Sophie Hardach reports for the NY Times:
[…] In 2009, Dr. [Kathleen] Wermke’s and her colleagues made headlines with a study showing that French and German newborns produce distinctly different “cry melodies,” reflecting the languages they heard in utero: German newborns produce more cries that fall from a higher to a lower pitch, mimicking the falling intonation of the German language, while French infants tend to cry with the rising intonation of French. At this age, babies experiment with a wide variety of sounds, and can learn any language. But they are already influenced by their mother tongue.
Today, Dr. Wermke’s lab houses an archive of around a half-million recordings of babies from as far afield as Cameroon and China, where a team of graduate students armed with recording equipment paced the corridors of a Beijing hospital around the clock. […] Quantitative acoustic analysis of these recordings has produced further insights into the factors that shape a baby’s first sounds. Newborns whose mothers speak tonal languages, such as Mandarin, tend to produce more complex cry melodies. Swedish newborns, whose native language has what linguists call a “pitch accent,” produce more sing-songy cries.
These studies underpin the lab’s broader effort to map the typical development of a baby’s cries, as well as vocalizations like cooing and babbling. Knowing what typical development looks like, and what factors can influence it, helps doctors address potential problems early on. […]
“Imagine you’re thrown into a new language environment, which is what happens with the newborn,” said Judit Gervain, a senior research scientist at the National Center for Scientific Research in Paris who studies early speech perception. “There’s just so much going on: There are all the words, there’s all the meaning, all the grammar, all the sounds, all of it. You can’t do it all, it’s just too much. One way prosody helps is it gives them nice little chunks that are the right size.”
In English, for example, a stressed syllable is often a cue for the start of a word, as in: English language. In French, a lengthened syllable signals the end of a sentence, as in: “Bonjour Madame!” Long before they can speak, babies begin to recognize patterns like these. “A lot has to happen before that first word is produced,” said Janet Werker, a developmental psychologist at the University of British Columbia who studies early language acquisition.
There’s lots more good stuff, including the soothing effect of maternal howling on infants. Thanks, Eric! (Incidentally, we discussed newborns and language in 2007 and 2011, and I note that the researchers featured in those posts were named Weikum and Werker; together with today’s Wermke, they constitute a cluster at least as impressive as the monosyllabic Indo-Europeanists — Rask, Bopp, et al. — we’ve discussed on occasion.)
On that basis, I’m not sure what the authors would make of my family’s linguistic heritage. The midwives nicknamed my firstborn “The Little Train” because of an ear-piercing, high-pitched scream, never a cry. Or does that say more about me than I care to reveal?
The news for Crystal is:
https://twitter.com/Lisa_Bizzle/status/1193220924327288834
…if you’re speaking slowly enough.
Huh. Sounds like this kind of research is only done by geographic Frisians.
In English, for example, a stressed syllable is often a cue for the start of a word, as in: English language.
How true is this? By my rough analysis of OED2, where a stress is indicated in polysyllables, it comes on the first syllable 67615 times (55%), and on another syllable 55974 times (45%).
But maybe it’s more true of the average speaker’s lexicon than the entire lexicon?
There is also a particular stress pattern for talking to babies. Maybe the babies learn the word divisions from this stress pattern but discard it later (except when they talk to babies).
Looking at words in the comments, the pattern of multi-syllable words having first-syllable stress seems to hold pretty strongly. One notable exception: verbs (that is, those with a two-syllable root). The word “reveal” brings to mind for me words like record where the noun has first syllable stress, and the verb second syllable stress. There are other exceptions, but first syllable stress seems to be a pretty solid pattern.
Thinking about the verbs thing, that shows one reason why usage would more strongly show a first syllable stress pattern than dictionary entries. -ing forms don’t have dictionary entries. They do have initial stress if they have a mono-syllable root.
You’d really need a decent-sized corpus, preferably a spoken-word corpus, to get a good sense of how common frontal stress actually is. (Check out the previous sentence, or this one for that matter.)
Okay so I’ve done this on the Brown Corpus using the CMU pronunciation dictionary, which produced 102,568 matches for polysyllables. 54% had primary stress on the first syllable, 6% had secondary stress on the first syllable, and 40% had an unstressed first syllable. So, not too far off from OED2 at all. A smaller twitter sample produced slightly different results: 38K polysyllables, in which the first syllable broke down as: P-62%, S-7%, U-31% (this is nearly identical to the only spoken corpus I have handy, which is much smaller).
@D-AW: do you have an example to hand of a word with secondary stress on the first syllable ? Offhand I can’t think of one, but then I’m not accustomed to thinking along such lines.
Wait: “autopoiesis” is one such, no ?
“Dodecahedron”. Now they’re coming thick and fast. But at the 6% level…
But the general corpus is fairly irrelevant; what matters is the words parents (mainly mothers, historically) use in talking to newborns, and I suspect these skew more towards monosyllables and penultimate-stress disyllables.
Of course, while we may think of monosyllables as having “first-syllable stress” more or less by definition (although I guess they’d also have final-syllable stress by definition), many extremely common “function word” monosyllables are unstressed in ordinary speech. This is particularly obvious when they have reduced vowels, but is not limited to that situation. You have to know where the word divisions are to pick them out. Take for example the catch-phrase “show me the money.” That’s often pronounced as SHOWmetheMONey. Two syllables stressed out of five (if the “me” was unusually salient — i.e. the point being communicated is “don’t show him the money, show it to me instead” — that would lead to stress being added or shifted to the “me.” At what developmental stage do we think fetuses/infants in Anglophone cultures know how many different words that five-syllable string is made of?
Along the same lines, at what developmental stage do we think that fetuses/infants demand that you show them your money ? Possibly quite early, given that teenagers can make such a demand without necessarily being able to count to five.
You might expect some differences in a frequency-weighted list, which it doesn’t sound like D-AW was working with, but just on a quick back of the envelope inspection here, the polysyllables in this list I googled up of the “200 most common” (according to some uncertain source) English words show only a very modest majority with first syllable stress – confounded by at least two that (in my idiolect) can go either way, viz. “present” and “increase”: note that if you treat those as pairs of different albeit etymologically-related words that happen to be spelled the same (which may well be the analysis that best explains why they are not pronounced the same), maybe all four resulting words would drop out of the top 200? https://teacherjoe.us/Vocab200.html
@JW Brewer
“present” and “increase” follow the verb/noun stress variation pattern I mentioned, with the noun getting first syllable stress, and the verb second syllable stress. Though increase perhaps with some variation.
You’re not going to hear “present” and “increase” in much nursery talk; think “Want a nice cookie?” and “Look, a rabbit!”
FWIW I pronounce “dodecahedron” with an unstressed first syllable and secondary stress on the second syllable. But if I understand Stu’s quandary aright there seem to my ear to be plenty of four-syllable-or-longer words with primary stress later in the word but secondary stress on the first syllable (as evidenced by the second syllable having perceptibly less stress than the first). Examples that came to mind that seem more common in ordinary speech that “dodecahedron” include “locomotive” and “intellectual.” At least in my idiolect.
Ellen K: Yes. They just seem to be the only ones following that common pattern to have made that top 200 list. That “present” is also reasonably common as an adjective is a further wrinkle, although maybe adjective-follows-the-noun-not-the-verb is the usual way that wrinkle plays out?
hat: I would think the noun “present” comes up fairly often in speech addressed to infants, esp if they are not first-born. Parents themselves may realize that the baby won’t grok the difference between a “present” being given to him/her and any other object placed within reach, but older siblings are often quite interested in giving the new arrival presents (and announcing them as such) and parents tend to tolerate that. I suspect the noun “present” is also pretty common in the lexicon of picture books meant to be read aloud to quite young children. The verb “present” with its different stress pattern may turn up a bit later, although if e.g. some toddler-focused picture book has a scene where the pig mayor of the talking-animal town presents a medal for bravery to a heroic squirrel I wouldn’t think the plot point odd or the verb “present” too grown-up for the audience.
hat: I would think the noun “present” comes up fairly often in speech addressed to infants, esp if they are not first-born.
Well, I was thinking more of the verb, but even the noun will not be frequent in those crucial first months. In any case, it has initial stress, so it doesn’t affect the point.
Come to think of it, it would be interesting if a study focused on any patterned differences between infants who are the only (relatively young) child in their household versus those who have somewhat older siblings and thus may be exposed from the very beginning to overhearing frequent dialogues between parents and a four-year-old. If the baby has a much older sibling in the house, like a teenager, the conversations overheard are probably more likely to resemble (in terms of prosody and phonology, if not in substance) conversations between the parents or other grown-ups.
This isn’t really a claim about words at all. It’s about prosodic rhythm and tone, about babies acquiring the elementary basic melodic structure of their mother tongue already in the womb. Words and word stress is only important to the degree that they define the general sentence melody.
Trond’s comment resolves my “quandary”, which I was not even aware of (I had imagined I was merely annoyed). The phenomena are more pragmatically understandable in terms of “acquiring elementary basic melodic structure”.
an example to hand of a word with secondary stress on the first syllable
Thirteen, fourteen, fifteen, sixteen, seventeen…
(Actually, Wiktionary says that those words differ in primary stress depending on what follows them. I didn’t know that.)
Yes, I brought that up a few years ago (et seq.)
In Han Solo’s first scene, Harrison Ford changes the stress pattern on seventeen between utterances, in a way that I have always found distinctly weird
@Stu Clayton
Here’s a random sample of 20, according to the CMU pronunciation dictionary:
intersected, dermatologists, solzhenitsyn, automatics, paleontologists, interrupts, supersonic, northwest, nationalization, litigation, carnegie’s, semifinal, arnaz, insincere, ciaobella, mongosut, aspiration, agrochemicals, vanlue, blockaders
@Trond Engen
The general is actually elided in the quoted paragraphs. It comes two paras above the stuff about English and French:
After they are born, young babies mimic many different sounds. But they are especially shaped by the prosody they heard in the womb, which becomes a handy guide to the strange sounds coming from the people around them. Through stress, pauses and other cues, prosody cuts up the stream of sound into words and phrases – that is, into speech.
I interpret this and the later claim about initial stress to mean that English babies learn where words begin (partially) by identifying initial stress. I just wondered how many words were actually like this. I’m sure Hat is right that the lexicon infants are exposed to skews towards monosyllables (penultimate stress disyllables I just don’t know about – you’d need a corpus I think to make good that intuition) but that seems to me another argument against the claim about word boundaries, since monosyllables can be stressed or unstressed according to various factors.
The business about word demarcation seems to be simply speculation rather than observation, unlike the rhythm and tone stuff.
In any case, I don’t think babies actually do distinguish words when they are first learning to speak: we’re all polysynthetic to begin with. Talking of “words” at that stage is just projection of adult speech habits onto children who haven’t actually acquired them yet.
It’s cute and all to talk of babies “crying in German”, but it’s not much more than a (pretty strained) metaphor. It confuses potential with achievement.
In actual developed adult language, for that matter, word demarcation is not a straightforward unproblematic “given.” If informal spoken French had no written form and no known history, it would probably be tagged as polysynthetic. And I have spent many a happy hour teasing out the actual patterns of word demarcation in Kusaal, despite the fact that stress falls regularly on the root syllables of open-class words in that language. The standard orthography systematically divides words inconsistently. Kusaal is hardly alone in that …
Syntactic and phonological “words” fail to match in many, many languages.
If informal spoken French had no written form and no known history, it would probably be tagged as polysynthetic.
That’s interesting. Can you elaborate?
“It would probably be tagged as not worth the effort needed to understand”, is how I would put it. 50 years of reading and hearing French, and there are still glitches in my signal processing. This does not happen to me with German. Ever. Never.
Of course the basic reason is that I never needed French in order to get along in the world, so I never cared that much nor put as much effort into it. It did not fly into my mouth like roasted miniature pigeons (Ger.loc.), so I said to hell with it.
JC has posted quite extensively on this (and I hope he’ll chip in), as have others here. Meanwhile:
https://www.academia.edu/2000636/Grammaticalization_of_polysynthesis_with_special_reference_to_Spoken_French_
A caveat is that there is no real accepted definition of “polysynthesis”; the most commonly cited attempt to give the term a precise single meaning was Mark Baker’s proposal of a “polysynthesis parameter”, which like many Chomskyana doesn’t really survive contact with data more extensive than those used to motivate the original proposal.
For the issue in question, though, this is not particularly important.
@DA-W: Well, sure. I just thought we lost track of that.
A few thoughts (I can’t quite string them together to a coherent opinion):
The number is interesting, but we should never expect 100%. As soon as a pattern is dominant enough to be default in a context, there’s value in breaking it for emphasis.
What David E. says about words. This is about marking out prosodic units. I guess we later learn to parse those as grammar and lexical meaning by noticing recurring (sub-)patterns. To the extent that we learn words, it’s when our parents help us along by pointing at things and saying “horse”.
If the study holds (or even if it doesn’t), it may contain an important insight about why languages develop distinct (but essentially arbitrary) prosodic patterns, rhythms and melodies.
I’m not sure if there’s anything there, though. How do they measure it? What is the signal-to-noise ration and the margin of error? Are the results replicable?
The signal-to-noise ration: “Please, sir, I want some more”.
Mummy!
we’re all polysynthetic to begin with
I would rather say holophrastic, as in [gɪmiˈjit]. This is certainly a morphological word when an anglophone toddler says it.
JC has posted quite extensively on this
“I loathe Mary”, but see the following comments by Marie-Lucie, Etienne, and the usual suspects, pretty much down to the (current) end.
Babies might not care about money, but they learn to communicate “give me the smartphone” very early. And the sadness when the smartphone stops singing or telling stories! (It takes longer to learn to navigate the smartphone, you see.)
It’s quite interesting what kind of things babies and small children are interested in.
My daughter knew how to operate a video recorder at age one. That was in the late nineties, when mobile phones were still mostly phones and not yet entertainement machines.
My little sister seemed to treat “where is a” as indivisible for a while. There’s no clear or consistent stress in the whole thing.
French polysynthesis on LH last year.
My little sister seemed to treat “where is a” as indivisible for a while
Kusaal actually has a monosyllabic lia “where is?”, which has no phonological resemblance to either of the verbs “to be” or to the word for “where”:
Fʋ ma la lia?
(2SG mother:SG ART be.where?)
“Where is your mother?”
alongside the perfectly acceptable alternative
Fʋ ma la bɛ yaani?
(2SG mother:SG ART exist where?)
Is this the relic of a lost locational be, replaced elsewhere by the existential be?
@Trond:
That is a very interesting thought. It would make a lot of sense as an explanation. I can’t find much evidence for the idea though (more’s the pity.)
Synchronically, lia is not a verb but a non-verbal predicator. It can’t take any of the verbal particles that mark tense and so forth, for example. But that wouldn’t, of course, mean that it couldn’t have originated as a verb.
All the Western Oti-Volta languages conflate “exist” and “be in a place”, and the forms are all obviously cognate with Kusaal bɛ. The copula verbs are all over the place etymologically though: for example, the Kusaal aen goes back to *ŋaya, but even the very closely related Mampruli has nyɛ and the not-quite-so-close Mooré has yaa; the only way to derive all of these from the same original form would be to invent ad hoc rules specifically intended for that purpose.
It would actually not be utterly implausible to suppose that these were all independent innovations and that Proto-WOV didn’t have a copula verb at all: there are non-verbal predicators for cases like Kusaal
Dau n la.
“That’s a man”
(Man:SG CATENATOR that.)
while predicative adjectival meanings were pretty certainly normally expressed with cognate stative verbs, as is still very frequently the case in the modern languages: Kusaal giŋ “short”, but
Dau la daa gim.
“The man was short.”
(Man:SG ART TENSE be.short.)
In Oti-Volta outside the Western subgroup, Nawdm, Buli and Moba similarly conflate “exist” and “be somewhere” in forms cognate with bɛ; and again the copula verbs aren’t cognate (nnìí, ale, tēé respectively.)
There are Niger-Congo languages further afield with “be” verbs that look like “lee” and the like (e.g. Proto-Bantu *dɪ̀), but you could probably match pretty much any CV syllable with some Niger-Congo verb “to be.” It’s a big and diverse group.
I see that the even-remoterer Gur language Chakali also has just a copula jáá versus an exist/be-in-a-place verb dʊ́á.
Yoruba has several copulas with different nuances (show-offs) but again conflates “exist” with “be in a place” (as wà, probably yet again cognate with bɛ); so too even Mandinka (exceedingly remotely related to Kusaal, if at all): copula mú, existence/be-in-a-place bé. The Dogon language Tommo So has several verbs for existence/be-somewhere but they differ in other respects than being limited to one or the other construction.
I think it’s the norm for the whole area. Not pan-West-African, though: the Songhay languages seem to distinguish existence verbs from location verbs, and Hausa does make a three-way distinction, though in a rather different way: it has particles rather than verbs for the copular and existential constructions, while “be in a place” is expressed in the same way as “be singing” or “be clever.”
Miya (the only other Chadic language that I can lay my hand on a decent grammar of this minute) goes with the flow, though: existence is expressed just by saying “X [is] there”, using the same locative adverb as you would for deixis. Much like English, in fact.