here it is certainly interesting that both Egyptian and Nahuatl have a word that means both snake and a human friend/partner/twin. Within Uto-Aztecan, Nahuatl is to my knowledge the only language that has this double meaning of the word for snake, and thus the double meaning cannot be reconstructed for PUA. 

And the phonological form is quite far from each other: really there isn’t even a single segment that can be considered a direct match between the two languages. One has to accept Stubbs’ complicated sound correspondences where multiple Semitic/Egyptian segments match a single segment in Nahuatl and in which vowels are irrelevant, and consonant segments can dissappear. In this particular instance, he has to insert an n to get an nw cluster that can correspond to rђ, and to get this n into the word he cites a 1976 paper by Joe Campbell with which I am very familiar (it is about Hueyapan Nahuatl, which is my main expertise among Nahuatl varieties). First of all, Campbell does not in fact argue that there is a historical /ŋw/ in any variety of Nahuatl, he only makes the argument that synchronically there can be posited an underlying hypothetical *ŋw phoneme that explains some irregularities in the grammar. Joe has confirmed to me several times that he did not mean to make a historical argument, but was only making a synchronic phonological argument in the style of the structuralism of the 1970s. Secondly, Joe does say that the hypothetical ŋw element is found in the word ko:wa, but he is talking about the verb ko:wa “to buy”, not the noun ko:watl “snake”. So it is a completely different unrelated word. The reference to Joe’s article to justify the claimed medial /ŋw/ is both a case of special pleading, and of misusing another scholar’s work in doing so.

Crocodile

The Egyptian Crocodile God Sobek
(photo Hedwig Storck, WikiCommons)

Egyptian sbk ‘crocodile, the crocodile-god Sobek’ and Classical Nahuatl sipak-tli ‘crocodile’.

Here again we have a good superficial likeness with actual near match of all three consonants – and Stubbs’ further argues that in Egyptian the voweling matches the Nahuatl form as well. We need to abstract, of course, from the fact that elsewhere the comparisons is between Nahuatl and Hebrew, Nahuatl and Maghrib Arabic, Nahuatl and Aramaic, and here Nahuatl and Egyptian (which is not even a Semitic language but related in the much larger Afro-Asiatic family).

Aztec crocodile god Cipactli,
from the Codex Borgia

The problem is that we cannot reconstruct this word for crocodile for proto-Uto-Aztecan, because no other Uto-Aztecan language has a documented cognate of the word sipaktli. There is a possibility that the word could be related to the Corachol word for caiman háaxi, where the first syllable likely means "water". The second syllable -xi could then be cognate to the si- syllable of sipaktli (and in fact the pa- syllable could be cognate to the ha- syllable of corachol, the order of the elements in the compound simply reversed). This would make sipaktli a likely compound word, in which case it cannot match the Egyptian triconsonantal root at all.  

Wine-skin/Prickly pear

"Hebrew nebεl ‘skin-bottle, skin’ in the common phrase of Hebrew nebεl yayin ‘skin of wine’; Syriac nbl3/3n’bl > Classical Nahuatl no’palli ‘prickly pear’ often used to make alcoholic beverage"; 

Here we have an ok phonological match (though it unexplainedly ignores the Nahuatl Saltillo segment), but a very bad semantic match. 

The Nahuatl word nohpalli quite simply does not mean prickly pear, it refers to the opuntia cactus, the prickly pear of which is called noochtli. Nohpalitl, also refers specifically to the edible ear of the cactus. Though mostly eaten as a fruit, the noochtli was used to produce a kind of fermented beverage called noochoctli. But the nohpal cactus is only tied to alcohol production in this very indirect way. Allowing this span from wineskin to cactus with a fruit used occasionally for fermented shows an very high degree of semantic latitude, and a cavalier approach to translation since nohpal is *not* the prickly pear or the part of the plant used for fermentation. 

Bury/Tamal

Semitic ṭmn > Aramaic ṭmr ‘hide, bury’ > Nahuatl tamal-li ‘tamale’

Stubbs argues that the Semitic root *tmn had the “references to ‘cooking underground or under ashes’ …which in Post-Biblical Hebrew also meant ‘put in an oven’” He makes the final l- in the Nahuatl fit by noting that “Aramaic changed n > r, as it often does (ben ‘son’ > bar ‘son’)”. 

But again, he doesn’t take the time to analyze the Nahuatl word, which does indeed come from a verb that can be reconstructed as *tɨma with the meaning “cook with steam/bake under ground”. But the final –l in tamal, is not a part of the root but a suffix, it is as mentioned before an old passive that is used to derive deverbal passive nouns, so a tamal is analyzable in Nahuatl and means “something steamed”. So again we have only two consonants out of three (or if counting the vowels two segments out of five) that actually match. 

Perfective prefix

Semitic perfective with wa- Nahuatl perfective with o-.

Stubbs notes that in some Semitic verbs a perfective can be formed by adding the prefix wa- and removing the last vowel of the verb. This is indeed quite similar to what happens in Western Nahuatl dialects, where an o- prefix and the loss of the last vowel creates the perfective of one verb class. But, this is limited to Western Nahuatl dialects, the Eastern dialects have neither the o- prefix nor the dropping of the stemfinal vowel in any verb class (they use a –k or –ki suffix instead). The vowel-dropping in preterit forms cannot be reconstructed for proto-Nahuatl, but is an innovation in the Western Branch. Karen Dakin has argued that Western Nahuatl ot the o- prefix as a borrowing from the corachol perfective prefix wa-. And it probably also shouldn’t be reconstructed for proto-Nahuatl.  So here we have a pattern that is superficially intriguingly similar, but once we know a bit about Nahuatl historical developments it disappears entirely.

Climb up/on top

Semitic rkb ‘mount, climb up on’ > CN tlakpa-k ‘above, on top’

Again Stubbs compares a triconsonantal root to a multimorphemic Nahuat word. The tla- in tlakpak is a prefix, that was originally ta-, and the root ikpa comes from the PUA root *kupa ‘hair’ or ‘head’ and has come to mean “top”.  The root rkb ‘mount climb up’ is not a very good match for kupa ‘hair/head’.

Quail

Hebrew śәlaaw ‘quail’, pl:salwiim; Syriac salway ‘quail’; Arabic salwaa ‘quail’; Samaritan šalwi > UA *solwi ‘quail’: CN sool-in ‘quail’; Mn sowi’ ‘pigeon’.

Here we have something that is superficially interesting again. But we have to note that ' reconstruction of the UA term combines the Mono word for "pigeon" with the Nahuatl word for quail to get an almost exact match for the Semitic word. But he has missed the obvious cognates in Corachol, namely Huichol xïau “codorniz” and Cora sa’uh. These forms, I reconstruct for proto-Corachol as *sauri, which is also the ancestral form for the Nahuatl word sool-in. Then we end up comparing slw with swr (if we admit the correspondence u/w), and we would have to posit a metathesis to sustain the argument.   

Understand/Grow

Hebrew hiśkiil, hiskal- ‘to understand, comprehend, make wise’ > CN iskal ‘to train’; CN iskal-ia ‘be discreet, prudent’.

Simeon's entry for Izcalia

Here we have something really bad. Namely an example of a massaged translation of the Nahuatl term that makes a large semantic stretch seem less problematic. Nahuatl iskalia does not mean "be discreet prudent". The intransitive iskali means “to grow” or "to revive" for example about plants that sprout. From this is derived the transitive iskalia that means "to make someone revive or grow" and which is used metaphorically in the sense of coming to one's senses for example after having passed out (used reflexively "make oneself revive"), and in the sense of nurturing and rearing a child to maturity (like one cultivates a plant). In Hueyapan one does respond to someone who says something silly by saying "ximoskali!", which is literally "come to your senses!". But is is related to reviving and coming back to life, and not to understanding or knowing. 

Molina's entry for Izcalia with no "prudence or discretion"

Stubbs' seems to have the "discreet, prudent" translation from the dictionary of Remi Simeon where "discreto, prudente" appears in a string of words used to translate iskalia (image inserted right). Simeon's dictionary is a derivative, mostly based on Alonso de Molina's dictionary, and interestingly the "discreet, prudent" doesn't appear as possible translations of the words in Molina's dictionary (image inserted). They were added by Simeon. The gloss "to train" for the putative word "iskal" is simply made up perhaps based on the metaphorical meaning of iskalia "to rear" a child. "Iskal" meanwhile, is not an actually existing Nahuatl word, since as I am sure Stubbs knows, all Nahuatl verbs end in a vowel. 

But of course when comparing, one has to try to analyze the word's core semantics instead of the simply choosing one of the potential translations that you like best. Here the semantics simply doesn't fit, and instead of analyzing the word's meaning Stubbs simply cherry picks the two of Simeon's glosses that fit best with his Hebrew word. 

Select/Take

Hebrew(BDB) brr ‘to select, choose’: CN kwia / kwiya ‘to consider s.th. one’s own, to keep’; CN kwi-lia ‘to take s.th.’;

This is also an example of a massaged translation, because Nahuatl kwi means simply to take something, kwia is a derived verb that means "to wrap something" and kwilia is the applicative of kwi that means "to take something from someone". Simeon again has a meaning of kwia not found in other dictionaries, namely the meaning of "keeping something borrowed". This could be a potential contextual extension of the sense of "taking something", with the added -a, to signify that it is used transitively. But given that there is no other source for this usage than Simeon's dictionary, that is not very accurate, and often supplies extra translations based on unknown evidence, it seems a bad idea to pick this specific meaning to compare with Hebrew.   

Apart from this semantic mismatch, the only element that actually matches is the b/kw. The r segment matches neither the l in kwilia because this is the applicative suffix. The y in kwiya suggested by Simeon is spuriously inserted because if there was the preterit would be *kwix, but is in fact kwih. Hence the *y cannot be used to match the r in the semitic form. 

Pronominals

One of the really eye-catching pieces of data presented by Stubbs is the correspondence between the semitic pronominal prefixes and the pronouns of Classical Nahuatl. He notes that the Aramaic verb 'to be' parallels the Nahuatl pronouns closely, being 'ehwe "I am", tehwe "you are" and yehwe "he is". And he notes that Maghrib Arabic analogized the first person plural n- to become first person singular as in Nahuatl. Here, I reproduce Stubbs' table from page 335.  

This kind of paradigm is the kind that historical linguists love, because it is a kind of relation that is relatively unlikely to arise by coincidence. But though it may be unlikely to arise by chance, it is not impossible. In this case in fact very possible.

In their reviews, both Elzinga and Robertson catch on this example  as the most convincing piece of evidence. Rogers though mentions the table critically, by pointing out the problem that Stubbs is here comparing individual varieties, and not proto-languages. Here Rogers points out a problem that in fact invalidates this example: Again the Nahuatl forms represent innovations within Nahuatl not retentions of earlier PUA forms, and hence cannot be used as examples of contact allegedly taking place thousands of years before proto-Nahuatl emerged.

The reason Stubbs' compares with Classical Nahuatl instead of PUA, or even instead of proto-Nahuatl is clear: this particular pronoun system is *only* found in Nahuatl in the entire UA family, and consequently cannot be reconstructed for the earlier stage. In fact, we can show that it is an innovation in proto-Nahuatl by comparison with the forms found in Corachol. In proto-Corachol the paradigm was 1p ne-, 2p pa- 3p pu-. Stubbs himself in his catalogue of UA cognates reconstructs pu- as the third person singular pronominal in PUA. From a system similar to that found in corachol, Nahuatl switched the second person form, apparently adopting the ti- prefix of the first person plural, also as the second person singular prefix. It kept the ni- prefix and it dropped the third person prefix altogether since it was redundant. The pu- pronominal stem was in fact only kept in the third person pronouns which in proto-Nahuatl I reconstruct as *yeha from an earlier *puha. The initial PUA syllable *pu becomes *hɨ- in proto-Corachol-Nahuatl, then ye- in proto-Nahuatl and then e- in eastern Nahuatl (except in the pronoun because Eastern Nahuatl had changed the pronoun to yaha, and y- was only dropped before e). The original pronoun furthermore did not have the -wa suffix, which is an innovation in the Western branch of Nahuan (so quite late, after the split of proto-Nahuatl). Consequently, the forms of the Aramaic copula and Mahgrib pronominal prefixes are complete red herrings, they only compare to the pronouns of Western Nahuatl in the postclassic period (i.e. after 900 AD). 

Moreover, the Maghrib first person form n- is an innovation while in Uto-Aztecan it is a retention, and it is probably even a retention from a much earlier stage of language evolution in the Americas given that many other language families in the Americas have ni- as the first person singular pronominal. So really if we were to allege contact, it would suggest that Maghrib Arabic was influenced by Nahuatl (or another Native American language), rather than the other way round. But of course this is really just a coincidence, and not evidence of contact at all.

If we were to compare only the oldest reconstructible stages of the two languages we would get:

               Semitic          Proto-Corachol-Nahua
1.sg         '-                    ne-
2p.sg.      t-                    pe- 
3p.sg.      y-                   pu-/Ø-

1p.pl.       n-                  t-
2p.pl        t-                   se-
3p.pl        y-                  me-

Here, nothing at all is shared between the two systems. (I presented my preliminary reconstruction of elements of proto-Corachol-Nahua at the Friends of Uto-Aztecan meetings in Tepic in 2018 including the pronominal prefixes, the proceedings are forthcoming [now published here]).

Starting from two systems that didn't actually share anything, Proto-Nahuatl swapped a previous 2p.sg prefix pa- to ti-, and the changed the form of the 3p.sg. prefix pu- to ye-, and Mahgrib swapped 1p '- to n-, and voilá: the systems align. 

In conclusion, the pronoun system that Stubbs compares with semitic is much too late for being evidence of a relation between PUA and Semitic, since it came into existence only several thousand years after the existence of PUA and the alleged contact between Semites, Egyptians and Uto-Aztecans. This is a case in which we can in fact show that similar pronominal paradigms have developed independently of each other. 

Conclusions: It’s a no from me

For all these reasons, I find the proposal to be very far from convincing. It seems to me that here, Stubbs is not at all doing the kind of careful comparative work that he is known for. The handling of Nahuatl data is highly problematic, with massaged translations that make words that have virtually no semantic link falsely appear to have the same meanings, with  apparently selective failure to pay attention to segmentation and morphological analysis in the Nahuatl, and inattention to other possible and plausible explanations even those found in his own previous work. 

I should note, that since many of the reconstructions of proto-Corachol and proto-Corachol Nahuatl are my own and most of them yet unpublished, Stubbs cannot of course be faulted for not knowing or accepting them. But if anyone in the world would be equipped with enough knowledge to investigate the history of these forms in Nahuatl on their own, it is Brian Stubbs. And really, it don't seem that he has even tried to look into the immediate history of any of the Nahuatl terms he cites. He has just assumed that it was conservative. Investigating alternative explanations of one's data is of course a basic part of establishing a hypothesis in a rigorous manner. 

I am not equipped to evaluate Stubbs’ usage of the data from many of the other UA languages, nor the Semitic or Egyptian, but given how the Nahuatl is treated, it cannot simply be assumed that it is being handled well. I pass the baton to the next scholars to check how he handles the languages of their expertise, both Uto-Aztecan and Afro-Asiatic.

J. S. Gould also showed us that we all as scholars are prone to the error of confirmation bias. Dearly held beliefs, whether religious, political or theoretical, are likely to color our interpretation of data and dull our critical sense about our own conclusions. I think as a discipline, historical linguistics, because it relies on our ability to see intricate patterns that others have not yet seen, is more prone to being influenced by our biases in interpretations than most other kinds of science. Really, I think historical linguistics is perhaps more of an art, though an art that should be approached with a scientific mindset. This proneness to confirmation bias of course no less affects me as an non-theist scholar with a distinct set of ideas about what happened in prehistoric Mesoamerica, than it does people of other diverse persuasions and ideas. But this is why these kinds of scholarly endeavors have to stand and fall on the evaluation of empirical data by people with different biases: we are all entitled to our interpretations, but no one is entitled to their own facts. 

What would be needed for a proposal like this to be convincing to me? First of all the question is, how much will be left once experts in different other languages involved weed out the infelicitous examples as I have done here. This sample suggests that quite little would be left after such a pass-through. But the next version of this proposal should also take some steps to remedy the basic methodological flaws:

• I would very much prefer non-linguistic (that is archaeological or genetic) corroboration of ancient trans-atlantic contact before I would entertain the hypothesis of contact between Uto-Aztecan and ancient Afroasiatic languages as a reasonable explanation of likenesses between the languages. 
• I would want a much more thorough description of the proposed borrowings, including vowels. Uto-Aztecans would have heard the semitic words with vowels, and borrowed them with vowels. So it is simply not possible to simply ignore the vowels as Stubbs does (exept when he finds one that accidentally fits). I would expect systematic vowel patterns for verbs, nouns etc. Also forms with partial matches, where only two of three consonants match, cannot be allowed.
• I would want attention paid to chronology. When did the supposed borrowings take place? Already before proto-Uto-Aztecan? In that case all proposed borrowings must be reconstructible to  PUA and to which ever layer of Afro-Asiatic or Semitic one believes was spoken at that period. It is not reasonable to cherry pick forms from the individual daughter languages and claim that they are retentions when the may as well be innovations (such retentions can only be posited after the relationship is established).
• As Chris Rogers’ points out a convincing proposal would in fact have much fewer correspondence pairs, of much higher quality, and preferably, they would be paradigmatically related. For example, what is closest to appear convincing for example is the pronominal forms, where the n, t- y- pattern looks highly intriguing (until you remember that the first person n- is pan-American and realize the y- doesn’t fit).

One thing that makes me uncomfortable is the fact that Stubbs worked on this simultaneously with working on his catalogue of Uto-Aztecan reconstructions. I cannot help but worry that the reconstructions of UA forms there, may be subtly compromised by Stubbs unconsciously trying to make them fit with his Semitic data. I will have to use it with a degree of apprehension in the future.

Texts Cited/Mentioned:

• Elzinga, Dirk. 2016. "Review of Exploring the Explanatory Power of Semitic and Egyptian in Uto-Aztecan" in BYU Studies Quarterly. 55(4):172-176
• Rogers, Chris. 2019. “A Review of the Afro-Asiatic: Uto-Aztecan Proposal”  Journal of Book of Mormon Studies, Vol. 28, 258-267
• Robertson, John S. 2017. "Exploring Semitic and Egyptian in Uto-Aztecan Languages".
• Interpreter: A Journal of Latter-day Saint Faith and Scholarship 25: 103-116
• Stubbs, Brian D. 2015. Exploring the Explanatory Power of Semitic and Egyptian in Uto-Aztecan. Grover Publications. 
• Stubbs, Brian D. 2011. Uto-aztecan: a comparative vocabulary. By Brian D. Stubbs. Blanding, Utah: Rocky Mountain Books and Publications