In reply to IPPurewater:
> How much do you mean by 'considerably' ?
Since it depends on the terrain and the visibility of satellites, and the direct paths, it's hard to put a hard-and-fast figure on it, but highclimber's example is pretty good; it
can be 50-100m.
Strictly, it's not multipath that's a problem (because multipath assumes that the direct signal
does get to the receiver); the real problem is where the direct path is blocked, but a reflected signal arrives instead. Imagine you're in a steep valley between two vertical cliff faces, a hundred metres apart. The direct paths to the satellites might not be visible, but the vertical faces might reflect the signal back to you, so the path length is quite a bit longer than the direct path. Since the position fix is calculated by triangulating distances to the known satellite positions, you can be out by as much as the additional path length (if you're unlucky with the satellites you can see indirectly).
In such cases, it's likely that the solution will detect a high error, since the path lengths won't be consistent, so watch the DOP display which will give you an idea that the receiver thinks its solution is questionable; DOP=Dilution Of Precision, a unit-free 'goodness' measure.
I've also seen my receiver (which has a SiRFStarIII chipset) give static offsets from my known location (using clear landmarks), with no multipath reflectors nearby. I still have no idea what caused that, but it snapped back into place some time later (I'd power cycled the receiver, and removed the battery, but it was adamant I was somewhere else).