Getting to Grammar: Somewhere over the rainbow, frequency grammars
I seem to have gotten myself mixed up in this big debate about learning grammar, so this post the first in an unnumbered series called "Getting to Grammar" where I lay out my strategy and respond to some of the other things banging around the language-learning blogosphere regarding grammar.
Geoff of Confessions makes a valid criticism of the state of modern grammars:
Read more...This problem can be mediated by carefully selecting your initial grammar. It should be good enough to give you a complete overview of the target language's grammar, but not so in-depth that you get bogged down in exceptions to exceptions to exceptions that you're only likely to encounter once in every five years of intense target language use. Once you've gotten what you can get out of your initial grammar, you can move up to a more in-depth one, and ultimately you can cap it off with a grammar aimed at native speakers. This, in a sense, makes three passes around the spiral, which combined with exposure should give you quite a solid understanding.
But, three doesn't make for much of a spiral and, as is, it hardly makes for an efficient process. You're likely to have already covered much of what is in any subsequent grammars you look at, and there'll be no easy way to determine where that more difficult rule is hiding; you'll have to pick through the stuff you know to find the stuff you don't. By the point you get to looking at these kinds of grammars, that review is probably not the best way to spend your time.
Moreover, the cut off point for grammars are typically determined somewhat arbitrarily, by how an editor or author feels or guesses. I've never heard of a grammar's cut off point for what is covered being backed by statistics of what rules, constructions, etc., are actually used.
Mediation of a problem is all well and good, but what really dances in my language-learning dreams are frequency grammars, or grammars that introduce rules based on how likely you are to encounter them. In your first pass around the grammar spiral, you'd cover rules that represent, say, 50% of the rules you'd typically encounter. This'll take you through much of the language, just as frequency lists do, but you'd still have quite a ways to go.
Once you're at 50%, you'd click a button (these would need to be electronic, of course) and then the grammar would suddenly cover, say, 70% of the rules you'd typically encounter. The additional 20% would show up in a different color so when you go through the grammar again, you'd know exactly what has been added since your last passthrough. And you'd keep going in increments up until 100%, with the highest percentiles representing the most obscure rules of grammar out there.
If anyone knows of anything even remotely close to this in any language, I'd love to hear about it.
Related: Don't ignore grammar, learn without grammar, or use it only for decoding
Geoff of Confessions makes a valid criticism of the state of modern grammars:
When I was in grad school, we talked about the spiral syllabus. Imagine a spiral staircase going up multiple floors: You keep coming back to the same points, but at a higher level each time. Unfortunately, conventional grammars don't do this. They typically are divided into, e.g., phonology, morphology, and syntax, with morphology broken down into nouns, adjectives, adverbs, verbs, etc. The treatments can get pretty exhaustive and the learner has to figure out how deep to dive in.How to mediate the problem, and the as-far-as-I-know non-existant solution of a frequency grammar, after the jump.
Read more...This problem can be mediated by carefully selecting your initial grammar. It should be good enough to give you a complete overview of the target language's grammar, but not so in-depth that you get bogged down in exceptions to exceptions to exceptions that you're only likely to encounter once in every five years of intense target language use. Once you've gotten what you can get out of your initial grammar, you can move up to a more in-depth one, and ultimately you can cap it off with a grammar aimed at native speakers. This, in a sense, makes three passes around the spiral, which combined with exposure should give you quite a solid understanding.
But, three doesn't make for much of a spiral and, as is, it hardly makes for an efficient process. You're likely to have already covered much of what is in any subsequent grammars you look at, and there'll be no easy way to determine where that more difficult rule is hiding; you'll have to pick through the stuff you know to find the stuff you don't. By the point you get to looking at these kinds of grammars, that review is probably not the best way to spend your time.
Moreover, the cut off point for grammars are typically determined somewhat arbitrarily, by how an editor or author feels or guesses. I've never heard of a grammar's cut off point for what is covered being backed by statistics of what rules, constructions, etc., are actually used.
Mediation of a problem is all well and good, but what really dances in my language-learning dreams are frequency grammars, or grammars that introduce rules based on how likely you are to encounter them. In your first pass around the grammar spiral, you'd cover rules that represent, say, 50% of the rules you'd typically encounter. This'll take you through much of the language, just as frequency lists do, but you'd still have quite a ways to go.
Once you're at 50%, you'd click a button (these would need to be electronic, of course) and then the grammar would suddenly cover, say, 70% of the rules you'd typically encounter. The additional 20% would show up in a different color so when you go through the grammar again, you'd know exactly what has been added since your last passthrough. And you'd keep going in increments up until 100%, with the highest percentiles representing the most obscure rules of grammar out there.
If anyone knows of anything even remotely close to this in any language, I'd love to hear about it.
Related: Don't ignore grammar, learn without grammar, or use it only for decoding
posted by Vincent at
12:06 PM
2 Comments
Links to this post
Subscribe to