"Today's morphology is yesterday's syntax"
« previous post | next post »
Wikipedia's entry on Univerbation tells us that "The univerbating process is epitomized in Talmy Givón's aphorism that 'today's morphology is yesterday's syntax'." Google hands the quote off to Gemini, at least in "AI mode", and explains:
The maxim "today's morphology is yesterday's syntax," coined by linguist Thomas Givon, means that grammatical word endings and affixes (morphology) evolve over time from separate, independent words once used together in phrases and sentences (syntax).
I use this familiar quotation in discussing the traditional "levels of language" in an introductory course. and so I wondered about the context of its source. Eventually I tracked it down to Givón's paper in the 1971 Proceedings of the Chicago Linguistic Society, "Historical syntax and synchronic morphology: An archaeologist’s field trip." That volume is not available on line, but I managed to get a copy via inter-library loan, and scanned it to produce the link above.
The paper starts.like this, with two footnotes on the title:
And it ends with section 5.4, which consists entirely of the quotation I was looking for:
5.4. Moral.- To paraphrase an old master: "Today's morphology is yesterday's syntax".
Footnote 1 offers us an attribution that's probably a joke —
1This trip was inspired in part by an an analect variously ascribed to Confucius but most likely emanating from the greatest of all time trippers, Lao Tse, who is reported to have remarked, on the occasion of being nformed that Chinese was an isolating language: "Weep not, my children, for today's syntax is tomorrow's morphology".
]
Talmy re-issued that paper, with some modifications, as a chapter in his 2015 book The Diachrony of Grammar, with the following endnote:
This paper was originally published in the proceedings of the CLS #7 meeting (Givón 1971a). My aim at the time was rather modest–explaining Greenberg's (1966) typological correlations between clausal word-order (VO vs. OV) and bound morphology (prefixal vs. suffixal, respectively). At the time, I had read neither F. Bopp (1833) nor H. Paul (1880) nor A. Meillet (1921) nor Jespersen (1917), and thus set out rather innocently to re-invent grammaticalization from scratch. The paper argued the grand theme that synchrony is largely explained by diachrony. The other major explanatory parameters of grammar–communicative function, the neuro-cognitive processor and language evolution–were grafted onto my evolving perspective a few years later (Givón 1979; see chs 2,3,4, below). While the main thrust of the original paper seems sound, I have removed several examples that seem, in retrospect, either dead wrong or irrelevant. I also dispensed with some of the more embarrassing formal discussion of ordered transformational rules and overly abstract syntactic structures.
Footnote 1 is then introduced with the disclaimer "The original footnote, perhaps forgiven in view of the Zeitgeist, went as follows:".
Leaving ancient philosophers out of it, we should add that today's phonetics is tomorrow's phonology, and today's pragmatics is tomorrow's semantics.
And so on…
Update — Jerry Packard sent a link to his 2008 article on "Chinese as an Isolating Language".
J.W. Brewer said,
September 6, 2026 @ 1:36 pm
I see that Prof. Givon was and is over a decade older than Prof. Liberman and his fellow Boomers, and whether the then-Zeitgeist really licensed someone of that advanced age to talk like that as of 1971 seems a close question. But perhaps an academic appointment at UCLA created additional complications?
I am separately, however, struck by "Lao-Tse" for the trippy* sage otherwise known as Lao-Tzu (standard Wade-Giles) and Laozi (PRC-preferred pinyin). This is not a hippie-era innovation, because you can find it in older texts. I'm not sure if it reflects a variant romanization scheme that has fallen out of more general use or a variant sense of how the character 子 was read. If pronounced in modern Mandarin as ㄗㄜ rather than the usual ㄗ, "Tse" is exactly what Wade-Giles would give you.
*Or, at a minimum, perceived as trippy by many folks in Western college towns circa 1971, who understood that by contrast Confucius was a pretty uptight bourgeois figure reminiscent of ones own parents.
ulr said,
September 6, 2026 @ 3:03 pm
There are more than just two transliteration schemes of Mandarin used in the past two centuries. There is a book (published in Germany, I think) listing dozens of transliteration schemes for the various European languages (German, English, French, Russian, …).
The 1961 Reclam edition (translated by Günther Debon) calls the author Lao-tse and the book Tao-Tê-King, with a note by the publisher saying that these forms are used because they are the "herkömmliche Schreibweise"; Debon himself used a different transliteration, where the putative author is Lau-dsĕ.
So my guess is that Givon originally read Lauzi in a German edition and remembered the conventional German spelling.
Jerry Packard said,
September 6, 2026 @ 3:37 pm
I have been an admirer of Prof. Givón and his iconic statement on grammaticalization for some time, but it is absolute fantasy that anyone at the time of Confucius/Laozi would have been bandying about the concepts -much less the words – isolating, syntax and morphology.
Stephen Goranson said,
September 6, 2026 @ 3:40 pm
Today's A is yesterday's B lost fashion to X is the new Y?
JPL said,
September 6, 2026 @ 7:53 pm
The origin of bound forms in the evolution of syntactic structure (the internal structure of the sentence) as developing from free forms at an earlier stage. (Although he seems to claim a similar process in the evolution of the morphological structure of lexemes.) It would be interesting to see how this hypothesis applies to polysynthetic languages, like the Native American languages in the American Northwest. Givon probably considered that. What did he say? (He was at Oregon.)
A related question concerns the historical evolution of the internal structure of the proposition. How did the distinctions in content evolve, what was the mechanism? (Like the distinctions in the elements of the proposition that will be effectively expressed by the nominal and verbal elements of the sentence structure.)
JPL said,
September 6, 2026 @ 7:56 pm
OOPS! That should be "… the origins of morphosyntactic structure …" (in the first sentence).
JPL said,
September 6, 2026 @ 8:15 pm
Oops again! That should be, "… the evolution of morphosyntactic structure …"
(I wonder if we could get an editing option in here.)
Olaf Zimmermann said,
September 7, 2026 @ 3:38 am
@ulr
You'll likely find that late 19th century German translations – by Jesuits, who else – simply called him 'Lao'.
Kai von Fintel said,
September 7, 2026 @ 6:19 am
"That volume is not available on line" … well, yes it is. The page at https://www.chicagolinguisticsociety.com/list.html links to a publicly accessible Google Drive with all proceedings from CLS 4 to the most recent proceedings from CLS 60. A treasure trove of linguistics.
Chris Button said,
September 7, 2026 @ 7:51 am
I haven't read the work by Givón, but i might add that it is a circular rather than linear process (as touched upon in the comments to Caveman Linguistics
I'm also reminded of Y. R. Chao's comment that the monosyllabic myth is "one of the truest myths in Chinese mythology."
Indeed. So why does so much work on historical linguistics seem oblivious to this reality?
Chris Button said,
September 7, 2026 @ 8:02 am
I think I mean "recurring" rather than "circular"
Jerry Packard said,
September 7, 2026 @ 8:05 am
For more on ‘isolating language’ in Chinese I’d recommend Packard, J.L. (2006). ‘Chinese as an Isolating Language.' In Brown, K. (Editor-in-Chief) Encyclopedia of Language and Linguistics – 2nd Edition. Oxford, UK: Elsevier. Volume 2, 355-358
KevinM said,
September 7, 2026 @ 11:38 am
But beware. Today's scholarly citation may be yesterday's joke footnote.
Jerry Packard said,
September 7, 2026 @ 12:27 pm
Yes so true.
JPL said,
September 8, 2026 @ 1:21 am
@Chris Button:
How about "cyclic"?
Chris Button said,
September 8, 2026 @ 5:17 am
@ JPL
Good one, although doesn't that require a return to the same state each time?
Johnn Finkbiner said,
September 8, 2026 @ 3:39 pm
@Chris Button: I don’t think so. The seasons are cyclic even though years are distinct.
JPL said,
September 8, 2026 @ 6:20 pm
@Chris:
I think the notion of the cycle can be applied to a process consisting of a finite series of changes that is repeated as a series, i.e., a dependency pathway that includes the "last" and the "first" again, and can focus on only the properties that you're interested in, not everything about each end state. So, "becoming more inflectional- becoming more isolating- becoming more inflectional- becoming more isolating", and so forth. I would call a change that involves "returning to the same state each time" a reversible event. Which form best describes the facts I wouldn't know.
Andreas Johansson said,
September 9, 2026 @ 2:32 am
The English "Saxon genitive" has gone from inflection to clitic ("the man I was with's wife"), so evidently yesterday's morphology can become today's syntax.
J.W. Brewer said,
September 9, 2026 @ 10:54 am
To rather belatedly follow-up on ulr's earlier comment, Givon need not to have gotten the "Lao Tse" spelling from a German source, because, for example, es.wikipedia asserts that "En español es muy utilizada la adaptación tradicional Lao-Tse (con las variantes Lao-Tsé, LaoTse, Lao Tse, etc.), similar a la transcripción francesa Lao Tseu y a la transcripción Wade Giles británica Lao Tzu." Note also the variation in the final vowel in Cyrillic versions: e.g., Serbian Лао Це versus Ukrainian Ла́о-цзи.
I suspect the variation may relate in large part to the fact that the standard current Mandarin pronunciation of the final syllable involves what traditional analyses call an "empty vowel" or "empty rime." Many sources seem to treat that as /ɨ/ for IPA purposes. But other sources suggest /ɹ̩ / or / ɯ / or say that Sinologists have devised their own glyph which unfortunately they have not yet convinced the IPA authorities to adopt. https://zh.wikipedia.org/zh-hk/%C9%BF And yet others say "joke's on you because it's a syllabic consonant, so stop looking for a vowel symbol because 'empty' really is the right analysis." None of this promotes straightforward consensus among varying romanization or for that matter cyrillicization systems.
David Marjanović said,
September 9, 2026 @ 11:13 am
The same. Indeed, some of the syntax of written French is the morphology of spoken French…
Ukrainian и has various central-ish values, so it's a less bad match than anything Serbian, a pretty classic five-vowel language, has to offer.
Do you have any examples? I've more often encountered the opposite, i.e. works on extant languages describing the origin of the current state as the current state.
Chris Button said,
September 9, 2026 @ 5:23 pm
The fancy term for that is a "synchronic" analysis.
Meanwhile a "diachronic" analysis looks at change over time.
What I meant to say is that phonetics lies behind sound change. Yet "historical linguists" aren't always as well-versed in phonetics as the "field linguists" out there actually listening to sounds.
I suppose "comparative linguist" is a nice catch-all term because it doesn't specify "historical" yet forces the field linguist to be a historical linguist to make any meaningful comparisons.
As for "philologists" (which I suppose is a historical linguist who only looks at texts/inscriptions), I think a knowledge of phonetics can help understand the mechanisms of the documented changes even if there are no actual recordings to be analyzed with spectrograms.
Michael Watts said,
September 10, 2026 @ 12:32 am
Boy, you're not wrong about that.
I was confused over the Packard paper's use of the label 'morpheme', and here's what wikipedia has to say:
This looks ridiculous to me. Native speakers of English will happily tell you that there is no relationship between the words "nation" and "nature". Those words don't share a root. They share an etymological connection, but that's a fact about history, not about the modern language, and as far as I can see it's impossible to defend the idea that native, nation, nature, or innate contains more than one morpheme. (There's a much stronger case for neonate.)
There is no "English root nat(e)". That's not how English works.
Bearing that in mind, I can't really follow the paper's idea that most Chinese words are dimorphemic. Examples aren't given, but I take this to be a reference to words like 喜欢, which is a completely atomic word meaning 'like' (in the sense of 'appreciate'), spelled with characters that in an earlier stage of the language could be used as independent words. But they can't now. 喜欢 is just a single word that's two syllables long.
I'm also bothered by the paper's claim that Chinese measure words are nounal affixes. They can be separated from the noun to which they are notionally "affixed" by arbitrary descriptive clauses. (I did manage to turn up an example from a native speaker: 至少找一个漂亮点的吧 "at least look for one [who is] a little prettier". Given the search requirements of "a measure word, then some other text, then 的", these are more challenging to find than you'd hope.) The noun itself has been dropped from this sentence, but if it were present it would come between 的 and 吧. I fail to see why we wouldn't just think of the measure words as independent words. But if we demand that they be affixes, that analysis surely requires them to be suffixes of the numbers they follow, not "phrasal prefixes" of the nouns that those numbers count. Let's forget about the concept of phrasal affixes.
—
Wrt traditional "empty rimes":
There was a comment on LL in the recent past complaining that 'the empty rime' "is not chartable". I don't really see what the difficulty is supposed to be. All vowels are colored by the sounds that surround them, so it seems straightforward to say that "si" and "shi" can reflect the same vowel phoneme even if when realized in those syllables you see different formants. (Those syllables sound pretty similar to me, actually, but I have trouble getting a good mental grasp on that vowel anyway. Don't take me as authoritative on that.)
To my understanding "xi" is also traditionally viewed as having an empty rime. I don't think identifying the rime of "xi" with the rime of "si" is supportable. I would identify it with the rime of "yi". (And infinite examples of yi/xi being viewed as a real rhyme are available in popular music.) So in that sense I'm against the concept of "the traditional empty rime".
Sticking with "shi" and its sister syllables, there are spoken accents that realize this vowel as what could reasonably be transcribed [ɹ̩], but in standard putonghua it isn't rhotic. Calling it a phoneme, /ɹ̩/, strikes me as weird.
I am not aware of any accent that produces syllabic [ʃ] or [s] with no vowel. There's always a vowel.
Andreas Johansson said,
September 10, 2026 @ 2:46 am
But the traditional view is that "xi" doesn't have the empty rime but that of "yi".
Michael Watts said,
September 10, 2026 @ 3:09 am
(Another thing that bothers me in the 2006 paper is the table listing the pronunciations of items meaning 'song', 'jar', and 'ghost' as kɤ, gɑŋ, and kuei. These appear to be the Mandarin words 歌 ge, 缸 gang, and 鬼 gui, which all begin with the same consonant, what I am informed is unaspirated [k]. The [k] is as close as Mandarin comes to [g]; why is there notionally a difference between 缸 and the other two?)
daniel said,
September 10, 2026 @ 5:53 am
today's morphology is yesterday's syntax" is such a clean way to put it fr. language really just be recycling itself over centuries. handy when you're digging into word origins too: translation dictionary
rashid said,
September 10, 2026 @ 5:56 am
Great post. I'd always heard the Givón quote but never knew it came from a single line at the end of a 1971 CLS paper or that it was framed as a joke about Lao Tse. Funny how a memorable aphorism can outrun its original context. Thanks for tracking down the source
Jerry Packard said,
September 10, 2026 @ 8:58 am
“I fail to see why we wouldn't just think of the measure words as independent words.”
Because a measure word cannot stand alone as an utterance.
“why is there notionally a difference between 缸 and the other two?)”
That is a typo. Thanks for catching it.
Jerry Packard said,
September 10, 2026 @ 9:12 am
@Michael Watts
The definition of morphine is a traditional definition, the smallest unit of meaning. A bound morpheme, like a root, is a morpheme but not a word. A word is a free morpheme, that is, a morpheme that can stand alone as an utterance or part of an utterance.
“I can't really follow the paper's idea that most Chinese words are dimorphemic. Examples aren't given, but I take this to be a reference to words like 喜欢, which is a completely atomic word meaning 'like’…”
Most words in Mandarin consist of two syllables, and most consist of two morphemes. 喜欢 is an exception: it is a single-morpheme word that consists of two syllables, because as you point out, neither 喜 nor 欢 is a morpheme.
David Marjanović said,
September 10, 2026 @ 1:37 pm
True, and I've seen some quite imaginary phonetics in works on historical linguistics (once I deciphered, in some cases, the single-use transcription systems).
Seconded!
And indeed, the official rules of Pīnyīn (that are not often followed in so much detail) actually demand that 一个 be transcribed yīge (without a space… though also without the tone sandhi), as if to claim that it's an indefinite article. This is not extended to any other combination of numeral + classifier, though.
Yeah, the term there is "clitics".
There's always something voiced, but it's not always a vowel. There's a paper that argues at some length that what si, zi, ci contain is a syllabic dental-ish approximant [ð̞̩] – the same sound as a Spanish intervocalic d, just syllabic. Personally (i.e. without equipment to measure native speakers with), I think that sounds reasonable for some accents, quite likely including the most prestigious one(s), even though I wouldn't have guessed it myself.
Syllabic approximants show up elsewhere in Mandarin. The syllables yi, wu, yu are not quite [i u y]; they're [j̩ w̩ ɥ̩].
Chris Button said,
September 10, 2026 @ 6:43 pm
I'm curious what meaningful phonetic distinction you are making here.
A phonological analysis could be /j/ [i], /w/ [u], /ɥ/ [y]
Chris Button said,
September 10, 2026 @ 8:40 pm
I'm curious what meaningful phonetic distinction you are making here.
A phonological analysis could be /j/ [i], w [u], ɥ [y]
Chris Button said,
September 10, 2026 @ 9:04 pm
… which is how bopomofo treats them
Michael Watts said,
September 12, 2026 @ 2:01 am
That is true of many words. Take prepositions. Why should the ability to stand alone be necessary for Chinese measure words, but not for them?
That's very interesting. My sense is that 一个 does indeed see use where it provides more of a sense of indefiniteness than a focus on a hard numeric count of one, much like a(n) in English. My instinct for the pronunciation would be yíge, with 一 pronounced to accommodate the citation tone of 个 at the same time that 个 isn't pronounced in citation tone. But I'm not confident in that.
The obvious problem for calling measure words affixes to numbers is that the measure word used is determined by the noun being counted and any measure word may occur with any number. As far as word order goes, though, the measure word is not tightly attached to the noun but sticks to the number like glue.
Michael Watts said,
September 12, 2026 @ 2:08 am
I was about to comment that this sounded absurd, and then I realized that it was saying approximant and not fricative.
With that in mind, my actual comment is that I don't see a difference between a "syllabic approximant" and a "vowel".
Chris Button said,
September 12, 2026 @ 7:40 am
@ Michael Watts
Indeed.
Take Cantonese m̩ and ŋ̍. Unlike j̩, w̩ as i and u etc., there are no standalone vowel symbols that could be used instead.
This is where it all gets blurry and frankly a bit silly. It's the consonant-vs-vowel dichotomy of the (western) linguistic tradition clashing with the obstruent-vs-sonorant reality of actual language. "Consonant" and "vowel" remain valid concepts, but they should not be the binary distinction that linguists struggle in vain to anchor analyses on.
At least, that's my take.
Chris Button said,
September 12, 2026 @ 8:40 am
Regarding the usefulness of the term "vowel", and by extension "semi-vowel", the syllable structures of Proto-Indo-European (PIE) and Old Chinese (OC) are perhaps useful to compare here.
In PIE, “j” and “w” are usually deceptively treated as “i” and “u” when occurring as the syllabic nucleus.
While a syllable like /pjt/ [pit] is possible in PIE and OC, a syllable like /pnt/ [pn̩t] is only possible in PIE.
The reason is because while "j” and "n" are both sonorants, "j" is an approximant (or more specifically a "semi-vowel"). PIE lets sonorants in general function as the syllabic nucleus, while OC only lets the approximant / semi-vowel subset of sonorants function as the nucleus.
Chris Button said,
September 12, 2026 @ 8:54 am
In the IPA, only the vowels i, u, ɯ, y get their own semi-vowel symbols j, w, ɰ, ɥ. Eventually, I think we can expect a dedicated semi-vowel symbol for the vowel "a" (or perhaps rather ɑ) to be added at least.
Jerry Packard said,
September 12, 2026 @ 9:41 am
“That is true of many words. Take prepositions. Why should the ability to stand alone be necessary for Chinese measure words, but not for them?”
The ‘stand alone’ criterion is not perfect and admits of exceptions (though not many); the best criterion for ‘free morpheme = word’ is that a free morpheme (= word) is a form that can fill a form class (= part of speech) slot by itself. The instances of ‘preposition’ that you speak of fit this criterion. The is called the ‘syntactic criterion’ for word, because it depends on form class slots generated by a theory of syntax. There are around 8 different criteria for defining ‘word’ – see Packard (2000) pp. 7-14.
Jonathan Smith said,
September 12, 2026 @ 11:58 am
Re: Mand. czs[h]i from the standpoint of teaching/learning, Point 1A is that these syllables are characterized by a single articulatory gesture (a principle often at work in Mandarin), i.e., the vocalic segments are homoorganic with the onsets; there is no movement toward e.g. an [ɨ]-type tongue position. BUT the vocalic segments ARE vocalic. Then you can have a transition where you hear/feel frication. I don't know what particular IPA-adjacent representation is best for the "vowels", but Li-Kim's (2014) general description seems to match the intuition: "Ultrasound […] showed that the segments are homorganic with the preceding dental and retroflex sibilants […] an acoustic study showed that they have a periodic waveform and clear formant structures with no inherent frication noise, indicating that they are not ‘voiced fricatives’."
Jonathan Smith said,
September 12, 2026 @ 1:00 pm
Re: thread title, prototypical morphology from the European POV involves inflectional/ derivational affixes, but e.g. Standard Chinese just does not have these. Instead, the great mass of bimorphemic words (most of them disyllabic) fall somewhere on a spectrum between "Latinate-root-pair"-like (cf. perceive, contract) and honest-to-god compound-like (cf. sleepwalk, doghouse), then throw in mixing and matching.
This means a word something like per and a cognate word part like -per- can be related in various complicated ways. E.g. lǐjiě 理解 'understand' might look like a "compound" of lǐ 'straighten up (e.g. hair, clothes)' and jiě 'untie', but it was not literally composed from the modern words… whereas e.g. lǐcǎi 理睬 'pay attention to' is (apparently) actually a newish product of the thoroughly Mandarin items lǐ 'heed' and cǎi 'heed').
NONETHELESS (on a typical analysis) all of the above are still bimorphemic. Same with xǐhuan: of course it's lexicalized and not meaningfully compositional, but in that it's no different than e.g. perceive. And really it's more transparent than that, since (pseudo)classical language among other factors means component parts do retain and/or recover limited kinds of independent application.
Chris Button said,
September 12, 2026 @ 3:41 pm
A "fricative vowel/approximant" is indeed oxymoronic. I think John Wells nails it:
https://web.archive.org/web/20211024004438/http://www.phon.ucl.ac.uk/home/wells/blog0703a.htm
Chris Button said,
September 12, 2026 @ 3:50 pm
[ð̞̩] as dental vs [z̞̩] as alveolar is really splitting hairs.
David Marjanović said,
September 12, 2026 @ 6:09 pm
That's what I was explicitly taught.
Just… listen. If you substitute one for the other, you have a noticeable accent – not the most obvious one, but after three sentences people will notice.
I do: the approximant, syllabic or not, is narrower. It's close to but outside the vowel chart.
This is where I think you're mixing phonetics and phonology. :-)
Another example of different languages treating "sonorants" differently is found in the pitch-accent systems of Ancient Greek and modern Lithuanian. Ancient Greek had a tone distinction for stressed syllables if their nucleus was a long vowel or a diphthong. Lithuanian has one if their nucleus is a long vowel, a diphthong, or a (short) vowel followed by /m n l r/. This seems like an argument to treat the Lithuanian diphthongs spelled e.g. ai, au as short vowels followed by /j ʋ/; but phonetically – I went to Vilnius in July – they really are diphthongs. There is no [j] or [w] or [ʋ] in them, even though Polish* is full of vowels followed by [j] even word-finally. Maybe that was different in prehistoric times and changed when *[ɔː] became uo [uɔ̯] and introduced the category of "diphthong" into the sound system.
* 110 years ago, approximately nobody in or close to Vilnius spoke Lithuanian, and I'm told it still shows.
Some say the symbols ʢ and ʕ already stand for the approximant neighbors of [æ] and [ɑ].
That may be the paper I had in mind. (I've read 2 or 3 on the issue and can't find them right now.)
Chris Button said,
September 12, 2026 @ 8:14 pm
Respectfully, to borrow your words…
If not, you're going to need to be extremely specific.
I would have thought "syllabic or not" is precisely the point from where you could have perhaps launched a specific phonetic argument.
I'm actually really struggling with this too:
And by ʋ, do you actually mean ʊ ?
Chris Button said,
September 12, 2026 @ 8:43 pm
Or because you have /j/ as the counterpoint rather than /ɪ/, you mean that /w/ is more precisely a labiodental [ʋ]?
Michael Watts said,
September 13, 2026 @ 1:52 am
OK, I'm happy to stipulate that the "stand alone" criterion makes no sense and shouldn't be used, and that it's better to identify words by the syntactic criterion.
But that doesn't even begin to address the question I asked, which was why we would want to apply the 'stand alone' criterion to measure words at the same time that we agree it shouldn't be applied to prepositions. It would make no sense if applied to prepositions and it also makes no sense when applied to measure words. If we hold ourselves to the fairly minimal standard of applying the same criterion to every category that we want to label, we'll immediately conclude that measure words are free morphemes, exactly like prepositions are.
But, in English, sleepwalk and doghouse are dimorphemic and perceive and contract are monomorphemic. If a morpheme is the smallest unit of meaning, per-, con-, ceive, and tract cannot be morphemes because they carry no meaning at all.
That's just the same confusion between history and reality that David Marjanović already mentioned. There is no sense in which the modern English word perceive contains more than a single meaningful element.
(Unless, I guess, you want to posit a "null" element potentially indicating e.g. finite non-3sg present tense to contrast with the second meaningful element present in perceived. But that leaves the point undisturbed; in that model perceive would contain one lexical morpheme and one inflectional morpheme, and so would perceived.)
Chris Button said,
September 13, 2026 @ 6:55 am
I was referring to dedicated semi-vowel symbols. For now, for there is no dedicated symbol for the pharyngeal glide, so the inverted subscript breve has to be used (which would be akin to using i̯ for j as a non-syllabic counterpart to i). Also, could you explain why you mention æ above?
Chris Button said,
September 13, 2026 @ 7:16 am
With "It's the consonant-vs-vowel dichotomy of the (western) linguistic tradition clashing with the obstruent-vs-sonorant reality of actual language. "Consonant" and "vowel", I'm making a more general statement about how "vowel" can be defined phonetically (i.e., no obstruction or friction).
But that same definition of "vowel" equally applies to "approximant" consonants as a subset of "sonorant". "Syllabicity" then becomes the recourse for distinguishing vowels from approximant consonants, but that is hardly adequate since all sonorants (approximant or not) can be syllabic.
So, an overarching consonant-vs-vowel distinction in linguistics smacks of an analytical tradition based on (western) ideas of an alphabet with neatly distinguished consonants and vowels. Step outside that tradition, and an overarching obstruent-vs-sonorant distinction seems far more useful in reality.
Chris Button said,
September 13, 2026 @ 7:23 am
Trying again ….
With "It's the consonant-vs-vowel dichotomy of the (western) linguistic tradition clashing with the obstruent-vs-sonorant reality of actual language. 'Consonant' and 'vowel' remain valid concepts, but they should not be the binary distinction that linguists struggle in vain to anchor analyses on," I'm making a more general statement about how "vowel" can be defined phonetically (i.e., no obstruction or friction).
But that same definition of "vowel" equally applies to "approximant" consonants as a subset of "sonorant". "Syllabicity" then becomes the recourse for distinguishing vowels from approximant consonants, but that is hardly adequate since all sonorants (approximant or not) can be syllabic.
So, an overarching consonant-vs-vowel distinction in linguistics smacks of an analytical tradition based on (western) ideas of an alphabet with neatly distinguished consonants and vowels. Step outside that tradition, and an overarching obstruent-vs-sonorant distinction seems far more useful in reality.
Jerry Packard said,
September 13, 2026 @ 8:59 am
@Jonathan Smith
“all of the above are still bimorphemic. Same with xǐhuan: of course it's lexicalized and not meaningfully compositional, but in that it's no different than e.g. perceive.”
喜欢 is a single-morpheme Mandarin word, as are 玻璃 ‘glass’, 葡萄 ‘grape’ and 乒乓 ‘ping-pong’. There are many monomorphemic multisyllabic Mandarin words, but most contain two syllables and two morphemes.
Many would argue that perceive is bimorphemic, because both per- and -ceive are meaningful and highly productive:
per-: perambulate, permeate, percolate, pervade, persist, perforate;
-ceive: conceive, perceive, receive, deceive,
Jerry Packard said,
September 13, 2026 @ 10:09 am
@Michael Watts
“ But that doesn't even begin to address the question I asked, which was why we would want to apply the 'stand alone' criterion to measure words at the same time that we agree it shouldn't be applied to prepositions.”
The reason is because the syntactic rule that constructs the syntactic slot ‘preposition’ (coverb, etc) may be minimally filled by a single form 在给跟对etc. The slot ‘measure’ is not syntactically generated as a syntactically-occurring free form: in other words, it must be augmented (e.g., with a number or determiner) before it is allowed to occupy a slot (namely the Determiner) form class slot.
It could hardly be otherwise: your alternative (that bound morphemes can be considered words) would imply that all syntactic structure building and processing would be morpheme rather than word-based. That’s just not the way we humans produce and understand language.
Chris Button said,
September 13, 2026 @ 1:34 pm
@ Jerry Packard
Without necessarily disputing or supporting your overarching point (I'll let others comment there), wouldn't 喜欢 need to go in a separate category from 玻璃, 葡萄, and 乒乓, whose conponents are inextricably bound as a loanword, a loanword, and an onompatopoeic word?
Jerry Packard said,
September 13, 2026 @ 5:31 pm
Yes they would fall into different categories of origin, but they are all single-morpheme, two-syllable words.
Michael Watts said,
September 14, 2026 @ 5:56 am
Meh. I think a consonant-vs-vowel distinction makes conceptual sense. I just don't think it can be justified on phonetic grounds. Consonant and vowel are phonological concepts. More specifically, returning to my earlier statement, I'm saying that the same sound is called an "approximant" when it appears in the phonological position of a consonant, and a "vowel" when it appears in the phonological position of a vowel. It's somewhat unfortunate that two of these four conceptual slots bear the same name, but I hope you can see what I mean.
I don't put much stock in the idea that the concept of an alphabet drives the distinction. Latin makes no alphabetic distinction between vowels and their related approximants, and when it needs to distinguish /kw/ from /ku/ the alphabet includes a new consonant solely for that purpose. (/kw/ is spelled QV where /ku/ is spelled CV.)
I think you can go plenty far with "/a/ is a legal syllable. Which other single syllables could rhyme with it, if they existed?" to develop the concept of a phonological consonant. If your writing system is a syllabary whose glyphs have no internal structure, that question will still make sense and I submit that it is extremely likely to occur to native speakers without prompting.
Really I was thinking about English prepositions, but I don't think that matters here. I wouldn't usually refer to "prepositions" when discussing Mandarin because Mandarin has two entirely separate word classes that correspond to different functions of the English one.
I can accept the idea that measure words can't occur without a number in front of them. (They can, if you want to be literal about it, but an analysis that says "when no number is explicitly realized, the number 1 is nevertheless implicitly present" is fine with me.)
Whether determiners can substitute for numbers is hazier. It's always possible to realize the number; by far the simplest analysis is that measure words cannot directly follow determiners, but that the implicit 1 is extremely common in that context.
However, I have a hard time following your argument. It's very easy to describe a syntax featuring grammatical slots that get filled by nothing but a measure word, and in fact this is so easy that it is how virtually all instruction in the language is done. You seem to be saying that these slots are somehow unworthy of the name and shouldn't count. Why? I want to observe that prepositions must also be licensed by some other part of a sentence – or in other words, full prepositional phrases can't stand as complete utterances any more than bare prepositions can – but that's just a guess as to something that you might have meant.
And I don't think it makes sense to say that measure words are grammatically bound to determiners as a way to justify calling them affixes of nouns.
I also have some qualms related to the grammatical reduplication of measure words, which seems like a questionable phenomenon if measure words are themselves inflectional affixes.
Well, "many" might make that argument. The only problem is that both claims are transparently false; per- and ceive are neither meaningful nor productive. I'm not saying they're not highly productive. They aren't productive at all. It isn't possible to coin a word using either element.
It's true that many words exist, in the case of ceive necessarily borrowed from French, that contain these elements. That is because their remote ancestors were productive multiple thousands of years ago in another language. But being productive in Latin doesn't make them productive in English. It doesn't even make them productive in French.
If you ask a fluent native speaker (who hasn't been specifically educated as to the historical answer) what the shared element in conceive, perceive, receive, and deceive means, you will get either "I don't know" or an answer that they invented on the spot. (And in that second case, the answer will most likely be obviously incorrect.) And the same thing will happen if you ask them about permeate, persist, and perforate. And precisely because these elements have no meaning, it's impossible to recombine them with other elements.
(I should note here that there are existing English morphemes spelled "de-" ["undo"] and "re-" ["again"] which can be productively prefixed to any verb, but those morphemes are not present in deceive and receive, as can be seen by both the meanings and the pronunciations of those words.)
This makes for a very sharp contrast with sleepwalk, where you're free to replace either element with something else and get a new word. (It works because both elements carry their own independent meaning!) It was trivial to find examples of "sleeplaughing" and "sleepdriving"; I also found "trancewalking" in the sense "walking while listening to trance music".
Jerry Packard said,
September 14, 2026 @ 11:30 am
@Michael Watts
Very good.
Chris Button said,
September 14, 2026 @ 8:48 pm
The Japanese pronunciation of "ikko" (一個) treats yīge as a single morpheme.
I suppose I was thinking more about how the external or onomatopoeic source of 玻璃, 葡萄, and 乒乓 renders any discussion of breaking them up a non-starter. Meanwhile the idea of breaking up 喜欢 at least has a leg to stand on even if it still topples over under scrutiny.
Jerry Packard said,
September 14, 2026 @ 9:41 pm
@Chris Button
I agree it is a non-starter. Should speakers of Mandarin start treating the components as individual morphemes that would be another story.
By saying that the components of 喜欢 have more of propensity to be used as single morphemes is to say that those components are less semantically opaque than the borrowings and onomatopoeics. In Packard (2000) pp. 217-223 I talk about ‘categories of lexicalization’, in which the semantic and grammatical opacity of word components are related to how much speakers are aware of the grammatical and semantic properties of the components.
Chris Button said,
September 15, 2026 @ 10:16 pm
Thanks. I fortunately had access via Cambridge online. I appreciated the logical breakdown into categories.
It also really made me think about etymology and its role. Most specifically, it made me think of part of Roy Andrew Miller's review of Samuel Martin's "The Japanese Language through Time".
Miller complains about Martin's analysis of yamadori "pheasant" as yama "mountain" + tori "bird" because what he really wants to know is the origins of yama and tori individually.
Jonathan Smith said,
September 16, 2026 @ 10:01 pm
Not sure what to say but xǐhuan is bimorphemic like (but more than) perceive. Once upon a time there was two words, then they was combined, with the same historical elements appearing in hundreds of other Mandarin words… like say huānxǐ. "More than" just above because, again, actual classical language / classicizing language / ad hoc goofiness / etc. where xǐ and huān are independent items.
Chris Button said,
September 17, 2026 @ 6:55 am
@ Jerry Packard
It also reminds me of the monumental 5-volume "A Burmese-English Dictionary" (1940-1981), which never made it beyond the single letter အ.
The letter အ can (among other occurrences) essentially create verbal nouns. So, with that single letter, the editors covered an enormous chunk of the Burmese lexicon!
Chris Button said,
September 17, 2026 @ 7:08 am
အ representing ʔə
Michael Watts said,
September 18, 2026 @ 2:35 am
Is it normal to include grammatical derivations as independent dictionary entries? Entries, if necessary, but e.g. the Latin dictionary on my shelf gives me this gloss:
That's the entire entry. It's warranted; there is no way to predict from fero that latus will be a related form. It's also abbreviated; Latin participles inflect like adjectives and most entries include the paradigm — the entry below this one is for the adjective lāt·us, -a, -um — but that information isn't given for the passive participle of fero. There's only one possibility for passive participles.
And even when a paradigm is given, you're expected to handle most of the grammatical analysis yourself. There's no entry for lātīs; it's your job to know that that's the plural dative and ablative form of lātus. If you wanted to make a verbal noun from lātus, that would be ferendum, and there is no entry for that form — despite the fact that it would be separated from fero by several entries — because it's your job to know how to do that.
Do German dictionaries have extremely long ge- sections? Why would this Burmese dictionary have inflated the အ- section like that?
Chris Button said,
September 18, 2026 @ 7:50 am
They tend to end up as individual morphemes. For example, "to do" becomes "work, occupation, job", and "to eat" becomes "food".
Michael Watts said,
September 18, 2026 @ 10:30 pm
That makes sense, but it also implies that the dictionary should have had no more problem accommodating those words than a dictionary of any other language would have accommodating the lexicon of its language. If lexicalized words of this derivation are so numerous, either they're taking away from other words elsewhere in the alphabet (and the task of making a dictionary covering the whole alphabet is more or less unaffected) or the lexicon of Burmese is implausibly large.
I should note for the record, given the thread we're in, that the very concept of a word with such a transparent derivation fusing into an "individual morpheme" would make it impossible to believe that perceive manages to retain two.
(And I find the discussion of semantic opacity in onomatopoeia weird. It seems clear to me that onomatopoeia are completely transparent? People may wonder why the verb for making the sound of a cow is "low", but they never wonder why the noun for the sound itself is "moo".)
Jerry Packard said,
September 19, 2026 @ 9:55 am
Onomatopoeia is semantically opaque in the sense that while it represents and may be somewhat evocative of sounds in the environment, the components of 喜欢 xihuan ‘like’ connect semantically to morphemes in the lexicon, while the onomatopoeic, e.g., 乒乓 pingpang ‘ping pong’ only connects to sounds in the environment, and does so imperfectly and arbitrarily at that: that is, “乒乓” sounds about as much like table tennis activity as does ‘tick-tock’ or ‘click-clack.’ Sound symbolism (phonesthemes) on the other hand has a direct, non-arbitrary connection (therefore transparency) to elements of meaning in the environment.
Yves Rehbein said,
September 21, 2026 @ 1:55 pm
I didn't think transparency is a concept in semantics. Etymological transparency is a different matter. In fact, etymology is two different matters in a trenchcoat or three.
Yves Rehbein said,
September 21, 2026 @ 2:12 pm
And I stand corrected: "In the field of lexical semantics, semantic transparency (in adjective form: semantically transparent) is a measure of the degree to which the meaning of a multimorphemic combination can be synchronically related to the meaning of its constituents." [WP] I still don't think I ever heard the problem stated this way in this day.
Michael Watts said,
October 5, 2026 @ 5:58 pm
Wrt semantic transparency, I like the example of "behead" (and other be- verbs generally).
The verb is well recognized; people know what it means. And the relationship to "head", the noun", is straightforward.
What's the prefix about? No one knows. There is clearly a relationship between "head" and "behead", but the nature of that relationship is completely opaque if you don't already know what "behead" means as its own lexical entry. My position in this thread has pushed for the idea that "behead" contains only one morpheme, and this is something of a strain. I'd be happier with the idea that "behead" is a single morpheme, but that it manages to contain the additional morpheme "head", which native speakers will perceive within it very clearly. I still can't endorse the idea that "be-" constitutes a morpheme. It has no meaning and the only effect of it that I can come up with at all is that verbs bearing that prefix are always transitive.
So there is something of a transparency failure here. In contrast, "redo" consists of two morphemes exhibiting a fully compositional meaning; "re-" means 'again', "do" means 'do', and "redo" means 'do again'. This is a multimorphemic word with perfect transparency.
For something with perfect nontransparency, I'd look to an idiom like "kick the bucket", which is a concept entirely unrelated to kicking or buckets. In that case there are three recognizable morphemes, but none of them make any contribution to the meaning of the "multimorphemic combination".
I'd still have to say that onomatopoeia are, in this sense, perfectly transparent. The fact that "click" means "a sound that sounds like 'click'" is self-explanatory.
ardj said,
October 6, 2026 @ 10:23 am
@Michael Watts
It is ages since I studied linguistics seriously, but I am puzzled by your dismissal of "be-". The OED gives a pretty full account of the develelopment of OE be-, with original sense of 'about', and of its use as prefix where sense 3 has a privative use. Have I misunderstood ?
(And Wikipedia has several ideas about "kick the bucket".)