The study was conducted in the understorey of primary rainforest on CI (10°25′ S, 105°40′ E) during the transition from dry to wet season (October–January). CI is oceanic (135 km2, maximum elevation 360 m) and dominated by a broad-leaved, structurally simple tropical rainforest. Many woody plants have fleshy fruits or arillate seeds (44%, n = 108 species) and the fruit/aril colours of species in the forest understorey are red–orange (50%), black (25%), white (10%), pink (10%) or yellow (5%), and diaspore diameters averaged 5.5 mm (range 2–15 mm, n = 20; P. T. Green 1994–2009, unpublished data).
High-density supercolonies of the YCA were always associated with highly elevated densities of exotic, honeydew-secreting scale insects (especially Tachardina aurantiaca
(Kerriidae) and Coccus celatus
(Coccidae), both broad host–plant generalists; O'Dowd et al. 2003
). YCA invasion has generated a mosaic of three forest states, primarily by extirpating a keystone native consumer (the red land crab, Gecarcoidea natalis
) (Green et al. 2008
). We chose five 1 ha sites of each state (i.e. YCA-invaded sites, where supercolonies had formed 1.5–2 years prior to the study; intact sites where supercolonies had never formed; and ghosted sites, which had not been invaded but where resident red crabs had been depopulated when previous crab breeding migrations intercepted supercolonies and were killed) to estimate the effects of ant invasion on avian frugivory. Sites were interspersed (0.7–10.7 km apart) across the island and spread across similar elevations (Davis et al. 2008
). The inclusion of ghosted sites allowed a comparison of impacts in the absence of both the native red crab and YCAs to those in the absence of crabs only, and in the presence of both species. This helped distinguish direct effects of ants on fruit handling from their indirect effects on resource levels and habitat structure.
Artificial fruiting displays and model fruits were used to measure fruit handling by birds (a,b). Displays were wooden dowels (0.8 cm diameter × 45 cm length), each with five holes 4 cm apart through which galvanized wire (0.7 mm diameter) was threaded. Model fruits (12 mm diameter) were moulded from non-toxic, odourless and water-resistant red Plastalina modelling clay and pressed onto the outer, up-facing ends of wires (10 fruits per display). Ten displays were attached to separate understorey plants at breast height and placed greater than 25 m apart within a 100 m × 50 m plot. One fruit survey was completed at each site in October, November and January. Each survey was over 4 days at five sites, staggered randomly within each month. Artificial displays allow standardization of the number and arrangement of fruits and placement within sites. The colour and size of model fruits can be controlled, as they neither rot nor desiccate, and records of visitation are retained. The number and type of peck marks could be assigned to the thrush (c) and white-eye (d) because of different bill morphologies and gape widths (thrush: 13.97 ± 0.52 mm, n = 3; white-eye: 6.47 ± 0.03 mm, n = 3). The diameter of model fruits was large relative to the gape widths of these two frugivorous birds to minimize chances of fruit removal and hence retain feeding attempts.
(a) Artificial fruiting display with model fruits, (b) red model fruits (scale bar, 1 mm) handled by the thrush ((c) left fruit; credit: Kee Seng Foo) or white-eye ((d) right fruit; credit: Tony Patisser).
We compared the handling of model fruits to real fruits of Schefflera elliptica (Araliaceae) at a different ant-invaded site. Ripe fruits were orange–red and 5.1 mm in diameter (n = 100). The presentation of displays was as for the fruit handling survey, except that 20 displays each of real and model fruits were used. We excluded ants from 10 displays of each by placing Tanglefoot bands at the base of each dowel. YCA numbers on each display were recorded after 2 days. After 4 days, handling rates were determined.
Bayesian analyses were used throughout (see the electronic supplementary material). The total number of fruits that had peck marks and the total number of peck marks made by each bird species per site were examined for variation among forest states using Bayesian analysis of covariance. For comparison of the handling rates of real and model fruits, and for the test for direct interference by ants on fruit handling, we analysed probabilities that fruits in a display would be handled, numbers of ants per display and peck marks per display of model fruits.