If You Cannot Read It, You Cannot Hear It: What LLMs Reveal About the Surprising Link Between Listening and Speed Reading

When you study English, you often hear this advice:

“If you want to improve your listening, listen to as much English as possible.”

But after studying English for years, I have come to feel that reading speed has an even greater effect on listening ability.

When you think about it, the reason is obvious.

If a sentence takes you five seconds to understand in writing but passes by in two seconds when spoken, there is no way you can process it in real time.

The problem is not your ears.

Your brain simply cannot process the English quickly enough.

If You Cannot Understand a Sentence in Writing, You Will Not Understand It in Speech

Imagine looking at an English sentence and thinking:

“What does this ‘which’ refer to?”

“What does ‘get’ mean here?”

Audio does not wait for you.

While you are still thinking about “which,” the speaker has already moved on to the next sentence. Naturally, you fall behind.

Before listening, therefore, you need the ability to grasp the meaning of an English sentence from left to right as soon as you see it.

There is no time to translate the sentence into Japanese first. You have to process English in English word order.

This is true of both reading and listening. The difference is that reading allows you to stop and proceed at your own pace, while listening forces you to follow the speaker’s pace.

In that sense, listening is a form of reading with forced scrolling.

That is why endlessly repeating listening exercises may be an inefficient approach when your listening skills have stopped improving.

Listening Is Also Prediction

Another important element is prediction.

The way modern generative AI—especially large language models—works provides a useful analogy.

An LLM such as ChatGPT generates text by using everything that has already been provided as context and estimating which word is most likely to come next.

For example, consider this Japanese sentence:

“I am a cat. As yet I have …”

This is the famous opening of Natsume Soseki’s novel I Am a Cat. The actual sentence continues, “As yet I have no name.”

An LLM that encountered this famous passage in its training data would assign a very high probability to “no name” following the first part. It is not selecting those words from nothing.

The context that came before dramatically changes the probability of what comes next.

Humans do something similar when listening to English.

When you hear “If I were you …,” you can expect something like “I would …” to follow.

When you hear “On the other hand …,” you know that a contrasting point or a different perspective is coming.

When you hear “I’m looking forward to …,” you may not know the exact next words, but the possible grammatical structure has already narrowed considerably.

The more English someone has encountered, the more patterns they have accumulated:

“These words often follow this expression.”

“Given the direction of this discussion, the speaker will probably go here next.”

They no longer need to analyze every word from scratch.

They process the English they are hearing while simultaneously predicting what is likely to come next.

This predictive ability is one reason they can keep up with fast native speech.

I think of it as building a small English-language LLM inside your own head.

How Do You Train That English-Language LLM?

This is where extensive speed reading becomes important.

If you listen to the same exercise dozens of times, you will eventually understand that particular recording. But you may simply have become familiar with that one passage.

To train the English-language LLM in your head, you need to feed it a much larger variety of English.

Read the news.

Read novels.

Read blogs.

Read articles in English.

The more you read, the more word combinations and sentence patterns you accumulate. The faster you read, the faster you become at processing them.

As a result, when you hear “If it were not for …,” you no longer have to think:

“First comes ‘if,’ then ‘were’—so this must be the subjunctive …”

You process the expression as a single meaningful unit. You may even be able to anticipate what kind of information will follow it.

That improves the speed at which you can process spoken English.

Speed reading, in other words, is not merely practice for reading English more quickly.

By processing a large volume of English at speed, you are also training an internal predictive model of what is likely to come next.

You Still Need to Train Your Ears

The sound itself remains a separate issue.

English contains sounds that are difficult for Japanese speakers to distinguish. Words run together. Sounds become weaker or disappear. “Did you” is not pronounced as two perfectly separate words in natural conversation.

You need listening practice to learn these features, and you should learn the correct pronunciation when you learn a new word. Otherwise, you may have to relearn its sound later.

But replaying the same passage dozens of times until you can finally decipher it is not the same as improving your general listening ability. If you cannot process the sentence when it is written in front of you, repeating the audio does not fix the real bottleneck.

First, become able to read it quickly. Then use the audio to connect the English you already understand with the way it actually sounds.

Is It Really Your Ears That Are Causing the Problem?

When you cannot understand spoken English, your first reaction may be:

“I need to listen to more English.”

But try looking at the transcript first.

If the sentence still takes time to understand without audio, the problem is not your ears. Your processing is already too slow at the reading stage.

If you understand the transcript immediately but cannot recognize the same sentence in speech, then you have identified a genuine sound-recognition problem.

This kind of diagnosis resembles troubleshooting an IT system.

An error may appear on the screen, but fixing the screen itself will not help if the bottleneck is somewhere else. You need to identify and address the actual source of the problem.

“It is a listening problem, so I should train my ears” sounds perfectly reasonable.

But if the bottleneck is the speed at which you process English itself, you are training the wrong component.

After years of studying English, I keep returning to the same conclusion:

If you want to hear English, first learn to read it quickly.

Taking off your headphones and reading may seem like a step in the opposite direction. In reality, it may be the shortcut you were looking for.