I recently started using a wanikani reordering script and I found that doing 1x1 reviews (where a review of a particular kanji is proceeded by its reading, etc.) makes my reviews go by much faster.
Has anyone found any downsides to this? In terms of memorization? I assume there must be a reason that Wanikani orders them randomly by default, but so far this method seems to work best for me. I just don’t want it to hurt my ability to learn what I need to know.
The one down side is that 1x1 sort of cuts your exposure to items in half, because each item comes up once (together) instead of twice (separate) per review session. I personally didn’t feel like that was an issue, but others did notice some negative effect.
If you find that to be the case, there’s something you can do instead of 1x1. It won’t make your reviews go a lot faster, but your memorization will be much better: Each time an item comes up, recite both the reading AND meaning out loud (or ‘vocalize’ it in your head if you want). Then look whether it’s asking for the reading or meaning, and answer accordingly. The key here is that you are mentally pairing reading and meaning, which helps create a strong association between the two. Some would argue that this works against “interleaving”, but I personally think that is a misapplication of what the research says about interleaving (specifically, I think it’s a misapplication regarding reading and meaning).
Anyway, I used 1x1 for most of my time on WK, and I loved it. But I also did a lot of Self-Study, which would have countered any negative effects from less exposure.
In addition to what rfindley said, I’d say another downside is that whenever you get an item wrong, you have to type in the correct answer immediately instead of after a few more other items. If I get the item correct, there’s no problem, but when I get them wrong, am I really remembering the correct reading/meaning if I saw the answer seconds ago? Like, if I’m not paying attention I might not actually be remembering the correct answer anymore even after just a few more items. (The key is to slow down and spend extra time on the items you get wrong, either then and there or after you’ve done all your reviews. Preferably both.)
That’s really the only downside to 1x1 for me (not that it bothers me much), I still go 1x1 because it is faster and predictable, especially since I know reading comes first each time and then the meaning (that’s my preference).
Were you self studying kanji outside of Wanikani or just Japanese in general?
I wrote the Self-Study Quiz script to drill my WK items.
(and the companion script Self-Study Hide Info, which makes it easy to quiz items by level from the Level pages on WK)
I do both. When I use WK on my laptop, it is 1x1 reviews and on my desktop, it is random. I haven’t noticed any real difference in how well the items reviewed seemed to stick. 1x1 is less prone to having items drop multiple SRS levels in a single session though since you won’t get it wrong the second time.
Not an expert on the science here but from what I understand, the more your brain struggles to focus on and remember something in the first place, the less likely you are to forget it.
Fun sidenote- researchers have developed a font that’s deliberately harder to read, to enhance this effect- check it out at https://sansforgetica.rmit.edu.au/
So the reason you find 1x1s easier is that your brain is doing much less work to recall them. That helps you in the moment, but also means that you miss out on some of the struggle, and some of the long-term learning potential.
The tricky question is ‘is this effect big enough that I should change what I’m doing?’. And the answer isn’t clear. So if it makes you enjoy your reviews heaps more to do them that way, you might accept the trade off in less efficient memorisation. The likely worst case scenario is that you’ll get them wrong a few more times and have more reviews and marginally slower progress. That’s not really a huge cost, in my book. Up to you!
Edit: been doing a bit more reading, and it seems that the font I linked to doesn’t actually work. Yay for science disproving research with better research!
TBH for me every single downside is outweighed by the fact, that this way you can think of both the reading and the meaning at the same time, which saves both time, and sometimes lets you catch a mistake before you make it. The sheer speed of reviewing this way is just nice.
I won’t claim to be an expert either, but I personally think the conclusion about difficulty making learning better is flawed… or at least limited in its applicability. The brain is very good at shortcutting things that it thinks (at the neurocircuit level) that it has already learned. And that’s often the case with the ‘easy’ task in a learning experiment. In other words, it’s not really true that the more difficult task is desirable, it’s just that the easier task is flawed. But most researchers in the ‘learning’ field won’t realize that because it’s somewhat outside of their normal expertise.
In many cases, though, you can design a better ‘easy’ task that will outperform a more difficult one. It “just” requires an understanding of what’s going on at the neurocircuit level.
For me, too. I don’t care about, say, 10% better memorization if I learn 50% less material in the same period of time
You raise interesting points and could well be right. If it’s a 10% effect, though, that’s huge- because there’s around 8000 items in WK and we review them each 8 times, being 10% less accurate at any point along the chain would add at least 6400 extra reviews over the course of completing wanikani. That’s weeks of work!
Even if it’s only a 2% improvement in accuracy (which you probably wouldn’t notice as you go along, I certainly wouldn’t) that’s more than 1000 reviews that you don’t have to do again later.
The sheer numbers of reviews is what nailed it for me- even though I’m not clear there is a huge benefit to doing reviews in larger batches, it isn’t annoying me enough to stop doing them that way, and I’ll cross my fingers for a potential accuracy benefit somewhere down the line. Others make different choices, which is also great, no criticism at all.
If you really cared, you could probably offset that loss by doing extra study on the items you’re less sure about.
Have been using 1x1 for the longest time now (always reading then meaning) and I can’t say I experience any downsides in retention but what I did realize is that I have a far stronger connection between meaning and reading and I focus less on the connection between meaning and kanji shape. Increasingly I’ll make careless mistakes by virtue of homophones, not visually similar kanji
Whatever the implications, all in all it’s just a more enjoyable mode and let’s me do reviews faster, which results in me doing reviews more often.
I think the interleaving theory increasing retention doesn’t hold up, since when you get the other part of the card, it is still in your short term memory. Just spamming flashcards within a very short timeframe isn’t effective in creating long term memory. You pretty much recall both the meaning and reading at the same time anyway, so why not just do them both simultaneously (or 1x1). That’s what Anki has been doing always.
I don’t really understand your logic here, care to elaborate? Why would doing cards in 1x1 be less work for the brain?
Biggest downside is getting a card you answered wrong back immediately (lapsing), which doesn’t give you another chance at remembering it after it has left your short-term memory. That’s why usually with Anki it’s best to set the time after lapse to about 20 min. But I guess you see the card with normal mode within a few reviews anyway, that’s more of a problem with WK itself.
Which can be massively outweighed by the extra time it takes…
Overall this is all pretty anecdotal, but I can tell you I did regular for tens of thousands of reviews and then switched to 1x1 for tens of thousands of reviews and had no drop in retention and notably quicker review sessions.
I quit 1x1 after about a month of trying it for this reason, but I loved the concept.
Now that WK has the extra review of missed items, this entire problem might be solved (as long as I can force myself to immediately jump into these after a review session to reinforce the “cheating”).
Thanks for asking, I didn’t mean anything too fancy here. Short term memory lasts about 15-30 seconds, so if you’re doing them back to back the other part of the meaning/reading pair will always be in your short term memory, and your brain has to ‘retrieve’ the meaning /reading only once per session.
If you do them interlaced then some of your reviews will be when the item has temporarily moved out of your short term memory so you have to ask your brain twice to ‘retrieve’ the item. This is (slightly) more work than just typing something in when it’s in your short term memory.
This makes intuitive sense to me too- doing them 1:1 feels easier (and is certainly quicker) partly because it’s less effort for your brain. Hope that clarifies?
The weakest part of my theory I think is whether putting in this extra effort actually has any retention benefit. I got curious and started reading papers and found that more effort might not, in and of itself, be all that helpful. More retrievals, however, should help. I also found something specifically on srs intervals - this study and other similar work looked at intervals of repetition for students learning lists of Swahili vocabulary.
The study was testing different spacing intervals for vocabulary, (eg if you’re going to review a word three times in a stack of 14 other words, is it better to see it at regular intervals or increasing intervals in the stack?). The interesting thing was they included an option that they called the ‘no spacing’ option where students had to correctly pair the English:Swahili meaning three times in a row, back to back, which is basically like the 1:1 option as far as I can tell. This did a little bit worse than leaving a few reviews in between retrievals. I’ve snipped the graph below, which shows a small effect.
To all those saying it didn’t make a difference to you, I’m pleased to hear that. I respect the volume of your anecdotal evidence (you’ve done waaay more reviews than I have!) but I’m not convinced that a small difference in accuracy would be easy to pick up - if I missed 2-5 more words per hundred reviews I’d probably think I was just having a rough day or something.
Tl;dr (sorry for long post!)
I’m still convinced that it’s plausible that memorising things 1:1 could be marginally worse for retention.
But the effect, if it’s real, is very small, and might not be worth worrying about. Doing an Srs system in the first place is already a highly efficient way to memorise things, so if you wanna go 1:1 then that’s great, I’ll defend you to the hilt.
Brains are awesome.
Source for graph and Further reading
I think you’re right on the whole, BUT if I get a kanji wrong, quite often I could type it correctly immediately afterwards, and if it doesn’t come back up soon enough, I still don’t recall it, and some have come back 2 or 3 more times before I can correctly enter the reading (it’s usually the reading).
Those are the ones that become leeches…
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.