This notion of a trap implies a grasshopper (for example) could detect and avoid the ants if they were not hidden. From what expert Frederick Prete told me about orthoptera this isn't likely; it's doubtful a grasshopper can see the minute workers, let alone change course to avoid them in mid-leap. The initial report was missing data I wanted to see, so I decided to look at the ants myself.
As I describe in Chapter 7 of AAA, I put the trap idea to the test in Tiputini, Ecuador. I hung a mosquito net over a shrub with an Allomerus colony, added a 100 orthoptera (katydids and grasshoppers) and sat inside for 5 mornings—a strange case of using a net to keep insects in instead of out. Even after the prey settled down, they were not selective in their movements. They hopped from where ants hid under a structure to where ants strolled in view to where there were no ants at all. I saw no evidence the grasshoppers noticed the ants when the workers were exposed, and whenever one landed among the workers, even those on or in the structures, it escaped unharmed.
Furthermore the ants at my study site did not patiently wait in ambush hour after hour at each opening, as might be expected if the structures are sit-and-wait traps. The image below was taken when conditions were calm. As you see, few workers were stationed at the entryways.
But after I tapped the branch, ants arrived at the entryways within 15 seconds.
This response wasn’t fast enough to allow the ants to catch prey. When I put a grasshopper on the arcade and released it immediately, in all 50 cases it jumped away unscathed. When I used forceps to hold a grasshopper to the arcade for 15 seconds—far longer than it would normally stay put with ants biting it—the workers had time to restrain its body (lower picture).
If the constructions aren’t traps, what is their primary function? The covers run continuously along the branches of the ant's shrub, and contain a column of ants commuting between the colony’s multiple nests, from which workers leave through the many entryways on their foraging expeditions. Such "arcades" likely serve primarily to protect the enclosed traffic against enemies, much as do the trail covers built of soil by marauder ants and driver ants.
One instance of a potential danger to the commuting workers was stopped by the defensive contingent at the arcade entryways. After a day of pulling leaping grasshoppers from my hair, I noticed interlopers of another species, a Pheidole or big-headed ant, climbing the plant to catch a wounded grasshopper missed by the Allomerus. Upon the arrival of the Pheidole, Allomerus ants began to guard each of the 20 or so arcade entrances nearest the commotion caused by the intruders. These guards, aided by nestmates roaming the arcade surface, caught and carried off one Pheidole that wandered onto their trail (below).
Of course defense and predation can be linked, but I think it fair to interpret this meal of a Pheidole as a lucky byproduct of defense—as might also happen with the occasional, hapless grasshopper that lands on any busy trail of aggressive ants.
The fact is, similar structures are built by many other plant-dwelling ants. Here, for example, is the carton-covered trail of an Azteca ant only a stone's throw from the colony of Allomerus.
Scientific notes. The original report: A Dejean, PJ Solano, J Ayroles, B Corbara, J Orivel 2005, Insect behaviour: Arboreal ants build traps to capture prey, Nature 434: 973.
My research was done on the shrub Cordia nodosa, on which the ants build arcades that look identical to those on Hirtella physophora, the host plant to Allomerus decemarticulatus in French Guiana where the original study was done. The orthoptera I used were 6 to 15 mm long, smaller than the ones illustrated in Nature. The inability of my ants to kill orthoptera suggests the workers might have less virulent stings where I studied them in Ecuador than in French Guiana, though the Nature article does not report how often potential prey landed on arcades, the frequency with which the ants seized those prey, or their ultimate kill rate.
This essay expands upon ideas on pages 94 and 95 of AAA, and the corresponding notes.