• diet;
  • foraging strategy;
  • morphological traits;
  • phylogenetically independent contrasts;
  • seed size


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendices
  • 1
    We analyse diet and propagule selection by the harvester ant Messor barbarus (L.) in Mediterranean grassland and scrubland in central Spain.
  • 2
    Diet was estimated by the identification of worker-transported prey in 34 colonies per system type, and compared with seed and fruit availability in the foraging areas. Propagules were characterized by six morphological traits: total weight; weight of seed content; the three main dimensions; and shape.
  • 3
    The effect of propagule attributes on selectivity was analysed after transforming data into phylogenetically independent contrasts. Propagules from a small number of species dominate the diet of M. barbarus in the study area, in terms of both frequency and contribution in seed weight. In grassland, prey selection depends on ln(prey length) and ln(prey weight) (R2 = 0·57). In scrubland, ln(prey length) explains 64% of selection. Long and heavy propagules are preferred.
  • 4
    This pattern of selection can be a mere effect of a time-saving foraging strategy, as apparent preference for long propagules can be expected even if workers forage in a non-selective way.
  • 5
    Messor ants are likely to play a role in the plant composition of Mediterranean grassland and scrubland, limiting the abundance of long propagules and thus indirectly favouring small-seeded species without dispersal appendages.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendices

Messor harvester ants are the main seed predators in therophyte grassland and certain types of scrubland of the Mediterranean Basin (López, Acosta & Serrano 1993; Cerdá & Retana 1994; Hensen 2002; Azcárate & Peco 2003). Both types of system show a high proportion of annual species (Peco 1989), which makes them particularly sensitive to ant–seed interactions. Foraging by ants can constitute a severe source of seed mortality for many species, limiting their recruitment odds (Andrew 1986; Louda 1989). In some cases, however, ants behave as vectors of dyszoochory (accidental seed dispersal), abandoning viable seeds on trails and refuse piles (Retana, Picó & Rodrigo 2004), where levels of plant competition and soil properties are altered by ant activity (Dean & Yeaton 1993a, 1993b).

The relevance of interactions between Messor ants and plants depends primarily on rates of seed or fruit removal. Previous studies show that harvester ants tend to concentrate their diet on a relatively small number of propagule species (Crist & Wiens 1994; Andersen, Azcárate & Cowie 2000; Wilby & Shachak 2000; Willott, Compton & Incoll 2000), which reveals the existence of traits that favour collection by ants. Prey-selection criteria can vary to some extent (Fewell & Harrison 1991; Crist & MacMahon 1992; Reyes-López & Fernández-Haeger 2002a, 2002b). However, certain morphological traits such as weight (Baroni-Urbani & Nielsen 1990; Baroni-Urbani 1992; Milton & Dean 1993; Detrain & Pasteels 2000); size (Rissing 1981; Campbell 1982; Crist & MacMahon 1992; Willott et al. 2000); shape (Pulliam & Brand 1975); or the possession of awns and other appendages (Schöning et al. 2004) explain a part of prey selection in harvester ants. Other non-morphological traits, such as nutritional or calorific content (Kelrick et al. 1986), chemical composition (Pizo & Oliveira 2000) and viability (Andrew 1986; Crist et al. 1992), can also influence preference by ants.

Seed predation can promote evolutionary changes on the design and reproductive strategies of plants (Harper, Lovell & Moore 1970; Louda 1989). Some authors have suggested that certain propagule morphologies can be favoured in communities where seed predation by ants is intense (Detrain & Pasteels 2000; Willott et al. 2000; Schöning et al. 2004). Likewise, plants can develop ant-attractive propagules in communities where myrmecochory or dyszoochory are important (Hughes & Westoby 1992). As Messor harvester ants are abundant in Mediterranean grassland and scrubland, it is plausible that plants composing these communities show traits or mechanisms capable of reducing seed harvesting, or at least seed consumption.

There are few data on seed selection by Messor barbarus (L.) in Mediterranean grassland. Detrain et al. (2000) analysed one single colony of M. barbarus in south-eastern France. According to these authors, ants collect seeds from a small number of species, and apparently prefer weighed ones (>0·4 mg). However, prey selection can vary between colonies (Traniello & Beshers 1991), and seed or fruit weight can be correlated with other traits not included in the study (Sánchez et al. 2002). There are no published studies on Mediterranean scrubland, although in semiarid scrubland of the Iberian Peninsula Willott et al. (2000) found a strong preference by Messor bouvieri for large seeds. Some research into seed selection by Messor spp. has also been conducted under laboratory conditions or using artificial seeds (Baroni-Urbani & Nielsen 1990; Baroni-Urbani 1992; Reyes-López & Fernández-Haeger 2002a, 2002b) and manipulating propagules from a single species (Schöning et al. 2004).

The aim of this work is to analyse diet and propagule selection by M. barbarus in Mediterranean grassland and scrubland under natural conditions. The study addresses two specific questions: (1) which propagule types comprise the diet of M. barbarus in Mediterranean grassland and scrubland?; (2) are propagule attributes good predictors of selection by harvester ants?


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendices

study site

Fieldwork was conducted in the Moncalvillo-Pedrezuela study site (2000 ha; 40°38′ N, 3°70′ E; 900 m height; 20 km north of Madrid). Climate is continentalized mediterranean, with a severe drought period in summer, 550 mm mean annual rainfall, and a mean annual temperature of around 13 °C. Soils are shallow, on siliceous substrata (mainly pre-Ordovicic gneiss), and vegetation is dominated by two types of clearly differentiated system: grassland and scrubland. Grasslands are the product of extensive livestock grazing over the past few centuries, and are dominated by therophyte species. Scrubland areas are located on abandoned grassland and dry farming croplands, and are dominated by Lavandula stoechas subsp. pedunculata. For a more detailed description of the site, see Azcárate & Peco 2003).

Messor barbarus is the main harvester ant in both types of system (Azcárate & Peco 2003). This species is very common in dry grassland and open scrubland of the western Mediterranean and northern Africa. Colonies make conspicuous trunk trails for food searching, collection and transport (López et al. 1993; Reyes-López & Fernández-Haeger 2001). Messor bouvieri is also an important harvester ant in the scrubland of the area. There are other seed-eating species in the study site belonging to the genera Messor, Aphaenogaster, Oxyopomyrmex, Goniomma, Tetramorium and Pheidole, although their abundance and harvesting capacity is much lower than those of M. barbarus, the only exception being M. bouvieri, which is common in the scrubland areas (Azcárate et al. 2003).


The diet of M. barbarus was estimated by the identification of worker-transported prey close to the nest entrance (Davidson 1980; Hahn & Maschwitz 1985; Gordon 1993; Milton & Dean 1993; Cerdá & Retana 1994; Detrain et al. 2000; Wilby & Shachak 2000; Willott et al. 2000). Sampling took place between March and November 1997. We took 34 observations per system type, each from one independent colony. Minimum distance between sampled colonies was 20 m. In each observation, we took the first 41 prey items brought to the nest hole by the ants, which required a gathering time of 2–3 min. Previous data about M. barbarus recorded in the study site show that summer (June–August) comprises 60% of the annual harvesting activity; spring (March–May), 30%; and autumn (September–November), 10% (Azcárate 2003), so sampling was distributed according to the relative contribution of each season. In spring we recorded 12 observations in grassland and 11 in scrubland; in summer, 19 and 20, respectively; and in autumn, three in both system types.

Prey was classified into five categories: plant propagules, other vegetal fragments, animal fragments, lichen fragments and mineral fragments. Differences in diet composition between system types were analysed comparing the arc-sin-transformed frequencies of each group following a t-test approach. We performed one test per prey group, and then a sequential Bonferroni correction for multiple tests (Rice 1989).

Only propagules were identified to specific level. We also distinguished between ‘seeds’ (seeds or single-seeded fruits in their simplest form, following Bekker et al. 1998) and ‘fruits’ (more complex or multiseeded propagules).

propagule availability

For every diet observation, we identified the main foraging area and set a circular 6 m diameter plot, where we placed 10 sampling points at random. Messor barbarus foragers are able to dig up shallow prey, and also to cut mature propagules from vegetation. Hence, in order to measure propagule availability, we extracted a 3 mm wide × 4 cm diameter cylindrical soil core per sampling point, along with the seed/fruit content of the standing herbaceous vegetation of the same area. The 10 subsamples belonging to the same plot were pooled for the laboratory analysis.

Prior to the evaluation of the propagule content, we dispersed the samples using a solution of 20 g sodium hexametaphosphate [(NaPO3)6] and 10 g sodium bicarbonate (NaHCO3) for 1 l water. Samples were kept in contact with dispersing solution for 2 h at a ratio of 100 ml for 7 g soil. They were then passed through a series of three sieves (2, 1 and 0·5 mm wide). After a preliminary analysis of diet data, we concluded that ants never collect prey <0·5 mm wide, thus the material crossing the third sieve was refused. Finally the three fractions resulting from the sieving process were scanned under a binocular microscope. In most cases seeds and fruits were identified at the specific level, with the aid of reference collections made with specimens from the study site. Those propagules showing any doubt about their viability (empty or broken ones) were rejected.

This procedure was adequate to estimate the availability of seeds produced by herbaceous vegetation, but did not include predispersed seeds of woody species. Therefore in scrubland plots we also counted the total number of predispersed Halimium capsules and Lavandula infructescences per plot. The availability of Lavandula fruits (calyces +1–4 mericarps) was estimated by collecting 30 infrutescences per plot. Data from the two techniques were pooled and transformed into average densities of each propagule type per dm2. As we were unaware of the origin of prey carried by foragers, we did not distinguished between predispersed and dispersed items in the propagule availability data set.

morphological traits of propagules

We selected the following morphological traits to describe each propagule type: total weight; weight of seed content; the three main dimensions (dim1, length; dim2, width; dim3, thickness); and shape (the variance of the three main dimensions, first divided by length; Thompson, Band & Hodgson 1993). In the case of seeds, we used the mean values published by Azcárate et al. (2002) and Sánchez et al. (2002). For fruits, we collected 30 units of each propagule type, then estimated the mean value of each morphological trait following the same procedure as in Azcárate et al. (2002). To assess the mean weight of the seed content, we first estimated the mean number of seeds per fruit, then multiplied this value by the mean weight of a single seed. The values assigned to each propagule type are shown in Appendix 1.

propagule selection

We evaluated selection of each propagule type by comparing its contribution in weight in the diet and availability data sets. The use of weight rather than frequency is a closer approximation of the actual importance of each prey type as a food resource. Other authors have also employed weight to evaluate prey preference in harvester ants (Kelrick et al. 1986; Reyes-López & Fernández-Haeger 2002a).

For each propagule type, i, we calculated the following selectivity index, Si:

  • Si = Nri/Ni

where Nri = number of observations in which the relative contribution in weight of propagule i is higher in diet that in availability; and Ni = total number of observations with information for the propagule i (present in diet and/or availability data sets). We calculated selectivity indices only when Ni ≥ 10. The minimum value of the index (Si = 0) means that the propagule i always occurs in diet in a proportion lower than in availability, or that it was found only in availability samples. In contrast, the maximum value (Si = 1) reflects that, in all cases in which the propagule i is present in the diet, its relative contribution in weight is higher than in the correspondent availability samples. The limit between positive and negative selection is at S = 0·5. The existence of some propagule types detected only in the diet data set impeded the use of some well known selectivity indices (Chesson 1983; Milton et al. 1993; Detrain et al. 2000).

The effect of propagule attributes on selectivity index was analysed by fitting multiple regression models. The selection indices, and all propagule traits except shape, were log-transformed to achieve normality of residuals. As individual species cannot be regarded as independent data points (Felsenstein 1985; Harvey et al. 1995; Harvey 1996; Martins & Hansen 1996), we obtained phylogenetically independent contrasts (PICs), following Felsenstein (1985). Accurate phylogeny information was not available, so we used current taxonomy to infer phylogeny (Appendix 2), as suggested by Martins & Hansen 1996).

In the analysis we included only those propagule types occurring in diet or availability data sets at least in 10 observations. In species showing more than one propagule type, we selected one at random.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendices


More than 91% of prey collected by M. barbarus in grassland are fruits or seeds (Fig. 1). We found 43 different propagule types, corresponding to 35 plant species (Appendix 3). In scrubland, plant propagules were also the main group of prey, although the frequency was significantly lower than in scrubland (77·7%, t66 = 4·4; P < 0·001; Fig. 1). In this type of system, ant diet included 51 different types belonging to 40 plant species (Appendix 3). Ants also collected a considerable amount of other plant fragments, particularly in scrubland (18·2%vs 6·1% in grassland; t66 = 3·5; P < 0·001; Fig. 1); a few animal remains (insects and excrement; >1% in both system types); some mineral particles in scrubland (2·4%); and, more rarely, lichen fragments.


Figure 1. Diet of Messor barbarus in grassland and scrubland. Mean frequency of the five groups of prey types considered for the study. Bars, standard deviations; N = 34 for both system types. Results of comparison t-test between grassland and scrubland are shown on top of bars. ns, P < 0·05.

Download figure to PowerPoint

Fig. 2 shows the main seeds and fruits of the propagule subgroup. In grassland the two most frequent prey types are the caryopsis (plus lemma and palea) of Vulpia muralis (28·7%) and the capsule (plus tepals) of Juncus bufonius (16·8%). In scrubland the commonest prey species is L. stoechas, which shows three propagule types: infrutescences (2·2%), fruits (19·1%), and mericarps (12·5%). Legumes of Ornithopus compressus are also frequent in the diet of scrubland ants (12·9%). The relative importance of each propagule type varies when considering its contribution in seed weight. From this point of view, the main prey in grassland is J. bufonius capsules (18·6%); Anthemis arvensis heads (14·0%); V. muralis caryopsis (8·5%); and Lotus hispidus seeds (8·5%). In scrubland, O. compressus contributes 24·1% of the total seed weight (legumes 20·1%, mericarps 4%). The three propagule types of L. stoechas together reach 28·4% (infructescences 10·7%, fruits 9·8%, mericarps 7·9%). Appendix 3 shows data for all the propagules found in the study.


Figure 2. Diet of Messor barbarus in grassland and scrubland. Mean frequency and contribution in seed weight of the main propagules collected by ants. Antarv, Anthemis arvensis; Hypgla, Hypochoeris glabra; Junbuf, Juncus bufonius; Lavsto, Lavandula stoechas; Leotar, Leontodon taraxacoides; Lothis, Lotus hispidus; Orncom, Ornithopus compressus; Rumace, Rumex acetosella; Triglo, Trifolium glomeratum; Vulmur, Vulpia muralis; Xolgut, Xolantha guttata. ac, Achene; cap, capitulum; caps, capsule; car, caryopsis; cyp, cypsela; f, fruit; inf, infrutescence; leg, legume; mer, mericarp; s, seed. Percentages refer to the plant propagule group.

Download figure to PowerPoint

propagule selection

Availability sampling found 79 propagule types in grassland and 71 in scrubland. We obtained selectivity indices for only 18 types in grassland and 22 in scrubland (Table 1), as the other types occurred in fewer than 10 observations of diet/availability (Appendix 3).

Table 1.  Selection indices shown by propagule types present in at least 10 observations of diet/availability
SpeciesPropagule typeGrasslandScrubland
  1. l/p, lemma and palea.

Agrostis castellanaCaryopsis + l/p0·00
Anthemis arvensisCypsella0·230·00
Anthoxanthum aristatumCaryopsis + l/p + fl0·18
Aphanes microcarpaAchene0·00
Asterolinon linum-stellatumSeed0·00
Cerastium spp.Seed0·00
Coronilla repanda ssp. duraMericarp0·13
Halimium umbellatumCapsule0·18
Hypochoeris glabraCypsella0·400·09
Juncus bufoniusCapsule + tepals0·86
Lavandula st. ssp. pedunculataMericarps + calyx0·54
Leontodon taraxacoidesCypsella0·550·23
Lotus hispidusSeed0·48
Moenchia erectaCapsule + calyx0·00
Montia fontanaSeed0·070·00
Myosotis spp.Mericarp0·00
Ornithopus compressusLegume0·93
Poa annuaCaryopsis + l/p0·00
Poa bulbosaCaryopsis + l/p0·000·09
Rumex acetosella ssp. angiocarpusAchene0·31
Scirpus setaceusAchene0·00
Silene scabrifloraSeed0·00
Spergula arvensisSeed0·000·00
Spergula pentandraSeed0·000·00
Tolpis barbataCypsella0·20
Trifolium arvenseSeed0·25
Trifolium campestreSeed0·170·21
Trifolium glomeratumSeed0·04
Vulpia ciliataCaryopsis + l/p0·17
Vulpia muralisCaryopsis + l/p0·740·39
Xolantha guttataCapsule0·30

In grasslands the preferred propagule types were J. bufonius capsules, Leontodon taraxacoides cypsellas and V. muralis caryopses. Most types present low or even null selectivity indices. After log-transformation and assessment of PICs, the normal distribution fitted the variable (K–S d = 0·21; P > 0·20). The existence of non-dichotomous nodes in the phylogenetic structure (Appendix 2) reduced the number of available contrasts to 14. All Pearson's correlation coefficients between ln(S + 1) and morphological variables were positive, and significant (P < 0·05) for ln(weight) (r12 = 0·61); ln(length) (r12 = 0·60); and ln(dim3) (r12 = 0·56). The best regression model (R2 = 0·57; F2,11 = 7·27; P = 0·01) for ln(S + 1) included two morphological variables: ln(weight) (β = 0·094; t12 = 2·29; P = 0·042) and ln(length) (β = 0·098; t12 = 2·26; P = 0·045).

In scrubland, the higher selection indices were shown by O. compressus legumes and L. stoechas calices (Table 1). We obtained 19 PICs from the 22 available propagule types. As in grassland, ln(weight) and ln(length) positively correlate with ln(S + 1) (r17 = 0·53 and r17 = 0·80, respectively). Fig. 3 shows a scatter plot for the best regression model (R2 = 0·64; F1,17 = 29·93; P < 0·001) found for ln(S + 1), which included ln(length) as a single dependent variable (β = 0·114; t17 = 5·471; P < 0·001).


Figure 3. Scrubland, propagule selection by M. barbarus. Phylogenetically independent contrasts in ln(S + 1) against contrasts in ln(Dim1). S, Selectivity index. Dim1 given in mm. N = 19 contrasts.

Download figure to PowerPoint


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendices


Seeds and fruits are the basis of the diet of M. barbarus in central Spain, which coincides with other observations for this genus (Hahn & Maschwitz 1985; Cerdáet al. 1994; Detrain et al. 2000). However, in scrubland ants collected a lower proportion of seed material, and took a relatively high amount of other plant fragments, which very often consisted of leaves and small branches belonging to Lavandula and Halimium. We lack data to explain this finding, which probably requires research on the chemical composition of the collected fragments, and more information on their use by the ants. Differences in seed availability between system types may also be involved. As in most studies of harvester ants (Hahn et al. 1985; Hobbs 1985; Cerdáet al. 1994; Vorster, Hewitt & van der Westhuizen 1994; Detrain et al. 2000), we found a small proportion of animal fragments in the diet, which are normally interpreted as a complementary protein source. The sporadic collection of soil particles could be explained by the presence of resin exudates on them (Fernández-Escudero & Tinaut 1993).

Most propagules collected by M. barbarus belong to a small number of species, in agreement with previous research into harvester ants (Crist & Wiens 1994; Andersen et al. 2000; Detrain et al. 2000; Wilby & Shachak 2000; Willott et al. 2000). In each system type, only two plant species comprise around half of the propagules taken by ants. However, some of the most frequent types (e.g. V. muralis caryopsis) show relatively small contributions when analysed in terms of weight of seed content. In other words, in at least some cases ants apparently focus their foraging effort on lowly profitable prey.

prey selection

Our results reveal a clear association between prey selection and prey length in both system types. The model estimated for grassland includes a second variable, prey weight, a trait that has been related to ant preference in previous research (Baroni-Urbani 1992; Detrain et al. 2000).

Selection for heavy prey cannot be justified on the basis of energy yield, as the variable ‘weight of seed content’, more closely related to the propagule calorific value, does not correlate with selectivity in either of the two system types. In fact, heavy propagules normally coincide with those possessing appendages for seed dispersal, which are later refused by ants and discharged into the chaff piles. Outstanding examples are caryopsis of V. muralis, fruits of L. stoechas, legumes of O. compressus, or capsules of J. bufonius. In the latter case we observed that some capsules transported by ants were actually empty of seeds.

In harvester ants, the balance between energy costs and benefits of resource collection appears to be positive, even for very small prey (Fewell 1988; Baroni-Urbani & Nielsen 1990). In this case the foraging strategy would reduce time costs rather than enhance the energy benefit per prey (Morehead & Feener 1998; Reyes-López & Fernández-Haeger 2001). Shorter trip times should increase the amount of prey carried into the nest, and reduce exposure to predators and desiccation risk. In fact, preference for long propagules can be explained as a mere effect of a time-saving foraging strategy. Regardless of their calorific content, long propagules are more easily detectable, and tend to be buried more slowly, than short ones (Peart 1984; Thompson et al. 1993), which makes them more accessible to ants. If so, an apparent preference for long propagules is expected even if workers forage in a non-selective way. Propagule enlargement can be a consequence of the possession of awns, pappus or other dispersal appendages. The acquisition of this type of prey suggests an increase in processing time inside the nest, although this type of task is likely to be less limited by environmental conditions or predation risk.

effects of m. barbarus on vegetation

Detrain et al. (2000) and Willott et al. (2000) have suggested that Messor ants could favour small-seeded species in communities where they are abundant. According to our data, risk of predation by M. barbarus in Mediterranean grassland and scrubland is higher for long propagules, and this can indirectly promote the occurrence of small-seeded species without dispersal appendages. Previous research shows that this type of species is over-represented in Mediterranean grasslands (Azcárate et al. 2002).

However, we lack evidence in favour of a hypothetical shortening of propagules in response to M. barbarus activity. Production of small seeds has been related to endozoochory (Malo & Suárez 1995) or mediterranean climate conditions (Azcárate et al. 2002), and can also be a simple consequence of allometric or phylogenetic constraints (Herrera 1992; Kang & Primack 1999; Guerrero-Campo & Fitter 2001). In addition, we lack detailed information on the phenotypic variability of each species; inheritance of seed attributes; or strength of the above-mentioned factors as selective pressures, making a discussion that is consistent in evolutionary terms difficult.

Still, M. barbarus ants are likely to play a role in the plant composition of Mediterranean grassland and scrubland. Both types of community have been modelled by human management (Joffre, Rambal & Ratter 1999), and thus can be considered young systems whose plant species composition has recently been selected from the regional flora. In this context, it seems plausible that M. barbarus activity limits the abundance of long propagules in Mediterranean grassland and scrubland, and thus indirectly favours small-seeded species without dispersal appendages.

Although most large-seeded species in Mediterranean grasslands are scarce (Azcárate et al. 2002), there are some remarkable exceptions. For instance, long-seeded species such as V. muralis and L. taraxacoides are widespread in Mediterranean grasslands, despite the fact that their propagules are consistently selected by M. barbarus. Research into the fate of these seeds is necessary, as it should not be assumed that all prey brought into nests by Messor ants are consumed (Retana et al. 2004). Mutualistic interactions such as dyszoochory could contribute to the success of certain long-seeded species in communities where Messor ants are abundant.

In scrubland, the risk of seed predation is more heterogeneous, favouring the existence of refuge habitats that are safe from the action of harvester ants (Azcárate et al. 2003). This could explain the high occurrence of largely predated species, such as O. compressus and L. stoechas. For the latter species, evidence also suggests the existence of ant–plant mutualistic interactions (unpublished data). The effects of other harvester ants, such as M. bouvieri (Azcárate et al. 2003), whose selection criteria have not been studied, may overlap with those of M. barbarus, making the final vegetation patterns more complex in scrubland than in grassland.

In summary, propagule morphological traits (length and total weight) are involved in selection by M. barbarus ants in both Mediterranean grassland and scrubland. While the evolutionary consequences of this behaviour are difficult to derive, it is likely that ants affect plant species composition in these system types by limiting the abundance of long propagules. More research is needed in order to corroborate these predictions, and to answer some questions that remain open. From the ants’ point of view, our data are consistent with a hypothetical relationship between collection and detectability, although the behavioural interpretation of the selection criteria is still incomplete. From the plants’ side, we need to understand the mechanisms that allow the success of certain long-seeded species in habitats where their propagules are collected in quantity by ants. In this sense, the effects of collateral ant–plant mutualistic interactions should not be discarded.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendices

We are grateful to C. Levassor for her help in the field work. The study was supported by the Spanish Ministry of Science and Technology (projects AMB 990382 and REN 2003-01562), the Spanish Ministry of Education and Culture (FPI scholarships to F.M.A. and A.M.S), and the Madrid Regional Government (FPI scholarship to L.A.).


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendices
  • Andersen, A.N., Azcárate, F.M. & Cowie, I.D. (2000) Seed selection by an exceptionally rich community of harvester ants in the Australian seasonal tropics. Journal of Animal Ecology 69, 975984.
  • Andrew, M.H. (1986) Granivory of the annual grass Sorghum intrans by the harvester ant Meranoplus sp. in tropical Australia. Biotropica 18, 344349.
  • Azcárate, F.M. (2003) Hormigas granívoras en pastizales y matorrales mediterráneos: interacciones y efectos sobre la vegetación. PhD thesis, Universidad Autónoma de Madrid, Madrid.
  • Azcárate, F.M. & Peco, B. (2003) Spatial patterns of seed predation by harvester ants (Messor Forel) in Mediterranean grassland and scrubland. Insectes Sociaux 50, 120126.
  • Azcárate, F.M., Sánchez, A.M., Arqueros, L. & Peco, B. (2002) Abundance and habitat segregation in Mediterranean grassland species: the importance of seed weight. Journal of Vegetation Science 13, 159166.
  • Baroni-Urbani, C. (1992) Factors affecting seed preference among Old World harvester ants of the genus Messor. Ethology Ecology & Evolution 2, 7380.
  • Baroni-Urbani, C. & Nielsen, M.G. (1990) Energetics and foraging behaviour of the European seed harvesting ant Messor capitatus. II. Do ants optimize their harvesting? Physiological Entomology 15, 449461.
  • Bekker, R.M., Bakker, J.P., Grandin, U. et al. (1998) Seed size, shape and vertical distribution in the soil: indicators of seed longevity. Functional Ecology 12, 834842.
  • Campbell, M.H. (1982) Restricting losses of aerially sown seeds due to seed-harvesting ants. Ant–Plant Interactions in Australia (ed. R.C.Buckley), pp. 2530. W. Junk, Dordrecht, the Netherlands.
  • Castroviejo, S. (ed.) (19862003) Flora Iberica. Real Jardín Botánico, CSIC, Madrid.
  • Cerdá, X. & Retana, J. (1994) Food exploitation patterns of two sympatric seed-harvesting ants Messor bouvieri (Bond.) and Messor capitatus (Latr.) (Hym. Formicidae) from Spain. Journal of Applied Entomology 117, 268277.
  • Chesson, J. (1983) The estimation and analysis of preference and its relationship to foraging models. Ecology 64, 12971304.
  • Crist, T.O. & MacMahon, J.A. (1992) Harvester ant foraging and shrub–steppe seeds: interactions of seed resources and seed use. Ecology 73, 16781779.
  • Crist, T.O. & Wiens, J.A. (1994) Scale effects of vegetation on forager movement and seed harvesting by ants. Oikos 69, 3746.
  • Davidson, D.W. (1980) Some consequences of diffuse competition in a desert ant community. American Naturalist 116, 92105.
  • Dean, W.R.J. & Yeaton, R.I. (1993a) The effects of harvester ant Messor capensis nest-mounds on the physical and chemical properties of soils in the southern Karoo, South Africa. Journal of Arid Environments 25, 249260.
  • Dean, W.R.J. & Yeaton, R.I. (1993b) The influence of harvester ant Messor capensis nest-mounds on the productivity and distribution of some plant species in the southern Karoo, South Africa. Vegetatio 106, 2135.
  • Detrain, C. & Pasteels, J.M. (2000) Seed preferences of the harvester ant Messor barbarus in a Mediterranean mosaic grassland (Hymenoptera: Formicidae). Sociobiology 35, 3548.
  • Felsenstein, J. (1985) Phylogenies and the comparative method. American Naturalist 125, 115.
  • Fernández-Escudero, I. & Tinaut, A. (1993) Alimentación no granívora en Messor bouvieri Bondroit, 1918 y Messor barbarus (L. 1767) (Hymenoptera: Formicidae). Boletin de la Asociacion Espanola de Entomologia 17, 247254.
  • Fewell, J.H. (1988) Variation in foraging patterns of the western harvester ant Pogonomyrmex occidentalis in relation to variation in habitat structure. Interindividual Behavioural Variability in Social Insects (ed. R. L.Jeanne), pp. 257282. Westview Press, Boulder, CO, USA.
  • Fewell, J.H. & Harrison, J.F. (1991) Flexible seed selection by individual harvester ants, Pogonomyrmex occidentalis. Behavioral Ecological Sociobiology 28, 377384.
  • Gordon, D.M. (1993) The spatial scale of seed collection by harvester ants. Oecologia 95, 479487.
  • Guerrero-Campo, J. & Fitter, A.H. (2001) Relationships between root characteristics and seed size in two contrasting floras. Acta Oecologica 22, 7785.
  • Hahn, M. & Maschwitz, U. (1985) Foraging strategies and recruitment behaviour in the European harvester ant Messor rufitarsis (F.). Oecologia 68, 4551.
  • Harper, J.L., Lovell, P.H. & Moore, K.G. (1970) The shapes and sizes of seeds. Annual Review of Ecology and Systematics 1, 327357.
  • Harvey, P.H. (1996) Phylogenies for ecologists. Journal of Animal Ecology 65, 255263.
  • Harvey, P.H., Read, A.F. & Nee, S. (1995) Why ecologists need to be phylogenetically challenged. Journal of Ecology 83, 535536.
  • Hensen, I. (2002) Seed predation by ants in south-eastern Spain (Desierto de Tabernas, Almería). Anales de Biología 24, 8996.
  • Herrera, C.M. (1992) Interspecific variation in fruit shape. Allometry, phylogeny and adaptation to dispersal agents. Ecology 73, 18321841.
  • Hobbs, R.J. (1985) Harvester ant foraging and plant species distribution in annual grassland. Oecologia 67, 519523.
  • Hughes, L. & Westoby, M. (1992) Effects of diaspore characteristics on removal of seeds adapted for dispersal by ants. Ecology 73, 13001312.
  • Joffre, R., Rambal, S. & Ratter, J.P. (1999) The dehesa system of southern Spain and Portugal as a natural ecosystem mimic. Agroforestry Systems 45, 5759.
  • Kang, H. & Primack, R.B. (1999) Evolutionary change in seed size among some legume species: the effects of phylogeny. Plant Systematic Evolution 219, 151164.
  • Kelrick, M.I., MacMahon, J.A., Parmenter, R.R. & Sisson, D.V. (1986) Native seed preferences of shrub–steppe rodents, birds and ants: the relationships of seed attributes and seed use. Oecologia 68, 327337.
  • López, F., Acosta, F.J. & Serrano, J.M. (1993) Responses of the trunk routes of a harvester ant to plant density. Oecologia 93, 109113.
  • Louda, S.M. (1989) Predation in the dynamics of seed regeneration. Ecology of Soil Seed Banks (eds M.A.Leck, V.T.Parker & R. L.Simpson), pp. 2551. Academic Press, New York.
  • Malo, J.E. & Suárez, F. (1995) Herbivorous mammals as seed dispersers in a Mediterranean dehesa. Oecologia 104, 246255.
  • Martins, E.P. & Hansen, T.F. (1996) The statistical analysis of interspecific data: a review and evaluation of phylogenetic comparative methods. Phylogenies and the Comparative Method in Animal Behavior (ed. E.P.Martins), pp. 2275. Oxford University Press, Oxford, UK.
  • Milton, S.J. & Dean, W.R.J. (1993) Selection of seeds by harvester ants (Messor capensis) in relation to condition of arid rangeland. Journal of Arid Environments 24, 6374.
  • Morehead, S.A. & Feener, D.H. (1998) Foraging behavior and morphology: seed selection in the harvester ant genus Pogonomyrmex. Oecologia 114, 548555.
  • Peart, M.H. (1984) The effects of morphology, orientation and position of grass diaspores on seedling survival. Journal of Ecology 72, 437453.
  • Peco, B. (1989) Modelling Mediterranean pasture dynamics. Vegetatio 83, 269276.
  • Pizo, M.A. & Oliveira, P.S. (2000) The use of fruits and seeds by ants in the Atlantic forest of southeast Brazil. Biotropica 32, 851861.
  • Pulliam, H.R. & Brand, M.R. (1975) The production and utilization of seeds in plains grasslands of southeastern Arizona. Ecology 56, 11581166.
  • Retana, J., Picó, F.X. & Rodrigo, A. (2004) Dual role of harvesting ants as seed predators and dispersers of a non-myrmechorous Mediterranean perennial herb. Oikos 105, 377386.
  • Reyes-López, J.L. & Fernández-Haeger, J. (2001) Some factors determining size-matching in the harvester ant Messor barbarus: food type, transfer activity, recruitment rate and size-range. Insectes Sociaux 48, 118124.
  • Reyes-López, J.L. & Fernández-Haeger, J. (2002a) Food storage in the nest and seed selectivity in the harvester ant Messor barbarus (Hymenoptera: Formicidae). Sociobiology 39, 123128.
  • Reyes-López, J.L. & Fernández-Haeger, J. (2002b) Composition-dependent and density-dependent seed removal rates in the harvester ant Messor barbarus. Sociobiology 39, 110.
  • Rice, W.R. (1989) Analyzing tables of statistical tests. Evolution 43, 223225.
  • Rissing, S.W. (1981) Foraging specializations of individual seed-harvester ants. Behavioral Ecology and Sociobiology 9, 149152.
  • Sánchez, A.M., Azcárate, F.M., Arqueros, L. & Peco, B. (2002) Volume y dimensiones como predictores del peso de semilla de especies herbáceas del Centro de la península Ibérica. Anales del Jardín Botánico de Madrid 59, 249262.
  • Schöning, C., Espadaler, X., Hensen, I. & Roces, F. (2004) Seed predation of the tussock-grass Stipa tenacissima L. by ants (Messor spp.) in south-eastern Spain: the adaptive value of trypanocarpy. Journal of Arid Environments 56, 4361.
  • Thompson, K., Band, S.R. & Hodgson, J.G. (1993) Seed size and shape predict persistence in soil. Functional Ecology 7, 236241.
  • Traniello, J.F.A. & Beshers, S.N. (1991) Polymorphism and size-pairing in the harvester ant Pogonomyrmex badius. A test of the ecological release hypothesis. Insectes Sociaux 38, 121127.
  • Tutin, T.G., Heywood, V.H., Burges, N.A., Valentine, D.H., Walters, S.M. & Webb, D.A., eds (196480) Flora Europaea. Cambridge University Press, Cambridge, UK.
  • Vorster, H., Hewitt, P.H. & Van Der Westhuizen, M.C. (1994) The effect of seed foraging and utilization by the granivorous ant Messor capensis (Mayr) (Hymenoptera: Formicidae) on the grassveld composition of the Central Orange Free State. African Entomology 2, 175179.
  • Wilby, A. & Shachak, M. (2000) Harvester ant response to spatial and temporal heterogeneity in seed availability: pattern in the process of granivory. Oecologia 125, 495503.
  • Willott, S.J., Compton, S.G. & Incoll, L.D. (2000) Foraging, food selection and worker size in the seed harvesting ant Messor bouvieri. Oecologia 125, 3544.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Appendices

Appendix 1

Table A1.  Morphological traits of the propagule types used to estimate regression models
SpeciesPropagule typeNo. seedsTotal weight (mg)Seed content (mg)Dim1 (mm)Dim2 (mm)Dim3 (mm)Shape
  1. l/p, Lemma and palea; No. seeds, average number of seeds per propagule type.

Agrostis castellanaCaryopsis + l/p 1 0·38 0·26 2·300·960·960·11
Anthemis arvensisCypsella 1 0·72 0·72 1·971·011·010·08
Anthoxanthum aristatumSpikelet 1 0·26 0·23 6·783·970·860·19
Aphanes microcarpaAchene 1 0·19 0·19 0·730·500·250·11
Asterolinon linum-stellatumSeed 1 0·29 0·29 1·050·860·430·09
Cerastium sp.Seed 1 0·05 0·05 0·570·500·160·15
Coronilla repanda ssp. duraMericarp 1 1·10 0·60 3·601·051·050·17
Halimium umbellatum
ssp. viscosumCapsule 616·67 8·57 5·874·074·070·03
Hypochoeris glabraCypsella 1 0·78 0·78 9·103·613·610·12
Juncus bufoniusCapsule + tepals16 1·20 0·46 5·752·202·200·13
Lavandula stoechas
ssp. pedunculataCalyx + mericarps 0·75 1·40 0·68 6·503·003·000·10
Leontodon taraxacoides
ssp. longirrostrisCypsella 1 0·24 0·2410·323·443·440·15
Lotus hispidusSeed 1 1·60 1·60 0·850·810·500·05
Moenchia erectaCapsule + calyx 5 0·45 0·21 5·253·403·400·04
Montia fontanaSeed 1 0·08 0·08 0·890·890·450·08
Myosotis spp.Mericarp 1 0·13 0·13 1·060·690·630·06
Ornithopus compressusLegume 632·3114·8122·112·211·200·23
Poa annuaCaryopsis + l/p 1 0·21 0·15 3·101·101·000·15
Poa bulbosaCaryopsis + l/p 1 0·13 0·08 2·510·790·790·16
Rumex acetosella
ssp. angiocarpusAchene 1 0·36 0·36 1·240·880·880·03
Scirpus setaceusAchene 1 0·05 0·05 0·900·590·590·04
Silene scabrifloraSeed 1 0·32 0·32 0·700·630·530·02
Spergula arvensisSeed 1 0·16 0·16 0·870·870·650·02
Spergula pentandraSeed 1 0·14 0·14 1·981·980·300·24
Tolpis barbataCypsella 1 0·10 0·10 3·361·441·440·11
Trifolium arvenseSeed 1 0·29 0·29 0·980·730·730·02
Trifolium campestreSeed 1 0·25 0·25 1·270·830·420·11
Trifolium glomeratumSeed 1 0·45 0·45 1·541·281·280·01
Vulpia ciliataCaryopsis + l/p 1 0·19 0·1211·230·510·340·31
Vulpia muralisCaryopsis + l/p 1 0·19 0·1020·650·540·810·31
Xolantha guttataCapsule30 1·90 1·32 3·602·802·800·02

Appendix 2


Fig. A1. Taxonomic structure followed to infer the phylogeny of the species included in the analyses. See Azcárate et al. (2002) for references used to compose the taxonomic tree.

Appendix 3

Table A2.  Propagule types recorded in the samplings of diet and availability
SpeciesPropagule typeDietAvailability No. occurrences (n = 34) Diet and/or availability No. occurrences (n = 34)
Mean frequency (%)Contribution in seed weight (%)No. occurrences (n = 34)
Agrostis castellanaCaryopsis 4·23 2·17 5 2 7 211
Caryopsis + l/p 614 614
Aira caryophylleaCaryopsis + l/p 3 5 3 5
Alyssum granatenseSilicula 0·05 0·03 1 1
Andryala integrifoliaCypsela 0·06 0·00 1 5 9 5 9
Capitulum 0·14 0·48 1 2 3
Anthemis arvensisCypsela 0·88 0·63 415131813
Capitulum 4·05 1·9214·02 6·71 4 5 4 5
Anthoxanthum aristatumSpikelet 0·41 2·91 0·13 1·02 2 6 316 517
Aphanes microcarpaAchene31 631 6
Arrhenatherum albumCaryopsis + l/p 1·19 1·84 1 5 6
Asterolinon linum-stellatumSeed 115 115
Bromus hordeaceusCaryopsis + l/p 2 2
Bromus tectorumCaryopsis + l/p 0·20 0·24 1 1 1 1
Capsella bursa-pastorisSeed 0·93 0·16 1 2 3
Silicula 0·53 0·90 1 3 4
Carduus tenuiflorusCypsela 1 1
Carex divisaUtriculus 8 8
Cerastium ramosissimumCapsule + calyx 1·63 1·58 1 1
Cerastium semidecandrumCapsule + calyx 0·23 0·18 1 1
Cerastium sp.Seed19 919 9
Capsule + calyx 0·08 0·03 1 7 8
Coincya monensis ssp. orophilaSeed 9 9
Coronilla repanda ssp. duraMericarp 0·93 0·64 21616
Corynephorus canescensSpikelet 3 3
Corynephorus fasciculatusCaryopsis + l/p 4 4
Crepis capillarisCapitulum 1 1
Cynodon dactylonCaryopsis + l/p 2 2
Cytisus scopariusSeed 1·88 3·09 5 1 5
Echium vulgareMericarp 2 2
Erodium cicutariumMericarp 0·46 0·11 0·52 0·07 3 1 5 7 1
Erophila vernaSilicula 0·13 0·37 1 4 4
Euphorbia exiguaSeed 8 1 8 1
Festuca rothmaleriCaryopsis + l/p 0·05 0·10 1 3 4 3 4
Galium parisienseMericarp 5 1 5 1
Halimium umbellatum ssp. viscosumSeed 1·40 2·67 1 2 3 2 3
Capsule 0·91 2·36 41111
Holcus setiglumisSpikelet 1 5 1 5
Hymenocarpos lotoidesSeed 0·06 0·01 1 3 3 1
Mericarp 0·31 0·06 2 2
Hypochoeris glabraCypsela 5·88 1·39 5·60 0·5310 314231722
Capitulum 1 1
Jasione montanaCapsule 2·36 0·15 1 2 3
Juncus acutiflorusCapsule + tepals 3 3 3 3
Juncus bufoniusCapsule + tepals16·7318·58181521
Juncus capitatusCapsule + tepals 2 2
Lamium amplexicauleMericarp 1 2 1 2
Lavandula stoechas ssp. pedunculataMericarp12·53 7·9714 130 129
Calyx + mericarps19·09 9·80171926
Infrutescence 2·2410·68 8 8
Leontodon taraxacoides sp. longirrostrisCypsela 3·50 0·63 1·41 0·20 9 5 7121513
Capitulum 0·40 1·48 1 1 1 1
Logfia minimaCapitulum 3 3
Lotus hispidusSeed 4·32 1·77 8·53 2·48 6 2 7 111 3
Legume 0·20 1·77 1 1
Lupinus hispanicusSeed 0·28 1·24 1 1 1 1
Melica ciliataCaryopsis + l/p 0·09 0·01 1 1
Merendera pyrenaicaSeed 3 3
Mibora minimaSpikelet 0·11 0·01 1 1 1
Micropyrum tenellumCaryopsis + l/p 9 9
Moenchia erectaCapsule + calyx 0·05 0·16 0·00 0·04 1 114 214 2
Molineriella laevisCaryopsis + l/p 0·26 0·01 1 1
Montia fontanaSeed 1·61 0·04 0·27 0·00 5 128142814
Myosotis sp.Mericarp  611  611
Ornithopus compressusMericarp 0·11 1·29 0·45 4·01 1 2  516  516
Legume 0·06 7·72 1·4920·08 113 2  114
Ornithopus perpusillusMericarp 3 3
Parentucellia latifoliaCapsule  2  2
Petrorhagia nanteuiliiSeed 3·19 1·19 3 4 4
Capsele 0·21 0·41 2 3 3
Plantago lagopusCapsule 0·10 0·04 1  3  4
Head 0·24 1·86 1  1
Plantago lanceolataSeed  2 5  2 5
Capsule 1·32 1·49 4  2 4  2 7
Poa annuaCaryopsis + l/p 16 1 16 1
Spikelet 1·87 3·26 2  2
Poa bulbosaCaryopsis + l/p 0·04 0·00 1 1210 1211
Spikelet 1·56 1·64 2  1  3
Polycarpon tetraphyllumSeed  1  1
Rumex acetosella ssp. angiocarpusAchene 4·23 1·49 4  810  813
Sagina apetalaCapsule  1  1
Sanguisorba minorAchene  2  2
Scilla autumnalisSeed 0·64 1·14 1  1
Capsule 0·36 1·11 1  1
Scirpus setaceusAchene 20 2 20 2
Scleranthus delortiiAchene 0·23 0·10 1  3 4  3 4
Senecio jacobaeaCapitulum 0·24 2·12 1  1  2
Silene gallicaSeed 4 4
Capsule + calyx 0·30 0·44 2 2
Silene scabrifloraSeed  111  111
Capsule 2 2
Spergula arvensisSeed 2615 2615
Capsule  2  2
Spergula pentandraSeed 2120 2120
Spergularia purpureaCapsule  3  3
Teesdalia coronopifoliaSilicula 0·23 0·10 0·28 0·01 1 1  1 1
Thapsia villosaMericarp 0·10 0·08 1 2 3
Tolpis barbataCypsela 0·30 0·03 3 12 9 15 9
Trifolium arvenseSeed 0·43 0·10 31112
Achene + calyx 0·99 0·90 2  2 4  4 4
Trifolium campestre s.l.Seed 2·24 3·77 1·46 0·78 6 9 2823 3024
Trifolium campestreAchene + calyx 2·16 0·83 0·93 0·17 4 5  2 8  612
Head 0·40 0·29 2·28 1·41 1 3  1  2 3
Trifolium dubiumAchene + calyx 1511 1511
Head 0·13 0·87 1 1
Trifolium glomeratum s.l.Seed11·18 6·9616 34 3 34 3
Trifolium glomeratumAchene + calyx 1·82 1·10 4  9 11
Head  2  2
Trifolium striatumSeed 0·21 0·04 1  4 4  4 5
Achene + calyx 0·16 0·58 0·46 0·55 2 3  2 7  4 9
Trifolium strictumSeed  1  1
Trifolium subterraneumSeed 0·11 1·15 1  4  5
Veronica arvensisSeed  1
Veronica vernaSeed 1 1
Capsule 0·82 1·58 2  2  3
Vulpia ciliataCaryopsis + l/p 1·14 0·07 5  211  212
Vulpia muralisCaryopsis  6 9  6 9
Caryopsis + l/p28·60 3·93 8·55 0·4622 9 2516 2618
Spikelet 0·26 0·14 0·32 0·06 2 1  2 1
Xolantha guttataCapsule 0·31 9·15 0·85 7·86 420  717  923
Unknown  3·31 3·27 5·27 2·321114 1214
No. propagule types 435143514351 7971 9086
No. species 334033403340 6459 7267
  545454 80  90
l/p, Lemma and palea.
Nomenclature follows Castroviejo (1986–2003), except taxa yet to be covered, which follow Tutin et al. 1964–80).