Go to one Akan-speaking area and ask for the word for head.
You'll get etiri.
Go to another and ask the same question.
You'll get eti.
Same word. Same meaning. Same language. The ending just isn't there.
Nobody's wrong. That's what dialects do — they wear words down in different places at different rates, and the -ri on the end of this one has dropped in some mouths and not others.
Small, ordinary, entirely unremarkable.
And it's carrying an enormous amount of weight in this corpus.
Stated plainly: in some Akan dialects a terminal -ri is lost.
The teacher gives two examples from the living language.1
Etiri → eti. Head, in Asante versus Akwamu and Akuapem.
Bidi / biri → bi. Black or dark. Ebiri mai, the black nation, contracting to ebi.
That's it. That's the rule.
It's a genuine feature of the language, the kind of thing any descriptive grammar records without comment.
Now here's where it stops being small.
The Coptic form of Asar's name is Ousiri.2
The Akan name for the same divinity is Awusi.
Those don't match. Ousiri has an ending that Awusi doesn't.
Apply the rule and they do match. Ousiri loses its terminal -ri, becomes Ausi, and lands on Awusi.3
The whole identification turns on that one step.
Without the rule, you have two similar-ish names and a gap. With it, you have a derivation.
This is worth being precise about, because it's the single most consequential logical step in the corpus's etymological layer.
The rule is real inside Akan. Etiri really does become eti. Nobody disputes that.
The question is whether a sound rule that operates within a language can be applied to a word coming into that language from outside.
Those are different things.
An internal sound change describes how a language's own inherited vocabulary wears down over time, in a regular way, across many words.
Applying it to a proposed borrowing assumes the borrowed word entered early enough and integrated fully enough to undergo the change — which is possible, and is also exactly what needs demonstrating rather than assuming.
The corpus treats the rule's existence as sufficient. It exists, therefore it applies.
There's a test that would settle a lot, and nobody runs it.
Does the elision get invoked everywhere it should, or only where a match needs rescuing?
If terminal -ri elision is a general feature being applied consistently across the whole comparative set, that's a principled rule doing honest work.
If it appears specifically at the points where two names would otherwise fail to line up, it's not a rule. It's a repair.
The difference is enormous and it's checkable — you'd walk the corpus's own examples and count.
I can't run that count from six sessions of transcript. What I can say is that in the material here, the elision surfaces at exactly one high-stakes juncture, and it is the juncture where the flagship identification would otherwise break.4
That's [n=1 OBSERVED]. It is not evidence of bad faith. It is a reason to want the count.
The teacher gives two more contraction examples in the same stretch, and they're doing similar work.
Ousiri → Ausi, via the -ri rule.5
Ausite → Ausi, a separate contraction he offers alongside.6
Tep → teep → tip / top. The word for head in Kemet, proposed as the source of English tip and top.7
That last one is a different kind of claim entirely — it reaches out of the Akan-Kemetic comparison into Indo-European, where the etymologies of top and tip are well documented and Germanic. It belongs in the ledger and it isn't defended here.8
But notice the shape. Each of these is a small phonetic adjustment that closes a gap. Individually each is plausible. Collectively they form a toolkit of gap-closers, and a toolkit of gap-closers can close any gap.
He's got the Akan dictionary open and he's showing you etiri.
Then he shows you eti, in a different dialect entry.
"So the same thing happens with Ousiri — Ousiri becoming Ausi in our kind."9
The demonstration and the application arrive in the same breath. He shows the rule working inside Akan, then applies it across the boundary, without marking that a boundary was crossed.
You're trying to evaluate a comparison someone's made between two words in different languages.
Useful habit: strip every adjustment and look at the raw pair.
What do the two words actually look like before anyone drops a syllable, shifts a vowel, or invokes a dialect?
If they match raw, that's interesting. If they only match after two or three operations, count the operations — because each one is a degree of freedom, and enough degrees of freedom will connect anything to anything.
You catch yourself doing it in your own reasoning about something unrelated.
You've got a theory and a piece of data that doesn't fit, and you reach for an adjustment. The adjustment is legitimate — it's a real thing that really happens.
The question you don't ask is whether you'd have reached for it if the data had fitted.
That's the same move, and noticing it in a hieroglyph argument is easier than noticing it in your own.
Lay the whole derivation out and count.
Start: W-S-R, the consonantal skeleton.10 Attested.
Step 1: Coptic supplies vowels — Ousiri / Ousiti.11 Attested, and this is the strong part.
Step 2: Terminal -ri elides. Ousiri → Ausi.12 Rule real within Akan; application across the boundary asserted.
Step 3: Initial vowel shifts. Ou- → Awu-. Not separately defended; treated as obvious.
End: Akan Awusi, the Sunday divinity.13
So: one attested step, one asserted step, one unexamined step.
Now compare that to how it's presented. In the session it arrives as a single smooth movement — Ousiri, Ausi, Awusi, same divinity — with the joints invisible.
The derivation might well be right. Sound changes across borrowings do happen and often look messier than this.
But the honest summary is: the identification rests on two unverified phonetic operations and is presented as a demonstration. That gap between what's established and how it's stated is the recurring defect in this layer of the corpus, and it's a defect of register rather than of observation.
If the same claim were made as "here's a plausible derivation requiring two sound changes I haven't independently tested" — it would be a decent working hypothesis, and nothing would be lost.
Strongest support. The rule is real. Etiri → eti and biri → bi are genuine Akan dialect facts, verifiable in the dictionary the corpus cites, and not invented for this argument.
The central tension. A rule's existence inside a language is not a licence to apply it across a proposed borrowing. The corpus never marks the boundary crossing, and the flagship identification depends on it.
A second. The elision surfaces at precisely the point where a high-stakes match would otherwise fail. [n=1 OBSERVED] — one instance, in one session. Not damning; a reason to count.
Open. Run the count. Walk every Akan-Kemetic pair in the whole Odwirafo corpus and record where phonetic adjustments are invoked. If they're distributed evenly, the rules are principled. If they cluster at the near-misses, they're repairs. That's a bounded piece of work and it would settle the credibility of the entire etymological layer in one pass.
Against The Last Version of the Language Still Has Its Vowels, this page names the weak joint in that page's flagship example.
Coptic is the strong instrument — attested, vocalised, standard. But Coptic gets you to Ousiri, and Ousiri is not Awusi. The elision rule is what bridges the last gap, and it is the least tested thing in the chain.
So the two pages together produce something neither says alone: the corpus's most rigorous instrument delivers you to the doorstep of its least rigorous one. The strength of step one is doing rhetorical work for step two.
Against Odwirafo's Comparative Linguistic Method, this page supplies the specific mechanism that page describes generally. Phonetic-similarity comparison is the method; terminal elision is one of the operations that makes the similarities close. Naming the operations separately is what lets you count them.
Cognitive Dissonance — Festinger
Festinger's finding was that a believer confronted with disconfirming evidence doesn't usually drop the belief. They add something — a supplementary assumption that lets the belief and the evidence coexist.
That's exactly what a phonetic adjustment does here. You hold an identification. A word doesn't fit. Rather than abandon the identification, you introduce a sound rule that accommodates the mismatch.
And the crucial part, which is what makes the move so hard to catch: the supplement is usually true. Terminal -ri elision genuinely happens in Akan. Festinger's subjects weren't inventing falsehoods either; they were reaching for real-enough considerations that happened to point the convenient way.
What makes it dissonance-reduction rather than reasoning isn't falsehood. It's asymmetric deployment: you reach for the supplement when the belief is threatened and not when it isn't.
The insight the pairing produces is a usable test, and it's a better one than "is this claim true."
Ask instead: would this adjustment have been invoked if the data had already fitted?
For terminal elision, the answer is checkable — count where it appears across the corpus. That converts a vague suspicion of bias into a bounded empirical question with a number at the end of it.
And it cuts fairly in both directions. If the count comes back even, the corpus is vindicated on exactly the ground its critics attack it. Most accusations of motivated reasoning are never made falsifiable enough to be withdrawn.
Sharpest implication. The most consequential claim in this corpus's etymological layer rests on a two-syllable sound change that nobody has counted. Not a large claim resting on a small one — a large claim resting on an uncounted one, which is a different and more fixable problem.
Generative questions.