I was recently watching the now famous Invisible Gorilla experiment. When this came out I loved the experiment and what it said about perception and selective vision. Watching it again - just now - I actually tried the experiment - counting the passes by the light colored team players, and in doing so noticed a fatal flaw in the experiment.
There are six players and one gorilla. The light colored team wears jeans and light colored shirts. The dark colored team, has one player wearing jeans and a dark shirt and two players wearing black pants and shirts. These two players, when diffused in your perphial vision (not focused) don't look much different from the gorilla - black fur pants and top, dark hair.
This appears to be a fatal flaw in the experiement. For it to be as scientific as possible, the light and dark team should have all players wear the same color jeans, and the appropriate colored top. Additionally there should be the same sample of dark and light color haired players across the teams.
I would expect that if the above conditions were met, the gorilla would be most distinguishing.
I'm not disagreeing with the hypothesis that we miss stuff because of a variety of reasons (it matches other stuff, we're paying attention to other variables). But I do think this landmark experiment could be improved. It's got biases built into it.