Do you mean the common tendency to lose the glideslope near the end of (but still on) an approach? (In which case the answer is: you're not scanning fast enough!)
Obviously the glideslope gets more sensitive (lateral deviation corresponding to angular deviation gets smaller) the closer you get to the antenna. I think most people see it drop out from under them on approach because the tendency is to fly the approach slightly high (the idea being that the more clearance you have from the ground, the safer you are).
Of course if you are above the non-precision MDA for any segment of an approach, obstacle clearance is assured, so there's nothing wrong with flying slightly below the glideslope until the very end of the approach, when your natural impulse to pull aft cyclic close to the runway tends to even things out.
That's the 'pilot error' interpretation. I'm always happy to blame something else, though, if there's a technical explanation for it...
