Should you practise speaking or listening? Part 1

Yesterday I was reading about language acquisition again. I was trying to understand both sides of the input/output debate.

Basically there is this massive debate among language learners about whether input alone (listening and reading) is enough to learn a language, or whether you should prioritise speaking.

Some people say you should speak as soon as you can and practise speaking as much and as often as possible.

Others say input is more important and that speaking early on and forcing yourself to speak above your level can be harmful.

Note that I’m comparing the two extremes here! I know there are other people who think both are important or who use other methods entirely.

My View

Anyway, after reading quite a few studies and articles and scratching my head, the view I settled on was that:

Yes, speaking is necessary, but it represents a tiny amount of what is needed to be able to speak a language fluently and accurately. It’s not so much the speaking that’s hard – it’s knowing what to say.

Most of the skill of speaking a language is knowing the meaning of words, how they can combine, how they change and so on; the speaking is a separate, smaller skill, which is used to express your model of the language.

It’s the opposite of riding a bike or swimming. With motor skills such as this, there’s a tiny bit of knowing what to do, which might help point you in the right direction, but it’s fundamentally something you need to experience yourself until you get a feel for it.

For language, knowing the language is the skill. You get a feel for it not by speaking, but by building an internal model via listening which accurately represents how native speakers speak (which you then express via the separate skill of speaking).

Models of Language Ability

When we know a language fluently, we have a model of that language inside our heads. We access this knowledge and use our productive skills to form sentences and then pronounce them.

model of language --> assemble sentence --> speak perfect sentence

For people who think input is all that’s needed and neglect speaking, it looks like this:

model of language --> assemble sentence --> magically do a thing you've never done before

For people who think speaking is all that’s needed, it looks like this:

blah blah not important --> magic --> speak perfect sentence

The input guys are wrong, because speaking is a complex physical and mental task of assembling sentences and then moving your tongue and mouth in a very specific way. If you don’t practise it, you won’t be able to do it.

The speaking guys are wrong because they are focusing too much on the outward manifestation of language ability and assume speaking itself is what makes you learn.

Getting Feedback

Unlike riding a bike, where you get immediate feedback if something is wrong (falling off), there is a lot of room for error in speaking.

You can make a lot of mistakes and still be understood.

Excuse me you, need train station now!! You tell me way please. Need way station quick train late. Thanks!

If this were a bike, you’d be in the ditch.

But when speaking, there is no ditch to fall into. The listener would understand you, and you’d still get the help you need.

To “fall in the ditch” when speaking, you’d have to say something totally incomprehensible:

Trin statiop wbowobw yayayaaaa help?

Even this absurdity isn’t totally impossible to understand.

By making mistakes frequently, you will start to make them more often. This feeds back into your model of the language and makes a big mess.

blah blah not important --> magic --> speak imperfect sentence
   <---            <---            <---            <---

The only good thing about speaking to someone using imperfect language is that you will make them produce accurate sentences which can feed back into your model and improve it (assuming they are a native or competent speaker). But if improving your language model was your goal, you could have just asked them to tell you a story, or watched a video or listened to a podcast instead.

Empty Field

If you put someone in an empty field with nothing but a bike and gave them a few hours, they could probably learn how to ride it.

But if you put someone in an empty field for a few decades and told them to learn Chinese, they obviously couldn’t. Language has to enter the brain first.

For the bike, the necessary information comes from the bike itself. For languages, the information comes from other speakers speaking (and nowadays also from videos, podcasts and books).

No Feedback

For most skills, you get immediate, quality feedback. You can get better at them by doing them.

If you draw a horse for the first time, you can immediately compare the drawing to your very precise internal model of what a horse looks like.

If you ride a bike and make a mistake, you lose your balance or fall off.

If you play the piano and press the wrong key, it sounds shit.

But if you say the sentence about the station, the listener will understand what you need and will help direct you to the nearest station. They’ll notice you’re not a native speaker, but that’s about it. Unless they’re very rude, they’re not going to start dissing your terrible language model – speaking gives you no feedback.

Don’t reinvent the wheel

By forcing yourself to speak above your level, you are effectively inventing your own language. If you’re lucky, it will probably be close enough to the native version to be understood, but it won’t be perfect.

Imagine you want to ask someone to take your photo.

Your model doesn’t yet contain the information “take a photo”, so you make an educated guess and choose another verb, or copy it from the model of your native language:

Can you ... make ... a photo of me please?

Unless you get immediate feedback, your brain will assume what you said was correct. From now on, “make a photo” will sound a little bit more correct than before. Since your tongue has said this before, it will be more available next time. Eventually it will sound totally normal.

The model is already available

Instead of inventing your own language, why not make your life easy by imitating the already existing one. If you’d have heard other people say “take a photo” 100 times, you would never think of saying “make a photo”. You will automatically know that it’s wrong, just like things sound wrong in a language you speak well.

When riding a bike, it’s hard to put the model into words because it’s a very precise motor skill which needs to be experienced first hand.

But for language, you can put the model into words because that’s exactly what the model is. Every correct sentence contains a little piece of the model. Hearing a correct sentence is the equivalent of riding a bike for 5 seconds.

The language is precisely the thing you are learning, which is why being exposed to it (input) is so effective. Being exposed to the language is just like being exposed to a situation in which you are riding a bike.

The confusion comes because languages consist of words and so does information about how to ride a bike, but they are not comparable at all.

A note about reading

So far I’ve been treating all types of input as equal, but there’s something important to mention about reading here.

Humans evolved language around 50,000–150,000 years ago as a skill based on listening and speaking.

You hear sounds other people make, and your brain works out how the sounds relate to reality and builds a mental model of correspondence between the two. You can then use this model to create your own sentences, which you express via the skill of speaking.

About 3,000 years ago, which is relatively recent by comparison, some very clever people worked out that you could make little marks on stone to store or transport words and sounds.

This was absolutely genius! Arguably the best invention of all time. But for us language learners it introduces some difficulties which need to be addressed.

Reading is a great way to improve your mental model of a language. It packs a lot of information about word meaning and grammar into a small space, and unlike spoken language, you can take as long as you like to understand it.

Unfortunately, there is something important missing here – sounds. Paper, wax and stone can’t actually store sounds. We can only use them to make a representation of the sounds.

This means that when you read, your brain conjures up the sound from your mental model instead, and if your mental model of the sounds of the language is incomplete, we’re back to the same feedback loop problem we saw before.

                                  read words
mental model of sounds --> hear sounds in your mind
      <---           <---           <---           

This means that, paradoxically, reading can actually harm your pronunciation.

The solution is quite simple: make sure you get input that will improve your model of the sounds of the language. In other words, listen to native speakers speaking, either in real life or through videos and podcasts.

If you don’t yet have a good understanding of the sounds of the language and how the orthography (writing) maps onto the pronunciation of words, you should avoid reading for now and focus on listening, or listening while reading at the same time.


Here’s a quick summary:

  • Build a mental model of the language which is as accurate as possible by exposing yourself to correct examples of spoken language. The mental model includes sounds, grammar and word meanings.
  • Practise forming your own correct sentences by accessing your mental model. Express these correct sentences by using your speaking skill.
  • Train this skill over time. Slowly you will be able to produce the correct sounds, assemble them into the right words and construct grammatical sentences of ever-increasing complexity.
  • Don’t speak above your level. If you don’t know how to say something, don’t invent your own language. Remember you are imitating an already established perfect model of the language. Unlike riding a bike, playing around with your mouth will not cause your mental model to improve.

Part 2 coming soon

This has been a fun post to write, and I hope it’s cleared some misconceptions up for you. In the second part, I’m going to talk about how you can apply this to your language learning.