English Primary and Secondary Stress

[Moose-tache]

I've been studying English stress patterns for a WIP, and I need to talk about it. And I need your help, because I have a lot of questions. I thought it would be good to have a separate thread, because if we really get into this topic, it could crowd the English questions thread. Anyway, I'll start with a theory that's been rattling around my head.

When you learn English stress as an ESL student, you are taught that there is a series of suffixes (most obviously Greek nominalizing -y) that create penultimate stress: phOtograph becomes photOgraphy, and hOmonym because homOnymy. This works well enough as a rule (assuming you can keep your Greco-Latin and Germanic -y suffixes separate), because it's fairly regular. But it doesn't make much sense, does it? By what mechanism does a suffix say "I shall draw the primary stress to a point two steps away from myself?"

The obvious possibility is that it's not the -y but the -graphy that's moving the stress around. Normally in a four-syllable Greek word you get a PuSu pattern (gonorrhea, for example), but "graphy" isn't one root. The -y suffix in this case comes from a Greek suffix -ia that carried stress, so the learned prats in charge of words like this might have felt that the second syllable in "graphy" was more deserving of stress. There are two ways you could express this: move the primary stress all the way to the end, or simply demote the first syllable in "graphy" to a syllable that carries no primary or even secondary stress. Since photography sounds pretty offensive in the context of English phonology, they went with the latter, causing stress to bunch up at the syllable immediately before the deliberately destressed "graph." This pattern was later extended by analogy to any word with nominalizing -y, even in words with no Greek etymology, like "photobomby."

Does this make sense? I haven't found any alternative theories in my search through the literature, but I am away from any good research library so I'm basically limited to Academia.org at the moment. If anyone knows of a better theory, let me know!

[Kuchigakatai]

When you learn English stress as an ESL student, you are taught that there is a series of suffixes (most obviously Greek nominalizing -y) that create penultimate stress: phOtograph becomes photOgraphy, and hOmonym because homOnymy.

Nah, we aren't taught shit with this much sophistication when it comes to phonology, or the relationship between spelling and pronunciation. No way! Unless your teacher is basically Finlay or Richard W I guess, but there's less than 0.1% chance of that, even when living in the US.


I don't know about the development of English stress, but I suspect the answer may lay in the stress movement of French borrowings, and Latin borrowings after being given a French-y shape (as English generally does), that happened in which e.g. Chaucer's "melodýe, Natúre, pilgrimáge" became the more Germanic "mélody, Náture, pílgrimage", and analogy based on those existing words.

I'm also not sure to what extent Greek words can be separated from how they were borrowed into Latin and French, since a lot of words should enter English via Latin (applying some local Frenchification too) or via Latin through actual French. E.g., the Dictionary of Medieval Latin from British Sources cites examples of (h)omonymus ('having the same name; ambiguous with more than one sense') from Alcuin, Ælfric and Bacon... The word family probably entered via Latin, although these learned words tend to have complicated histories.

[Moose-tache]

I don't know about the development of English stress, but I suspect the answer may lay in the stress movement of French borrowings, and Latin borrowings after being given a French-y shape (as English generally does), that happened in which e.g. Chaucer's "melodýe, Natúre, pilgrimáge" became the more Germanic "mélody, Náture, pílgrimage", and analogy based on those existing words.

I thought of this: photographY with secondary stress on the second syllable retracting to simply photOgraphy with the secondary stress now primary. That's basically what's going on in terms of deep structure in my theory. But the problem with final stress actually being realized phonetically is that there is a series of Latinate suffixes that attract secondary stress to themselves, and keep it to this day, like ize. So I think the avoidance of final primary stress was always operant in English, and demoting graph in photography was a way to encode stress in the final y without having to phonetically stress it. If instead the y just received secondary stress, it would presumably remain that way in modern speech, because that is perfectly allowed. But for me at least, it carries no stress, primary or secondary.

So we end up with three categories.
1) suffixes that carry no stress: adjectival -y, for example - These just follow normal trochaic or dactylic cadence.
2) suffixes that attract secondary stress to themselves: -ize - These move secondary stress, sometimes causing adjacent syllables to become unstressed.
3) suffixes that speakers wanted to mark as (primary) stressed without sounding weird: nominalizing -y - these cause the previous syllable to become conspicuously unstressed; maybe originally the y was secondarily stressed, but the conspicuous demotion bled into the final syllable to create a dactyl.

Then there is a whole series of Greco-Latinate suffixes that draw primary stress to the syllable before, like -ic. I feel that if I can come up with a set of rules that works for all of these suffixes I will become the Quisatz Haderak.

[Estav]

Then there is a whole series of Greco-Latinate suffixes that draw primary stress to the syllable before, like -ic. I feel that if I can come up with a set of rules that works for all of these suffixes I will become the Quisatz Haderak.

There is a lot of prior work on that class of suffixes. Chomsky and Halle's The Sound Pattern of English resorts to analyzing -ic in these cases as underlyingly -ical with omission of the -al. "The Architecture of the English Lexicon" by Jonathan B. Alcántara (1998) instead analyzes -ic as underlyingly /-icə/ (page 3-117).

[zompist]

I expect there's a lot more categories than that. I started looking but realized it would be more of a project.

I looked at all this just a bit when doing my spelling project. There are a lot of weird exceptions-- sometimes English makes a big deal of historically long vs short syllables, and then some words just laugh at those rules. Why do we say rapIDity, but scienTIfic? Why does TItillate have a short i, Isolate a long one?

One comment though: you can look at Greek stress... but don't bother, no English speaker ever does. We don't, after all, say mesopotaMIa, epitoME, physiCS (the stressed syllable here is lost), peloPONnese, comeDY, tiMOthy.

[Moose-tache]

There are a lot of weird exceptions-- sometimes English makes a big deal of historically long vs short syllables, and then some words just laugh at those rules. Why do we say rapIDity, but scienTIfic? Why does TItillate have a short i, Isolate a long one?

That first pair is actually regular according to the silly, ad hoc rules of English stress: nominalizing -y forces antepenultimate stress, while -ic forces primary stress onto the previous syllable (or, if you believe that -ic is short for -ical, then -ical forces antepenultimate stress before it gets abbreviated). As for inconsistent rendering of vowel length, I don't know. Both of those words entered English after the GVS, so it probably just reflects which Georgian Fancyboy's pronunciation was more popular. As you say, I'm not expecting Greek or Latin stress rules to give us a clear indication of what early Modern English speakers were doing, even when they (as Kuchigakatai pointed out) clearly wanted to ape continental pronunciation to some extent. The Latin stress rule about heavy and light syllables, for example, seems to have no impact on English. That's good for me, because I've got enough on my plate without having to figure out what a bunch of overeducated fops thought "habeas" was supposed to sound like in a dead language.

Another fun complication is that suffixes that move primary stress are largely immune to secondary stress, but suffixes that do not affect primary stress can, in some cases, take secondary stress. I gave the example of -ize above, but this also applies to -fy and -ate. You see the pattern. That last creates minimal pairs of verbs and nouns/adjectives, like "animate."

[Moose-tache]

Update! I had originally had the intuition that -graphy, -cracy, etc. were the suffixes, rather than -y. But that would be stupid, I thought. Then I remembered the world "allomorphy." This shows no motion of primary stress. So maybe combinations of some Greek roots plus y act as suffixes that attract stress to the immediately preceding syllable, just like other Greek, Latin, or French suffixes like -ic, -tion, etc. This still leaves the question of by what phonetic mechanism the stress was moved, but at least it would be just one rule that is needed instead of two.

[hwhatting]

There is a lot of prior work on that class of suffixes. Chomsky and Halle's The Sound Pattern of English resorts to analyzing -ic in these cases as underlyingly -ical with omission of the -al. "The Architecture of the English Lexicon" by Jonathan B. Alcántara (1998) instead analyzes -ic as underlyingly /-icə/ (page 3-117).

This is just generative phonology at its usual making up of unattested "underlying forms" to make things look as if they adhere to its stated general rules - instead of just saying "you have to remember that suffix -ic has this or that effect" they say "you have to remember that suffix X is actually underlying suffix Y and then our ingenious rule 22 applies".

[zompist]

There is a lot of prior work on that class of suffixes. Chomsky and Halle's The Sound Pattern of English resorts to analyzing -ic in these cases as underlyingly -ical with omission of the -al. "The Architecture of the English Lexicon" by Jonathan B. Alcántara (1998) instead analyzes -ic as underlyingly /-icə/ (page 3-117).

This is just generative phonology at its usual making up of unattested "underlying forms" to make things look as if they adhere to its stated general rules - instead of just saying "you have to remember that suffix -ic has this or that effect" they say "you have to remember that suffix X is actually underlying suffix Y and then our ingenious rule 22 applies".

Worth investigating: Sound Pattern of English came out in 1968. That's also when the really good weed arrived at MIT.

That is... I agree with you; I think C&H mistook historical for generative processes. Though the ic/ical thing isn't even historical. It's mind-boggling that a child would come up with *ical as an underlying form for -ic, and nonsensical that every child does so. At least with generative syntax Chomsky had the excuse of believing that the grammar itself was innate. He can't say that the history of English suffixes is innate.

[zompist]

1) suffixes that carry no stress: adjectival -y, for example - These just follow normal trochaic or dactylic cadence.
2) suffixes that attract secondary stress to themselves: -ize - These move secondary stress, sometimes causing adjacent syllables to become unstressed.
3) suffixes that speakers wanted to mark as (primary) stressed without sounding weird: nominalizing -y - these cause the previous syllable to become conspicuously unstressed; maybe originally the y was secondarily stressed, but the conspicuous demotion bled into the final syllable to create a dactyl.

I've been going through suffixes, and I think I have at least these categories:

* suffixes that take primary stress-- boring for your purposes
* suffixes with primary stress just before: percentile, enormous, expectable
* suffixes with primary stress way before: patricide, democracy, geomancy, iconoclast, contemplative, communism, polyglot, Holocene, gaseous, hexagon, audacity, normality

The problem is, I don't have a source at hand for where the secondary stress is, and I'm not sure I trust my ear (or that my ear would agree with yours). Some have clear SS: geomancy, iconoclast, contemplative, polyglot, hexagon. Most of the rest seem to be to just fall off evenly, or have no secondary stress-- especially commendable, audacity, gaseous, normality. I don't really perceive photography as having secondary stress at all (the a is [ə]). I guess allomorphy does, in that I can't reduce the second o (*"allumurphy").

[Richard W]

I'll try to summarise a paper I read on English stress assignment about forty years ago. One should remember that for much of English vocabulary, the written form is primary. Those who don't read and write don't control such words and their Latinate derivations. Indeed, the opening theme of the paper was the question of how their was such widespread agreement on how to say 'Chappaquiddick'. Thus the paper was very much about reading rules, so vowel length is relevant and is as used in the writing system.

English words are split into feet, with possibly in initial extra-metrical syllable consisting of an open syllable with a short vowel. Each foot has one stressed syllable. The last foot may have one to four syllables. (Though the paper didn't mention it, many people have a constraint against tretrasyllabic feet, which may be worked around phonetically by syncopation.) Other feet have one or two syllables, with the default being two if possible.

The stress goes on the last heavy unignored syllable, subject to the limit on the length of the syllable. The suffix on an adjective is ignored, and the final consonant (if any) of a noun is ignored. If the final consonant of the penultimate syllable is a resonant, it may be ignored. This allows the initial stress of messenger and passenger, and I think there are some more learned words that it also affects. As these are reading rules, exceptions are possible.

There are some morphological exceptions - the adjective suffix -ic forces penultimate stress, and for those happy with tetrasyllabic feet, -ity forces antepenultimate stress. They also shorten the vowel that will be stressed, unless it is 'e' or 'u'. (Contiguous digraphs seem to be exempt.) I can't remember whether these rules were covered in the paper.

There's also a related, converse rule for adjective suffixes written with two syllables whose first syllable may be reduced to yod. The final stem of the vowel is lengthened (as in the correspondences for English spelling) if it is 'a', 'o' or u', but shortened if it is 'i' or 'y'. I've noticed that the rule can be extended analogically - for me, vowel shortening applies to Einsteinian.

Finally, when testing these rules, remember that, as the Zompist himself said, English comes in both the free and paid for versions.

[Moose-tache]

Noam Chompsky is the Tolkien of linguistics, in that his imaginative descriptions of fictional languages continue to inspire young readers to this day.

[Moose-tache]

Zompist

I think it is likely that secondary stress especially will be idiosyncratic. Even when people agree, the stress may be lexically rather than phonetically determined in some cases. For example, I can’t think of any reason why the last vowel in “habitat” can never be reduced for me, but the last vowel in “halibut” must always be reduced, other than lexical stress assignment. But there are clearly rules, as with primary stress, spelling, morphology, and every other aspect of English that defies perfect regularity.

Richard W

Your description sounds a lot like the Myers paper from 1987 that is quoted by Alcantara in the above-linked dissertation. Could that 35 year old paper be the 40 year old one to which you refer? In any case, the metrical foot description works well in some cases. The other idea the GG people put forward is the idea of “resyllabification,” basically that sometimes primary or secondary stress causes CVC.CV to be reinterpreted as CVCC.V, triggering existing vowel shortening rules. It all kind of works within certain contexts, but fails on a number of levels. Even the regular rules of metrical feet (ignore this nasal, but not that nasal, this foot doesn’t count, but this one can be extra morae) are probably not operant in the minds of speakers.

What’s probably happening is that there are archetypes and lexically determined defaults, and a series of rules to override them. Most efforts to determine English stress assume that the number of archetypical stress patterns is one, which is clearly false. But it’s also clear that speakers can apply rules by analogy, so they do exist. I gave the example of “photobomby,” which demonstrates that the rule dictating words with nominalizing -y is operant, and “allomorphy,” which demonstrates that it is not universal.

[zompist]

I think it is likely that secondary stress especially will be idiosyncratic. Even when people agree, the stress may be lexically rather than phonetically determined in some cases. For example, I can’t think of any reason why the last vowel in “habitat” can never be reduced for me, but the last vowel in “halibut” must always be reduced, other than lexical stress assignment. But there are clearly rules, as with primary stress, spelling, morphology, and every other aspect of English that defies perfect regularity.

Sounds right to me.

I gave the example of “photobomby,” which demonstrates that the rule dictating words with nominalizing -y is operant, and “allomorphy,” which demonstrates that it is not universal.

I don't get out much, so I don't think I've heard either word in the wild. How do you pronounce them?

[Moose-tache]

I gave the example of “photobomby,” which demonstrates that the rule dictating words with nominalizing -y is operant, and “allomorphy,” which demonstrates that it is not universal.

I don't get out much, so I don't think I've heard either word in the wild. How do you pronounce them?

"Photobomby" is pronounced pho-TOB-bomby, same as "photography." I've only ever heard it spoken as a joke, but the joke relies on people knowing the rule, and how it would be applied to novel words. "Allomorphy," to me, is ALL-omorphy, same as "allegory," i.e. the default stress pattern of simplex English nouns of four syllables, or the same stress pattern you would get if you added a non-stress-affecting suffix, like "allomorphly." In both cases the primary and secondary stress might be hard to differentiate, but they're easily distinguished from unstressed syllables by their lack of reduction.

[zompist]

"Photobomby" is pronounced pho-TOB-bomby, same as "photography." I've only ever heard it spoken as a joke, but the joke relies on people knowing the rule, and how it would be applied to novel words. "Allomorphy," to me, is ALL-omorphy, same as "allegory," i.e. the default stress pattern of simplex English nouns of four syllables, or the same stress pattern you would get if you added a non-stress-affecting suffix, like "allomorphly." In both cases the primary and secondary stress might be hard to differentiate, but they're easily distinguished from unstressed syllables by their lack of reduction.

OK, thanks. I would not have guessed the first one-- I would have modeled it on "phonotactics". But it all depends on which pattern people apply.

[Estav]

I gave the example of “photobomby,” which demonstrates that the rule dictating words with nominalizing -y is operant, and “allomorphy,” which demonstrates that it is not universal.

The stress pattern of allomorphy, necromancy, rhinoplasty is usually treated as regular for -y compound nouns ending in -CCy where CC is a consonant cluster of the type that attracts penultimate primary stress in -al or -ous words (or equivalently, in Latin penultimate syllables). Actual speaker behavior with unfamiliar -y nouns does not necessarily follow the observed generalizations.

[Travis B.]

My instinct is for photobomby to be prononounced as /ˌfoʊtoʊˈbɒmi/~/ˌfoʊtəˈbɒmi/ myself - pronouncing it as /foʊˈtɒbəmi/~/fəˈtɒbəmi/ does not feel natural to me at all.

[Linguoboy]

My instinct is for photobomby to be prononounced as /ˌfoʊtoʊˈbɒmi/~/ˌfoʊtəˈbɒmi/ myself - pronouncing it as /foʊˈtɒbəmi/~/fəˈtɒbəmi/ does not feel natural to me at all.

Ditto. I'm having difficulty producing the latter pronunciation, in fact. I keep wanting to make it rhyme with "lobotomy".

[Richard W]

Your description sounds a lot like the Myers paper from 1987 that is quoted by Alcantara in the above-linked dissertation. Could that 35 year old paper be the 40 year old one to which you refer?

No. I lost access to printed journals in 1982, and only recovered merely occasional on-line access in the 90's. One thing I should have learnt, but didn't, was to make a proper note of sources. It's just possible I've kept some notes on the paper - I should have a look.