The 4 channels (left, center, right & surround) are folded into 2 channels (left & right). The 2 tracks of the dual bilateral track represent the 2 encoded channels, known as Left total (Lt) and Right total (Rt).
The right channel is added to Rt; the left one is put in Lt. The center channel is reduced by 3dB and then added to both Rt and Lt. The surround channel is also taken down 3dB and added to Rt and Lt but first both channels are shifted 90° in opposite directions so that there's a 180° difference between them.
To retrieve the center channel during playback the decoder looks in Lt and Rt for an identical signal which it extracts. To extract the surround channel the decoder looks for identical signals that are 180° out of phase.