Anim. Behav., 1998, 55, 1583–1594 An abrupt transition in colony founding behaviour in the ant Messor pergandei SARA CAHAN, KEN R. HELMS & STEVEN W. RISSING Social Insect Research Group, Department of Biology, Arizona State University (Received 31 December 1996; initial acceptance 21 April 1997; final acceptance 15 September 1997; MS. number: 7803) Abstract. Transitions in behaviour across a continuous distribution of organisms can provide valuable information on how variation in behaviour is maintained. We used analyses developed for interspecific hybrid zones to examine geographic variation in colony founding strategy in the desert seed-harvester ant, Messor pergandei. Newly mated females initiate new colonies either alone (haplometrosis) or cooperatively with other foundresses (pleometrosis). The incidence of these founding strategies were surveyed across the species’ range and found to occur in geographically distinct regions joined by a narrow transition zone. Foundresses collected from haplometrotic sites were more likely to display aggression and found solitary nests than foundresses from pleometrotic sites, suggesting that geographi- cal variation in metrosis is due to genotypic divergence. Foundresses from transitional sites were generally not aggressive and tended to co-found nests in the laboratory, yet rarely formed associations in the field. Such an abrupt shift in behaviour indicates that variation in colony founding strategy is maintained by selection rather than the result of secondary contact of neutral characters. Level of aggression displays a wider cline than founding strategy and is likely under selection only when accompanied by active strategy preference. 1998 The Association for the Study of Animal Behaviour Although species are often defined as cohesive approach, based on the dynamics of hybrid zones, evolutionary units (Templeton 1989), intra- which is applicable to continuous populations. specific phenotypic variation across populations Hybrid zones occur when divergent populations is common (Ehrlich & Raven 1961; Slatkin 1987). or species interbreed across a contact zone. If Geographical variation can originate from alternative characters are selectively neutral, a fixation of alternative neutral characters due to steep cline produced by such contact will decay at isolation (Mayr 1942), or from geographically a rate proportional to the rate of gene flow across varying selective pressures causing local adapta- the contact zone, leading to a broad phenotypic tion, as suggested for a variety of variable gradient relative to individual dispersal distance behaviours (e.g. Reichert 1981; Magurran & (Barton & Hewitt 1989). In contrast, selection Seghers 1990). Testing these alternatives is not acting in opposition to dispersal will prevent clinal straightforward, however, as local populations decay and maintain steep transitions between may have accrued multiple phenotypic differences alternative traits (Barton & Hewitt 1985). As which prevent direct comparisons of fitness of strength of selection increases, the corresponding alternative types. One solution to this problem cline should narrow, so different traits displaying is to examine behaviours across many isolated character transitions at the same location should populations and correlate differences with show cline widths proportional to the level of variation in proposed environmental factors such selection acting on each trait (Hagen 1990; as predation intensity or resource availability Sanderson et al. 1992; Nu¨rnberger et al. 1995). (Endler 1995). In this paper we propose a second Such selection estimates have been generated for a variety of characters, including morphology, Correspondence: S. Cahan, Social Insect Research Group, Department of Biology, Arizona State plumage coloration, and life history characters University, Tempe, AZ 85287-1501, U.S.A. (email: (Szymura & Barton 1986; Mallet et al. 1990; [email protected]). Moore & Price 1993), and are supported in direct 0003–3472/98/061583+12 $25.00/0/ar970699 1998 The Association for the Study of Animal Behaviour 1583 1584 Animal Behaviour, 55, 6 tests of proposed selective agents (e.g. Maccallum conducted a large scale survey of colony founding et al. 1995). behaviour to determine the sites and widths of Although hybrid zone theory was developed for behavioural transitions and thus infer the strength interspecific contact zones, its use is not restricted of selection acting to maintain differences in to cases where groups are highly divergent (Mallet colony founding strategy. et al. 1990). Hybrid zone analysis can therefore be a particularly useful tool for studies of intra- specific geographical variation. One such vari- METHODS able trait whose adaptive significance has not been resolved is colony founding behaviour in ants. We surveyed M. pergandei starting nests from During colony reproduction, females leave the 20 February to 10 March 1995 at 19 sites along a colony to mate and start a new nest, either alone 200-km transect from western Arizona to south- (haplometrosis), or cooperatively with one or central California (Table I). Most sites were more co-foundresses (pleometrosis). Cooperation located ca 45 km apart; sampling intervals were may be adaptive because it reduces nest search reduced to 3.75 km within one transitional area to time (Pfennig 1995) or because it confers a increase survey resolution. At each location, we colony-level competitive advantage (Rissing & excavated starting nests and censused foundress Pollock 1987, but see Pfennig 1995). Geographical numbers. No effort was made to control for variation in colony founding strategy has been proximity to adult nests, which has been shown to demonstrated in several ant species, including affect group size, although nests were generally the desert seed-harvester ant, Messor pergandei further than 1 m from active entrances (Pfennig (formerly Veromessor pergandei; Ryti 1988; 1995). When possible, we excavated nests from Rissing et al., in press). This species ranges natural desert areas, primarily ephemeral stream throughout the Sonoran and Mojave deserts of beds (washes) and other low areas. If sufficient the southwestern United States. In south-central numbers of nests (>10) could not be found in Arizona, over 85% of foundresses initiate group natural conditions, starting nests were also nests (Pollock & Rissing 1985). Foundresses surveyed from disturbed areas with substantial collected from these areas readily form associ- bare soil which attract large numbers of ations in the laboratory without respect to foundresses (e.g. along unpaved roads). relatedness, and show a preference for larger Solitary founding in other ant species is group sizes (Hagen et al. 1988; Rissing & Pollock generally mediated by aggressive behaviour 1988; Krebs & Rissing 1991). In contrast, only towards other foundresses attempting to use the solitary nests were found at a location in southern same nest site (e.g. Mintzer 1990). To test California (Ryti 1988). Whether differences in whether aggressive behaviour is associated with incidence of cooperative founding result from solitary founding in M. pergandei, we collected genetically based variation in behaviour or from foundresses from solitary nests at 11 sites. We a facultative response to local environmental placed foundresses from all but one site (site 1) in conditions is unknown. pairs in covered 5.5 cm (diameter)#1.5 cm In this study we investigated maintenance (height) plastic petri dishes. Foundress pairs were of geographical variation in mode of colony observed for a 5-min period. Pairs were consid- foundation in M. pergandei. To do this, we ered aggressive if fighting behaviours (biting identified three behaviours that may influence and/or grappling) were observed at least once, and founding decisions: the level of aggression non-aggressive if no fights occurred. Sample sizes between newly mated queens, queen preference ranged from 9 to 33 pairs per site. The containers for solitary or cooperative colony founding, and we used for foundresses from site 1 were some- group size preference. For the first two, we what smaller than those used for other sites; data assessed whether variation within and between from site 1 are illustrated but excluded from sites was due to intrinsic differences among queens statistical analyses. by testing foundress behaviour under uniform We identified regions of homogeneity in found- laboratory conditions. Group size preference ing strategy and aggression with a simultaneous was identified by analysing the distribution of test procedure (STP) for multiple comparisons of foundress numbers per nest in the field. We then categorical data (Gabriel 1966). This procedure Cahan et al.: Variation in colony founding 1585 Table I. Summary statistics on foundress numbers per nest in Messor pergandei from study sites in south-central California to western Arizona Mean number of foundresses Number per nest Site Location of nests (range) 1 Whitewater Canyon Rd, 1 km N of I-10 (CA) 43 1.00 (1) 2 2 km N of Deep Canyon Research Station (CA) 177 1.00 (1) 3 Berdoo Canyon Rd (CA) 88 1.00 (1) 4 24 km S of Oasis (CA) 67 1.02 (1–2) 5 Aqueduct Rd, 6.8 km E of Fargo Canyon Rd (CA) 74 1.00 (1) 6 Painted Canyon Rd (CA) 42 1.00 (1) 7 Aqueduct Rd, 10 km E of Fargo Canyon Rd (CA) 64 1.09 (1–2) 8 Aqueduct Rd, 13.2 km E of Fargo Canyon Rd (CA) 73 1.36 (1–5) 9 Aqueduct Rd at Pinkham Wash (CA) 72 1.99 (1–8) 10 1.5 km N of North Shore (CA) 23 1.04 (1–2) 11 E mouth of Box Canyon (CA) 21 1.14 (1–2) 12 Frontage Rd Gas Pumping Station (CA) 124 3.67 (1–15) 13 Coachella Canal Rd, Canal crossing no. 26 (CA) 11 2.82 (1–7) 14 Maniobra Valley (CA) 13 3.77 (1–9) 15 Red Cloud Wash (CA) 17 3.24 (1–7) 16 Eagle Mountain Rd, 2.6 km N of I-10 (CA) 20 6.70 (1–18) 17 Highway 177, 23 km N of I-10 (CA) 10 2.70 (1–5) 18 Ford Dry Lake (CA) 19 3.58 (1–10) 19 Ehrenberg, AZ 36 7.39 (1–28) uses an R#C table G-test of independence to goodness-of-fit (Cohen 1960; Johnson & Kotz identify heterogeneity within the entire sample.