Friday, August 10, 2012

what i think about science today

Science is about useful narratives. Those that are backed up with lotsa numbers and/or clear testability/falsifiability are all well and good, but any narrative is worth listening to if it helps us see what we haven't seen before. That doesn't mean we shouldn't demand the sharper-edged claims as much as possible, but nor does it mean that we should miss out on more imaginative approaches just because they don't always immediately offer open-and-shut testability. Often if we give them a chance, they will in fact do so in the end, just a few further exploratory steps down the line. The best intellectual efforts are those that understand and respect not just the value of rigor, but also the unlimited potential of an ever-generous curiosity. That's science to me.

Friday, March 23, 2012

20120323




we use the old to handle the new

this is how ancient languages (and most all languages are ancient)
handle the new modern world

this is how we all handle new information
in conversations: new, buffered via the old and familiar

this is how a learner handles a new language,
improvising towards the new
with whatever is old to them, familiar to them, easy to them, readily
used by them.

this is how we use the old to handle the new,
the familiar to handle the novel.

[it is your family, your elders, showing you how to go out into the bigger world of the unknown]

Thursday, March 8, 2012

Event structure and programming languages

So here's an idea I'm working on: that a programming language based on event structure---more precisely, the linguistic/cognitive representational model thereof---will be intuitive and easy for a wide range of programmers to use.   Now breaking down events into sub-events has long been a fundamental task for much of programming.  To say the least.  The advantage of basing yet another such system on natural language/cognition is that it should align our computational breakdowns of event structures (i.e. chiefly as instructions on sub-events to execute) how we naturally perceive them. This alignment isn't always super-obvious, since languages tend to lump some sub-events together into chunks while parsing others out (see the work of Leonard Talmy for lots on this, and I may revise with examples therefrom).  But these kinds of analyses of the sub-elements of event structure are in my experience quite easy to learn---perhaps since they simply aim to reflect the cognitive structuring what we all already implicitly know. More on this later.

Ghost phonology in Farsi false codas

So the deal with Farsi is that it has false codas. Glottal stops only in the most artificial speech: otherwise it's just the same pre-cluster lengthening, just before no actual cluster. Ghost phonology again, like Irish mutations: the (morpho)phonlogical persistence of the past.

Wednesday, August 10, 2011

Why we have to pay for language revitalization

It's more than reasonable to ask why there should ever be government (a.k.a. taxpayer) support for revitalization of indigenous languages. After all, language shift is an individual choice, right? And what was done in the past isn't my current responsibility, right?

Well, all that would make more sense if not for two points. First is that more or less by logical definition, the currently threatened indigenous languages of this continent survived the early colonial period fairly well on their own: it's recent, no earlier than 20th-century policies that have helped push most languages over the brink.

Second is that I am pretty sure that for practically all of these languages, we can demonstrate that government power was brought to bear (directly or indirectly) down on and against the free choice of indigenous peoples to maintain their own languages in their own homes and communities. Since that is the case, denying governmental responsibility for a large part of why these languages are currently threatened is the same as, say, denying governmental responsibility for waging a past war which it now admits was wrongly motivated and/or unjustified. It is not fun for taxpayers, but nonetheless, reparations/restitution are a requirement in such cases for any government that wishes to be considered civilized and responsible.

Look at it this way: if the government came and burned down your house, especially the library you'd been building up your whole life, all the while telling you that they knew what was best for you in doing so, you would expect reparations from that government. Say, some credible amount at least towards rebuilding that library. And that government couldn't simply say, "Well, WE in particular didn't burn down your house and library. That was OTHER people who just used the power of this government to pull it off." Governments are responsible for the consequences of the actions of their predecessors. Otherwise we could totally just not worry about the national debt. So when you have a clear case of a government stepping in and actively infringing on the basic individual rights of a group, using their institutional power to coerce them into doing what the government wants and thinks is best for them, then I think you have an argument for more responsibility than just ceasing to carry out that illicit policy. There's a clear debt owed, due to the immoral and unconstitutional imposition on personal (and by extension, minority-group) rights.

Or to put it even more simply: if I burn down your house and library, my responsibilities to you don't end just because I promise not to do it again.

So this is, above all, a question of responsibility and accountability, like it or not.

Friday, April 1, 2011

How a morphosyntactician teaches phonetics

I suppose this is sort of a disingenuous title, since I am a documentary linguist by profession. And that entails a fairly broad range of background training in all the subfields of linguistics, in order to ensure that we can do adequately comprehensive coverage of all the features of the languages we seek to document.

But the reason I choose this title is because even the ostensible morphologist and syntactician (proof, such as it is, of this allegiance: my dissertation is titled "Referential-Access Dependency in Penobscot") that I am is thoroughly distressed at how we tolerate thin to dismal training in practical phonetics for all but those who go on to specialize in that field.

One might argue that this is just the normal course of specialization: we syntacticians can slack on producing and perceiving uvular ejectives and retroflex vs. alveolar contrasts because, well, we can always call in a phonetician when the need arises.

But this is not really true, for at least two reasons. The first is the obvious fact that one cannot collect remotely accurate syntactic data without solid control of any and all phonological contrasts: if tone is how they mark relative clauses, one cannot beg off on the difficulty of hearing tone.

The second and more important is this: the deeper I go into the syntax, the more I see how much of the surface syntactic patterning that we attribute to abstracted and empty features might better be accounted for by attending to prosody.

Rank unawareness of how much the phonological form of morphemes, of phrases, and so on, contributes to the surface orderings and surface wellformedness of language, can lead to some needless formal floundering, positing features that do not really do anything except make the bits show up in the surface form they actually do. In short: any work in syntax that is going to deal with PF had better, well, deeply understand all that "P" can involve.

There's a case to be made in the opposite direction, i.e. for how phoneticians and phonologists would do well to attend deeper than they already do to morphology and syntax. This is largely because the latter domains set up and drive the environments of combination and interaction in which we identify the most illuminating phonetic and phonological phenomena. But the dependency is really quite asymmetrical: phonetics in particular does not radically depend on syntactic form for all its data---whereas there is no access to syntactic data at all except through the intermediary of phonetic-phonological form.

My suggestion: have every linguistics major master the full set of contrasts (and their phonetic details) in the White Hmong system. With that done, students will have command of a rich tonal system (including breathy and creaky phonation contrasts) and some of the more challenging segmental contrasts: preglottalization, prenasalization, aspiration, extensive coronal (dental, retroflex, palatal) and dorsal (velar, uvular) features, among others. Toss in some training in pharyngeals, ejectives, length contrasts, and some of the more baroque vowel systems of the world (English, Khmer, Scandinavian), and students should be well set to handle most of what the world's phonetic-phonological diversity can throw at them.

This seems an oddly specific choice, but it is grounded in the practical fact that White Hmong is rather well documented, and fairly accessible worldwide, as a major language of the Hmong diaspora, who are substantially represented at least in North America, Europe, Australia, and of course, China and Southeast Asia.

More in the next installment on exactly HOW to teach the (supposedly) "tin-eared" to approach the challenge of non-mother-tongue sound contrasts....

Wednesday, October 13, 2010

My currently-stewing model of morphology-syntax-discourse (and phonetics and phonology)

This is a very rough sketch, but I thought I'd toss it out as is to see what responses it engenders....

Morphology, syntax, and discourse are all the same thing: structure builders. What qualitative differences exist between them emerge strictly from the scale-derived type of elements they are combining; their fundamental combinatorial principles are the same.

One qualification to this is that morphology is actually two things: a combinatorial component, and a lexical component. These are roughly the morphosyntactic and the morphophonological. The latter in particular covers aspects of morphology that are not directly structure-building, namely, paradigmaticity effects...which are probably attributable to acquisition and retention constraints and strategies.

It's useful in this to see that traditional syntax, as understood through generativist tree diagrams, is quite explicitly the interface between the generic-encyclopedic and the discourses-specific. Functional structure can be of either kind: the former realizes event-argument structure and aspect (both verbal and nominal; nominal aspect is measures, quantification, etc.), and the latter realizes all the features interpretable only relative to the specifics of the discourse, i.e. of the speech act itself. This includes voice, tense, mood and modality, pronominal features, focus, topicality, and clause type.

To understand this clearly: man bite dog is a generic, non-discourse-specific event-argument structure. What we currently call light elements---i.e. stackings of minimal predicates---are enough to constrain the semantics to this realm. Aspectuality doesn't change this: man having bitten dog, man regularly biting dog: all of these are still generic, encyclopedic, non-discourse-specific concepts.

Add in voice, and we begin to have discourse-determined priorities:

man bite dog
dog bitten by man

Add in tense, mood, etc., and we definitely have discourse-determined material, since the semantics added are calculated with respect to NOW, with respect to our REAL WORLD, etc.

man did bite dog
dog would be bitten by man

Definiteness of argments, relativization of arguments, pronominalization of arguments, ellipsis of arguments: all of these refer to pre- or elsewhere-established reference of arguments...which is of course discourse. Same again of course for focus, topicalization, etc.

the man did bite the dog

Clause-type is of course exactly the same thing as the above, even subordination, as subordination indicates NOT being the discourse-Main proposition. Same again for imperatives, which of course are discourse-specific par excellence.

man, bite the dog!
the man that did bite the dog

Phonetics and phonology of course do relate to this: prosody in particular tracks word-level and phrase-level structure quite intensely, and has a rather obviously substantial role at the discourse level. For example, the prosodic weight, and the prosodic specification (clitic, etc.) of morphemes determines their distribution, their position, their availability, their well-formedness in a given configuration. Does the same thing at the phrasal level, acting as the real agent behind the parameterization of word order, among other things. And at the discourse level, it of course manifests quite a bit of discourse morphology, i.e. old/new information and topic/focus and topic/comment contrasts, and very often interrogativity and a thousand other emotional and affective stances.

Phonetics and phonology also share properties with the morphology-syntax-discourse complex in that they too have combinatoriality and configurationality, and often the same modeling tools for the latter work well for the former. Locality, for example. We often think of this as the syntax of phonology, etc., but it's better understood as syntax and phonology both drawing from a common pool of cognitive-computational processes and constraints thereon.