Seed Preferences of Three Harvester Ants of the Genus Pogonomyrmex (Hymenoptera: Formicidae) in the Monte Desert: Are They Reﬂected in the Diet?

ECOLOGY AND POPULATION BIOLOGY Seed Preferences of Three Harvester Ants of the Genus Pogonomyrmex (Hymenoptera: Formicidae) in the Monte Desert: Are They Reﬂected in the Diet?

1,2 1 GABRIELA INE´ S PIRK AND JAVIER LOPEZ DE CASENAVE

Ann. Entomol. Soc. Am. 104(2): 212Ð220 (2011); DOI: 10.1603/AN10093 ABSTRACT Harvester ants play an important ecological role as seed consumers in arid areas. We performed choice experiments to study preferences of Pogonomyrmex rastratus (Mayr), Pogono- myrmex mendozanus (Cuezzo & Claver), and Pogonomyrmex inermis (Forel) (Hymenoptera: For- micidae) for seeds of six native species (three grasses, two forbs, and a shrub) in the central Monte desert, Argentina. We hypothesized that ant diet composition would reßect ant preferences. Thus, based on previous ant diet studies, we expected that 1) the three species would prefer grass to nongrass seeds, 2) P. inermis would have a lower preference for nongrass seeds than P. rastratus and P. mendozanus and 3) relative abundance of seeds in the diet would be positively associated with relative preference. In general, relative abundance of seeds in the diet was positively associated with relative preferences. Ants preferred grass seeds, but P. inermis did not have a lower preference for nongrass seeds. We also analyzed the relationship between preferences for seeds of the six species and their size and morphology, and we found higher preferences for seeds of intermediate size but no relationship with seed morphology. The overall match between seed preferences and diet composition could increase the chances of ants affecting the abundance and composition of some seed resources in the Monte desert, with important community implications.

RESUMEN Las hormigas cosechadoras de semillas cumplen un papel ecolo´gico de gran importancia en las zonas a´ridas. Realizamos experimentos para evaluar las preferencias de Pogonomyrmex rastratus (Mayr), Pogonomyrmex mendozanus (Cuezzo y Claver), y Pogonomyrmex inermis (Forel) (Hyme- noptera: Formicidae) por semillas de seis especies nativas (tres pastos, dos dicotiledońeas herbaćeas, y un arbusto) en el Monte central, Argentina. Nuestra hipo´tesis fue que la composicioń de la dieta de estas hormigas es el reßejo de sus preferencias. Por lo tanto, basańdonos en trabajos previos sobre la dieta de estas especies, esperamos que 1) las tres especies preÞrieran semillas de pastos sobre las dema´s semillas, 2) P. inermis presentara una menor preferencia por las semillas de dicotiledońeas herbaćeas y arbustos que P. rastratus y P. mendozanus, y 3) que la abundancia relativa de semillas en la dieta estuviera asociada positivamente con las preferencias. En general, la abundancia relativa de semillas en la dieta estuvo asociada positivamente con las preferencias. Las hormigas preÞrieron semillas de pastos, pero P. inermis no presento´ una menor preferencia por semillas de dicotiledońeas herbaćeas y arbustos. Analizamos tambień la relacioń entre las preferencias por las semillas de las seis especies estudiadas y su taman˜ o y morfologõá y encontramos una mayor preferencia por semillas de taman˜ os intermedios, pero ninguna asociacioń con la morfologõá. Finalmente, la coincidencia entre las preferencias y la dieta podrõá incrementar la probabilidad de que las hormigas afecten la abundancia y composicioń de de semillas en el desierto del Monte, con importantes implicancias para la comunidad.

KEY WORDS antÐseed interaction, choice experiments, desert communities, diet, granivory

Harvester ants are a common component of arid and ecological role as seed consumers (Ho¨ lldobler and semiarid ecosystems, where they play an important Wilson 1990, MacMahon et al. 2000). Several studies have shown that they can inßict severe seed losses, 1 Grupo de Investigacioń en Ecologõá de Comunidades de Desierto, with a greater impact on the more preferred species Departamento de Ecologõá, Gene´tica y Evolucioń, FCEyN, Univer- (Brown et al. 1979, Reichman 1979, Crist and Mac- sidad de Buenos Aires, Piso 4, Pab. 2, C. Universitaria, C1428EHA Mahon 1992). This could eventually alter the rela- Buenos Aires, Argentina. tive abundance of plant species, causing changes in 2 Correspondingauthor:LaboratorioEcotono,INIBIOMA,CONICET- Universidad Nacional del Comahue, Pasaje Gutie´rrez 1125, 8400 the structure of plant communities (Inouye et al. 1980, Bariloche, Rõó Negro, Argentina (e-mail: [email protected]). Samson et al. 1992, Espigares and Lo´pez-Pintor 2005).

0013-8746/11/0212Ð0220$04.00/0 ᭧ 2011 Entomological Society of America March 2011 PIRK AND LOPEZ DE CASENAVE:SEED PREFERENCES OF HARVESTER ANTS 213

Although harvester ants collect a wide range of diet composition of these ants mirrors their seed pref- seeds, most experimental studies have shown that they erences, we expected that 1) the three species would have marked preferences for certain species (Kelrick prefer grass to forb and shrub seeds; 2) P. inermis et al. 1986, Gross et al. 1991, Crist and MacMahon 1992, would have a lower preference for forb and shrub Reyes-Lo´pez and Fernańdez-Haeger 2002, Nicolai et seeds than P. rastratus and P. mendozanus; and 3) al. 2007). In many of these studies, grass seeds are relative abundance of the six offered species in the diet among the most preferred, although seeds of other of P. rastratus, P. mendozanus, and P. inermis should be families are sometimes chosen [e.g., Purshia tridentata positively associated with their relative preference. (Pursh) DC. [Rosaceae], Kelrick et al. 1986; Trifolium We also hypothesized that these ant preferences subterranean L. [Leguminosae], Reyes-Lo´pez and would be associated with seed size and morphology. Fernańdez-Haeger 2002]. Several seed traits have We expected 1) a positive association between ant been found to inßuence forager choices; among them, seed preferences and seed size and 2) a higher pref- nutritional quality (Kelrick et al. 1986, Crist and Mac- erence the more elongated the seeds. Mahon 1992), concentration of secondary compounds (Carroll and Janzen 1973, Whitford 1978, Davidson Materials and Methods 1982, Buckley 1982), size (Davidson 1977, Kelrick et al. 1986, Crist and MacMahon 1992, Willott et al. 2000, Study Area. The study took place at the Biosphere Heredia and Detrain 2005), and morphology (Pulliam Reserve of N˜ acun˜ ań (34Њ 03ЈS, 67Њ 54Ј W), located in and Brand 1975, Azca´rate et al. 2005). Regarding size, the central portion of the Monte desert. The main ant preferences have been positively correlated with habitat is the open woodland of Prosopis flexuosa DC. seed size in several studies (Kelrick et al. 1986, Crist where individuals of this species and of Geoffroea and MacMahon 1992), probably because of the higher decorticans (DC.) B. are scattered within a matrix of energy reward they offer. For seed morphology, there perennial tall shrubs (Ͼ1 m height, mostly Larrea is evidence that elongated seeds that bear conspicuous divaricata (DC.) Coville, but also Condalia micro- awns, hairs, and other projections are preferentially phylla Cav., Capparis atamisquea Kuntze, Atriplex harvested by several ant species (Pulliam and Brand lampa Gillies ex Moq., and Larrea cuneifolia Cav.), low 1975, Azca´rate et al. 2005) because they could be shrubs [Lycium spp., Junellia aspera (Gillies ex Hook), easier for ants to handle than rounded seeds with and Acantholippia seriphioides (A.Gray) Moldenke], smooth outlines. and perennial grasses [e.g., Trichloris crinita (Lag.) Pogonomyrmex rastratus (Mayr), Pogonomyrmex Parodi, Pappophorum spp., Sporobolus cryptandrus mendozanus (Cuezzo & Claver), and Pogonomyrmex (Torr.) Gray, Aristida spp., Digitaria californica inermis (Forel) (Hymenoptera: Formicidae) are (Benth.) Henr., Setaria leucopila (Scribn. & Merr.) K. three sympatric harvester ant species that occur in the Schum.]. Annual forb cover (e.g., Chenopodium papu- central Monte desert, Argentina. P. mendozanus (de- losum Moq., Phacelia artemisioides Griseb., Parthenium scribed in Cuezzo and Claver 2009) was referred to as hysterophorus L.) is highly variable from year to year. P. pronotalis in our previous publications based on The climate of N˜ acun˜ ań is dry with hot summers and Claver and Fowler (1993). All three species have small cold winters. Mean annual temperature is 15.9ЊC colonies of 300Ð1,100 individuals, with Ϸ70% adult (1972Ð2004) and mean annual rainfall is 333.5 mm workers (Nobu´ a Behrmann et al. 2010). Their activity (1972Ð2004), with high interannual variation. Seven- season spans from mid-spring to early autumn (Oc- ty-Þve percent of the annual rainfall occurs in spring toberÐApril), and they exhibit diurnal activity with and summer (OctoberÐMarch), and seed production temporal changes responding mainly to soil temper- of almost all plants is restricted to summer. For a atures (Pol and Lopez de Casenave 2004). Their diet complete description of the study area, see Lopez de (as determined by quantiÞcation of the type and abun- Casenave (2001). dance of seeds brought to the nest by foragers) con- Sampling Design. Seed preferences of P. rastratus, sists mainly of grass seeds and only P. rastratus and P. P. mendozanus, and P. inermis were assessed in the mendozanus occasionally harvest forb and shrub seeds Þeld by performing choice experiments under a paired (Pirk et al. 2004, 2009; Pirk and Lopez de Casenave comparisons design (David 1988, de Vries 1998). Ant 2006). Seed size seems to be an important trait for colonies were offered one pair of seeds of different choice by these ants. Seeds of intermediate size prevail species at a time in all possible species combinations. in the diet, and a slight size match between seed size This design was chosen over cafeteria experiments and ant body size was detected, with P. mendozanus (Gross et al. 1991, Crist and MacMahon 1992) because carrying larger seeds than P. rastratus, followed by P. when more than two seeds were offered, individual inermis, matching body-size differences of the ant foragers generally removed one of them without eval- species (Pirk and Lopez de Casenave 2010). If the uating the complete set. The only way to make sure documented diets of these ants reßect their foraging that ants were actually choosing between alternatives preferences, this would result in a consistent removal was by placing two seeds close to each other (G.I.P., of a few preferred species, increasing the potential unpublished data). effects of granivory on seed banks. We performed Experiments took place at Þve colonies of each choice experiments in the Þeld to describe ant pref- species that presented high activity levels and an ac- erences for seeds of six native species (three grasses, tive response to seed offer on any given sampling two forbs, and one shrub). Under the hypothesis that occasion (February, April, and December 2003; Jan- 214 ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA Vol. 104, no. 2

Table 1. Traits of seeds used in choice experiments performed in the Monte desert, Argentina

Species Family Wta (mg) Lengtha (mm) Widtha (mm) Shape indexb Aristida spp.c Poaceae 0.74 7.70 0.43 17.9 Pappophorum spp.c Poaceae 3.15 2.65 2.48 1.27 S. leucopilac Poaceae 1.59 1.74 1.05 1.66 C. papulosumd Chenopodiaceae 0.18 1.01 0.91 1.11 P. hysterophorusd Compositae 0.61 2.05 1.44 1.42 L. divaricatae Zygophyllaceae 7.09 5.18 2.33 2.22

a Average measures from 20 seeds per species. Seeds were measured as they were offered to ants (see Sampling Design). b Length/width. c Grass. d Forb. e Shrub. uary and February 2004). Experiments on individual were offered as they are generally found in the soil colonies were completed in only one of these occa- seed bank (Peralta and Rossi 1997), which is probably sions. In each experiment, seeds of Þve or six species how ants usually Þnd them. Grass seeds were offered were offered in all 10 or 15 possible combinations of with their vegetative bracts (palea, lemma, and two plant species, respectively. A trial consisted in glumes), C. papulosum without its calyx and pericarp, placing two seeds on the ground, 1 cm apart and within P. hysterophorus with ßower remains and L. divaricata 10 cm from the nest entrance, during periods of high with its hairy mericarp. foraging activity (Pol and Lopez de Casenave 2004). Data Analysis. The proportion of trials in which a It was considered completed when a seed was re- grass seed was removed Þrst when offered with a forb moved and taken to the nest by a forager or 1 h after or a shrub seed was calculated for each species sep- seeds were offered. The Þrst seed to be removed was arately. A chi-square test was performed per colony to assumed to be more preferred than the remaining establish whether seeds removed Þrst were indepen- seed. Five trials were planned for each species com- dent of their type (grass seeds or nongrass seeds). bination (i.e., 50 or 75 trials per colony in total for Þve Yates correction for continuity was applied as sug- and six plant species, respectively), and pairs of seeds gested by Zar (1996). Also, a one-way analysis of were put out at the colony one at a time. The com- variance (ANOVA) was performed to compare pro- pletion of all trials in a colony generally took several portion of nongrass seeds taken in the Þrst place days (up to 3 d) and depended on the response of the among ant species, by using colonies as replicates. colony. However, for some particular seed combina- To rank plant species according to ant preferences, tions, fewer than Þve trials were performed because data were analyzed in paired comparisons matrices. neither seed was removed within 1 h even though This type of analysis has been widely used and im- seeds were actively visited. For several pairings, we proved by ethologists for the study of dominance re- had only a single trial in which no seed removal oc- lationships in a social group of individuals (Appleby curred within 1 h. 1983; Boyd and Silk 1983; de Vries 1995, 1998; de Vries Seeds offered in the experiments came from three et al. 2006). Three subsequent steps were carried out: grasses (Aristida spp., Pappophorum spp., and S. leu- 1) a paired comparisons matrix was constructed for copila), two forbs (C. papulosum and P. hysteropho- each colony, 2) a linearity index and its statistical rus), and a shrub (L. divaricata) (Table 1). Their seed signiÞcance were calculated for each matrix to test if abundance in the soil seed bank of the open woodland preferences for seeds were hierarchical, and 3) a car- (number of seeds meterϪ2)is30forAristida spp., 252 dinal index was calculated to assign a preference value for Pappophorum spp., 104 for S. leucopila, 4,620 for C. to each seed type. papulosum, and 15 for P. hysterophorus (Marone et al. Paired comparisons matrices were 6 ϫ 6or5ϫ 5 1998; seed abundance of L. divaricata was not mea- matrices (for six or Þve species, respectively), with sured). They were chosen because they are native each cell ij representing a pair of seed species, and its seeds that differ in shape and size and are represented value, the number of times species i (row) was pre- in different proportions in these antsÕ diets (i.e., some ferred to species j (column) in all trials involving i and are highly and some are poorly represented in the j. For combinations in which we had only a single trial diet) (Pirk et al. 2004, 2009; Pirk and Lopez de Ca- in which no seed removal occurred, a value of 0.5 was senave 2006). They were collected in the study area assigned to both cells involving the two species, be- the season before the performance of the experiments. cause this situation was considered a tie based on a They were stored in paper bags, placed inside vacuum single trial (de Vries 1995). Likewise, in cases in which sealed plastic bags with silica gel, and kept at 4ЊC until seeds remained unremoved after some completed tri- inspection. Only mature, sound seeds (i.e., that did not als, 0.5 was added to the corresponding cells (see crumble when probed with forceps and that lacked Tables 2Ð4). signs of fungal attack or other damage) were selected Before ranking seeds according to ant preferences, under a stereoscopic microscope. Seeds were manip- it is necessary to verify that preferences are hierar- ulated with forceps to avoid direct hand contact. Seeds chical (Boyd and Silk 1983). When all the elements in March 2011 PIRK AND LOPEZ DE CASENAVE:SEED PREFERENCES OF HARVESTER ANTS 215

Table 2. Preference matrices for the P. rastratus colonies Table 3. Preference matrices of P. mendozanus colonies stud- studied in the Monte desert, Argentina ied in the Monte desert, Argentina

LandauÕs LandauÕs PR342a Pa Se Ar Pr Ch linearity PM1a Ar Se Pa Pr La Ch linearity index, index signiÞcance of h Pa 4 5 5 3.5 h ϭ 1 Ar 45555 h ϭ 0.80 Se 1 3 5 5 Se 1 4 5 1.5 5 p ϭ 0.052 Ar 0 2 5 5 Pa 0 1 1.5 5 5 Pr 0 0 0 1.5 Pr 0 0 1.5 0.5 0.5 Ch 0.5 0 0 0.5 La 0 0.5 0 0.5 0.5 Ch 0 0 0 0.5 0.5 PR355a Se Pa Ar Ch Pr PM2a Ar Se Pa Ch Pr La Se 4 3 5 5 h ϭ 0.95 Pa 1 4 5 5 Ar 5 5 5 4.5 5 h ϭ 0.74 Ar 2 1 3 5 Se 0 2 0.5 4.5 1 p ϭ 0.120 Ch 0 0 2 0.5 Pa 0 0 4 3.5 5 Pr 0 0 0 0.5 Ch 0 0.5 0 0.5 0.5 Pr 0.5 0.5 0.5 0.5 0.5 PR408b Ar Pa Se Pr Ch La 0 0 0 0.5 0.5 Ar 3 5 5 5 h ϭ 0.95 PM3b Ar Pa Se Pr Ch Pa 2 4 5 5 Se 0 1 0.5 1.5 Ar 4455 h ϭ 0.85 Pr 0 0 0.5 1.5 Pa 1 5 5 5 Ch 0 0 0.5 0.5 Se 1 0 0.5 5 Pr 0 0 0.5 0.5 PR543b Pa Ar Se Ch Pr Ch 0 0 0 0.5 Pa 5 4 5 5 h ϭ 0.95 PM12c Se Pa Ar La Pr Ch Ar 0 4.5 4 5 Se 1 0.5 1.5 3.5 Se 34555 h ϭ 0.89 Ch 0 1 0.5 0.5 Pa2 5555 p ϭ 0.051 Pr 0 0 0.5 0.5 Ar 1 0 3.5 4.5 5 La 0 0 1.5 0.5 0.5 PR708a Ar Pa Se Pr Ch Pr 0 0 0.5 1.5 Ch 0 0 0 0.5 1.5 Ar 3 4 4.5 5 h ϭ 0.95 Pa 2 4 4 3.5 PM167c Se Pa Ar Pr La Ch Se 1 1 4.5 4 Pr 0.5 1 0.5 0.5 Se 3 5 4.5 5 5 h ϭ 0.97 Ch 0 0.5 1 0.5 Pa 2 3 4.5 5 5 p ϭ 0.022 Ar 0 2 4.5 5 5 Each value represents the number of trials in which a seed of the Pr 0.5 0.5 0.5 1.5 3.5 species in the row was preferred to a seed of the species in the column. La 0 0 0 0.5 0.5 Species are arranged in decreasing order of preference. Ar, Aristida Ch 0 0 0 1.5 0.5 spp.; Ch, C. papulosum; Pa, Pappophorum spp.; Pr, P. hysterophorus; Se, S. leucopila. Each value represents the number of trials in which a seed of the a Sampled in January 2004. species in the row was preferred to a seed of the species in the column. b Sampled in December 2003. Species are arranged in decreasing order of preference. Ar, Aristida spp.; Ch, C. papulosum; La, L. divaricata; Pa, Pappophorum spp.; Pr, P. a paired comparisons design can be arranged into a hysterophorus; Se, S. leucopila. a Sampled in February 2003. strict hierarchical order, the hierarchy is said to be b Sampled in April 2003. linear. A perfectly linear hierarchy requires all possi- c Sampled in January 2004. ble relationships between three elements to be tran- sitive (i.e., if seed A is preferred to B and B is preferred to C, then A is preferred to C) and all relationships After testing for linearity, a cardinal index was cal- between two elements to be asymmetrical (i.e., one culated to assign a preference value to each seed type seed should be preferred to another seed) (Boyd and in each matrix. DavidÕs score (David 1988, de Vries Silk 1983). To test for linearity, LandauÕs linearity 1998), corrected for chance and normalized (de Vries index, h (Landau 1951), was calculated for each matrix et al. 2006), was chosen (Supp. 2 [online only]). Ac- (Supp. 1 [online only]). This index ranges from 0 to cording to de Vries et al. (2006), its crucial advantage 1, where one indicates complete linearity and 0 indi- is that it takes into account the relative importance of cates that all species are equally preferred. The sig- each element. Thus, a seed removed when offered niÞcance of the linearity index h was determined by with a highly preferred seed is weighted heavier than means of a randomization test (de Vries 1995) in the a seed removed when offered with a lowly preferred program MatMan (version MfW 1.1; earlier version seed. It varies between 0 and N Ϫ 1, where N is the described in de Vries et al. 1993). This test was per- number of compared species, with higher values in- formed only in 6 ϫ 6 matrices because in smaller dicating higher preference levels. matrices there is always a probability higher than 5% To test whether relative abundance of the six of- that any linear hierarchy is occurring by chance fered species in the diet of P. rastratus, P. mendozanus, (Appleby 1983). and P. inermis was positively associated with their 216 ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA Vol. 104, no. 2

Table 4. Preference matrices of P. inermis colonies studied in relative preferences, we conducted Spearman rank the Monte desert, Argentina correlations (data were non-normal) between mean DavidÕs scores of each species of seeds and their mean LandauÕs PI1a Pa Ar Se Pr La Ch linearity index, relative abundance in the diet of each ant species. Diet signiÞcance of h data came from a previous study (Pirk et al. 2009), Pa 5 5 5 5 4.5 h ϭ 0.89 where it was estimated on four occasions in each of Ar0 3445 p ϭ 0.051 three activity seasons. For this analysis we used mean Se 0 2 4 2.5 1.5 relative abundance of seeds in the diet across occa- Pr 0 1 1 0.5 0.5 sions. Finally, Spearman rank correlations were con- La 0 1 0.5 0.5 0.5 Ch 0.5 0 0.5 0.5 0.5 ducted to study the signiÞcance of the association between ant preference and seed weight, and ant PI11b Ar Pa Ch Se Pr preference and seed shape (length/width) (Table 1). Ar 4555 h ϭ 0.85 Pa 1 5 5 3.5 Se 0 0 0.5 0.5 Results Pr 0 0 0.5 0.5 Ch 0 0.5 0.5 0.5 Grass seeds were removed Þrst in a higher propor-

c tion of trials than forb and shrub seeds when a seed of PI355 Ar Pa Se Pr Ch La each type (a grass and a nongrass seed) were offered Ar 32555 h ϭ 0.77 together (96.3, 99.1, and 94.8% for P. rastratus, P. men- Pa2 3555 p ϭ 0.120 Se 3 2 2.5 5 5 dozanus, and P. inermis, respectively; Table 5). The Pr 0 0 1.5 0.5 0.5 chi-square tests per colony were all highly signiÞcant Ch 0 0 0 0.5 0.5 (P Ͻ 0.001 in all cases). Also, the three ant species had La 0 0 0 0.5 0.5 similar preferences for forb and shrub seeds (one-way ϭ ϭ ϭ PI362d Ar Pa Pr Se Ch La ANOVA: F 1.53; df 2, 12; P 0.25). All ant species presented a linear hierarchy in their Ar 55455 h ϭ 0.91 Pa0 5455 p ϭ 0.022 preferences for seeds (Tables 2Ð4). Values of Lan- Pr 0 0 5 4.5 5 dauÕs linearity index were high in all colonies and Se 1 1 0 3.5 0.5 statistically signiÞcant (P Ͻ 0.05) or nearly signiÞcant Ch 0 0 0.5 0.5 0.5 (e.g., P ϭ 0.051 or P ϭ 0.052; Tables 3 and 4) in most La 0 0 0 0.5 0.5 colonies where statistical signiÞcance could be eval- PI482c Ar Pa Se Pr Ch La uated (i.e., in 6 ϫ 6 matrices; see Data Analysis). Only Ar 54555 h ϭ 0.89 colonies PM2 (Table 3) and PI355 (Table 4) had Pa0 4555 p ϭ 0.051 nonsigniÞcant values of LandauÕs linearity index. PM2 Se11 455 presented four ties (the highest number of ties re- Pr 0 0 1 0.5 0.5 corded in all studied colonies, along with colony Ch 0 0 0 0.5 0.5 La 0 0 0 0.5 0.5 PM1), the great majority of which occurred between less preferred species. PI355 presented the only in- Each value represents the number of trials in which a seed of the transitivity case: Aristida spp. was preferred to Pap- species in the row was preferred to a seed of the species in the column. pophorum spp. which in turn was preferred to S. leu- Species are arranged in decreasing order of preference. Ar, Aristida copila, but S. leucopila was preferred to Aristida spp. spp.; Ch, C. papulosum; La, L. divaricata; Pa, Pappophorum spp.; Pr, P. hysterophorus; Se, S. leucopila. However, in each of these combinations, the preferred a Sampled in February 2003. species was removed only in three of Þve trials. In b Sampled in April 2003. colonies where statistical signiÞcance could not be c Sampled in January 2004. d studied (all P. rastratus Õ colonies, Table 2; PM3, Tables Sampled in February 2004. 3; PI11, Table 4), values of LandauÕs linearity index were high (between 0.85 and 1) and for P. mendozanus and P. inermis, preference patterns were similar to

Table 5. Percentage of trials in which a grass seed (Aristida spp., Pappophorum spp., or S. leucopila) was removed ﬁrst when offered together with a nongrass seed (P. hysterophorus, C. papulosum, or L. divaricata) in choice experiments performed in the Monte desert, Argentina

P. rastratus P. mendozanus P. inermis ␹2 a ␹2 a ␹2 a Colony % 1 N Colony % 1 N Colony % 1 N PR342 100 28 28 PM1 97.4 34.1 38 PI1 91.9 25.9 37 PR355 93.3 21.2 30 PM2 100 31 31 PI11 100 28 28 PR408 100 21 21 PM3 100 25 25 PI355 97.7 39.1 43 PR543 95.8 20.2 24 PM12 97.7 39.1 43 PI362 86.8 20.6 38 PR708 92.3 18.6 26 PM167 100 42 42 PI482 97.8 41.1 45 Mean Ϯ SE 96.3 Ϯ 1.6 99.1 Ϯ 0.6 94.8 Ϯ 2.4

Chi-square tests per colony were all signiÞcant (P Ͻ 0.001). a N is the number of trials involving one grass and one nongrass seed per colony. March 2011 PIRK AND LOPEZ DE CASENAVE:SEED PREFERENCES OF HARVESTER ANTS 217

(a) 4 contrast with P. rastratus, preference of P. mendozanus and P. inermis for L. divaricata seeds (the heaviest in 3 the set) was estimated and was very low. Thus, for these two species, preferences were highest for seeds 2 of intermediate size (Fig. 2). Seed preference was not DSnorm 1 correlated with seed shape in any ant species (Spear- ϭ ϭ ϭ man rank correlation: rs 0.40 N 5, P 0.51 for P. 0 ϭ ϭ ϭ rastratus; rs 0.60, N 6, P 0.21 for P. mendozanus; Ar Pa Se Pr Ch ϭ ϭ ϭ and rs 0.26, N 6, P 0.62 for P. inermis). There were both relatively elongated and rounded seeds (b) 5 among the most preferred species (Fig. 2). 4 3 Discussion 2 DSnorm In the central Monte desert, P. rastratus, P. mendo- 1 zanus, and P. inermis prefer grass to forb and shrub

0 seeds. Grass seeds were removed Þrst in almost every Ar Pa Se Pr Ch La trial in which the alternatives were forb or shrub seeds. These results support our Þrst expectation that grass (c) 5 seeds are more preferred than forb and shrub seeds by

4 these ants, matching their diet (Pirk et al. 2004, 2009; Pirk and Lopez de Casenave 2006). They also agree 3 with other experimental studies that show that harvester ants have marked preferences for certain seed DSnorm 2 types, many of which are grass seeds (Kelrick et al. 1 1986, Gross et al. 1991, Crist and MacMahon 1992, Rey 0 et al. 2002, Grimbacher and Hughes 2002, Reyes- Ar Pa Se Pr Ch La Lo´pez and Ferna´ndez-Haeger 2002, Nicolai et al. Fig. 1. Mean preferences (ϩSE) of (a) P. rastratus, (b) 2007). P. mendozanus, and (c) P. inermis colonies for seeds of Aris- Preferences for forb and shrub seeds are similar tida spp. (Ar), Pappophorum spp. (Pa), S. leucopila (Se), P. among ant species. This does not support our second hysterophorus (Pr), C. papulosum (Ch), and L. divaricata expectation that P. inermis has a lower preference for (La), calculated with the normalized DavidÕs score corrected forb and shrub seeds than P. rastratus and P. mendo- for chance (DSnorm). N ϭ 5 colonies, from choice experiments performed in the Monte desert, Argentina. zanus. Although these seeds have low DavidÕs scores and have an overall low representation in the diet of the three ant species, P. rastratus and P. mendozanus those of other colonies of the same species with a occasionally include them in high proportions (up to signiÞcant linearity index. 78 and 87% of the diet, respectively) at the beginning According to DavidÕs score, the relative preference of the season when grass seed abundance is low (Pirk for Aristida spp., Pappophorum spp., and S. leucopila et al. 2009). This suggests that when preferred species varied among colonies of P. mendozanus and P. ras- are scarce, these ants take high proportions of less tratus, but the three species were always most pre- preferred seeds, as the optimal foraging theory preferred. In contrast, P. inermis displayed a lower pref- dicts (Stephens and Krebs 1986). Other harvester ants erence for S. leucopila, which was even less preferred also show a higher selectivity in the diet when pre- than P. hysterophorus in one colony (PI362, Fig. 1). ferred items are more abundant in the environment Among nongrass seeds, C. papulosum, P. hysteropho- (Whitford 1978, Mehlhop and Scott 1983, Crist and rus, and L. divaricata had similarly low DavidÕs score MacMahon 1992, Wilby and Shachak 2000). In con- values in all colonies (Fig. 1). trast, P. inermis has a lower activity than P. rastratus Relative percentage of seeds in the diet of P. ras- and P. mendozanus when abundance of grass seeds in tratus and P. inermis was positively correlated with ant the environment is low. It displays a more strict diet, ϭ preferences (Spearman rank correlation: rs 0.90, including only preferred species (Pirk et al. 2009). ϭ ϭ ϭ ϭ N 5, P 0.04 for P. rastratus and rs 0.93, N 6, Several seed attributes could account for observed P ϭ 0.01 for P. inermis; Fig. 2). This association was not ant preferences. In our experiments, the three most signiÞcant for P. mendozanus, for which the relative fre- preferred seed species by P. mendozanus and P. inermis quency of preferred S. leucopila in the diet was very low were of intermediate weight, whereas the least pre- ϭ ϭ ϭ (as in P. rastratus)(rs 0.60, N 6, P 0.21; Fig. 2). ferred species were the lightest (C. papulosum and P. Seed preference was positively associated with seed hysterophorus) and the heaviest (L. divaricata). Al- weight only in P. rastratus colonies (Spearman rank though L. divaricata seeds probably provide the high- ϭ ϭ ϭ correlation: rs 0.90, N 5, P 0.04; Fig. 2). In both est energetic beneÞts, they are more than twice as P. mendozanus and P. inermis, correlations were not heavy as Pappophorum spp. seeds, the next largest and ϭ ϭ ϭ signiÞcant (rs 0.20, N 6, P 0.71 for P. mendozanus highly preferred species (7.1 versus 3.1 mg, respec- ϭϪ ϭ ϭ and rs 0.03, N 6, P 0.96 for P. inermis). In tively). Thus, energetic costs associated with handling 218 ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA Vol. 104, no. 2

Pogonomyrmex rastratus Pogonomyrmex mendozanus Pogonomyrmex inermis 4 Pa 4 Ar 4 Ar Se Pa Pa Ar Se Se

Pr 2 2 Pr 2 Ch Pr La Ch DSnorm Ch La 0 0 0 010203001020300 10203040 %diet % diet %diet

4 4 Ar 4 Se Pa Ar Pa Pa Ar Se Se 2 Pr 2 Pr La 2 La

DSnorm Ch Pr Ch Ch 0 0 0 012340246802468 Seed weight (mg) Seed weight (mg) Seed weight (mg)

4 4 Ar 4 Pa Pa Ar Pa Se Ar Se Se 2 2 Pr 2 Pr

DSnorm Ch Pr La La Ch Ch

0 0 0 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 Seed shape (length/width) Seed shape (length/width) Seed shape (length/width) Fig. 2. Association of P. rastratus, P. mendozanus, and P. inermis Õ seed preferences (DSnorm) with the relative percentage of that seeds in the diet, weight, and shape (length/width) in the Monte desert, Argentina. (Ar, Aristida spp.; Pa, Pappophorum spp.; Se, S. leucopila; Pr, P. hysterophorus; Ch, C. papulosum; La, L. divaricata). and carriage of L. divaricata seeds could be too high mandibles and of lifting and transporting large loads and morphological restrictions could arise (i.e., me- may arise. In various ant communities workers maxi- chanical limitations of Þt between seeds and ant man- mize the rate of energy return to the colony by match- dibles; Kaspari 1996). Preferences of P. rastratus in- ing the size of the seeds collected to their own body creased with seed size, but this species was not offered size (Davidson 1977, Kaspari 1996). Seed morphology, seeds of L. divaricata, which most likely impose similar in contrast, does not explain ant preferences for the six restrictions as in P. mendozanus and P. inermis (P. offered species. Pulliam and Brand (1975) suggested rastratus is larger than P. mendozanus but smaller than that elongated seeds could be preferred to round seeds P. inermis; Pirk and Lopez de Casenave 2010). In with smooth outlines, because the former would be another preference experiment involving the same ant easier to handle. However, in this study, both elon- species, in which small (0.4 mg) and large caryopses gated and rounded seeds were among the most pre- (1.3 mg) of Pappophorum spp. were offered, large ferred species. Finally, other seed attributes could caryopses were the most preferred caryopses (Pirk alternatively account for ant seed preferences ob- and Lopez de Casenave 2010). This is consistent with served here. Grass seeds have a low concentration of the results presented here because large caryopses are secondary compounds compared with seeds of other within the size range of the most preferred seeds. Also, when seed size in the diet was analyzed, seeds of plant families (Dõáz 1996). Because secondary com- intermediate size also prevailed in the diet of the three pounds could be toxic to ants (Carroll and Janzen species and there was a slight size match because P. 1973, Whitford 1978, Buckley 1982, Davidson 1982), mendozanus carried larger seeds than P. rastratus, this could be the cause of a higher preference for grass which in turn carried larger seeds than P. inermis, seeds in relation to forb and shrub seeds. Also, nutri- matching body-size differences (Pirk and Lopez de tional quality of the seeds could explain these differ- Casenave 2010). Thus, seed size inßuences ant seed ences. Percentage of soluble carbohydrate, for exam- choice. This is consistent with other studies in which ple, has been found to be positively associated with preferences of harvester ants were positively corre- seed preferences (Kelrick et al. 1986, Crist and Mac- lated with seed size (Kelrick et al. 1986, Crist and Mahon 1992) and probably varies among the six stud- MacMahon 1992), up to a point. As seed size increases, ied species. Chemical analyses of the seeds used in this mechanical limitations of Þt between seeds and ant study need to be performed to determine the impor- March 2011 PIRK AND LOPEZ DE CASENAVE:SEED PREFERENCES OF HARVESTER ANTS 219 tance of secondary chemistry and nutritional quality and composition of some seed resources in the Monte in determining ant seed preferences. desert, with important community implications. Diet composition of these ant species fairly reßects their preferences. Overall, the most preferred species had a higher representation in the previously de- Acknowledgments scribed diets of these ants (Pirk et al. 2004, 2009; Pirk and Lopez de Casenave 2006). However, there were We thank H. de Vries for useful comments on data analysis some discrepancies between preferences and diet. and for help with performing the linearity randomization Setaria leucopila was highly preferred by the three tests in MatMan. F. Milesi and L. Marone also gave valuable species, especially by P. rastratus and P. mendozanus, advice. V. Cueto and A. Mangione read earlier versions of this manuscript. Financial support was supplied by Consejo but had a low representation in the diet. Also, L. Nacional de Investigaciones Cientõ´Þcas y Tećnicas divaricata was lowly preferred but had a relatively (CONICET), by Agencia Nacional de Promocioń Cientõ´Þca important representation in the diet of P. mendozanus. y Tecnolo´gica (ANPCyT) of Argentina (ultimately through Few studies have compared harvester ant diet and PICT 2196), and by Universidad de Buenos Aires (ultimately preferences for native species. Gross et al. (1991) through UBACyT X/120). This is contribution 72 of the found that several species of the genera Chelaner, Desert Community Ecology Research Team (Ecodes) of Meranoplus, and Pheidole stored only grass seeds in IADIZA Institute (CONICET) and Facultad de Ciencias Ex- their granaries, whereas legume seeds were particu- actas y Naturales (FCEyN) (Universidad de Buenos Aires). larly abundant in a tropical savanna woodland in Aus- tralia. Cafeteria experiments also showed a clear preference for grass seeds over legume seeds. Crist and References Cited MacMahon (1992) studied diet and preferences of Appleby, M. C. 1983. The probability of linearity in hierar- Pogonomyrmex occidentalis (Cresson) in a shrub chies. Anim. Behav. 31: 600Ð608. steppe in Wyoming, and they found some discrepan- Azca´rate, F. M., L. Arqueros, A. M. Sańchez, and B. Peco. cies: some highly preferred seeds were poorly repre- 2005. Seed and fruit selection by harvester ants, Messor sented in the diet. These discrepancies were attrib- barbarus, in Mediterranean grassland and scrubland. uted to aspects related to ant foraging behavior and Funct. Ecol. 19: 273Ð283. seed availability. Ants of this species may have little Boyd, R., and J. B. Silk. 1983. A method for assigning car- opportunity to select among seeds in natural condi- dinal dominance ranks. Anim. Behav. 31: 45Ð58. tions because of their low seed encounter rates per Brown, J. H., O. J. Reichman, and D. W. Davidson. 1979. foraging trip (Crist and MacMahon 1992). Similarly, Granivory in desert ecosystems. Annu. Rev. Ecol. Syst. 10: 201Ð227. the high preference but low representation of S. leu- Buckley, R. C. 1982. Ant-plant interactions: a world review, copila in the diet of Pogonomyrmex ants in the Monte pp. 111Ð141. In R. C. Buckley (ed.), Ant-plant interactions desert could be due to the low availability of these in Australia. W. Junk, The Hague, The Netherlands. seeds. In the open woodland of N˜ acun˜ ań the standing Carroll, C. R., and D. H. Janzen. 1973. Ecology of foraging crop of S. leucopila seeds is low compared with other by ants. Annu. Rev. Ecol. Syst. 4: 231Ð257. grass seeds [e.g., Stipa ichu (Ruiz & Pav.) Kunth, T. Claver, S., and H. G. Fowler. 1993. The ant fauna (Hyme- crinita, and Pappophorum spp.], yet similar to other noptera: Formicidae) of the N˜ acun˜ ań Biosphere Reserve. species which were highly represented in the diet (i.e., Naturalia 18: 189Ð193. Crist, T. O., and J. A. MacMahon. 1992. Harvester ant for- Aristida spp. and D. californica; Pol et al. 2010). Den- aging and shrub-steppe seeds: interactions of seed re- sity of S. leucopila individuals is also low but still not sources and seed use. Ecology 73: 1768Ð1779. lower than that of D. californica (Pol et al. 2010). Cuezzo, F., and S. Claver. 2009. Two new species of the ant However, its spatial distribution differs, as its density genus Pogonomyrmex (Hymenoptera: Formicidae) from and soil seed abundance is higher under tree canopy Argentina. Rev. Soc. Entomol. Argent. 68: 97Ð106. (Rossi and Villagra 2003, Marone et al. 2004), whereas Davidson, D. W. 1977. Foraging ecology and community the other grass species abound in open mesohabitats organization in desert seed-eating ants. Ecology 58: 711Ð or under shrub canopy (R. Pol, G.I.P., and L. Marone, 724. unpublished data). Because ants forage very little un- Davidson, E. A. 1982. Seed utilization by harvester ants. Ant-plant interactions: a world review, pp. 1Ð6. In R. C. der tree cover (R. Pol and J.L.d.C., unpublished data), Buckley (ed.), Ant-plant interactions in Australia. W. the encounter rate of S. leucopila seeds by foragers Junk, The Hague, The Netherlands. could be lower than that of other species, resulting in David, H. A. 1988. The method of paired comparisons. C. a lower representation in the diet. GrifÞn, London, United Kingdom. Preferences found in this study are overall coinci- de Vries, H. 1995. An improved test of linearity in domi- dent with diet composition of these ants, indicating nance hierarchies containing unknown or tied relation- that their feeding pressure is clearly directed to grass ships. Anim. Behav. 50: 1375Ð1389. seeds. Because these ants have relatively high seed de Vries, H. 1998. Finding a dominance order most consistent with a linear hierarchy: a new procedure and review. removal rates (6 and 5 ϫ 104 seeds per colony through- Anim. Behav. 55: 827Ð843. out the activity season, for P. rastratus and P. mendo- de Vries, H., W. J. Netto, and P.L.H. Hanegraaf. 1993. Mat- zanus, respectively; Pirk and Lopez de Casenave Man: a program for the analysis of sociometric matrices 2006), this match between diet and preferences could and behavioural transition matrices. Behaviour 125: 157Ð increase the chances of ants affecting the abundance 175. 220 ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA Vol. 104, no. 2 de Vries, H., J. M. Stevens, and H. Vervaecke. 2006. Mea- Reserva de la Bio´sfera de N˜ acun˜ ań (Mendoza, Argen- suring and testing the steepness of dominance hierar- tina). Bol. Ext. Cient. IADIZA 3: 1Ð24. chies. Anim. Behav. 71: 585Ð594. Pirk, G. I., and J. Lopez de Casenave. 2006. Diet and seed Dıáz, M. 1996. Food choice by seed-eating birds in relation removal rates by the harvester ants Pogonomyrmex ras- to seed chemistry. Comp. Biochem. Physiol. 113A: 239Ð tratus and Pogonomyrmex pronotalis in the central Monte 246. desert, Argentina. Insectes Soc. 53: 119Ð125. Espigares, T., and A. Lo´pez-Pintor. 2005. Seed predation in Pirk, G. I., and J. Lopez de Casenave. 2010. Inßuence of seed a Mediterranean pasture: can ants modify the ßoristic size on feeding preferences and diet composition of three composition of soil seed banks? Rev. Chil. Hist. Nat. 78: sympatric harvester ants in the central Monte desert. 615Ð622. Ecol. Res. 25: 439Ð445. Grimbacher, P. S., and L. Hughes. 2002. Response of ant Pirk, G. I., J. Lopez de Casenave, and R. Pol. 2004. Asoci- communities and antÐseed interactions to bush regener- acioń de las hormigas granõ´voras Pogonomyrmex prono- ation. Ecol. Manage. Rest. 3: 188Ð199. talis, P. rastratus y P. inermis con caminos en el Monte Gross, C. L., M. A. Whalen, and M. H. Andrew. 1991. Seed central. Ecol. Austral 14: 65Ð76. selection and removal by ants in tropical savanna wood- Pirk, G. I., J. Lopez de Casenave, R. Pol, L. Marone, and F. land in northern Australia. J. Trop. Ecol. 7: 99Ð112. Milesi. 2009. Inßuence of temporal ßuctuations in seed Heredia, A., and C. Detrain. 2005. Inßuence of seed size and abundance on the diet of harvester ants (Pogonomyrmex seed nature on recruitment in the polymorphic harvester spp.) in the central Monte desert, Argentina. Austral ant Messor barbarus. Behav. Process. 70: 289Ð300. Ecol. 34: 908Ð919. Ho¨lldobler, B., and E. O. Wilson. 1990. The ants. Harvard Pol, R., and J. Lopez de Casenave. 2004. Activity patterns of University Press, Cambridge, MA. harvester ants Pogonomyrmex pronotalis and Pogono- Inouye, R. S., G. Byers, and J. H. Brown. 1980. Effects of myrmex rastratus in the Central Monte desert, Argentina. predation and competition on survivorship, fecundity, J. Insect Behav. 17: 647Ð661. and community structure of desert annuals. Ecology 61: Pol, R., G. I. Pirk, and L. Marone. 2010. Grass seed produc- 1344Ð1351. tion in the central Monte desert during successive wet Kaspari, M. 1996. Worker size and seed size selection by harvester ants in a Neotropical forest. Oecologia 105: and dry years. Plant Ecol. 208: 65Ð75. 397Ð404. Pulliam, H. R., and M. R. Brand. 1975. The production and Kelrick, M. I., J. A. MacMahon, R. R. Parmenter, and D. V. utilization of seeds in plains grassland of southeastern Sisson. 1986. Native seed preferences of shrub-steppe Arizona. Ecology 56: 1158Ð1166. rodents, birds and ants: the relationships of seed attri- Reichman, O. J. 1979. Desert granivore foraging and its im- butes and seed use. Oecologia 68: 327Ð337. pact on seed densities and distributions. Ecology 60: Landau, H. G. 1951. On dominance relations and the struc- 1085Ð1092. ture of animal societies: I. Effect of inherent character- Rey, P. J., J. L. Garrido, J. M. Alcańtara, J. M. Ramı´rez, A. istics. Bull. Math. Biophys. 13: 1Ð19. Aguilera, L. Garcıá, A. J. Manzaneda, and R. Fernańdez. Lopez de Casenave, J. 2001. Estructura gremial y orga- 2002. Spatial variation in ant and rodent post-dispersal nizacioń de un ensamble de aves del desierto del Monte. predation of vertebrate-dispersed seeds. Funct. Ecol. 16: Ph.D. dissertation, Universidad de Buenos Aires, Buenos 773Ð781. Aires, Argentina. Reyes-Lo´pez, J. L., and J. Fernańdez-Haeger. 2002. Food MacMahon, J. A., J. F. Mull, and T. O. Crist. 2000. Harvester storage in the nest and seed selectivity in the harvester ant ants (Pogonomyrmex spp.): their community and ecosys- Messor barbarus (Hymenoptera: Formicidae). Sociobiol- tem inßuences. Annu. Rev. Ecol. Syst. 31: 265Ð291. ogy 39: 123Ð128. Marone, L., V. R. Cueto, F. A. Milesi, and J. Lopez de Ca- Rossi, B. E., and P. E. Villagra. 2003. Effects of Prosopis senave. 2004. Soil seed bank composition over desert flexuosa on soil properties and the spatial pattern of un- microhabitats: patterns and plausible mechanisms. Can. J. derstorey species in arid Argentina. J. Veg. Sci. 14: 543Ð Bot. 82: 1809Ð1816. 550. Marone, L., B. E. Rossi, and J. Lopez de Casenave. 1998. Samson, D. A., T. E. Philippi, and D. W. Davidson. 1992. Granivore impact on soil-seed reserves in the central Granivory and competition as determinants of annual Monte desert, Argentina. Funct. Ecol. 12: 640Ð645. plant diversity in the Chihuahuan desert. Oikos 64: 61Ð80. Mehlhop, P., and N. J. Scott. 1983. Temporal patterns of seed Stephens, D. W., and J. R. Krebs. 1986. Foraging theory. use and availability in a guild of desert ants. Ecol. Ento- Princeton University Press, Princeton, NJ. mol. 8: 69Ð85. Whitford, W. G. 1978. Foraging in seed harvester ants Nicolai, N., J. L. Cook, and F. E. Smeins. 2007. Grass com- Pogonomyrmex spp. Ecology 59: 185Ð189. position affects season shifts in seed preference by Wilby, A., and M. Shachak. 2000. Harvester ant response to Pogonomyrmex barbatus (Hymenoptera: Myrmicinae) in spatial and temporal heterogeneity in seed availability: the Edwards Plateau, Texas. Environ. Entomol. 36: 433Ð pattern in the process of granivory. Oecologia 125: 495Ð 440. 503. Nobu´ a Behrmann, B. E., F. A. Milesi, J. Lo´pez de Casenave, Willott, S. J., S. G. Compton, and L. D. Incoll. 2000. For- R. G. Pol, and B. Pavan. 2010. Taman˜ o y composiciońde aging, food selection and worker size in the seed har- la colonia de tres especies de hormigas del geńero Pogono- vesting ant Messor bouvieri. Oecologia 125: 35Ð44. myrmex (Hymenoptera: Formicidae) en la porcioń cen- Zar, J. H. 1996. Biostatistical analysis, 3rd ed. Prentice Hall, tral del desierto del Monte, Argentina. Rev. Soc. Entomol. Upper Saddle River, NJ. Argent. 69: 117Ð122. Peralta, I. E., and B. E. Rossi. 1997. Guõá para el reconocimiento de especies del banco de semillas de la Received 3 June 2010; accepted 30 November 2010.