Bad Habits

I saw a magazine cover that had a teaser for an article by Dr. Oz. The list of things I could learn in his “Healthy-Life Handbook” included “Facts You Must Know,” “Tests You Need Most,” and this:

Bad Habits to Break

Let’s see, what would be a bad habit to break? Getting some exercise every day — it would be bad to break that habit, if I had it. Eating lots of fruits and vegetables — breaking that habit would be a bad thing, too. What else? Does not smoking count as a habit? It would be bad to break that habit by quitting not smoking.

When I looked inside the magazine at the article itself, the actual habits in the list consisted of eating fat-free stuff that has added sugar; taking pills to stop pain instead of finding its root cause; sitting too much; and overrelying on technology.

In other words, Oz wasn’t talking about habits it would be bad to break; he was talking about bad habits that you should break. In fact, the wording on the list title inside the magazine made this clear: It called them “Bad Habits You Should Drop.” Probably a lack of space on the magazine cover led to the ambiguous wording I saw there. The reading that the editors intended for bad habits to break corresponds to this parse:

The adjective bad modifies the noun habits, and that whole chunk is modified by to break. (The label Inf/NP means an infinitive phrase has a noun phrase gap. To break bad habits would be an Inf, but without that direct object, it’s Inf/NP.) I’ll call this the intersective reading, because the meaning is the intersection of two sets: habits that are bad, and habits that you should break. This interpretation implies that there are bad habits that you shouldn’t break, which might have been one factor that pushed me in the direction of the other reading.

That other reading corresponds to a different parse:

Here, the adjective bad and the infinitive phrase to break work together to form a meaning something like “X such that breaking X is bad”. This sort of discontinuous adjective phrase wraps around the noun it modifies, habits, and we end up with “habits such that breaking them is bad”. In a 1983 paper, Michael Jones gives the name “property fusion” to this kind of adjective-infinitive meaning.

So I read “bad habits to break” with the “property fusion” meaning, instead of the intersectional meaning, and the result was a completely different set of habits. But if the phrase had instead been good habits to develop, the ambiguity would have been hardly noticeable. The fusional meaning of “habits that it would be good to develop” and the intersective meaning of “good habits that are also habits you can/should develop” are for all practical purposes the same. Or are they? What would be some good habits to develop that are actually impossible to develop?

Correlatively Comparatively Speaking, Part II

On Fritinancy, Nancy Friedman commented on a poster for a walk for breast cancer. Here’s the poster, lifted from Nancy’s blog post:

Nancy’s reaction:

As I see it, the line needs a second relative pronoun to be properly parallel in structure: “The more of us who walk, the more of us who survive.”

She wanted another who in there so that the two parts of this comparative correlative would be maximally parallel, but in fact, there are some speakers who wouldn’t even put a who in the first part. As for me, I’m not even sure what I would do. (Maybe I should do a search on this blog and see if I’ve generated any data that would say.)

The uncertainty comes from the fact that comparative correlatives like the one Nancy found are a little different from others. In many comparative correlative clauses, the comparative part — the X-er — corresponds to a gap in the remainder of the clause. This gap might be a direct object gap, as in (1) below; an indirect object gap, as in (2); or a prepositional object gap, as in (3). It can even be a predicative adjective gap, as in (4), adverbial gap, as in (5).

1. DO gap: the more [I learn __]
2. IO gap: the more people [you give __ a break]
   (if you allow extraction from ditransitive VPs)
3. PrepObj gap: the more people [we talk to ___]
4. PredAdj gap: the happier [we'll be ___]
5. Adv gap: the more [we get together ___]

Interestingly, all these kinds of comparative clauses can also have a relative pronoun before the gappy part of the clause, as if it were an actual relative clause. Even the gaps for predicative adjectives and adverbs can take a relativizer, as long as it’s that. Instead of making up examples this time, here are some from Google:

6. DO gap: The more people who [you can get ___ to dine with us that day]
7. IO gap: the more people that [you give __ a break]
   (OK, I did make this one up)
8. PrepObj gap: the more people that [you can connect with ___]
9. PredAdj gap: “The Smarter That [I Think I Am ___], the Dumber [I Get ___]“
10. Adv gap: The faster that [the boat goes ___]

Example (9) is interesting in that it’s like Nancy’s example: a relativizer in the first clause (the smarter that I think I am), but not in the second (the dumber I get). But let’s leave aside relative pronouns for the moment and talk about the main difference between Nancy’s example and other comparative correlatives. It’s easier to see if we put some brackets in them and gap labels, the way we did with the others:

11. Subj gap: the more of us who [___ walk]
12. Subj gap: the more of us [___ survive]

In these examples, the comparative phrase the more of us corresponds to a subject gap in the remainder of the clause. In (11), this linkage is handled by the relative pronoun who. In (12), it isn’t. If you think of comparative clauses as relative-clause structures, then probably you don’t like (12), because in English, you typically can’t delete relative pronouns that connect to a subject gap. (The exceptions are in sentences such as There was a farmer had a dog.) But if you never thought of comparative clauses as a kind of relative clause — in other words, if you just thought of them as the, plus a phrase containing a comparative adjective/adverb/determiner, plus a clause missing that same kind of phrase — then there should be no problem with (12).

If you’re one of the speakers who are OK with (12), and in general don’t think of comparative correlatives as a species of relative clause structure, I suspect that you still might be comfortable uttering comparative clauses like the more of us who walk. The reason involves a third kind of comparative correlative that I haven’t been talking about. However, that’s a whole ‘nother kettle of worms, which will have to come in a separate post. See you then!

Man Accused of Assaulting Officer, K9 Indicted

My wife showed me a nice crash blossom yesterday. It went like this:

Mentally filling in the usual missing pieces of headline syntax, I arrived at this interpretation:

A Delaware County man is accused of assaulting an officer and a K9 has been indicted.

My wife and I agreed that that really sucked for the dog, who was taking the heat for that Delaware County man’s misdeed.

Of course, here is the intended reading, with slightly different missing pieces restored:

A Delaware County man who has been accused of assaulting an officer and a K9 has been indicted.

This crash blossom was made possible by two of the usual culprits in headline ambiguity. First, there’s the copula omission (i.e. omission of forms of be). In ordinary language, man accused of assaulting an officer wouldn’t be ambiguous; it would just be a noun, man, modified by an adjectival passive phrase, accused of assaulting an officer. Or, if you want, a “reduced relative clause”: who has been accused…. But because headlines leave out the copula, we’ve been trained to turn phrases like this into full clauses by inserting is or has been. Of course, that’s exactly what does have to happen for indicted: We have to supply our own is or has been, but being led to do the same thing for man accused paves the way for the second part of the misunderstanding.

The second part is based on the replacement of and with a comma. If the headline had read Delaware County man accused of assaulting officer and K9 indicted, the easiest parse would have been the correct one, in which and joins just officer and K9 as the objects of assaulting. Instead, the comma made the easiest parse the one in which we have two entire clauses, with K9 as the subject of the second one.

Dummy Prepositions

I dropped by a weekly discussion group at OSU yesterday, to hear Carl Pollard talk about the version of categorial grammar he’s currently developing. When it came to prepositions, he made a distinction between prepositional phrases that actually referred to a location (as in I saw a mysterious figure on the roof); and those that might as well just be plain noun phrases for all the meaning the preposition contributes. The example Carl gave was depend, which takes an on-PP as a complement, as in depend on me. He proposed not even calling on me in this example a prepositional phrase; instead, its syntactic category (its “tecto” in Carl’s jargon) would be simply be an “On Phrase”.

It can be tricky identifying these “dummy” prepositions. It’s easy enough to discard clear cases of meaningful prepositions, in verb phrases like walk to school, but it gets harder as the prepositionals become metaphorical, in phrases such as stare at him. Furthermore, you have to avoid “intransitive prepositions” (sometimes called particles), in phrases like tie up the prisoner. You might mistake up for a dummy preposition because it certainly doesn’t seem to contribute any spatial meaning. The trouble is that it also doesn’t take an object. Although it might look like the prisoner is the object, of up, if you replace the prisoner with a pronoun, you quickly realize that up isn’t taking it as an object. If it were, a phrase like *tie up him would be grammatical, just like stare at him is. Instead, the phrasal verb has to “wrap” around its pronoun direct object: tie him up. So to get a true dummy preposition, you want a preposition that contributes no spatial meaning, and also takes an object. The on after depend meets these requirements.

To further demonstrate that this kind of meaningless PP was a different thing than an ordinary PP, Carl ran it through a classic ambiguity test (which I’ve described here), having a single on-PP function in both ways at once:

It’s on Mt. Everest that I live and depend.

I laughed, and Carl said, “I knew you’d get that!” And to fellow syntactician Bob Levine, who was turning around in his seat to look at me: “Neal can coordinate anything!”

“That sentence wasn’t grammatical for you, was it?” Bob asked.

“No,” I answered. “That’s why I’m laughing!”

Gotta love that linguist humor. Where else would It’s on Mt. Everest that I live and depend work as a comedy one-liner? If you’ve got some others, let’s hear them!

Illegal Immigrants

With the recently re-ignited debate over the term illegal immigrant, I have heard all the arguments against using the term, including:

1. It is politically divisive or inflammatory.
2. It presumes guilt before due process has been done.
3. It is inaccurate in characterizing people who entered legally but overstayed their visa, or did not come here of their own accord.
4. It is nonsensical, because illegal refers to acts, not to people.

I will grant (1), and add that the same applies to the euphemistic undocumented immigrant (and the dysphemistic illegal alien). I will also grant (2), but add that this is fixable with the well-accepted use of alleged in cases where there is doubt. I will also grant (3). But as for (4), this argument is just plain silly, and grasping at straws.

I will grant that when illegal modifies a noun, that noun usually refers to an action. I will further grant that when it does modify a noun that refers to a thing, it usually means that the thing is illegal to possess, as in illegal drugs and illegal weapons. Using those collocations as analogies, we would expect illegal immigrant to mean an immigrant that it is illegal for someone to possess–in other words, a victim of human trafficking. That, of course, is not the meaning that it has.

In fact, that is a good argument (in addition to arguments about dehumanization) for abandoning the term illegal alien. However, that still doesn’t mean that illegal immigrant is nonsense. When the noun is the agentive form of a verb, and the adjective is the morphological analog of a manner adverb, there is a common, productive rule of semantic composition that gets you to the accepted meaning. Let me illustrate with an example unburdened by controversy. If I were to say, “Sandy is a deep thinker,” it would be willfully obtuse to say, “Hey, wait a minute! People can’t be deep!” If I were to tell you, “Lee is a beautiful dancer,” I could be telling the truth even if Lee’s face, when covered by a paper bag, could still make clocks lose two minutes per hour. In short,

dances beautifully : beautiful dancer :: thinks deeply : deep thinker :: immigrates illegally : illegal immigrant

Object to the term illegal immigrant on ethical, political, or legal grounds if you want to. But don’t resort to claiming the term embodies sloppy semantics, when it’s the most natural way to refer to someone who immigrated illegally. That just makes it look like you’ll accept any old argument that favors your side, and weakens the more valid ones.

Update, Oct. 16, 2012: Changed list item #2 from “were born here” (which I’ve known since elementary school automatically confers citizenship) to what I meant to say: “entered legally but overstayed their visa”.

Found in the Wild

On the August 30 episode of Kevin Allison’s Risk! podcast, I heard two examples of syntactic phenomena that I’ve written about before, that supposedly don’t occur much in actual written or spoken English. Not to say that they never happen, but they’re rare enough to have caught my attention.

As Allison says during each episode, Risk! is the podcast where “people tell true stories they never thought they’d dare to share.” It’s labeled “Explicit” on iTunes, and I should note that they don’t mean explicit in the way that a good mathematical proof or instruction manual should be. They mean sexually explicit, and some of the episodes truly are. Allison himself did a story spanning three episodes called “Kevin Goes to Kink Camp,” which I didn’t care to listen to past the middle of the Part 2. But if you want to hear a depressing yet hilarious story featuring not only sex, but also excrement and vomit, there was this other episode that’s got to be from sometime in August, but I can’t seem to find it again. Other stories are completely family-friendly, like the one from a couple of years ago involving a standardized test and a squirrel. Sometimes I’ll take a risk (as Allison likes to recommend) and listen to the latest episode in the car while Doug and Adam are with me, and hope it’ll be one of the clean episodes like that one. Sometimes it is.

Anyway, like many podcasts, Risk! has sponsors, which Allison promotes wholeheartedly. I liked when he talked up one sponsor, an online sexual accessories store, in a gravelly, old-tar sailor’s voice, telling us, “Yer gonna buy yer lube an’ yer condoms anyway, so ye might as well get ‘em from….” In recent episodes, the sponsor hasn’t been nearly as interesting: an online purveyor of postage. But Allison gushes over it gamely, and on the August 30 episode, he said

There’s a lot more mailing that should have been being done before that is being done now….

A nice example of a past perfect progressive passive, a kind of verb cluster that I’ve also written about in passing in this post, and as the main topic in this Visual Thesaurus column.

Shortly after that utterance, Allison gave his usual spiel on how to take advantage of a special offer on that website, and make sure that his show got credit for referring you:

So go to [sponsor] before you do anything else, click on that little radio microphone at the top of the home page, type in R-I-S-K, and get going.

So we have four main clauses, coordinated:

1. Go to [sponsor]
2. click on that little radio icon,
3. type in R-I-S-K
4. get going

Then there’s one subordinate clause: before you do anything else. The way Allison says it, there’s no pause between the first main clause and this subordinate clause, and there is a pause between anything else and click. So it sounds like the before clause modifies Go to [sponsor]. That could work, if he truly means for me to visit this website. On the other hand, the utterance makes more sense if the before clause is modifying click on that little radio icon. If you click the radio icon before you do anything else on that website, Risk! gets the credit.

In short, we have a phrase that could look backwards to modify something, or forwards to modify something. It’s a forward/backward attachment ambiguity, better known (to some at least) as a squinting modifier! Here’s what the Merriam-Webster Dictionary of English Usage has to say about squinting modifiers:

[T]he squinting modifier is more of a theoretical possibility — with, it must be admitted, a catchy title — than a real problem.

Maybe so, but there it is, in the wild!

Un-, Non- and Not

When we watched the Olympics opening ceremonies earlier this month, Adam got his first look at Rowan Atkinson’s Mr. Bean character. He liked it enough that we watched the movie Bean with him and Doug a few days later. I have never seen Adam laugh so hard as when Mr. Bean’s attempted repair of Whistler’s painting Arrangement in Grey and Black was revealed. Some time later Adam referred to Rowan Atkinson by his character’s name, and Doug got on his case about it. Adam defended his choice by remarking:

Rowan Atkinson is an uninteresting name. Mr. Bean is.

Using my ordinary rules for expanding out VP ellipsis, I look for the thing earlier in the sentence that can appear as the complement of is, and end up with this:

Rowan Atkinson is an uninteresting name. Mr. Bean is [an uninteresting name].

Huh? That’s no good. If Adam had intended to say both names were uninteresting, I would have expected a too at the end, and maybe even an and connecting the two clauses. For me, one way of expressing Adam’s thought would be:

Rowan Atkinson is an uninteresting name. Mr. Bean isn’t [an uninteresting name].

But it’s hard to process those two negations, isn’t [uninteresting], one of which is unspoken. More likely, I would have said:

Rowan Atkinson is not an interesting name. Mr. Bean is [an interesting name].

So in Adam’s sentence, the negative prefix un-, which should be stuck tight to the adjective interesting, has wiggled loose to roam its clause at large, as if it were the free-standing negator not. How strange. Of course, we all know that Rowan Atkinson is the more interesting name, regardless of Adam’s statement.

Not long after that, I heard the same thing happen with another negative prefix. Someone on the news on the radio was talking about super PACs and the unlimited amounts of money that are flooding this year’s political campaigns, and he mentioned

groups that purport to be nonpartisan but often are

Again, going by the usual rules, we get nonsense:

groups that purport to be nonpartisan but often are [nonpartisan]

For that to make sense in isolation, the but would have to be an and, and of course even that would make no sense as a description of the current political funding scene. What the speaker meant was,

groups that purport to be nonpartisan but often aren’t [nonpartisan]

Once again, it would have been hard to process those two negatives, one of which isn’t even spoken. He could have also phrased his thought like this:

groups that claim they are not partisan but often are [partisan]

That sentence took a little more tweaking in order to turn the non into a not, and I could well imagine that the speaker began with a common-enough phrase, purport to be, and then just couldn’t find a smooth way to finish the thought in the time he had. So now I wonder: Were Adam’s and the pundit’s sentences mistakes, things they would not have said if they had had a few more seconds to plan their utterances, or are negative prefixes truly free to behave this way in their grammars? How about in yours?

Waste or Recycle, Please

Elliot Anderson sent me this picture from a trip to Disneyland or Disney World (he didn’t say which), and told me, “I didn’t know whether I should waste or recycle.”

As I wrote in response:

I think either of these messages would be OK on its own, but together, they generate the kind of puzzlement you experienced. When I’m in the post office and I see the trash can labeled “Waste,” I get a micro-chuckle about imagining that they’re encouraging me to waste stuff, before coming back to the message of “(Put your) waste ([noun] here).” I would do the same thing at Disney World for a trash can by itself. But right next to a recycle bin with a similar message, you get the interference. Unlike waste, recycle cannot (yet) be used as a noun, so you’re forced to interpret the message as an explicit command. And then by association, you want to do the same with the garbage can, essentially being forced to give it the stupid reading.

I Forgot to Go to the Store and Get Any

A few days ago, as I was pulling into the garage, I suddenly said to myself,

Well, crap! I forgot to go to the store and get any club soda.

How annoying. We had run out of club soda two days before, so my wife couldn’t make any more of her favorite drink: club soda with cranberry juice (the sweetened kind, with lime flavor already in it). I couldn’t make any more for myself, either, and when that happens and there’s no iced tea made, it gives me an unfortunate excuse to continue my love-hate relationship with Coke.

Of course, that’s not what compels me to write about my utterance here on the blog. As with my last post, I was interested in a negative polarity item (NPI), in this case, the word any. You can’t say things like,

*I got any club soda.
*I want to get any club soda.
*I went to the store and got any club soda.

There has to be a negation or question or something similar involved; for example,

I didn’t get any club soda.
Do you want any club soda?

The verb forget counts as something similar, with its implicitly negative meaning of “not remember,” so you can certainly say,

I forgot to get any club soda.

So because my sentence had forgot as its main verb, there should be nothing surprising about having the NPI any somewhere in the complement to forgot, right? But in that case, why doesn’t this next sentence work?

*I forgot to scoop out the litterboxes and get any club soda.

At least, I don’t think it works. Do you? And the reason it doesn’t is the same reason that you can’t say something like

*Club soda is what I forgot to scoop out the litterboxes and get [ ].

For you to make a relative clause out of club soda, which plays a part in only one of the coordinated verb phrases, those verb phrases have to have some sensible relation to each other. Go to the store and get club soda go together as two steps in a single undertaking. On the other hand, scoop out the litterboxes and get club soda don’t have any relation to each other. Unless…

• …you keep bottles of club soda buried in your litterboxes.
• …scooping out the litterboxes is something that always happens right before you get club soda.
• …scooping out the litterboxes sets a Rube Goldberg apparatus in motion that results in the delivery of club soda.

In those situations, that sentence would work, and so would I forgot to scoop out the litterboxes and get any club soda, I think. This is interesting. I hadn’t read or thought about NPI licensing as something that could be relevant to these coordinations that require a special relationship between the coordinated items.

P.S. I see that when I view the preview for this post using Chrome, the words continue, housecleaning, and filled are hyperlinked to spammy sites. I’ve been using Firefox up until now, and I see that these tacky ads don’t show up in that browser. Good on you, Firefox, and Chrome, I’m very disappointed.

Negative Polarity and Dr. No

When I was a kid, I gradually became aware of a book and movie character named James Bond. There were no VCRs or movie rental places then, so my first opportunity to see a Bond movie was when Moonraker came out. I was eager to go, and thought the movie was kind of cool when I saw it, even though I couldn’t follow the plot very well. I attributed that to being young. A couple of years later, I saw For Your Eyes Only and still couldn’t follow the plot. After a couple more Bond movies in the theaters, and a few on TV, I concluded that the problem wasn’t me; it was the movies. More recent entries have been boring as well as hard to follow. But now, Doug and Adam are old enough to want to be interested in James Bond. They haven’t seen enough of them yet to realize that most of them are boring and overlong, with plots that don’t make sense. Unfortunately, their mother hasn’t realized this yet, either, even after Goldeneye and the new Casino Royale, so I’ve had to sit through a few of them on family movie night. The earlier ones haven’t been too bad; I liked Goldfinger. The problem is that if the boys like these movies, they’re gaining an inaccurate impression of the true nature of Bond movies, and as a result will probably want to see Skyfall when it comes out.

Anyway, the latest Bond movie we saw was one of the ones with Sean Connery: Dr. No. In this one, Bond meets a woman on a secluded beach, collecting shells. Bond fans will know that her name is Honey, and she’s played by Ursula Andress. She’s apprehensive as Bond approaches, and this dialogue ensues:

Bond: I promise I won’t steal your shells.
Honey: I promise you you won’t, either.

Did you hear what went wrong there? When it’s used as an adverb, either is a negative polarity item (NPI), occurring only in sentences involving a negation, questions, or contexts focusing on a limitation. If you want to say either in an ordinary affirmative sentence, you have to use too instead, as in these examples:

I don’t like the Bond movies with Pierce Brosnan. I don’t like the ones with Daniel Craig, either.
My wife likes Sean Connery. She likes Roger Moore, {too, *either}.

Sometimes you can get away with using a too in a negative context, where you’d expect either. John Lennon pulled it off in “Imagine”, when he sang

Nothing to kill or die for / And no religion, too.

I also remember an old Justin Wilson bit, in which a recurring line is

I don’t know. And that ain’t all, I don’t give a damn, too!

But either needs two things in order to be grammatical, at least in my idiolect: a negation, and a proposition that is semantically similar to one that has already been uttered (or otherwise understood by the speakers). When both of these occur in one clause, everything’s fine, as in I don’t like the ones with Daniel Craig, either. We have the negation don’t, and the whole proposition is semantically similar to the one that came before it: I don’t like the Bond movies with Pierce Brosnan.

What about in the sentence from Dr. No? The negation is in the embedded clause: you won’t [steal my seashells]. But there is no semantically similar proposition in that embedded clause. Honey didn’t say something like I promise you you won’t seduce me, or steal my seashells, either. Where we do get the similar propositions is at the upper clause level: Bond promises X, and Honey promises X, too.

It reminds me of sentences I’ve actually uttered myself, involving tag questions. But when I said these, I noticed, and identified them as things I didn’t mean to say. The Dr. No sentence was presumably part of a written script. What do you think? Is I promise you you won’t, either an error, or something that’s a part of other people’s grammar but not mine?