The telegraph sounder

A sounder is an electromagnet with a moving armature. Current arrives, the armature strikes downward against the anvil - a bright click. Current stops, a spring pulls it back - a second, quieter click. Between the two lies the duration of the element.

An operator therefore heard no long and short tones but pairs of clicks with different spacing. That sounds like a detail, but it is the reason for almost everything that looks odd about American Morse.

Why American Morse looks the way it does

The original 1844 version has characters with internal gaps: C was not one continuous pattern but two elements with a space inside. At the sounder that was no problem - you were listening to gaps anyway, and one more gap was one more gap.

Over radio it became a disaster. Noise fills every gap, signals smear over distance, and an internal space is no longer distinguishable from a character boundary. That is why Gerke's 1848 revision prevailed, in which every character is one continuous pattern. More under American Morse code.

Here the two can be compared directly: the same text, once in the code of 1844 and once in today's.

The three builds

A sounder in a resonator box was the standard in a telegraph office: the wooden box amplifies the sound and warms it. A bare sounder without a box is brighter, shorter and rattlier. A relay in the next room is the same mechanism heard through a wall - muffled, but still readable, and precisely why one office could work several lines at once.

The proof

Under the device stands the text as it can be read back from the clicks - not the input, but what a listener would get from the spacing alone. With the international code the two agree: the click representation carries the full information, and the sounder is not a sound toy but a complete receiver.

With American Morse they do not agree, and that is the more interesting case. There the internal gap is 1.5 units and the character boundary 2 - too close to separate from spacing alone. An operator could do it, because he knew the context and was used to the other side's fist. Over radio, with noise and smeared signals, neither helped any more. That is what the code failed on, and here you can have it demonstrated.

The threshold between short and long, incidentally, is not given but estimated from the spacings present - exactly as a person does on hearing an unfamiliar fist for the first time. The same method sits inside the decoder.