Massive Input vs SRS: the Inverse Ratio Effect

Inverse ratio “rule”: the greater your input the “looser” your SRS can be

When I say “massive input vs SRS”, I am not trying to imply an opposition between the two because I believe that they are (for most kinds of memory) excellent partners. However:

SRS is the thief of immersion time.

This is a saying known to the ancients. Well, maybe not in quite that form, but it doesn’t take a sage to know that time spent on A can’t be spent on B.

And with a lot of vocabulary etc. to learn, SRS can become time-consuming. Some people advocate dropping SRS altogether in favor of pure massive input. My view is that this is actually the best approach if it works for you, but my experience is that it doesn’t work for most people.

That may be because input isn’t massive enough. But sufficiently massive input may be impractical anyway. And I suspect it isn’t just that.

The problem is, of course, that SRS is “rigged” to feed you vocabulary in a way suited to helping you remember it, and wild input isn’t. So one can tend to forget words between natural exposures.

But natural exposures are the real ground for learning. We encounter words as real meaning units in real emotional contexts, with their contextual nuances and knowing the kind of person that uses a particular word in a particular situation. We aren’t “memorizing” all this, but we are absorbing it and this is what eventually amounts to knowing a word rather than knowing its dictionary definition.

SRS is just a pinning device for holding words in place between real exposures.

This is the approach I advocate.

And we can use it to prune down our SRS.

How?

The thing to remember, which gets very obscured in other methods, is that abstract learning through SRS or anything else is (I’m stressing this point)…

a means to an end, not an end in itself.

This has important implications for many things. Including how we approach SRS.

If learning words through SRS were an end in itself, we would have to be much stricter about it. This means failing more words, which means bringing them back into play more often and increasing one’s daily SRS time considerably.

Before I realized this, like many androids, I took the view

90% right is wrong.

And if SRS were one’s means of learning rather than a handmaiden to real learning, that would be correct.

But, provided one is doing at least reasonably massive input, then in many cases

50% right is right.

Why? Well let’s think about what end SRS is a means to. And this may vary, but let’s take what I think is the most common case.

SRS is a means to pinning words so that they will be recognized when we read them.

Let’s take some scenarios:

We see the word on the front of a card and…

know how it is pronounced and know that it can have one of two related meanings – but can’t be sure which.

Pass (usually).

Why? The question to ask is “when I encounter it will the context make it clear which of the two it is?” If so, SRS has done its job sufficiently for its role, which is supporting real exposure to the extent of understandability. We are not trying to learn the word from SRS – that is the job of massive exposure.

Know the meaning but aren’t sure which of two pronunciations it is.

Pass (usually). Possibly press the “hard button”. Fail if one feels a need to get the pronunciation fixed at this stage.

Why? If you are mostly using material with furigana or anime subtitles you will be reminded of the pronunciation on future encounters. Even if not, I am personally more inclined to look them up on the fly* (but that’s me).

Are confused between two words/meanings

Often pass.

Let’s take an example. You don’t recall which of ポッチャリ and ポチャリ means “plump” and which means “splash”.

First question: if you read or hear one, what is the likelihood of not knowing which meaning is intended in context? As an android I can tell you. Precisely 0.037%. Pass.

Second question: How much do you want to be able to use one or both in conversation? If answer “not very much” then pass.

Pretty sure of the general area of meaning. Can’t actually define the word.

Could well be a pass. 

With context the word would probably be understandable.

These principles also apply to putting words into Anki in the first place. There is a tendency to collect words like a squirrel to a greater extent than necessary (certainly for me).

As you get more used to kanji you are often pretty sure of the meaning and pronunciation of a word from kanji plus context. Now, you may want to Anki it just to remind yourself that the word exists for two reasons:

1. Because you want to use it on the fly

and/or

2. Because you want to recognize it when you hear it, with no kanji to help.

This is reasonable and I do it sometimes. The thing to bear in mind is that this is a trade-off. You are trading expanding your SRS time for some immersion time.

You are going to encounter the word again (or if not you don’t need to learn it).

You are also going to get better (and faster) over time at reconstructing the kanji in your mind when listening even for unknown words (from a mixture of context and knowing what onyomi are likely in this case – Japanese people do this all the time – because knowing Japanese does mean kanji-thinking).

You don’t need to worry that everything will drop off if you don’t SRS it. Some of it will. But it will get other chances. Some of it won’t, especially as you get more proficient.

Sometimes in output you will get words wrong (even though you know the components you forget the order – or you misremember exactly which two kanji it was and use a similar concept-kanji for one of them, or use the wrong reading. This happens. It does improve over time. How much time do you want to invest in ironcladding against these little errors in the short term vs. moving on with immersion?

I can’t answer that question for you, but I can suggest that you ask it rather than simply assume everything needs to be SRSed.

The point of this post isn’t to give specific instruction so much as to suggest a way of thinking strategically about the question.

Textbook Japanese tends to inculcate an exam-based view that abstract memorization is an end in itself (or a means to pass exams, which comes to much the same thing).

The ANKI/Heisig focus of the main AJATT-influenced immersion schools ironically can have a very similar effect. Especially since some “second generation” AJATT-related immersionists are prone to put more stress on the “method” than on the real immersion it was originally intended to support.

And if either of these approaches is your preferred one, of course you should ignore what I am saying here.

If not, if you are aiming for what I call direct or organic immersion, then the thing to remember is that SRS is only there as a tacking-stitch to hold things in place between exposures

And shape your strategy accordingly.

SRS is a good servant, but a very bad master.

___

* Personally I find non-furigana’ed text fun for testing pronunciation (a bore if you don’t know most already of course, but if that’s the case furigana’ed text is currently best). Yes, I am reading, not “studying”, but I like words and enjoy seeing if I can guess unknown combinations of known kanji and very often can. I’m happy to do a quick look-up for the game of it – but that’s me. Often I feel sure enough not to.

You also find that an instinct for readings develops. For example I can usually guess in new compounds whether 物 is もの, もつ, or ぶつ. I don’t (yet) have a rule for helping with it, but one gets a “feeling” for what is likely.

This article first appeared on my private Patreon feed

2 thoughts on “Massive Input vs SRS: the Inverse Ratio Effect

Comments are closed.