This study was conducted on four P. megacephala colonies in Cameroon (on the campus of Yaoundé University, 3°50′N, 11°30′E), and in Mexico (Puerto Morelos, Quintana Roo; 20°N, 86°W) on a huge colony whose territory has spread over several kilometres along the Caribbean coast [16]. The colonies we studied in Yaoundé very likely correspond to the P. megacephala colonies referred to as ‘Cameroon 109’ in Taylor [17] as they were found in the same areas (although these colonies extended for hundreds of meters along the road side, different colony recognition was observed, as workers from different regions will fight each other during confrontation tests; A.D. pers. obs.).

Pheidole megacephala is native to tropical Africa and has become pantropical [18] after having been unwittingly dispersed by human commerce throughout the tropics and sub-tropics. Because it is able to produce dominant supercolonies in introduced areas [3,4,17], it is one of the invasive ants that has most successfully and severely reduced native arthropod abundance both in disturbed habitats and in tropical rain forests [9,19].

As it is the case for most species in this genus, the worker caste is dimorphic, with no intermediary body size between the small minors (approximately 2 mm long) and the big-headed majors (or soldiers; 3–4 mm long). Also, workers have an atrophied sting that they use to lay scent trails, but not to subdue prey or competitors [20]. This species nests in the ground, in termite mounds or in the crevices of tree bark and, like most other invasive ants, is omnivorous. It has always been assumed that P. megacephala is a good predator, because arthropod abundance and diversity decline in areas where it has been introduced [3,17,21,22]; however, this has also recently been demonstrated in its native range [23,24].

To examine P. megacephala's response to potential competing ant species, we reared colonies of the most frequently encountered species in Cameroon, and in a non-invaded area in Mexico adjacent to the invasion zone. The ants tested were reared in the laboratory in 22×2.5cm test tubes supplied with a water source and opening into hunting arenas composed of wooden boxes (50×30×7cm) covered with a plate of glass (see Table 1 for the list of species reared). Depending on the size of the workers from the tested species, one to three test tubes were connected to the hunting area. The mouths of the tubes were stopped with a cork pierced with a hole permitting the reared ants to pass through. One side of each hunting arena was pierced with small holes permitting the P. megacephala workers and soldiers to enter (these holes were plugged with cotton prior to the field experiments). Each time 200–250 workers plus brood (or entire colonies for certain species) were reared for three days prior to the experiments.

For each experiment, we placed one box plus the connected test tubes in the field, on the natural territory of a P. megacephala colony, 3–4 m from a nest entrance (several nest entrance sites were tested in both Cameroon and Mexico). We then monitored P. megacephala workers for two hours following the installation of the introduced ant colonies, or less if the time required for P. megacephala to find and completely destroy the ant colonies was under two hours. The encounters between P. megacephala and workers from colonies of the tested species therefore occurred in the hunting arenas. We observed and scored the four following situations. A: All of the individuals of the tested colony were killed in less than one hour following the discovery of the introduced nest by a P. megacephala scout; B: the same result in between the first and second hour; C: only some of the individuals were killed during the total duration of the experiment (monitoring of each introduced colony for two hours); or D: none of the individuals were killed. Then, a statistical comparison between the results found in Cameroon and Mexico was conducted on the frequency at which species resisted P. megacephala raids from both localities using a contingency table and the Chi-square test for trends (GraphPad Prism 4.0 software).

In all tested cases, the experimental device/colony was discovered in less than 20 min by P. megacephala scout workers (only minor workers) in both study sites. In Cameroon, when these scouts discovered the experimental devices, they entered the hunting arenas and then immediately returned to their nest to recruit nestmates in all cases. Recruited P. megacephala individuals gathered and retrieved refuse from the competing species' nests (prey debris and dead workers deposited in a corner of the foraging arena). In Cameroon, seven species out of twelve resisted the raids without losing a worker (High in Table 1); three others resisted well overall, but in some instances their colonies were killed in more than an hour (Rather high in Table 1). Leptogenys sp. A was the most vulnerable species tested and the only species in the native range with all of its colonies killed, sometimes in less than one hour.

In Mexico, the colonies of eight species out of 11 were always (Low in Table 1; n=5) or often killed (Medium in Table 1; n=3) by P. megacephala. In contrast, only Dorymyrmex pyramicus, Dolichoderus bispinosus and Solenopsis geminata always resisted the raid with little or no loss of life.

It so happens that in both Cameroon and Mexico, the guards of the colonies of some species were ineffectual when faced with small P. megacephala workers, and hardly reacted when the latter entered the hunting arena or even the mouths of the test tubes (i.e., their nest entrance). On the contrary, species that resisted well first counter-attacked, and then stayed just behind the holes pierced on one side of the hunting arena, ready to bite any intruder. Even, D. pyramicus workers, which are relatively small, sometimes exited the hunting arena to attack the intruders outside of the nest.

The overall result is a significant difference between Cameroon and Mexico concerning the frequency at which species resisted P. megacephala raids (Table 1).

The negative impact on native species by invasive ants in areas where they have been introduced is hypothesized to be related to a high level of aggressiveness, because the invaders come from species-rich environments where heightened aggressiveness is adaptive [4,9,11,12]. Indeed, in central Africa, P. megacephala colonies can compete with territorially dominant arboreal ants such as Crematogaster africana and Oecophylla longinoda in the low canopies of cocoa tree plantations [10,17]. In addition, the success of invasive species is facilitated when they are able to take advantage of a ‘niche opportunity’, or the combination of a so-called ‘escape opportunity’ and a ‘resource opportunity’ [1]. An ‘escape opportunity’ arises when native species do not abound or are not effective in keeping out introduced species, a situation that is generally true for islands, but not for the Mexican study site, which is species-rich, at least in ants, and has one of the highest densities of army ants ever noted [16]. A ‘resource opportunity’ arises when the resources that a species needs are highly available, a situation difficult to apply to the Mexican study site as numerous food sources and nesting sites, if not all, were exploited by native ants before P. megacephala was introduced, as shown by a study conducted in an adjacent non-invaded area [16]. It is therefore probable that the P. megacephala colony first successfully invaded the disturbed habitat of the small harbour of Puerto Morelos, far from competition with native ants or predation by army ants, before spreading into the village of Puerto Morelos, and then over a wider, natural, undisturbed area.

The present results highlight the capacity of P. megacephala to raid other ant colonies in both its native and introduced ranges, a behaviour also noted in Linepithema humile and Solenopsis invicta in their introduced range [4,25]. It is likely that, in addition to its ability to achieve unicoloniality, raiding other ants constitutes a prerequisite for P. megacephala for becoming an ‘ecologically dominant’ species in areas where it is introduced. This is particularly true on islands where organisms (including ants) offer only little resistance to invasive species due to their high taxonomic endemism [1,4,7,26,27]. Furthermore, invasive species not only eliminate native species, but also disassemble their communities where each species occupies and maintains its position after a long co-evolutionary process [28]. Comparing these situations for P. megacephala's in both the native and introduced range has given us new insight into ‘how’ disassembling ant communities can initially occur: in its native range, the ant community is structured in large part due to numerous species that are able to resist raids, a trait that we have now shown is reduced or absent in a number of species in the introduced range.

During successful P. megacephala raids in the native and introduced range, the guards of the attacked colonies seem stunned, even when the P. megacephala workers do not come directly into contact with them. Since the same thing has been noted for non-ant prey, it has been suggested that the workers release a secretion from their mandibular glands [24], but secretions from the pygidial gland could also be involved [29]. The size of the attacked workers does not play a role, as relatively large Poneromorphs workers are inefficient guards when confronted with P. megacephala raiders.

Moreover, the efficacy of P. megacephala scout workers is enhanced by the fact that they recruit nestmates when they discover the landmarks of competing ant colonies, avoiding actually having to come into contact with alien workers, and potentially being attacked and even killed; the same is also true when they perceive termite scents [23,24].

In conclusion, this study suggests that P. megacephala's heightened ability to raid the colonies of most of the ant species it encounters in its introduced range permits it to very rapidly destabilize the native ant communities that constitute a first bulwark of defence against invasive ants.