Software monitoring is quite simple in practice.
Let's say you have a guitar, plugged into your interface, which sends to Logic. The interface is also connected to your speakers.
Now, when you want to record your guitar, you'll obviously want to hear it through the speakers. So somehow, your guitar signal coming in on the audio interface's input, has to be sent to it's output, so you can hear it on the speakers.
Ok. So, what we could do, in the software for the interface, is "hard wire" the input directly to the output - now when you play the guitar, it instantly comes out of the speakers. And when you play Logic, that's also coming out of the speakers (the audio interface has an internal mixer that mixes everything together and sends it to the outputs).
In this scenario, we are using what's called "direct monitoring", because the input is going directly to the outputs, inside the audio interface. In this configuration, Software Monitoring should be turned OFF, as we do not wish the monitor through Logic ('cos we're already hearing our guitar, innit!)
To record, just select a track, pick the correct input, and start recording.
Ok. One positive to this approach is that we get minimal latency on our guitar part - good for playing.
A downside is that because we are not monitoring through Logic, we can't add Logic's effects to the guitar part in real time (although we can
after a part has been recorded, obviously). So no snazzy Guitar Amp Pro and reverbs etc on our guitar. Boo! Hisss!!
If we want to play and hear Logic's FX, we must change our routing. We disconnect the direct connection from our guitar input to the interface's output. Now when we play guitar, we can't hear anything at all!
But because Logic is also going to the outputs, we can send our guitar signal, from our input,
through Logic, and therefore back out to the speakers. Logic passes the signal right through. And in the process of doing this, we can add all our Logic effects! Yay!
Ok, so we turn Software Monitoring ON, and record enable our track, still set to record from our guitar input. Play the guitar, and you'll see Logic's meters move,
and hear what you are playing. Even better, we can add a Guitar Amp Pro plugin to that track, and now hear the whole guitar sound.
(When recording, we will still only be recording the dry signal, so we have the flexibility to change settings on the Guitar Amp Pro plugin even after recording).
The downside to this approach is that, because the signal has to go into Logic and back out again, you will get some latency. How much you get will depend on your hardware, your buffer settings, and what type of thing you are playing - some applications the latency isn't really noticeable at low settings, other times (maybe drums or guitar playing) you might hate it - in this instance, use direct monitoring, and external effects if you need them.
Does that help make things a bit clearer?