Preferences for sugars and amino acids and their conditionality in a diverse nectar-feeding ant community


and present address: Nico Blüthgen, Department of Animal Ecology and Tropical Biology, Biozentrum, Am Hubland, D-97074 Würzburg, Germany. E-mail:


  • 1Feeding preferences of nectarivorous ants for sugars and amino acids were studied in an Australian tropical rain forest using artificial nectar solutions. Fifty-one ant species were recorded feeding on the solutions.
  • 2Preferences among carbohydrates were principally concordant between ant species. In paired tests, sucrose was often preferred over fructose, glucose, maltose, melezitose, raffinose and xylose, respectively. Attractiveness of sucrose baits increased with concentration.
  • 3Many ant species preferred sugar solutions containing mixtures of amino acids over pure sugar solutions. However, preferences among seven pairs of single amino acids in sugar solutions differed substantially between ant species, including several cases where different ant species displayed significant opposite choices.
  • 4Ant selectivity between solutions was significantly reduced when different ant species co-occurred on the same bait. Preferences for single amino acids were also reduced when colonies fed extensively on the same compound prior to the experiment for 2 days.
  • 5Our results indicate that both interspecific variability in gustatory preferences and conditional effects such as competition and colony requirements affect resource selection in multispecies communities. These processes may be crucial in niche partitioning of species-rich nectarivore assemblages.


Niche partitioning is a basic concept in community ecology (Schoener 1974). Within functional groups using the same type of resource, niche partitioning may involve spatio-temporal heterogeneity or other modes of differentiation of resource components. Trade-offs in life-history parameters such as dispersal abilities and non-equilibrium processes such as disturbances may promote differentiation and prevent competitive exclusion. These concepts are particularly crucial and controversial in tropical ecosystems, where great numbers of species coexist that apparently belong to the same functional group (Novotny et al. 2002; Sheil & Burslem 2003). In the absence of competition, however, niche specialization may be relaxed. Such conditionality in resource selectivity as a consequence of asymmetric competition has been demonstrated in feeding experiments with three hummingbird species (Pimm, Rosenzweig & Mitchell 1985).

The functional group of nectarivores is highly diverse in many ecosystems. For floral nectar, flower shapes provide the most obvious feature in association with niche differentiation, e.g. tubular flowers constrain the nectarivore spectrum and promote specialization (Pyke 1982). Visual and olfactory cues or flowering phenology are other examples, although universal flower/pollinator syndromes may be conceptually too simplified and trends towards specialization have been over-emphasized in the past (Waser et al. 1996; Ollerton & Cranmer 2002). In extrafloral nectar, which is usually presented openly, potentially constraining niche dimensions may be fewer than in flowers, although trichomes and surface structures may limit the visitor spectrum to some extent (Davidson, Snelling & Longino 1989; Federle, Rohrseitz & Hölldobler 2000). Furthermore, only behaviourally specialized insects that respond to specific cues may exploit efficiently honeydew excretions of herbivores. One important aspect, however, has been poorly explored so far: the composition of the nectar or honeydew itself may determine preferences of, and partitioning among, nectarivores. Concentration and composition of sugars as the main nutrient component in nectar has been correlated with responses of flower visitors (Percival 1961). Pioneered by Baker & Baker (1973, 1983), amino acids and other substances in nectars have been investigated and certain syndromes have been recognized and related to floral visitors (but see Gottsberger, Schrauwen & Linskens 1984). Studies on actual substance preferences or physiological requirements of nectarivores remain scant (Gardener & Gillman 2002), but provide powerful hypotheses to explain differential resource use. For example, the absence of invertase in some taxa may prevent sucrose digestion and correlate with preferences for monosaccharides (Martínez del Rio 1990). Trisaccharides common in honeydew and attractive to some insects may be harmful to others (Zoebelein 1956; Wäckers 2000). The uptake rate of sugar solutions may depend on interactions between physical properties and mouthpart structures, and trade-offs between viscosity and energy content may occur (Adler 1989; Hainsworth, Precup & Hamill 1991). Physiological requirements of nectarivores may also affect reproductive success as detected for butterflies (O’Brien, Fogel & Boggs 2002).

Ants are the most common animals consuming extrafloral nectars and honeydew (Buckley 1987; Koptur 1992) and they may be common on some flowers, although conspicuously absent on others (Janzen 1977), sometimes as a consequence of repellent substances in nectar or floral tissue (Ghazoul 2001). A number of studies have examined sugar preferences of ants (Ricks & Vinson 1970; Sudd & Sudd 1985; Vander Meer, Lofgren & Seawright 1995; Cornelius, Grace & Yates 1996; Koptur & Truong 1998; Völkl et al. 1999; Tinti & Nofre 2001). Lanza's work demonstrated that ants prefer some mixtures of sugar and amino acids over others or over pure sugar solutions; mixtures that mimic extrafloral nectar after herbivore attack are particularly attractive (Lanza & Krauss 1984; Lanza 1988, 1991; Lanza et al. 1993). Most studies have focused on one or two ant species in isolated experimental situations (but see Koptur & Truong 1998; Kay 2002). However, tropical ant communities at natural nectar sources are highly diverse (Schemske 1982) and typically involve many species that co-occur on the same plant (Blüthgen et al. 2000). Competition between ants can be intense and asymmetrical, resulting in hierarchical communities with competitively superior and inferior species. Dominance hierarchies were found in ant communities from the temperate zone (Fellers 1987; Savolainen & Vepsäläinen 1988) as well as tropical ones (Majer 1976; Jackson 1984; Majer 1993; Dejean & Corbara 2003), including the ant community in tropical northern Queensland where this study was conducted (Hölldobler 1983; Majer et al. 2001; Blüthgen & Fiedler 2002). However, very little is known about the importance of interspecific competition or resource partitioning in nectar use by ants (Blüthgen et al. 2000; Hossaert-McKey et al. 2001).

The goal of this study was to examine preferences of ants for sugars and amino acids by experiments with artificial nectar solutions using a multispecies approach in the ants’ natural environment. First, we asked whether ant species differ in their gustatory preferences and which sugar and amino acid characteristics are subject to interspecific variation. Secondly, we tested whether ant foraging selectivity changed due to previous consumption or through the presence of competing species.

Materials and methods

study site

This study was carried out in Cape Tribulation, Far North Queensland, Australia (16°07′ S, 145°27′ E, 80 m a.s.l). Study sites include the Australian Canopy Crane site, the property of the Environmental Research Station and adjacent forest areas. These sites comprise lowland rain forest characterized by a high abundance of lianas and an average canopy height of 25 m (Complex Mesophyll Vine Forest, Tracey 1982), secondarily reforested areas and beach forest. Forests were in an early stage of recovery from category 3 cyclone ‘Rona’ in February 1999 when large parts of the canopy had been severely damaged. Average rainfall is about 3500 mm per year, 60% of which occurs in the wet season between December and March Mean daily temperature ranges from 22 °C (July) to 28 °C (January) (Turton, Tapper & Soddell 1999).

experimental setup

Sugar and amino acid solutions (c. 1·7 mL) were offered in standard microcaps (2 mL Eppendorf). The solutions were available to insects through a cotton thread that functioned as a wick (each thread led from the base of the solution to the outside; lids were closed). Wicks outside the lid were c. 3 cm long and usually quickly soaked with solution near the lid or throughout their entire length. A similar setup was established earlier by Lanza and co-workers (Lanza 1988, 1991; Lanza et al. 1993). In our experiment, tubes were offered pairwise; each pair was tied together in upright position to a tree trunk at breast height with a plastic tape, wicks were pointing away from the trunk. The order of the two solutions was altered between neighbouring pairs. Selected trees were located at least 8 m apart from each other in order to minimize pseudoreplication by multiple testing of the same ant individuals or colonies. Two types of experiments were performed.

Experiment 1 was carried out in wide rain forest areas (13.01–04.03, 2001; 01–06.08, 2001). This design attempted to represent a large part of the forest and its characteristic ant community with a minimum of intracolony replication (at the cost of limited sample size for individual species and tests). A total of 663 trees were haphazardly selected along paths regardless of observed ant activity, and only one treatment pair per tree was installed. Paths were used repeatedly for different tests, but time lags between trials at each location were at least one week, and different trees were selected each time. Experimental trials were set up in the afternoon (c. 15.00–16.00). Ants were subsequently counted at each tube four to five times, including afternoon before dusk (c. 17.00), night (21.00), the following morning (10.00) and afternoon (24 h after the start of the experiment).

Experiment 2 aimed to provide a more detailed picture of preferences of colonies from selected species rather than representing the entire community. Experiments included secondary and beach forest besides mature rain forest (20.04–06.06, 2002). A total of 53 trees (separated by ≥ 8 m) was selected on which at least one common target ant species was observed to be active; these trees were used repeatedly for different tests. On each tree, 10 treatment pairs were installed in one to three vertical rows (design varied with trunk diameter and architecture) with at least 10-cm spaces between pairs. Experiments started at different times during the day; intervals between surveys were at least one hour and at least one survey was performed at night (total three to five surveys).

Different mixtures of amino acids or single amino acids were solved in sugar solutions. Compositions of the sugar-only controls and mixed solutions are given in Table 1. Amino acids in Experiment 1 had constant molarity (total 50 mmol/L amino acids in Mix A to Mix D; 100 mmol/L amino acids in Mix E, Mix F and all single amino acids). Experiment 2 was based on constant concentration (w/w): total sugars were always 15 g and total amino acids 1 g per 100 g solution in all single or mixed amino acid solutions. Mix A and Mix B each contained five amino acids of light and heavy molecular weights in similar proportions; Mix C was composed of five light and Mix D of five heavier amino acids. Mix E contained all 10 amino acids as before; Mix F was a commercially available product for human muscle training. Mix G mimicked the amino acid composition of a typical honeydew sample as consumed by Oecophylla smaragdina (Fabricius) ants in the study area (excreted by membracids on Caesalpinia traceyi lianas; Blüthgen & Fiedler 2002). Mix H was a mixture of the same amino acids as in Mix G but in equal concentrations. All sugars used were pure D(+)-isomers, except D(–)-fructose and sucrose (commercial white cane sugar). Solutions of trisaccharides (as well as their respective sucrose controls) were presented openly inside the open lids of the tubes, because they were insufficiently dissolved in the threads; these experiments were limited to 5 h well before re-crystallization of trisaccharides became visible (although variable hygroscopic effects cannot be excluded).

Table 1.  Composition of sugar-only control and mixed amino acid solutions used in preference tests
Substance + Abbrev.Solution:Experiment 1 (mmol/L)Experiment 2 (mg/g)
SugarMix AMix BMix CMix DMix EMix F2SucrMix GMix H
  • 1

    All amino acids offered as pure l-isomers except Met, Phe, Ser and Val (available in mixed DL-form, but only as l-form in Mix F).

  • 2

    Commercial powder (‘Muscle Gain’, Musashi Ltd, Australia); composition as labelled on package; also contains 7·7 g glucosamine HCl and 2·6 g sugar-based ‘lemon flavour’ per 100 g powder.

 GlucoseGluc500500500500500500500     –     –
 FructoseFruc500500500500500500500     –     –
Amino acids1
 AlanineAla     –     –  10  10     –  10     –      0·3    0·7
 ArginineArg     –  10     –     –  10  10    6     0·9    0·7
 AsparagineAsn     –     –  10     –  10  10     –      0·1    0·7
 CysteineCys     –     –     –     –     –     –     –       –     –
 Glutamic AcidGlu     – 10     –     –  10  10     –      0·5    0·7
 Glutamine      –     –     –     –     –     –    8·1      –     –
 GlycineGly     –  10     –  10     –  10  12     0·1    0·7
 HistidineHis     –     –     –     –     –     –    7·4     4·2    0·7
 IsoleucineIle     –     –     –     –     –     –    7·3      –     –
 LeucineLeu     –     –  10  10     –  10    9·3     0·8    0·7
 LysineLys     –     –     –     –     –     –  10      –     –
 MethionineMet     –  10     –     –  10  10    1·2     0·1    0·7
 PhenylalaninePhe     –     –     –     –     –     –    2·6     0·8    0·7
 ProlinePro     –     –     –     –     –     –     –      0·2    0·7
 SerineSer     –  10     –  10     –  10     –      0·8    0·7
 Taurine      –     –     –     –     –     –    8·6      –     –
 ThreonineThr     –     –     –     –     –     –    4·8     0·1    0·7
 TyrosineTyr     –     –  10     –  10  10    3·6     0·6    0·7
 ValineVal     –     –  10  10     –  10    7·3     0·7    0·7

A pilot study was performed on preferences for 15 synchronously offered amino acid solutions by the two dominant ant species (three colonies of Anonychomyrma gilberti (Forel), five of Oecophylla smaragdina). Pairs of amino acids in the final experiments were deliberately chosen in order to represent large interspecific variation in preferences as indicated by the preliminary study. For two solution pairs, experiments were repeated after five days for three colonies each of A. gilberti and O. smaragdina in order to test changes in preferences. Prior to the second experiments, these colonies were given access to large amounts (> 100 mL) of a 4% (w/w) solution of the single amino acids previously preferred in the pairwise tests (serine in the first, leucine in the latter species) for over 48 h. Note that all amino acid solutions mentioned before contained sugar like natural nectar sources. Two additional tests were performed using only asparagine or phenylalanine (2% w/w) in water (pairwise tests against water as a control). In addition to tests of variable composition, we tested four concentration levels of sucrose (5%, 20%, 35%, 50% w/w) simultaneously, using the same methods like above except that only five replicates were installed per tree.

Total sugar concentration (°Brix) was checked before and after selected experiments with a handheld refractometer (Eclipse, Bellingham & Stanley, Tunbridge Wells, UK). As expected, droplets taken from the threads after the experiment (24 h) had a higher sugar concentration than the solution as filled into the vial before the trial, but the increase from initial 15% (w/w) solutions was small (1·4% ± 1·6% (w/w), n= 41) and did not vary significantly between different sugars (sucrose, glucose and fructose, maltose; anova: F2,38 = 1·7, P= 0·19). Furthermore, enzymatic activity on threads (possibly from various sources including ant and microbial activity) may be effective, but had a limited impact on composition during the duration of the experiment. Due to invertase activity, 15% sucrose solutions had a median of 0·3% (w/w) glucose (n = 10; > 2% in two cases) on threads and only 0·1% (w/w) in the vials (n = 5) after the experiment (measured with glucose indicator paper; Glucostix, Bayer Sydney).

data analysis

Ant workers that attended the wick of each solution were counted separately for each ant species (A) and vial pair replicate (V) during each survey (S). Each count on the wick of the first solution was denoted as XAVS, on the second wick as YAVS, and their difference as DAVS= YAVS– XAVS. The mean number and difference during all valid surveys was obtained (denoted as XAV, YAV and DAV, respectively). The preference of the ant species (A) for one solution over another was denoted as DA and obtained as the mean DAV across all nA vial pair replicates where this ant occurred (thus d.f. =nA−1). Data from surveys where the less attended tube was empty or lacking the wick were excluded prior to analysis (a few wicks were bitten off by various animals).

For each solution pair, two hypotheses were tested: first, does one solution receive greater visitation than the other? This was tested for each ant species and resource combination with sufficient sample size (nA ≥ 5) using paired t-tests on all paired XAV vs. YAV. Secondly, do ant species differ in their preferences? An analysis of covariance (ancova) was applied to detect interspecific variation of DA between ant species (where nA ≥ 5). The mean number of ants per vial pair and survey was taken as covariate to control for variation in absolute recruitment and crowdedness at the baits. This method was conservative in all cases (F-values were greater when covariates were not included; data not shown). All statistics were limited to the 23 most common ant species from both experiments (including all species that occurred at least on 10 vial pairs in Experiment 1 or 25 vial pairs in Experiment 2).

Surveys where (1) the respective ant species was the only arthropod species on the respective vial pair were distinguished from (2) cases where different species attended the vial pair simultaneously. Henceforth, these cases are referred to as single occurrences (1) vs. co-occurrences (2), respectively. For above preferences in Experiment 2 (but not for Experiment 1), each DAV was calculated only from single occurrences. A two-way ancova was performed to investigate differences between preferences in the absence of competitors, calculating DA only from (1), vs. bait selectivity in co-occurrence situations, DA from (2). This analysis was restricted to solution pairs where preferences of most ant species were largely equivocal; species with odd preferences (different sign of DA from (1)) and the rarely co-occurring A. gilberti were excluded from this comparison to avoid a bias through their fixed preference. As above, the inclusion of the covariate (mean number of ants per vial pair) led to smaller effects.

In order to test changes in foraging choices during the duration of the experiment, we compared DAVS for the first survey where this ant was present (if this was before the third general survey) with DAVS from the final survey where both vials had remaining solution using paired t-tests. As for changes due to co-occurrence, these tests were restricted to species with equivocal preferences (but including A. gilberti) and the same selection of solution pairs.

Statistics were performed using Statistica 5·5 software package (Statsoft Inc., Tulsa, USA). Sequential Bonferroni corrections followed Hochberg's (1988) procedure.


Fifty-one ant species were recorded on the experimental solutions (including three similar nocturnal Camponotus species that were not always collected for later identification; these species were pooled in the following analyses). For preference analyses below, the 23 most common ants were selected (Table 2). These focal ant assemblages accounted for 90·8% and 98·5% of all ant visits recorded in Experiments 1 and 2, respectively. Some species were highly abundant and occurred on more than 10% of the experimental trees (Pheidole platypus, Camponotus‘nocturnal’, Paratrechina vaga). Pheidologeton affinis was relatively uncommon, but recruited hundreds of workers to vials in several cases and covered them completely with soil particles; they showed the same behaviour towards other baits such as fruits and meat placed on the ground. Because ant numbers were difficult to quantify in this species, it was excluded from the preference analyses below. Arthropods other than ants made up only 15·4% vs. 2·8% of the visits (including Aranae, Blattodea, Chilopoda, Coleoptera, Collembola, Diptera, Ensifera, Isopoda and Opiliones).

Table 2.  Most common ant species on sugar and amino acid solutions
Ants (Formicidae)Abbreviation (used in Figures)Subfamilyn
  • 1

    Includes two species from C. novae-hollandiae group and C. (Colobopsis) macrocephalus (Erichson). Ant subfamilies: D = Dolichoderinae, F = Formicinae, M = Myrmicinae, P = Ponerinae. N = number of experimental trees visited.

Anonychomyrma gilberti (Forel)AnoD  67
Camponotus vitreus (Smith)Cam.vitF  22
C. ‘nocturnal’ (3 spp.)1Cam.nocF128
C. sp6 (gasseri gp.)Cam.6F    6
Crematogaster cf. fusca SmithCre.fusM  54
C. cf. pythia ForelCre.pytM  20
C. sp3Cre.3M    4
Echinopla australis ForelEchF    6
Leptomyrmex unicolor EmeryLepD  10
Monomorium floricola ForelMonM  45
Oecophylla smaragdina (Fabricius)OecF  53
Paratrechina vaga (Forel)ParF  92
P. minutula (Forel)Par.minF  23
Pheidole platypus CrawleyPhe.plaM186
P. cf. athertonensis ForelPhe.athM    8
Pheidologeton affinis (Jerdon)PhgM  10
Polyrhachis foreli KohoutPol.forF  32
Rhoptromyrmex wroughtonii ForelRhoM  14
Rhytidoponera spoliata (Emery)RhyP  25
Tapinoma melanocephalum (Fabricius)TapD  37
Technomyrmex albipes (Smith)TecD  67
Tetramorium insolens F. SmithTem.insM  15
T. validiusculum EmeryTem.valM  50

sugar and amino acid preferences

In Experiment 1 (Fig. 1), none of the ant species showed significant choices between sugar solutions with amino acid mixtures of similar molecular weight (Fig. 1a), or between sugar plus single amino acids vs. sugar only (Fig. 1f–h), except for a significant discrimination against asparagine and serine by Pheidole platypus (Fig. 1e, h). Interspecific variability was not significant in these cases (ancova, Table 3). However, interspecific differences between heavy and light amino acid mixtures and rich amino acid mixtures (Mix E, Mix F) vs. pure sugar solutions were pronounced; the latter two remained significant after sequential Bonferroni correction (Table 3). Some ants significantly preferred light amino acids over heavy ones (Fig. 1b) and pure sugar over Mix E (Fig. 1c) (Crematogaster cf. fusca, P. platypus) or vice versa (Technomyrmex albipes). The more complex amino acid mixture (Mix F) was particularly favoured by Oecophylla smaragdina (Fig. 1d).

Figure 1.

Amino acid preferences of ant species on broad community level (Experiment 1). Preference values are the mean difference of mean visitation on solution (2) minus that on solution (1). On the left side of the dotted line (indicating no preference), solution (1) is preferred, on the right side solution (2). Composition of solution mixtures given in Table 1 and Methods (amino acids always in addition to sugar). Bars indicate ± SEM; sample sizes after each bar are number of vial pairs; those species shown only where n > 1. Significant preferences indicated by asterisks (*P < 0·05, **P < 0·01, ***P < 0·001) according to paired t-tests (applied to all species where n≥ 5). Variation between listed ant species with n≥ 5 was tested using ancova (Table 3).

Table 3. ancova results for interspecific variation in amino acid and sugar preferences (see Figs 1–3). Significance levels indicated by asterisks as * P < 0·05, ** P < 0·01, *** P < 0·001. Significant effects after sequential Bonferroni correction in bold type
Solution (1): (2)ancova
Mix A: Mix B  F4,68   =  0·4
Mix C: Mix DF11,159   =  2·3*
Sugar: Mix E  F9,87   =   4·1***
Sugar: Mix F  F5,45   =   4·6**
Sugar: Asn  F4,65   =  1·1
Sugar: Gly  F6,69   =  0·5
Sugar: Ser  F2,15   =  3·0
Sucrose: Mix GF12,373   =   4·0***
Sucrose: Mix H  F3,73   =  1·3
Mix G: Mix H  F4,79   =  0·5
Mix G: His  F4,121   =  1·6
Mix G: Phe  F5,39   =  1·8
Ala: Pro  F2,50   =  0·1
Arg: Glu  F2,44   =  3·7*
Asn: Phe  F2,78   =   22·1***
His: Cys  F2,55   =   38·1***
Leu: Gly  F5,91   =   15·1***
Met: Val  F4,55   =   5·5***
Ser: Thr  F4,50   =  2·4
Sucrose: Gluc + FrucF13,299   =   3·5***
Sucrose: GlucoseF11,229   =  1·3
Glucose: FructoseF11,248   =   5·8***
Sucrose: Xylose  F5,113   =  1·4
Sucrose: MaltoseF10,146   =   4·9***
Sucrose: Raffinose  F9,151   =  1·8
Sucrose: Melezitose  F8,162   =   4·3***

In Experiment 2, ants either showed a preference for sucrose plus amino acid mixtures over pure sucrose, or they were non-selective (Fig. 2a,b). For the complex honeydew mimic (Mix G) and Mix H, several species showed high and significant discrimination. Ant species did not discriminate significantly (Fig. 2c) between the honeydew mimic (Mix G) and the mixture of the same amino acids in equal proportions (Mix H) (hence both solutions were pooled in the analyses of changes in bait selection below). The original honeydew source that functioned as a model for Mix G was attended by O. smaragdina (Blüthgen & Fiedler 2002). This ant species (and Monomorium floricola) significantly preferred Mix G over a solution of its single most abundant amino acid, histidine (Fig. 2d). However, O. smaragdina did not discriminate between one of its most preferred single amino acids, phenylalanine (see Fig. 2h), and the honeydew mimic (Fig. 2e; only Crematogaster cf. pythia preferred Mix G). Significant interspecific differences with contrary choices of some species were found for all pairs of single amino acids (Fig. 2g–l) except alanine vs. proline (Fig. 2f). For example, O. smaragdina preferred phenylalanine over asparagine, leucine over glycine and methionine over valine, while A. gilberti showed the opposite preferences for these pairs. The latter species also preferred alanine over proline, histidine over cysteine, and serine over threonine, while O. smaragdina did not discriminate significantly between these pairs.

Figure 2.

Amino acid preferences of ant species from selected colonies (Experiment 2). See legend of Fig. 1 for details.

Ants did not show any preferences for amino acids when offered without sugar. Solutions of asparagine in water were not significantly preferred over pure water (mean ± SEM number of ants; water: 0·5 ± 0·1, Asn: 1·2 ± 0·5, t23 = 1·6, P= 0·12). Furthermore, phenylalanine in water received a similar visitation as the control (water: 0·8 ± 0·1, Phe: 0·8 ± 0·2, t26 = 0·2, P= 0·83). Both amino acids were mainly offered to colonies of ant species that showed a high preference for the same substance when solved in sugar (O. smaragdina: Phe; A. gilberti: Asn; see Fig. 2).

Ant responses to different sugar solutions (Fig. 3) were more similar than they were to solutions containing amino acids. Many ant species significantly preferred sucrose over glucose and fructose (Fig. 3a), or over glucose alone (Fig. 3b), other ants were non-selective among these sugars. Glucose was frequently preferred over fructose (Fig. 3c), although A. gilberti had the opposite preference. Xylose was barely consumed by ants and significantly less attractive than sucrose (Fig. 3d). The disaccharide maltose (Fig. 3e) and the trisaccharides raffinose (Fig. 3f) and melezitose (Fig. 3g) were significantly less attractive to many ant species, although some species showed no preferences here. Some ants (especially A. gilberti) only consumed considerable amounts of melezitose after sucrose controls had been completely emptied.

Figure 3.

Sugar preferences of ant species (Experiment 2). See legend of Fig. 1 for details.

Ants generally favoured higher sugar concentrations over lower ones; most species showed a consistent increase in visitation with sucrose concentration (Fig. 4, including all species with n > 5). Ant attendance varied significantly between concentrations (ancova, Table 4), and the concentration effect differed between species (significant interaction term). In general, ants discriminated all four concentration levels offered (Tukey's test: all P < 0·01), although individual species may be less selective, e.g. Crematogaster spp. did not differentiate between the highest levels (35% vs. 50%).

Figure 4.

Visitation of nine ant species among four levels of sucrose concentration (mean number of ants per vial ± SEM).

Table 4.  Two-way repeated-measures ancova results for preferences of nine ant species among four sugar concentrations
Species  8181  0·8        0·77
Concentration  354674·3< 0·0001
Species × concentration2454616·8< 0·0001

conditional changes in bait selection

For eight solution pairs, differences between ant visits on the two solutions were more pronounced during the final survey compared to the first survey where each ant species occurred (paired t-tests, Table 5). However, this increase in selectivity was relatively small for many solutions and only significant for the discrimination against xylose and melezitose after sequential Bonferroni correction. A small and non-significant decrease in sucrose preferences over glucose and fructose was found.

Table 5.  Changes in bait choices during the experiment. Mean differences of ant visits between two solutions (D, see Methods) compared between the first survey (Dfirst) and final survey (Dfinal) (paired t-tests). Significance levels *P < 0·05, **P < 0·01, ***P < 0·001. Significant effects after sequential Bonferroni in bold type
Solution (1) : (2)Preferences Dfirst: Dfinalt-test
Sucrose : Mix G/H      2·9 : 4·1t191   =  2·6*
Mix G : His −0·6 : − 1·0    t71   =  0·9
Mix G : Phe −0·5 : − 1·2    t36   =  1·9
Sucrose : Gluc + fruc −2·0 : − 1·5t144   =  0·9
Sucrose : Glucose −2·8 : − 2·7t128   =  0·2
Glucose : Fructose −1·1 : − 1·5    t87   =  1·2
Sucrose : Xylose −4·7 : − 8·7    t42   =   3·2**
Sucrose : Maltose −3·0 : − 3·7    t49   =  1·2
Sucrose : Raffinose −2·7 : − 3·7    t32   =  1·4
Sucrose : Melezitose −1·3 : − 2·7    t63   =   3·4**

Preferences between amino acids changed after ants were fed large amounts of one of the compounds for over 2 days. A. gilberti initially preferred serine over threonine (Fig. 2). After access to serine for two days, the direction of the preference remained unchanged, but its extent was significantly reduced (mean D± SEM before and after surplus feeding: Dbefore=−5·2 ± 0·7, n= 27; Dafter=−2·5 ± 0·6, n= 24; ancova: F1,48 = 7·6, P < 0·01). Similarly, O. smaragdina preferred leucine over glycine (Fig. 2); preferences diminished after feeding extensively on leucine, but the effect was only marginally significant (Dbefore=−0·7 ± 0·3, n= 30; Dafter= 0·3 ± 0·3, n= 11; ancova: F1,38 = 3·5, P= 0·07).

Co-occurrences between ants often affected each ant's choice between alternative solutions (Table 6; note that the rarely co-occurring A. gilberti and those ant species that showed contrary choices from most species were a priori excluded from each comparison, and tests did not include solution pairs with highly variable preferences, see Figs 1–3). In all cases, overall selectivity between solutions tended to decrease when other ant species were present, thus ant visits were more evenly distributed between the alternatives. This trend was significant after Bonferroni correction for preferences between mixed amino acid solutions/sucrose, sucrose/glucose and sucrose/glucose + fructose.

Table 6.  Changes in bait selectivity through simultaneous co-occurrence of different ant species on the same vial pair. Sample size (n) and mean differences of ant visits between two solutions (D) ± SEM given for single occurrence (n1, D1) and co-occurrence (n2, D2) (ancova). Significance levels *P < 0·05, **P < 0·01, ***P < 0·001. Significant results following sequential Bonferroni in bold type
Solution (1): (2)n1n2D1D2ancova
Sucrose : Mix G/H350225    2·5 ± 0·2    1·5 ± 0·2F1,572   =   7·7**
Mix G : His105136 −1·3 ± 0·3 −0·5 ± 0·3F1,238   =  2·4
Mix G : Phe  46  50 −0·7 ± 0·3 −0·5 ± 0·3    F1,93   =  0·2
Sucrose : Glucose + fructose287141 −2·0 ± 0·2 −0·5 ± 0·2F1,425   =   13·0***
Sucrose : Glucose220149 −1·9 ± 0·3 −0·3 ± 0·1F1,366   =   9·7**
Glucose : Fructose235127 −1·0 ± 0·1 −0·5 ± 0·2F1,359   =  6·8**
Sucrose : Xylose100  45 −2·7 ± 0·3 −2·2 ± 0·4F1,142   =  0·6
Sucrose : Maltose151  67 −2·5 ± 0·3 −1·5 ± 0·5F1,215   =  6·4*
Sucrose : Raffinose140  78 −2·5 ± 0·4 −1·2 ± 0·2F1,215   =  4·2*
Sucrose : Melezitose154  42 −1·2 ± 0·2 −0·6 ± 0·7F1,193   =  1·7


sugar preferences

Ant species showed similar carbohydrate preferences and also consistently preferred higher concentrated sugar solutions over lower ones. Ranking of the three main nectar sugars was concordant across many ant species: sucrose > glucose > fructose. The remaining species did not show significant preferences among these sugars, and only A. gilberti preferred fructose over glucose. Preferences of sucrose over both monosaccharides, variable preferences among glucose and fructose, or non-significant preferences were also reported in other studies on ants (Ricks & Vinson 1970; Vander Meer et al. 1995; Cornelius et al. 1996; Völkl et al. 1999; Tinti & Nofre 2001; but see Koptur & Truong 1998). Invertase is widespread among ants and allows them to digest sucrose (Ayre 1967; Ricks & Vinson 1972). Xylose was not attractive to any of the ants (Vander Meer et al. 1995), and maltose was usually less attractive than sucrose (see also Cornelius et al. 1996; Tinti & Nofre 2001; but Vander Meer et al. 1995). Trisaccharides (raffinose and melezitose) that are common in honeydews in our study site (Blüthgen & Fiedler, in press) and elsewhere (Auclair 1963; Völkl et al. 1999) were significantly less attractive than sucrose for most ants, although some species did not show a significant discrimination. For instance, O. smaragdina was equally attracted to sucrose and melezitose, while A. gilberti significantly preferred sucrose. Both species commonly fed on honeydew. The most common coccid and membracid honeydew sources of O. smaragdina at the study site (Blüthgen & Fiedler 2002) were dominated by melezitose and sucrose (Blüthgen & Fiedler, in press). In contrast, cicadellid honeydew frequently consumed by A. gilberti lacked melezitose and was dominated by fructose, raffinose and melibiose (Blüthgen & Fiedler, in press). Our findings are not concordant with preferences for melezitose and raffinose over sucrose reported by three studies on European Lasius niger ants (Duckett 1974; Völkl et al. 1999; Tinti & Nofre 2001), but they support a study of three tropical ant species where no significant preferences for melezitose over sucrose were found (Cornelius et al. 1996). Melezitose has low phagostimulatory effect or nutritional value to other insects (Wäckers 1999) and can even be toxic (Zoebelein 1956). Besides important osmotic functions for the homopterans, trisaccharides may thus reduce the suitability of honeydew for homopteran parasitoids as host signal or food source (Wäckers 2000). Our results obtained with a broad taxonomic and ecological range of ant species suggest that trisaccharides in honeydew have no general ant-attracting role as proposed earlier (Kiss 1981). Rather certain honeydew-foraging ants secondarily evolved the ability to tolerate or assimilate these sugars (e.g. O. smaragdina). The ant's enzymatic or microbial equipment could be a constraint on their ability to effectively exploit honeydew sources, and typically only a fraction of the nectarivorous ant community attends homopteran aggregations (Davidson 1997; Blüthgen et al. 2000). Apart from behavioural and physiological constraints, however, restricted honeydew foraging may result from active competitive exclusion through territorial dominant ants (Hölldobler 1983; Jackson 1984; Dejean & Corbara 2003), since a broad spectrum of nectarivorous ant genera can be at least qualitatively categorized as trophobiotic (see Fiedler 2001).

amino acid preferences

In contrast to carbohydrates, interspecific variability in preferences among sugar solutions with different single amino acids was much more pronounced. Nevertheless, most ants preferred solutions containing mixtures of amino acids over sugar alone, with few notable exceptions of contrary or nonpreferences (e.g. Pheidole or Crematogaster species in some cases). These results correspond with previous studies where complex nectar mimics were used (Lanza 1988, 1991), while more simple combinations of amino acids were often less attractive than sugar-only controls (Lanza & Krauss 1984). Ant species were found to differ in their preference for amino acid mixtures in the field (Lanza 1988; Lanza et al. 1993) or under laboratory conditions (Lanza 1991). Currently, we have only limited information how discriminations of single amino acids translate into preferences for more complex mixtures typical for natural honeydew and nectar sources, although such mechanisms would be crucial for our understanding of plant–animal interactions. Certain ‘key’ compounds may be as attractive to a foraging ant as a complex mixture (such as phenylalanine in O. smaragdina). Moreover, our observation that preferences for complex amino acid mixtures over sugar-only solutions were relatively uniform across ant species in our study, suggests that positive effects of the amino acid mixture outweigh potentially repellent functions of some single amino acids (but see Lanza & Krauss 1984). Interestingly, sugars are important in the preference of amino acid solutions, as sugar-free amino acid solutions were not found to be attractive to ants in our study. Other studies also reported that several amino acid solutions without sugar were not accepted by ants (Ricks & Vinson 1970), although such effects may vary between species (see Kay 2002). Synergetic effects between sugars and amino acids may be common and were recently detected for glucose and glycine in the response of ant taste receptors, where the amino acid may enhance the sweetness perception (Wada et al. 2001).

variability and conditionality

Differences in amino acid preferences may help to explain field observations of nectar and honeydew source partitioning in the ant community (Schemske 1982; Blüthgen et al. 2000; Apple & Feener 2001; Hossaert-McKey et al. 2001). Multiple factors potentially cause different dietary choices in ants, although little is known about their importance. For instance, physiological causes may be distinguished from ecological factors, and both may interact. Physiological factors may involve species-specificity in taste reception or digestive systems (Ayre 1967; Davidson 1997). Ecological variability may be found for energy budgets and nutrient requirements. Since most ants are omnivores and nectar is rarely or never their sole diet (Stradling 1978), the need for nitrogen may vary with the relative importance and complementary composition of other sources. Consequently, ant species that commonly collect nectar and honeydew may exhibit higher preferences for sources rich in proteins or amino acids (Kay 2002), which may reflect the nitrogen limitation of their diet (Yanoviak & Kaspari 2000). Requirements for certain compounds may also vary due to the ants’ highly specific chemical communication or defence system (Hölldobler & Wilson 1990).

Our results suggest that both physiological and environmental factors may be important for dietary choices. After ant colonies have been feeding large amounts of a preferred amino acid solution for two days, its attractiveness was significantly reduced. During shorter time intervals in the experiments, none of the preferences was altered rather than strengthened. Therefore, habituation, previous experience or changes in colony requirements may be involved in foraging decisions. Choices among sugar concentrations and between sucrose vs. amino acid solutions may change according to colony demand and resource availability (Sudd & Sudd 1985; Cassill & Tschinkel 1999; Kay 2002), and our experiments indicate that such processes may even apply to single substances. Moreover, the ants’ selectivity for one solution over another as expressed in the absence of other species was often significantly reduced when two or more species co-occurred on the same baits. This observation indicated an effective influence of interspecific competition on dietary choices (see also Blüthgen & Fiedler, in press). The effect of interspecific competition was independent of increased intra- and interspecific ant crowding at the baits (we controlled for crowding by using the number of ants as a covariate in the analysis). Conditional effects on nectar preferences have been rarely demonstrated previously. Studies on three hummingbird species showed that competition and experience influenced foraging decisions among artificial feeders with variable sugar concentration (Pimm et al. 1985; Sandlin 2000). Moreover, in temperate-zone ants, choices between fish and syrup baits were affected by asymmetric interspecific competition (Savolainen & Vepsäläinen 1988).

In conclusion, gustatory preferences for amino acids (and to a lesser extent for carbohydrates) vary substantially among different ant species and may be linked to nectar and honeydew resource partitioning. Dietary choices are conditional with respect to resource availability and active competition. Such processes may be fundamental to the territorial mosaic-like distribution of dominant and submissive ants that are common to many tropical ecosystems (Dejean & Corbara 2003).


We are indebted to Mascha Bischoff and Silke Engels for their intensive help in the field. Identification of many of the ant species was provided by Brian Heterick, Rudy Kohout, Hanna Reichel and Steve Shattuck. The Australian Canopy Crane company (ACC) and the Cape Tribulation Environmental Research Station (Austrop) provided logistical support. Special thanks are due to Nigel Stork (ACC, Rainforest CRC) and Hugh Spencer (Austrop). We thank Janet Lanza for stimulating discussions, Klaus Fischer and two anonymous referees for helpful comments on earlier versions of the manuscript. Financial support was kindly provided by the Deutsche Forschungsgemeinschaft (Fi 547/9–1) and by a doctoral fellowship of the Studienstiftung des deutschen Volkes to N.B.