Analysis of highly analytic languages, oligosynthetic langs...

[Kuchigakatai]

Here is a relevant background discussion between and bradrn on this same topic...

Also a related quip, probably poorly written, I made on this too, which I didn't make further replies about until now (my first reply to bradrn here).

Independently of these discussions from months ago, here I'm replying to a post from Circeus' recent thread, Eliminating verbal adjuncts. I didn't want to hog Circeus' thread for this, so I opened a separate one.

In Cantonese at least, there’s a reasonable argument that these are actual verbs rather than prepositions: any word in the first position of an asymmetric serial verb clause is forbidden from taking an aspect suffix, no matter whether it’s a ‘verb’ or a ‘preposition’.

I think I'm failing to understanding something implied in your comment here. Isn't the fact these "minor verbs" (to use Aikhenvald's term) can't take aspect marking an argument in favour of them being prepositions? (I just want to understand the camp that thinks the opposite as I do... How does this favour them being verbs?)

While we're at it, some definitions by Aikhenvald, if anyone that happens to be reading this thread finds them useful:

In terms of their composition, serial verbs divide into symmetrical and asymmetrical types.

Symmetrical serial verb constructions consist of two or more verbs chosen from semantically and grammatically unrestricted verb classes. Their semantics covers sequences of sub-actions or concomitant actions related to each other; the order of components tends to be iconic. Symmetrical serial verbs tend to become lexicalized.

Asymmetrical constructions include a “major” verb from an unrestricted class and a “minor” verb from a restricted verb class. They may express various grammatical categories, such as direction, orientation, aspect, change of state, adding an argument, and increasing valency. The order of components does not have to be iconic. The minor component tends to grammaticalize into an exponent of aspect or modality, directionality, etc. Then the erstwhile serial verb will lose its status as such.

EDIT: I just found a particularly impressive SVC from Cantonese (Matthews 2006), with several adjuncts in one sentence:

keoi⁵ gon² faan¹ lai⁴ wan² jan² heoi³ bong¹ lei⁵ waan⁴ cin² bei² ngan⁴hong⁴
3s rush return come seek person go help you return money give bank
He came rushing back looking for someone to help you pay back the money to the bank

It's a nice example, etymologically in particular, but I can't help to notice how some of those should probably be paired up as compounds... Some of them have well-established grammatical behaviour that should be mentioned too. I don't know whether Matthews bothers in that source, but that gloss strikes me as a bit more impressive than it needs to be, even if it's technically correct in some way. Specifically:

- 返嚟 faan1lai4 is pretty much a disyllabic term meaning 'to come back', even if literally being return-come, that is, go/come.back-come, sure. Here it serves as a directional complement of 趕 gon2 'to rush'. There are no morphemes that can be added between 返 faan1 and 嚟 lai4, although in other situations 返 faan1 and 嚟 lai4 have independent existences (when 返 faan1 alone means 'to come back, return to [a place]', or when 嚟 lai4 alone means 'to come', in either case the verbs being able to take aspect markers and direct objects on their own...)
- 搵人 wan2-jan2 is a verb-object compound meaning 'look for someone'. That this is one word is particularly noticeable in the tone of the latter morpheme: normally 人 'person' is jan4, but here it has become jan2 because Cantonese often does this switch to a tone #2 in disyllabic compounds of very high frequency, when the second syllable has a lower-register tone (such as tone #4 which is low-falling, [jɐn˨˩]), especially if it's a verb-object compound or a disyllabic noun-noun compound noun. A similar example is 搵錢 wan2-cin2 'to make money' (literally seek-money), in which 錢 'money', normally cin4, has been changed to cin2. Independently, say as a monosyllabic subject, or when modified by a relative clause and nothing else, 錢 cin4 'money' cannot become cin2.
- 去 heoi3 'go' has an established grammatical use, indicating the beginning of a purpose adjunct. You could think of it as a function word meaning 'in order to'. The meaning seems diachronically pretty transparent from 'go', but it's also established... Formal Mandarin allows 用以 yòngyǐ (literally use-take, with an archaic -yǐ "-take" morpheme) to express 'in order to', but spoken Cantonese hardly bothers (also spoken Mandarin). That it is fairly transparent doesn't stop these things from being fairly lexicalized...
- 還錢 waan4-cin2 'to pay back a loan' is another verb-object compound, also with the cin4 > cin2 morphophonological change of highest-frequency compounds.

The following gloss reduces the impressiveness of the gloss, but IMO reflects what is going on grammatically better (and is also more useful to learners, for what it's worth...):
佢趕返嚟搵人去幫你還錢畀銀行。
keui5 gon2-faan1lai4 wan2-jan2 heoi3 bong1 lei5 waan4-cin2 bei2 ngan4hong4
3SG rush-come.back seek-person in.order.to help 2SG pay.back-money to bank
'S/He rushed back looking for someone to help you pay back (the) money to the bank.'

At the same time, it seems clear to me there is a serial verb construction going on here... There isn't any morpheme that is unambiguously a non-verbal function word, and bong1 'help' all the way past the middle sentence is still referring back to keui5 'he/she' from the start of the sentence as its subject.

Regarding 去 heoi3 'to go; away; in order to', I suppose a somewhat similar discussion could be had about 'to' as a preposition meaning 'towards sth, for someone', and the 'to' of 'I help them to do a contribution', except that this English word is clearly a very unstressed, functional, grammatical word in both cases... But is 去 heoi3 here a content word or a function word, a verb or some kind of preposition? Does the difference matter in Cantonese?

As 2+3 Clusivity says here (in a thread I linked to above), intuitively at least it seems problematic to link several pretty different syntactic uses under one category... Like 上 shàng being 'go up, climb up; attend [an event, institution]' (as a transitive verb), 'on X, to the top of X' (as a (pseudo-?)postposition), 'up, upwards' (as a directional complement of verbs, perhaps an 'adverb' but often fairly reasonably called a verb...).

[Kuchigakatai]

So yeah, what do you guys think? Is it valuable to think of this situation as having many monosyllabic morphemes, or some compounds and some monosyllabic morphemes? I guess the topic of Chinese wordhood is thorny for many good reasons...

[bradrn]

Thanks for the updated Cantonese gloss! It’s good to get corrections from someone more knowledgable about the topic.

So yeah, what do you guys think? Is it valuable to think of this situation as having many monosyllabic morphemes, or some compounds and some monosyllabic morphemes? I guess the topic of Chinese wordhood is thorny for many good reasons...

This seems like a perfect opportunity to invoke the typology of wordhood I created a couple of days ago:

… my suggestion is there are three different ways to define ‘word’ (rather than the two which are usually suggested), with their definitions as follows:

• Phonological words have their own stress and minimal word length, and are the domain of (some) phonological rules.
• I have already quoted Julien’s definition of grammatical words. [Let me insert it here for convenience:]

  First, [a property] that distinguish[es] grammatical words from smaller elements [is] … the relative freedom of position within the phrase or the clause. If a string of one or more morphemes has [this property], it is minimally a grammatical word. Further, a grammatical word cannot be freely interrupted by other free forms. A string that can be interrupted by words or phrases is a phrase, unless there are very specific restrictions on the inserted material and the resulting complex behaves as a word on other relevant tests.

• A lexeme is a conventionalised expression for a particular lexical meaning; that is, they are expressions which are used to express a particular meaning more commonly than would otherwise be expected. (This is the extra term I want to introduce; as far as I can tell, it has not been rigorously defined before, though I have seen people refer to the need for it, e.g. in Pawley’s writings on Kalam.)
  • Additionally, some lexemes are non-compositional as well as conventionalised; we could distinguish an extra class of non-compositional lexemes, though I’m not yet convinced of its utility. (On the other hand, what is useful is knowing that anything which is non-compositional must be a lexeme.)

In the terms of this typology, what we have here are compositional lexemes: commonly-used collocations of words with a conventionalised meaning, which nonetheless have internal structure and are analysable in generative terms. This seems clearly applicable to cases like faan¹lai⁴, waan⁴-cin² etc. It’s also what’s happening in Kalam: after I had that discussion with you, I managed to find another paper of Pawley’s, in which he quite clearly states that Kalam SVCs are very lexicalised, to the extent that each time he encounters a new one he writes it in the dictionary. (Don’t remember the paper, sorry.) English also has such lexemes (though not quite as many), e.g. ‘mountain climbing’.

That is: yes, these are conventionalised expressions, of the sort you’d write in the dictionary, which at the same time are composed of monosyllabic morphemes in a regular generative manner.

[Kuchigakatai]

Thanks for the updated Cantonese gloss! It’s good to get corrections from someone more knowledgable about the topic.

Is the gloss not Stephen Matthews's? I mean, he's a real scholar of Cantonese grammar, I wouldn't really dare to say I'm correcting him here, even if I find his gloss (if it is his) to be easily misleading for readers who're not familiar with modern Chinese in general. In some way it is correct after all.

In the terms of this typology, what we have here are compositional lexemes: collocations of words with a conventionalised meaning and wide use to express a particular meaning, which nonetheless have internal structure and are analysable in generative terms. This seems clearly applicable to cases like faan¹lai⁴, waan⁴-cin² etc. It’s also what’s happening in Kalam: after I had that discussion with you, I managed to find another paper of Pawley’s, in which he quite clearly states that Kalam SVCs are very lexicalised, to the extent that each time he encounters a new one he writes it in the dictionary. (Don’t remember the paper, sorry.) English also has such lexemes (though not quite as many), e.g. ‘mountain climbing’.

So that paper said the opposite you were arguing for in that old thread... It is also in line with what I think is normally the case in highly analytic languages, namely having lots and lots of pretty lexicalized phrases. Things get awfully subtle between Mandarin synonyms sometimes, even though the two compounds in question that I might be looking at appear to be simple and transparently derived from their components.

看 kàn 'to see sth/sb' is often used with the meaning 'to visit sb'. 看望 kànwàng, in which 望 wàng is a morpheme that doesn't appear independently but only in compounds meaning something like 'visit/see' or 'hope', also means 'to visit sb', and is listed thus in dictionaries, but (in Taiwan at least) it has the connotation you're visiting someone at a hospital...

By the way, I'd like to hear what seems convincing about Cantonese "coverbs" (minor verbs in asymmetrical SVCs) being verbs due to not being able to take up aspect.

[bradrn]

Thanks for the updated Cantonese gloss! It’s good to get corrections from someone more knowledgable about the topic.

Is the gloss not Stephen Matthews's? I mean, he's a real scholar of Cantonese grammar, I wouldn't really dare to say I'm correcting him here, even if I find his gloss (if it is his) to be easily misleading for readers who're not familiar with modern Chinese in general. In some way it is correct after all.

Aargh, I hate ambiguity… I meant more knowledgable than me, not Matthews!

In the terms of this typology, what we have here are compositional lexemes: collocations of words with a conventionalised meaning and wide use to express a particular meaning, which nonetheless have internal structure and are analysable in generative terms. This seems clearly applicable to cases like faan¹lai⁴, waan⁴-cin² etc. It’s also what’s happening in Kalam: after I had that discussion with you, I managed to find another paper of Pawley’s, in which he quite clearly states that Kalam SVCs are very lexicalised, to the extent that each time he encounters a new one he writes it in the dictionary. (Don’t remember the paper, sorry.) English also has such lexemes (though not quite as many), e.g. ‘mountain climbing’.

So that paper said the opposite you were arguing for in that old thread...

Yes, it did. I was wrong there.

(Though admittedly Pawley hasn’t said anything about productivity as far as I can see, and I don’t even remember him justifying his practice of writing SVCs in the dictionary! So it could easily be possible that Kalam does allow novel SVCs, it’s just that Pawley is misanalysing them… in which case I would actually have been correct in that thread.)

It is also in line with what I think is normally the case in highly analytic languages, namely having lots and lots of pretty lexicalized phrases. Things get awfully subtle between Mandarin synonyms sometimes, even though the two compounds in question that I might be looking at appear to be simple and transparently derived from their components.

As you may have gathered, I’ve been reading a lot about SVCs lately, and this is a recurring pattern with them as well. I remember reading a paper on Mwotlap saying (if I’m remembering correctly) that only 10% of SVCs used in conversation are novel. Kalam does seem an outlier in having no novel SVCs.

看 kàn 'to see sth/sb' is often used with the meaning 'to visit sb'. 看望 kànwàng, in which 望 wàng is a morpheme that doesn't appear independently but only in compounds meaning something like 'visit/see' or 'hope', also means 'to visit sb', and is listed thus in dictionaries, but (in Taiwan at least) it has the connotation you're visiting someone at a hospital...

Hmm… an interesting case. What other compounds does wàng appear in, and what sort of semantic difference does it make when it is included vs. when it is excluded?

By the way, I'd like to hear what seems convincing about Cantonese "coverbs" (minor verbs in asymmetrical SVCs) being verbs due to not being able to take up aspect.

I can only quote Matthews’s arguments here:

It is often argued that in such structures bei² (or its counterpart gei in Mandarin) must be a preposition. One such argument (Zhang 1990: 314) holds that V₂ cannot take aspect marking, which is indeed true of bei² … It can, however, take a verbal particle such as faan¹ ‘back’, one of a number of
particles which follow verbs (Matthews and Yip 1994: 213) … Another proposal maintains that bei² must be a preposition because its object cannot be extracted … The flaw in this argument is that the object of a ‘minor’ verb in an asymmetrical SVC, such as gan¹ in [the sentence ngo⁵ gan¹-hoi¹ keoi⁵ hok⁶ jing¹man²] … cannot be extracted either.

Personally, I find this argument convincing, though of course I know little about Cantonese.

[Kuchigakatai]

In the terms of this typology, what we have here are compositional lexemes: collocations of words with a conventionalised meaning and wide use to express a particular meaning, which nonetheless have internal structure and are analysable in generative terms. This seems clearly applicable to cases like faan¹lai⁴, waan⁴-cin² etc. It’s also what’s happening in Kalam: after I had that discussion with you, I managed to find another paper of Pawley’s, in which he quite clearly states that Kalam SVCs are very lexicalised, to the extent that each time he encounters a new one he writes it in the dictionary. (Don’t remember the paper, sorry.) English also has such lexemes (though not quite as many), e.g. ‘mountain climbing’.

So that paper said the opposite you were arguing for in that old thread...

(Though admittedly Pawley hasn’t said anything about productivity as far as I can see, and I don’t even remember him justifying his practice of writing SVCs in the dictionary! So it could easily be possible that Kalam does allow novel SVCs, it’s just that Pawley is misanalysing them… in which case I would actually have been correct in that thread.)

Honestly, Kalam strikes me as quite a bit similar to Mandarin grammatically (even if it is even more analytical than Mandarin...), that I bet novel SVCs do occur, at least for minor verbs in asymmetrical SVCs with meanings such as 'according to' and 'except for', and maybe some phrases of directions like 'from... to...'.

(As I'm sure you know, minor verbs can be subject to lexicalized selection, especially the most basic ones with meanings like 'from' and 'through', pretty much like prepositions in English: "thank you for your help" could be "come.from your help thank you" or "be.about your help thank you", in which you have to use those specific minor verbs (like how we don't say "thank you on your help" in English).)

Hmm… an interesting case. What other compounds does wàng appear in, and what sort of semantic difference does it make when it is included vs. when it is excluded?

Now that I had some sleep, I think I should've given a better example. 上去 shàngqu lit. 'go up; on; up' + 'go, go away' can be a straightforward directional complement:

登上去 dēng-shàngqu 'climb up'
坐上去 zuò-shàngqu 'sit on sth'
貼上去 tiē-shàngqu 'post sth up by pasting it' (typically a poster, 貼 tiē 'paste sth', up on a wall)
挂上去 guà-shàngqu 'hang sth up' (like a painting on a wall)

But in a few verbs it has the more idiomatic use of forming verbs of perception generally followed by a complement phrase afterwards:

看上去 kàn-shàngqu 'to seem, appear' (as in 'it seems beautiful/dark/green/Japanese/ideal/stupid, it seems to be my friend', 看 kàn 'see sth')
摸上去 mō-shàngqu 'to feel' (as in 'it feels rough/soft/smooth', 摸 mō 'feel sth with one's hand; stroke sth tenderly; grope sth; steal sth')
聽上去 tīng-shàngqu 'to sound' (as in 'it sounds nice/terrible/melodic/one of Mozart's concertos', 聽 tīng 'hear sth')

This is a much better example than wàng.




I guess I have time to answer your question about wàng, but it's a bit irrelevant... My main point is that there are both pretty transparent, novel SVCs in Chinese, but also pretty idiomatic ones too.

希望 xīwàng 'to hope sth, wish for sth' (very common verb, but both roots are necessarily bound roots)

失 shī 'to lose sth'
失望 shīwàng 'to lose hope, to be in despair'

絕 jué (mostly a bound root in some adverbs like 絕大 juédà 'greatly, very' and 絕不 juébù 'never, not at all', sometimes a resultative 'not stop, not be cut' like 繚繞不絕 liáorào-bù-jué "linger-not-stop" 'ever-lingering [pollution...]')
絕望 juéwàng 'to be in despair' lit. "stop/cut-hope"

仰 yǎng 'to look up, move [one's face/head] up'
仰望 yǎngwàng 'to gaze up (full of hope)' (maybe could think of it as 'to look up and hope')

聲 shēng 'sound'
聲望 shēngwàng 'prestige, cachet' (only a noun)

望著 wàng-zhe 'to gaze at sth' (with the 著 -zhe '(particle for stative durative aspect)' necessarily attached; Mandarin has some verbs where aspect markers have been attached lexically, another example is 忘了 wàng-le 'to forget', cf. Latin meminī 'to remember (sth)' which is always in "perfect tenses")

While looking for examples (on the LINE online dictionary and the JuKuu service) I found that in very formal Mandarin it is sometimes used as a transitive verb on its own, 'to hope sth'. I don't want to say I was wrong about 望 wàng being a bound root though, because formal Mandarin just... borrows so much from Literary Chinese, and readily abbreviates like that language does to some extent. You could think of this usage as an abbreviation of 希望 xīwàng above.

此事望迅即處理。
cǐ shì wàng xùnjí chǔlǐ
this thing, hope immediately handle
'We hope this matter will be immediately handled.' (using a formal word for 'this', normally 这 zhè, and a very formal word for 'immediately, right away', normally 立即 lìjí or colloquially 馬上 mǎshàng, the latter literally 'on a horse, on horses')

望 wàng does have an independent existence as a verbal measure word for glances though:

大衛仰臉回首一望，大吃一驚。
Dàwèi yǎng-liǎn huí-shǒu yī wàng, dà chī yī jīng
David turn.up-face return-head one glance, big(ly) eat one surprise
'David turned his face back up for a glance, and was greatly surprised.'
(吃驚 chī-jīng "eat-surprise" is a perfectly normal word for 'to be surprised'.)

I can only quote Matthews’s arguments here:

It is often argued that in such structures bei² (or its counterpart gei in Mandarin) must be a preposition. One such argument (Zhang 1990: 314) holds that V₂ cannot take aspect marking, which is indeed true of bei² … It can, however, take a verbal particle such as faan¹ ‘back’, one of a number of
particles which follow verbs (Matthews and Yip 1994: 213) … Another proposal maintains that bei² must be a preposition because its object cannot be extracted … The flaw in this argument is that the object of a ‘minor’ verb in an asymmetrical SVC, such as gan¹ in [the sentence ngo⁵ gan¹-hoi¹ keoi⁵ hok⁶ jing¹man²] … cannot be extracted either.

Personally, I find this argument convincing, though of course I know little about Cantonese.

That is unrelated. Here he's talking about a certain special kind of construction of some monosyllabic verbs that are placed right after the main verb ("V₂" in his notation here), because it isn't clear if the main verb and this following monosyllabic verb actually form a single compound verb, a verb + preposition, or are indeed in an SVC. If they're a single compound verb, the alleged direct object should be able to be extracted to the start of the sentence (topicalized), because direct objects can pretty much always do that, but since this isn't possible, that proposal said these "V₂"s are prepositions after all, and distinct from minor verbs. But Matthews points out that if this is compared to minor verbs, and if minor verbs are indeed verbs rather than prepositions, then this behaviour of those monosyllabic "V₂"s is irrelevant as an argument, since neither minor verbs nor these "V₂"s alleged to be prepositions can have their NP extracted out.

我哼畀你聽。
o5 hang1 bei2 lei5 teng1
1SG hum give 2SG hear
'I'll hum it for you.' (More literally "I hum for you to hear"; and note 畀 bei2 is often used to mark recipients / indirect objects, rather like 'to, for')

It sometimes seems a bit unclear if this is actually "hang1bei2" instead, i.e. one compound verb. (Personally I'm in the camp that these V₂ words like bei2 are the same as the rest of the "minor verbs", or "prepositions" if you will, whatever word class classification you prefer, so there's no compound such as "hang1bei2".)

EDIT: Actually, I think I'm misreading this quote a fair bit, and confused myself over it, partly because I don't understand the context. I don't get why that proposal Matthews criticizes said these V₂ words are prepositions distinct from minor verbs. What article or book is this from?

[bradrn]

So that paper said the opposite you were arguing for in that old thread...

(Though admittedly Pawley hasn’t said anything about productivity as far as I can see, and I don’t even remember him justifying his practice of writing SVCs in the dictionary! So it could easily be possible that Kalam does allow novel SVCs, it’s just that Pawley is misanalysing them… in which case I would actually have been correct in that thread.)

Honestly, Kalam strikes me as quite a bit similar to Mandarin grammatically (even if it is even more analytical than Mandarin...), that I bet novel SVCs do occur, at least for minor verbs in asymmetrical SVCs with meanings such as 'according to' and 'except for', and maybe some phrases of directions like 'from... to...'.

It’s possible… Pawley does mention these sorts of SVCs now that I check, though again he doesn’t comment on productivity. One difference from Mandarin is that Kalam only has about 150 verbs (most very generic), and its SVCs don’t seem to have too many strict rules for argument sharing, which reduces productivity.

Hmm… an interesting case. What other compounds does wàng appear in, and what sort of semantic difference does it make when it is included vs. when it is excluded?

Now that I had some sleep, I think I should've given a better example. 上去 shàngqu lit. 'go up; on; up' + 'go, go away' can be a straightforward directional complement:

登上去 dēng-shàngqu 'climb up'
坐上去 zuò-shàngqu 'sit on sth'
貼上去 tiē-shàngqu 'post sth up by pasting it' (typically a poster, 貼 tiē 'paste sth', up on a wall)
挂上去 guà-shàngqu 'hang sth up' (like a painting on a wall)

But in a few verbs it has the more idiomatic use of forming verbs of perception generally followed by a complement phrase afterwards:

看上去 kàn-shàngqu 'to seem, appear' (as in 'it seems beautiful/dark/green/Japanese/ideal/stupid, it seems to be my friend', 看 kàn 'see sth')
摸上去 mō-shàngqu 'to feel' (as in 'it feels rough/soft/smooth', 摸 mō 'feel sth with one's hand; stroke sth tenderly; grope sth; steal sth')
聽上去 tīng-shàngqu 'to sound' (as in 'it sounds nice/terrible/melodic/one of Mozart's concertos', 聽 tīng 'hear sth')

This is a much better example than wàng.

This doesn’t strike me as odd at all. Directionals very often have idiomatic meanings — it’s not something that is unexpected for this category. As such I just see these as normal directional SVCs.

I guess I have time to answer your question about wàng, but it's a bit irrelevant... My main point is that there are both pretty transparent, novel SVCs in Chinese, but also pretty idiomatic ones too.

希望 xīwàng 'to hope sth, wish for sth' (very common verb, but both roots are necessarily bound roots)

失 shī 'to lose sth'
失望 shīwàng 'to lose hope, to be in despair'

絕 jué (mostly a bound root in some adverbs like 絕大 juédà 'greatly, very' and 絕不 juébù 'never, not at all', sometimes a resultative 'not stop, not be cut' like 繚繞不絕 liáorào-bù-jué "linger-not-stop" 'ever-lingering [pollution...]')
絕望 juéwàng 'to be in despair' lit. "stop/cut-hope"

仰 yǎng 'to look up, move [one's face/head] up'
仰望 yǎngwàng 'to gaze up (full of hope)' (maybe could think of it as 'to look up and hope')

聲 shēng 'sound'
聲望 shēngwàng 'prestige, cachet' (only a noun)

望著 wàng-zhe 'to gaze at sth' (with the 著 -zhe '(particle for stative durative aspect)' necessarily attached; Mandarin has some verbs where aspect markers have been attached lexically, another example is 忘了 wàng-le 'to forget', cf. Latin meminī 'to remember (sth)' which is always in "perfect tenses")

As for these, as you say, some of them are perfectly normal SVCs, while others have idiomatised. (In my typology, I called those ‘non-compositional lexemes’.) I don’t see any particular problem with that.

I can only quote Matthews’s arguments here:

It is often argued that in such structures bei² (or its counterpart gei in Mandarin) must be a preposition. One such argument (Zhang 1990: 314) holds that V₂ cannot take aspect marking, which is indeed true of bei² … It can, however, take a verbal particle such as faan¹ ‘back’, one of a number of
particles which follow verbs (Matthews and Yip 1994: 213) … Another proposal maintains that bei² must be a preposition because its object cannot be extracted … The flaw in this argument is that the object of a ‘minor’ verb in an asymmetrical SVC, such as gan¹ in [the sentence ngo⁵ gan¹-hoi¹ keoi⁵ hok⁶ jing¹man²] … cannot be extracted either.

Personally, I find this argument convincing, though of course I know little about Cantonese.

EDIT: Actually, I think I'm misreading this quote a fair bit, and confused myself over it, partly because I don't understand the context. I don't get why that proposal Matthews criticizes said these V₂ words are prepositions distinct from minor verbs. What article or book is this from?

Dixon and Aikhenvald, Serial Verb Constructions (Chapter 2). He’s arguing that the minor verb (underlined) in constructions such as these is not a preposition:

lei⁵

you

lo²

take

di¹

PL

saam¹

clothing

bei²

give

keoị⁵

3sg

Bring her some clothes.

[Vardelm]

Dixon and Aikhenvald, Serial Verb Constructions (Chapter 2). He’s arguing that the minor verb (underlined) in constructions such as these is not a preposition:

lei⁵

you

lo²

take

di¹

PL

saam¹

clothing

bei²

give

keoị⁵

3sg

Bring her some clothes.

I'm a bit out of my depth here, but to me the difference seems like a very thin line. That line would be that a "minor verb" could be also used as a major verb, whereas a preposition can take the same place as a minor verb, but can't be used by itself as the main verb of a sentence.

[bradrn]

Dixon and Aikhenvald, Serial Verb Constructions (Chapter 2). He’s arguing that the minor verb (underlined) in constructions such as these is not a preposition:

lei⁵

you

lo²

take

di¹

PL

saam¹

clothing

bei²

give

keoị⁵

3sg

Bring her some clothes.

I'm a bit out of my depth here, but to me the difference seems like a very thin line. That line would be that a "minor verb" could be also used as a major verb, whereas a preposition can take the same place as a minor verb, but can't be used by itself as the main verb of a sentence.

Ah, but it can be the main verb: Ngóh béi chín léih, lit. ‘I give money you’.

[Vardelm]

Ah, but it can be the main verb: Ngóh béi chín léih, lit. ‘I give money you’.

Yeah, so I would agree w/ Matthews.