, 150:61

A field test of the directed deterrence hypothesis in two species of wild chili


    • Department of ZoologyUniversity of Florida
  • Joshua J. Tewksbury
    • Department of BiologyUniversity of Washington
  • Martin L. Cipollini
    • Department of BiologyBerry College
  • Tomás A. Carlo
    • Department of BiologyUniversity of Washington
Plant Animal Interactions

DOI: 10.1007/s00442-006-0496-y

Cite this article as:
Levey, D.J., Tewksbury, J.J., Cipollini, M.L. et al. Oecologia (2006) 150: 61. doi:10.1007/s00442-006-0496-y


The directed deterrence hypothesis posits that secondary metabolites in ripe fruit function to deter fruit consumption by vertebrates that do not disperse seeds, while not impacting consumption by those that do. We tested this hypothesis in two species of wild chilies (Capsicum spp.). Both produce fruits that contain capsaicinoids, the compounds responsible for the pungency of chilies. Previous work suggests seed-dispersing birds but not seed-destroying rodents consume chili fruits, presumably because rodents are deterred by capsaicin. However, fruit removal from chili plants by rodents and other mammals has not been previously explored. Because laboratory rodents can develop a preference for capsaicin, it is quite possible that wild rodents are natural consumers of chili fruits. We monitored the fate of 125 marked fruits of Capsicum chacoense and 291 fruits of Capsicum annuum. For both species, essentially all fruit removal occurred during the day, when rodents are inactive. Video monitoring revealed fruit removal only by birds, mostly by species known to disperse chili seeds in viable condition. Furthermore, these species are from taxonomic groups that tend to specialize on lipid-rich fruits. Both species of chili produce fruits that are unusually high in lipids (35% in C. chacoense, 24% in C. annuum). These results support the directed deterrence hypothesis and suggest that fruiting plants distinguish between seed predators and seed dispersers by producing fruits that repel the former and attract the latter.


BirdsCapsaicinChiliDirected deterrenceFruit


Many of the most intricate and fundamental interactions between plants and their consumers are mediated by secondary metabolites, compounds with no known physiological role in the plants that produce them (Bernays and Chapman 1994; Coley and Barone 1996; Adler et al. 2001). In seeds, leaves, and unripe fruit, the primary function of secondary metabolites is to deter consumption by granivores and herbivores. In ripe fruits, their function is more complex because fruit consumption can be either beneficial or detrimental to plants, depending on whether the consumer disperses or destroys seeds. Thus, fruiting plants face the evolutionary challenge of packaging their seeds in pulp with compounds that simultaneously attract seed-dispersing frugivores and deter seed-destroying frugivores (Janzen 1977; Herrera 1982; Cipollini 2000). The directed deterrence hypothesis (Cipollini and Levey 1997b) posits that some plant species have met this challenge through the production of secondary metabolites that are inhibitory toward organisms that are likely to destroy seeds, but relatively non-toxic and non-deterrent to seed-dispersing vertebrates.

The directed deterrence hypothesis remains largely untested. Among studies that have examined responses of both seed predators and seed dispersers to isolated fruit secondary metabolites (not extracts), most have concluded that fruit metabolites deter all consumers, not just seed predators (Cipollini and Levey 1997a, c, 1998; Levey and Cipollini 1998; Cipollini 2000; Izhaki 2002; Tsahar et al. 2002). The only metabolites that so far appear to be selectively toxic to seed predators are capsaicinoids (mostly capsaicin and dihydrocapsaicin), the compounds responsible for pungency in chili fruits (Iwai et al. 1979; Cordell and Araujo 1993). Feeding trails with captive animals have demonstrated that capsaicin is an irritant to rodents and not to birds (Norman et al. 1992; Mason and Clark 1995; Tewksbury et al. 1999; Tewksbury and Nabhan 2001). Furthermore, this difference in sensitivity appears general across all birds and mammals, due to taxon-specific differences in the vanilloid receptors responsible for pain perception (Mason et al. 1991; Jordt and Julius 2002). In the field, deterrence of mammals is likely advantageous to chili plants because granivorous rodents are the most common mammals in habitats where chilies occur. Many common species of birds in the same habitat are seed dispersers. Indeed, seed-dispersing birds are well-known consumers of wild chili fruits, whereas nocturnal rodents appear to avoid them (Tewksbury and Nabhan 2001).

We suggest that acceptance of the directed deterrence hypothesis for chilies is premature. Records of avian consumption are limited to casual observations, short-term (4-h) trials of chili fruits on plates on the ground, and some video monitoring (Tewksbury et al. 1999; Tewksbury and Nabhan 2001). More problematical, field data on mammalian consumption of chili fruits are restricted to 2 nights of trials in which small piles of chilies were placed on the ground (Tewksbury et al. 1999; Tewksbury and Nabhan 2001). Although no removal of chili fruits was detected, fruits were not monitored on plants, where consumers would be most likely to search for fruit. Directed deterrence posits that chemistry is used as a pre-dispersal filter, changing the identity of organisms consuming fruits on the plant. Thus, trials conducted with fruit on the ground are not appropriate tests of this hypothesis. Further, because removal of chili fruits is rare and episodic (Tewksbury et al. 1999; Tewksbury and Nabhan 2001), short-term trials will often fail to detect it.

Many experiments with laboratory rodents have revealed the irritant properties of capsaicin (Hilker et al. 1967; Rozin et al. 1979; Prescott 1999; Shumake et al. 2000; Simmons et al. 2001), but a deeper exploration of the literature yields at least four reasons to suspect rodents and other mammals may frequently consume chili fruits in the field. First, field studies on the application of capsaicin to reduce damage to crops and electrical wiring by rodents and other mammals often conclude that capsaicin either has no effect or an effect that quickly disappears over time (Swihart and Conover 1991; Andelt et al. 1994; Wagner and Nolte 2000; Bosland 2001; Santilli et al. 2004). Second, laboratory studies show that mammals become desensitized to capsaicin when they are exposed to it in a way that simulates repeated feeding bouts (Green 1989, 1991; Karrer and Bartoshuk 1991; McBurney et al. 1997; Prescott 1999). Third, rodents and primates that are initially repelled by capsaicin can develop a strong preference for it (Rozin et al. 1981; Rozin and Kennel 1983a, b; Galef 1989; Dib 1990; Prescott and Stevenson 1995). Fourth, wild-derived strains of mice are highly variable in their sensitivity to capsaicin (Furuse et al. 2002).

The primary purpose of this study was to document the natural consumers of two wild chili species, Capsicum annuum in Arizona (USA) and Capsicum chacoense in Bolivia. We identified consumers by conducting censuses of fruits on plants at dawn and dusk and by using video cameras trained on chili plants during the day. The fruit censuses allowed us to monitor a large number of fruit under natural conditions for daytime (bird) and nighttime (mammal) removal. The cameras allowed us to continuously and unobtrusively observe chili fruits and to identify consumers.

We also looked for clues about consumer identity by examining fruit traits. C. annuum and C. chacoense fruits are small (5–10 mm), red when ripe, and ovate—all traits that tend to characterize bird-dispersed species (Janson 1983; Debussche and Isenmann 1989; Herrera 2002; Pizo 2002). Their nutrient composition is potentially important because the nutrient content of pulp is often linked to differences in disperser assemblages (McKey 1975; Snow 1985; Fuentes 1994; Witmer and Van Soest 1998). In particular, mammals tend to avoid lipid-rich fruits (Martin et al. 1951; Debussche and Isenmann 1989; Jordano 1995; Corlett 1996; Bollen et al. 2004) and at least some taxa of birds tend to specialize on them (McKey 1975; Moermond and Denslow 1985; Snow 1985; Place and Stiles 1992; Fuentes 1994; Witmer 1996; Witmer and Van Soest 1998). In the New World, these avian taxa include flycatchers (Tyrannidae), thrushes (Turdidae), and mimids (Mimidae) (Loiselle and Blake 1990; Stiles 1993; Witmer 1996; Witmer and Van Soest 1998). In contrast to cultivated species of Capsicum, fruits of C. annuum and C. chacoense feel oily when crushed. If they are indeed lipid-rich, they would be less likely consumed by mammals and more likely consumed by flycatchers, thrushes, and mimids.

Materials and methods

Study sites

In Arizona, we worked at the Tumucacori Chili Reserve, Santa Cruz County (31°33′N, 111°04′W). Vegetation is predominantly semi-desert grassland and mesquite woodland (Brown 1982). The population of approximately 500 C. annuum var. glabriusculum plants at this site represents the northernmost extent of Capsicum. Fruiting individuals are 0.5–1.5 m in height and occur disproportionately under the canopies of desert hackberry (Celtis pallida) and netleaf hackberry (Celtis reticulata) (Tewksbury et al. 1999). Crop sizes of ripe fruits are typically small (< 100 fruits) but can range to 1,000. When ripening, chili fruits of both study species pass through a pre-ripe stage during which they are bright orange and difficult to remove from the peduncle. When ripe, they are bright red and easily removed from the peduncle. Red fruits can persist for more than a month.

In Bolivia, we worked at Rancho San Julian, Santa Cruz (19°77′S, 62°70′W). The vegetation is xeromorphic, with approximately 75 C. chacoense in a forest stand dominated by Schinopsis, Aspidosperma, Ziziphus, and Prosopis. In stature, C. chacoense is similar to C. annuum, but less bushy and with smaller fruit crops. Several species at our site produced fruit concurrently with C. chacoense, including Celtis, Vallesia, Capparis, and Rivina. C. chacoense was not obviously associated with these species.

At both sites, we monitored fruit removal at the height of the fruiting season (March and April in Bolivia; September–November in Arizona). In Bolivia, we counted ripe fruits on each of ten plants at dawn and at dusk. Chili fruits are displayed at the end of peduncles that contain bracts. These bracts remain intact when a fruit is plucked, thereby allowing us to verify fruit removal. Because we found no evidence of fruits being removed and then dropped under the parent plant, we assume that all removed fruits were consumed and we use the terms “removed” and “consumed” interchangeably. Further support for this assumption comes from analysis of video tapes (see Results). After each fruit census, we removed all empty bracts.

In Bolivia, we monitored from five to 14 ripe fruits (average = 9.4; SD = 3.0) on each of ten plants for approximately 8 days (average = 7.7; SD = 1.2). In Arizona, we followed a similar protocol, monitoring from three to 55 fruits (average = 22.4; SD = 3.4) on each of 13 plants for exactly 11 days.

For video monitoring, we positioned high definition 8-mm video cameras 2–3 m from chili plants and filmed for approximately 12 h/day, including dawn and dusk. We operated six cameras simultaneously on six different plants, often on the same plants that we monitored by fruit counting. Each camera was positioned to record animal activity in all or most of the plant’s crown. In the morning, we determined if any fruit had been removed from a given plant during the previous 24 h. If there had been no fruit removal, we continued filming the same plant. If there had been fruit removal, we changed camera location and started to film a new individual.

Tapes were reviewed by fast-forwarding until animal activity was seen, at which point we watched in real time. We recorded the species of animal and how many fruits it consumed per foraging bout. We did not review tapes that did not contain at least one incidence of fruit removal. If a fruit was present at dusk 1 day and was not present the following morning, we assumed it had been removed at night, even though it was often possible for birds to have removed fruits between when we departed one day and returned the next.

Nutritional analyses

C. annuum and C. chacoense fruits were flash-frozen on dry ice and lyophilized. After removal of seeds, the dried pulp was ground with a mortar and pestle, passed through a no. 40 mesh screen, and stored at −20°C. Pulp samples were analyzed for:

  1. 1.

    Total reducing (simple) sugars via methanol extraction, followed by the anthrone assay (Smith 1981), using a 1:1 glucose:fructose standard.

  2. 2.

    Total nonstructural carbohydrates (TNC) via α-amylase extraction, followed by the anthrone assay (Spiro 1966; Smith 1981), using the same standard.

  3. 3.

    Total lipids via petroleum ether extraction and gravimetric analysis (Williams 1984).

  4. 4.

    Total protein via the Bradford assay (Jones et al. 1989), using a ribulose 1,5-biphosphate oxygenase-carboxylase standard.

  5. 5.

    Ash via residual mass following combustion at 525°C for 24 h.


All assays were run in triplicate and were recorded as average percent of dry pulp mass. Fiber was estimated for each sample by subtracting average values for TNC, lipids, proteins, and ash from 100%. Complex carbohydrates (e.g., starches, hemicelluloses, pectins) were estimated by subtracting simple sugars from TNC.


In Bolivia, 24 marked fruits of C. chacoense were removed during 699 fruit-days (one fruit monitored for 1 day = one fruit-day) and in Arizona, 111 fruits of C. annuum were removed during 3,201 fruit-days. Likewise, video monitoring yielded 125 instances of removal in 1,840 h of taping C. chacoense plants and 45 instances of removal in 362 h of taping C. annuum plants.

Removal of marked fruits at night was essentially non-existent for both species. In Bolivia no fruits disappeared between dusk one day and dawn the next day. In Arizona only one of 111 removed fruits disappeared between dusk and dawn. For this fruit, we could not rule out the possibility that a bird visited the plant before we did that day. Alternatively, it could have been dislodged by deer, which were common in the area.

Video monitoring revealed that all fruit removal during the day was by birds. Birds almost always immediately swallowed the fruits they removed and the few times they did not, they flew out of sight carrying the fruit. We never observed a bird reject a fruit, once it had plucked it from the plant. One or two species of birds at each site were responsible for the vast majority of fruit removal (Fig. 1). In Bolivia, the most common consumer was the small-billed elaenia (Elaenia parvirostris), accounting for 49% of total fruits consumed. (Some individuals may have been Elaenia albiceps, which is similar in appearance but rare in the region; Jahn et al. 2002.) Creamy-bellied thrush (Turdus amaurochalinus) was another common visitor at C. chacoense plants in Bolivia, accounting for 36% of consumed fruits. A total of six additional species, including three flycatchers, accounted for the remaining 15% of fruit consumption.
Fig. 1

Fruit consumption of Capsicum annuum and Capsicum chacoense by birds recorded on video tape swallowing fruits. Also shown are study site locations, approximate range boundaries for C. annuum (hatching) and C. chacoense (cross-hatching), and the approximate range of the genus Capsicum (light gray)

In Arizona, curve-billed thrashers (Toxostoma curvirostre) were by far the most common consumers of C. annuum (69% of fruits consumed), followed by northern mockingbirds (Mimus polyglottos; 18%; Fig. 1). Blue-gray gnatcatchers (Polioptila caerulea), rock wrens (Salphinctes obsoletus), and white-crowned sparrows (Zonotrichia leucophrys) were rare consumers (4% each).

Nutrient content

Lipid content of ripe chili fruits was high, averaging 35% of pulp dry mass in C. chacoense and 24% in C. annuum (Table 1). Simple sugars (mostly glucose and fructose) constituted approximately 16% dry mass in C. chacoense and 24% in C. annuum. Protein content was 10% and 12% in C. chacoense and C. annuum, respectively. Fiber was slightly higher in C. chacoense than C. annuum and complex carbohydrates were higher in C. annuum than in C. chacoense.
Table 1

Nutritional content of Capsicum annuum and Capsicum chacoense fruits from Arizona and Bolivia, respectively. Values are mean percentage dry mass of pulp ± 1 SE


C. annuum

C. chacoensea

Simple sugars

24.4 ± 0.2

16.4 ± 2.6

Complex carbohydrates

7.7 ± 1.1



11.5 ± 1.2

9.6 ± 1.4


23.7 ± 0.3

34.8 ± 0.2


25.1 ± 1.8

32.5 ± 2.3

aValues in table are for pungent C. chacoense fruits. Values for non-pungent C. chacoense fruits are: simple sugars = 15.9 ± 2.4; complex carbohydrates = 0.9 ± 0.2; protein = 9.7 ± 3.8; lipids = 28.7 ± 0.3; fiber = 36.7 ± 1.3


cIncludes free amino acids

dCalculated by difference. Ash content = 7.6%


Our results provide the first support for the directed deterrence hypothesis in naturally occurring fruits. The prevalence of consumption by birds and absence of consumption by mammals was consistent across two species of Capsicum on different continents and exposed to non-overlapping communities of birds and mammals. Capsaicin may therefore provide a mechanism by which plants can prevent consumption of their seeds by mammalian seed predators without decreasing the probability of consumption by avian seed dispersers.

We now address five additional conditions that must be met for strong support of the directed deterrence hypothesis. First, birds that consume chili fruits must pass chili seeds in viable condition. Indeed, seeds of C. annuum defecated by their most common consumer in Arizona, curve-billed thrashers, germinate at the same rate (ca. 50%) as control seeds (Tewksbury and Nabhan 2001). In Bolivia, we have fed C. chacoense fruits to captive individuals of the two most common consumers, elaenias and creamy-bellied thrushes. Of seeds defecated by elaenias, 70–80% germinated within 60 days—approximately the same frequency as seeds removed directly from fruits (J. Tewksbury and D. Levey, unpublished data). We did not attempt to germinate seeds defecated by creamy-bellied thrushes, but none appeared damaged in any way. In another study that used similar species and examined the effect of gut passage on germination, seeds of two species of Solanum defecated by American robins (Turdus migratorius) were nearly 100% viable (Cipollini and Levey 1997c). More generally, germination of seeds in fleshy fruits is typically enhanced by gut passage of fruit-eating birds (Traveset 1998). Taken together, this evidence suggests that the most common consumers of C. annuum and C. chacoense fruits are legitimate seed dispersers.

Second, small mammal consumers of fruits and seeds must be common at our study sites. This condition is necessary to dismiss the explanation that we failed to detect removal by small mammals because they were not present. Rodents are abundant and diverse in arid regions of southwestern North America, where they comprise a major guild of granivores (Heske et al. 1994; Brown et al. 2001). At our Arizona study site, the two most abundant species that consume seeds and fruits are the cactus mouse (Peromyscus eremicus) and desert woodrat (Neotoma lepida); non-Capsicum fruits are frequently consumed by them (Tewksbury et al. 1999; Tewksbury and Nabhan 2001). Because both of these rodents are good climbers (Caras 1967; Hoffmeister 1971), C. annuum fruits on branches would be accessible to them. The small mammal community of our Bolivian study site is totally unstudied. Experiments comparing removal of seeds in cages (protected from rodents) to removal of uncovered seeds have revealed higher removal of uncovered seeds, a difference that we attribute mostly to the foraging of granivorous rodents (J. Tewksbury and D. Levey, unpublished data). In nearby chaco habitat, at least ten species of small rodents in seven genera have been recorded (Anderson 1997; Taber et al. 1997; Chu et al. 2003). At least three of these genera (Graomys, Oligoryzomys, and Oryzomys) are generally considered to be arboreal and to consume seeds (Eisenberg and Redford 1999).

Third, removal of chili fruits by birds must not be unusually rare for a bird-dispersed shrub. Although the directed deterrence hypothesis does not require total deterrence of seed predators or total lack of deterrence of seed dispersers, it is important to rule out the possibility that all types of fruit consumers are generally deterred (general toxicity hypothesis; Cipollini and Levey 1997c). In this respect, the low removal rates of fruits by birds at both of our study sites (ca. 3% per day), might indicate that fruit-eating birds avoid consumption of chili fruits—that capsaicin is generally toxic. We do not believe this to be the case because similarly low removal rates typify bird-dispersed shrubs with small crop sizes (Willson and Whelan 1993; Li et al. 1999; Delgado Garcia 2002; McCarty et al. 2002; Tsahar et al. 2003; Kwit et al. 2004) and because ripe chili fruits can persist for > 1 month and almost all are eventually consumed (J. Tewksbury and D. Levey, unpublished data). Furthermore, birds are physiologically incapable of sensing capsaicin (Jordt and Julius 2002).

Fourth, small mammals must be likely to destroy any chili seeds that they consume. If this were not the case, these mammals would not be seed predators and the directed deterrence of capsaicin would not apply to them. When cactus mice and packrats are offered seeds from pungent Capsicum fruits, they totally avoid them (Tewksbury and Nabhan 2001). However, when the same species are offered naturally non-pungent Capsicum fruits, they consume them and destroy the seeds (Tewksbury and Nabhan 2001). We have been unable to test treatment of Capsicum seeds by Bolivian small mammals but we expect similar results. With a related species of rodent (Peromyscus maniculatus) that consumes fruits and seeds from two species of Solanum (which is in the same family as Capsicum), essentially no seeds are passed intact (Cipollini and Levey 1997c).

Finally, capsaicin is expected to occur in Capsicum plants primarily where it will be most likely to deter seed predators—i.e., in and around the seeds. If it occurred in abundance throughout Capsicum plants, its primary function would unlikely be directed toward seed predators (Ehrlen and Eriksson 1993; Cipollini and Levey 1998). Unlike many secondary compounds, which occur throughout the plants that produce them (Ehrlen and Eriksson 1993), capsaicin indeed occurs at high concentrations only in fruit and it is concentrated around the seeds (Suzuki et al. 1980; Fujiwake et al. 1982).

Fruit nutrient content

Nutritionally, chili fruits of both species are extraordinarily rich in lipids. Values of 24 and 35% lipid (C. annuum and C. chacoense, respectively) are approximately 6–9 times higher than the average of 21 other fruit species in the same family (Solanaceae; Jordano 2000), and higher than the averages of species in 22 additional families with vertebrate-dispersed fruits, including families characterized by lipid-rich fruits (Jordano 2000). As typical of fruits with high lipid content (Jordano 1995), the carbohydrate content of chili fruits we assayed was relatively low, 16–24%, compared to an average of 65% across 25 families of fruiting plants (Jordano 2000). Protein, which tends to vary independently of lipid and carbohydrate content across families was comparable to other solanaceous fruits (9–10%) and higher than the averages of 24 of 25 additional families (Jordano 2000).

Field and laboratory studies demonstrate the importance of fruit nutrient content in diet selection of fruit-eating birds (Levey 1987; Jordano 2000; Schaefer et al. 2003). Some species prefer fruits that are high in carbohydrates and low in lipids, whereas others prefer fruits low in carbohydrates and high in lipids (Fuentes 1994; Witmer 1996; Witmer and Van Soest 1998; Levey and Martínez del Rio 2001). The former tend to have rapid transit of digesta through the gut, whereas the latter have slower transit times and feed frequently on insects or seeds (Fuentes 1994; Witmer and Van Soest 1998; Levey and Martínez del Rio 2001). This general pattern fits our results. Most avian species that we observed consuming wild chili fruits are omnivores that consume mostly insects or seeds, at least during part of their annual cycle (Martin et al. 1951). Our observations also agree with previous work documenting that thrushes, mimids, and some species of flycatchers specialize in lipid-rich fruits (Loiselle and Blake 1990; Stiles 1993; Fuentes 1994; Witmer and Van Soest 1998). Indeed, these taxa comprised the vast majority of chili fruit consumption records at both sites (Fig. 1).

Further tests of the directed deterrence hypothesis

A unique opportunity exists for testing the directed deterrence hypothesis in Bolivia, where at least three species of Capsicum are polymorphic for production of capsaicinoids (Tewksbury et al. 2006). In these species, individuals that bear only pungent fruit can be found near individuals that produce only non-pungent fruit. Also, the ratios of pungent to non-pungent individuals vary geographically (Tewksbury et al. 2006). The directed deterrence hypothesis predicts that mammals would more likely consume fruits from non-pungent individuals than from pungent individuals and that birds would consume pungent and non-pungent fruits equally. This prediction assumes that fruit nutrient profiles do not co-vary with pungency. For at least one species, C. chacoense, this is indeed the case; proportions of lipids, proteins, simple sugars, and complex carbohydrates do not differ between pungent and non-pungent fruits (Table 1, footnote a).

It remains unclear how common directed deterrence may be among other fruiting plants. Although it is difficult to record all fruit consumers of a given species, it appears that many bird-dispersed fruits are left alone by granivorous rodents (e.g., squirrels in north temperate forests). This pattern is consistent with the directed deterrence hypotheses, suggesting that a primary reason directed deterrence is rarely documented is because it is rarely studied.


We thank Robert Dobbs, Meribeth Huzinga, Dan Cariveau, and Melissa Simon for field help in Bolivia and Arizona. Rebecca Neal, Sarah Stephan, and Christopher Worrell helped with the nutritional analyses. Chris Whelan and an anonymous reviewer provided constructive comments that greatly improved the manuscript. This work was supported by the National Geographic Society and the National Science Foundation (NSF DEB-0129168).

Copyright information

© Springer-Verlag 2006