Table 1.

Mixed‐effects survival model, testing how the time an ant i takes to return to a site s after the end of the vth visit, ${\text{RT}}_{v\to v+1}^s$, is influenced by (i) the duration of the vth visit, that is, the dwell time $D_v^s$, (ii) the reproductive caste of the ant, (iii) the community to which it belongs, and (iv) the spatial distribution of brood of different developmental stages

Predictor	HR	SE	z	P
Dwell time, $D_v^s$	1·07	0·00295	21·9	***
Community	1·3	0·0327	7·94	***
Reproductive caste	1·91	0·0607	10·6	***
N eggs at site	1·02	0·00271	8·06	***
N small larvae at site	1·03	0·00578	4·27	***
N large larvae at site	1·04	0·00825	4·51	***
N pupae at site	1·17	0·013	11·8	***

For non‐categorical predictors, the hazard ratio (HR) indicates the instantaneous risk of a return visit to s, relative to the baseline hazard. For the categorical predictors, caste and community, the HR indicates respectively, the instantaneous risk that a queen returns to site s relative to a worker, and the instantaneous risk that an ant in the ‘nurse’ community returns to s, relative to an ant in the ‘other’ community. Colony and ant identity were coded as random factors, with ant identity nested within colony identity. Two colonies were excluded from the analysis because their spatial interaction network had only 1 community, hence ants could not be labelled according to their community. The model was based upon 50 187 site visits, of which 29 391 were uncensored site‐returns.

The ‘***’ indicates P $<{10}^{-4}$.