Compression lowers the level of a signal on a moment-to-moment basis. So if you have, say, three vocalists bussed to one compressor, whenever any one of those vocalists becomes loud, all three of the voices will be lowered. That can be a good thing or a bad thing depending on what you are trying to accomplish.
I generally find that compressing all of the vocals together works best if I've already compressed each of them individually. That tends to mitigate the undesirable aspects of this effect, preventing any one vocalist from drowning out the others.
I also find that compression is least visible when it is applied in layers anyway. Instead of setting one compressor for a 12 peak reduction, set two compressors for 6db each - using different ratios and thresholds so that the effect is less noticeable. I tend to use very low ratios anyway.