We define the characteristics of plume events that can be called superplumes. Using the surface area of the smallest oceanic plateau generated during the Cretaceous superplume event, we define a cutoff surface area for superplume flood basalts of > 410,000 km2. We show that the maximum widths of feeder dikes of plume lavas are linearly related to the square root of the surface area covered by their flood basalts. From this we derive a cutoff: the widest feeder dikes of a superplume must be >74 meters wide. All high Mg rocks as defined by Isley and Abbott (in press) are superplume rocks. Layered intrusions formed by superplumes have either high platinum group element abundances and/or high Cr. We use all of the data from the superplume proxies: flood basalts, dike swarms, high Mg rocks, and layered intrusions to define the duration of superplume events over Earth history. Over two thirds of the superplume periods last less than 8 million years. However, there are a few very long superplume periods, with durations of 79, 32, and 22 Ma, producing a very long tail to the distribution. We infer that a second geological process is producing the long tail, perhaps resonances between free and forced core nutations or times with many large impact events or a plate tectonic process. Finally, we use our data on maximum dike widths and flood basalt surface area to construct estimates of the overall surface area covered by lava during superplume events over the last 2.9 Ga. We find that the largest Precambrian superplume events were at least ten times larger than the largest Phanerozoic superplume event. We also find evidence for many superplume events earlier than 2.9 Ga, but due to a lack of data on maximum widths of feeder dikes, we are unable to estimate the relative sizes of most of these events.