Trouble remembering meaning AND reading?

Do you ever find yourself having trouble recalling both meaning and reading for some Kanji after a batch of lessons? Sometimes I find myself instantly remembering the mnemonic for the meaning, but the reading one just doesn’t want to come with it.

Why does this happen? One of my classes in university this year has been learning & memory, and since it helped me find a strategy to overcome the meaning + reading struggle, I wanted to share this ☆ also writing this is a great way to procrastinate studying for my finals. anyways

How does remembering and forgetting work?

When forgetting something, the information isn’t simply gone. Unlike a computer, a brain can’t delete information. So when you know you learned a Kanji but can’t recall it, the meaning and reading are still saved somewhere in your brain. The problem is probably connected to the retrieval cue!

This retrieval cue in the case of learning Kanji is the symbol of the Kanji itself that appears during reviews. During the lessons, you’re actively linking the retrieval cue to the contents you’re learning, aka meaning and reading. Mnemonics support building this connection. Remembering something stands and falls with the successful retrieval of information, which is what you’re doing during reviews.

This is just a very rough overwiev to dive into the topic. Pls forgive me for not going into more specific scientific detail

Interference Theory: Why it's hard to remember both meaning & reading

Today, scientists still stand at the beginning of unraveling the mysteries of The Brain™. There’s still uncertainty going on about how the long term memory and forgetting actually works. However, there are theories explaining forgetting that are supported by empirical evidence, and we’re going to look into one of those: the Interference Theory.

As you can probably guess from the name, this theory is about two bits of learned information interfering with each other, making the recall of one of them or even both harder. Interference happens when these two bits of information are connected to the same retrieval cue, which is in our case, can you guess? The Kanji!

During lessons, you’re building two associations for one Kanji:
Kanji ➜ Meaning
Kanji ➜ Reading

When you have this pattern of learning A ➜ B and learning A ➜ C right afterwards, interference happens. There are two types of interference:

Proactive Interference happens when the learned association learned earlier (A ➜ B) is blocking the recall of the association learned afterwards (A ➜ C). In WK terms, you remember the meaning but struggle with the reading which you learn right afterwards during lessons. This is a result of the older representation (Kanji ➜ meaning) being stronger than the newer one (Kanji ➜ reading).

Retroactive Interference is kind of the opposite: it happens when the association learned more recently (A ➜ C) is blocking the one learned earlier (A ➜ B). In our case, this means recalling the reading while struggling with the meaning. This might happen because relearning (Kanji ➜ reading) undermines the older association (Kanji ➜ meaning), because the automatic retrieval of Kanji ➜ reading might be hard to suppress, or simply because there’s not enough capacity for recollection (aka being tired and unfocused).

There are several empirical studies on this topic showing that subjects who learned both A ➜ B and A ➜ C have higher error rates than control subjects who just learned one of these representations. Obviously, we still want to learn both since we need to know meaning and reading of a Kanji. (duh). So, how do we go about this?

A loophole for avoiding interference

One of the things that make WaniKani amazing is that we don’t just learn simple associations between Kanji and their meaning and reading. Instead, WaniKani provides us with Mnemonics to semantically strengthen the links between learned associations. Taking this into consideration, the learning pattern looks like this:

Kanji ➜ Mnemonic A ➜ Meaning
Kanji ➜ Mnemonic B ➜ Reading

The issue here can be that sometimes, mnemonic A and B respectively are both linked to the Kanji (in a way that the story matches the Radicals), but have little to do with each other. This brings us back to the pattern of A ➜ B and A ➜ C, (B & C being two entirely different things,) which causes interference. However, there’s a way to overcome this.

To avoid interference, you can actively change the learning pattern during lessons (without much of an effort) to look something like this:

Kanji ➜ Mnemonic Compound ➜ Meaning
Kanji ➜ Mnemonic Compound ➜ Reading

The difference here is that looking at the Kanji makes you recall one single Mnemonic Compound that acts as a retrieval cue for both meaning and reading. To create a Mnemonic Compound, you can simply link what’s happening in the story for the reading mnemonic to the story for the meaning mnemonic which you learned right before. Or you adjust the meaning mnemonic to match the reading mnemonic. Or, if you dislike both, you think of an entirely new story to create a mnemonic. The point here is telling it to yourself as one coherent story. In detail, the pattern here looks like:

Kanji ➜ Mnemonic Meaning ➜ Mnemonic Reading

That way, you will have the same retrieval cue to activate meaning and reading, which are both part of the same story. In the beginning, you might have to remember the meaning first to get to the reading, but it will act as a bridge to get there. After a few reviews however, the dirt tracks you have for links in the beginning will become paved roads and – you know how it is – the retrieval will be seemingly automatic.

[TL;DR] During lessons, you learn two representations for the same Kanji: meaning and reading. This causes interference, which can make it harder to remember one of them. Instead of learning two separate Mnemonics for meaning and reading, try to connect them into one Mnemonic Compound by putting them together into one coherent story. This way, you can avoid interference and turn two separate retrieval cues needed for sucessfull recollection into one.


Hi Annacelina, it’s super interesting - at least for me :wink: Do you have a link to share to such scientific sources / data ?

I can’t agree more, at least based on my own experience !
The more integrated my stories are, taking into account various dimensions of a kanji, the stronger the connections and the easier the retrieval. But obviously it comes with a price : the time needed to find a good mnemonic that is highly integrated / powerful / using various senses, etc.


This is what I’ve always been suggesting to people when they talk about having trouble with mnemonics or with recall. I’m happy that I got to learn the scientific basis behind my intuition today. Thanks for sharing this information! :slight_smile:


Interesting. I didn’t think about it that way.

I find myself being able to recall the meaning more often than the reading but I thought that was because once we learn the meaning we are done. We don’t have to rely on mnemonics for the actual meaning. Because we dominate English, we don’t have to make an effort to recall the word atmosphere. We don’t have to remember the story for atmos and then the story for phere.

But we do have to remember the story for 空 and 気 and within each story we need to remember the actual word from which we need to extract the reading.

I hope I made sense. OP, if you could share a link based on that study that you mention it would be great!

1 Like

I find I can recall both meaning and reading for kanji that is significant for me. Perhaps that is because I can use it in my everyday thinking. I find I can also decode Kanji for the meaning. Sky spirit is air but recall the reading not yet. I have copious notes to support my shortcomings. I will now try to combine meaning and reading stories as I learn new Kanji. Thank you for your explanation. I thought it was only me!

1 Like

My take on the meaning + reading conundrum has been for several years already to associate phonetics with meaning directly so that when I hear (or am able to read) a word, I instantly know what it means and comprehend the emotion behind the word without having to think about it.

It’s a little weird with the existing WK mnemonics for me, therefore, because they pull the phonetics of an unrelated word (and inherently to me, its meaning) and attach it to the kanji, which now becomes emotionally colored or biased in a specific direction.

The ones that do make more sense are the ones which at least partially try to overcome the interference issue, like connecting the reading せつ to sabres and cutting. One starts with the word “sabre”, derives the initial せ, links that to the act of severing (even aligns phonetically!) or cutting and finishes with the sound of cutting - せつ.

The issues start when too many mnemonics leverage the same phonetic component with the same meaning for different purposes. Mrs ちょうwalks down the 丁 and is a 長, but how many other things can she do?


Hi Andy,

Agree, that is the ideal scenario, and in some cases it works out very well. In my system, I am using a character called Setsuna (I Am Setsuna - Wikipedia), who happens to have a saber. At least in my mind she has one, because that’s how I decided she would be !

I see it the other way round (unless I misunderstood) and actually as a solid help to memorization.
For me, keeping the same character for the same ON reading makes sure I can more easily recall the reading when going through the story in my mind. It also helps me in creating the story more quickly because I don’t have to find a character for it : it is always the same.
It means indeed that I will have, eventually, dozens of stories with the same character, but so far I haven’t seen any ‘interferences’ resulting from this, provided that all the other elements of the stories are different in order to make each story unique. Which they are because the KUN reading is different, the meaning is different, etc.


I agree this might’ve been the intention here - using a limited set of identifiable characters to use as anchors for mnemonic stories. That is effective, but as you said it requires the stories to be unique, which is easy enough, but it’s way harder to avoid the problem of ちょう phonetically being associated with only 1 group of emotions (anger, happiness, etc.) to make its meaning more readily recognizable.

I usually try to expand phonetics to a word or phrase to limit this problem.

I want to expand on what I wrote above, because I think it wasn’t very clear and I accidentally confused the meaning of kanji in compounds, which is a secondary matter.

The way I see a more phonetics-first approach is the following:

  • Creating an artificial character or reusing an existing character wouldn’t work, because:
    • that character already comes with a set of traits which would have to be used in mnemonics or worked around to ignore them.
    • attaching extra traits/features to a character, to me, is a lateral process and requires mnemonic stories to be wildly unique to avoid interference.
  • The core for a kanji reading would be a phonetic component, for instance ちょう + its general feeling/emotion, for instance “chief, long”
  • Kanji with overlapping readings can be stacked and derived from each other phonetically and emotionally (鳥「ちょう」 - 白鳥 is a bird with a “long” neck, 社長「しゃちょう」 - a “chief” of the company, 丁「ちょう」 - streets tend to be “long” )
  • English words with precise or imprecise pronunciation are a no-go, because:
    • imprecise pronunciation is variable and doesn’t map accurately to the Japanese pronunciation.
    • precise pronunciation might be linked to an English word in which case we’re adding into the mix an extra component which biases the mnemonic.
  • A workaround to the above would be using Japanese mispronunciation of English words, because if done consistently it still kind of keeps the mnemonics on the Japanese side of things.
  • It probably makes sense to link multiple words and kanji in a phonetics-heavy network where each component enforces the overall meaning, but is an easily recognizable element of the network.

I would say the above is kind of similar to the children’s board game where different items would share a color - each item is clearly different from the other, but because they’re of the same color, one intuitively considers them to be a group.


I think it’s really a matter of personal preference and of what each of us struggles with the most. For people who have a lot of trouble remembering readings, it may help to have a consistent anchor for them. Personally, however, I find using a system of mnemonics far too restrictive because that means I’ll have to force those anchors to fit into my mnemonic. I mean, it’s a good exercise in creativity, but sometimes there’s no logical link between the anchor and what needs to be remembered, so I find it more a hindrance than a help. I didn’t remember readings using anchors for Chinese anyhow. I prefer to individualise any ideas that come to me for remembering something related to a kanji.


On this I totally agree, there’s only one way : the one that works for you :slight_smile:

Indeed, and that is probably where I come from. I started with RTK long before WK, so I was relatively good on meaning / recognition / writing… but I was a total numb on pronunciation :grimacing: since RTK doesn’t help at all on ON/kun acquisition. At least the first volume, the second book covers that part but I haven’t read it yet, still lying on my shelf. So I built my system with the pronunciation at its core, more than the meaning I guess.

So that is the path that I have chosen : to create a system based on pre-defined rules and stick to it ‘at all cost’, or at least ‘as much as possible’. The difficulty lying in having rules flexible enough to match all possible cases and have the least number of exceptions.

it’s indeed a benefit of the approach, it really forces you to be ‘more creative’ and not jump on the most immediate association. Kind of creativity under constraint.

and this is exactly the other side of the coin ! Having a systematic approach, trying to respect a set of pre-defined rules, I end-up sometimes spending lots of time (and I mean, really a lot of time) to make the mnemonics work. So yes, more hindrance than a help :frowning: Usually I manage to make it work eventually, so looking at the end result, it’s positive, but if you measure the time spent to get there, it’s probably not worth !

You’re right, that’s exactly how I proceed, but I actually turn it to my advantage. Having existing traits of a given character makes it much easier to remember the story. What’s important is to really use as many as possible specificities of the existing character : feelings, emotion, voice, social status, age, etc.

yes indeed, it’s a consequence, if stories are not unique enough, interferences will occur at recall time, and you end-up mixing things. To prevent this, the key for me was indeed to put even more effort in the details of each mnemonics, and play with all the traits, to ensure differentiation. The price to pay being, of course, the initial time investment…

This is brilliant, and indeed it works very well on the examples you gave :grinning: Were you able to replicate
that systematically for all ON readings
? or a certain number of them ?

I’m with you on this one. After experimenting quite a bit, I came to the conclusion that for KUN readings, the sound-alike method basically doesn’t fit well. There are just too many combinations possible for KUN readings and relatively very few good matches (at least in English alone… if you have several languages under your belt and cam mix them, probably you can make it work).
By contrast, for the ON readings, I found that the sound-alike technique works much better, since there is only a limited number of items (around 300).

I’m not sure I understood, sorry… are you referring to “loan words” or something else ? Can you expand on that ?

Hope what I wrote is still understandable, I’m switching between tasks / screens / meetings so not very efficient I guess :grimacing:


Hahaha I can totally see this being reasonable and practical, but I just can’t think of actual characters for that purpose :frowning: .

I think the other on’yomi readings for which I build similar connections are こう (something related to school or public matters), しょう and しょく (usually related to food somehow), because these so far have been the most common.

I could mix other languages into this as well, but for Japanese I prefer to study in English so it’s either the “Japanglish” pronunciation or I don’t even try. Rather, what I do with kun’yomi readings is I repeat them and try to build an emotional attachment to the sounds and a connection between the reading and the meaning via core emotions. For instance, when thinking of 疲れる, I imagine myself working hard and being tired so the つかれ sound is associated with exhaustion and this will be reinforced with 疲れさま and other words. Not to be confused with the つく (making) and つか (using).

Yes, the つく for “attach” might overlap with つく for “making”, but I pronounce them (at least in my head) a little differently.

I was referring to both loan words and Japanese pronunciations of English works like “other” would be pronounced あざ, “ball pen” - ボールペン, etc.


Hey guys sorry for my late reply, it’s kinda like me to just drop the information here then dip for a few days lol.

@YannickFrance & @Kinozato , I’d love to share my material and notes from the class but since they’re entirely in German idk if they would help. (However if you by chance happen to speak German, just tell me and I’ll send them to you!)

I think the original study on interference theory was conducted by postman in the '60s or '70s. I didn’t find a link to the original Postman Paper, but there are several short articles (like this one: or youtube videos out there that explain the concept in more detail than I did.

This sounds super cool, have you heard of synesthesia before? The concept of words being emotionally colored sounds a lot like it.

It’s so interesting to read what the learning process is like for others, thank you all for sharing!


I heard the term a long long time ago, but never looked much into it.

Actually, Japanese is to me the only language other than English for which I used this approach. Never considered it when learning French or German, but then again I never took those two too seriously :sweat_smile:.

1 Like

I also have problems with the need to ‘unremember’ radical meanings that I’ve learned through other study.

1 Like

This has been what I’m doing since the start. I wasn’t aware that there are people who actually try to remember all the mnemonics. oO Like, isn’t that even harder than just remembering meaning/reading on it’s own?

No what I do is, I read both mnemonics, and then create my own one sentence mnemonic from them.
Take for example 形

The mnemonics are:
You’re walking at night when you discover a lantern covered entirely in hair, making it a very strange shape. In fact, this is a form you’d never expect a lantern to take.

You stare at the shape until you finally recognize it: cake (けい)! Well, it’s hairy, glowing cake, but it’s still shaped like a big slice of cake.

So, I just compound it into one sentence: The fuzzy lantern is the shape of cake.
See, it contains both the radicals, the meaning and reading, while being a much smaller sentence.

There are of course kanji where this is harder.But maybe that’s because english is not my first language.

What I have more serious issues with is, when we get to the vocabulary readings, and then they introduce a different reading for the same kanji, but the mnemonic is not at all related to the previous reading’s mnemonic. That conflicts in your head very much.

1 Like

I’ve subconsciously been linking meaning and reading mnemonics together simply because whenever I’m trying to recall one, I work to recall both at the same time.

My struggle now is trying to learn how to reverse engineer what I’ve learned so I can recall the meaning from the reading or the reading from the meaning without needing the visible kanji there to prompt my brain. I know that’s not really the point of wanikani though.