I've been thinking a lot about L2 input and learning rate and tools like L-R, L1 vs L2 Subtitles, my collection of 'learning' dialogs, etc. It has occurred to me that there is a rough formula for how well these things work: ----------------------- Learning rate = Volume * Parsing% * Meaning% Volume: words per unit time Parsing: % of words you individually parse out and really "hear" / "see" Meaning: % of words/messages you get the meaning for (one way or another) ----------------------- Some examples that I think fit the model: Beginner trying to listen to L2 radio all day? Volume: Very High Parsing: near 0% Meaning: near 0% Result: nearly nothing Japanese Anime watcher with L1 subtitles and near zero Japanese listening skill (who quickly ignores the L2 audio as a wall of incomprehensible sound) Volume: High Parsing: 0% Meaning: High Result: nearly nothing L-R'ing L2 audio, L1 text Volume: Very High Parsing: Low to Moderate Meaning: Moderate to High (depending on how "parallel' the translation is) Result: moderate-good progress, increasing as parsing ability improves L-R'ing L2 audio, L2 text Volume: Very High Parsing: High to Very High Meaning: *** Depends a) Cognates b) L2 Vocab level c) familiarity with the story/text Result: Highly dependant on ***, increasing as parsing ability and vocab improves Grinding Learning Textbook dialogs over and over Volume: Very Low Parsing: ~100% (depending on usage) Meaning: ~100% Result: low, DECREASING as vocab level saturates the volume of text Intensively watching a video and pausing to look up every word in the subtitles you don't know: Volume: Low (depending on existing vocab level, and look-up tools) Parsing: High Meaning: Moderate to High (again, depending on level and look-up tools) Result: low to moderate ----------------------- Of course there are lots of other factors that can effect these inputs. For example, motivation can effect both Volume (you get bored and do less of it or get excited and do more) and Parsing (bored: you don't really paying attention to the sounds, excited: you are hanging on every word or perhaps replay bits you miss.) And in languages with diglossia issues or non-phonetic written languages, things get a bit more complicated as "parsing" audio and "parsing" written text can be very different things. For example, in Cantonese, the Standard Written Chinese subtitles you find almost everywhere do not match the audio, at times they are little more help than the English subtitles as far as listening comprehension... i.e. they only provide meaning and do not help with "parsing" the audio. I'm sure this fits in the "obvious" bucket for most experienced language learning folk, but I thought I'd share anyhow as it was a small "Aha!" moment for me.
This appears to be specific to vocabulary, rather than learning in general. It's an interesting idea, and you might get something useful out of the basic concept, but in general I disagree. How about adding some units? Let's say LR = (80 w/m)(.25)(.25) = 5 w/m In verbal form, if I study 80 w/m, I can parse, or separate out and recognize the fact that 25% of them are specific words, and I know the meaning of 25% of the words that I can parse….then I'm learning 5 w/m I don't think so. If those were the only factors involved with learning, you might be able to say your learning rate is proportional to that product. But imo, there are hundreds of factors, and if it were possible to represent a learning rate by an equation, it would almost certainly be more complicated than a simple product.
I'm finding it useful to evaluate my listening and adjust my activites accordingly. For example, if I find my parsing is low, I might do some transcripting, or just turn of video.
Oh, I'm certainly aware that this is a very rough model ("formula" was probably a poor word choice), and I realize that even the concept of "learning a word" isn't simple. I would never say that this is something that could be boiled down to actual numbers in a meaningful way, only that things like "Volume", "Parsing" and "Meaning" are multiplicative factors that clearly effect input such that if any of them drop to zero, your results follow directly. And I believe this applies to more than just vocabulary; each time you witness and understand a usage of grammar, an idiom, a colocation... you have an opportunity to learn some bit about the language. To steal a term from advertising, we could call each of those opportunities an "impression". Just like in advertising, where each impression doesn't necessarily turn into sale (in fact a very low % do), each language impression doesn't mean that you will "learn" that item, particularly long term. But all else being equal, the more impressions you get, the more things will stick. Thinking about it a bit more, I'd modify the model to: Impressions = RawVolume * Parsing * Meaning * Affective Filter Now, how impressions turn into "Learned stuff" is, I agree, quite complicated (forgetting curves, interference, partially learned bits and all that jazz)... but I would still assert that, all else being equal, the more Impressions you get per unit time the faster you learn. (Notice the use of the greatest scientific cheat: "all else being equal") Anyhow, this is just fun musings as opposed to serious theory. But I am curious how it jives with other folks' mental models of these elements.
