Ants build efficient transportation networks

Argentine ant colonies have many nests which they connect using an extensive network of chemical trails. We studied the efficiency of these trail networks using lab experiments where ants were given the task of connecting several nests. We found that ants consistently connected nests using the shortest amount of trail. The ants built these efficient networks using a process of elimination where inefficient and redundant trails gradually disappeared from the network.

HFSP Program Grant Award holders Toshiyuki Nakagaki, David Sumpter, Martin Middendorf and Madeleine Beekman and colleagues
authored on Mon, 05 September 2011

Transportation networks are an important part of human society. Efficient transportation networks allow us to distribute resources and people to different locations.  The argentine ant (Linepithima humile) also relies on a transportation network to connect its many nests. Argentine ant colonies consist of hundreds of nests, all of which are connected through chemical trails. As ants walk between nests, they deposit small droplets of an attractive chemical (a ‘pheromone’) on the ground.  These droplets form a chemical trail that allows ants to find their way between nests.  Like human engineers, ants face the problem of building networks that are efficient. Long meandering trails may cost the colony considerable amounts of energy since ants burn precious calories as they walk; travel on trails is also dangerous, so having short trails can reduce the risk of predation.  Thus, there’s reason to believe that ants may have evolved mechanisms to allow them to construct efficient trail networks. However, unlike humans, ants have no leaders and no form of centralised control.  Can ants build efficient trail networks despite the lack of leadership? And if so, how do they do it? We set out to answer these questions using captive colonies of argentine ants.


Figure A composite picture showing ant trail networks after 6 hours. Inset photo shows a diagram of the shortest possible way to connect three points arranged in a triangle.

We challenged ant colonies with the problem of connecting either three or four nests. We placed ant nests beneath large circular tubs (our ‘arena’). Each nest was connected to the arena via a stick that passed through a hole drilled into the base of the arena. Nests were arranged in either a square or a triangle configuration. We chose these simple configurations because optimal network solutions were relatively easy to calculate. We took pictures of the arena every 2 minutes and ran the experiment for a total of 6 hours. To make trails more visible, we superimposed 10 minutes worth of photographs to create composite images in which the trails were clearly visible.

Amazingly, most ant colonies created minimal networks (see figure), connecting their nests using  the shortest amount of trail. This is surprising given that ants have tiny brains and lack centralised leadership. Nevertheless, they were able to rapidly create highly efficient networks. To find out how ants accomplished this feat, we examined time lapse photographs of network formation. We found that ants initially built a complex, inefficient network, but within hours, dead-end and redundant trails disappeared until the final, efficient network was achieved.  The process underlying this ‘process of elimination’ is based on the fact that the ant’s trail pheromone has a relatively fast evaporation rate. As a result, long trails, which take longer to traverse, do not get reinforced as much as short, direct trails. Thus, the ants are using a self-organised, decentralised process to build short, efficient networks.

Our work has shown that despite their lack of centralised control and tiny brains, ants are capable of creating efficient transportation networks using a process of elimination.  The knowledge gained from this study could help human engineers and computer scientists in the design of self-organised networks. 

Text by Tanya Latty


Structure and formation of ant transportation networks. Latty T, Ramsch K, Ito K, Nakagaki T, Sumpter DJ, Middendorf M, Beekman M.  J R Soc Interface. 2011 Sep 7;8(62):1298-306.

Pubmed link