Koch and Farnsworth are not two methods you pick between. They change different things, so you use both at once.
Koch decides which characters you learn, in what order, and when the next one arrives. Farnsworth decides how much silence sits between the characters. One is a syllabus. The other is a spacing setting. Asking which is better is like asking whether to follow the recipe or set the oven temperature.
Any training program worth using lets you set both at the same time, and the normal beginner setup is Koch's ordering with Farnsworth's spacing. What follows is what each one actually claims, the arithmetic behind the spacing, and which single number to change when you get stuck.
Try it with your ears. MorseLab plays any text back as tone, flash or vibration, free and offline.
The difference in one table
| Koch | Farnsworth | |
|---|---|---|
| Controls | Which characters are in the lesson, and when one more is added | The length of the gaps between characters and between words |
| Never touches | The speed. Characters are sent at your target speed from session one | The characters. Each one is sent at full speed, unaltered |
| The rule | Start with two, add one only at 90% accuracy | Send at full character speed, pad the gaps until you can keep up |
| The setting it maps to | The character set in front of you | Two speeds: character speed and overall speed |
| The problem it solves | Overload. Twenty-six new sounds at once is not learnable | Panic. No time to write down the letter you just heard |
Read the "Never touches" row twice. Neither method slows a character down. That is the thing they agree on, and it is why they sit together so comfortably.
Koch: the syllabus, not the speed
The Koch method comes from Ludwig Koch, a German psychologist working in the 1930s. It is a rule about progression and nothing else:
- Begin with exactly two characters, sent at your full target speed.
- Practise only those two until you copy them at 90% accuracy.
- Then add one more. Never two.
- The speed never drops. If you are struggling, add characters more slowly.
The traditional first pair is K and M:
| Character | Code | Said aloud |
|---|---|---|
| K | -.- | dah-di-dah |
| M | -- | dah-dah |
Operators say the sounds rather than reading them: a dot is a dit, a dash is
a dah. In written form a dot is ., a dash is -, letters are separated by
a space, and words by /.
K and M are a deliberate pair. Both start with a dash, so the opening sound tells you nothing, and both are short, so neither has a long, obvious rhythm to lean on. You are forced to hear the whole shape. Here is a lesson-one stream you can paste into a translator and play:
-.- -- -.- -.- -- -.- -- -- -.- -- -.- --
The usual first eight, in order:
| Order | Character | Code |
|---|---|---|
| 1 | K | -.- |
| 2 | M | -- |
| 3 | R | .-. |
| 4 | S | ... |
| 5 | U | ..- |
| 6 | A | .- |
| 7 | P | .--. |
| 8 | T | - |
Programs differ on the exact list past the first few, and that is fine. The principle is the part that matters: confusable characters go in early, so telling them apart is trained rather than dodged. An order that began with the shortest characters would leave you hearing E, I, S and H (one dot, two, three, four) as a counting exercise, which is the habit you are trying not to build.
Notice what Koch does not specify. It says nothing about how fast the text flows overall, nothing about tone pitch, nothing about how long a session runs. Those are left open. Farnsworth fills in one of them.
Farnsworth: the spacing, not the syllabus
Farnsworth spacing keeps every character at full speed and stretches the silence around it. The letter R at 18 words per minute sounds exactly like the letter R at 18 words per minute, whether the next character arrives in a fifth of a second or in two seconds.
This gives you two speeds instead of one:
- Character speed — how fast the dots and dashes inside a character are sent. This is your target, and it stays put.
- Overall speed — how fast the text as a whole comes out, once the padded gaps are counted. This is the one you lower.
The reason this works is that recognising a character and writing it down are separate jobs. Slowing the character down damages the first job, because the rhythm you learn at 5 words per minute does not exist at 18. Padding the gap only helps the second.
Where the padding goes, in numbers
Speed is measured against the reference word PARIS, which is exactly 50 units long including the gap that follows it. One unit is the length of a dot, a dash is 3, a gap inside a character is 1, between characters 3, between words 7. And dot length in seconds = 1.2 ÷ WPM.
PARIS is .--. .- .-. .. .... Break the 50 units down:
| Part | Units |
|---|---|
P .--. including its internal gaps | 11 |
A .- | 5 |
R .-. | 7 |
I .. | 3 |
S ... | 5 |
| Four gaps between the five letters, 3 each | 12 |
| The word gap that follows | 7 |
| Total | 50 |
So 31 units are character and 19 units are spacing. Farnsworth sends the 31 at your character speed and inflates the 19 until the whole word takes as long as the overall speed demands.
That gives one line of arithmetic:
Spacing time per word, in seconds = 60 ÷ overall WPM − 37.2 ÷ character WPM
The 60 is the 50 units of PARIS multiplied by the 1.2 in the dot formula. The 37.2 is the 31 character units multiplied by the same 1.2. Split the result in the same 3 : 7 ratio the standard uses, so each gap between characters gets 3/19 of it and the word gap gets 7/19.
Worked through at 18 WPM characters and 5 WPM overall:
- One word takes 60 ÷ 5 = 12 seconds.
- The characters take 37.2 ÷ 18 = 2.07 seconds.
- Spacing left over: 12 − 2.07 = 9.93 seconds.
- Each gap between characters: 3/19 of 9.93 = 1.57 s.
- The gap between words: 7/19 of 9.93 = 3.66 s.
Check it: 2.07 + (4 × 1.57) + 3.66 = 12.0 seconds. The dot is still 1.2 ÷ 18 = 67 ms, and a normal gap between characters would be 3 units, or 200 ms. You are getting 1.57 seconds instead: almost eight times the room, with the characters untouched.
The same character speed at a range of overall speeds:
| Overall | Gap between characters | Gap between words | Character gap in dot units |
|---|---|---|---|
| 5 WPM | 1.57 s | 3.66 s | 23.5 |
| 8 WPM | 0.86 s | 2.00 s | 12.9 |
| 10 WPM | 0.62 s | 1.45 s | 9.3 |
| 13 WPM | 0.40 s | 0.94 s | 6.0 |
| 15 WPM | 0.31 s | 0.71 s | 4.6 |
| 18 WPM | 0.20 s | 0.47 s | 3.0 |
The last row is the finish line. When overall speed equals character speed there is no padding left, the gaps are the standard 3 and 7 units, and you are copying real Morse. Full timing tables are in timing and WPM.
Two things fall out of that table. The gaps shrink fast at first and slowly later, so a 3 WPM step at the bottom takes away close to seven times more thinking time than the same step at the top. And the ratio between the two gaps never changes: a word gap is always 7/3 of a character gap, which is why the "end of word" feeling still trains correctly at 5 WPM.
Using both, which is the actual answer
Nothing in the two methods collides, so a session uses both without any compromise:
| Session choice | Set by |
|---|---|
| Which characters are in this lesson | Koch |
| When the ninth character joins the eight | Koch, at 90% accuracy |
| Character speed | Your target, and Koch says never lower it |
| Overall speed | Farnsworth, lowered until you can write |
| When to raise the overall speed | You, as copying gets comfortable |
A starting point that works for most people: characters at 18 to 20 WPM, overall at 5 to 8 WPM, two characters, fifteen minutes a day. Copy onto paper, score it honestly, and only then decide what to change.
Once you have the first eight, copy words rather than random letters, with the same settings. K, M, R, S, U, A, P and T give you these:
.--. .- .-. -.- / -- .- ... - / - .-. .- .--. / ... - .- -- .--.
That is PARK MAST TRAP STAMP. Words carry rhythm and your brain starts predicting, which is what fluency is. More drill formats are in Morse code practice.
I built MorseLab partly because these two numbers are so often welded together in other places. Its Koch lessons add one character at a time at full character speed, the spacing is a separate control, and the drills, games and quizzes score you so the 90% gate is a measurement rather than a feeling. It works offline, so a commute with no signal is a practice session.
When you stall, change exactly one number
Almost every plateau is one of four situations, and each has one correct fix.
| Symptom | Change | Leave alone |
|---|---|---|
| You are counting dots in your head | Raise character speed | Everything else |
| You hear the letter but cannot write it before the next | Lower overall speed | Character speed |
| You keep swapping two specific characters | Drop back a lesson and drill that pair | Both speeds |
| It is comfortable and boring | Raise overall speed a step | Character speed |
The one move that is never right is lowering the character speed. It feels like progress and it is a rebuild: you are learning a rhythm that will not exist at your target speed, and you will have to unlearn it. If 20 WPM characters really are too much, choose 15 WPM as your target and keep it there, rather than sliding around.
Common mistakes
Treating them as a choice. They control different variables. There is no version of this where you have to give one up.
Slowing the characters instead of the gaps. The single most expensive mistake in Morse learning, and exactly the one Farnsworth spacing was designed to remove.
Setting an overall speed you never raise. Farnsworth padding is scaffolding. If you are still at 5 WPM overall after two months, the goal has quietly become comfort. Take a step even if accuracy dips for a few days.
Adding a character before 90%. The gate is not decoration. Adding a ninth character while the seventh is shaky slows the whole thing down.
Judging accuracy by feel. Count errors on paper across a run of fifty characters, or use a scored quiz. "It felt okay" is not 90%.
Studying the chart instead of the sound. The alphabet chart is a lookup table for writing Morse down, not a study method. Learning by eye builds a translation step you will spend months removing.
Expecting it in a day. You can memorise the chart in a day. Copying by ear is a listening skill that builds between sessions, so it takes weeks of daily practice. The honest version of that answer is worth reading before you plan anything.
Questions people ask
Is the Koch method better than the Farnsworth method?
Neither is better, because they are not competing. Koch is a rule for which characters you study and when you add another. Farnsworth is a rule for how much silence sits between characters. Use Koch to build the lesson and Farnsworth to set the pace of it.
What is the Koch method in one sentence?
Start with two characters at your full target speed, and add one more each time you can copy the current set with 90% accuracy.
What is Farnsworth spacing?
Characters sent at full speed with stretched gaps between them, so the rhythm inside each character is correct while you still get time to write. It is described by two numbers: character speed and overall speed.
What character speed and overall speed should I start with?
Characters at 18 to 20 WPM, overall at 5 to 8 WPM is a common and sensible start. Set the character speed to whatever you want to end up copying at, then lower the overall speed until you can keep up on paper.
Why not just slow the whole thing down?
Because a character sent slowly is a different sound. At 5 WPM the dot is 240 ms and you have time to count elements, so you learn to count. At 18 WPM the dot is 67 ms and counting is impossible, so you have to hear the shape. Padding the gaps lets you practise the fast shape at a slow pace.
How do I know when I have hit 90%?
Copy fifty characters onto paper without stopping, then check them against the source and count the errors. Five or fewer means add a character. Six or more means run the same set again in the next session.
Do I have to use the standard Koch character order?
No. Orders differ between programs, and the exact sequence matters much less than the principle: two characters to begin with, similar-sounding pairs early, one addition at a time.
Does any of this apply to sending rather than copying?
The character speed rule does. Send characters at proper speed with proper internal timing from the start, and give yourself long gaps between them while you think. Sloppy character timing is the hardest sending fault to fix later, and it is the one every receiving operator hears.
Can I learn by reading Morse rather than hearing it?
You can learn the chart that way, and it is useful for writing Morse down. You cannot learn to copy that way, because copying is a listening skill. Start with sound from session one, as set out in reading Morse by sound.
Get it: MorseLab — free on iPhone and Android. The link opens your own app store.
Keep reading
- How to Learn Morse Code — the full plan these two methods sit inside.
- Timing and WPM — the 1:3:7 ratio and the PARIS standard in detail.
- Reading Morse by Sound — how to stop counting elements and start hearing shapes.
- Morse Code Practice — drill formats that train ears rather than eyes.
- Can You Learn Morse Code in a Day? — what one honest day actually gets you.
