Check out this study I just found from 1981. Seems like P.S. 98 in NYC was right next to some elevated train tracks. When they tested the kids on the side of the building next to the tracks and compared them to the kids on the quiet side of the building, they found that the kids on the noisy side did considerably worse (about .3 - 1 SD, which they interpret as about 3 months to a year behind in reading). They then installed rubber railroad ties and soundproofing materials on the noisy side of the building and re-tested the decibel level (it was significantly decreased). Subsequent reading tests found no differences in achievement between the two sides of the building.
It's not clear that the kids on the two sides were comparable in 1978 (the author says he was assured by the assistant principal that they were, but no other tests were done to check this), but it doesn't seem plausible that they would be so different one year and not the next. It's also not clear whether the kids did better because the noise disrupted them during the testing, because it disrupted their learning during the year, or both.
Either way, it seems pretty incredible that such small changes can have such huge effects. And I'm sure that schools all over the country are trying to find such miracle cures during this testing season -- some schools ask parents to bring food, my school didn't allow kids to use the bathroom so that nobody would be in the hallway making noise, and I'm sure a million other ideas have been tried as well.