Wednesday, July 30, 2008

odd stress placement

When I was on hold recently I was constantly told to "Please remain on the line. One of our customer care representatives will be with you momentarily." Almost immediately I noticed that the sentence was stressed very strangely (and this was an actual recorded voice, not a computer generated message). I would put a pause between those two sentences, making them two separate utterances for prosodic purposes. For the first I would put a primary accent on "please" and a secondary accent on "line". For the second I would put a primary accent on "care" and a secondary accent on "with". However, this was not at all the case in this recording.

Instead, the speaker did not have a pause or any kind of intonational reset where the orthographic period is. She seemed to parse it into "Please remain on the line one of our customer. Care representatives will be with you momentarily." (At least, that's how I would represent orthographically the prosodic pattern she used.) In the first "sentence", primary accent was on "please" (no surprise there), but the secondary accent was on "our". In the second "sentence", the primary accent was on the third syllable of "representatives", while the secondary accent was on "with". Adding further to the oddity was the fact that the first intonational phrase fit perfectly into 3/4 time, complete with minor accents on the first beat, with "please" taking two beats: "Please -- re-/ main on the / line one of /our customer", and then of course it started to break down. But I found it exceedingly odd that, like myself, the speaker parsed the utterance into two intonational phrases, but that her phrases did not correlate in any way with meaning or clausal structure. There has to be some sort of flagrant alignment violation here, and I don't like it one bit. Luckily I'll never have to deal with it again, since I was on hold to cancel my account.

Monday, July 28, 2008

another Jay Leno headline

Another humorous "headline" from the Tonight Show recently was the misspelling "hors devours" instead of "hors d'oeuvres". This is what is referred to as Cupertino (another linguistic term generated by Language Log), or a computer generated incorrection. Many people use spell checkers on their work, and some have automated spell checkers to replace misspelled words. Problems arise when the words are not actually misspelled, but rather unknown to the spell checking dictionary. If this were a human based error, we'd expect something closer to the original. Certainly no one, no matter how confused by spelling, would think "hors devours" is the correct spelling, especially since the got "hors" right despite its two silent graphemes. And let's be honest, who can remember how to spell "hors d'oeuvre"? The only reason I can manage to do so is because I'm aware of the French grapheme "oe", which keeps me from mixing up the order of the vowels, and because the metathesis (switching of segment order) of v and r in French loans is not uncommon (cf. Brett Favre).

(I'll refrain from discussing the addition of English plural morphology on a word that is already plural in the original language.)

Friday, July 25, 2008

Low humor

A lot of humorous signage and announcements arise from our preference for low attachment (attaching a NP to the lowest possible node, or in a more linear sense, having it complement the most recent verb, preposition, etc.) I heard a nice one a little while ago on Jay Leno's Headlines segment on the Tonight Show. It was a wedding announcement that mentioned the couple's traditional Hawai'ian wedding, complete with "the blowing of the conch shell and Hawai'ian minister". The intended reading is [[the blowing [of [the conch shell]]] and [Hawai'ian minister]], with two separate NP's joined by the conjunction "and". The humorous reading results from the fact that we don't want to attach "Hawai'ian minister" high up on the tree where it is sister to "the blowing of the conch shell". We want to attach it way down low to "blowing", giving us something like [[the blowing [of [the conch shell] and [Hawai'ian minister]]]].

Wednesday, July 23, 2008

new ECM verbs

Prescriptivists seem to be troubled by the advent of new and innovative ECM verbs. First, a little background. ECM stands for Extraneous Case Marking, and applies to verbs like "believe" or "expect" which take a CP complement (CP = Complementizer Phrase; in standard prescriptivist grammar essentially any complete clause) but assign accusative case to the subject of that clause. To quote my syntax teacher's example, "Max expects Maria to word letters carefully." In this example we have what is essentially a complete clause as the complement of "expects" (with the exception of the infinitive verb "to word", but that's outside the scope of this post). "Maria" is clearly the subject. Yet if we replace "Maria" with a pronoun, we're going to choose "her", not "she". This means that the NP in subject position of the complement clause is being assigned accusative case. How the heck is this possible? "Expect" is an ECM verb! It can assign case across a CP boundary (something verbs generally aren't supposed to be able to do).

Nowadays, though, it seems that lowly prepositions are taking CP complements (or has this always been the case?). Many of us have heard or uttered something like "I was surprised by them winning the race". Prescriptively, of course, this is "wrong". It should be "I was surprised by their winning the race", where "their winning the race" is a NP versus the CP of "them winning the race". Clearly there's something going on here, though, because plenty of people say things with this structure. My wager would probably be on the analysis of "be surprised by s.t." as a single verb, and then giving that verb ECM marking. Try as I might I can't think of very many good examples of this construction, even though I hear it all the time, so I may post a follow up later.

Monday, July 21, 2008

The logicians have lost.

As some people interested in English history know, there was a movement a few hundred years ago to make English a logical language, in the sense of obeying the rules of predicate logic. One example of this is the attempted eradication of double negation. In symbolic logic, ~p means "it is not the case that p is true" regardless of what p is. It could be a whole series of statements, and the one negation negates them all. This was not true in English until the logicians made it so. Many people still use double negation, but now it's a marked variant, a non-standard dialect. Another example is the use of the nominative case for verbs of being, i.e., "It is I". Logically, the logicians said, the copula there ("is") represents "=", and thus the word preceding and following it should have the same case. It seems, however, that ultimately the logicians have lost.

The token that really brought this home to me is the Godspeed You! Black Emperor song "motherfucker = redeemer". I was puzzling over what the song title could mean, and realized that one reading that was certainly not possible was that every motherfucker is a redeemer and that ever redeemer is a motherfucker, which in logic and math is what the symbol "=" is used for. 2+4=6 is a truth, no matter how you look at it. However, using the equal sign in English generally means that the left hand item is equivalent to the right hand item, but not vice versa. The copula works the same way. If I say "Computational linguists are jerks", I don't mean that every jerk is a computational linguist, but I probably mean that every computational linguist is a jerk (I didn't say it was an accurate statement; it's just an example). Sorry, logic. Natural languages don't really like you.

Friday, July 18, 2008

The scope of "next"

When I say "next Tuesday" I mean the Tuesday of next week. So if today's Monday, I don't mean tomorrow. Similarly, if I say "next Saturday", I mean the Saturday of next week. On Monday, I don't mean the day five days from now, I mean the day twelve days from now. I've found that this is not true of everyone, and that there's quite a split in how people perceive this usage. It seems the two main interpretations are "the next X that occurs" and "the X of next week". So to some people, saying "next Saturday" means "the next Saturday that occurs" which may often be the Saturday of this week. On the other hand, to people like me, "next Saturday" always means the Saturday of next week; using the word "next" cannot refer to any day this week. Needless to say, this causes problems.

The OED gives us, under the entry for "next":

Applied (without preceding the) to days of the week, with either the current day or (in later use; orig. Sc.) the current week as the implicit point of reference.
Thus (for example) next Friday may mean ‘the soonest Friday after today’ or ‘the Friday of the coming week’. The latter may be indicated contextually, e.g. by contrast with this, but it is not always clear which meaning is intended.


So the key question here is what kind of scope "next" has (here I don't really mean semantic scope so much as temporal scope). For some people, the frame of reference is the day, for some the week. It seems the key distinction is that last sentence from the OED quote: people who distinguish "this Friday", "next Friday" are going to use the week as the frame of reference, whereas for someone who doesn't use "this X" for days of the week isn't going to have any kind of week association with the word "next"; it will mean what it means in ordinary speech, i.e., the next X that occurs, without any intervening time.

Thursday, July 17, 2008

When good coarticulations go bad

In English we have two /l/ sounds, commonly called the "light" or "clear" l and the "dark" l. They exist in complementary distribution, with the light l in onsets and the dark l in codas (and syllable nuclei, though in these cases the l is underlyingly in coda position). The clear l is a regular lateral approximant, with the tip of the tongue resting on the teeth or alveolar ridge, depending on pronunciation, and the dark l identical in apical (tip) placement, but with co-occurring velar constriction by the lamina (the blade of the tongue).

However, Tom Brokaw for some reason just doesn't like clear l's. Not only does he pronounce all his l's (including those in onset position) dark, he doesn't even articulate the apical feature of the sound, instead using only the back of his tongue for the velar articulation, resulting in what can sound at times like a French "r" or Arabic "gh".

On a side note, my understanding of the two different l's in English was an extremely important step in my pronunciation of Spanish, which only has clear l's. Try it yourself: say "lamp" and "awl". The former is a clear l, the only l in most languages. The second is a dark l.

Wednesday, July 16, 2008

British h-dropping

I don't know a whole lot about British pronunciation, but my impression was that h-dropping (i.e., failing to pronounce the h at the beginning of a word) is considered "low class", and is not present in RP (Received Pronunciation, "the Queen's English"). It is also my understanding that RP is the standard dialect used in broadcasting, public speaking, etc., or at least it was until recently (I seem to remember John Wells blogging about Estuary English overtaking RP in public settings in the past 10-20 years). However, I noticed in a news clip from the 70's or 80's that the newscaster failed to pronounce his h's. There was one particular example that caught my ear. The clip was about Gary Glitter, a pop star from a few decades ago, who was arrested. The newscaster said, "Gary Glitter 'as been arrested." In rapid speech I probably wouldn't even have noticed the h-dropping (after all, even in American English we would probably drop the h in that situation), except that in the newscaster's non-rhotic dialect the "r" at the end of "Glitter" jumped out at me. Since he pronounced the "r", I had to surmise it was in onset position, which means there couldn't have been an h.

So I guess my question is for any British English speakers, or anyone else who knows: what's the deal with h-dropping? Is it common among broadcast speech?

Tuesday, July 15, 2008

More on alignment constraints

It seems that "jurisdiction" can be syllabified in two ways: juris.diction or juri.sdiction (leaving aside the other two syllable boundaries). The first is what I thought to be my own pronunciation (as it turns out that's only the way I perceive my own pronunciation because I perceive the morpheme boundary between "juris" and "diction"), while the second is probably the common way people pronounce the word. The NoCoda constraint strikes again! Alignment constraints want us to align morpheme boundaries with syllable boundaries, and since "jurisdiction" comes from a combination of Latin juris (the genitive of jus, 'law') and dictio (from dicere, 'say, speak'), theoretically the syllable boundary should be between the "s" and the "d". However, the NoCoda constraint is ranked high enough in English that we would rather sacrifice alignment than have a coda in the preceding syllable. We also have a lower ranking Ident-IO(vc) (input and output segments should have the same value for [voice]) constraint, since we would rather say jur.i.stic.tion than attempt the unwieldy jur.i.sdic.tion.

Monday, July 14, 2008

Cherri-gac

I've been keeping up with a show called "The Next Food Network Star", in which a number of cooks compete for their own show on the food network. In one of the episodes the contestants had to create and market their own pre-packaged food product. One of the contestants chose to make a chocolate sauce containing cherries and cognac, and marketed it as "Cherri-gac", pronounced ˈtʃeɹiˌjæk. What I thought was interesting about the spelling is the perception of the "g" and /j/, even though it comes before the "n". The original French pronunciation would be koɲɒk, with a probable American phonemicization of kɒnjæk, unless the speaker really has a palatal nasal in their idiolect.

Thus the "gn" sequence in "cognac" is interpreted as a phonetic [nj] sequence, and apparently it didn't bother this contestant that the [j] sound comes after the "n" while the orthographic "g" comes before the "n". I thought this was rather strange because (if I can try to remember back before I started being interested in orthography and pronunciation) I think my original interpretation of "gn" sequences in French and Italian was that the "g" was silent, and the palatalization of the "n" was just a quirk of those words in those languages. Clearly this is not the only way people view that digraph. Since he associated the "g" with [j], it made sense to him to spell his product as he did.

Friday, July 11, 2008

Another reason for universal linguistics education

To the phonetician, nothing is more amusing that seeing a band that has decided to bring umlauts into its name. It makes sense. Linguistics is such an understudied field that the very name of the field is (at least on one occasion in my own experience) confused with a type of pasta. To the layperson, umlauts, accents, and other diacritics are merely decorations. Sure, they know that in some arcane science they have some meaning, but nothing that they need to pay attention to. I think we should all go around pronouncing these ridiculous band names as they should be pronounced according to the orthography. It might turn some heads to pronounce Mötley Crüe as møtli kɹy (or perhaps as møtɬi crye), or Mötörhead as møɾøɹhɛd or møtørhead.

Thursday, July 10, 2008

Sszark the Burning

As anyone who has played the video game Diablo II can attest, the best part is the names of the unique characters. These include the venerable Frozenstein, Puke Pus the Sharp, and one that hadn't caught my eye until recently: Sszark the Burning. The real question here is, how do the game designers want us to pronounce this? Is that a long "s" at the beginning, followed by a [z]? Maybe the "ss" is a syllable nucleus, as can be the case in Blackfoot: ss.zark. Or is it an "s" followed by the Hungarian "sz" for /s/, giving a long s cluster at the beginning: ssaɹk. It's a mystery to me, but either way it's a great name.

Wednesday, July 9, 2008

Why we need linguistics education for all

In last month's issue of Cigar Aficionado, while reading an article on cachaca (Brazil's national liquor of choice, distilled from sugar cane but by a different method than rum), this gem of pronunciation advice caught my eye, concerning the three different c's in the word:

The first is hard, the second is the soft blend of "ch" as in "chagrin," and the third is a combination of "s" and "z" like the "c" in "facade": ka-SHAH-sa.

The first description is well-known to any speaker of English; we often talk about "hard" c's (/k/) and "soft" c's (/s/). The second description is a little confusing ("blend" of what? "c" and "h"?), but fairly transparent with the example word. The third description, however, really threw me for a loop. A combination of "s" and "z"? They seem to have picked the one feature of phonetics that truly is on or off, without any gradations (yes, there are several types of voicing, but it all comes down to either the vocal folds are vibrating or they aren't). It's clear enough what sounds they mean : /s/, as evidenced by the sample word "facade." But what the heck were they trying to describe by saying that this normal "s" sound is somewhere between /s/ and /z/? Maybe unbeknownst to me everyone else pronounces "facade" with a breathy voiced "z".

Tuesday, July 8, 2008

I hate lazy spammers

If you're going to phish for important information, at least get your syntax right. I was told recently in an email, "Verify your account now to avoid it closed." The intent of the message is obvious, and skitters just on the edge of being grammatical (maybe "avoid it being closed"?), but clearly this is an inexpert English user rather than a typing impaired spammer. Come on guys, have pride in what you're doing, even if that's trying to trick people in giving up their credit card information.

Any guesses on the writer's native language?

Monday, July 7, 2008

literally, technically

Semantic bleaching is something that happens all the time. In fact, I just used it: I don't mean that at every infinitesimal moment in time semantic bleaching is occurring; I merely mean that it is not uncommon to see its effects. Essentially semantic bleaching is the lessening of the force of a word. For instance, even "extremely" these days doesn't mean much. People have an interesting way of dealing with this. One of the strategies I hear most often is to use "literally," as in, "I could literally eat a horse I'm so hungry." Clearly this person does not mean that they desire to sit down and consume an entire horse. What they mean is that we use hyperbole so much in everyday speech that even saying one could eat a horse does not express the extreme hunger that person is experiencing. Hence the use of literally. My professor Tony Mattina proposed that people use "literally" essentially as the opposite of what it actually means: to mean "metaphorically". However, I think the difference is more subtle. Not all the usages I hear are strictly metaphorical. Rather, I think people are using the word "literally" not to mean literally, but simply as an intensifier.

Another (and I believe related) difference in word usage is with the word "technically". On a linguistics forum I frequent a new member posed a question about "technically". He recently voted for candidate A because he did not want candidate B to win. Thus in his mind he was technically voting against candidate B, rather than for candidate A. However, his friend argued with him, saying that since technically it is not possible to cast a vote against a candidate, he technically voted for candidate A. My take on this is that the friend is the correct one, at least in the strictest usage of the word. To me, "technically" refers to procedure, i.e., what is objectively happening at any given moment. Thus if John shoots Bill, technically all he's doing is pulling the trigger of a gun. Of course, if you believe in the slippery slope argument, where do we draw the line? Perhaps I should really say John's brain is firing electrical impulses that cause his index finger to contract.

Thursday, July 3, 2008

4th of July weekend

No new posts until Monday because of the 4th of July weekend. I will be busy smoking a butt of pork and drinking mint juleps, but fear not, I'm sure I'll still be thinking about Optimality Theory.

Wednesday, July 2, 2008

Any X is X who X

I was struck by an odd (to my ear) phrasing by Richard Gere is a preview for what looks to be a terrible romantic comedy/drama: "Any man is a fool who doesn't appreciate you." I'm sure I've heard this construction before, but it definitely rubs me the wrong way. The problem is the isolation of the relative clause from the NP it modifies. I would always phrase such a thought as "Any X who X is X." Searching for "any * is * who *" on Google returns several hits of the same phrasing, so clearly it's not a rare construction. After careful consideration, I'm unable to make heads nor tails of it syntactically. The incorrect interpretation would be the following:

With "any" the interpretation doesn't make much sense, but replace it with "every" and there could be real ambiguity, at least in print. The only way I could see generating kind of construction syntactically is some sort of movement, where the CP starts out under the NP "any man" and then moves lower down. This is also a good example of the cognitive preference for low attachment, i.e., we want the CP "who does not appreciate you" to be attached to the lower NP, not the higher one the speaker wants us to attach it to.

Tuesday, July 1, 2008

Noun Incorporation

My wife brought to my attention a spectacular example of noun incorporation in pop culture: twinfanticipating. Apparently it was used to describe Angelina Jolie, who is currently pregnant with twins. It is, of course, a blend of three words: twin, infant, and anticipating. We couldn't agree on how to pronounce it. I opted for ˌtwɪnfənˈtɪsɪpejtɪŋ, preserving the stress of twin, infant, and anticipating, but using the vowel quality in infant for the second vowel, rather than the vowel quality of anticipating. My wife, on the other hand, insists that it should be ˌtwɪnfænˈtɪsɪpejtɪŋ, with the same conservation of stress, but using the vowel quality of anticipating rather than infant. My argument is that anticipating already has four syllables to itself, so infant should at least get that second syllable, even if it has to share both of them.

As for the word itself, the meaning is fairly obvious: anticipating twin infants. It is also a prime example of true noun incorporation. One of the hallmarks of true noun incorporation is that it decreases the valency of the verb. While non-verbal noun incorporation is debatable (where's the line between true incorporation and compounding, or are they the same?), verbal noun incorporation always includes a decrease in verbal valency, i.e., the number of arguments the verb takes decreases. In this case, the original phrase "anticipating twin infants" has a valency of 2: the anticipator and the twins. However, "twinfanticipating" has a valency of 1: only the anticipator is involved. We see this kind of incorporation all the time in Salishan and Wakashan languages, in which, for example, you could have a verb that refers to buying meat, so that one would say something like "I did some meat-buying this weekend." "Meat" ceases to become a separate argument of the verb and instead becomes incorporated inside the verb as a bound morpheme that contributes meaning to the specificity of the verb. Recently I've heard many English speakers do this as well: "online-shop" instead of "shop online." More interestingly, I was told by a waitress to "overlook" their list of daily specials when I was on my way to a conference a couple months ago (she meant for me to look over the list, not overlook it).