Max Headroom in the phonetics lab

This report has been circulating, in slightly less detailed form, on the BBC. It describes researchers 'reading off' speech signals in a coma patient. Sadly, but not unexpectedly, the BBC story is a trimmed-down version of the New Scientist story (already under 700 words), with a few quotes from other brain researchers thrown in for good measure, and, of course, filled with suggestions that the man in the coma is about to be turned into Max Headroom. This is not entirely true.

First of all, from the NS article, it looks like the Boston University Speech Lab is responsible for this research, but I can't find any reference to the project on their website. The description is of an invasive technique, implanting electrodes directly in poor MH's brain. (His real name seems to be Eric, but one would think they'd have changed it for the purposes of reporting the research anyway, so I'm going to stick with MH.) Sadly, their computer can't recognize 'N-N-New Coke' in the output of this little bundle of neurons, but it can apparently distinguish /u/, /o/, and /i/, and do so with 80% accuracy. Evidently they tell MH to 'think hard about saying /u/' and he complies in his comatose state. Until the details come out, I'll be a bit skeptical about what's really going on.

Nevertheless, what is there on the BU speech lab website is interesting: first, brain imaging data for speech production, most notably the apparent location in the brain of certain articulatory signals. Second, a computer model which they use to situate what they think these groups of neurons are doing. I won't evaluate either one, but if we believe their brain maps, and if we believe they've stuck the electrodes in the right place, then it seems like they are really reading off something like articulatory information.

This is somewhat interesting once you realize that the actual form that articulatory information takes is still up for debate. On the one hand, there's the fairly obvious theory that, when we speak, we just send instructions to the articulators. Of course, if sounds are stored in this articulatory format, then perception must involve something like the Motor Theory of speech perception, which means that you have some (presumably built-in) hardware for matching up speech sounds that you hear with the gestures that produced them. This is how you match up the sounds you hear with stored forms, which just tell you what to do.

On the other hand, a goal-based theory of production (subscription to Journal of Phonetics required) says something like the reverse. The information you need to send to the motor system is (mostly) a bunch of acoustic targets. You might also have articulatory targets, but the key thing is that you can just send 'I want a low f2' to the low-level system and it will work it out automatically, presumably, again through some built in mapping. So when we map out what parts of the brain are lighting up when we say /i/ etc, we shoul consistently see things corresponding to these more abstracted features. If we don't, then we can't tell whether the goal-based theory is right.

Of course, this is not so easy to tell for /i/, since the mapping between acoustics and articulation for the vowel space is fairly trivial, but if we had enough brain data we should in principle be able to tell; do the bits that light up for particular sounds in production seem to correlate with acoustics or articulation? Clearly, the articulatory signals have to be there, but if we can't see the acoustics, we don't have any reason to believe in the goal-based theory. This is factoring out any methodological concerns, of course, which I take it are acute with brain imaging. But it would be interesting to look (and if anyone wants to hunt through the stuff on the BU Speech Lab web site and try and find data bearing on this question feel free).

Bemoaning the Loss of English Subjuctive

The New York Times published an article today about efficiency in the workplace. The book is called The 4-Hour Workweek by Timothy Ferriss. This is what Po Bronson had to say about it:

'“It’s not saying the ‘20-Hour Workweek,’” Mr. Bronson explained. “That would be something that lots of people can live. It’s 40 hours a week versus four. It’s very important in the tech world that consequences are exponential, not geometrical.”'

So what does he mean? Does he mean that in the tech world, concequences are defined as being exponential, not geometric (declarative reading)? Or that the tech world isn't going to care unless the consequences are exponential (subjunctive reading)?

Now, if English still had a subjunctive in regular use, I would know: if he meant this subjuctively he would have said "It’s very important in the tech world that consequences be exponential, not geometrical." Since he didn't, this is declarative. Unfortunately, I think he might actually have intended this to be read subjunctively.

Now, I know what you're thinking: wait a minute, aren't linguists supposed to be all desciptivist and shun prescriptivism? Well, to that I say two things:

1. I come from an intensely prescriptivist background. It's not that easy to throw off, you know!

2. The main reason I'm posting this is to talk about language change. What happens when a change causes confusion? Doesn't it usually ultimately fail? Or, something comes in to take its place?

It seems to me that the loss of the subjunctive is too far along to reverse itself. Once language change gets going in earnest, there doesn't seem to be any stopping it. But if situations like this keep coming up, where I can't figure out from context which of two ambiguous meanings was intended, can we expect to see something come in to take its place? Maybe a slightly awkward injection of a modal, like "It’s very important in the tech world that consequences should be exponential, not geometrical"?

Syntax or Lexicon?

Ok, if we're gonna get this blog started again, I thought a good way to do this would be to ask a question.

So, I've got a question. A rather vague "how do you feel about this approach?" question, that's been a bit of a topic in my syntax class.

I've recently been trying to make my way through"The Normal Course of Events" by Hagit Borer. One of her main themes is that several phenomena traditionally accounted for through lexical specification should actually be accounted for through structural/syntactic means.

For instance, Borer argues that the difference between mass nouns and count nouns, which is traditionally accounted for by stating that each noun in the lexicon is specified as either mass or count[1], is actually structurally represented, so that count nouns have more functional structure than mass nouns.

This is extended to the verbal domain as well. So the difference between unergative and unaccusative verbs, argues Borer, is not a lexical difference, such that "run" is specified as an "unergative" verb, while "arrive" is specified as an "unaccusative" verb, but is actually a structural difference such that unergative verbs have more functional structure than unaccusatives.

And again, the difference between telic and atelic verbs is that telic verbs have more functional structure than atelic verbs.

So what do you think about this approach that consists of taking all of the information out of the lexicon and instead representing it structurally? Like it or don't like it? Good move or bad move? Why or why not?

As for me, going with my gut feeling, I kind of like the idea that everything is structurally represented (so that even a notion like "verb" and "noun" is derived in the syntax). But then I wonder: what does such an approach mean for languages like Yucatec Mayan?

In Yucatec Mayan, there are (at least) two types of intransitive verbs: inherently telic, and inherently atelic intransitives. The inherently telic verbs, when unmarked, are interpreted as (hence the name) telic. They need to be overtly marked by an atelic morpheme to be interpreted as atelic. The inherently atelic verbs, on the other hand, when unmarked, are interpreted as atelic, and they require an overt telic morpheme to be interpreted as telic.

So if Borer is arguing that atelicity is universally represented by less functional structure, what does it mean to have these inherently telic verbs, where an overt morpheme marks atelicity, and telicity is unmarked?

Now, I have yet to fully read Borer (and have the suspicion that her theory actually may be able to account for this [2]) but at first glance this seems problematic, IF I assume that overt morphological marking corresponds to more functional structure.

So yes, two questions: What do you think of having all the information represented structurally, as opposed to being listed in the lexicon, and is the assumption that overt morphological marking corresponds to more functional structure valid?

--------------------------------------------------------

[1] On a random note...why is "e-mail" a count noun, but "mail" a mass noun?
[2] My paper topic for this syntax class! It's getting somewhat ridiculous: I don't know how to get around using the following phrase "inherently atelic (i.e. possible unergative) intransitives," but really, there ought to be a limit to the number of morphological negations you can use in a single noun phrase...

The Ig Nobel Prize in Linguistics

"A University of Barcelona team for showing that rats are unable to tell the difference between a person speaking Japanese backwards and somebody speaking Dutch backwards."

Source: http://news.bbc.co.uk/2/hi/science/nature/7026150.stm

Presumably this means they could tell the difference between Dutch and Japanese spoken forward...

We've got to get going on this blog again!

Another thing...

So I tried out my analysis below on some members of my reading group last night, while waiting for everyone to arrive, and everyone thought I was jumping the gun with the "quantifying over deictic spheres" instead of "worlds." Which is fine, because what I was really trying to get at, I think, is the idea that BF may lack an existential quantifier. Which would then explain why whenever we try to elicit things like "a man," "some men," "a few men," we get elaborate round-about ways of conveying these. (The few quantifiers over sets that we've found so far all seem to be interpretable as universals, like 'all' (duh) and 'both'...)

So now what I'm asking: Does anyone know someplace else where someone might expect an existential quantifier, in a language like English?

And just because I like to complain, UBC says that this article on (supposed) existential polarity items in Chinese is (supposedly) available full-text online, but the referred site keeps asking me for $32.00! Now I might have to do something ridiculous, like actually, physically, go into the library and find the journal...

Trying to piece together an analysis

Do you know how you can see bits of a picture, and then your mind will fill in the empty places, so that you think you're seeing a complete picture? Or you can see the first and last letter of a word, where the middle letters are all mixed up, and your brain will unmix-it-up for you? I've been hoping that my brain would manage to fill-in and unmix-up a half-baked analysis that's been mouldering in my mind for something like 8 months. Unfortunately, my brain hasn't quite managed to do it yet. So I'm asking for help! Or assurance that I'm not crazy.

Now, this research is what I've been focusing on for the past two years, and it's very dear to me, but I've been reluctant to post on this stuff before. This is because the relevant data really belongs to my language consultant, not me, and I don't think it's right to just splash it all over the internet. But I think I've managed to write it in a way that avoids this problem since I don't actually provide any data.

So here is my full-of-holes-and-faulty-logic working analysis:

Assumption 1: In IE languages like English, semantic truth-values are encoded through Tense, or IP, in the clausal (verbal) domain. (Kearns 2000)

Messy Assumption 2: The semantic equivalent of truth-values in the nominal domain is existence, or referentiality. (Because I am, er, semantically-challenged, I'm not sure what the difference between existence and referentiality is. )

Now, here is a hole. Is this assumption justifiable? I figure that nominal elements don't really have truth-values, but something more like an existential, or referential value. Or looking at it the other way, a noun has (or doesn't have) a referential value, and the clausal/verbal is considered 'true' if event being referred to exists, and is 'false' if the event doesn't exist. Does that make sense? Man, I really wish there were more undergraduate classes in semantics. But anyways, assuming that the above, er, assumption [i], is justified, I'll move on with the holey analysis...

Generalization 1: NPIs are known for having existential narrow-scope - they are non-referential (Progovac 1994, Uribe-Etxebarria 1996)

Assumption 3: Blackfoot lacks the syntactic node Tense, utterances instead being anchored deictically via a Participant (as in Speech Act Participant) node (Ritter & Wiltschko 2005).

Now, here's where my (poor excuse of an) analysis get's really vague and holey. I have a load of questions. If Blackfoot doesn't have the syntactic node Tense, how are semantic truth-values encoded? Or are they even encoded? Because there's this idea that several languages do not ASSERT information, but instead PRESENT information (or so I gathered from my LING 447 evidentials seminar...but have yet to find a citable reference) Does the Participant node encode truth-values, or some other semantic, perhaps speech act participant-related, property being encoded?

BUT anyways, if I jump over that giant hole in my analysis, and assume that there is some kind of Participant-related-semantic feature encoded by Participant node in the clausal/verbal domain, the simplest stab at what the nominal equivalent would just be Participant. Right? Because while it doesn't really make sense for nouns to have a +/- truth-value, they can certainly have a +/- SpeechActParticipant value.

Now putting this all together, one might predict that Blackfoot NPIs wouldn't have an existential property within the scope of negation, but instead have a parallel SpeechActParticipant property within the scope of negation. (Which, of course, I already know is the case, and I'm just pretending is a prediction for presentation's sake...)

So, despite the giant holes, does this analysis make sense? Is it interesting? I hope it's interesting...

And a question from this: How might such a semantic 'SpeechActParticipant' property manifest in other parts of the language? What kind of predictions does this analysis make? How cool would it be if this language didn't quantify over worlds, but instead over deictic spheres? Or something like that...
----------------------------------------

[i] I know Giannakidou's 1998 notion of semantic (non)veridicality (which is directly related to truth-values) translates (non)veridicality onto the nominal domain (with determiners and quantifiers) in terms of whether the denotation of the NP is nonempty or not...which kind of feels like the same sort of thing I'm trying to say. Except, of course, the semantic terminology is a giant barrier in my understanding. See now, I really could benefit from a 'Semantics Awareness Week'...(Hey, they have those wristbands for just about anything. Why not this?)

Linguistic Ecumenism

Hi. I've been MIA for a bit, and haven't posted yet, so I thought I'd better say hi, and thanks for inviting me to be part of the team and stuff.

Today I'm going to talk about linguistic ecumenism.

In my two years at the fringes of the linguistics academic community, one thing I notice over and over again is that different fields of linguistics don't talk to each other. I believe the reasons for this include the following:

1. Time. Who has time to read a lot of research in her own field, much less the less-relevant fields?

2. Disdain. It is easy to get dogmatic about something you know a fair bit, but not a huge amount, about. My first exposure to linguistics was Chomskian*, and the arguments for it are quite conviincing. The arguments against it are not given. There are plenty of structural linguists, who, with good reason, think sociolinguistics is premature: how can we study language variation when we don't even know what language is? And there are plenty of sociolinguists who, also with good reason, think structural linguistics is oversimplifying to the point of failing to capture anything about language. Etc.

3. Ignorance. Linguists who know nothing of a given subfield of their discipline are likely to fail to see the value of it, or even fail to notice its existence.

Another thing I've noticed is that there are many students, especially at the undergrad level, who are still interested in subfields of linguistics that are not their specialty. I've spoken to many young linguists on this very topic, and while some are somewhat agreeable for the sake of being friendly or polite, and throw in half-joking comments like "well, sure, except for [insert derided subfield/theory here]", many are in full agreement. Enough that I say we start an explicit movement.

THE ECUMENICAL LINGUISTICS MANIFESTO

A spectre is haunting linguistics -- the spectre of Ecumenism. All the powers of old linguistics have entered into a holy alliance to exorcise this spectre: prof and chair, Chomsky and Labov, MIT dogmatics and UCLA postmodernists.

Where is the theory in opposition that has not been decried as foolish by its opponents at MIT?

Two things result from this fact:

I. Ecumenism is already acknowledged by all traditional academic powers to be itself a power.

II. It is high time that Ecumenical Linguists should openly, in the face of the whole linguistic world, publish their views, their aims, their tendencies, and meet this nursery tale of the spectre of ecumenism with a manifesto of the movement itself.

To this end, imaginary Ecumenical Linguists of various institutions have assembled in my head and sketched the following manifesto, to be published in English, IPA and possibly French if we feel like it.

We have laboured under the pretense that separate subfields of linguistcs are in conflict with one another. No more! We must recognise that linguistics is a science, and therefore (1) is virtually always wrong, and (2) must by definition be constantly open to refutation, alternative theories and new data. A good scientist is excited by problematic data and opposing viewpoints. A good scientist struggles against her ego in the service of the search for knowledge. Brothers and sisters, the ego must not win!

We, the new generation of linguists, must rise up against this dogmatic separation! We must fight for the time to keep up on research from outside our tiny specialties! We must fight down the egoistic need to be right, and embrace opposition, debate, and cooperation!

ECUMENICAL LINGUISTS OF THE WORLD, UNITE!



* A math prof once told me that the greatest compliment a mathematician can get is for his name to be used as an ordinary noun: uncapitalised and subject to morphology. Chomsky's halfway there.