General Electric Review, Ocober, 1920, page 808:

Duplex  Radiophone  Receiver  on  U.S.S.  George  Washington
By  HARROLD  H.  BEVERAGE
RADIO  ENGINEERING  DEPARTMENT,  GENERAL  ELECTRIC  COMPANY
 
Duplex Arrangements at New Brunswick
    At New Brunswick, Mr. Burke Bradbury made arrangements for duplex operation by setting up the receiving apparatus at a point about four miles from the radio station, and sending the amplified received currents back to the radio station over an existing telephone line connecting New Brunswick with the Marconi receiving station at Belar.
    Fig. 2 shows the connections used for duplex operation at New Brunswick. It will be noted that the received currents are introduced in series with the toll line leading from New Brunswick, enabling the party talking from Washington or any other point to hear the incoming radiophone speech over the same wires which transmit his own speech to New Brunswick, as in an ordinary land wire connection. It is also evident that, with this connection, the speech and signals picked up by the receiving apparatus at New Brunswick will modulate the alternator output and be re-radiated again at New Brunswick's wave length. Anyone listening on New Brunswick's wave length would, therefore, hear both sides of the conversation. This explains a point which puzzled many amateur operators, who reported that they heard the George Washington radiophone on 8000 or 13,600 meters, whereas the wave length of the radiophone on the George Washington was 1800 meters. The wave lengths of 8000 and 13,600 meters were both used at New Brunswick for radiophone tests at various times. The writer often heard short wave spark signals while listening to the New Brunswick radiophone on the George Washington, the signals being picked up by the receiving apparatus at New Brunswick and being re-radiated in the manner described.
Figure 2