In case this is your first visit to SQLBlog.com, Linchi Shea does very methodical analysis of the performance and behaviour of disk arrays. His most recent listings specifically explore file fragmentation and its affects on performance on storage. The posts sort of give one the impression that there is no significant impact on performance when the disk storage is a SAN. I'd be inclined to agree, and his data certainly backs that up. But in many in the posts he does compare direct attached storage, where we do see degradation.
I'm writing this little blurb because my hunch is, at least from what I've encountered, the majority of SQL installations use DAS or "quasi-SANs" (Linchi, I'm sure you've run into some of these devices that claim to be SANs but are sorely lacking). In those cases, especially ones where databases are unmonitored and auto-grow occurs on a regular basis (or transaction log auto-grow/manual shrink accordianism), Linchi actually demonstrates that there will be performance degradation. This is one of the most common performance issues I run into out in the "jungle". Fragmentation.
My message - those of you on direct storage or if you aren't certain of fragmentation affects, error on the safe side and defragment. Or at the very least investigate how physically fragmented your arrays are. Please.