International Morse code sends each character as a short pattern of dots and dashes, and its timing rule is brief: a dash lasts three dots, the gap inside a letter one dot, the gap between letters three and the gap between words seven. That rule is in the ITU's Recommendation ITU-R M.1677-1 of October 2009. We read it in full and searched its text. It has no words-per-minute figure, no word PARIS, no Farnsworth spacing and no SOS. The speed convention sits in an ARRL timing standard of 1990, whose author wrote that ARRL "couldn't find any definitive specification" for Farnsworth timing.
Every dash lasts as long as three dots under the ITU's spacing rule
M.1677-1 says it "confirms the International Morse code characters and the operational provisions applying to their use", and that versions of the code "have been in use since 1844". Section 2 measures everything in dots and never gives a dot a length in seconds. The document's only mention of speed is procedural: "The call shall always be made at hand speed".
The character list holds the 26 letters, an accented e (..-..), the ten figures and 14 written signs, from the full stop to @. On our count that is 51 entries with a printed character, on 50 distinct signals, because the multiplication sign is listed with the letter X's signal. The same list carries six signals with no printed character, among them Error, which is eight dots.
Each dot takes 1.2 seconds divided by the speed in ARRL's standard
The PARIS rule is written down in Jon Bloom's "A Standard for Morse Timing Using the Farnsworth Technique", QEX, April 1990, pages 8–9: "For purposes of specifying code speed, the 'PARIS' 50-unit standard is used." Bloom cites the convention to the 1989 ARRL Handbook and gives the unit as u = 1.2 / c, with u in seconds and c in words per minute. At 20 words per minute a dot lasts 60 milliseconds.
We recomputed the counts from the ITU table. The five letters of PARIS take 31 units of signal and in-letter gaps, the four letter gaps take 12 and the word gap 7, which makes 50. CODEX comes to 60. E is the shortest letter at one unit, and J, Q and Y the longest at 13 each. The figure 0 runs to 19. Typing tests have their own standard word, traced in our guide to the five-keystroke typing word.
Farnsworth gaps occupy the time a slower overall speed adds
Bloom defines Farnsworth timing as "sending the characters at a faster speed than the words", then concedes: "There is no standard for Farnsworth timing." ARRL's choice, for its own training material and the W1AW station, was to send characters at 18 words per minute and stretch only the spacing below that speed. In a PARIS word the 31 units inside letters stay at character speed. The 19 units of gaps become a delay, ta = (60c − 37.2s) / (sc) seconds for overall speed s, shared out three nineteenths to each letter gap and seven nineteenths to the word gap.
Our figures from Bloom's formulas, with 18-word-per-minute characters (normal letter gap 0.200 seconds):
| Overall speed (WPM) | Letter gap (s) | Word gap (s) | One PARIS word (s) |
|---|---|---|---|
| 5 | 1.568 | 3.660 | 12.000 |
| 10 | 0.621 | 1.449 | 6.000 |
| 13 | 0.402 | 0.939 | 4.615 |
Each row closes on exactly 60 seconds divided by the overall speed, our check on the arithmetic.
Our converter held each letter gap for a unit too long
Until 21 October 2026 our text to Morse converter added a gap after every element, three more units for each space and seven for the slash marking a word break, so letters sat four units apart and words 14. A run of PARIS cost 61 units a word, so its "20 WPM" played at 16.4 on our arithmetic. Encoding turned a character the table lacked into a bare "?", and decoding stripped every question mark, so ..--.. came back empty.
The revision made that day schedules one, three and seven units. E E now takes 9 units, and PARIS followed by another word costs 50. Typing 2% now sends the ITU's 2-0/0, café × gives é its own signal and × the letter X's, and a character with no signal stays in square brackets, counted beneath the output.
The Berlin distress call runs to 23 units when read as a single signal
The 1906 Berlin radiotelegraph conference adopted the distress signal. The ITU History Portal holds a scan of that text as printed by the US Government Printing Office in 1912, in English. Article XV of the Service Regulations says the signals "are those of the International Morse Code". Article XVI says ships in distress "shall use the following signal", printed as ...---... and "repeated at brief intervals".
The print shows nine evenly spaced elements, marks no letter breaks and gives the signal no name. Read as one signal at ITU spacing, that is 23 units. Sent as the letters S, O and S, it takes 27. Both counts are our arithmetic; the regulation sets no spacing for the signal. The convention was signed on 3 November 1906 and came into force on 1 July 1908.
Wabun signals last as long as European signals, and 31 share a pattern
Japan's Radio Station Operation Rules, 無線局運用規則, carry the code as a statutory table, Appended Table 1 on e-Gov, in a Japanese part and a European part. Its notes give the ITU's spacing: a dash is three dots, with gaps of one, three and seven.
The Japanese part lists 48 kana in iroha order, plus the voicing mark 濁点 (..) and the half-voicing mark 半濁点 (..--.), 50 entries in all. By our count, 31 of those 50 reuse a pattern the ITU gives to a written character. A single dot is ヘ in Japanese mode and E in European mode. The syllabaries are counted in our kana guide. The European part has 50 entries on 49 distinct signals, lacks the ITU's é and its service signals, and calls @ the 単価記号, the unit-price sign.
A percent sign takes three characters under the ITU's substitutions
For characters with no signal, section 3 of M.1677-1 sends a percent sign as 0/0, joined to its number by a hyphen ("For 2%, transmit 2-0/0"), and a per-mille sign as 0/00. Quotation marks may be sent as the apostrophe twice "where code converters are used". The table gives each letter a single signal, so a capital and a lower-case letter arrive the same.
@ is already in M.1677 (05/2004), the first edition as ITU publishes it today, beside a note that France's General Committee on Terminology approved "arobase" in December 2002. The exclamation mark, ampersand, semicolon, dollar sign and underscore are absent from M.1677-1, so any signal sent for them is a convention outside the ITU text. Kana live in a national table, and Chinese went to numbers: the Chinese telegraph code used a codebook.
View author profile → · Editorial policy
