In a series of lectures published in the comments section to the previous 2 posts, I discovered and detailed the Yeniseian origins of the Japonic language family. That's not to say that Japanese is a Yeniseian language, but that the Wa people used to speak a Yeniseian language, and they carried over many of their lexical items into the Altaic language that they shifted to after Yeniseian -- probably a Para-Mongolic language -- and have retained them even into the Japonic stage of their history, after incorporating Emishi / Ainu speakers along the way.
But the numerous examples I gave are mostly nouns and adjectives, and some highly important closed-class items like pronouns, kin terms, and particles / affixes. There are very few verbs in the examples -- until now!
The reason I couldn't draw Yeniseian origins for Japanese verbs is that I was trying to capture too many segments from the Japonic form, but it turns out that the final 2 segments in Japonic verbs -- Cu (some consonant, and "u") -- are just verb markers, not part of the semantic core of the word. I knew that already about the final "u" for Japanese verbs, but I did not appreciate the fact that the final consonant is not semantically crucial, it's only there for phonotactic reasons.
And that reason I didn't know that, is that nobody else knows that! So I couldn't just read about it in a Wikipedia article, Wiktionary entry, or even scholarly article, unless it's an obscure / forgotten one, entirely in Japanese, from the 15th C...
So in the interest of not only revealing the nature of Japanese verbs, but also to help connect them back to earlier Yeniseian forms that have been carried over after language shift, I'm going to start a new post on this topic.
The source of Proto-Japonic verbs is Wiktionary's list. See also some helpful intros from Wikipedia on the topic of Japanese conjugation and Japanese five-way vs. one-way verbs ("godan" vs. "ichidan").
First, to briefly summarize what *is* already known about Japanese verbs. They characteristically end in "u", not some other vowel, and not a consonant (coda consonants are banned in P-J). Other parts of speech are not this regularized in their ending, Japonic really wanted to clearly mark its verbs.
Before this final "u" is a consonant. Verbs fall into 2 classes. By far the most common are those where this consonant is considered the final segment of the stem of the verb. Less common are those where this consonant is "r", but the verb stem is considered to end in the vowel just before that, always followed by "ru". These classes are called consonant-stem verbs vs. vowel-stem verbs.
Example of a consonant-stem verb -- "kaku" ("to write"), has the stem "kak-".
Example of a vowel-stem verb -- "miru" ("to see"), has the stem "mi-".
The consonant-stem verbs are also known as "godan" or 5-way verbs, since the vowel that follows the stem comes in all 5 different Modern Japanese vowels, depending on the tense, aspect, etc. that the verb is being used it. So "kak-" is followed by "a", "i", "u", "e", and "o", before other suffixes and particles are added on, to indicate whether it's a negative, a command, the infinitive, the past, and so on.
The vowel-stem verbs are also known as "ichidan" or 1-way verbs, since their stems end in a vowel, and this is the one and only vowel in their conjugation pattern (before the negative, command, infinitive, etc. suffixes are added onto them).
Since the vowel-stem verbs have stems that end in a vowel, and the unconjugated verb must end in a vowel as well, and Japonic phonotactics prefer syllables that are "CV", a consonant must come between the end of the stem and the final "u". In every single case, this consonant is "r" -- showing that it is a dummy consonant, not semantically crucial to the particular verb it occurs in. For example, "miru" and "tomeru", whose stems are "mi-" and "tome-".
Now, for the part that is not understood, or if it was once understood, has not been transmitted to the present. This preference for verbs ending in "ru" is not unique to the vowel-stem verbs, where it is obligatory -- the most common final consonant for the consonant-stem verbs is "r", which means they also end in "ru"!
Can any ol' consonant be the final segment of a stem for the consonant-stem verbs? I skimmed through the list of P-J verbs and could immediately see that "r" was the most common, followed by "k", with others rare, and some non-existent. How is this not widely known??? There's a massive preference for and against certain segments as the final consonant of consonant-stem verbs.
I present these in the following list of 61 consonant-stem verbs and 16 vowel-stem verbs, which are the entire list of P-J verbs at Wiktionary. First they're split into the consonant (godan) vs. vowel (ichidan) stem verbs. And within them, they're ordered by the frequency of the final segment, from common to absent. Wiktionary lists "woi-" as "wo-", but it has an "i" in some forms, so it's like the others ending in a vowel sequence of "Vi". The vowel "ə" is AKA "o2".
Godan
R, 21
ar-
asar-
kir-
kukOr-
kur-
mapar-
nar-
nenpur-
ninkir-
panpakar-
pikar-
sir-
sur-
ter-
tukur-
tur-
ur-
watar-
wəntər-
yar-
ər-
K, 18
arik-
ik-
isonk-
k-
kak-
kik-
mak-
muk-
nonk-
sik-
suk-
tonk-
tontok-
uk-
unkok-
yak-
ək-
əyənk-
P, 9
ap-
asump-
ip-
kup-
op- ~ əp-
sinənp-
turunp-
tənp-
yukəp-
T, 5
kat-
mat-
mət-
ut-
ət-
S, 4
kərəs-
s-
əs-
ətəs-
M, 2
nəm-
yOnkam-
N, 2
in-
sin-
Y, 0
W, 0
Ichidan
A, 12
pukor-a-
wasur-a-
ank-a-
mak-a-
int-a-
nant-a-
nis-a-
wosam-a-
koy-a-
moy-a-
kuw-a-
uw-a-
I, 4
ai-
mi-
poi-
woi-
E, 0
O, 0
U, 0
Again, all of the vowel-stem verbs end in "ru" in the uninflected form. But the most common final segment for consonant-stem verbs is "r", yielding the ending of "ru" in their uninflected forms as well! This preference for "r" dwarfs nearly all other possible choices for consonants. In a close 2nd place is "k". In a distant 3rd place is "p", and even further behind are "t" and "s", with "m" and "n" rounding out the list of those that are actually present but rare, while "y" and "w" do not appear at all.
So, far from the final consonant of consonant-stem verbs being open to any ol' choice, there is clearly a strong preference for "r" and "k" and against most of the others. Next we investigate why this is, and draw conclusions for the study of etymologies -- very important to know if the final consonant in consonant-stem verbs is semantically crucial or vacuous! Turns out, it's vacuous, and that was what prevented me from discovering Yeniseian origins to Japonic verbs. Thankfully, that can now be corrected!
* * *
So why is "r" the default consonant for the final syllable -- "ru" -- of the uninflected form of Japonic verbs? When the Wa people were shifting from Yeniseian to Altaic / Japonic languages, their morphology changed from polysynthetic to agglutinative, i.e. where you stick or "glue" basic building blocks together in a long chain.
There are only so many building blocks, which introduces the homophone problem -- how can you tell that some suffix is a verb ending vs. a particle connecting two words vs. a noun ending vs. a prefix vs. anything else that a building block could be?
Polysynthetic languages are fusional, where all these changes to a stem are not building blocks concatenated together, but "bound" morphemes that can only appear in certain contexts and not on their own. English has very little of this left, although it is still fusional. A better example are the Romance languages that we studied in school.
In Spanish, the verb "to sing" is "cantar", and the stem is "cant-". Unlike Japanese, though, this stem does not merely receive a series of building blocks stuck on the end, one block for each bit of information added. Rather, the entire rest of the meaning comes from just one ending, which therefore come in dozens of forms.
One of these "bound" endings is "-o", yielding "canto", "I sing". This is the present tense, first person, singular number, indicative mood, simple aspect -- all of those bits of additional information, "fused" into a single morpheme that cannot appear on its own, but only on a verb stem. If you want "We sing", the ending is "-amos", yielding "cantamos". There is no clear relation between "-o" and "-amos" -- opacity is a feature of fusional morphology, whereas agglutinative morphology is transparent, since each bit of meaning has its own building block. That's not even to mention the dozens of other endings that "cant-" can receive, to fill out all those differences in tense, person, number, mood, and aspect!
Well, although agglutinative morphology is far more transparent than the opacity of fusional morphology, that comes at a trade-off with the homophone problem. In Spanish, there are almost no homophone problems across the dozens of endings that "cant-" can take, and none of them are homophones with other words throughout the language -- "amos", "aste", "arian", etc., are not free-standing words that could be confused with these fusional verb endings.
This makes the detection of word boundaries crystal-clear for fusional languages -- "Oh, I just heard '-aste', which can only be a verb ending, so that's the end of a word. And it followed 'cant-', which is a verb stem, so that must be the beginning of the word."
Agglutinative languages like Japanese have a huge potential homophone problem, since building blocks can hypothetically be combined in any which way, so how do you know which meaning is intended? Is "mi" the stem of a vowel-stem verb, is "mi" a noun", is "mi" a particle, is "mi" a suffix, or prefix, or pronoun? It could be any of those things, and in fact in Japanese it *is* several of those things.
Agglutinative languages also have the problem of detecting word boundaries, since the building blocks can appear in initial, medial, and final position, hypothetically. They aren't like "aron" in Spanish, which is a fusional verb ending that cannot appear as a prefix or as a stem of a noun or verb. Unlike hearing "aron" in Spanish and immediately knowing it's the end of a word, hearing "mi" in Japanese gives you zero information about whether it's the beginning, middle, or end of a word, since that building block can appear in all places!
Therefore, agglutinative languages have to try to impose a set of rules about which building blocks can appear at the beginning, middle, or end, and for verbs vs. nouns vs. adjectives. It mitigates some of the confusion about homophones and word boundaries, but doesn't entirely solve it.
How can they do this, if their whole morphology is about gluing together any ol' string of building blocks? Well, that's just the semantic side, where "mi" could refer to various things.
* * *
Enter, phonotactics! That is, the rules or constraints on what sounds, sound sequences, sound structures, etc., words can take in the language.
Getting right to the point, Japonic phonotactics prohibit "r" in initial position. It only occurs initially in words borrowed from Chinese, English, or other non-Japonic language. So it works perfectly to indicate medial or final position, eliminating initial position. When you hear "ru", you know it's the middle or end of a word. Again, doesn't totally solve the problem, but it helps.
That also means that "rV" cannot be a particle, since that would mean "r" in word-initial position. Japanese has tons of connecting particles, but none of them begin with "r". So when you hear "ru", you know it's not a particle, not a prefix or beginning of a stem, so it's either a suffix or a later part of a stem.
Why did the vowel have to be "u"? To avoid homophone problems, it seems: "ra" was already taken as a pluralizing noun suffix, "re" was already taken as a nominalizing suffix, and "ri" for and adverbial suffix. "Ro" seems to have been dispreferred on phonotactic grounds, it's by far the least common "r" syllable in OJ. So that leaves "ru" (3rd-most common "r" syllable in OJ) as the best choice for default verb ending.
From "ru", the default vowel ending for verbs became "u", even for verbs whose final syllable has some other consonant, like "kak-". In OJ, the most common syllable for a consonant does not always end in "u", e.g. "p" and "t". So it's not consonant-by-consonant phonotactics that determines the final vowel of the verb -- it's a default vowel that was originally established for "r", and it was copied for all other verbs in order to standardize the final segment of verbs -- "u".
Why didn't nouns and adjectives receive such rigidly standardized endings? Cuz they don't add on as many building blocks as verbs do in Japanese. Nouns are not inflected for number (mostly) or gender. They only take case suffixes. Adjectives do not have to agree with the nouns they modify, for number, gender, or case. Since verbs are going to take on all sorts of suffixes, it's more necessary to know where the end of the stem is, that all these blocks are being stacked on top of.
Which consonants other than "r" are allowed for consonant-stem verbs? This is both phonotactics and avoiding homophones, especially with suffixes and particles. And this is the only place where the 5-way conjugation of verbs is relevant -- since many vowels are going to follow the final consonant of the stem, this presents phonotactic and homophone problems for that consonant plus any of the vowels. That's a lot of problems to avoid, and only a few consonants can do so.
Starting with the consonants that are absent from consonant-stem verbs -- the approximants "y" and "w". From the P-J stage, the syllables "wu" and "yi" were banned in order to dissimilate approximants from their vocalic counterparts. Since both "i" and "u" would appear after "y" and "w" in the 5-way conjugation pattern, this would result in the syllables "yi" and "wu" appearing, and that's illegal. So, no consonant-stem verbs can end in "y" or "w". In fact, if they were intended to be the final consonant of a consonant-stem verb, they were given an invariant "a" afterward, and put into the 1-way / vowel-stem class instead. We'll get to those later.
The nasals "m" and "n" would not result in illegal syllables when conjugated, but they would involve too big of a homophone problem. "N" plus any vowel is already a highly common suffix or particle, especially the possessive particles "no" / "na" and the locative particle "ni", as well as the diminuitive noun suffix "ne". "M" would yield a bunch of homophones for common concepts like "seeing", "body", and "three", as well as the non-unique topic marker particle "mo", and "mu" was already a verb suffix (volitional, etc.).
Originally, "o" was not one of the vowels following the last consonant of the stem, so these homophones aren't so crucial, but they indicate the homophone problem nonetheless, which "r" does not have.
The problem with "s" is that "su" is already a verb unto itself, one of the very few that is monosyllabic, and a very ancient and common verb -- "to become" and "to do". Hearing this at the end of some other verb whose consonant stem ended in "s" would make it sound like the final "su" was this standalone verb being used as the 2nd element of a compound verb, or as an auxilliary verb for the part of the stem preceding "s". WAY too confusing. Evidently so confusing that over time, Japanese standardized "su" into "suru" with the most preferred verb ending, "ru", just to prevent any confusion.
I'll get to the handful of exceptions to these rare consonant verbs, in a bit, but briefly they seem to be compounds where the rare consonant was not originally the end of a stem, but the start of the 2nd element of a compound. So they're not such exceptions anyway.
"T" would yield "tu", which was already a common particle (genitive), as well as a verb suffix (completion), and a counter suffix for nouns. "To" was already taken by connecting word "and".
"P" would yield "pu", which was already a verb suffix (ongoing, repeating). Worst of all, "pa" was the topic marking particle.
That leaves only "k" as able to compete with "r". The only particle it would yield a homophone of is "ka", but that's a sentence-final particle, not one that connects words, so it's no problem for the end of a verb. However, "ka" was an adjective suffix in OJ. "Ki" as a certainty suffix for verbs came later during Classical Japanese, not in P-J or OJ. And "ku" was a not-so-common nominal suffix in OJ, but was also a standalone common verb in P-J, "to come", later standardized into "kuru". Same problem with "ku" as with "su".
"K" is not as flawless as "r", but a quantum leap above the other consonants, so "k" is more for the overflow, after a stem ending in "r" has already been taken -- the rest of the stem is fine, just alter "r" to "k", and presto, a new verb with no homophone problems.
E.g., since "sir-" and "sur-" already exist, "sik-" and "suk-" can be used instead. The "su" in "sur-" has to do with "rubbing", while the "su" in "suk-" has to do with "liking, loving". The "si" in "sir-" has to do with "knowing", while the "si" in "sik-" has to do with "spreading out". Rather than conflate the unrelated "su"s and "si"s into homophonous "suru"s and "siru"s, use the default "r" for one and the over-flow "k" for the other. Bingo.
* * *
Before exploring this matter of the less common consonants being an over-flow when the more common consonants were already taken, let's address the matter of how many of these verbs are morphologically atomic vs. complex.
Well, any time there's a nasal + obstruent, it just means there's a morpheme boundary, and this will later get rendaku-fied in OJ. Rarely is there any evidence for the nasal.
Also, Japanese morphemes in general and verbs especially want to be bisyllabic / bimoraic, or perhaps monosyllabic / monomoraic, not more than 2 syllables / moras. Atomic verbs could be hiding inside verbs of 3+ syllables, though, while not being attested on their own. So I'm inclined to only count 4 verbs under stem ending in "p" -- those whose infinitive is 1 or 2 moras ("apu", "ipu", "əpu", "kupu"). The others are complex, and the "p" is really an initial or medial consonant to a later element of a compound, not the final consonant of an atomic stem.
Then restricting the analysis to atomic verb stems, the idea that less common consonants are an over-flow for already taken more-common consonants, predicts that the less common verbs should have more common counterparts -- the ones they are trying to avoid homophony with, by using a less common over-flow consonant. This excludes monosyllabic verbs, since their sole consonant is not a dummy consonant, and you can't tell whether or not there's a competing form -- what precedes this consonant is nothing, not a sequence of vowels and consonants.
Let's see...
Of 10 "k" verbs, 6 have higher-ranking counterparts ("kik-", "sik-", "suk-", "uk-", "yak-", "ək-"). Also, "kak-" may have a counterpart hiding inside of either "panpakar-" or "pikar-".
Of 4 "p" verbs, all have higher-ranking counterparts, and 1 has both of them! ("əp-")
Of 5 "t" verbs, 4 have higher-ranking counterparts, some with more than one.
Of 1 "s" verb, it indeed has every higher-ranking counterpart ("əs-").
Of 1 "m" verb, it does not have any higher-ranking counterparts. This is "nəm-" meaning "to drink", and perhaps it was allowed its very rare final nasal in the stem, as part of onomatopoeia, where eating and drinking tend to have nasal consonants -- both consonants are nasal here, in fact, just like "nom-nom" in English.
Of 2 "n" verbs, both have multiple higher-ranking counterparts.
So yes, the less common consonants are over-flow choices, for when the better choices are already taken, and homophony must be avoided.
And in all cases, to reiterate the main point, these final consonants in the stem are NOT semantically crucial, they are vacuous and only chosen on the basis of phonotactics and avoiding homophony. Important to bear in mind when trying to find earlier ancestors of these verbs...
* * *
What about the 1-way / vowel-stem verbs? Here again we see them acting as either over-flow for already taken forms, or to avoid phonotactic prohibitions.
By far the most common vowel-stem verbs end in "a", and they are transcribed as "-a-", to hint that the "a" is just a dummy consonant. If it weren't, then the other consonants could show up as well -- but they don't. Only "i" shows up as well, and it's rare, and most of those involve vowel sequences, not consonant + "i".
This "a" seems to have been "hard-coded" to prevent the 5-way (or earlier, 4-way) vowel pattern from spawning. Every conjugated form will have "a".
Well, that avoids the ban on approximants "y" and "w" -- both of those are fine followed by "a". Indeed, unlike the consonant-stem class, where "y" and "w" are totally absent, they make up 4 of the 12 entries for vowel-stem ending in "a", just before the dummy "a". Since they couldn't end in "y" or "w" when illegal vowels could result under the 5-way / 4-way pattern, just hard-code the following vowel to be "a", and problem solved!
I'm inclined to think that means these approximants *are* semantically meaningful, since they went through the extreme measure of hard-coding a dummy vowel afterward, to allow the approximant to be preserved. If it were just an over-flow choice of consonant, there are better choices -- the nasals, the sibilant, or "t" or "p", which were generally not chosen. They really wanted the "y" and "w" to stay in these cases.
That goes for the other 2nd-to-last consonants in the vowel-stem class -- which means there's a sick inversion going on! The final consonants in the consonant-stem class are vacuous, while the latest-occurring consonants in the vowel-stem class are meaningful! The 5-way class is really "consonant stem for phonetics, but the preceding vowel for semantics", and the 1-way class is really "vowel stem for phonetics, but the preceding consonant for semantics". Neat.
Another 4 of the 12 vowel-stem verbs are from consonants that are rare in the consonant-stem class (2 "t"s, 1 "s", 1 "m").
Only 2 of 12 vowel-stem verbs are from the super-common "r" in the consonant-stem class, and another mere 2 of 12 from the super-common "k" in the consonant-stem class. In the vowel-stem class, "r" and "k" are not so dominant at all -- combined, they are as common as "y" and "w", which are totally absent in the consonant-stem class!
The over-flow pattern shows up here again. The atomic "kor" in "pukor-a-" could be hiding in "kukOr-" from the consonant-stem class. The atomic "sur" in "wasur-a-" already appears in "sur-". The atomic "mak" in "mak-a-" already appears in "mak-".
As for the vowel-stem verbs ending in "i", 3 of them involve vowel sequences, which are generally a feature of P-J nouns, not verbs. E.g., "tai" = "hand", whose forms have an "a" sometimes and an "e" other times. And the 1 vowel-stem verb that does not have a vowel sequence, "mi-", is cognate with a P-J noun that does have a vowel sequence, "mai" = "eye". Perhaps "miru" began as "mairu", which would make it fit better with the other "i" vowel-stem verbs, and for whatever reason the "a" was deleted.
None of these 4 would have a homophone in the consonant-stem class if they had an "r" hard-coded after their vowel stem, e.g. "ai-" could be altered to "air-" and not compete with an existing consonant-stem verb of that form. Ditto for the others.
Here it seems more like phonotactics play a role -- atomic stems / infinitive verbs cannot be more than 2 moras. "Airu", "poiru", and "woiru" all have 3 moras, so they can't be hard-coded into "r"-ending consonant-stem verbs. In fact, their descendant or variant forms will not have 3 moras either -- "eru", "hiru", and "oru / iru".
Again that suggests that "mi-" used to be "mai". Otherwise it's unusual, since its infinitive, "miru", has only 2 moras as is desired -- that should result in it being treated as a consonant-stem verb ending in "r", "mir-". But if it was originally "mai-", then "mairu" would have 3 moras, break the rule, and get lumped into the "i"-ending vowel-stem verbs, along with "ai-", "poi-", and "woi-".
* * *
I'll get to etymologies in a separate post, or perhaps in the comment section to this post. The most important thing before that is laying out this foundation, about what segments are semantically meaningful vs. vacuous. It turns out, a lot of those consonants are meaningless, so they don't need to be captured in an etymology, only the first "CV" syllable.
Sadly, that makes the etymologies less convincing, since a 3-segment etymology is more convincing than a 2-segment one. But that's just the way the cookie crumbles with Japonic historical linguistics...
Aside from showing a number of Yeniseian origins for Japonic verbs, I'll also draw a parallel between the phonotactic structure of their verbs, and how it parallels Japonic.
Briefly, none of the Proto-Yeniseian verbs -- hardly any words at all -- begin with a nasal. Yeniseian is polysynthetic and prefixing, while Japonic is agglutinative and suffixing -- so it's the same process. Yeniseian verbs avoid initial nasals, Japonic verbs avoid final nasals (in the stem). Japonic is rife with suffixes and particles beginning with nasals, while Yeniseian has several prefixes for verbs that begin with a nasal.
So they share this avoidance of nasals in the part of the verb that gets the most modification during conjugation. And since those particles are carry-overs from Yeniseian to Japonic, as I showed earlier, this is not a coincidence across language families. Yeniseian verbs avoid initial nasals in their stem, and Japonic as the end of their stem, for cognate reasons.
I've also discovered another sound correspondence between Proto-Yeniseian and Proto-Japonic, but I'll get to that one later as well. But briefly, P-Y "tɬ" corresponds to P-J "p", at least in initial position. Not very phonetically expected or motivated -- why not alter "tɬ" to simply "t" or "s" or even "r"? -- but it is what it is.
Well, let me end with at least one etymology! P-Y "cej" means "to rip". P-J lacks "c", but can shift its location to velar "k", as I showed in previous examples. The coda consonant is not allowed, and turns into its vocalic counterpart "i". Since vowel sequences are not preferred, and the 2nd usually takes over the 1st, that gives "ki" as the stem.
Whaddaya know? "Kir-" in P-J means "to cut"! Infinitive: "kiru". And that final "r" in the stem is semantically vacuous, it's only there to make verbs adhere to the standard of "ru" being the final syllable. That means only "ki-" is meaningful in Japonic -- and perfectly matches the expected form that would derive from P-Y "cej". QED!