Hacker News new | comments | show | ask | jobs | submit login
"I never said she stole my money" has 7 different meanings depending on the stressed word
154 points by sama 2736 days ago | hide | past | web | 62 comments | favorite
Just thought that was interesting.

Yup, another favorite of mine from linguistics is:

    "We gave the monkeys the bananas because they were [adjective]"
What is "they" referring to, and the adjective describing? Well, it depends - "ripe" would describe bananas, "hungry" would describe monkeys. The syntax alone is ambiguous!

"They gave the monkeys the bananas because they were [adjective]"

now it can describe a third noun...

"They gave the monkeys the bananas in bunches because they were [adjective]"

I had an English teacher in high school who was highly attentive to this sort of construct. He would return many a paper with "VPR" marks sprinkled throughout for 'vague pronoun reference' and points deducted accordingly. Certainly a common error.

While I hated these teachers when they were doing this, I've grown to appreciate their advice. I wish more did stuff like this.

Do you mean teachers, or people in general?

Do you mean you-singular, or are you addressing everybody in the thread who used a pronoun that could have referred to teachers. I think he meant that they fed the monkeys bananas because the teachers were hungry.

I don't remember having teachers who focused on that, and I really wish one of my coworkers had.

After a few minutes of talking to him it's very hard to figure out what he's saying because he'll have dropped nouns all together in favor of pronouns. This can get very confusing when talking about technical stuffs.

Well, you could of course change the context. If we glom "We didn't give the rhinoceroces anything because they weren't ripe for a feeding, but ..." onto the front of your sentence then "ripe" starts to apply to the monkeys.

This one has given me problems so many times. Never got an answer.(I'm not a native speaker)

Jerry Weinberg uses this technique in "Exploring Requirements: Quality Before Design" as a technique for exploring vague requirements. When you get a short verbal requirement thrown at you, repeat it back with different words stressed, and ask whether that's the right meaning or what information might be missing. It's a great technique for flushing out extra information and avoiding misunderstandings.

While we're on the topic of wordsmithing, "verbal" means using words. You probably mean "oral".

Sounds difficult to pull off without sounding like a jerk.

The phrase "Time flies like an arrow" is another example. "Time flies" could be a type of insect with an affinity for a particular arrow (or each for its own arrow), or you might be being instructed to measure a fly's speed in the manner that one measures that of an arrow (or to time only those flies that resemble an arrow), or it could have the conventional metaphorical meaning.

cf. "Fruit flies like a banana."

Garden path sentences along with other linguistic puzzles are awesome (and this little puzzle is the reason why Noam Chomsky and other linguists can all be cultural critics):


Discussions of garden path sentences deserve a reference to Emo Philips: http://en.wikiquote.org/wiki/Emo_Philips

The Structure and Interpretation of Computer Programs touches upon this. Somewhere in chapter 4, they implement a nondeterministic interpreter with CPS and then use it to (among other things) find all possible parses of sentences such as these.

This makes me think it might be easier to get humans to speak a language that computers can understand rather than trying to get computers to understand our current languages :)

It might be. I'm reminded of a story I read about Palm designing their early PDFs. At that point, they had only experienced marginal success interpreting written text. There were a lot of ambiguities, and the actual style varied a lot from person to person, so it tended to fail outside of carefully controlled tests.

Eventually, someone hit on the idea of forcing the users into a custom set of symbols that was close enough to normal to be usable, but was different enough that it forced everyone into a common style that was easier to parse.

Of course, the tricky part is to come up with a method that is expressive enough to be useful without ending up with one of those "natural" programming languages.

Nice discussion, but doesn't explore nearly enough. As Don Gause and I showed in http://www.geraldmweinberg.com/Site/Exploring_Requirements.h..., you can stress more than one word in a sentence. In fact, you can stress two, three, or any number up to the number of words in the sentence.

This sounds really weird, but if you say these sentences out loud, you can make a different meaning from each variation. Try it:

<i>I</i> <i>didn't</i> say she stole my money. That is, I was one of the few who didn't.

I didn't <i>say</i> she <i>stole</i> my money. I wrote a letter about how she found my money.

See if you can get all the way up to all 7 words stressed. There are lots of cases (which you can computer using the binomial theorem).

You get different implied meanings by stressing different words in the famous (Buffalo)+ sentences, too.

THE buffalo FROM Buffalo WHO ARE buffaloed BY buffalo FROM Buffalo ALSO buffalo THE buffalo FROM Buffalo.

The word "buffalo" has officially lost its meaning after reading that article.

One of my favorite examples of ambiguity when it comes to punctuation is the following

"When hunting lions, hide in the bushes." vs. "When hunting, lions hide in the bushes."

More than seven, since you can stress more than one word.

when doing nlp is there a standardized system or practice for encoding this data (which word(s) where emphasized)?

what is the logical framework for this? at first I thought you could only have one emphasized word per statement, but though its a bit more nuanced but you can also have two at a time (ie "my" and "money")

Yeah, this stuff is done with what basically amounts to an annotated typed logic: http://en.wikipedia.org/wiki/Focus_(linguistics)


Maybe I'm missing something. Does this not apply to just about every sentence in every language?

This is applicable to most sentences. For example: "Make Something People Want"

Creative creatives create creative creations, with language!

Unlike the "buffalo" one, this sentence is at least parseable.

Additionally, in Czech almost any permutation of words in a sentence is allowed. You can then get even more variances of meanings.

This subject is featured in a Seinfeld episode: "These PRETZELS are making me thirsty!" "These pretzels are making ME thirsty!"

Just had to spell this out to a colleague, so thought I may as well post it. It should be "didn't say", rather than "never said", but the gist is the same ;)

I didn't say she stole my money - someone else said it

I didn't say she stole my money - I didn't say it

I didn't say she stole my money - I only implied it

I didn't say she stole my money - I said someone did, not necessarily her

I didn't say she stole my money - I considered it borrowed, even though she didn't ask

I didn't say she stole my money - only that she stole money

I didn't say she stole my money - she stole stuff which cost me money to replace

According to training I got as a volunteer in Washington, this drives native Vietnamese speakers crazy. I think it may confuse older Koreans as well. My grandma didn't get it at all.

I think this is a part of what makes English one of the best languages to tell a lie in.

Pauses can similarly change the meaning.


This applies to many other languages. I can translate all this to Russian and this phenomenon will hold: "Я не говорил что она украла у меня деньги" -- the same 7 meanings depending on stress.

In Arabic

انا لم اقل بأنها سرقت مالي

Ana lam aqul bi'anaha saraqat maali

If you stress any particular word, you will have to insert or remove parts of the speech in order to conjugate and decline words as necessary:

(1) Ana lam aqul bi'anaha saraqat maali

(2)Laysa ana man qal bi'anaha saraqat maali .. or (2.b) Lestu ana man qal bi'anaha saraqat maali

(3) Ana ma qultu bi'anaha saraqat maali

(4) Ma qultu bi'anahaa saraqat maali

(5) Lam aqul bi'anahaa hiya man saraqa maali (or "man saraqat maali" when implying an all female list of suspects.)

(6) Ana lam aqul inahaa saraqat maali

(7) Ana lam aqul inahaa saraqat maaliya (anaa)

Sometimes Arabic is way too precise and makes puns a little harder.

I think this translation is more accurate and makes for a better proof: Я не говорил что она украла мoи деньги; otherwise, a great point.

YARGH. Okay so this is one of the primary areas of research that one of my professors from university studies ( http://www.ling.ohio-state.edu/~croberts/ ).

None of the explanations that people have written here are entirely correct or reflect what's actually going on.

The phenomenon that's being discussed here is something called Contrastive Stress. It is a part of an interesting area of research on Linguistic Focus (http://en.wikipedia.org/wiki/Focus_(linguistics)) that sits at the juncture of Semantics (meaning in abstract), Pragmatics (meaning in context), and Autosemgental Phonology (mental representation of sounds and their production of non-phoneme related stuff).

So, contrastive stress ties directly into the notion that sentences are stated in response to either explicit or implicit questions. In fact the location of the emphatic stress is directly related to what question the speaker is trying to answer.

So, you can actually do this with nearly any sentence, simply by placing the stress on a different word.

Badger's account is mostly correct, but it's tied in a little closer with grammar than his examples actually intimate. I'd say that there's a much wider range of possible candidates for say:

Did you say she stole your money? I didn't say she stole my money, i know/saw/heard/thought/wrote/hinted/testified/dreamed it!

"It" in this context is the entire grammatical structure "she stole my money". The reason why this is important is because of the notion of what can be stressed and what the stress is actually applied/scoped to.


You can do this with other sentences as well:

George W. Bush is the 43rd President of the United States.

"Is jeb bush the 43rd president of the united states?"

No, George Bush is the 43rd President of the United States.

"Is George Washington the 43rd President of the United States?"

No, George Bush is the 43rd President of the United States.

"George Bush isn't the 43rd President of the United States, right?"

No, George W. Bush is the 43rd President of the United States.

"Was there more than one 43rd president of the united states?"

No, George W. Bush is the 43rd President of the United States.

"Was George W. Bush the 44th President of the United States?"

No, George W. Bush is the 43rd President of the United States.

"Was George W. Bush the 43rd Vice President of the United States?"

No, George W. Bush is the 43rd President of the United States.

(okay, so "of" is a function word that we can't contrast against anything else)

"Was George W. Bush the 43rd President of the United Arab Emerates?"

No George W. Bush is the 43rd President of the United States.


So this really is a general phenomenon that you not only see every day, but really use on a constant basis. Every sentence has grammatical stress, it's how we know what other people are focusing on when they speak.

I can see how your explanation is more complete, but is there anything wrong in badger7's post?

No, i'm not saying that Badger's post is incorrect (it correctly defines the phenomenon). Some of the other posts on this thread are all over the place, and don't describe the phenomenon or what underlies it.

So sorry if i gave the impression that Badger7 is wrong, he's not, he just doesn't explain what's going on, or what the full scope of the phenomenon is :)

The similarity between this response and the sample ones made me laugh.

"Is there anything wrong in badger7's post?" "No, I'm not saying Badger's post is incorrect"

The double-negative makes alternative emphases difficult, but I'm tempted.

I took it as it was meant. My colleague doubted the basic assertion, so I was providing an example of each - it wasn't intended to be comprehensive :)

(okay, so "of" is a function word that we can't contrast against anything else)

"Was George W. Bush the 43rd President from the United States?"

Yeah i guess that's one way to do that. There's not much of a semantic distinction there, so that comes off as more of a sentence that's focusing on stressing a syntactic difference not a semantic one. But point taken :)

My memory of these terms is weak, but isn't this metalinguistic contrast? You're not using the word 'of' when you stress it here, you're mentioning it.

um, I reckon you can put constrastive stress on function words, as in in the street against on the street and President of the United States against President (of UAE) in the United States.

Re: #7 - She could have stolen your heart :'(

that sounds painful

it didn't seem painful from what I saw in Indiana Jones

Of course neither did getting blasted in a nuclear explosion as long as you have a fridge to hide in.

The stress adds information. #2 is the basic meaning (it is true for all of the others as well), a stress on any other word adds information about what you did way/imply.

Incidentally, some still bear alternative interpretations: e.g., #3 could be interpreted as "I only thought it".

#2 has a more specific meaning: it implies that someone else has claimed that you did say she stole your money.

So what does it mean if you don't put emphasis on any of the words? Or more than one? Also, I can't stop reading the comments without putting emphasis on random words now. I can almost make any sentence say what I want it to, or at least distort the intended meaning.

That's known as "neutral stress" and can just be an assertion of fact. The topic/question under discussion may be undefined, for example.

Sentences can also have multiple stresses, and you may have multiple topics under discussion. Sentences still have a primary stress, but can have secondary stresses as well.

Who sang what to whom?

Pat sang "I feel pretty" to Jackie!

Implicit here are three stresses, which may indicate the following: Pat sang, can you believe it? I didn't know pat could sing! Pat sang "i feel pretty", who would have know that Pat likes show tunes? Pat sang "i feel pretty" to Jackie. Jackie was the wrong person to sing to, because Jackie hates the music from West Side Story.

Actually, I would consider the meaning of all the sentences to be the same. The implications are all different, but that's really not surprising or profound in itself. I mean, you could also say that sentence 10 different times referring to 10 different "she's," but that's not at all surprising, is it?

Nope, analogy fail. The meaning is absolutely different. If someone says to you "She took how much of your money?", and you reply "I didn't say she stole my money" that means something completely different than "I didn't say she stole my money". Who the pronoun refers to is completely external to the parsing of the sentence.

No, they meanings are really the same. You're telling the person that you did not do something. That something, no matter which word you stress, is the acting of saying she stole your money.

This is one reason why NLP is hard.

That's so sad.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact