Crazy idea: Using the Koch-Farnsworth method for ear training from day one

[SpanishInput]

Just wanted to throw a few ideas here to see what everybody thinks.

Despite all the "speak from day one", I believe it's more important to listen from day one.

A few years ago I came across the Koch method for learning to listen to Morse code. In traditional methodology, you listen to really slow code and slowly build up speed. But Koch realized that it's really hard to re-train yourself to listen to natural speed code once you've trained yourself to listen to slow code. Also, some people come up with mnemonics to remember each Morse letter, and these mnemonics get in the way. So in the Koch method, which is favored nowadays, you listen to natural speed code (15-20 WPM) from day one. The trick of the Koch method is that you only listen to two letters in the first day: K and M. Once you can transcribe these two letters at natural speed with 90% accuracy, you move on to the next lesson and add an extra letter. Koch's approach of listening to full speed code from the beginning is favored by people today because you get used to the rhythm of each letter instead of trying to translate the sounds to dots and lines in your head, or even worse, trying to translate the sounds to dots and lines and then translating that to mnemonics. With Koch's method you go straight from sound to letters, because it just doesn't give you time for anything else. I've seen videos of people saying how training with slow code at the beginning has actually harmed them and now they can't stop mentally translating when they listen to code.

I'd say the equivalent in the language learning world could be to listen to short sentences at full speed from day one, but only listen to sentences with just a very tiny dose of new words. Your teacher/tutor could do this for you during a dictation exercise. The same words could be recycled again and again in the same dictation session.

There's an adaptation of Koch's method called Koch-Farnsworth. In this adaptation, again you listen to Morse letters at full speed, but there's a space between letters to allow you to digest and mentally process the information. You begin with a relatively large gap between letters, and you slowly close this gap as you progress.

The equivalent in the language learning world could be to give the learner a few seconds between sentences, so he could process what has been said in each sentence. And in higher levels, slowly close this gap between sentences.

In today's training software using the Koch-Farnsworth method for Morse, the learner actually needs to type what he hears.

The language-learning equivalent could be to actually require the learner to type everything. Some years ago there was a thread here about the transcription technique, and some forum members said it's highly effective.

What do you think about these ideas?

[rdearman]

This is very interesting to me. I might try something like it ... Have to think about how to implement it. So your suggesting that you start with some short sentence and build up?

Run!
I run
Spot runs
See spot runs
See spot run and jump
See spot and Sam run and jump
Spot and Sam run and jump over the fence
Spot and Sam run in the field and jump over the fence
Running and jumping over the fence both spot and Sam fell down

Like this?

[galaxyrocker]

I've sort of developed a similar method to this, working with transcribing audio. It's not automated, but I used it to great success in training my Irish listening. It basically involved listening and transcribing as I was listening, doing it in short intervals to make sure I had it correct. Once it was correct, and once I could hear that it was correct, I'd move on to the next interval (say, 10 seconds, or sentence based depending on length), etc. You have to have something to check yourself with, but I found it extremely helpful. I need to do something similar with French, as I went from listening to documentaries (not many issues when I focus) to just a random youtube video of 'top 15 non-fiction books' and the difference in comprehension was staggering.

[jeff_lindqvist]

Back in February 2017:

There's actually a trick to learning how to process (limited!) full speed input in several dozen hours instead of a couple thousand. I first heard about this in the context of the Koch method for teaching Morse code

[cito]

Back in February 2017:

There's actually a trick to learning how to process (limited!) full speed input in several dozen hours instead of a couple thousand. I first heard about this in the context of the Koch method for teaching Morse code

EMK has always been ahead of the curve!!

[Le Baron]

It could be argued that this is pretty much what happened when we were tiny tots and someone was holding us and repeating single words right into our faces. Then two words...etc. Building up.

I'm sceptical about this being achieved as an adult at "full speed input in several dozen hours". A trick indeed. By adulthood you're listening out for lots of other things. On the other hand if anyone has some results of trying this with success I'd be interested. I too would like to reduce the workload.

[BeaP]

It might be a good idea. The question is always: Is there a better idea we've already been using?

One can use books like Ci vuole orecchio (Alma), Hören und Sprechen (Hueber) or Comprehension Orale (CLE), even in tandem with a good book on phonetics.

A unit usually contains a dialogue, then an exercise that asks for the gist, then other exercises that ask for details. If you check your answer after each exercise, as you go deeper you're able to concentrate on minor things. I've found this type of training very effective. Even when I hardly understand anything in the first round, things become really clear by the end, and I don't even need to check the transcription. These exercises seem really natural for me (this is how we listen to information IRL), and listening to the same dialogue multiple times also improves speaking skills.

So, if we make our own materials with this new idea, will the material be more practical, natural, effective? Will we save time?

I just ran into this video yesterday. I'm linking it to show that the book format is not essential here. This tutor does something similar with a scene from a Netflix show. Although she changes the speed (I don't think it's a good idea) and presents new things instead of prompting them, the approach is the same. Several rounds, increasing number of details.

[jeffers]

I think there's a lot of merit to using this method, but with one caveat: there is no such thing as "full speed" or "native speed".

With Morse code there is a standard speed for normal communication. But with languages you can't define "native speed" with a fixed value. Sometimes native people speak faster, and sometimes people speak more slowly, and much of this depends on the style of the speaker and the context in which they are speaking. In addition, some natives speak quickly but enunciate clearly, while others speak slowly but slur their speech. So there is a lot more variety in spoken language, and the learner would do well to get used to a wide variety. The basis of the Koch method for training for Morse code is that "Koch realized that it's really hard to re-train yourself to listen to natural speed code once you've trained yourself to listen to slow code". However, with language listening you ultimately need to be prepared to hear at any and all speeds. For this reason, while listening to full on native speeds from the beginning is useful, switching the kinds of things you listen to will also be useful.

There is one additional complication: the amount of content per minute (ideas, vocabulary, etc) is another measure of speed, but it will also vary wildly from context to context. For example, compare a minute of a documentary to a minute of a conversation between friends. The documentary will be spoken more slowly, but the number of ideas expressed and the size of the vocabulary will likely be many times higher.

So I like the general idea being presented in this thread, but knowing how different real speech is from Morse code, any listening programme should also take in the need for variety, and even the abilty to handle switches in style and pace.

Practically, here is what I'm doing (in theory) to address these needs:
1. I listen to a wide variety of types of audio: films, podcasts, documentaries, audiobooks. Some things are more difficult in terms of content, some more difficult in terms of vocabulary, some more difficult in terms of speed, others more difficult in terms of clarity. I strongly believe that variety from the beginning is essential.
2. I agree that typing what you hear is a useful way of learning. I'm using online dictation, typing what I hear, to improve my ear. However, for dictations I am working from easier to harder. The reason for this is mainly that I think it's useful and necessary for the content and duration to build up slowly. So the lower level lessons are slower, but the important factor for me is that they are on a narrower range of topics with a narrower range of vocabulary, and I will eventually work up to more "native style" audio in my dictation practice.

[Dragon27]

The basis of the Koch method for training for Morse code is that "Koch realized that it's really hard to re-train yourself to listen to natural speed code once you've trained yourself to listen to slow code". However, with language listening you ultimately need to be prepared to hear at any and all speeds. For this reason, while listening to full on native speeds from the beginning is useful, switching the kinds of things you listen to will also be useful.

If you're able to comprehend the natural native level speech (which is "fast"), then the slower kind of speech should pose no problem (speed-wise, you still have to learn to understand less articulate speech, which is a separate factor), or would require minimal training (getting used to). But the analogy here is not with the kind of slower speech that the nature documentary narrator employs in order to not overwhelm the listeners with info, the "slow code" speed is like a deliberately drawn-out speaking speed aimed specifically at the beginners in the audio-materials for the language learners. It should be natural, not necessarily fast, but not so slow that the learner will be inclined to cheat (i.e. translate in their head, instead of developing a direct understanding of spoken phrases).

The basis of the Koch method (as I understood it from the article linked in emk's post) is that instead of going through the intermediate phase that the teachers think the learners should necessarily go through (creating a look-up table in your head and using it to decipher the string of Morse code) you can just immediately start building the reflexes "hear the code - get the message". The slow code, as understood by the author of the article, is 5 wpm. A typical learner just memorizes all the characters and when they hear the code they transform it into letters by remembering the correspondence table in their head. This kind of strategy works more or less fine with the slow code, but all this thinking (remembering, etc.) is very awkward and takes too much mental resources, so when the learners reaches the threshold of about 10 wpm the entire thing breaks down and the learner gets stuck on a plateau. Some lucky learners break through the plateau because they transition into a qualitatively different way of comprehending the Morse code message - by reflex, without thinking about what the letters the Morse code they hear stand for. Copying at a normal speed of 15-20 wpm is impossible with all the thinking in the way, it should be automatic, relaxed. So learning to copy code at that speed (starting with a narrower set of characters) forces you to develop the correct reflexes/habits from the start.

[luke]

The trick of the Koch method is that you only listen to two letters in the first day: K and M. Once you can transcribe these two letters at natural speed with 90% accuracy, you move on to the next lesson and add an extra letter.

This reminds me of how I learned to type, except the first letters were juj. There, one finger moving to an adjacent key was done over and over and the keys and fingers varied, but all of that was learnt to the point of "muscle memory" before we really started typing more complicated things, like paragraphs.

One analogy to language learning I might draw here is, get used to hearing one particular speaker at native speed. That's like the juj of typing. As you progress, add more speakers.

With that in mind, @SpanishInput would be a good model.