I'm working on derivational affixes for my conlang. I want some of them be stress shifting. After looking at examples from natural languages, I have some ideas of how that would work. But what I can't tell is if there are any qualities that would predict whether an affix is likely to be stress shifting or not (in languages that include both).
Here are some things I'm wondering:
Are derivational affixes more likely to shift stress than inflectional ones?
Are their certain phonetic properties that are correlated with stress shifting affixes, as opposed to stress preserving ones?
Are multi-syllabic affixes more likely to shift stress than monosyllabic ones?
Is there anything about grammaticalization that predicts stress shifting (such as, an affix originally having been a word with a certain stress pattern)
Does the location of the affix (suffix / prefix) correlate with the likelihood to shift stress?
Basically, I want to know if there's any pattern in natural languages behind which affixes tend to be stress-shifting. I assume there aren't any, which would give naturalistic conlangers a lot of flexibility in choosing which affixes can shift stress. But I'm struggling to find any research that confirms this.
Would you happen to know? Any information would be much appreciated.
I don't think there is a clear pattern, but there are some very vague tendencies/hypotheses:
A relatively old idea is that affixes closer towards the root are more likely to undergo phonological processes, including stress shift, than affixes further away from the root. In some languages these 'outer affixes' are called clitics. The diachronic explanation is that inner affixes were grammaticalized earlier and had more time to undergo reduction, IIRC.
The second tendency is that prefixes are less integrated prosodically. So prefixes are less likely to trigger stress shift. Of course, the picture is skewed bevause there are far more right-aligned stress languages than left-aligned stress languages. Consider, however, the following Standard Indonesian (as spoken by Toba Batak speakers) data, where prefixes play a role in right-aligned stress. Stress shifts if a suffix is attached to a bisyllabic stem, but stays the same if a prefix is attach to a monosyllabic stem, thus creating an exception to penultimate stress.
[although Indo-European languages offer a lot of counterexamples in which prefixes have triggered a stress shift - at least in Celtic, Italic and (in nouns but not always verbs) Germanic languages]
I would think about why an affix might affect stress (or not).
Firstly, I'd suggest that we split this up into not two but four situations, forming a sort of continuum:
1: stress is unpredictable, or affixes are irrelevant to stress rules (eg stress is on first syllable and all affixes are suffixes; or, stress is always on a heavy syllable and no affix happens to have any heavy syllables)
2: stress follows regular rules that 'should' be affected by affixes, but the rules actually ignore the affixes
3: stress follows regular rules that are regularly affected by the presence of affixes
4: the presence of affixes affects stress but in a way that does not seem to follow the regular rules of stress assignment in roots
A language "should" have either situation 1 or situation 3, because these are regular. 1 can be generated regularly from 3 simply by setting rules and affix patterns that do not affect one another; so, 1 is actually an example of 3 (because stress is regular), but it looks superficially like an example of 2 (because stress is not affected by affixes).
If a language has 2 or 4, these are irregular, and there should be some sort of reason for it.
Why would a language have situation 2 (i.e. affixes that are invisible to stress rules)? Well, we can split this up into two groups of situations: one where the affixes have always been invisible, and one where they have become invisible over time.
If an affix has always been invisible, that means it must have always had in some way a status that differs from those of syllables within the root. This difference could be of either of two kinds: it could be an absolute difference (these syllables are treated as fundamentally different from root syllables due to some internal property) or it could be a relative difference (these syllables are different because they are not within the root).
In case of the relative difference, the obvious reason why affixes would be treated very differently from the same syllables occuring within the root is that the root boundary is relatively important compared to the word boundary: in some ways, the root is acting like it's still a complete word, even though in other respects the affixes are part of the word (if the 'root' is strong enough and the 'word' including the affixes is weak enough, then we'll say that the root IS the word and the affix is actually just a clitic; but this is a continuum, and the area between "affix is part of word" and "affix is not part of word" is where you get these relative-difference 'situation 2' situations).
Why would that be? In part it may be arbitrary. But we can also generally say that the more recently the affix has been attached, the less likely it is that it has already come to be fully treated as part of the word. I also wonder whether the phonological and syntactic salience of the root boundary may act to keep affixes 'outside', or conversely to invite them in. Eg maybe if the original limit of the word has been reinforced with special rules (eg initial and final fortition) maybe it's more likely that affixes remain in some way 'outside' the word in some respects?
In terms of absolute difference, some languages have a clear distinction between syllables that are in some way "full" and those that are "minor". Minor syllables commonly are unable to bear stress, and they may have other properties as well (they may not be allowed the same range of phonotactics, for instance). If affixes originate as minor syllables and to some extent retain that status once incorporated, they may not be allowed to take stress, which would leave the stress potentially unaffected by the affixes. [although it's also possible to have minor syllables that can't take stress but that can shift the stress]
Of course, these two solutions can contradict one another. For instance, it seems to be a thing in Indo-European that many small words, particles, were incapable of taking stress. Because of this, they tended to attach themselves to adjacent words - because they had to be included in the stress 'feet' of other words (as they couldn't take stress themselves), which moved some of the 'barriers' at the edge of words to include these clitics, which weakened the barriers as a whole, which tended to make those clitics come to be considered as part of the word, which, in turn, tended to make them affect stress rules, and thus end up being stressed, even though their unstressability is presumably the whole reason why this whole process kicked off.
Finally, as regards situation 2, there will be languages where affixes WERE visible to the stress rules, but became invisible. Why would that be? Presumably (as this is not a regular process) it was by analogy. The position of stress is part of the word, and hence can be regularised between different forms of the word by analogy, even where this leads to a violation of the formerly regular stress allocation rules. Generally, where the rules lead to strange, rare and unpredictable forms they are more likely to be analogised; where the rules give outcomes that are frequent, predictable and not unusual they are less likely to be analogised. Analogy can also operate between roots: if most combinations of roots and affixes regularly lead to no change in the stress, but certain shapes of root combined with certain affixes do, it's not unreasonable for these rare outcomes to be analogised away.
------------
And then we can ask the same (or equivalent) questions about situation 4. Why are affixes having MORE affect than the regular rules would predict? Either they always did, or this developed.
If they always did, presumably we're looking at an absolute difference in syllables (because I don't see why the relative word position would make an affix have more effect on the root than the root itself does). And here it's easy to see that if an affix originates in a root the affixed syllables may have a different status from at least some syllables in the root:
- the affix may come with its own stress. If the stress rules assume one stress per word, and suddenly you add a stressed affix, giving two stresses per word, then your regular stress rules no longer apply (at least synchronically). Diachronically, the same regular rule may apply, but because the stressed affix gives a different input for the rules than an unaffixed rule can the outcome may not look to follow the same regular rule.
- or, perhaps a language specifies which syllables CAN bear stress. Then, even an (originally) unstressed affix could affect the rules simply by being able to bear stress. For instance, if in each root only the first two syllables can bear stress, and you then add a new affix formed from an old root, the affix may still be able to bear stress (so that, for instance, a rule like 'place stress on the rightmost syllable capable of bearing stress', to pick an easy example) would lead to a stressable suffix attracting the stress.
These are just examples.
Alternatively, why might irregular stress attraction develop from a formerly regular situation? Well, sound change! If there's a sound change it can remove the regular conditions for a stress shift, but may not actually remove the stress shift.
Does the location of the affix (suffix / prefix) correlate with the likelihood to shift stress?
….
If i recall correctly from looking at a few online stress databases, the answer to your question that I quoted, depends on which natlang you’re working on.
In some languages the whole word, prefixes+wordstem+suffixes, must be re-analyzed to determine what kind of stress goes where;
In some the prefixes+wordstem must be reanalyzed as a unit, to distribute the stresses thru it, but the suffixes are not involved (though the stresses on the suffixes might be influenced by those on the prefixes+wordstem);
In others the wordstem+suffixes must be reanalyzed as a unit to (re-)distribute the stresses, but the prefixes are not involved (though the stresses on the prefixes might be influenced by those on the wordstem+suffixes);
In other natlangs the stresses in the wordstem are immune to both the prefixes and the suffixes, though distribution of the stresses on the affixes might be influenced by those on the wordstem.
….
If there’s any pattern of which natlangs are likelier to follow one of these types of systems, vs which natlangs are likelier to follow one of the other types of systems, I don’t recall ever seeing any discussion of that. I would imagine there is some tendency toward such a pattern, linked to the language family, or the linguistic area, or the languages’ typologies, somehow; but I don’t really have any hint (better than a wild surmise) what it would be like.
….
I wish I could supply you with citations for the above. If I ever can, I’m likely to post it here on your thread.
….
I hope I helped some! Or, at least, reassured you that help is out there, somewhere!
Thank you to everyone so far. This is all helping me think about the problem more deeply.
One issue is my conlang has phonemic stress. So it's hard for me to apply some set of changes to the "normal" stress patterns, because I haven't defined any. But maybe I am misunderstanding how lexical stress works. Am I right in thinking that languages with phonemic stress do have a certain stress patterns that they prefer, but that one process of another produces words that violate the pattern? Or are there languages where stress is truly unpredictable, word to word, as if chosen at random. What little I've read seems to indicate that linguists disagree on how to classify phonemic stress, which unfortunately doesn't help me.
Salmoneus wrote: ↑04 Aug 2023 01:48 Why would that be? In part it may be arbitrary. But we can also generally say that the more recently the affix has been attached, the less likely it is that it has already come to be fully treated as part of the word.
I was also wondering this. Could it be that stress shifting affixes tend to be more recently grammaticalized, and thus less likely to be analyzed as part of the base? I wish I could find research that looked that that, with a large enough sample size that I could generalize from it.
eldin raigmore wrote: ↑04 Aug 2023 04:30 I wish I could supply you with citations for the above. If I ever can, I’m likely to post it here on your thread.
Completely understood! I'm struggling to find research into this myself. Seems like a lot has been written on the theoretical aspects of stress shifting (as in, what are the cognitive stages we go through when applying stress), as well as many attempts to describe stress shifting within certain languages. But I can't find any cross-linguistic research about it, which is what I find most helpful for conlangs. But your recollections are helpful. Thank you!
(EDIT: I was misusing the term 'lexical stress', i think. I believe I meant phonemic stress. Corrected.)
collect_gluesticks wrote: ↑05 Aug 2023 21:47
Thank you to everyone so far. This is all helping me think about the problem more deeply.
One issue is my conlang has phonemic stress. So it's hard for me to apply some set of changes to the "normal" stress patterns, because I haven't defined any. But maybe I am misunderstanding how lexical stress works. Am I right in thinking that languages with phonemic stress do have a certain stress patterns that they prefer, but that one process of another produces words that violate the pattern? Or are there languages where stress is truly unpredictable, word to word, as if chosen at random. What little I've read seems to indicate that linguists disagree on how to classify phonemic stress, which unfortunately doesn't help me.
I doubt that there are many languages where stress is truly random, because people like patterns, and language is a game of finding patterns in things. But it's not impossible, I guess.
Then there are languages in which language is both predictable and unpredictable, in different senses. English is an example. In an English text, apparently, nearly 95% of words encountered will have stress that can be predicted by certain rules (if you count by all possible words, rather than actual corpus frequency, it'll be a bit lower, but still very high). But most native speakers do not know those rules consciously. And few ESL teachers will teach English-learners those rules, because it's probably easier just to learn English stress as though it's random (beyond 'if in doubt, first syllable!').
Diachronically, we can assume that every case of lexical stress evolved from regular stress - because eventually lexical stress tends to get analogised out, so it's unlikely lexical stress would remain in perpetuity, so a language with lexical stress probably hasn't had it in perpetuity, so it probably evolved from regular stress in some way. And as a result there's likely to be traces of diachronics in synchronic stress allocation rules.
In the case of English, lexical stress can basically be blamed on five factors:
- the Germanic prefix rule, whereby prefixes are unstressed in verbs but stressed in nouns
- the wholesale adoption of Latinate derivational suffixes, bringing with them Latinate stress rules triggered by those suffixes
- the continual borrowing of vast numbers of loanwords from different sources, bringing their stresses with them (but gradually analogising them away over time).
- zero derivation partially (and irregularly) undoing the Germanic prefix rule
- ad hoc analogising between words deriving their stress from different sources
Another major source of irregularity may be compounding, although I'm not sure that's an issue in English.
Salmoneus wrote: ↑04 Aug 2023 01:48 Why would that be? In part it may be arbitrary. But we can also generally say that the more recently the affix has been attached, the less likely it is that it has already come to be fully treated as part of the word.
I was also wondering this. Could it be that stress shifting affixes tend to be more recently grammaticalized, and thus less likely to be analyzed as part of the base? I wish I could find research that looked that that, with a large enough sample size that I could generalize from it.
I don't know whether stress-shifting (or stress-defining) affixes tend to be more recently grammaticalised, but I'm confident that more recently grammaticalised affixes will tend to be less integrated into the regular stress rules.
For instance, in Germanic, the more firmly (and presumably anciently) attached nominal prefixes DO 'shift stress' - because they're the first syllable, and the regular rule is 'stress the first syllable' - whereas the more weakly and recently attached verbal prefixes DON'T 'shift stress', because although they're the first syllable the regular rule doesn't include them (they're not treated as fully part of the word for stress purposes).
Salmoneus wrote: ↑06 Aug 2023 01:39
I don't know whether stress-shifting (or stress-defining) affixes tend to be more recently grammaticalised, but I'm confident that more recently grammaticalised affixes will tend to be less integrated into the regular stress rules.
Thank you for your reply. It clears a lot up for me.
I think you've confirmed for me that languages with phonemic stress can still have strong tendencies in how stress is assigned. I'm going to more clearly define what those tendencies are in my conlang. Fortunately, I only have a few hundred words, and there are some emergent stress patterns already present. (I just hope I haven't been subconsciously applying English stress patterns to my conlang this whole time.)
![[<3]](./images/smilies/heartic.png "<3"){kind=small}
Papuan languages, Morphophonology, Lexical Semantics 
