Actually you can get a compressor that will handle any spikes, but it doesn't sound good that way. If you are using a computer, you can try experimenting with some of the software compressors, and then take what you learn back to the hardware. The reason this helps is that the good software compressors have loads of settings, and you can change the settings very precisely over a humongous range. ( I.E. attack time = 0.0001 etc ) They also have presets, like "classic vocal" that allow you to see what all the settings are for a typical vocal compression etc, and they have a chart of the compression to volume, with a graphical readout of what's going on, and peak holding meters. If you don't have a pro soundcard, you won't want to actually use the software compressor, but I believe them to be an excellent learning tool. Some of them actually have some pretty good documentation too. The best companies, I think, are Waves, DSP-FX, and Sonic Foundry. (Steinberg interfaces piss me off.)
Iain.