At 5:58 a.m. on the 28th of June 1992, the ground began to tremble in the Mojave Desert, California. John Lighton and Frances Duncan, in the midst of a study on ant locomotion energetics, were sitting near a creosote bush collecting data on their computer when the magnitude 7.4 earthquake struck. It was an impressive event: `our car was bouncing on its springs', Lighton recalls. The startled researchers had the presence of mind to key a time stamp into their computer and continue their study. Lighton and Duncan realised that they had a perfect opportunity to test the anecdotal reports that animals can predict earthquakes, which have crept into the popular press in recent years(p. 3103).

Lighton and Duncan had been measuring desert harvester ants' metabolic rate to calculate their cost of transport. Every day at the crack of dawn, as the ants began to forage, the pair laid a compressed-fibre board in the ants' path to separate the ant trail from the CO2 coming from the ground. The ants happily marched through a respirometry chamber nestled on the board, so the researchers could measure the ants' CO2 production. A video trained on the ant trail allowed the pair to count the number of nest-bound and outbound ants and measure each ant's size. Finally, they measured air temperature at `ant height' by placing a thermocouple 3 mm above the ground.

With this suite of measurements taken before, during and after the earthquake, Lighton and Duncan could test whether the ants reacted to the multiple quakes, aftershocks and possible precursors on E-day – the day the earthquake struck – relative to other, earthquake-free days. First,they tested anecdotal reports of ant nest evacuations during earthquakes. To detect any mass exodus, they measured the ratio of nest-bound ants to total traffic (nest-bound plus outbound foragers) on E-day and subsequent days. To their surprise, the earthquake and its aftershocks had no effect on ant traffic dynamics. Next, they examined whether ants ran slower or faster during the earthquake than on other days. They found that temperature accounted for 87% of the ants' running speed, and the relationship between running speed and temperature on E-day was identical to the relationship on subsequent earthquake-free days. Reasoning that seismic activity might cause ants to reallocate tasks and keep ants of certain size inside the nest, Lighton and Duncan looked for differences in foraging ants' body size. Again, there was no difference between E-day and other days. In a last-ditch attempt to find some sign that the ants weren't completely oblivious to the shaking earth, they looked for anomalies in metabolic rate on E-day – and failed to find any. `We were convinced that the ants would show some abnormal behaviour, so we were amazed to find that the ants kept slogging on as usual', Lighton says.

Lighton and Duncan concluded that ants can't predict – and apparently don't even react to – earthquakes. But their study highlights an important issue: what to do if you realise during data collection that you can answer a different hypothesis than the one you set out to test. To test the untestable, like the claim that animals can predict earthquakes, we have little choice but to rely on serendipitous events. After all, how likely is it that you'll get funding to scrutinize animals while patiently waiting for an earthquake to strike?

Lighton, J. R. B. and Duncan, F. D. (
). Shaken, not stirred: a serendipitous study of ants and earthquakes.
J. Exp. Biol.