Learning how to learn Chinese through self-experimentation | Hacking Chinese
Language research is notoriously difficult to perform properly simply because variables in a real classroom are hard to keep constant. Let’s say that we want to examine the effectiveness of a new learning strategy. In order to be able to say anything with certainty, we need a fairly large sample and we need to find a way of testing the new strategy that isn’t biased in any direction.
One problem is that in order for random factors to balance each other out, we need large, random samples. If the study becomes too large to conduct by one single teacher, suddenly different teachers are involved, too. This quickly escalates into the realm of meaninglessness. The problem is that the variation between countries, schools, teachers and individual students is bound to be large, perhaps large enough to render the results meaningless for any given individual.
If we could prove that the learning strategy was 5% more efficient than another, that doesn’t really help you, does it? Other factors will influence you more than the actual strategy you use. For instance, if you like the method or not might make much more of a difference than the effectiveness of the method itself.
Self-experimentation and n=1
How can an experiment with only one participant (n=1) be of any value? Most scientific journals will (rightfully) say “no”, because they don’t really care about you as an individual. But you care about your own learning, don’t you? Therefore, self experimentation can be of tremendous help if that single student is yourself.
If you conduct the experiment to learn more about yourself and the way you learn, no-one requires you to follow any rules, but it does make sense to follow the scientific method in general, because otherwise your results won’t be reliable even for the single student.
In case you haven’t done much research recently, here is a crash course in the scientific method:
How to conduct these kinds of experiments in Anki
I’ve had many debates with people who think there are better flashcard software than Anki and one of the reason I maintain Anki is superior is because of it’s flexibility. In Anki (especially in Anki 2), you can manage and edit your flashcards in bulk in almost any way you wish. Split, merge, tag, anything you want.
You want to add a tag to half the cards and add a star after the final character for all those cards (so you know which characters to read aloud)? Easy. Do you want to temporarily split all your cards into two decks, track statistics for these two decks separately and then merge them again when the experiment in done? Sure. In fact, Anki already has statistics for hourly performance.
This is what my graph looks like:
There are some 113063 reviews behind those bars, so whatever differences we can see here are statistically very significant. It seems my hypothesis was wrong (obviously, I knew the answer when I wrote the article, my hypothesis was the one I had before I saw this graph). It seems I remember best when reviewing around lunch. Interesting. Not surprisingly, I have the lowest retention rate late at night.
This case happened to be rather easy because Anki already has a built-in function to check hourly retention rates, but what if you want to check something else, like if reading aloud is helpful or not? As I mentioned above, the best thing you can do is create a new deck with all words you intend to read aloud. As far as I know, you can’t sort randomly (that’s not sorting, technically), but you can sort on an arbitrary value, such as the initial letter of the English or some other factor which should be random for practical purposes, albeit not technically. Then grab half the words and change them to a new deck. Conduct the experiment and see if the groups differ. Of course, you don’t need to do this with all cards if you don’t want to, testing with a few hundred might be enough.
The two other examples I brought up yielded fairly unexpected results, actually. It turned out that both reading words aloud and colouring the tones had an effect, but that it was very small indeed; I expected the effect to be much bigger. However, remember that even though that might be the case for me, we shouldn’t extrapolate that result to you, because you might be different.
This isn’t “real” science (you probably won’t get your results published)
This isn’t” real” science, but that doesn’t mean that scientific thinking isn’t necessary. In fact, this kind of thinking is always good to adopt, even if you’re not conducting any experiments at all.
For instance, I said above that you can’t really experiment on different ways of preparing for the HSK simply because if you succeed, you can’t try again. However, you can still apply the same kind of thinking.
The results might not be reliable either, because there are many things going on that are beyond your control. Perhaps you will learn better because you know that you want one method to be better and therefore try a little harder. Perhaps you unconsciously make a hundred other small adjustments. Therefore, I wouldn’t care too much about small differences achieved over a short time with few words. If you have results like my hourly retention rates above, though, that were produced based on over 100,000 reviews, the results are probably not based on other random factors.
More than words
I’ve spent most of this article talking about vocabulary and how to pronounce characters and words, but that just happens to be a convenient example. You can conduct similar experiments in many other areas. Identify the problem, formulate an hypothesis as to how the problem can be solved, try your solution, evaluate. Repeat. Doing this will help you understand both yourself and Chinese. It will bring benefits well beyond the realm of language learning as well.
For anyone who is interested in reading more about self-experimentation, I want to recommend Seth Robert’s blog, “Personal Science, Self-Experimentation, Scientific Method”. Apart from reading his articles, you can find many interesting links and suggestions for further reading. I also suggest Scott Young’s blog, which is about much more than self-experimentation and offers clear thinking on many different topics.
This content was originally published here.