Learning Morse Code - Pointless Hammering

The Koch method is often pushed as the best (or even the only) way to learn More Code. It certainly works for some, but not, as I found out the hard way, for everybody.

Reflections on Koch

Personally I think the Koch method has one very serious flaw; it progresses by adding one character at a time. That can be mind-numbingly tediously discouragingly slow. A large part of reading Morse is in distinguishing one character from another, but you can't get that distinguishing right until you're dealing with the full set.

Maybe what Koch needs is a progression that doesn't so hopelessly over-emphasise the early characters. Perhaps learn two characters, then another two, and then put those together as a set of four. Then learn another set of four, and put those two sets together as a set of eight. Then learn another eight the same way and put them together as a set of sixteen. Or go up in threes, sixes, twelves, or maybe even fives, tens, twenties. Anything to get all forty (or so) characters as quickly as possible without swamping the process with over-emphasis on the early characters.

It really doesn't help that the various Koch training programs all seem to have their own idiosyncratic sequences, so swapping between them is more complicated than it should be.

There's also the matter of checking what you think you've heard against what was sent. In order to check it you need to record what you think you've heard. You'll either have to write it down or type it in, and it's going to be random characters, not words. The speed at which you can make that record will put a top speed limit on the speed at which you can run the exercises. Work out what that limit is, because trying to run the exercises any faster than that will be a waste of time. For me that limit is about 18 words per minute.

It took me far too long to realise the Koch course wasn't working for me. With hindsight I can see that if I'd reviewed my progress after a month or three I'd have realised it wasn't working. I certainly should have given Koch no more than six months.

Eventually I abandoned the Koch method altogether, after wasting well over two years trying to learn with it, and made myself a CD about an hour long with pairs of Morse characters at about 15wpm in predictable sequences, which I listened to in random bits again and again. That gave all the characters equal exposure, and meant I had to sort each one out against all the rest in one hit. That, however, is most definitely not the Koch way.

My time with Koch went something like this:

The following was written in February 2012, and represents a check-point on my Morse journey, before I realised that the Koch method was, for me, a dead end.

Still Hammering Away

I'm still trying to learn Morse code, but it's taking me a very long time, which makes me think I'm not going about it the right way. Morse is a skill (or a set of skills), not simple knowledge, so it can't be learned from a book; it has to be learned by use, by repetition, by practice. In some ways learning Morse code is a bit like learning to play a musical instrument. I've not done particularly well on that score, either. In other ways it's a bit like learning a new language, or learning to read and write.

That brings me to the first bit of contradictory advice. One school says "learn to read Morse code first, and then learn to send it", while the other says "learn both reading and sending at the same time". This one is fairly easy to resolve. If you're in a classroom with other students then the second approach is possible, but if your only option for conversation in Morse is on-air then you have to be able to read a fair bit before you start sending. As Morse code classes seem to be a thing of the past now, the only practical option is to learn to read the code first. Sure, I can practice sending Morse as well, but not on-air (and therefore not in conversation) until I stand a fair chance of understanding what I'm hearing.

I've been concentrating on learning to read Morse code. The most widely promoted method for doing so is one developed in the 1930s by a German psychologist called Ludwig Koch. His approach is intended to build reflexes by teaching, at first, only two characters. When those are learned to a sufficient degree of accuracy, another character is added, and the process repeated. With computers, this process is fairly easily automated, and most of the programs and websites providing Morse training use Koch's method, so it's clearly the favourite. Thus far, there's not much contradictory advice. Different programs use slightly different character orders, so it's worth picking one and sticking with it. I've been using one of the website, LCWO.net because it also does a passable job of recording progress, and it's usable on good variety of different machines.

The contradictory advice starts rolling in when you ask exactly how you should use Koch's method. Take, first, the question of speed. Koch's method involves listening to fast code right from the start. The most commonly recommended speed is 20 wpm (words per minute). If you're a good typist you'll probably be able to record what you've heard just fine, even though it'll be random characters rather than words. If you're writing your copy on paper then you may be pressed to keep up. I know I can't type or write random character copy at 20 wpm. I can just about manage 15 wpm with pen and paper. Now, you can't check your copy unless you have a record of it, so you're effectively limited to the maximum speed at which you can take copy. Say that's 15 wpm. You have two choices. Either you slow down the rate at which Morse code is sent to you to 15 wpm, or you pull a trick invented by Donald Farnsworth, and send the individual characters at 20 wpm, but spread them out until the overall speed is 15 wpm. This second approach is favoured by many, but it has one big problem if you're trying to learn to read the code by reflex. It leaves big gaps between the letters. I find that these gaps are long enough for me to consciously think, say, "Was that an N or an A?" and once you start down that route you're doomed. However, the alternative, slowing everything down to 15 wpm, gives more time for analysing dots and dashes, and once you start thinking "was that two dots or three?" you're even more doomed. There's a balance that needs to be struck; a combination of character and word speeds that gets the code sent fast enough to avoid both dot and dash analysis and thinking time between characters, and yet slow enough for verifiable copy to be taken. For me, the dot and dash analysis can cut in with letters sent even faster than 20 wpm, so my only option is to try not to do it. I can, however, avoid the thinking time between characters by not allowing any extra time between characters, so my usual approach is to choose a character speed just slower than the speed at which I expect to be able to write copy, and not have a different slower word speed. My exact choice has varied from lesson to lesson, depending a bit on the likely mix of characters in that lesson. If short characters like E and T are more numerous than long ones then it's better to use a slightly slower speed. However, I've tried to avoid using speeds lower than 12 wpm.

Next, there's the question of copy verification, and when to move on to the next character. Koch's original concept involves moving on as soon as 90% accuracy is achieved. That's a bit more vague than it first appears. Is that measured over fifty characters, a hundred, five hundred, or more? The standard "word" is considered to be five characters long. At 20 wpm a minute of Morse code would be a hundred characters, so maybe that's a sensible minimum? It certainly works well enough for early lessons when there are only a few different characters, but by the later lessons when there are thirty or more different characters in the mix there might easily only be two or three occurrences of the new character in a random selection of a hundred characters, so a longer exercise must be used, and a higher accuracy threshold applied. Here more contradictory advice creeps in. Some folk say you can't be sure you know the new character until you get 100% perfect copy. Some folk say a minute of copy is enough for each attempt, while others say you should copy for up to five minutes. However, if the copy time is longer then the verification stage will be more tedious and more error-prone. For early lessons I've settled on a minute of copy time per attempt, and I'll move on to the next character as soon as I pass that 90% threshold, but I'll disregard exercises where the new character has been seriously under-represented.

How long should it take to learn each new character? Koch is said to have been able to teach pupils to read Morse code in under 14 hours. Even if he only taught them the alphabet that's little more than half an hour per letter. If he taught the alphabet, the numbers, and a minimal set of punctuation codes then it's nearer 20 minutes per character. I think they must have been exceptionally good students. The website I've been using, LCWO.net, keeps a record of attempt scores. I've put together an animated gif of its graphs of my progress through lessons 2 to 11. The horizontal red line's at 90% accuracy. In the first few lessons I was trying to get myself consistently over 90%, but I gave up on that after about lesson 3 because it was patently obvious I'd never get to the next lesson if I tried to get to the point at which I was always over 90% on a lesson. As for aiming for 100%... well, there'll be flying pigs by then. The data that went into those graphs represents at least 20 hours of actual listening to code. That doesn't include any of the attempts I didn't try verifying, or the other exercises I've done with just two or three characters in order to try to burn in the difference between them, which probably amount to at least another 5 hours. That's just the "listening to code" bit, not counting the time taken to verify the characters read, check over the errors, and so on, which will have taken at least as long again (and probably longer). All that to learn just twelve characters (and, if my progress with the next lesson is any indication, not learn them very well, either). What I've not been able to determine is how my experience compares with the average.

If, as I suspect, I'm making far slower progress than average, then I'm bound to ask why. Are there things I'm doing wrong? If so, what might they be?

There are things I've heard other folks mention that might hold clues. Some folks talk about varying the pitch of the Morse. I've found that Morse becomes increasingly unpleasant to listen to as its pitch rises. I can just about tolerate a pitch up to 600 Hz, but anything over that becomes painful, and 500 Hz is much less unpleasant. However, at lower frequencies it becomes harder to read. I havn't really found a pitch that is comfortable to listen to and relatively easy to read, but, at least with the computer training programs I've tried, I've always had to pull the pitch down from the default 600 Hz to somewhere nearer 500 Hz. I've also found that I don't tolerate Morse above a certain volume at all well. My computer's volume control must be below 40% when I start LCWO's player, or the sound of the Morse goes straight through my head. Some folks say that Morse is relaxing, and they have no particular trouble listening to (and reading) Morse for several minutes at a stretch. I find it not in the least relaxing, but rather the opposite. Reading it demands intense concentration from me. If I stop concentrating then I miss characters (or even whole words), and after a minute or so of trying to read Morse I simply have to take a break. The inevitable side effect of this is that, wheras some folk have no trouble at all practicing Morse for half an hour a day, I'll be feeling brain-fried after a few minutes. On a good day I might manage as much as ten minutes practice time.

What might I do differently? Should I just give up on the exercise?

Addendum: The answer turned out to be "Dump Koch in the bin, and try something else!"

© M0LEP (Originally written in February 2012. Last updated December 2015.)


This page is stored on Zen Internet's broadband customer webpages server.