English questions

[bradrn]

I came across the following striking comment in Haspelmath’s (in my opinion excellent) paper The indeterminacy of word segmentation and the nature of morphology and syntax (https://doi.org/10.1515/flin.2011.002, highlighting not in original):

[…] can we define the morphosyntactic word as a ‘maximal uninterruptible string of morphs’? Unfortunately, this is not possible because the definition is too loose: There are many uninterruptible combinations that are not normally considered words. For example, in (11), the combinations linked by a plus sign are not interruptible by anything else:
(11)

1. both+my parents
2. even+Kim understands it
3. very+good food

To rule out the A+B combinations in (11) as complex words, one would need to invoke another criterion, e.g. the criterion of non-selectivity […] Since both and even combine with words of diverse classes, they would not count as affixes, despite being very tightly (and uninterruptibly) combined with their hosts in (11a–b) (in 11c, not even the non-selectivity criterion gives the desired result, because very is selective, combining only with adjectives; I know of no way of arguing against inflectional prefix status of very).

Now, I know Haspelmath to be an excellent and careful analyst in general, and I tend to agree with his conclusion in this article that we can’t yet really distinguish morphology and syntax… but can it seriously be true that there is no way to argue that very is not an affix? I know I can’t think of a reason why it shouldn’t be an affix, but then again I’m somewhat terrible at this sort of reasoning.

[Travis B.]

I came across the following striking comment in Haspelmath’s (in my opinion excellent) paper The indeterminacy of word segmentation and the nature of morphology and syntax (https://doi.org/10.1515/flin.2011.002, highlighting not in original):

[…] can we define the morphosyntactic word as a ‘maximal uninterruptible string of morphs’? Unfortunately, this is not possible because the definition is too loose: There are many uninterruptible combinations that are not normally considered words. For example, in (11), the combinations linked by a plus sign are not interruptible by anything else:
(11)

1. both+my parents
2. even+Kim understands it
3. very+good food

To rule out the A+B combinations in (11) as complex words, one would need to invoke another criterion, e.g. the criterion of non-selectivity […] Since both and even combine with words of diverse classes, they would not count as affixes, despite being very tightly (and uninterruptibly) combined with their hosts in (11a–b) (in 11c, not even the non-selectivity criterion gives the desired result, because very is selective, combining only with adjectives; I know of no way of arguing against inflectional prefix status of very).

Now, I know Haspelmath to be an excellent and careful analyst in general, and I tend to agree with his conclusion in this article that we can’t yet really distinguish morphology and syntax… but can it seriously be true that there is no way to argue that very is not an affix? I know I can’t think of a reason why it shouldn’t be an affix, but then again I’m somewhat terrible at this sort of reasoning.

Just for the record, very can also combine with adverbs - The man very quickly scarfed down the burger..

[bradrn]

I came across the following striking comment in Haspelmath’s (in my opinion excellent) paper The indeterminacy of word segmentation and the nature of morphology and syntax (https://doi.org/10.1515/flin.2011.002, highlighting not in original):

[…] can we define the morphosyntactic word as a ‘maximal uninterruptible string of morphs’? Unfortunately, this is not possible because the definition is too loose: There are many uninterruptible combinations that are not normally considered words. For example, in (11), the combinations linked by a plus sign are not interruptible by anything else:
(11)

1. both+my parents
2. even+Kim understands it
3. very+good food

To rule out the A+B combinations in (11) as complex words, one would need to invoke another criterion, e.g. the criterion of non-selectivity […] Since both and even combine with words of diverse classes, they would not count as affixes, despite being very tightly (and uninterruptibly) combined with their hosts in (11a–b) (in 11c, not even the non-selectivity criterion gives the desired result, because very is selective, combining only with adjectives; I know of no way of arguing against inflectional prefix status of very).

Now, I know Haspelmath to be an excellent and careful analyst in general, and I tend to agree with his conclusion in this article that we can’t yet really distinguish morphology and syntax… but can it seriously be true that there is no way to argue that very is not an affix? I know I can’t think of a reason why it shouldn’t be an affix, but then again I’m somewhat terrible at this sort of reasoning.

Just for the record, very can also combine with adverbs - The man very quickly scarfed down the burger..

I do realise that — but is that enough to disqualify it from affix-hood or inflection-dom? I’m not sure.

But consulting Huddleston & Pullum on this topic, they note very can also occur as an adjective: this very room, or the very edge of the cliff. Whether this should be considered the same word or not, it’s hard to tell — but if we do analyse it that way, that seems like quite a strong argument against it being an affix.

[Travis B.]

But consulting Huddleston & Pullum on this topic, they note very can also occur as an adjective: this very room, or the very edge of the cliff. Whether this should be considered the same word or not, it’s hard to tell — but if we do analyse it that way, that seems like quite a strong argument against it being an affix.

I had forgotten about those usages as well, and yes, if those are the same word then they are a very good argument for that word's non-affix-hood.

[Estav]

I came across the following striking comment in Haspelmath’s (in my opinion excellent) paper The indeterminacy of word segmentation and the nature of morphology and syntax (https://doi.org/10.1515/flin.2011.002, highlighting not in original):

[…] can we define the morphosyntactic word as a ‘maximal uninterruptible string of morphs’? Unfortunately, this is not possible because the definition is too loose: There are many uninterruptible combinations that are not normally considered words. For example, in (11), the combinations linked by a plus sign are not interruptible by anything else:
(11)

1. both+my parents
2. even+Kim understands it
3. very+good food

To rule out the A+B combinations in (11) as complex words, one would need to invoke another criterion, e.g. the criterion of non-selectivity […] Since both and even combine with words of diverse classes, they would not count as affixes, despite being very tightly (and uninterruptibly) combined with their hosts in (11a–b) (in 11c, not even the non-selectivity criterion gives the desired result, because very is selective, combining only with adjectives; I know of no way of arguing against inflectional prefix status of very).

Now, I know Haspelmath to be an excellent and careful analyst in general, and I tend to agree with his conclusion in this article that we can’t yet really distinguish morphology and syntax… but can it seriously be true that there is no way to argue that very is not an affix? I know I can’t think of a reason why it shouldn’t be an affix, but then again I’m somewhat terrible at this sort of reasoning.

The problem with arguing about things like "is X an affix" or "is X a syllable" is that we don't have definite prior knowledge about what these things actually are (or whether they even have a 'real' existence), so it becomes a matter of opinion about whether arguments sound convincing or not.

In the case of 'very', I would argue that the fact that it can attach to phrases composed of multiple coordinated adjectives ("very happy and satisfied" can have the sense "very happy and very satisfied", implying the bracketing [very [happy and statisfied]]) distinguishes it from the behavior of un- in phrases like "unhappy and satisfied", which in non-contrived circumstances cannot have the sense "unhappy and unsatisfied" and so presumably cannot typically be bracketed as *[un-[happy and satisfied]]. But I don't know how I could support my feeling that this attachment at the phrasal level is unprefixlike behavior.

[Travis B.]

In the case of 'very', I would argue that the fact that it can attach to phrases composed of multiple coordinated adjectives ("very happy and satisfied" can have the sense "very happy and very satisfied", implying the bracketing [very [happy and statisfied]]) distinguishes it from the behavior of un- in phrases like "unhappy and satisfied", which in non-contrived circumstances cannot have the sense "unhappy and unsatisfied" and so presumably cannot typically be bracketed as *[un-[happy and satisfied]]. But I don't know how I could support my feeling that this attachment at the phrasal level is unprefixlike behavior.

I was thinking of just that! The fact that you can say [very [happy and satisfied]] is a strong argument for very being an independent word rather than an affix.

[elgis]

In Tagalog, "very beautiful and very kind" can be translated as "napakaganda at napakabait" or "napakaganda at bait". The prefix napaka can modify a word it's not attached to. Although perhaps one could argue that napaka should be a word on its own.

[bradrn]

The problem with arguing about things like "is X an affix" or "is X a syllable" is that we don't have definite prior knowledge about what these things actually are (or whether they even have a 'real' existence), so it becomes a matter of opinion about whether arguments sound convincing or not.

This is exactly the point Haspelmath’s article is trying to make!

But I don't know how I could support my feeling that this attachment at the phrasal level is unprefixlike behavior.

You can’t… Haspelmath even gives an English example: pro-choice and -gun control, plus examples from Turkish, German, Hungarian and Japanese. Apparently the term is ‘suspended affixation’.

[zompist]

For example, in (11), the combinations linked by a plus sign are not interruptible by anything else:
(11)

1. both+my parents
2. even+Kim understands it
3. very+good food

Say what?

both of my parents
even that dimwit Kim understands it
very memorably good food

For the first one ("both my"), these have both been analyzed as Det, and Dets are notoriously hard to stack ("the my books"). Still, as shown, it can be done.

I haven't read the paper, but I'd certainly agree that words and affixes are not neatly separated. These aren't good examples of that, however.

[elgis]

The paper mentions this about uninterruptibility:

As Mel’čuk (1993: 173–174) notes, the semantic relations must remain intact if one applies this test, because otherwise almost any string of morphs could be shown to be interruptible. Another requirement is that one must be sure that the interrupting element is not itself an affix.

Maybe the author considers of as an affix, so it doesn't interrupt both+my.

Regarding very+good, I think the very has different meanings in "very good food" and "very memorably good food."

[bradrn]

The paper mentions this about uninterruptibility:

As Mel’čuk (1993: 173–174) notes, the semantic relations must remain intact if one applies this test, because otherwise almost any string of morphs could be shown to be interruptible. Another requirement is that one must be sure that the interrupting element is not itself an affix.

Maybe the author considers of as an affix, so it doesn't interrupt both+my.

I think it’s more that ‘of’ is the only word which can interrupt here — and ‘interruptability’ of such a limited sort is hardly interruptability at all.

[zompist]

Maybe the author considers of as an affix, so it doesn't interrupt both+my.

You could make a case for prepositions being affixes or clitics, but in general one shouldn't rescue a doubtful idea by adding in another one.

Regarding very+good, I think the very has different meanings in "very good food" and "very memorably good food."

On the other hand, he's cheating a bit by using a value adjective, which generally has to come first in a string of adjectives. Compare "very French food": you can easily make it "very good French food".

To put it another way, there are some strong syntactic order restrictions in English. Determiners are one example; the verbal complex is another. These seem very different from saying that that "unhappy" can't have anything in between the morphemes.

[bradrn]

Regarding very+good, I think the very has different meanings in "very good food" and "very memorably good food."

On the other hand, he's cheating a bit by using a value adjective, which generally has to come first in a string of adjectives. Compare "very French food": you can easily make it "very good French food".

But not without violating the criterion that ‘semantic relations must remain intact’! Otherwise you could claim that pretty much anything could be interrupted by anything.

To put it another way, there are some strong syntactic order restrictions in English. Determiners are one example; the verbal complex is another. These seem very different from saying that that "unhappy" can't have anything in between the morphemes.

I’m not sure I understand what you’re saying here.

[zompist]

Regarding very+good, I think the very has different meanings in "very good food" and "very memorably good food."

On the other hand, he's cheating a bit by using a value adjective, which generally has to come first in a string of adjectives. Compare "very French food": you can easily make it "very good French food".

But not without violating the criterion that ‘semantic relations must remain intact’!

Well, I don't know where he gets that, and I don't have access to the paper, so I'll have to leave it there.

[bradrn]

On the other hand, he's cheating a bit by using a value adjective, which generally has to come first in a string of adjectives. Compare "very French food": you can easily make it "very good French food".

But not without violating the criterion that ‘semantic relations must remain intact’!

Well, I don't know where he gets that, and I don't have access to the paper, so I'll have to leave it there.

That was quoted by elgis:

As Mel’čuk (1993: 173–174) notes, the semantic relations must remain intact if one applies this test, because otherwise almost any string of morphs could be shown to be interruptible. Another requirement is that one must be sure that the interrupting element is not itself an affix.

I also have a PDF copy of the paper, uploaded here: http://bradrn.com/files/Haspelmath2017.pdf

[elgis]

I came across the following striking comment in Haspelmath’s (in my opinion excellent) paper The indeterminacy of word segmentation and the nature of morphology and syntax (https://doi.org/10.1515/flin.2011.002, highlighting not in original):

[…] can we define the morphosyntactic word as a ‘maximal uninterruptible string of morphs’? Unfortunately, this is not possible because the definition is too loose: There are many uninterruptible combinations that are not normally considered words. For example, in (11), the combinations linked by a plus sign are not interruptible by anything else:
(11)

1. both+my parents
2. even+Kim understands it
3. very+good food

To rule out the A+B combinations in (11) as complex words, one would need to invoke another criterion, e.g. the criterion of non-selectivity […] Since both and even combine with words of diverse classes, they would not count as affixes, despite being very tightly (and uninterruptibly) combined with their hosts in (11a–b) (in 11c, not even the non-selectivity criterion gives the desired result, because very is selective, combining only with adjectives; I know of no way of arguing against inflectional prefix status of very).

Now, I know Haspelmath to be an excellent and careful analyst in general, and I tend to agree with his conclusion in this article that we can’t yet really distinguish morphology and syntax… but can it seriously be true that there is no way to argue that very is not an affix? I know I can’t think of a reason why it shouldn’t be an affix, but then again I’m somewhat terrible at this sort of reasoning.

I think very+adjective can be separated by an interjection, as in "this very um... delicious meal." The paper does point out that it's possible to have an interrupting element in the middle of a word:

Another problem is that some combinations which are usually considered as single words can actually be interrupted by free forms, in particular in incorporation patterns. For example, Pawnee allows both (12a) and (12b), and in the latter the element rīks ‘arrow’ interrupts the word in (12a). (These data are also discussed in Julien 2002: 35).

Pawnee (Caddoan; Boas 1911: 31)
a. tā-tu-kut
‘I have cut it for you’
b. tā-tu-rīks-kut
‘I have cut your arrow’

I agree absobloodylutely. Another example is unfuckingbeliavable. But unlike un-, very can stand on its own.

A: This is a very um... tasty sandwich.
B: How tasty is it?
A: Very.

[bradrn]

… unlike un-, very can stand on its own.

A: This is a very um... tasty sandwich.
B: How tasty is it?
A: Very.

Good observation! And that’s enough to bring this discussion to a close, since pretty much everyone (even Haspelmath in that paper) agrees that free occurrence is sufficient to establish wordhood.

[zompist]

I also have a PDF copy of the paper, uploaded here: http://bradrn.com/files/Haspelmath2017.pdf

Thank you!

It's a good overview, and I think he makes a good case of how difficult the "word" concept is. He notes himself that this is not a new observation. (I've discussed it myself, e.g. here.)

My main objection is that he's shown that the "word" concept is tricky, but not that it's meaningless. He talks about fuzzy categories, but doesn't consistently apply this concept to his own examples. As just one example:

so for
example Italian has piang-eva e piang-erà [cry-PST.3SG and cry-FUT.3SG]
‘she cried and will cry’, but not *piang-eva e -erà. But it is well-known that
coordinated prefixes and suffixes are not uncommon in English and related
languages [:] pro- and anti-war

However the affix-word continuum works, these are at very different levels. I'd say that English pro and anti are very close to being full words already-- indeed, pro is hardly a prefix in "Are you pro or con?" This example doesn't at all demonstrate his point.

I do think the passage you originally cited is also a weak point. His claim as stated is simply wrong. I don't think uninterruptability alone can define words, and his points about agglutinative languages are a useful caveat; but it seems to me he dismisses the issue too quickly, and again doesn't take fuzzy categories into account. If there are different levels of interruptability, then he can't simply point to one level of it and conclude that uninterruptability isn't a useful criterion.

I'd also note that his point is sharpest when aimed at Chomskyans and at atttempts to generalize across languages. He doesn't make a good case that (say) you shouldn't have separate morphology and syntax sections in your Latin reference grammar. But outside GG, I don't think any linguists maintain that reference grammars are actually reflective of the mental structure of a language, whatever that is.

[bradrn]

It's a good overview, and I think he makes a good case of how difficult the "word" concept is. He notes himself that this is not a new observation. (I've discussed it myself, e.g. here.)

Yes, indeed… I’ve often thought of Yingzi when reading in this area.

My main objection is that he's shown that the "word" concept is tricky, but not that it's meaningless. […] I'd also note that his point is sharpest when aimed at Chomskyans and at atttempts to generalize across languages. He doesn't make a good case that (say) you shouldn't have separate morphology and syntax sections in your Latin reference grammar. But outside GG, I don't think any linguists maintain that reference grammars are actually reflective of the mental structure of a language, whatever that is.

I don’t disagree with this criticism. He gets closest in the final sections, when he talks about clustering: if it could be shown that these criteria cluster in some way, that would indicate that it would be possible to define a meaningful ‘word’ concept. But no-one has come close to showing that yet… on the contrary, the article rather shows that no-one has come close to giving a sensible cross-linguistic definition for ‘word’!

Relatedly, I think ‘word’ as a concept is fine when defined language-internally. But so does he! On this point, his criticism is that if ‘word’ can only be defined language-internally, then saying ‘all languages have words’ is a uselessly vague statement. Personally, I take the view that ‘word’ is cross-linguistically meaningless until proven otherwise — just like every other term in linguistics.

He talks about fuzzy categories, but doesn't consistently apply this concept to his own examples. As just one example:

so for
example Italian has piang-eva e piang-erà [cry-PST.3SG and cry-FUT.3SG]
‘she cried and will cry’, but not *piang-eva e -erà. But it is well-known that
coordinated prefixes and suffixes are not uncommon in English and related
languages [:] pro- and anti-war

However the affix-word continuum works, these are at very different levels. I'd say that English pro and anti are very close to being full words already-- indeed, pro is hardly a prefix in "Are you pro or con?" This example doesn't at all demonstrate his point.

I actually don’t think you’re disagreeing with the article there. I interpret that and similar examples as illustrating that many kinds of quite different morphemes have been lumped together under the single word ‘affix’.

(There’s also the converse point: that many similar kinds of morphemes have been called ‘words’! I think he missed out a good opportunity to demonstrate some of the supposedly ‘isolating’ languages from, say, Sino–Tibetan and Austronesian, where constructions that look like long particle strings could just as easily be analysed as single grammatical words.)

I do think the passage you originally cited is also a weak point. His claim as stated is simply wrong.

Indeed, hence why I asked about it!

I don't think uninterruptability alone can define words, and his points about agglutinative languages are a useful caveat; but it seems to me he dismisses the issue too quickly, and again doesn't take fuzzy categories into account. If there are different levels of interruptability, then he can't simply point to one level of it and conclude that uninterruptability isn't a useful criterion.

On the other hand, the mere fact that different levels of interruptability exist should suggest that by itself, ‘interruptability’ alone isn’t useful as a criterion. People have made this point more strongly with phonology: the phonological hierarchy is typically postulated to be something like ‘foot < word < phrase’, but in fact there seem to be more (or possibly less) levels than that, which don’t even necessarily line up with each other, so using them as a criterion is cross-linguistically a bit pointless. (See e.g. https://doi.org/10.1111/lnc3.12364 — incidentally another great article on wordhood!)

Also, I don’t really understand what kinds of ‘taking fuzzy categories into account’ you’d like to see here. He’s looked at ten different criteria, and demonstrated that each is neither sufficient nor necessary to define the kinds of things which people have called ‘words’. If those criteria are themselves fuzzy or ill-defined, that just reinforces his point.

[zompist]

Relatedly, I think ‘word’ as a concept is fine when defined language-internally. But so does he! On this point, his criticism is that if ‘word’ can only be defined language-internally, then saying ‘all languages have words’ is a uselessly vague statement.

It's really hard to make any universal statement about languages! It's always worth remembering that Sign languages exist. Does ASL have words? Maybe, maybe not, someone would have to make a case for it.

On the other hand, the mere fact that different levels of interruptability exist should suggest that by itself, ‘interruptability’ alone isn’t useful as a criterion. People have made this point more strongly with phonology: the phonological hierarchy is typically postulated to be something like ‘foot < word < phrase’, but in fact there seem to be more (or possibly less) levels than that, which don’t even necessarily line up with each other, so using them as a criterion is cross-linguistically a bit pointless. (See e.g. https://doi.org/10.1111/lnc3.12364 — incidentally another great article on wordhood!)

Haspelmath is deeply into cross-linguistic comparisons, so that bothers him a lot more than it bothers me.

Also, I don’t really understand what kinds of ‘taking fuzzy categories into account’ you’d like to see here. He’s looked at ten different criteria, and demonstrated that each is neither sufficient nor necessary to define the kinds of things which people have called ‘words’. If those criteria are themselves fuzzy or ill-defined, that just reinforces his point.

Fuzzy doesn't just mean vague. Fuzzy things can be quite scientific... color, for instance. That is, perceptual colors can be neatly analyzed as the intersection of three continua. Or for that matter vowels, which (as phones, not phonemes) take three or four.

The thing is, "make very precise definitions with 100% binary values, then use those for definitions" is precisely not dealing with continua. For the most part, he's showing that binary definitions don't work well. (There's a page where he literally shows these as pluses and minuses.) Again, Chomskyans get hung up on those pluses and minuses, but they're probably wrong about that.

I don't know what an approach using continua would look like, though if anyone wants to use it as a dissertation topic, I'd read that. I expect you'd get a lot of correlations rather than ironclad laws, and that you'd get a lot of clumping that doesn't turn out to be universal.