The remarkable ecological success of social insects is often attributed to their advanced division of labour, which is closely associated with temporal polyethism in which workers transition between different tasks as they age. Young nurses are typically found deep within the nest where they tend to the queen and the brood, whereas older foragers are found near the entrance and outside the nest. However, the individual-level maturation dynamics remain poorly understood because following individuals over relevant timescales is difficult, hence previous experimental studies used same-age cohort designs. To address this, we used an automated tracking system to follow >500 individuals over >100 days, and constructed networks of physical contacts to provide a continuous measure of worker social maturity. These analyses revealed that most workers occupied one of two steady-states, namely a low-maturity nurse state and a high-maturity forager state, with the remaining workers rapidly transitioning between these states. There was considerable variation in the age at transition, and, surprisingly, the transition probability transition was age-independent. This suggests that the transition is largely stochastic rather than a hard-wired age-dependent physiological change. Despite the variation in timing, the transition dynamics were highly stereotyped. Transitioning workers moved from the nurse to the forager state according to an S-shaped trajectory, and only began foraging after completing the transition. Stochastic switching, which occurs in many other biological systems, may provide ant colonies with robustness to extrinsic perturbations by allowing the colony to decouple its division of labour from its demography.

Experimental set-up

To investigate long-term worker behavioral maturation, we continuously recorded three queenright Camponotus fellah colonies for ~100-days. The environmental conditions and nest dimensions were identical to those described in Mersch et al (2013). Thus, colonies were housed in a double-chamber setup, with a nest box (26x16cm) that was dark 24 hours a day, connected (via a tunnel) to a foraging arena (26x16cm) under a 12:12 light-dark and temperature cycle. Underfloor reserves of water and dilute sugar solution (3g/100ml) were connected to the foraging arena via cotton wicks. Ten dead flies were also provided weekly in the foraging arena.

Automated ant tracking

The location of every worker within the nest and foraging boxes was determined by gluing a unique ARtag barcode to the thorax of each individual in the colony, and recording their movements using monochrome digital video cameras (4560x3048 pixels at 2 frames/sec) coupled to a machine-vision system (Fiala, 2005; Mersch et al., 2013). Physical contacts between workers were extracted based on the proximity and orientation of workers relative to one another.

Worker age determination

For several years prior to the experiment, and during it, newly-eclosed workers were paint-marked on a weekly basis. All worker ages given in the manuscript reflect the time elapsed since the day when the worker was painted, hence the actual age of a `5 day-old' ant is in the range 5 to 5+7=12 days old.  The head, thorax and gaster of each worker was marked with a paint code indicating the week of birth.
Codes were applied after immobilizing the ants by placing them in a petri dish placed on ice (dish temperature 2^oC). Cooling typically lasted <5 minutes, and painted ants recovered within few minutes. Each newly-painted ant was photographed with a hand-held digital microscope (MAOZUA, 5MP USB Microscope) and comments about their physical appearance were referenced in a table. These photographs and comments were later used to distinguish between ants with the same colour code (the same age) that lost their identification tags during the same week.

Each .txt.gz file contains a time-ordered edge-list representing all pairwise physical contacts between ants over a one week period. These files are named according to the colony identity and the week.

The accompanying .xlsx files provide both the AprilTag number, and the paint-code of each individual.

The Col_Ant.csv file contains the date of birth of each ant.

The Experiment_description.csv file gives the start and end times for each week of tracking for each colony.

The README_for_contacts.txt file contains a description of the columns in the contacts files.