One can't just do a generic experiment and expect good results. You start with a theory, set up tests and conditions, and then run experiments to prove that theory, and modify the theory to fit the results of the tests. (Or, if one is a BAD scientists/experimenter, makes the facts match the theory. But I digress...)

But when one is focused on the results of the tests - a perfectly legitimate point of view - what about the non-significant occurrences? I think that's what this is about. Humans are made to focus ("zoom lens") or take in a big picture ("wide angle") - but we can't do both at once. I'll compare it to a complicated drawing by Mr. Byrne. I can see the details (e.g., the Lizard's tail) or I can see the full diorama (Batman fighting a spate of Spider-Man's foes) - but I can't appreciate both at once.

Of course, if someone is checking all results of an experiment, then the odd elements should show up too. That's a little hard when time is a factor, though... unfortunately, much research is for a directed purpose, so "free time" to check those other results is restricted.

And when I notice four or five of Mr. Byrne's drawings fit together to form one image... well, there's my serendipity.