Part 1: Introduction and *h₁n̥gʷnís
Let’s get right into it.
This is one of the clearest traces we have of PIE’s internal history: r/n stems are pretty rare outside of the Anatolian languages, indicating that they’re the leftovers of an older system (just what I am looking for), and the trace of an old sound change is obvious. They’re all also neuter nouns, so another point for animacy distinction.
Since it ends in a syllabic consonant, I can slot in an **ə in front of *-r̥, being a reduced / unstressed **a. Then I just wind back *r < **n and we have **-an and a new step to our sound-change list
(This is going to need more specificity in the future, because we’ve got those pesky *-mn̥ stems to deal with.)
Word progress: -an
Word progress: -’tan
Traditional recon!PIE operates on the assumption that every syllable has a vowel slot, which can potentially hold *e, *o, the long versions of those vowels, or nothing at all. These are called “grades”, and are used to describe ablaut patterns: a slot is in such-and-such a grade in these cases, then shifts to another slot in other cases, it’s such-and-such grade when stressed, and this-or-that grade when unstressed. It’s all terribly complicated.
*a doesn’t fit into the ablaut patterns at all, so a lot of reconstructioneers will say “oh it’s an allophone of *e when adjacent to *h2” because that does fit ablaut patterns, and every other case will be written off as a marginal phoneme inherited from nursery talk or loanwords.
This runs into two problems: Problem 1 is that the traditional ablaut patterns might not be true in the first place. Paul Kiparsky (2010) suggests an alternative that he calls the Compositional Model, where stress placement previously treated as arbitrary ablaut patterns are the result of a process of several underlying rules applied in a specific order. This theory has not caught on with PIE studies at large (likely because it throws out like 50% of the entire field), but I find it a lot more appealing, understandable, and reasonable than the patterns and paradigms.
Problem 2 is that there are languages with only two vowels out there - conveniently located right next door in the Caucasuses - but those vowels are never /e/ and /o/. All the languages that have only two or three vowels will distinguish them by height, not by backness.
You will notice that all of these languages have /a/. In fact, the only languages I know of that don’t have /a/ are Arapaho and its closest relatives, and even then they have /i/ and /u/ to pick up the slack and it’s clear by comparison to the rest of the family that they had /a/ at some point and later lost it.
So either real!PIE had a vowel system that has never before been documented in a human language, or recon!PIE is using misleading symbols out of tradition.
(You can guess which side I come down on)
Kümmel (2012) posits that at least in early PIE, *e was **æ ~ **a and *o was **ɑ, which was itself descended from ***ā. This is, ultimately, the base I will be using because it requires the least amount of jumping through hoops (I will still be jumping through hoops, and I don't have all the edge cases hammered out yet, so for now all we need to care about is that *ó is **â (I’m using circumflexes for when it is both long and stressed)
Word progress: -â’tan
1.4 w
I have two options available
Word Progress: p-
*h₂ is the reconstructioneer’s magic bullet. You can drop it in anywhere to explain anything: it can turn *e into *a, it can turn short vowels long, it can aspirate voiceless stops, it explains ablaut paradigms; it can be a consonant, it can be a vowel, it can be a consonant that acts like a vowel or a vowel that acts like a consonant. Depending on who you ask it, it can be reconstructed as *q, *qq, *x, *χ, *ħ, *ɦ, *h, *ʕ or *a. I am certain there’s a paper out there claiming with utmost sincerity that it can turn lead into gold and resurrect the dead.
Kümmel (2022) once again saves the day with what I think is the most convincing argument. Short version:
Finally getting back to the word at had, this one comes with easy Anatolian attestation via Hittite paḫḫur and Luwian pāḫur. That double form is unique to *h₂, so I’m going to add a bog-standard unvoiced uvular fricative **χ. *e is **a, naturally.
(I could have *h₂ = **q, but I am saving **q for the time being.)
Word Progress: paχ-
Let’s get right into it.
1. *wódr̥
This is the most common recon!PIE word for water, surprising absolutely no one. It's not the only word, but it is the generic one vs the more active "body of water" found in *h2ep. I’m actually going to do this one backwards.1.1 -r̥
*wódr̥ is a fun one, because it’s what’s called a heteroclitic stem: In the “strong” cases (nominative, vocative, and accusitive) it ends in *-r̥, but in the “weak” cases (everything else, but genitive is the standard example), we find an *-n̥ where we’d expect to find an *-r̥ (ie *wódr̥ / *wédn̥s).This is one of the clearest traces we have of PIE’s internal history: r/n stems are pretty rare outside of the Anatolian languages, indicating that they’re the leftovers of an older system (just what I am looking for), and the trace of an old sound change is obvious. They’re all also neuter nouns, so another point for animacy distinction.
Since it ends in a syllabic consonant, I can slot in an **ə in front of *-r̥, being a reduced / unstressed **a. Then I just wind back *r < **n and we have **-an and a new step to our sound-change list
Final N > R Shift: **n > *r when word final, following an unstressed **a / **ə
(This is going to need more specificity in the future, because we’ve got those pesky *-mn̥ stems to deal with.)
Word progress: -an
1.2 d
Plain voiced stop, no special circumstances, this is going to be **t’ (or, since the stress is on the first syllable, **’t.)Word progress: -’tan
1.3 ó
No avoiding it now; it’s time to rip off the bandage and talk about vowels.Traditional recon!PIE operates on the assumption that every syllable has a vowel slot, which can potentially hold *e, *o, the long versions of those vowels, or nothing at all. These are called “grades”, and are used to describe ablaut patterns: a slot is in such-and-such a grade in these cases, then shifts to another slot in other cases, it’s such-and-such grade when stressed, and this-or-that grade when unstressed. It’s all terribly complicated.
*a doesn’t fit into the ablaut patterns at all, so a lot of reconstructioneers will say “oh it’s an allophone of *e when adjacent to *h2” because that does fit ablaut patterns, and every other case will be written off as a marginal phoneme inherited from nursery talk or loanwords.
This runs into two problems: Problem 1 is that the traditional ablaut patterns might not be true in the first place. Paul Kiparsky (2010) suggests an alternative that he calls the Compositional Model, where stress placement previously treated as arbitrary ablaut patterns are the result of a process of several underlying rules applied in a specific order. This theory has not caught on with PIE studies at large (likely because it throws out like 50% of the entire field), but I find it a lot more appealing, understandable, and reasonable than the patterns and paradigms.
Problem 2 is that there are languages with only two vowels out there - conveniently located right next door in the Caucasuses - but those vowels are never /e/ and /o/. All the languages that have only two or three vowels will distinguish them by height, not by backness.
- Abkhaz has either /ɨ/ and /a/ or /ə/ and /a/, depending on who you ask.
- Kabardian and Adyghe have /ə/, /a~ɐ/, and /ā/.
- Ubykh and Arente have /ə/ and /a/
You will notice that all of these languages have /a/. In fact, the only languages I know of that don’t have /a/ are Arapaho and its closest relatives, and even then they have /i/ and /u/ to pick up the slack and it’s clear by comparison to the rest of the family that they had /a/ at some point and later lost it.
So either real!PIE had a vowel system that has never before been documented in a human language, or recon!PIE is using misleading symbols out of tradition.
(You can guess which side I come down on)
Kümmel (2012) posits that at least in early PIE, *e was **æ ~ **a and *o was **ɑ, which was itself descended from ***ā. This is, ultimately, the base I will be using because it requires the least amount of jumping through hoops (I will still be jumping through hoops, and I don't have all the edge cases hammered out yet, so for now all we need to care about is that *ó is **â (I’m using circumflexes for when it is both long and stressed)
Word progress: -â’tan
1.4 w
I have two options available- Option 1: *w in this environment is **w, final word is **wâ’tan
- Option 2: Long high vowels (**ī, **ū) split into (**ya, **wa) > (*yo, *wo) when word-initial. Final word is **ʔû’tan.
2. *péh₂wr̥
This is the other recon!PIE word for fire, which was much more common than *h₁n̥gʷnís and represented fire as a substance (-r̥ on the end is a sure sign of a neuter noun)2.1 p
*p is **p, no issues here.Word Progress: p-
2.2 éh₂
At last, my nemesis reveals itself.*h₂ is the reconstructioneer’s magic bullet. You can drop it in anywhere to explain anything: it can turn *e into *a, it can turn short vowels long, it can aspirate voiceless stops, it explains ablaut paradigms; it can be a consonant, it can be a vowel, it can be a consonant that acts like a vowel or a vowel that acts like a consonant. Depending on who you ask it, it can be reconstructed as *q, *qq, *x, *χ, *ħ, *ɦ, *h, *ʕ or *a. I am certain there’s a paper out there claiming with utmost sincerity that it can turn lead into gold and resurrect the dead.
Kümmel (2022) once again saves the day with what I think is the most convincing argument. Short version:
- The *h₂ & *h₃ laryngeals inherited into the Anatolian languages were a fortis-lenis pair (likely voiced/unvoiced) of uvular consonants - probably fricatives, but stops is possible in the extremely early stages.
- This applies only to Anatolian languages and Homsar Hol, since it’s clear that they work differently in Strongmadian and Strongsadian PIE.
- I am operating under the assumption that the *H series changed so radically between Homsar Hol and later Strongmadian and Strongsadian PIE that they were essentially independent sound systems.
- If there’s attestation in an Anatolian language (ie, if ḫ shows up), there’s definitely a laryngeal there (either *h₂ or *h₃, they used the same symbol for both)
- If an Anatolian cognate shows no sign of *h₂ but it’s reconstructed with one anyway, the reconstruction is incorrect and there was just normal **a.
- If there’s no Anatolian cognate at all, we go by secondary evidence and vibes.
- Voiceless aspirates in Indic languages are decent evidence of a laryngeal being there, though I am less certain on what kind.
- Long vowels on their own are not sufficient evidence, since there are other ways for long vowels to form than just through laryngeal deletion + compensatory lengthening.
- Example: *muHs (“mouse”) is typically reconstructed with an unknown laryngeal to explain why the descendants have a long /u/ - I am going to apply Szemerényi's law and Occam’s Razor and say that either A) the original form was *mus-s and Sze’s law kicked in as normal or B) it was originally *mu-s and Sandall & Byrd (2014) are correct.
- Greek Triple Reflex is insufficient justification on its own - if there’s no other evidence, I’m saying it was probably just a regular-degular vowel. Maybe if the vibes are good I’ll pick and choose.
- If the only justification for there being a laryngeal present is vowel coloration, it’s suspect.
- *h₃ had a backing effect, not a rounding effect; if it turned *e into *o because of labialization, then the *Kʷ series would have done it as well.
- If all other options are exhausted and there is no way to rule out a laryngeal or identify it, the dice decide. 1-2 are *h₁, 3-4 are *h₂ , 5-6 are *h₃.
Finally getting back to the word at had, this one comes with easy Anatolian attestation via Hittite paḫḫur and Luwian pāḫur. That double form is unique to *h₂, so I’m going to add a bog-standard unvoiced uvular fricative **χ. *e is **a, naturally.
(I could have *h₂ = **q, but I am saving **q for the time being.)
Word Progress: paχ-
2.3 wr̥
Here we get another heteroclitic stem; the *r̥ is **n, but we do have another variable with the vowel. This ending is attested in Hittite as -war, so I'm going to make things easy for myself and go with **-wan; since it's an unstressed vowel we get a nice clean shift of **-wan > **-war > *-wr, with Anatolian languages either inheriting **-war whole cloth, or just adding an /a/ back in.Final Word: paχwan
3. Dictionary entries
- ʔû’tan (IN): Water as a general substance, regardless of size, state, drinkability, or other factors. Generic, non-branded water.
- páχwan (IN): Fire; typically a controlled fire (campfire, cooking fire, fire for illumination, etc)
Since I started with words that were all a root + a suffix, I can add some derivational suffixes as well.
- ʔank’ʷaní = ʔank’ʷa + (a)n + í
- ʔû’tan = ʔû’t + an
- páχwan = páχ + u/w + an
Which, after tweaking the existing definitions, end up as
- -(a)n (SUF) - Generic inanimate / passive nominalizer; turns root verbs into nouns
- -i (SUF) - Forms nouns of animacy / activity
- -ani (SUF) - Forms animate / active (but not agent) nouns from verb roots
- -u/w (SUF) - Forms action nouns from verb roots
- -wan (SUF) - Forms object nouns from verb roots
Or to break it down:
- páχ- = "to warm"
- páχ-w- = "(the action of) warming"
- páχ-w-an = "warming thing"
4. Corrections and Revisions
Despite liking Glen Gordon's "labiovelars come from reduction of an older /u/" theory, it doesn't really jive with how I'm going to be handling vowels anymore: that step has been axed. Since only one word was effected, that barely means anything.
5. Conclusion
Well, I'm having fun. Don't know about all y'all, but I'm having fun. So much idle musing and scribbling on sticky notes is finally starting to become something coherent - honestly, I think it's the format that does it. Much easier to get my thoughts in order when I force myself to describe it to an audience step-by-step.It's honestly a really cool feeling seeing things start to snowball - with the right groundwork and basic rules laid down, the rest just kinda flows naturally.
I think I have one more of these in the tank for the time being, then doubtlessly more stuff later on down the line. Don't want to overdo it and scare everyone away.
