Evolving tense from a tenseless proto-language

Question

I'm trying to come up with a verb system as confusing as Georgian, where expressing TAM involves the combination of many affixes whose individual meaning has been lost to time.

I want the daughter language to be able to conjugate for a distinct 1) present, 2) future, 3) aorist (=perfective past), 4) imperfect (=imperfective past), and 5) perfect. Georgian can conjugate for all of those and more, and as far as I understand, Proto-Kartvelian is thought to have been tenseless, only marking aspect - so it's surely possible.

So, here's my starting point. Let's say that in the proto, verb stems were all default perfective. But taking PIE as an inspiration, let's say it could express perfective vs. imperfective vs. stative. Then you'd need a way to derive imperfective and stative forms. Let's say the stem is rendered as a noun/participle and then appended suffixed with the copula (-> stative) or "go" (-> imperfective).

Okay, so, 3 aspects... but 5 conjugations that somehow need to get squeezed out of it. Which conjugation does each stem turn into?

• The perfective stem could yield either the aorist, or the future,
• the imperfective stem could yield either the present, or the imperfect, and
• the stative stem could yield either the present, or the perfect

For at least two of the pairs here, there's two different things it needs to yield. What is supposed to cause them to evolve to be marked differently?

I've sort of given up on using the WLG for TAM ideas because it's so incomplete; The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World (Bybee, Perkins, and Pagliuca, 1994) is more comprehensive. What does it suggest? Present and imperfect can both derive from the imperfective, great, but how to differentiate them if I'm trying to derive both simultaneously? Future can evolve from "to be" - wait, I was already using that - or from "to go" - wait, I was already using that too!

The thing I keep coming back to is something like the PIE augment, which marks the past tense deriving from a word for "then; at that time". But that breaks the initial goal of the individual affixes' meaning being impenetrable! The augment you can definitely pick out and say "this particular part is what makes it past tense". That's just not how Georgian verbs work and I'm trying to pull off something like Georgian.

But I'm kind of out of ideas. How do languages starting only from aspect, manage to differentiate them into a variety of morphologized TAM conjugations?

Answer 1

Okay, so, 3 aspects... but 5 conjugations that somehow need to get squeezed out of it. Which conjugation does each stem turn into?

I think part of your issue is that your process isn't really how languages develop - they don't skip from neatly encoding a single TAM category to a messy fusional system or paradigm with gaps.

What happens is a long process of grammaticalisation. All languages can express all the meanings of tense/aspect/mood/evidentiality; what distinguishes them is which categories are grammaticalised, and how.

For example, English is a tense-prominent language, with all verbs being either PAST or NONPAST. We have one aspect affix (-ing), and a bunch of auxiliary verbs for modality, voice, and the future. But unlike modality, evidentiality is pretty much completely non-grammaticalised, you use true content words rather than function words.

So I'd suggest you try modelling the process of grammaticalisation:

1. First determine how people would express tense using content words.
2. Then reduce the morphology, and do some phonetic erosion. Turn words into clitics, and then into affixes.
3. As you're reducing your morphology, you'll see some sounds that are more likely to be dropped, like liquids and semivowels, whereas stops are more likely to be kept. (But of course other phonological changes might occur which mean the stops change or drop off too.) Or a dropped sound could result in an effect like palatalisation.
4. Perhaps one marker gets reduced to a single vowel, which itself could be lost depending on the vowels of the aspect affixes or roots. Now you have an explanation for why some stems don't distinguish between tense differences that other stems do.
5. Just because there is a gap doesn't mean that speakers won't want to be able to make that distinction some times. What will they do? They'll use content words! And the cycle begins again. For example, English lost its second person singular pronoun thou. But sometimes you need to be clear whether you're referring to one or many, so people would say things like "you all". In some English varieties that has been grammaticalised as y'all, though it's still considered informal. But who knows what will happen in the future!

For extra realism this cycle should be happening for each grammatical category at the same time, but in different stages. (Actually even for different forms within the one category.) If your tense forms are being cliticised, then maybe the modality forms are only at the auxiliary stage. Meanwhile people need/want to be able to express evidentiality or some different types of aspect that the grammar doesn't currently distinguish, so some semantic bleaching of their respective content words is happening...

Answer 2

The evolution from Pre-Proto-Indo-European (the language reconstructed by internal reconstruction from Proto-Indo-European, the language reconstructed by the comparative method from the Indo-European languages) to Old Irish is a good illustrative example (with illustration along the way from Greek and Latin) of one way to do this.

Pre-Proto-Indo-European has two conjugations, with each verb belonging to just one of the two classes: eventive (which has both active and mediopassive endings) and stative (which has no voice distinctions).

Early on, the eventive verbs develop a set of imperative endings through unclear means, but they are clearly closely related to the previous "indicative" endings.

At some point, lexical aspect (aktionsart) develops among the eventive verb, with some verbs being felt to be imperfective, and others perfective.

The imperfective eventive verbs then start being marked with the hic-et-nunc "here and now" particle (-i in the active and -r in the mediopassive) when describing present tense events.

We have now arrived at the situation we see in Proto-Indo-European. Imperfective verbs have two sets of endings, a "primary" set for present tense events, and a "secondary" set shared with perfective verbs for past tense events.

Now we start to interpret the previously lexical aspect distinction as inflectional with what was previously various methods of deriving verbs of one lexical aspect from another as ways of inflecting a single verb.

The imperfective verb gives the present and imperfect "tense" (both with "present" or imperfective aspect), the perfective verb gives the aorist, and the old stative is reinterpreted as a perfect.

At this point, other derived verbs become grammaticalised giving subjunctive, optative, and future "tenses". The original sense of the subjunctive and optative is unclear, and these were likely grammaticalised at a very early stage (at the latest only shortly after Proto-Indo-European itself), but the future is derived from what was previously a means of forming desiderative verbs (i.e. verbs with the sense "to want to").

Now we have the following:

• Present (active & mediopassive voices; indicative, subjunctive, optative, and imperative moods)
• Future (active & mediopassive voices; indicative, subjunctive, optative, and imperative moods)
• Imperfect (active & mediopassive voices)
• Aorist (active & mediopassive voices; indicative, subjunctive, optative, and imperative moods)
• Perfect (indicative, subjunctive, optative, and imperative moods)

By applying aorist endings to the perfect stem, we can then form a pluperfect (as in Greek & Latin). By applying future endings to it we can form a future perfect (as in Latin), and by applying imperfect endings to the future stem we can form a conditional (as in Old Irish).

Around this stage, Greek produces a genuine passive (as opposed to mediopassives) in certain tenses with a new suffix, possibly derived from a reflex of *dʰeh₁- "to do". Greek also introduces the augment from an earlier particle, explicitly marking past tense verbs as such (essentially the opposite of the earlier hic-et-nunc particle).

Meanwhile Italic & Celtic reinterpret the mediopassive as a true passive (with the deponents left behind as a relic of the earlier sense). Italic then replaces several of these forms with new ones formed by suffixing the stem with a reflex of *bʰuH- "to become".

Italic & Celtic also then merge the perfect and aorist, with some verbs (all of them in Italic) taking a new set of endings formed from combining both sets of endings.

Then, to get to Old Irish it's a process of accreting as many clitics as you can:

• Conjunction
• Relative Particle
• Negative Particles
• Object Pronouns
• Perspectivity/Augmentation (from a prepositional preverb but now giving either a retrospective/perfect or potential sense depending on the tense of the verb)
• Lexical preverbs (often without well defined semantics themselves)
• Subject pronoun (the nota augens)

At this point you have a verbal complex containing up to 12 different elements which can occur in many (but not all) combinations with many elements having little in the way of obvious semantics of their own, but only really making sense when the entire complex is considered as one.

It starts looking a lot closer to Georgian than the more typically Indo-European Greek and Latin seen earlier did, and much more so than the Pre-Proto-Indo-European we started with. In fact, David Stifter, a specialist in Old Irish has gone so far as to describe it as polysynthetic (although in doing so he does use a broader definition than is typical).

I'd recommend reading his grammar of Old Irish (or his chapter in the Celtic Languages volume from the Cambridge Languages Survey) for a fuller account of Old Irish and its derivation from earlier Proto-Celtic, as well as Willi's Origins of the Greek Verb for an account of the development from Pre-Proto-Indo-European up to the typical Indo-European level of Greek and Sanskrit (but also Proto-Celtic).

---

The key point here is that you've got a lot of time. You don't need to derive everything in one go, but can instead do so in many stages, and even use essentially the same process multiple times (note the use in Greek of both the hic-et-nunc particle and the augment), especially if the sense of the earlier process has been lost.