Competitive mechanisms underlying the displacement of native ants by the invasive Argentine ant

Ecology, Jan, 1999 by David A. Holway

Interactions between Argentine ants and native ants were staged as follows. I placed a bait on an index card within 20-30 cm of the nest entrance of each native ant colony. I let the native ant colony recruit to the bait until at least 30 workers were present (20 workers for the poorly recruiting Formica spp.) and recorded the number of native ants once recruitment stabilized. I then placed an Argentine ant laboratory colony 20-30 cm away from the bait, removed the exit hole plug, and replaced it with a 10-cm plastic tube. The tube allowed Argentine ant workers to leave their container and explore the vicinity of the bait. I let Argentine ants interact with native ants for 15-20 min, depending on the resistance offered by the native ant colony. At the end of 15-20 min, I counted the remaining native ants at the bait, aspirated all of the Argentine ant workers, and returned them to their nest container. Once all Argentine ants were removed, I waited 30 min and counted how many native ants recruited back to the bait.

I matched each of the 15 laboratory colonies with each of the seven native ants in this study no more than once. Thus, I conducted 15 replicate experiments for each species of native ant (except for the Formica spp., for which I conducted 12 replicates). For each species, I conducted experiments evenly spread throughout the duration of the experimental period (late May-early July 1995). All native ant colonies were used only once and were at least 100 m from invasion fronts. To analyze the trends in native-ant worker abundance before, during, and after Argentine ant introduction, I used repeated-measures multivariate analysis of variance (MANOVA) and single degree-of-freedom polynomial (linear and quadratic) contrasts as recommended by Gurevitch and Chester (1986). I ran a separate MANOVA for each species of native ant. The linear contrasts tested the null hypothesis that the number of native ants before Argentine ant introduction was not different from the number of native ants present 30 min after Argentine ants were removed. The quadratic contrasts were more relevant; these contrasts tested the null hypothesis that the average number of native ants before and after Argentine ant removal was not different from the number of native ants in the presence of Argentine ants. I also used simple linear regressions to test if the ability of Argentine ants to displace native ants depended on the number of Argentine ant workers and queens in each laboratory colony. The dependent variable in this latter analysis was the percentage reduction in native ants (averaged across species) after introduction of each of the 15 laboratory colonies; arcsine transformations were used here because the dependent variables were proportions.

For each short-term introduction experiment, I also collected data on one-on-one interactions between Argentine ant workers and native ant workers. During each experiment I recorded no more than five one-on-one behavioral interactions, each separated by at least 3 min. For each interaction I recorded the initiator, the mechanisms employed by both species (i.e., physical aggression or chemical defensive compounds), the outcome (i.e., win or lose), the number of each species present, and the temperature. For data analysis these observations were pooled across replicates for each species of native ant. I analyzed the data on one-on-one interactions in several ways. First, I used pairwise G tests to determine if the proportion of one-on-one interactions won by Argentine ant workers differed from the proportion of one-on-one interactions won by workers of each species of native ant. Second, I used simple linear regressions to test if the proportion of one-on-one interactions won by native ants depended on native ant body size. Body mass was estimated using mean dry mass of a sample of 10 workers of each species (20 for M. ergatogyna) oven-dried at 60 [degrees] C for 48 h (Tschinkel 1993). Lastly, I used logistic regression to test if the outcome of one-one interactions (native ant worker won or lost) depended on temperature, initiator, number of native ants present, and number of Argentine ants present. The logistic regressions were performed using LOGIT 2.0 (SYSTAT 1990).