Measure of the tails of a frequency distibution when compared with a normal distribution. Similar to skewness, kurtosis measures the tailedness of the probability distribution as opposed to the peakedness in the center. A univariate normal distribution has a kurtosis of 3. Distributions having higher kurtosis have fatter tails or more extreme values (a phenomenon called 'leptokurtosis'), and those with lower kurtosis have fatter middles or fewer extreme value (a phenomenon called 'platykurtosis'). From the Greek 'kyrtosis,' convexity.