The judge in the climate trial doesn't listen to who shouts loudest. Instead, he watches who prompts whom. Nine 'witnesses', from carbon dioxide to cosmic rays, testified from 1984 to 2005. Their statements were recorded every month. The judge doesn't trust mere coincidences, because two witnesses might echo each other without knowing who started first. He looks for a one-way flow of information.

The records were cleaned of slow trends, leaving fast fluctuations. Then, for each pair of 'suspect and temperature', they calculated who prompts whom, with delays from zero to ten months. To filter out random coincidences, the records were shuffled a hundred times to see if the signal persisted. If the prompt disappeared after shuffling, it meant the real data had a genuine signal.

The judge delivered a verdict: carbon dioxide, methane, and volcanic aerosols prompted temperature almost equally, accounting for 24%, 19%, and 23% of the total information flow. Ultraviolet and El Niño also prompted, but more weakly. Cosmic rays, nitrous oxide, and total solar radiation remained silent. Previously, correlations could not distinguish cause from effect: a warmer ocean also releases more CO₂. The judge hears the direction of the prompt. This method does not require model assumptions, so it can serve as an independent check for climate models. Now the judge is ready to hear new cases.