"Initially, we looked at these huge OH and HO 2 signals found in the clouds and asked, 'What is wrong with our instrument?,'" said William Brune, a meteorologist at Penn State. "We assumed there was noise in the instrument, so we removed the huge signals from the dataset and shelved them for later study." The data were collected from an instrument on a plane flown above Colorado and Oklahoma; the instrument analyzed chemical changes thunderstorms and lightning make in the atmosphere. The researchers determined the signals were really hydroxyl and hydroperoxyl and worked with colleagues to see if these signals could be reproduced by sparks and subvisible discharges in the laboratory. The team then performed a re-analysis of the thunderstorm and lightning dataset.