The Great Proto-Indo-European Thread

[Travis B.]

Umm Germanic nouns sometimes have umlaut for encoding plurality, not ablaut.

[Chagen]

Both processes involving the alteration of a core part of the root as opposed to suffixation. To an English speaker, the processes underlying both goose:geese and break:broke are the same thing, even if they come from different processes historically.

Anyway a better criticism of my hypothesis would have been the broken plurals of Arabic which I think are even more common than the non-broken plurals AND productive.

[Neek]

I think you mean "typically" instead of "invariably", as they are definitely not completely closed, not even in Western European languages. Pronouns can be borrowed (e.g. English 3rd person plural) or invented (many honourific pronouns are abbreviations of addressing formulas). Prepositions can be created from nouns directly, or shortned from compound PPs (e.g. Dutch "in de richting van" (in the direction of) ->"richting" (direction), starting to replace "naar").

Yes, I meant "typically." I guess I meant "invariably" more in the sense that there's a low probability of IE languages allowing new lexemes into the pronominal class, not that the class itself was completely immutable. English's third person plural is an interesting case, and one that I haven't forgotten about. It's still undergoing a decent respect of change, and for at least 700 years has also been used to refer to singular arguments in addition to plural arguments. Novel second person plurals are also of great interest, since they are showing that mutability of this class is waning in recent times.

It's true new nouns can borrowed when there's a word lacking, but typically they're derived from others (e.g. by compounding) when not borrowed? And depending on the type of derivational morphology a language has, forming new verbs is easy (e.g. English allowing any noun being "verbed" if there's a semantic need). Or where you talking about PIE only?

I was speaking of older than 1000 years ago. That's my fault for not specifying. The popularity of the weak verb classes, met with a large and rapid erosion of the morphology, allows English for greater flexibility in shifting nouns to verbs. But I am speaking without much data, and further research will give better insight if me talking out of my ass will hold water. It's a hunch that's worth investigating, and even if it's wrong, the conclusions drawn may still be helpful.

[Tropylium]

(…) there're only so many ways to form verbs from any other morphological category, and obvious structural limits (for instance, you cannot verbalize a pronoun or preposition or adverb in IE languages.)

On the other hand, nouns tend to be easier to nominalize. Just slap on noun endings, thematic endings, -t- or -n- participial forms (as well as others). Infinitives, too, are quite novel in each language in how they're formed. Even Ancient Greek can't agree on how to properly form an infinitive, suggesting that entry is pretty low. It was probably the same case in PIE, that nominalization is haphazard because nominalization is an easier process, and hence less resistant in keeping up with the ablaut.

There's also that loanwords tend to be nouns, which means that suffixes acquired in / extracted from loanwords also tend to be nominalizing; indeed, nominalizing denominal ones. This helps in its part for languages ending up with a wider palette of nominal suffixes than verbal ones, especially vague "diminutive" ones (a type of derivation that often seems to have no verbal counterpart at all). And the formation of new verbs will therefore again have to rely more on inherited morphophonology…

[Zaarin]

What is the earliest recorded IE language that lost vowel length? Gothic?

EDIT: Ah, so Gothic does have vowel length. First question still stands.

[KathTheDragon]

Do you mean that doesn't have vowel length at all, or lost the original PIE vowel length? If the latter, Hittite. All synchronic long vowels are later developments.

[Zaarin]

Do you mean that doesn't have vowel length at all, or lost the original PIE vowel length? If the latter, Hittite. All synchronic long vowels are later developments.

Lost the PIE vowel length. I'll check out Hittite.

[Zaarin]

Another question. As I look over pronominal paradigms of various Indo-European languages, it seems many of them have been quite...inventive when it comes to pronouns (and Fortson says as much). Any advice on developing pronouns for an a posteriori IE language? It seems much less straightforward than nominal declensions and verb conjugations (which is a quagmire I haven't even gotten into yet, actually).

[KathTheDragon]

What specifically are you referring to? There are lots of ways that the pronominal system suffered innovation in the various daughters.

[Zaarin]

Well, Fortson talks about how innovative the daughter languages were with the inherited PIE pronouns, but he's offering a descriptive grammar of PIE, not a guide to glossopoets trying to create a language from that grammar. In particular, I want the language to have a Western European feel and I've already borrowed a number of features (both phonological and grammatical) from the shared features of Italic and Celtic. I've been studying the Latin pronouns, but unfortunately the Celtic evidence to compare it to is rather...sparse--and the Germanic data isn't horribly helpful either. I suppose what I'm rather loquaciously asking is if there's any good overview on the specific ways in which Indo-European languages have innovated their pronominal systems. Fortson gives a list of descendant forms but he doesn't really go into them in detail.

[KathTheDragon]

So, the main thing seems to have been the introduction of a full system of cases (notably in the plural, which was heavily impoverished), and then the replacement of the genitive by a genitive adjective. That pretty much covers Latin and Germanic, I think.

[hwhatting]

In the 1st and 2nd persons, there has been a lot of analogical levelling between the various suppletive stems used in the paradigms for the same person; dual forms have gone over to the plural. For all persons, various deictic elements have been added to the stems or endings in many IE languages; various demonstrative pronouns have becom 3rd person pronouns (it's not even clear whether PIE had a regular 3rd person pronoun), quite often leading to suppletive paradigms (just look at English he, she, it, they which represent four different PIE pronoun stems).

[Zaarin]

So, the main thing seems to have been the introduction of a full system of cases (notably in the plural, which was heavily impoverished), and then the replacement of the genitive by a genitive adjective. That pretty much covers Latin and Germanic, I think.

In the 1st and 2nd persons, there has been a lot of analogical levelling between the various suppletive stems used in the paradigms for the same person; dual forms have gone over to the plural. For all persons, various deictic elements have been added to the stems or endings in many IE languages; various demonstrative pronouns have becom 3rd person pronouns (it's not even clear whether PIE had a regular 3rd person pronoun), quite often leading to suppletive paradigms (just look at English he, she, it, they which represent four different PIE pronoun stems).

Thanks, that makes sense.

[Šọ̈́gala]

Has anyone ever suggested that PIE had an affricate series? I mean, I believe Beckwith suggested something like that, which is where I got the idea, although I can't find the passage I'm thinking of. I seem to recall Beckwith cited the idea as originating with someone else. But ... other than that?

This occurred to me because apparently PIE had a range of dissimilation-like patterns affecting the consonants in roots, but the only assimilation-like pattern is that (what are traditionally reconstructed as) voiced aspirates cannot co-occur with voiceless stops. An obvious explanation is that at some point in pre- or early PIE, voicing assimilation affected voiceless stops within roots. However, in that case, it is mysterious why traditional plain voiced stops neither cause nor are affected by this assimilation.

As reconstructed, stops are the only phonemes where voicing is phonemic. So, one might start to wonder if traditional plain voiced stops (*d, *ǵ, *g, gʷ) were actually not stops at all, i.e. perhaps they were affricates (*t͡s, *kʲ͡ç, *k͡x, *k͡ʷxʷ). Being neutral for voicing, they were not affected by voicing assimilation, nor did they cause it. This would fit in with Beckwith's explanation for the infamous missing *b. There would originally have been a corresponding labial affricate, but it merged with *w in all dialects (p͡ɸ > ɸ > ʍ > w); this might have been motivated by the absence of a preexisting phoneme similar to /ɸ/~/f/ (this logic works if one or more of the laryngeals was a fricative similar to /x/).

More details that follow from these assumptions: taking the dentals as exemplary, suppose original *t *t͡s *d (rather than traditional *t *d *dʰ). In Late Common IE, all dialects have *t > *tʰ allophonically, and then, after voicing assimilation, *t͡s > *t (tenuis). Meanwhile, in the dialects ancestral to Indic, Hellenic, and Italic if not others, *d > *dʱ allophonically (perhaps motivated by voiceless aspirates that became voiced due to assimilation, e.g. assume a root *pewd > pʰewd > (assimilation) bʱewd > (analogy) bʱewdʱ). So, Late Common IE would have *tʰ *t *d~dʱ. Subsequently all dialects except Germanic (and, I guess other taihun dialects, e.g. Armenian?) voice the tenuis series to enhance contrast with the voiceless aspirates, *t > *d. In Indic, Hellenic, and Italic, *t͡s > *t > *d continues to contrast with the original voiced series because the latter has developed the murmur, which is now phonemic. In other non-taihun dialects, the original affricates have simply merged with the original voiced stops.

The point of this is to give a single explanation that explains the neutrality of the second series with regard to voicing assimilation and also the missing labial sound in that series.

[KathTheDragon]

It's a nice theory, but the diachronic somersaults you have to make are... implausible. It is honestly more likley that the *d series was an ejective or an implosive, or some such.

[jmcd]

Also, "stops are the only phonemes where voicing is phonemic" is not particularly remarkable crosslinguistically. About half the language surveyed by WALS with voicing only have voicing in the plosives.

[Tropylium]

What is the earliest recorded IE language that lost vowel length?

On that note, how many IE languages still have the original vowel length contrast? West Germanic (sort of, heavily modified) and Baltic do, but Romance lost it, Slavic lost it, Greek lost it, modern Indo-Aryan generally lost it…

[KathTheDragon]

I presume you mean by that that (most) originally long vowels are still long, and (most) originally short vowels are still short? I'm not overly familiar with Celtic, but might Irish qualify? Or has it too turned length into quality?

[Šọ̈́gala]

Also, "stops are the only phonemes where voicing is phonemic" is not particularly remarkable crosslinguistically. About half the language surveyed by WALS with voicing only have voicing in the plosives.

Right, that's a feature. If the goal is to explain why voicing assimilation doesn't affect this set of phonemes, then not being plosives is a good explanation because it's normal for non-plosives to be unmarked for voicing.

Incidentally, is voicing assimilation at a distance (e.g. two stops separated by a vowel) attested in other languages? Not that my speculations above hinge on this question. It's widely speculated that some kind of assimilation was happening between the *t and *dʰ series, with the *d series (as well as all non-plosives) unaffected.

[Cedh]

Incidentally, is voicing assimilation at a distance (e.g. two stops separated by a vowel) attested in other languages?

Some Berber languages have long-distance voicing assimilation of sibilants, e.g. in Tamajaq the causative prefix s- has the allomorph z- when the verb stem contains /z/:
s- + əχrək > səχrək
s- + əntəz > zəntəz

A more general type of long-distance voicing assimilation is attested in Ngizim (Chadic), Chaha (Semitic), and Zulu (Bantu), although there it applies (mostly?) within roots, where it is detectable as a harmony process only through comparison with related languages that have cognates with non-identical voicing.

Other laryngeal features (aspiration, glottalisation etc.) can of course also cause long-distance assimilation or dissimilation; Grassmann's Law is a well-known IE example of the latter.

[Tropylium]

Right, that's a feature. If the goal is to explain why voicing assimilation doesn't affect this set of phonemes, then not being plosives is a good explanation because it's normal for non-plosives to be unmarked for voicing.

Which is a good point, but also, it does not point to (af)fricatives specifically. Kummel's implosive theory arguably fits in better with this.

Other laryngeal features (aspiration, glottalisation etc.) (…)

Not "etc"; those are all there is. Phonetically you can draw finer-tuned distinctions like "tense voice" versus "creaky voice", but phonologically all languages make do with just ±voice, ±spread ("aspiration") and ±tense ("glottalization").

[Zju]

Other laryngeal features (aspiration, glottalisation etc.) (…)

Not "etc"; those are all there is. Phonetically you can draw finer-tuned distinctions like "tense voice" versus "creaky voice", but phonologically all languages make do with just ±voice, ±spread ("aspiration") and ±tense ("glottalization").

Isn't nasal spreading in the same category of distant assimilations though? There's rhinoglottophilia, so you could even go out a limb and call it glottalic.

[Zaarin]

So over the course of Indo-European history, athematic nouns and verbs were commonly analogized into thematic nouns and verbs, which happened at different rates in different language groups. So did it ever occur that in the course of such a shift that only part of a declension would shift? In my a posteriori Indo-European language, I've found I have a handful of irregular nouns that (through sound changes rather than analogy) have merged with the thematic declensions in the singular but remain athematic in the plural. I rather like the irregularity, but I'm wondering if there were similar cases in natural Indo-European languages.

[KathTheDragon]

In Hittite, the thematic and athematic declensions merged in a number of case forms, mostly through analogical extension of endings.

[Zaarin]

Cool, thanks.