Or rather, at least take pity. I’ve been wading through Eisner’s Straitjacket recently, and I’m wondering if somebody would be kind enough to tell me what a prosodic edge (represted as a single bracket), as opposed to an interior, is?
Welcome to SDMB’s. I haven’t the foggiest notion as to what OT or Straitjacket is/means other than it has to do with linguistics.
At least you now have a courtesy reply. Hopefully a real linguist will see one reply to you inquiry and be inquisitive enough to reply himself.
From Jason Eisner’s Abstracts, Doing OT in a Straitjacket (1999):
A universal theory of human phonology should be clearly specified and falsifiable. To turn Optimality Theory (OT) into a complete proposal for phonological Universal Grammar, one must put some cards on the table: What kinds of constraints may an OT grammar state? And how can anyone tell what data this grammar predicts, without constructing infinite tableaux?
In this talk I’ll motivate a restrictive formalization of OT that allows just two types of simple, local constraint. Gen freely proposes gestures and prosodic constituents; the constraints try to force these to coincide or not coincide temporally. An efficient algorithm exists to find the optimal candidate.
I will argue that despite its simplicity, primitive OT is expressive enough to describe and unify nearly all current work in OT phonology. However, it is provably more constrained: because it is unable to mimic deeply non-local mechanisms like Generalized Alignment, it forces a new and arguably better account of metrical stress typology. I will even discuss a proposal for constraining it further.
Eisner actually isn’t that scary, once you get past the fact that he can’t write.
Essentially, what Optimality Theory is all about is finding a sufficiently weak way to express how sound constraints in language work. The old way of thinking about sound patterns in human language (and the way lots of quite intelligent professionals still maintain) was to write oodles of Rules to explain trends of behavior. Rules tended to be very powerful, and to keep them from predicting things they shouldn’t they were often restricted to a single or group of languages. English, for example, doesn’t tend to have more than two adjacent syllable. We certainly write words that have more than two conson
Ack, pardon the double post; I hit enter before finishing, and since I’m unregistered I’m afraid I can’t edit it back to normalcy. In any case, while we can certainly write an english word with more than three adjacent consonants (I know I said two, but I just thought of some exceptions), we very rarely articulate more than three consonantal sounds at one time. The reason, according to old-school phonology, that we don’t get four or even five consonants together, like Polish does, is because english has a rule that, in plain terms, states something like “No clusters of four or more consonants”. (in technical linguistics we’d actually do something completely different, because it’s been repeatedly shown that grammar can’t count, at least the way the human brain handles it. But technical linguistics has been known to send many a good man or woman running for their sanity and lives. ^^)
So, we have rules and life is good. Why not stick with them? Well, for starters, they’re very specific, and since a large number of phonologists and theoretical linguists in general start with the theory that there is a universal grammar (UG), i.e. a set of universal Things regarding the way language can be constructed which every human is born with, many theoretical researchers like to avoid making language-specific statements about typology whenever possible. Second, rules are far, far too powerful. No sooner will we make a rule than somebody will find an exception, so we write a new, more specific rule, to cover this exception. The concept that the human language faculty intentionally comes up with big rules that don’t even work half the time, and lots of smaller rules to make sure the big ones function, is just silly. There are many interesting and ungodly complex ways of getting around this, but in the early 90s a computer geek and a linguist came up with a very interesting alternative: Optimality Theory.
In 1993, Alan Prince and Paul Smolensky proposed that instead of a rule-based (derivational) phonology, we should be looking at a constraint-based phonology, e.g. instead of saying “There exists a rule stating that the english plural marking is ‘z’, and only becomes ‘s’ in certain contexts,” you would write a constraint (or, infinitely more likely, a series of constraints) that said something like “Sibiliants (fancy term for ‘s’ and sounds like it) must harmonize with their preceeding consonant in voicing”. For our purposes, this basically says “If ‘s’ comes after a sound that uses the voice box, like ‘d’, then it too must be pronouned with voice, and become ‘z’”.
(at this point I’d like to apologize to every linguist on the board for vastly oversimplifying complex things. And yes, I know that the underlying representation of the english plural marking is actually ‘z’ and the OT constraints that predict the proper pattern look nothing like what I just said, but bear with me. ^^)
Now the essential thing to realize about using constraints instead of rules, is that their implementation revolves around this concept: every word has an Input form and an Output form. There is exactly one Input for a given meaning (only one word per concept), and an infinite number of possible Outputs for every word. Given this set of infinite outputs, the grammar will select the optimal output form for the word.
So, we’ve just come up with a system three times as complicated as it needs to be in order to awkwardly accomplish the same task that Rules were already doing. The advantage, however, lies in one simple fact: constraints can be shown to be universal. It is very feasible to propose that every (or nearly every) phonological constraint exists in every language, and the only difference is the order in which constraints are obeyed. As such, we might propose that English and Polish both have a “Consonant clusters must not have more than three sounds” rule, but that Polish has a different constraint which is valued more highly than the consonant cluster constraint, and it’s this constraint which lets the Polish use up to five consonants in a given cluster.
It’s fun to look at language in this way since it’s quite a lot like life: it supposes that everything always violates lots of constraints, no matter what you do you’re going to break the rules… the trick is making sure you break the small ones.