Barbara Partee has a Log post that so baffled me I have to repost the topic here. It begins:
Today in Seth Cable’s seminar on Montague’s Universal grammar, he gave out a problem set that included the task of adding “Neither Mitt smokes nor Barack smokes” to the little fragment of English that had been developed. And in the discussion of the problem set, it turned out that I was the only one in the class who seemed to have any doubts about whether the sentence “Neither Mitt smokes nor Barack smokes” was grammatical. My own intuition was that it had to be “Neither does Mitt smoke nor does Barack smoke”, though that sounded a little funny too.
My first reaction (since it’s not April 1) was that this is the kind of morass people get into when they spend too much time theorizing about constructed sentences; I couldn’t imagine that any native English speaker, uncontaminated by a linguistics PhD, could possibly think “Neither Mitt smokes nor Barack smokes” was an acceptable English sentence, and “Neither does Mitt smoke nor does Barack smoke” sounds more than just “a little funny,” it sounds like a construction that hasn’t been in use since the 16th century. But reading the comment thread, it seemed that some people disagree, and other examples (just as bad from my point of view) were proposed — e.g., Suzanne Valkemirer said, “If the sentence read ‘Neither does Mitt smoke nor does Barack drink,’ you would presumably find nothing odd about it.” Nothing odd?! I’m not sure whether I’ve fallen irrevocably out of touch with English or these people are all existing in an alternate linguistic universe, so I thought I’d turn to the Varied Reader and as: does “Neither Mitt smokes nor Barack smokes” work for you? What about the other proposed sentences? (N.b.: If you’re confused by my post title, S is a linguistic abbreviation for “sentence.”)
My reactions to some of these are very different depending on whether the example is seen as an introduction to a subject or a response to one, and I wonder if that is confusing some others as well.
For example, the sentence which the Cambridge Grammar gives as a sample of ungrammaticality –
Neither are we trying to keep out immigrants, nor are we favouring the well-to-do
– would be fine if it followed something like ‘We are certainly not trying to prevent local workers from accessing affordable housing’.
(I think. The more I think about this, the weirder it all looks.)
What’s the grammatical way of saying this then?
I can only come up with “Neither Mitt smokes nor does Barack smoke”
As several people said in the Log thread, the normal sentence is “Neither Mitt nor Barack smokes.”
Neither Mitt nor Barack smokes.
Yes, exactly.
I wouldn’t say that the original was ungrammatical; just weird and clunky, which is not the same.
Incidentally, the vaguely irritating choice of personal names of politicians in the examples seems to be characteristic (for some reason) of the linguistic subculture that seeks Universal Linguistic Enlightenment by introspecting about sentences in English.
It’s difficult to think of worse placeholder names than Barack and Mitt; they both invoke various socio-political sentiments that overwhelm the grammatical point you’re trying to make. (And of course Obama is famously a smoker). I’m still not sure what her point was.
I guess “either Mitt wins, or Barack wins” is perfectly fine.
Mitt doesn’t smoke, and so doesn’t Barack
“Neither Mitt smokes nor Barack smokes” is just wrong. “Neither does Mitt smoke nor does Barack smoke” is wronger. “Neither Mitt smokes nor Barack” works for me. (Of course “Neither Mitt nor Barack smokes” is better.)
Both “…neither of producers nor consumers” and “We are neither trying …, nor are we favouring …” are faulty parallelism; whether this is a question of grammar or mere style I know not. I would forgive either in spontaneous speech (though [ii] is suspiciously longwinded to be spontaneous as opposed to written by a sloppy speechwriter). “Neither are we trying to keep out immigrants, nor are we favouring the well-to-do” sounds very odd — archaic plus something else — but I’m not sure its oddness amounts to ungrammaticality.
In the CGEL sentence quoted by Jen, the clause introduced by “neither” is the 1st of 2. If we take that sentence and add at the beginning the words Jen suggested, then that clause is now the 2nd of 3, so the “neither” is acting as another “nor”. “Nor” before a clause acts like “and” and negates the clause’s verb. “And” is OK before one of the conjuncts in a list, but not if it’s the first one. Perhaps this explains Jen’s reaction?
I admit that I have to suppress a ‘spontaneous combustion’ interpretation for the sentence— but otherwise it doesn’t seem so bad to me.
“Is either presidential candidate a smoker?”
“No, neither Mitt nor Barack smokes”
“Well, one of them has been smoking back here. Which one of them was it?”
“Sorry, neither Mitt smokes nor Barack smokes”/”Sorry, neither does Mitt smoke nor Barack smoke”/”Sorry, neither Mitt smokes nor does Barack smoke”/”Sorry, neither does Mitt smoke nor does Barack smoke”
They all seem grammatical to me here. They’re certainly clunkier but also more emphatic.
What’s the grammatical way of saying this then?
In my own native-Russian-speaker version, the first emendation option I thought of was “Neither Mitt smokes nor Barack does”, but I’m not actually sure whether that’s supposed to be grammatical in regular English either.
I agree that “Neither Mitt nor Barack smokes” is much better.
I wouldn’t say that the original was ungrammatical; just weird and clunky, which is not the same.
Seconded for the longer alternate versions; I’m legitimately not sure about the shorter original.
It’s difficult to think of worse placeholder names than Barack and Mitt
Sajid and Annunziata. /s
But yeah, the grammaticality judgements would flow much better with less exotic names, such as, I dunno, John and George (or indeed Alice and Bob). I had to specifically focus on the grammaticality-determining part of my mind to avoid getting distracted by the exotic name “Barack” (and the only slightly less exotic name “Mitt”).
I guess “either Mitt wins, or Barack wins” is perfectly fine.
It is, and now I wonder why.
They all seem grammatical to me here.
They may all be grammatical in a formal sense but I have a hard time imagining a native speaker of English producing any of the examples you give.
As a 30-something BrE speaker, I definitely agree with LH: both the given sentences from the Log post sound totally ungrammatical to my ear, both for myself and for any other dialect I’m familiar with.
Can “Neither […] nor […]” co-ordinate whole sentences at all? All the ways of doing that I can think of sound ungrammatical or at best very archaic. They can co-ordinate lots of kind of phrases (“Neither Mitt nor Barack smokes”, “Barack neither smokes nor drinks”, “Mitt smokes neither in public nor in private”, “Barack neither smokes Gauloises nor drinks vodka”) but I can’t think of any way to co-ordinate complete sentences with them, in modern English.
Jen’s example pattern comes closest — “Barack does not take drugs. Neither does he smoke, nor drink.” — but even that both seems somewhat archaic and doesn’t allow a standalone “neither … nor …”, only as part of a longer co-ordination.
I agree with our host’s judgments. Maybe for some people any sentence that’s grammatical with either…or is also grammatical with neither…nor? That’s not true for me.
Then I am not mad! I have not fallen irrevocably behind the speeding vessel of the English language!
(It’s interesting, it occurs to me, that I use “mad” = ‘crazy’ only when I have Poe in mind.)
“Mitt don’t smoke nor Barack neither”. B Partee is a Good Thing and a relatively lucid Montagovian but these things are relative for a reason
have a hard time imagining a native speaker of English producing any of the examples you give.
Now you have me doubting my command of my own native language because I didn’t just want to suggest it was grammatical but also natural. But, at risk of exposing myself further, here’s my thought. Suppose you want to make the point emphatically. You could say: “Neither MITT nor BARACK smokes”. But then you either end bathetically on an unemphasised “smokes” or you stress it which would also feel, to me, odd, as if you were contrasting smoking with something else. So instead you might say: “ Neither MITT SMOKES nor BARACK SMOKES”. That way you preserve the parallelism in the stressed parts. And the duplication then feels natural to me.
By pure chance, I was just reading a paper of Geoffrey Pullum’s, (Consigning Phenomena to Performance: A Response to Neeleman, 2013) which seems germane to this to some extent; he says at one point:
Take the section of my article in which I point out that GES [generative-enumerative syntax] grammars define only a binary grammaticality distinction. They classify everything as either perfect or nonexistent – the only alternative to being grammatically impeccable is being nonlinguistic garbage and not having grammatical properties at all. But in reality, I point out, ungrammaticality seems intuitively to be a matter of degree: some utterances are much more ungrammatical than others.
@Ian P: I can just about buy that expression, but it strikes me as a rather stiff formulation that’s unlikely to come out in casual conversation. If I wanted to be emphatic I think I would say “No, Mitt doesn’t smoke and Barack doesn’t either” or, for added vehemence, “Neither of them smoke, you [expletive deleted].”
So instead you might say: “ Neither MITT SMOKES nor BARACK SMOKES”. That way you preserve the parallelism in the stressed parts. And the duplication then feels natural to me.
Doesn’t work for me under any circumstances. Idiolects!
Under most circumstances I would find the sentence perfectly understandable, but very surprising if coming from someone whose first language is English. If English were the speaker’s second language, and if our relationship made it appropriate, I would tell them that the sentence was unidiomatic and suggest (the screamingly obvious, to my mind) alternative of “Neither Mitt nor Barack smokes.”
That being said, I can see Ian Preston’s point, that under very narrow conditions the demands of emphasis might bring a native speaker to this. So I guess I am saying I don’t find it ungrammatical, but rather unidiomatic in the vast majority of circumstances.
Alice & Bob FTW.
I’ve recently (the past few weeks) given up neither…nor. It seems archaic, and that’s not an effect I’m after most of the time, but the reason I gave it up was that I kept having to rearrange the surrounding words to get it to sound right. If the Log people can’t see that there’s something wrong with all but one of their examples, then ‘Can’t see the wood for the trees’ comes to mind.
…worse placeholder names than Barack and Mitt
Jan1May: Sajid and Annunziata. /s
Because foreign innit I’d assumed he was SaJEED JavEED but he goes by SADjid JAVid. The Rees-Moggs remind me of the Addams Family, but only to look at. In real life they’re about as cool as Fred & Wilma Flintstone.
I can use neither…nor with all kinds of phrases, but I don’t think it works to coordinate complete sentences in my idiolect. I would have to drop neither and negate the first clause in some other way, while putting the verb first in the clause after nor (and if the verb phrase is the same, it would be weird to repeat it): Mitt does not smoke, nor does Barack (smoke). But in reality I’d use and neither in place of nor.
My reaction to the LLog post was: haven’t people in Linguistics departments got something better to do? Like going out and recording some of the world’s dying languages; or protesting about climate change; or getting their Senators to impeach Trump. Or flossing.
LLOG is barely even a language blog anymore. It’s just endless cranky screeds against the Chinese Communist Party and how their writing system sucks.
To me “Neither X nor Y verbs” is idiomatic, the other two are understandable but not what a native speaker says or writes. (Without a definition of “grammatical” I don’t know if the other two are “ungrammatical,” but I would expect 90% of copyeditors to strike them out and suggest a different phrasing).
LLOG is barely even a language blog anymore. It’s just endless cranky screeds against the Chinese Communist Party and how their writing system sucks.
Unfair! There are guest posts about how Taiwanese is related to Old Norse. What’s that if not language?
Unfair! There are guest posts about how Taiwanese is related to Old Norse. What’s that if not language?
If only. Most of it is like “Xi Jinping garbles quotations from Tang poetry”.
LLOG is barely even a language blog anymore. It’s just endless cranky screeds against the Chinese Communist Party and how their writing system sucks.
And it used to be endless long-winded explanations about what exact mistake did Trump make in one of his recent speeches, and (occasionally) why it wasn’t even a mistake, and just misinterpreted.
Things were so much better in the good old days of the 2000s. I really should restart Language Log Twelve Years Later… as soon as I figure out how to update the name, anyway.
https://en.wikipedia.org/wiki/Montague_grammar
And David Foster Wallace!
“There is in my opinion no important theoretical difference between natural languages and the artificial languages of logicians; indeed, I consider it possible to comprehend the syntax and semantics of both kinds of language within a single natural and mathematically precise theory.”
Montague effectively waving two bold middle fingers at Kurt Godel and his namby-pamby “logical proofs”
“It has been claimed that Mitt smokes, and it has been claimed that Barack smokes. Neither does Mitt smoke, nor does Barack smoke; but Bernie totally does!”
That should work, shouldn’t it? Clash of registers perhaps.
But in isolation anything but “Neither Mitt nor Barack smokes” is glaringly unidiomatic at best.
That’s flat-out ungrammatical, isn’t it?
I do like the posts that showed he never said bigly, but big-league.
@nemanja:
As an exercise in armchair anthropology, one can try imagining what it must be like to be someone who encounters this sort of thing and thinks “Wow! what an illuminating insight!”, rather than … not.
There’s a reason why these people talk about “fragments” of languages (in fact, to be charitable, it shows some insight, I suppose.)
@david, indeed – the more time passes the more I realize DFW was drawn to this type of bullshit like a moth to the flame. The belief that the everything in existence can be explained by neatly organized and easily understandable theorems is pernicious – you think you’ve gained insight when instead you’re just mired in confirmation bias.
“Neither Mitt nor Barack smokes ”
For me this requires a plural verb:
“Neither Mitt nor Barack smoke.”
It’s similar to
“Neither of them smoke. ”
Neither does Mitt smoke, nor does Barack smoke; but Bernie totally does!”
That should work, shouldn’t it?
Not for me. Not even a little bit.
I don’t understand how David Marjanović’s additions are supposed to make it better. CGEL as quoted in the Log post seems right to me when it says neither “cannot normally occur initially in a coordination of main clauses.” I’m just not sure what the cases not covered by “normally” would be.
@Kieth Ivey: My instinct is that “normally” in a case like that means something like: “As far as we can tell, this rule holds essentially without exception, but it would be no particular surprise if some respected author like Cormac McCarthy* is found to violate it.”
* Example author selected because of this Language Log post.
“As several people said in the Log thread, the normal sentence is “Neither Mitt nor Barack smokes.”
Totally agree. But imagine this, if you found yourself in a situation where you uttered “Neither Mitt smokes, nor….” how would you finish the sentence?
In my case, I’d probably say “nor does Barrack”
However, “nor Barrack smokes” would also be OK, as would
“nor Barrack smokes either” and maybe even
“nor Barrack”
“nor Barrack smokes either”
Wow. No, totally unacceptable and barely even comprehensible.
To me, the issue is very straightforward. When used as a coordinating conjunction (of FANBOYS fame) nor comes after a negated verb, not a negated noun. Therefore, the natural two-clause sentence would be Mitt doesn’t smoke, nor does Barack. As a single clause sentence, Neither Mitt nor Barack smokes is also correct. No other suggestions strike me as normative English.
I think I was making a topic-and-comment sentence out of it, and that’s a lot harder in English than in German.
I’d never find myself uttering “Neither Mitt smokes”.
edit: unless there are two Mitts, and neither of them smokes!
But imagine this, if you found yourself in a situation where you uttered “Neither Mitt smokes, nor….” how would you finish the sentence?
What David Marjanović said. That is not a possible opening to a sentence in my idiolect (unless there are two Mitts).
David Marjanović said: I’d never find myself uttering “Neither Mitt smokes”.
Even to a question?
or in response to a statement like “Either M or B smokes”
I agree that the sentences aren’t the happiest – idiomatically. For example, I would probably use “Mitt’s smoking” or “Mitt’s a smoker” rather than “Mitt smokes”. But, it is still comprehensible, it is possible, and I would go so far to say it’s grammatical as well. If it is not grammatical, why is that?
If it is not grammatical, why is that?
Because you can’t join two different sentences with “neither.” In my dialect, obviously, and clearly in others’ as well.
Note that the whole thing arose out of Montague grammar, which “is based on formal logic, especially higher-order predicate logic and lambda calculus […] Montague’s thesis was that natural languages (like English) and formal languages (like programming languages) can be treated in the same way:”
Which is, in my opinion, completely loony.;
languagehat says: Because you can’t join two different sentences with “neither.”
Not entirely sure what are the two sentences. Are you saying that sentence 1 is “Mitt smokes” and sentence 2 is “B smokes”. In that case, the joining word is “nor” as opposed to “neither”. Or is it the case that the combination of “neither…nor” can’t be used to join two sentences. Am I missing something?
What happens in this case:
Is Elizabeth laughing or crying?
SENTENCE 1: Laughing.
SENTENCE 2: Crying.
NEITHER NOR STATEMENT: Neither laughing nor crying.
What happens in this case:
Is Elizabeth laughing or crying?
SENTENCE 1: Laughing.
SENTENCE 2: Crying.
NEITHER NOR STATEMENT: Neither laughing nor crying.
Those are not “sentences.” Those are participles.
And, no, you certainly cannot join actual “sentences” (which must contain a subject and a verb) by “neither…nor.”
What laowai said.
Even to a question?
What question would prompt you to begin a sentence with “Neither Mitt smokes” (unless, as mentioned above, there are two Mitts, in which case you don’t have neither…nor joining two sentences)?
I’m a bit confused by the definition of “grammatical.”
Although a non-native speaker, I share the majority view that the only natural sentences are “Neither Mitt nor Barack smokes” and “Mitt does not smoke and neither does Barack” or “Mitt does not smoke, nor does Barack.”
However, perhaps in the good company of Barbara Partee, I have trouble seeing all the alternatives proposed as equally ungrammatical. “Neither Mitt smokes nor Barack smokes” sounds flat-out wrong. Instead, “Neither does Mitt smoke, nor does Barack” sounds like a rhetorical inversion of “Mitt does not smoke, nor does Barack.” The rhetorical artifice is surely inappropriate and more than faintly ridiculous in such a mundane sentence; but does that suffice to make it ungrammatical?
Or am I failing to understand that this kind of oratorical flourish is no longer acceptable in English under any circumstances?
An example from Google Books that still sounds good to me, though it is admittedly from 1827:
“M. de RAYNAL imagines that the colour of the Cochineal is to be ascribed to the red fig on which it lives; but that author has been misinformed; for neither does the Cochineal feed upon the fruit, but only upon the leaf, which is perfectly green; nor does that species of Nopal bear red, but white figs.”
That seems to be the kind of topic-and-comment sentence I was aiming at.
It also seems to be a Radio Yerevan joke.
I suspect non-native speakers, having often been exposed to a greater percentage of out-of-date texts (and, of course, not having had a constant, years-long immersion in the current language), assimilate the grammar and other linguistic attributes of those texts and can’t feel them as “wrong” in the way native speakers do. (Compare the notorious inability of foreign learners to feel the impact of the curse words they learn.)
Definitely. Also texts that are not so much out of date as just literary or otherwise construed, e.g. scientific writing (which often isn’t by native speakers to begin with).
Neither are we trying to keep out immigrants, nor are we favouring the well-to-do
Sounds perfectly fine to me.
How about:
Either you are trying to keep out immigrants, or you are favouring the well-to-do.
Everyone says that ‘nor’ is archaic. Perhaps I’m not typical, but ‘nor’ is perfectly fine for me, although perhaps more typical of written English.
Victor Mair is certainly using Language Log as a platform for exploring and expounding his own ideas about the Chinese language and script, as well as his dislike of the CCP. For me, though, it’s worth reading the blog for all the interesting stuff he puts up about Chinese and recent Chinese usage. Mair seems to pick up a lot of information on trends in Chinese from his students and followers. If your focus is more on European languages and English (or the Middle East and Africa) then a lot of it might not be as interesting.
Turning to the main point of this post, I agree that the generativist obsession with playing around with sentences often goes too far. The idea is to figure out exactly what the rules are and where the boundaries lie, but it gradually becomes farcical. One of the dangers is that you actually start to find weird sentences more and more acceptable the more you repeat them.
Keith Ivey: What question would prompt you to begin a sentence with “Neither Mitt smokes”
Q: Why’s there an ashtray full of butts on the table in here; who smokes?
A: Neither Mitt smokes nor Hillary, so it must be Barack.
Compare the notorious inability of foreign learners to feel the impact of the curse words they learn.
REALLY upsetting someone by angrily calling them a cod(fish) in Norway. It can be as rude as calling them a cunt in England. Do the reverse and they’d think you were a bit peculiar but they wouldn’t get cross. I’d be interested in reading an explanation for this localised effect, if there is one.
Partee has a comment saying she agrees “Neither Mitt nor Barack smokes” is much better, but that’s beside the point. The point may or may not involve whatever a “root phenomenon” is.
Yeah. I would never have heard of Ultraman without his post on the straw-mud horse.
Now excuse me, I have to go squirm on the floor laughing.
@ajp
Even in answer to your question, for me it has to be: “Neither Mitt nor Hilary smokes…”???? Neither mitt smokes is only parseable as “there are two mitts and neither smokes”, no matter whether the next word is “nor” or not.
Neither Mitt smokes nor Hillary, so it must be Barack.
That sounds OK to you? It sounds impossible to me. Has to be “Neither Mitt nor Hillary smokes, so it must be Barack.”
Neither Mitt smokes nor Barack drinks.
You’d never heard of Ultraman? EVERY kid in China and Japan knows Ultraman.
It sounds impossible to me. Has to be “Neither Mitt nor Hillary smokes, so it must be Barack.”
Not as good, I agree. I’d probably say it your way. Impossible? It’s a way of starting a list of the people who don’t smoke. But Bathrobe’s right that the more you stare at them the harder it becomes to judge.
Neither mitt smokes is only parseable as “there are two mitts
Yeah, I see that, but Paddy, maybe you need both parts to parse any neither…nor setup?
I have now achieved the ESL learner’s satori on the subject of neither / nor: Don’t use it, some native will always say you are wrong.
Well, the only time I was in China was in 2006, and I’ve never been to Japan. The real story here is why Ultraman isn’t better known in the rest of the world.
Correction: I think I’ve encountered the name once since first reading that post. But that’s all.
Thy godfadirs wyff thow shalt not take; Neither a woman hir godmodir housband; Neither thy gossep to haue to thy make.
This is middle English but the modern equivalent, e.g.,
you shall not marry your godfather’s widow; nor [shall] a woman [marry] her godmother’s husband….
works better for me than “neither Mitt smokes nor …”. It is because the bits in square brackets are supplied automatically for me without thinking.
My godfather’s still around, his wife is eighty and her mother over 100, but you never know the future and it’s good to be prepared. On the square brackets I have stared & stared and can’t decide what I think. I feel I’d get the gist of ‘Neither Mitt smokes nor Hillary’ in a spoken conversation so to that extent it “works for me”. Whether it’s grammatically perfect is something else, but I didn’t think that was the original question.
AJP: Neither Mitt smokes nor Hillary, so it must be Barack.
Hat: That sounds OK to you? It sounds impossible to me. Has to be “Neither Mitt nor Hillary smokes, so it must be Barack.”
AJP: Not as good, I agree. I’d probably say it your way. Impossible? It’s a way of starting a list of the people who don’t smoke.
Speech is a bundle of loose ends and false starts. Much is possible, even perfectly natural, when said. I think the above is natural when you start listing non-smokers, realize that this might go on for a while, and feel that you can’t withhold the verb to the end of the list without losing the listener. “Neither Mitt … smokes … nor Hillary … nor John, or George, or the other John … I think Bill smoked, but Hillary emptied the ashtray … so I guess it must be Barack.”
At some point there you’ll switch from ‘nor’ to ‘or’, which I suppose joins the elements syntactically by equating them to eachother before comparing them as a group to whatever went before the last ‘nor’. Maybe that makes it easier to follow.
(If you needed another non-native opinion, that is.)
Yes, I agree with Trond. It’s more a conversational than written usage so there might well be something wrong with its syntax.
It doesn’t strike me as conversational at all, but this may be a US/UK thing, or maybe just an idiolectal difference. I’m increasingly used to being an outlier.
Don’t use it, some native will always say you are wrong.
Only if you’re using it to join two sentences. Two noun phrases, two verb phrases, two prepositional phrases, two predicate adjectives are all fine.
Yes, I can see it being idiolectal. My speech is probably more muddled than yours.
What I find fascinating about examples like these is how the category of grammatical English utterances is constructed by our brains. It is a well-known observation that the conditions of grammaticality are unnecessarily strong, in that there are utterances that are clearly comprehensible but also ungrammatical. This indicates that there are more restrictions on grammatically than are really needed for communicative purposes. What is more, native speakers (at least) can learn, with pretty good person-to-person consistency, which constructions are ungrammatical, even when they are parallel to perfectly grammatical constructions. As Kieth Ivey notes, neither can be used to coordinate almost anything except complete sentences, and either can be used to coordinate complete sentences. Yet a native speaker has apparently been exposed to enough of a corpus with neither not being used that way to know it is unacceptable.
As I was thinking about this yesterday, I learned something new about what can be learned just from exposure to a vast corpus of English. My RPG gaming group finished a campaign yesterday, and we were having a meta discussion of what to do next. The gamemaster mentioned that there were now computerized resources for generating game data, and he pointed us to the Talk to Transformer Web site, where you enter a prompt text, and it continues it for you, based on what it knows of existing texts. We all tried it for a while, and we found that it typically produces texts that are topical, grammatical, and comprehensible, yet which do not sound like something a human would actually compose. They end up being repetitive and sometimes self contradictory, putting them in a sort of uncanny valley of writing.
To some extent, our results may have been atypical, since we were mostly entering famous quotes, which the program was directly familiar with. I tried, “Do you want to play questions?” from Rosencrantz and Guildenstern Are Dead, and it gave back a response formatted as play dialogue, showing that it, to some extent, “recognized” the source. A friend entered the first two lines of “The Raven,” and the site came back with the next two lines, followed by a passage about a Cat rapping to be let in. However, unquestionably the most disconcerting example was when I tried, “We are the priests of power.” from Nineteen Eighty-Four. It just followed up with twenty-five more repetitions of, “We are the priests of power.”
I tried the Talk to Transformer with Trump says, “I’m mad about Noel Coward and got:
@Brett:
Surely the moral of this particular example is that native-speaker judgments of acceptability aren’t always consistent? My own impressions of these sentences are not at all the same as those of some of the other L1 speakers above – and (more or less by definition) we’re all correct, despite disagreeing with each other.
As a corollary, grammaticality is not necessarily a binary acceptable/unacceptable thing, but can be a cline (which is a problem for at least some of the more brain-dead versions of Chomskyism.)
Geoffrey Pullum comes up with a nice example where there is basically no right answer to the question of grammaticality, in formal-register sentences like
You should marry who(m)ever pleases you most.
He points out that native speakers generally cannot say which form is correct.
An ordinary-register example where there seems to be no acceptable form (again cribbed from Pullum) is
Either my brother or I am likely to be there.
Either my brother or I are likely to be there.
Either my brother or I is likely to be there.
(This is from his paper The Truth about English Grammar: Rarely Pure and Never Simple, in which I myself get a laudatory, though alas, anonymous, mention.)
http://www.lel.ed.ac.uk/~gpullum/LTTCpaper.pdf
Talk to Transformer
Ha! My son spent the better part of the summer reading (and laughing loudly to) pretend sub-reddits generated by our virtually intelligent overlords. Is this the same software with news rather than Reddit as input?
I tried “Kusaal is spoken in the far northeast of Ghana”, confidently expecting to fry “Talk to Transformer”‘s brain, but got
Kusaal is spoken in the far northeast of Ghana. It’s a regional language spoken by about 8% of the country’s population, and it is related to Swahili and Yoruba, among others. The city is home to two universities, with a population of over 250,000.
which is not bad at all, really, especially if you substitute “district” for “city”, which is the only real giveaway that the algorithm has no actual understanding of the text at all. If it had stopped after one sentence I would have been seriously impressed.
It didn’t do quite so well with “Hausa is very widely used as a lingua franca in West Africa”, producing
Hausa is very widely used as a lingua franca in West Africa. For example, the French and British are both fluent speakers of it, as are many Africans.
Hausa is also spoken by a number of other African nations, such as Eritrea, Nigeria, Sudan, Chad, the Democratic Republic of the Congo and the Republic of Guinea.
As a language spoken by large parts of Africa, Hausa has also become the most commonly spoken language in Africa.
Contents show]
History
In the year 1335, the Arab invasion of northern Mali caused much fear and chaos. The Muslim invaders were so determined to take over the country that they set up a system of Shariah courts that were based around the law of the sword.
The Muslims used the Muslim calendar as the basis for their calendar, meaning that the Muslim calendar differed from that of the local people in a number of different ways. It was this difference that caused many local people to have a hard time accepting the
Not much better than Wikipedia, really. Oh, well.
Trond Is this the same software with news rather than Reddit as input?
They imply the input is from more than one source since it’s not “trained” on any one, though it’s obviously seizing chunks of something. Their own choice of language makes me wonder how they got into this, he said haughtily:
I used TTT to compose this homily. I started with the first clause of John 1:1, and seeded each successive part by using the last sentence plus any additional sentence fragment of the previous part. When I got tired of it, I added the heading “Conclusion” to my text, and then seeded TTT with the remaining clauses of the same verse in order to bring it back on topic. After deleting the last sentence fragment, my uncle became Robert.
I sent it to Daphne Preston-Kendall, who’s studying Indo-European linguistics and (less formally) theology; she said TTT was actually a pretty good theologian, and the homily was definitely better than the one she heard last Sunday.
Yes, I remember it being OpenAI. I don’t know what selection of texts TTT is trained on, but it gave me this:
… which at least Google doesn’t recognize as something straight out of a book.
I used TTT to compose this homily.
I like the reference “Catechism of the Catholic Church No. 1064, cited by St. Augustine.”
Which reminds me of Bob Dylan, which makes me wonder what would happen if you seeded it with Dylan lyrics.
Perhaps the AI was written in Dylan.
Hey mister tambourine man, play a song for me. What do you say?”
“I say, ‘Hoo, that’s a fine thing you’re doing.'”
“I say, ‘Hoo, that’s a fine thing you’re doing!'”
“Hoo-eee, you’re very kind. What about a song?”
“I say, ‘Well, maybe that’ll be the last thing you play. Why don’t you do it next.'”
“What do you want me to play next?”
“Well, I don’t know, I just want you to say, ‘Hoo, that’s a fine thing you’re doing.'”
“Hoo, that’s a fine thing you’re doing.”
“I say, ‘Hoo, that’s a fine thing you’re doing!'”
“I say, ‘Hoo, that’s a fine thing you’re doing!'”
TTT continuing the long tradition of Dylan covers being significant improvements on the original.
Don’t be mean to Robert. He’s John Cowan’ s uncle.
…we’re all correct, despite disagreeing with each other.
Linguistics is a Buddhist parable.
When my wife and her then-boyfriend were living in TB’s mother’s basement in the late 60s, his mother was curious about the kind of music they played down there, so she asked him to bring up a record that she might enjoy. They brought one of their favorite Dylan singles (I don’t know which one). She listened to it carefully and then said, “Well, it’s a very fine song. But don’t you think it would be better if someone else had sung it?”
When Gale told me this story, I told her I agreed with TB’s mother. Fortunately she had known me long enough by then not to kick me out of the house over it.
My wife agrees with you and TB’s mother.
“Neither MITT SMOKES nor BARACK SMOKES” strikes me as a register clash. That kind of vocal stress for emphasis and “neither…nor” just don’t go together in the same register, for me.
I do get the impression that with a lot of these examples, whether we find them acceptable depends on the context we imagine (which includes register, since register relates to context).
My wife agrees with you and TB’s mother.
Well, mine doesn’t. The greatest, most expressive voice in the history of the world and you see what the poor man has to put up with. Still.
I input “Is there a doctor in the house?” and got:
There is.
It is time for a heart check.
There is a doctor in the house.
We are going to be late.
What is the time?
We are going to be late.
What time is it?
We are going to be late.
How long is it before the concert starts?
We will be late.
How long is it before the concert starts?
The concert starts.
When the concert starts, we are going to be late.
We are going to be late.
What time is it?
We are going to be late.
How long is it before the concert starts?
The concert starts.
When the concert starts, we are going to be late.
We are going to be late.
We are going to be late.
We are
Sounds like a mashup of Beckett and TS Eliot.
It’s acceptable to leave out the first “smokes” in the original sentence, “Neither Mitt nor Barack smokes” or “Mitt does not smoke, nor does Barack”. Linking a repeated verb with “nor” is clumsy and grates on my sensitive ears. OED on “nor”:
II. In senses of neither.
5. And —— not; neither. In later use normally with inversion of subject and verb.
Categories »
a. Following an affirmative clause, or in continuing narration. Now chiefly poetic.
I did the obvious thing and seeded TTT with “neither Mitt nor Barack smokes”, and the results look quite coherent— but driven mostly by the associations those particular names have in the training data (which seem to include an association between smoking and cannabis):
Changing the names to generic Alex and Bertie (I’ll refrain from my usual grumbling about the pointlessly gendered Alice and Bob) stops the pot association from coming up:
I’ll refrain from my usual grumbling about the pointlessly gendered Alice and Bob
Not pointless at all: it allows the natural he and she to be used in Alice-and-Bob stories rather than repeating their names constantly. Quine’s example “Il retira le manuscrit de la serviette et le/la jeta dans le mer” illustrates how (if you’re lucky) you can disambiguate whether it was the manuscript or the briefcase that was thrown into the sea with a pronoun.
Proper languages have at least eight genders. What’s the use of anaphoric pronouns when they’re all the same?
Tried famous first lines and got:
“It was the best of times, it was the worst of times and in the end, it was about as good as it gets.”
That’s deep
If only Dickens had thought it through like that, he could have brought the novel to a satisfying conclusion on Page One.
(Come to think of it, it would do very nicely as an entry in Private Eye’s First Drafts spot.)
Apocryphal Gospel of Transformer:
Now when Jesus was born in Bethlehem of Judaea in the days of Herod the king, behold, there came wise men from the east to Jerusalem. When they saw the child with Mary his mother, and the power of the Lord overshadowing him, they made known to the disciples what was done. And the disciples were filled with awe and were thinking within themselves, “Who is this?” And immediately all were amazed, saying, “Truly this was the Son of God.” And the disciples were in awe and saying, “Can this be the Son of David?” And while they were still thinking these things, behold, the Herodian soldiers came and surrounded the child with his mother, and guards with them. Then the Magi—two brothers—who were men of the country—and they came and stood before the child. And they came and said to him, “How is it that we who have seen you can see you?” But…
Proper languages have at least eight genders. What’s the use of anaphoric pronouns when they’re all the same?
”[I]n the Yimas creation legend, there is an important prop which is introduced and then ignored for a reasonably long stretch of discourse, about 3 pages; it is then reintroduced, but only through the use of its corresponding pronominal prefix. This is because, being Class VII dual, a relatively rare combination, there is no chance of confusion with any other referent.”
How is it that we who have seen you can see you?
Evidently one of those Gnostic gospels. (Though perhaps a bit too coherent?)
thrown into the sea with a pronoun
On reflection that sounds like one of those Roman punishments, like being tied in a leather bag with a dog, a snake, a monkey, and a chicken and then thrown into the water, as was done to parricides. Alternatively, they were thrown to the wild beasts in the arena.
Proper languages have at least eight genders.
Proper languages have at least eight genders. And so does the language in which it is written, and the language that it is spoken.
In addition, the languages in which a language is spoken also have grammatical gender. A noun in French can be masculine, feminine, neuter or neutral. It is often assumed that these are only grammatical rules, not real differences in the genders of the nouns in the language. It turns out that these differences are real.
These gender differences are also reflected in the names of words. In a paper by D.E. Piersma, it is shown that there are four main gender differences. These include:
• a) nouns with gender, e.g. man, woman, and child, are always masculine. The gender of pronouns (e.g. “he” and “she”) is also masculine.
• b) adjectives are either masculine or feminine.
• c) verb conj […]
Alas, we cannot actually read Piersma’s paper (or any of his other work), any more than we can read Winterlake’s book on the Earl of Essex.
The first line of A Tale if Two Cities was one of the first things I tried also, but the output it gave me was less interesting.
Alas, we cannot actually read Piersma’s paper (or any of his other work)
I trust you flipped an eight-sided die to choose a pronoun.
I think a die would have to be rolled, so you must mean “flipped an eight-sided coin”.
I gave up on polyhedral dice a long time ago in favor of ordinary cubical dice plus a pair of icosahedral dice for generating numbers from 0 to 99. I keep them in a long narrow jar….
When I read this, I was enlightened.
@John Cowan:
I think it’s worth considering why we need Alice-and-Bob stories in linguistic examples; it’s not as if gendered pronouns are the only way to disambiguate between discourse referents.
I’m not saying there aren’t any contexts when gendered referents might be appropriate (discussing anaphora might be one of those), but the problems their common use gives rise to have been explicitly pointed out at least as far back as Macaulay, M., & Brice, C. (1994). Gentlemen prefer blondes: A study of gender bias in example sentences. In M. Bucholtz, A. Liang, L. Sutton, & C. Hines (Eds.) Cultural performances:
Proceedings of the third Berkeley women and language conference (pp. 449–461). Berkeley,
CA: University of Berkeley.
XKCD on Alice and Bob.
The third edition of [Advanced] Dungeons & Dragons* tried something like the “Alice and Bob” approach for discussing character stats, classes, and abilities. For each character (sub)-class, they created an “iconic” character to be used whenever an example was needed. So, rather than having to deal with questions of whether an abstract character should be male of female, all discussions of, say, fighter abilities, would involve the specific individual Regdar.
* I only know this second hand. I have never personally played anything later than early second edition AD&D, or even perused any of the Third through Fifth edition books, so I have no idea how long this idea lasted, or how it was received among players/fans.
Since lambda calculus has been mentioned in this thread, I’ll put this here for those who might be interested: To Dissect a Mockingbird: A Graphical Notation for the Lambda Calculus with Animated Reduction, by David C Keenan. I confess I can make neither heads nor tails out of it.
Alice-and-Bob stories in linguistic examples
Those are for cryptography: linguists talk about John and Mary mostly. I admit I prefer Alice and Bob, partly because of their other alphabetical relatives. Charlie and Dan are generic, like Alice and Bob, but when they are not required by the plot, Charlie may be the (password) Cracker. Similarly, Eve is the Eavesdropper, Mallory the Malicious Attacker, and Olivia the Oracle. Some come in groups: Peggy is the Prover, Victor and his sister Vanna are the Verifiers, and Trent the Trusted Arbitrator. Sybil is not alphabetical: she is an attacker who uses a large number of false identities. I am sure there are more.
In an e-mail, I just instructed one of my graduate students to take a look at one of my previous papers, because it explains the two possible symmetry structures of localized divergenceless fields.* However, I wanted to check to see how precisely my previous student and I had expressed things, and rereading the manuscript, I found that part of the description used a rather odd locution (which I now remember grappling with when I wrote it eight years ago).
This triple coordination seems fine to me, albeit perhaps unnecessarily stilted. However, putting a “Neither” at the beginning would make it hopelessly ungrammatical. It also seems like the variation of “nor will” and “nor does” (instead of “nor does” twice) makes it sound a lot less awkward.
* I apologize for the obvious typo in the arXiv version. Equation (12) should have the B field, not E.
** Imagine there are carets on the Greek letters too.
Must one really say divergenceless rather than divergence-free, non[-]divergent, or something else less barbarous than the nearly vowelless /daɪʼvɚgṇsḷs/?
I would have gotten rid of the second nor with its associated second inversion by saying “Likewise, taking the derivative does not” etc.
@Hat lambda calculus … To Dissect a Mockingbird: …. I confess I can make neither heads nor tails out of it.
(Just noticed that comment from last year.)
I’d call myself rather well up on lambda calculus/use it all the time in actual programming. Happy to answer your questions/but it’s an extraordinarily simple to understand yet powerful formalism.
When I was learning it, lots of tutorials/StackOverflow answers pointed to Mockingbird [**].
I never could make heads nor tails either. It seems to wrap the elegant simplicity in rococo mystification.
[**] To be a little more precise, Mockingbird is model of Combinatory Logic [***] — which is isomorphic to lambda calculus, but you won’t grasp that it is nor how, until you understand both — in which case looking at Smullyan or Keenan first isn’t likely to help.
[***] Due to Moses Schönfinkel, independently discovered by Curry — which I mention chiefly because of the entertainment value we could have got from talking about ‘Schönfinkeling’ rather than ‘Currying’.
Thanks… I think! But I do like Schönfinkeling.
This notation was developed as part of an attempt to place the lambda calculus on an even deeper foundation. Both the attempt (so far unsuccessful) and the notation were inspired by George Spencer-Brown’s (1969) controversial book Laws of Form.
To my knowledge, that makes two people whose boats were floated by GSB: Keenan and Luhmann. Laws of Form is not so much controversial as unintelligible. You can interpret and cite it without fear of controversy. It stands mute and inglorious in a bookshelf just behind me.
Whereas I read Luhmann all the time, in transports of illumination. I’ve never met anyone else in meatspace who has read Luhmann. There seems to be a connection here.
@John Cowan: Divergence free would be fine, but I don’t like the sound of it as much. However, I don’t think you can use anything that involves “divergent” or “diverging.” The divergence is indeed a measure of how much a vector field is diverging out from each point (sources having positive divergence, sinks negative), but I think it is too specific a mathematical measure to be adequate described by anything other than the proper name divergence; other locutions that might be identical in meaning etymologically do not seem to be satisfactory. Moreover, while there are other terms that are routinely used to described a field that satisfies ∇ · B = 0 (incompressible, solenoidal, or transverse), they are based on physical analogies that I did not feel were appropriate in the abstract context I was discussing.
I’m currently trying to chew my way through The Homotopy Type Theory Book (the HoTT book to insiders) and I shall gain a small sliver of satisfaction by mentally correcting to schönfinkeling throughout.
I spent a large part of a summer about 45 years ago trying to understand Combinatory Logic. Intuitive it’s not.
EDIT: Hmm, I refreshed the page after the WordPress malfunction, not more than 15 minutes ago I think, and I swear that Hat’s comment from three hours ago was not there yet. And neither were Stu’s or Brett’s. This is a response to AntC’s latest.
Yeah, sorry about the malfunction. Songdog fixed it with his usual acumen and celerity.
The z-derivative of a divergenceless field remains divergenceless
That rang a bell: “the polhode rolls without slipping on the herpolhode lying in the invariable plane”, from Goldstein’s Classical Mechanics.
@Y: Indeed, see here for my previous comments (and a link to a seminar announcement that quotes Goldstein directly).
I had a vague memory that it had come up before, but in the interest of laziness gambled that it had not.
@ Lars Matthiesen :
Workshop on Homotopy Type Theory/ Univalent Foundations,
Vienna, Austria, April 22-23, 2023
https://hott-uf.github.io/2023/
Gonna be wet & wild, baby! All sorts of ancestral spirits will be about.
Unicode… pretends to have you covered: θ̂, φ̂.
I didn’t know -less gets reduced that far, but I guess it makes sense, especially for people with the rabbit-abbot merger.
But /g/?
Laws of Form is not so much controversial as unintelligible.
wp has tried. I got a vague feeling it’s trying to do something like the Sheffer Stroke — which appeared about half way through.
Having laboured (years ago) to convince myself the traditional Boolean operators can be reduced to Sheffer, I promptly wiped that bit of memory for re-use.
I see mention of lattices. I wonder if it can help me prove something (so far elusive) about a particular sub-class (non-Boolean) of lattices that has a very practical application. Or perhaps I’ll go mad trying.
But /g/?
Surely a typo.
Ah, on re-reading I see one source of confusion in Keenan (paraphrasing Smullyan)
So these birds are calling _names_/descriptions. Whereas mockingbirds mimic _songs_. We at least need a bird that can pronounce/mimic words. Ideal would be the Mynah:
religiosa would fit our dreadful puns nicely: Alonzo Church, inventor of lambda calculus.
Unfortunately there’s no verb ‘to mynah’, so we’ll have to work with parroting parrots. Conveniently we then don’t have to introduce so early in the piece anything spelled with ‘m’, and especially not the ghastly contortion oMega — upside down ω looks like ‘m’ my foot!
Our first parrot after the Idiot-bird can be NZ’s second-most famous parrot, the Kea (rather than Kestrel) then KIwi (admittedly not a parrot, but rather than KIte — also not a parrot, neither is mocking-bird). And there’s plenty NZ endemic birds — many with onomatopoeic names in Te Reo. Plus
… sorry, no: Keenan’s diagrams just aren’t working for me. Altogether too silly. Is this how reading too much GSB screws with your mind?
/g/ was indeed a typo, or rather a thinko, and I am indeed an abbot-rabbiter.
@jack morava, thanks. I won’t be able to attend, but the proceedings of the previous seven workshops are online and replete with rabbit holes, (Also my mathematical maturity is a bit rusty, and I need a bit more background material than the speakers are likely to supply before I can see that stuff is obvious).