Proper referencing of source material is absolutely critical in scientific writing. This is something we discussed in this week's Writing-Intensive-Learning-Course workshop, in the context of plagarism - something SFU has an unfortunate reputation for.
It's also a very time-consuming and drudging part of my writing. My rough-draft PhD proposal still contains far too many (REF) in places where actual papers need to be. The drudgery part of this comes from digging through the pile of PDFs on my hard drive to find the one that most accurately says what I think it says. The time-consuming part comes from the digging through the hard drive, and using search engines to find papers I don't yet have. But the really critical part comes when I want to make a factual statement, but either I cannot support it, or I'm actually wrong in saying it.
This happened at least once during the time I was writing my M.Sc. thesis. I wanted to say "guppies do X" but in fact they had never been observed to do X, and had been observed to do Y, a mutually-exclusive behaviour I hadn't thought of. Fortunately I caught that error, and it reinforced for me the importance of carefully checking all of my references when I write.
The horrible boringosity of this activity drives me to do other things (ie, procrastinate). For example, I close Word, open IE and post to my blog. Maybe I'll just cheat and try to reference my potential PhD supervisor's work as much as possible. *sigh* Back to work...