Hierarchy of morphology, auxiliaries, and suppletion of verbal accidents?

Question

I would like to make a hierarchy of verbal accidents that would have the following features.

1. For any two accidents in the hierarchy, if a language marks only one of them by lexical suppletion, it is significantly likelier to be the one earlier in the hierarchy.

2. For any two accidents in the hierarchy, if a language marks only one of them by an auxiliary word, it is significantly likelier to be the one later in the hierarchy.

3. For any three accidents in the hierarchy, if a language marks both the earliest one and the latest one by morphology, it is significantly likely to also mark the intermediate one by morphology.

4. For any two accidents in the hierarchy which a language marks both of by morphology,

   • if only one of them is by stem-alteration it is probably the earlier one in the hierarchy;
   • if only one of them is by affixation it is probably the later one in the hierarchy.

5. For any two accidents in the hierarchy which a language marks both of by affixes, if both affixes are on the same side of the root, then it is significantly likely that the one earlier in the hierarchy is marked closer to the root.

I am aware that it might not be possible to fully satisfy all of these desiderata with just one hierarchy; and that perhaps the "hierarchy" would be only partially-ordered rather than "linearly" ordered. That's OK; I just want to get as close as I can.

To start with I'm going to consider just the following accidents:

1. Aspect
2. Modality, mode, and/or mood
3. Polarity
4. Tense
5. Voice

Pluractionality, evidentiality, mirativity, validationality, etc., I might work in later. Also, "modality/mode/mood" might be two (or more?) accidents instead of one; so might voice. And possibly not every language would have every accident; for that matter maybe some languages don't have a "verbs" "part-of-speech", though I gather most have something close enough.

My original hypothesis, which might be complete hogwash, or might be only partially true, is that the hierarchy is more-or-less the following:

Aspect > Voice > Tense > Polarity > Modality/mode/mood

And here's why;

Of those five accidents, the one that has the most influence on the actual meaning of the verb is Aspect.

Of those five accidents, modality/mode/mood tends to be about what the speaker expects the addressee to do with the clause, or how the clause fits into the discourse, or how the speaker feels about the clause, etc.; and tends to apply to the whole clause rather than just the verb. It doesn't really change the meaning of the verb itself.

And tense is intermediate between aspect and m/m/m.

Up 'til that, I'm pretty sure my reasons are opinions shared by most professional linguists; the thing I'm not sure of is that it has any significant influence on most languages' morphology.

As for the rest of the hierarchy; I guess that Voice is between Aspect and Tense.
I guess that Polarity is between Tense and Modality/Mode/Mood.

---

Does anyone here know of anyone who actually knows anything about these questions? Can anyone here point me toward a resource (preferably a URL) that answers some of them? Would anyone here care to speculate about any of it that no resource (as far as you know) definitively answers?

Thanks.

Answer 1

Take a look at the first eleven tables in A Siberian Link with Na-Dene Languages, which show the order of constituents in a verbal complex.

Included are various Dene-Yeniseian languages as well as Sumerian, Burushaski, Abkhaz, and Ha (a Bantu language).

Unfortunately, it's very hard to test your hypothetical order against them, since TAM (tense/aspect/mood markers) tend to be clumped together--and sometimes repeated throughout the complex!--whereas negation is only marked in Burushaski, and it's not on the same side as the other relevant markers; and voice doesn't seem to be marked in any of them.

It seems to suggest that perhaps causative markers (and possibly stative resultative markers, although the evidence is really from one family) are often very close to the root.

Answer 2

Guglielmo Cinque wrote a book called Adverbs and Functional Heads: A cross-linguistic perspective wherein he (ambitiously) tries to construct a cross-linguistic universal hierarchy of functional categories. He relies on adverb placement, for example facts like the following:

(1)  John luckily mistakenly kissed Mary
(2) *John mistakenly luckily kissed Mary
(3)  John mistakenly tenderly kissed Mary
(4) *John tenderly mistakenly kissed Mary

The contrast between (1) and (2) shows that adverbs related to the speaker's perception of the event (luckily) must precede adverbs related to the subject's state of mind (mistakenly). And subject-oriented adverbs precede adverbs describing the event itself (tenderly), as (3) and (4) show. He has many, many more categories of adverbs than just these, and by pairwise comparisons he puts them all into an ordered sequence. He then argues that each adverb is associated with a functional head, and that the heads are also so ordered.

Mark Baker proposed the mirror principle in this paper (JSTOR link). It says that, of two affixes, the one which is syntactically closer to the stem must be morphologically colser to the stem as well. So in a structure like the following, wherein Z is the root and X and Y are suffixes, we predict Z-Y-X and not Z-X-Y.

   XP
  /  \
 X   YP
    /  \
   Y    Z

Cinque's universal hierarchy plus the Mirror Principle plus some version of the Head Movement Constraint (no skipping positions as you move up the tree gathering affixes) gets you what you want. If some relatively high functional head is expressed as a verbal affix, the stem must have moved at least that far, so the lower head must also be affixed to the root, closer than the higher one.

I don't know Bybee's work and my recollection of the details of Cinque is hazy, but they seem to roughly align in the order you say.

Answer 3

I don't know of any existing research on precisely this question. However, there are some categories in WALS that may be of interest. You might look at:

Exponence of Selected Inflectional Formatives

Position of Tense-Aspect Affixes

Suppletion According to Tense and Aspect

Verbal Number and Suppletion

As far as I can tell, there is no data in WALS specifically regarding the order of affixes in verbal morphology. However, the chapters linked above present a wealth of data about the kinds of affixes used, together with links to the relevant papers, which should provide a sufficient base for doing this research on your own.

Answer 4

I am sure that the book Morphology: a study of the relation between meaning and form by Joan L. Bybee (1985) has something about the relative order of inflectional affixes.

I think it was:

valence > voice > aspect > tense > mood > person-and-number

which leaves out "polarity" but includes both "valence" and "person-and- number".

However as I recall Bybee uses two of the three terms "modality, mode, mood" with non-synonymous meanings; I can't remember exactly which one meant what.

Anyway that kind of says something about what the hierarchy I was looking for actually is (descriptive adequacy); I don't know how to achieve explanatory adequacy, though.

This post to CONLANG talks about which accidents are likelier shown by inflection, which are likelier shown by some other means, and which are likelier not shown.

This post says:

Bybee also investigates, for each pair of markings, which one tends to be closer to the stem in case they occur on the same side of the stem. She also investigates which markings tend to be fused with the stem. This involved investigating both which features tended to be marked by stem-changes instead of affixes, and which affixes tended to get changed depending upon which stem they were being applied to.

In every language derivational markers tend to be closer to the stem than inflectional markers; and whichever marker is closest to the stem is most likely to be fused with it. Within those restrictions, the hierarchy tended to be valence, voice, aspect, tense, mood, person-and-number.

Person-and-number agreement were often combined in one morpheme. Person-agreement is never derivational, it is only inflectional.