Strict coevolution requires that interactions among organisms be species-specific. We assessed the relative roles of host- and habitat-specificity in determining the match between a genus of myrmecophytic trees and a guild of obligate plant-ants in the moist tropical forests of Madre de Dios, Peru. Four locally coexisting but habitat-restricted Cecropia species were cultivated in screen tents until all plants had developed myrmecophytic traits. Saplings were then placed within replicate blocks of each of two habitat types: riversides and small forest light gaps. Colonization events were recorded every 3 d between June and August of 1992, and queens were later removed from stem internodes for identification and brood censuses. A similar experiment, conducted in September through November of 1993, included just two species of Cecropia hosts. Effects of host species and habitat on queen colonization rates were evaluated by log-likelihood goodness-of-fit tests and contingency table tests. For three ant species, we also conducted queen preference experiments to compare queen behaviors across a range of host plants. Differences among ants in the extent of habitat-specificity vs. host-specificity provide evidence for multiple evolutionary routes to obligate association with Cecropia. Habitat-specificity exceeded host-specificity in Azteca ovaticeps (Dolichoderinae), for which queen preference experiments revealed no significant discrimination among hosts. This extreme riverside specialist is thought to have descended from generalist live-stem nesters in second-growth habitats. In Azteca australis; host-specificity was strong, and in this species only, directed toward hosts where brood production was most successful. Conflicting habitat associations in the two experiments indicated the weakness or absence of a consistent habitat affiliation in Azteca australis and suggested that colonization frequencies were influenced instead by proximity to foundress sources. Close relatives of A. australis live in exposed carton nests, which may have been positioned ancestrally on key resource plants, e.g., those producing lipid- and amino-acid-rich pearl bodies. Pachycondyla luteola (Ponerinae) exhibited both strong habitat and host associations and may have undergone pair-wise coevolution with its forest-gap-dwelling primary host. Queens of Camponotus balzani (Formicinae), possibly a recent and secondary associate of Cecropia, were overrepresented in forest gap habitat but were host generalists, underrepresented only on a host with extremely small internodes. Apparently greater host-specificity in C. balzani at later stages of colony establishment may be due to differential post colonization mortality on the various hosts. Attack of ant queens by parasitoid wasps was strongly concentrated in the linear riverside habitat and weak to absent in the patchily distributed forest gap habitat. Due to lower rates of either parasitoid attack or other forms of queen mortality, Camponotus balzani experienced greater success in the forest gap habitat, where it was overrepresented in colonization experiments. Historical coincidences and preadaptations appear to have strongly influenced pairings between Cecropia species and their obligate plant-ants and account for much of the 'apparent' niche partitioning observed in the system. Species-specificity seems to be determined mainly by coincident habitat affiliations of ants and plants ('coordinated dispersal') and by preadapted capacities of ants to distinguish among host-plant species. Multiple mechanisms for species-specificity may be characteristic of relationships in which associates disperse separately from one another (i.e., show horizontal transmission). Our results are consistent with the view that coadaptation and co-cladogenesis are more likely in systems where dispersal of associates is tightly coupled. Key words: ant-plant; ants; Cecropia; coevolution; colonization; coordinated dispersal; ecological fitting; habitat-specificity; host-specificity; mutualism; parasitoid wasps; preadaptation., INTRODUCTION A key challenge of coevolutionary studies is to understand the ecological and evolutionary factors driving interacting species to undergo progressive bouts of reciprocal evolutionary change (Ehrlich and Raven 1964, [...]