If you are recording an acoustic instrument or voice with no pre-existing recorded material as an accompaniment, then you probably don't need to worry about monitoring. Just make sure you've made the right connections and you should be ready to record without reading this section. However, if a musician is playing an instrument (it doesn't matter what kind) while listening to some pre-existing material, then it is important that some mechanism exists to allow her to hear both her own playing and the accompaniment. The same is true in a slightly different way if the instrument makes no sound until the electrical signal it creates has been amplified and fed to some loudspeakers. Listening to the performance in this way is called monitoring. So, if you are recording an electrical or software instrument/signal, and/or the musician wants to listen to existing material while performing, then you need to ensure that signal routing is setup to allow monitoring. You have 2 basic choices:

Hardware monitoring uses the capabilities of your audio interface to route an incoming signal (e.g. someone playing a guitar into a microphone) to an output connection (for example, the speaker outputs, or a dedicated analog monitoring stereo pair). Most audio interfaces can do this, but how you get them to do so, and what else they can do varies greatly. We can divide audio interfaces into 3 general categories: relatively simple, typically stereo, devices that allow the signal being recorded to be routed back to the main outputs (most "consumer" audio interfaces fit this description, along with anything that provides an "AC97-compliant CODEC") multichannel devices that allow a given input channel to be routed back to its corresponding output channel (the main example is the RME Digi9652) multichannel devices that allow any input channel, along with any playback channel, to be routed to any output channel (the RME HDSP and various interfaces based on the envy24/ice1712 chipsets, such as the M-Audio Delta 1010, EZ-8 and various Terratec cards)

For interfaces in the first category, there is no standard method of getting the signal routing correct. The variations in the wiring of hardware mixing chips, and the capabilities of those chips, means that you will have to get familiar with a hardware mixer control program and the details of your audio interface. In the simple cases, simply increasing the level named "Line In" or "Mic" in the hardware mixer control program will suffice. But this is not a general rule, because there is no general rule. The following diagram shows a fairly typical AC97-based audio interface schematic: Notice: there are multiple input connections, but only one can be used as the capture source it is (normally) possible to route the input signals back to the outputs, and independently control the gain for this "monitored" signal it may or may not be possible to choose the playback stream as the capture stream

For the only interface in the second category, the RME Digi9652 ("Hammerfall"), the direct monitoring facilities are simplistic but useful in some circumstances. They are best controlled using JACK hardware monitoring. When using one of the interfaces in the third category, most people find it useful to use hardware monitoring, but prefer to control it using a dedicated hardware mixer control program. If you have an RME HDSP system, then hdspmixer is the relevant program. For interfaces based on the envy24/ice1712/ice1724 chipsets, such as the Delta1010, Terratecs and others, envy24ctl is the right choice. Both programs offer access to very powerful matrix mixers that permit many different variations on signal routing, for both incoming signals and the signals being played back by the computer. You will need to spend some time working with these programs to grasp their potential and their usage in different situations. The following diagram gives a partial view of the monitoring schemantics for this class of audio interface. Each input can be routed back to any output, and each such routing has its own gain control. The diagram only shows the routings for "in1" to avoid becoming completely incomprehensible.

Much simpler than hardware monitoring is "software monitoring". This means that any incoming signal (say, through a Line In connector) is delivered to software (such as Ardour) which can then deliver it back to any output it chooses, possibly having subjected it to various processing beforehand. The software can also mix signals together before delivering them back to the output. The fact that software monitoring can blend together incoming audio with pre-recorded material while adjusting for latency and other factors is the big plus for this method. The major downside is latency. There will always be a delay between the signal arriving at your audio interface inputs and it re-emerging from the outputs, and if this delay is too long, it can cause problems for the performer who is listening. They will sense a delay between pressing a key/pulling the bow/hitting the drum etc. and hearing the sound it produces. However, if your system is capable of low latency audio, its likely that you can use software monitoring effectively if it suits your goals.