Well, since we have no accurate idea of how the spiders really work, we can only guess, right?

But consider this: the spider does need to infer something about the structure of a page to determine its contents. If the html is broken, it's certainly possible that it will interpret incorrectly or even miss parts of your page. Not likely, but *possible*.

So why NOT do it right? It's only a small amount of extra effort and it might make a difference.