Monday, November 10, 2014

Good questions we'd love to have answers to

Hubert Haider and Henk van Riemsdijk had a terrific idea. Inspired by Hilbert they decided to put together a book that identifies interesting open questions in syntax. In the original invite, Hubert and Henk suggested that syntacticians revive the lost art of posing interesting questions:

Some are particularly talented for asking, some are specifically talented for answering. We tend to restlessly thrust answers upon our scientific community. But – more often than not – questions would have been even more welcome than answers. Why not share questions?
Let’s try the possible, namely repeating in linguistics what David Hilbert dared for mathematics with his famous 23 problems.[1] His questions ranged greatly in topic and precision, but most of them have been propounded succinctly & precisely enough to facilitate a clear answer; compare for instance Hilbert’s 20th problem.[2]

To aid thought, Hubert and Henk proposed some well-formedness conditions on good Hilbertian questions (I elaborate a little on their guidelines below):

Precision must be the aim. Even if facts are occasionally cloudy, the questions should be crystal clear. A commenting paragraph will be necessary for most of the questions and references to literature. Surely, you are aware of the saying: Even a fool can ask more questions in five minutes than twelve wise syntacticians can answer in their life time.
Here is your task if you are willing to take up the challenge: Select up to four of the following categories and submit one syntactic[3] question for each of the categories chosen.
i.   Generalization that holds of cross-linguistic facts. In other words, formulate what should become known as “Y’s generalization” or “Y’s problem” (with Y as the variable for your name).
ii. Generalization over facts of a particular language (= language-specific case of (i.)).
iii. Theoretical issue: a novel theoretical claim
iv. Your favourite question: “The question I always wanted to find the answer for,  …”

Hubert and Henk’s brainchild was quickly adopted in conception. The two asked a ton of linguists (this is a lot, btw) if they would participate and they received a rousing response, or at least a lot of intellectual IOUs. However, if the responders were like me, this was a lot like agreeing to attend a conference two years in advance (meaning, as reality set in and the work required became clear, you became less and less enthusiastic). The end of the story is that fewer delivered on their implied promises than Hubert and Henk had hoped and so they decided to kill the project. Here’s where Faculty of Language (FoL) comes in.

On reading their last note (more than a little disappointment there) I asked if they would be interested in pursuing this project in a more “stately” manner. Rather than shooting for 100 good questions all at once, we run this as a blog project and we allow the questions to accumulate over blog-time. This would allow the questions to be solicited sequentially. Deadlines would be avoided and people could decide to contribute after seeing how this thing was done. After all, nothing similar within linguistics have ever been tried[4] and, IMO, Hilberts good at asking good, timely, answerable questions are not that thick on the ground, not even in the very talented pool of syntacticians.

So, the good news. Hubert and Henk decided to relocate the project onto FoL. They will still manage the content (with some help from me). We hope that those that Hubert and Henk solicited earlier to contribute will still play and submit their favorite open questions. However, we’d like to open this up to anyone out there that wants to contribute. All the questions will be vetted, but I hope that all readers of FoL will feel free to propose questions for general consideration. To make this manageable, I propose that no submission be more than 4 pages. Please submit in a standard format (e.g. Word) as this makes editing and posting easier.

An important addition: we also propose opening up the scope so that questions from any domain of generative linguistics are invited. So phonologists, computationalists, psycho types (both parsing and acquisition), semanticists, morphologists and, of course, syntacticians are ALL welcome to submit. The only restriction is that these be questions that fit comfortably within the generative tradition. Hubert, Henk and I will review these and decide what to go with. I am sure we will ask for outside advice as well.

Now many of you may be thinking that this is actually a pretty tough kind of paper to write. I agree. So I thought I’d end here by mulling over the questions: what makes a problem Hilbertian? The short answer is, it’s complicated. Here’s a longer answer.

As you all know, Hilbert was not a linguist. He was a mathematician (and a pretty good guy apparently. I heard that the reason that all mathematicians wear their sandals with socks is that Hilbert wore his sandals with socks. I also read that he was Emmy Noether’s champion (he got her into the PhD program and found her a job (though not a great one given her obvious talents) at a time when being “feminist” was not at all fashionable in German academic circles). Now, I need not observe that linguistics is not math. The most obvious difference is that linguistics is an empirical discipline so that it will be very hard to specify a question that will get a dispositive answer. There are no such questions in the sciences.

However, that said, there are questions that can get close to being Hilbert like. These will tend to be based on theoretical assumptions and will probe features of their conceptual structure. Such theory-facing questions can be pretty cleanly formulated. So, do not resist the temptation to formulate questions making pretty hefty theoretical background assumptions. These kinds of questions come in at least two flavors.

The first kind of question is asks can we get there from here? In other words, given such and such assumptions is it possible to model this and that effect and if so how. So, for example, given phases can we model islands and if so how (the answer, btw, is yes, but not in any interesting manner so far as I can tell). Note saying that it can be done, does not mean that it should be. Nor need the answer be always positive. It is useful to discover that you can’t get there form here too. Such questions can be crisply stated and answered precisely because they are theory-facing. Whether the answers are also true or identify the right way to go is still a very open question. But, I believe that there are many questions of this kind still out there and that are worth posing. 

A second flavor involves unification of our axioms. So Chomsky asked if we could unify structure building and movement and answered yes (hint: via a certain conception of merge). I’ve asked how to unify movement and control (I concede this has gathered less praise than Chomsky’s, but them’s the breaks). Again, it is important to separate the question of whether this can be done (and how) from the question of whether this should be done (i.e. do the empirical facts warrant the unification).

Note that both kinds of questions can be supplemented with suggestions of how to test whether the proposals have empirical legs. This kind of addendum is very welcome. So, for example, something along the lines of: Question Q can be tested by looking for Gs with property P.  If this can be specified as well, you have hit the Hilbert jackpot.

Of course, these are not the only questions interesting to GGers.  There are many others. But, they are harder to state as crisply. Nor should we expect them to be. I’ve asked a few of them on this blog (e.g. Why Morphology?). One asked as a reply to Hubert and Henk’s last letter by Maria Rita Manzini is: Do we want late or early lexical insertion? Here, framing the problem crisply so that it will be answerable is quite a bit harder, involving both theoretical and empirical considerations, with the latter being more prominent. However, these are great questions, the challenge being to ask them in more than one sentence.[5] 

At the risk of going too far out on a limb (go ahead and cut it, see if I care) here are some diagnostics for a Hilbertian question in GG:

1.     It builds on something that we know a lot about both empirically and theoretically, i.e. we have lots of relevant facts and interesting analytic attempts to explain them.
2.     The question implicates (bears on, has consequences for) one of the leading ideas of GG (bears on some MP question, or Plato’s Problem, or some basic assumed feature of our theoretical apparatus (e.g. bears on what a movement rule might be, or what a derivation is or…)).

The first desideratum should make it possible to ask a question that’s not too vague. The second indicates that we have a question that is worth asking and answering. So, it would be great if any submission included a clearish exposition of the problem and a little discussion of why the problem is interesting.

That’s it. Take my above comments as suggestions on how to proceed, nothing more. I am sure that there are many other ways of approaching this, as your submissions will show us.

Oh one more thing. What to call the series? I was thinking “Generative Hilberts.”
Why this? Because ‘Hilbert’ sounds a bit like ‘Filbert,’ which is a nut (hazelnut actually and they are delicious!!). Nuts are very nutritious, are good for hoarding, taste great and need cracking. So are good questions.

Henk suggested ‘Tharl,’ short for “Towards a Hilbertian Agenda for Research in Linguistics.” ‘Tharl’ sounds like the name of an often angry Norse God that requires constant care and feeding to prevent his going on a dangerous rampage. In other words, a good name for inspiring you to send in a constant stream of questions.

For now, I will leave the series name open. Let me know if you have any suggestions.

Before ending, I want to again thank Hubert and Henk for having initiated this project and for agreeing to move it to FoL while continuing to help manage it. I also hope that you all agree to play. Hubert and Henk are right to point out that questions are often as (if not more) important to a vibrant field than answers. Let me add that comments to posted questions are very welcome and, I believe, important. If they are like comments generally posted on FoL they should allow the questions to be significantly sharpened and clarified, to the benefit of us all. Personally, I am greatly looking forward to this. Let the questioning begin!

[1] David Hilbert (1900): Mathematische Probleme. In: Nachrichten der Königlichen Gesellschaft der Wissenschaften zu Göttingen, mathematisch-physikalische Klasse 3:253–297. Göttingen. Vandenhoeck & Ruprecht.
[2] Do all variational problems with certain boundary conditions have solutions?”  The question is resolved.
[3] Of course, this includes the interfaces with phonology, semantics, and pragmatics (information structure), and brain & processing issues in relation with psycho-, neuro- and biolinguistics.
[4] This is not quite true. As you all know, great minds think alike. If you take Henk and Hubert as adding up to one Newton, then they had their Leibniz in Andrea Moro. He ran a Hilbert like questions in ling conference in Pavia November of last year. Here’s the link:
As should be clear, when three such distinguished people have the same kind of idea, the time is ripe to pursue it.
[5] Henk correctly reminded me that one can go further still. To quote him:
I believe that today’s problems are yesterday’s mysteries, and I am not talking about last year’s, but, say, 100 or 200 years ago. From this perspective I would not be adverse to also including mysteries that we would like to see transformed into problems as soon as possible.
SO, all you mysterians out there wishing to push us towards enlightenment are more than invited to contribute as well.

1 comment:

  1. Very nice initiative. How do we send our questions? Is there an email address?