Data communications over telegraph circuits 1889
In some cases a telegraph circuit was the equivalent of just a wire, albeit with relays and carrier channels and other stuff in the middle, but with no code structure implied. In other cases telegraph circuits included regenerative repeaters, which imposed a 5-unit start-stop code on the system. That is why IBM early on built card-to-tape and tape-to-card converters. This allowed customers to transmit data through TWX or other teletypewriter systems which might include regenerative repeaters. I am not familiar with how they encoded the card data; so they could have used an error detecting code with the data appearing as gibberish on a teleprinter. Or they could have converted card code to a near equivalent in teleprinter code so that the card data would print readably on a teleprinter.
If there were no regenerative repeaters you could transmit any code structure you wished, so long as you didn't exceed the rated speed of the circuit. So you could make e.g. a card-to-card transceiver system that used a code of your own choosing.
In addition to the material in Western Union Technical Review you might be able to find an article "Simplified Printing Telegraph Switching and Integrated Data Processing" by J. B. Booth and R. H. Klich, AIEE Transactions on Communication and Electronics, March 1956, also in Electrical Engineering magazine, April 1956. --
Data communications over telegraph circuits 1890
Things evolved over time. Until about 1950 they had a lot of duplexed ground-return single wire circuits that were time-division multiplexed and ran Baudot. Then they...
jhhaynes at earthlink dot net