2008, bridges were removed and dead ants were collected from nest boxes and foraging
trays and stored in alcohol until they were counted to assess mortality from each colony.
Statistical analysis
All statistical analyses for the colony-level competition experiment were
performed in JMP 7.0.2 (SAS Institute Inc., Cary, NC, 2007). The following response
variables were used in analyses for each colony or pair of colonies (in the case of
resource consumption): mortality at the end of the experiment (number of dead ants not
stuck in Tanglefoot), mealworm and sugar water consumption (averaged across the 3
pieces of each bait given to each colony), total recruitment across all time periods and
time until recruitment (when 2 or more ants of the same species were present) for each
bait type, the total number of ants fighting and the total number of ants in the other
species’ nest box summed across all observations, and final bait control (the species
present on each bait type after food was replaced on day 5). We tested for correlations
between all response variables. For all variables other than bait dominance, the data were
analyzed as follows: control colonies were compared using ANOVA’s with species,
standardizing factor (biomass or worker numbers), and the interaction between the two as
factors. Post-hoc analyses on significant interactions were conducted using Tukey5 s HSD
tests. To compare between species in the competition treatment, the difference between
the two species was calculated and the mean was tested for significant deviation from 0
using a 1-group, 2-tailed t-test. The effect of standardizing factor was analyzed using a
one-way ANOVA. The following transformations were performed to meet assumptions
of normality: mealworm consumption for control colonies and total ants fighting were
square root transformed, and mortality and mealworm and sugar water recruitment were
43