You need to take into account that signals in the network move at finite speed.
When the colliding parties are spaced apart, the later sender will detect the collision earlier than the other (as the earlier sender's signal has propagated further). The transmission period needs to be long enough so that the earlier sender reliably detects the collision as well. The jamming signal's form on a shared wire is designed to enable easy collision detection.
Note that early CSMA/CD networks used coaxial cable with a shared wire that did not allow instantaneous collision detection, unlike twisted pair or fiber where transmit and receive use separate channels.
Of course, CSMA/CD is all but obsolete and modern networks are completely collision-free.