Record Breaking and Temporal Clustering

Casual observation suggests that athletics records tend to cluster over time. After prolonged periods without new records, a record breaking performance spurs other athletes to increase effort and thereby repeatedly set new standards. Subsequently, record breaking subsides and the pattern repeats itself. The clustering hypothesis is tested for the mile run, the marathon, the world hour record and long jump. For all four disciplines, the null hypothesis of non-clustering is rejected at the 4% level or below. A theoretical rationale for this phenomenon is provided through a model of social learning under limited awareness. The agents are assumed to be unaware of the true limits to performance and to take the current record as the upper bound. The observation of a record breaking achievement spurs the agents to try harder and thus temporarily increase the probability of new records. Subsequently, record breaking trails off and the process is repeated.