The hornet's nest of grammatical complexity

[zompist]

one can pretty easily make a comprehensible English sentence even if the syntax is utterly mangled

True, English is a good language to speak badly, as they say, [...]

How people react to mangled language would be a study in itself, and probably have a lot of surprises.

In general I think native English speakers are very judgmental about bad English-- just as much as French speakers are about bad French.

On the other hand, I've had very pleasant experiences trying out Italian, Spanish, Portuguese, and Mandarin. But here, being a US tourist or in-law helps; my understanding is that Spaniards and Italians are far less welcoming to (say) Latin Americans.

[bradrn]

one can pretty easily make a comprehensible English sentence even if the syntax is utterly mangled

True, English is a good language to speak badly, as they say, but dealing on a daily basis with non-native English speakers of all parts of the world on various fora, I can say that mangled syntax, and the wrong choice of related words, can make even topic-constrained questions a real challenge to understand.

I agree, but I do think it’s interesting that you feel the need to call out incorrect word choice specifically. I wonder if the syntax would continue to be as confusing if they used more appropriate words.

creating an acceptable sentence in X is more difficult than doing so in Y.

I think this is a point that we should not leave unexamined. Grammar is not just about what sequences are incorrect. In English, there are countless strings of words that would make a correct sentence, just not the one you're trying to convey. In the sentence "Children, have you run races" any of the words could be removed and leave behind a correct sentence, at least in some dialect. This makes it trivially easy to make a sentence in English that will mean something, whereas in German or Spanish the lack of concord will quickly render any randomly generated sentence meaningless. But if anything this is a sign of greater complexity, not lesser.

Really? Having fewer rules about which words are required makes a language more complex? I’m struggling to see the logical connection here.

As for the rat's nest of modals not counting as grammar because they are "lexicon," I ask you if an L2 learner of English would understand the syntax of the word "can" simply by being told it's the English version of poder. The whole point is that there is a lot of grammar going on with the piddly function words of English.

You misunderstand me. I never said in any way that modals are not grammar because they are lexicon. Indeed, my point was that there probably isn’t any firm dividing line between the grammar and the lexicon, and most words and constructions combine both. My argument would perhaps be better rephrased as follows: the complexity of language for L2 speakers lies mostly in the semantics of individual words, rather than in the grammar of constructions. (Indeed, the syntax of ‘can’ isn’t terribly complicated.) This seems basically the same as what zompist is saying:

Non-linguists talking about "grammar" usually mean only the stumbling blocks you get in the first 3 months of language learning: different writing systems; those horrible tables of morphology; basic stuff like gender. But it doesn't take 3 months to learn a language (in school)-- more like 3 years. And most of what you're learning is constructions and lexicon.

[Moose-tache]

Really? Having fewer rules about which words are required makes a language more complex? I’m struggling to see the logical connection here.

The point is that rules aren't just about what's required, but about how to interpret something.

Since emotional investments around natlangs are unavoidable, I will illustrate with Binglish and Blerman.

Blerman marks every noun and adjective for number and gender, so if I see a phrase like "igneous(masc.pl) rocks(fem.sg)," I can stop thinking about it immediately; I know it's a typo. It has no possible meaning. The rules only allow certain combinations to be grammatical, so this one is gibberish. Meanwhile, Binglish allows basically any combination of words, so "Igneous rocks" could mean "rocks formed by volcanism" or "that new club on 4th street, Igneous, is really good!" In so far as English speakers share a rulebook on how best to interpret the Byzantine network of possible combinations, English does have rules for how to parse sentences. But these rules are infinitely more complex than "Does it match the one allowable form? If yes, accept one allowable meaning; if no, return to sender." Rules dictating which of an infinite set is the best one, rather than rules dictating which single combination is allowable, are still rules. You can try to shunt all of that to the side and say "Aw, that's just pragmatics," but try that when teaching English grammar to an ESL student, and you'll quickly find yourself out of a job.

[alice]

Of course, it's worth establishing what "minimal grammatical complexity" actually is and what it looks like.

[bradrn]

Blerman marks every noun and adjective for number and gender, so if I see a phrase like "igneous(masc.pl) rocks(fem.sg)," I can stop thinking about it immediately; I know it's a typo. It has no possible meaning. The rules only allow certain combinations to be grammatical, so this one is gibberish. Meanwhile, Binglish allows basically any combination of words, so "Igneous rocks" could mean "rocks formed by volcanism" or "that new club on 4th street, Igneous, is really good!" In so far as English speakers share a rulebook on how best to interpret the Byzantine network of possible combinations, English does have rules for how to parse sentences. But these rules are infinitely more complex than "Does it match the one allowable form? If yes, accept one allowable meaning; if no, return to sender." Rules dictating which of an infinite set is the best one, rather than rules dictating which single combination is allowable, are still rules. You can try to shunt all of that to the side and say "Aw, that's just pragmatics," but try that when teaching English grammar to an ESL student, and you'll quickly find yourself out of a job.

I think you’re conflating ‘complexity’ with ‘ambiguity’. The former term, as we’ve been using it, relates to the difficulty of creating an intelligible sentence to say what you want to say. The latter relates to the difficulty of understanding a sentence you’ve seen or heard. (If you really want to keep the complexity terminology, you could call them ‘production complexity’ and ‘comprehension complexity’ respectively.) I tend to see these as opposed: the more overspecified a sentence is, the easier it is to understand what was said.

[zompist]

I think you’re conflating ‘complexity’ with ‘ambiguity’. The former term, as we’ve been using it, relates to the difficulty of creating an intelligible sentence to say what you want to say. The latter relates to the difficulty of understanding a sentence you’ve seen or heard. (If you really want to keep the complexity terminology, you could call them ‘production complexity’ and ‘comprehension complexity’ respectively.) I tend to see these as opposed: the more overspecified a sentence is, the easier it is to understand what was said.

I don't think it's very clear that these are two separate things. For computer languages, when you have a generator, you pretty much also have a parser: that is, rules for one process are also rules for the other. For human languages, whether you learn them as an adult or as a child, it's normal to progress in both and to end up mastering both at about the same time.

[Ares Land]

How people react to mangled language would be a study in itself, and probably have a lot of surprises.

In general I think native English speakers are very judgmental about bad English-- just as much as French speakers are about bad French.

Oh, not in my experience at least. Native English speakers are usually very understanding. (Except for US Customs, but I understand they're rude to everyone.)

[jal]

I agree, but I do think it’s interesting that you feel the need to call out incorrect word choice specifically. I wonder if the syntax would continue to be as confusing if they used more appropriate words.

Well, it mostly comes with the package: seldomly do they use the right words with bad grammar or vice versa, so even if the right words with bad grammar would be better understandable, out in the wild that almost never happens.


JAL

[jal]

It's so cute when Spanish speakers try to claim that Spanish is difficult. It makes me want to tell them that American 8th graders take Spanish because they think it's the easiest foreign language to learn. (Which given the usual alternatives, it is.)

Yeah, they were both adament that Spanish was much more complex than English and "Chinese", because it has genders and morphologically encoded moods etc., then called me linguistically incompetent for disputing their claims :).


JAL

[zompist]

Oh, not in my experience at least. Native English speakers are usually very understanding.

If you speak as well as you write English, that's fine. We're often suckers for a light accent.

(There's a Minecraft Youtuber I follow who has the most delightful accent... some sort of combo of Swedish and Polish.)

I probably put it too strongly though. I was just reacting to the idea that English speakers don't care if you mangle the language. I think English speakers kind of expect everyone to learn English, and well enough to hold a conversation. Whereas (say) Italian speakers are charmed that you learned their little ol' language.

[hwhatting]

I think we're getting somewhere now. As I said, it's not the case that morphology-rich languages necessarily have less complex syntax or auxiliary verb systems or fewer issues of idiomaticity or register. It's not easier to form a correct sentence in Spanish or Russian than in English once you have mastered the morphology. Maybe many learners are so happy that they have mastered these things that they don't care anymore about idiomaticicity, nuance, or register, but that's the same mistake like thinking English is easy because it's low on morphology.
The point is rather, as Moose pointed out, that morphology-rich languages lower ambiguity by overspecifying. In general, I'm not tripped up by English much anymore because I have sufficient practice - sometimes there are garden paths, but they normally get resolved by context. But if there is one language among those I'm learning that I often have difficulties decoding, it's Bahasa Indonesia, which has even less morphology than English. I can look up all the words in the dictionary, but still frequently cannot decipher the meaning of the sentence. This happens much less frequently to me in morphology-rich languages (here the problem is more frequently "what word is that a form of", especially in languages with highly irregular morphologies like Arabic or Classical Greek).

[bradrn]

I think you’re conflating ‘complexity’ with ‘ambiguity’. The former term, as we’ve been using it, relates to the difficulty of creating an intelligible sentence to say what you want to say. The latter relates to the difficulty of understanding a sentence you’ve seen or heard. (If you really want to keep the complexity terminology, you could call them ‘production complexity’ and ‘comprehension complexity’ respectively.) I tend to see these as opposed: the more overspecified a sentence is, the easier it is to understand what was said.

I don't think it's very clear that these are two separate things. For computer languages, when you have a generator, you pretty much also have a parser: that is, rules for one process are also rules for the other. For human languages, whether you learn them as an adult or as a child, it's normal to progress in both and to end up mastering both at about the same time.

Not sure I agree. Even from a CS perspective, there are very few libraries which will give you both a parser and a printer from one grammar. And then there’s the dictum of ‘being liberal in what you accept and strict in what you output’, which presupposes that these are two entirely different modules.

From a linguistic perspective, there's also plenty of reason to believe that these are separate. For one thing, there’s the common phenomenon of passive mastery of a language, where people exposed to a language can understand it but not speak it, which in and of itself is pretty strong evidence that these are two separate things. And then there’s the converse, where an L2 speaker can produce sentences but can’t understand what others say. (This is what hwhatting reported with Indonesian.)

The point is rather, as Moose pointed out, that morphology-rich languages lower ambiguity by overspecifying. In general, I'm not tripped up by English much anymore because I have sufficient practice - sometimes there are garden paths, but they normally get resolved by context. But if there is one language among those I'm learning that I often have difficulties decoding, it's Bahasa Indonesia, which has even less morphology than English. I can look up all the words in the dictionary, but still frequently cannot decipher the meaning of the sentence. This happens much less frequently to me in morphology-rich languages (here the problem is more frequently "what word is that a form of", especially in languages with highly irregular morphologies like Arabic or Classical Greek).

This is a very interesting comment — I had never thought of languages like Indonesian as being so difficult to understand. But this is still a reflection of ambiguity, not complexity. Indeed, it is precisely because Indonesian is simple to speak that it is so difficult to understand.

[zompist]

From a linguistic perspective, there's also plenty of reason to believe that these are separate. For one thing, there’s the common phenomenon of passive mastery of a language, where people exposed to a language can understand it but not speak it, which in and of itself is pretty strong evidence that these are two separate things. And then there’s the converse, where an L2 speaker can produce sentences but can’t understand what others say.

No, it's evidence that the skills need to be practiced separately-- not that the linguistic knowledge is different.

How do you think a child learns how to talk? Well, if you're Chomsky you think it's almost all genetic, but let's ignore him. The child learns by listening. The underlying knowledge is the same, or deeply intertwined. The kid learns how to say "Mommy" after hearing "Mommy" hundreds of times. Same with every more difficult construction. You learn to recognize the sounds and the meaning, and with that very information, you know how to produce them.

Your position is like that of the cartoonist Walt Kelly, who had characters who could write but not read, and others who could read but not write. It's a joke, because we know that's not how reading and writing work. If you can do one, you can do the other. But of course the physical parts of each process need to be practiced, especially the production part.

Now, I'm very skeptical about the conduit metaphor-- that there is some sort of Mentalese that's dehydrated into English by a production component, passed to the listener, and rehydrated back into Mentalese. However, whatever is in the speaker's head when they start to speak, is probably very close to what's in the listener's head when they finish... otherwise we'd question if the communication happened, or the two people know the same language.

And again, this is exactly how computer languages work. You can specify how the language works, and that specification is used by the programmer to write code, as well as by the compiler writer to process the code.

I'll grant you that details of the processes differ. The articulatory process for pronouncing a [t] is not very much like the acoustic process for hearing a [t]. But, you know, they happen at the same time, and babies repeat them over and over, practicing both listening and pronouncing.

[zompist]

For one thing, there’s the common phenomenon of passive mastery of a language, where people exposed to a language can understand it but not speak it, which in and of itself is pretty strong evidence that these are two separate things. And then there’s the converse, where an L2 speaker can produce sentences but can’t understand what others say.

Also, let's look at these situations more closely. What is lacking? I'd argue, mostly process-independent linguistic knowledge.

When you can't understand what the other person says, this is usually clear: they are using words or constructions you don't know. If you knew them, you'd probably recognize them. Let's take an example: you know some French, and you're reading a French comic:

Heu... grâce à ma super-ouie, j'peux tout d'suite te dire que ça va être coton!!

There's a lot of possible stumbling blocks here, but they're all things you might not have known, e.g.
* why the author insists on leaving out some of the e's
* possibly "grâce" if you take it as English "grace"
* maybe the word "ouie"?
* the idiom "être coton"

(If you're curious, it means "Uh, thanks to my super-hearing, I can tell you right away that it'll be tricky!")

But once you know and understand these things, there's no reason you can't say them, too.

As for being able to speak when you can't understand all spoken speech-- you can of course say things yourself because you know what you're saying, you cannot use a construction you don't know. Other people will say things that are new to you all the time, that's precisely why they're hard to follow.

Another difficulty can be understanding what people say in real time. Grabbing another French book from my shelf, the first word is "Doukipudonktan?". It happens to be a nice representation of what it can sound like when you're unsure of your listening skills. There's a bunch of sounds, but they refuse to put themselves into words.

It takes practice to get this skill into shape. But it's also affected by what you know. This one is "D'où qu'ils puent donc tant?" (How is it they stink so much?) The thing is, the first thing you don't know will make the rest unrecognizable too: maybe the "d'ou que" construction, maybe the verb "puer", maybe "tant" which is a simple enough word but doesn't work like its English equivalents.

Now, the person who can understand a language but not speak it. I'm a little skeptical about this person, because I suspect the truer description is that they know some of the language-- maybe highly restricted in register and subject. They can't speak well mostly because of what they don't know: trying to say something, they'd immediately run into things they are not sure about. Plus, I suspect, there's the fear of feeling foolish, common in adult speakers.

Listening to a language, the listener can also guess a lot; and it's the speaker who has to remember all the morphology. But that's just another way of saying that the listener doesn't know the language that well-- or they wouldn't have to guess and stumble over the endings.

[vegfarandi]

Plus, I suspect, there's the fear of feeling foolish, common in adult speakers.

That's definitely a big part of it.

For me personally, I know Spanish well enough that I can read most wikipedia articles without much trouble and the word look-up feature in MacOS and iOS is fantastic when I see unfamiliar words. But listening is a very different story. I have a very hard time understanding most Spanish in the Americas. Then Elite came out on Netflix and I realized my ear is only trained on Madrileño. All these years since high school and I still understand the Spaniards' prestige dialect just fine.

[Linguoboy]

I was an adult before I fully appreciated the difficulty learning to distinguish count nouns and mass nouns poses for L2-speakers of English.

Or, if their first language is German, the entire concept of adverbs as something distinct from adjectives.

One of my standard rejoinders to "English is so easy!" is, "So what are the rules for adverb placement in English?" (This is one of the top sources of errors when I'm correcting the English of intermediate to advanced L2-speakers.)

[Travis B.]

I was an adult before I fully appreciated the difficulty learning to distinguish count nouns and mass nouns poses for L2-speakers of English.

Or, if their first language is German, the entire concept of adverbs as something distinct from adjectives.

One of my standard rejoinders to "English is so easy!" is, "So what are the rules for adverb placement in English?" (This is one of the top sources of errors when I'm correcting the English of intermediate to advanced L2-speakers.)

I'm a native speaker who's shelled out for the paid version of English, and I still can't spell out any clear rules for adverb placement in English!

[Kuchigakatai]

One of my standard rejoinders to "English is so easy!" is, "So what are the rules for adverb placement in English?" (This is one of the top sources of errors when I'm correcting the English of intermediate to advanced L2-speakers.)

Indeed... I still distinctly remember studying the syntax of English adverb placement for various afternoons, about a couple years after moving Canada, after I got an essay back with a bunch of corrections in precisely this area. This, along with the uses of prepositions, is actually the topic for which I was helped by explicit grammar learning the most (for most everything else, sheer exposure did most of the job).

[bradrn]

From a linguistic perspective, there's also plenty of reason to believe that these are separate. For one thing, there’s the common phenomenon of passive mastery of a language, where people exposed to a language can understand it but not speak it, which in and of itself is pretty strong evidence that these are two separate things. And then there’s the converse, where an L2 speaker can produce sentences but can’t understand what others say.

No, it's evidence that the skills need to be practiced separately-- not that the linguistic knowledge is different.

Ah, I see the problem. I’m not arguing that they require fundamentally different linguistic knowledge to perform. I was just saying that they’re different skills, and it’s entirely possible to be good at one but bad at the other.

In any case, I note that we’ve drifted off the original topic a bit. We started off by disputing whether ‘complexity’ and ‘ambiguity’ are separate things, and we still haven’t resolved that. As far as I’m concerned, this has little to do with whether speaking and understanding a language require different linguistic knowledge: a language can be complex, not complex, ambiguous or not ambiguous even if you speak and understand it perfectly.

Now that I think of it, perhaps I should give some examples of what I mean by those terms. This would be an example of a language which is ambiguous but not complex:

1, lion eat
‘as for me, I ate a lion’ or ‘as for me, I ate the lion’ or ‘as for me, a lion ate me’ or ‘as for me, the lion ate me’ or ‘as for us, we ate a lion’ or ‘as for me, I ate with the lion’ or ‘as for me, I ate lions’ or ‘as for me, I’m eating a lion’ or ‘as for me, I’ll eat a lion’ or…

This would be an example of a language which is complex but not ambiguous:

1.SG.MASC-TOP eat-hand-VEN-PST-PFV-1.SG.MASC>3.SG.FEM lion-INDEF-SG.FEM-ACC
‘as for me, I [masculine] ate a lioness with my fingers some distance away’

[zompist]

In any case, I note that we’ve drifted off the original topic a bit. We started off by disputing whether ‘complexity’ and ‘ambiguity’ are separate things, and we still haven’t resolved that.

I'm not sure I agree with the claim that English is more prone to ambiguity. Yes, we have a whole thread of lots of amusing examples. But-- we're English speakers, we enjoy that and we read enough stuff that we can note down fun examples. If it was really the case that every other sentence of English was ambiguous, the examples wouldn't be so remarkable.

Now, Sanskrit is highly morphologically marked: complex verbs, a full case system, gender, etc. So you or Moose would expect it to lack ambiguity, right? Yet there's an epic poem, Dvisandhana, which cleverly uses synonyms to simultaneously retell the Mahabharata and the Ramayana.

Or take Chinese, where poets love to play with understatement and ambiguity. E.g. Wáng Wéi's most famous poem:

空山不见人
empty mountain not see person
但闻人语响
however hear person words/speak sound
返景入深林
return brightness/view/situation enter deep/thick forest

You could write a book on how to translate these 15 syllables-- indeed, Eliot Weinberger and Octavio Paz wrote such a book. Is Chinese the summit of ambiguity? But you could also write this:

有一座山，空无一人。 我们看不到任何人，但我们听到了声音。 黄昏的光芒穿透了森林深处，再次照耀在森林地面上的青苔上。
There is a mountain that is empty. We couldn't see anyone, but we heard voices. The light of dusk penetrated the depths of the forest and shone again on the moss on the forest floor.

(I'm using Google Translate for the last bit, so I guarantee nothing.)

That is-- if you're actually speaking Mandarin, not trying for poetic concision, you can very easily put in all the things the poem leaves out: narrative time, who is looking around, how exactly the light was shining.

One more example, from Portuguese. It's no Sanskrit, but it has gender and complex verbs. Here's one verse of a song:

É pau, é pedra
É o fim do caminho
É um resto de toco
É um pouco sozinho...

It's a stick, it's a stone / it's the end of the road
It's a rotted stump / it's a little alone

And it goes on like that, a little tone poem about early fall. The method is about the same as the Chinese poem. There's a verb (é = is), but no subject. So using a language with morphological tense, gender agreement, and usually explicit subjects, doesn't prevent you from writing like Wáng Wéi.

Yes, impressionistically, English speakers knowing English well and knowing how to play with English think that English is prone to ambiguity. I'd rather see a much more empirical study. How ambiguous is a real English text (not a text made up for maximum ambiguity)? How does that actually compare to Sanskrit or Mandarin or Portuguese, or other similar languages?