"Duplex" means two-way, and "full" means simultaneously — option 1. The three standard transmission modes differ exactly in this respect:

Option 3 describes half duplex, the classic distractor: both directions are available, but they take turns. Options 2 and 4 describe simplex operation, in which the channel never reverses at all.

How full duplex is achieved physically — the two directions must be separated somehow, and there are three common ways:

Space division: two separate physical paths, as in a four-wire circuit or the separate transmit and receive pairs of Ethernet cabling. Simplest and most robust.

Frequency division (FDD): the two directions occupy different frequency bands, as in a mobile phone, where the uplink and downlink are separated by a fixed duplex spacing and a duplexer filter keeps the transmitter out of the receiver.

Time division (TDD): the directions alternate in rapid time slots. Strictly this is half duplex at the physical level, but the slots are short enough that the users perceive a simultaneous two-way link — the arrangement used by DECT cordless phones and by many modern data systems.

The throughput consequence. A full-duplex link can carry its rated capacity in each direction at once, so a "100 Mbps full duplex" Ethernet connection moves 100 Mbps each way and 200 Mbps in total. It also eliminates the collisions inherent in shared half-duplex media, which is why switched full-duplex Ethernet displaced the old shared-hub arrangement entirely.

A related distinction worth keeping separate is that duplex describes direction, while synchronous and asynchronous describe timing; a link may be any combination of the two.

Hence, full duplex permits transmission in both directions at the same time.