Ecology, 80(6), 1999, pp. 2074±2084 ᭧ 1999 by the Ecological Society of America

MUTUALISTIC INTERACTIONS BETWEEN VIRESCENS (), A POLLINATING SEED-CONSUMER, AND SCHOTTII (CACTACEAE)

J. NATHANIEL HOLLAND1 AND THEODORE H. FLEMING Department of Biology, University of Miami, Coral Gables, Florida 33124 USA

Abstract. Pollinating seed-consuming interactions are rare, but include ®g±®g wasp and ±yucca interactions, both of which are thought to be coevolved. Conditions favoring such mutualisms are poorly known but likely include and pollinators whose life cycles are synchronized. In this paper, we describe a new pollinating seed-consumer mutualism between a Sonoran Desert , (senita cactus), and a pyralid moth, Upiga virescens (senita moth). We compare this mutualism with the yucca mutualism in terms of life history traits, active pollination, and selective abortion. Senita cactus ¯owers were pollinated nearly exclusively by nocturnal senita , but a few halictid bees also pollinated ¯owers. Only 40% of ¯owers set fruit during the years of study, apparently due to resource limitation. All phases of the senita moth's life history were associated with the senita cactus. During ¯ower visitation, female senita moths col- lected , actively pollinated ¯owers, and oviposited one egg. After ¯owers closed, emerging larvae bored into the tops of developing fruit, where they consumed seeds and fruit tissue. However, not all seeds/fruit were consumed by larvae because only 20% of eggs produced larvae that survived to be seed/fruit consumers. Senita cactus and senita moth interactions were mutualistic. Moths received food resources (seeds, fruit) for their progeny, and cacti had a 4.8 bene®t-to-cost ratio; only 21% of developing fruit were destroyed by larvae. Life history traits important to this mutualism included low survival of senita moth eggs/larvae, several moth generations per ¯owering season, host speci®city of senita moths, active pollination, oviposition into ¯owers, and limited seed/fruit con- sumption. Active pollination by senita moths in the presence of co-pollinators supports the prediction that active pollination can evolve during a period of coexistence with co-pol- linators. The specialization of both senita and senita moths in the presence of co-pollinators makes the senita mutualism quite remarkable in comparison with ®g±®g wasp and yucca± yucca moth mutualisms. Key words: life history; Lophocereus schottii; mutualism; oviposition; pollination, active; seed consumption; senita cactus; senita moth; Sonoran Desert; specialization; survivorship; Upiga vires- cens.

INTRODUCTION resources (i.e., seeds/fruit) for their progeny and in some cases obtain resources (i.e., nectar, pollen) for commonly oviposit in ¯owers and many in- themselves. Based on net effects of the interaction on sects pollinate ¯owers. However, oviposition in ¯owers the population (usually measured as seed/fruit and subsequent predispersal seed consumption by lar- vae of adults that pollinated ¯owers is extremely rare. set), however, such interactions could potentially take Known cases of both pollinating and seed-consuming the form of predation, commensalism, or mutualism. interactions between plants and their pollinators in- Interactions between vulgaris and clude and Hadena moths (Pettersson moths are predatory due to ineffective pollination and 1991a, Pettersson 1992a), and larval consumption of seeds (Pettersson 1991a, b). moths (Thompson and Pellmyr 1992, Davis et al. moths are effective pollinators but are 1992), Trollius spp. and globe¯ower ¯ies (Pellmyr commensalistic with Lithophragma in years or popu- 1989, Pellmyr 1992), and ®g wasps (Janzen 1979, lations where co-pollinators are abundant because their Wiebes 1979), and Yucca and yucca moths (Riley positive effects on seed set are masked by co-polli- 1892). Interactions between seed-consuming pollina- nators (Thompson and Pellmyr 1992, Pellmyr et al. tors and plants confer a positive effect upon the pol- 1996). In contrast, the other three known pollinating linator population because adult pollinators assure food seed-consuming interactions are mutualistic, and each has been referred to as a highly specialized, coevolved system: (1) and Chiastocheta ¯ies Manuscript received 21 July 1997; revised 19 May 1998; accepted 14 August 1998. (Pellmyr 1989, 1992), (2) ®gs and ®g wasps (Janzen 1 E-mail: jholland@®g.cox.miami.edu 1979, Wiebes 1979, Bronstein 1987, Kjellberg et al. 2074 September 1999 POLLINATING SEED-CONSUMER MUTUALISM 2075

1987, Addicott et al. 1990), and (3) yucca and yucca U. virescens on ¯owers of L. schottii, and (4) survi- moths (Riley 1892, Aker and Udovic 1981, Addicott vorship, life cycle, and life stage associations between 1986, Pellmyr et al. 1996). Trollius europaeus interacts U. virescens and L. schottii. mutualistically with three of Chiastocheta ¯ies that pollinate and oviposit into ¯owers (Pellmyr 1989, METHODS 1992). Actively pollinating ®g wasps (Herre 1996, Ma- Study species and study area chado et al. 1996, West et al. 1996, Anstett et al. 1997) are mutualistic with Ficus because wasps assure a food The Lophocereus contains two species that are resource for progeny and simultaneously confer a pos- restricted to the Sonoran Desert, with populations of itive effect on seed set. Similarly, Yucca interact mu- L. schottii (Englem.) extending as far north as southern tualistically with yucca moths that actively pollinate (Lindsay 1963). The senita cactus attains ¯owers and oviposit into locules where larval growth heights of 2±4 m (Parker 1989) and has many branches and development occur. Fig±®g wasp and yucca±yucca radiating from its base. Senita produce small whitish- moth mutualisms differ from other seed-consuming pink ¯owers that were presumed to be hawk moth pol- pollination interactions in that high plant speci®city linated (Gibson and Horak 1978). Cactus branches bear occurs among pollinators and co-pollinators are absent many areoles (spine-bearing pads) that can produce one (Miles 1983, Bronstein 1987, Addicott et al. 1990, Pell- or more ¯owers per night. Flowers open at sunset and myr and Thompson 1992, James et al. 1993), and in remain open for 6±12 h. Each ¯ower contains a single that pollen collection and pollination are active, not stigma and 109 Ϯ 4.9 anthers (mean Ϯ 1 SE; n ϭ 30 passive (e.g., Riley 1892, Bronstein 1992, Pellmyr plants). Corollas are 6.1 Ϯ 0.2 mm (n ϭ 30 plants) 1997). wide at the mouth of the ¯ower, 9 mm long, and 2.1 Studies of specialized pollination mutualisms indi- Ϯ 0.1 mm (n ϭ 30 plants) wide just above the ovary. cate that the community context in which pairwise in- Lophocereus schottii is self-incompatible and repro- teractions occur between plant and pollinator can affect duces both vegetatively and sexually (Parker 1989, the outcome of interactions in part due to the presence Fleming and Holland 1998). or absence of co-pollinators and predators of pollina- The senita moth, Upiga virescens (Hulst) (Lepidop- tors (Pellmyr 1989, 1992, Thompson and Pellmyr 1992, tera: Pyralidae), is the only known species of Upiga Herre 1996, West et al. 1996, Thompson 1997). In ad- (Munroe 1972). Its range includes southern Arizona dition to community context, life history traits are in- and Sonora and Baja California, Mexico. It is a small creasingly recognized as important for understanding moth with a forewing length of 7±9 mm. The last ab- the outcome of population interactions in communities dominal segments of females are covered ventrally with (Addicott et al. 1990, Bronstein 1994, Polis et al. 1996). many long scales (Fleming and Holland 1998). In order for a plant and pollinator to develop an obligate We conducted this study from April through July in mutualism, appropriate life history traits must be pres- 1995 and 1996 on the Gulf of California (29Њ N, 110Њ ent in plant and pollinator populations (Addicott et al. W) ϳ9 km northeast of Bahia Kino, Sonora, Mexico. 1990, Waser et al. 1996). Specialization of a plant on Annual rainfall at Bahia Kino is 200 mm. Topography a pollinator is predicted to occur when the plant is (1) of the area includes sandy ¯atlands and steep hills large and long lived, (2) has many reproductive epi- (200±450 m above sea level). Study sites (Seri and sodes, and/or (3) has an effective pollinator whose pop- Polilla Flats) were 2 km apart with adult L. schottii ulation dynamics are predictable (Waser et al. 1996). densities of 7 and 20 plants per hectare, respectively. Specialization of a pollinator on a plant is predicted to For further description of the Central Gulf Coast region occur when (1) the pollinator's generation time is not of the Sonoran Desert and our study site, see Shreve longer than the duration of a ¯owering season and (2) and Wiggins (1964) and Fleming et al. (1996). synchrony occurs between ¯owering phenology and a Flower phenology, nectar production, and seed set pollinator's life cycle (Addicott et al. 1990, Waser et al. 1996). We studied ¯ower phenology of L. schottii for the In this paper, we report a mutualism involving pol- same 20 plants in 1995 and 1996 by counting the num- lination and seed-consumption interactions between a ber of open ¯owers at 1-wk intervals, beginning the pyralid moth (Upiga virescens) and a columnar cactus ®rst week of April. Our observations of ¯ower phe- (Lophocereus schottii), and the third known occurrence nology stopped in early July, about three-quarters of active pollination. Results of this study enabled us through the ¯owering season. In 1996, we measured to evaluate theory on the importance of life history ¯oral nectar using 2-␮L capillary tubes for n ϭ 56 traits in the ecology and evolution of pollinating seed- bagged and n ϭ 54 unbagged ¯owers among 29 cacti. consuming mutualisms by comparing the senita mu- Bagged ¯owers were covered with bridal-veil netting tualism with other such mutualisms. Speci®cally, we (1-mm mesh) before opening; the next morning, netting studied (1) phenology, seed set and seed germination was removed and nectar measured. To determine the in L. schottii, (2) patterns of ¯ower visitation and ¯ower number and viability of seeds resulting from pollina- visitation behavior of U. virescens, (3) oviposition by tion, we collected three fruits from each of 30 plants 2076 J. NATHANIEL HOLLAND AND THEODORE H. FLEMING Ecology, Vol. 80, No. 6 and counted seeds. We assessed seed viability by plac- were staggered in time; we marked cohort B after com- ing all seeds from a fruit on moistened ®lter paper in pleting the study of cohort A. For cohort A we tagged a petri dish and counting the number of seeds that 229 ¯owers bearing an egg and then randomly collected germinated within 12 d. Mean number of seeds per fruit a subset (9±35) of the ¯owers/fruits over 9 d. We did and percentage germination per fruit were calculated not follow cohort A through pupation because cohort by averaging the means of 30 plants. We conducted size was too small to allow for sequential destructive pollinator exclusion and hand-pollination studies to de- sampling for Ͼ9 d. For cohort B, we tagged 589 ¯owers termine the contribution of nocturnal (senita moths) bearing an egg and then randomly collected a subset and diurnal (bees) pollinators to fruit set and whether (12±72) of ¯owers/fruits/areoles over 28 d. For each fruit set was limited by pollen. Pollinator exclusion collection we dissected corollas, fruits, and/or areoles studies included three treatments (Fleming and Holland to determine the presence of U. virescens. Survivorship 1998): (1) control ¯owers open to all pollinators, (2) was determined based on population proportions (Var- nocturnal ¯owers available to pollinators at night but ley et al. 1973), and standard deviations of population covered with bridal-veil netting before sunrise, and (3) proportions were calculated. To further quantify larval diurnal ¯owers available to pollinators in the morning presence in cactus branches, we dissected the terminal but covered with bridal-veil netting during the night. 35 cm of eight cactus branches from eight plants. For Stigmas of hand-pollination ¯owers received fresh pol- each branch we counted the number of reproductive len from another plant. Hand-pollination ¯owers were areoles (n ϭ 1375 areoles), number of larval galleries not covered with netting. behind areoles, and determined the presence or absence of a parasitic wasp based on wasp pupa or pupal case. Moth activity period and ¯ower visitation To determine if larval pupation occurred in aborted We determined activity periods and time of ¯ower fruit in the ®eld, we collected and dissected n ϭ 783 visitation of senita moths by conducting hourly cen- aborted fruits from underneath n ϭ 28 cacti and ex- suses of ¯owers on each of six nights. We counted the amined them for presence or absence of larvae, pupae, total number of ¯owers on each of 30 plants and cen- or pupal cases. sused a subset (ϳ1±100 ¯owers per plant) of ¯owers on each plant to determine the proportion of ¯owers RESULTS occupied by senita moths. Censuses began in the eve- Phenology of Lophocereus schottii ning at 1900 MST and ended at 2400 or 0100, when Ͻ5% of ¯owers were occupied by moths. During each Lophocereus schottii ¯owered from the ®rst week of census, senita moth matings were noted and all ¯ower April through at least early July (Fig. 1), and usually visitors (senita moths and others) were recorded. All ¯owers through early August (Shreve and Wiggins proportions were arcsine square-root transformed prior 1964). In both 1995 and 1996, mean ¯ower production to statistical analysis. Repeated-measures ANOVA was was Ͻ10 ¯owers per cactus per night prior to mid- used to test for the effect of time on proportion of April and Ͼ40 ¯owers per cactus per night by July ¯owers occupied by moths. Linear regression analyses (Fig. 1). Although differences in peaks and troughs of were used to determine whether peak proportions of ¯ower production occurred between years, overall phe- ¯owers occupied by moths were associated with total nological pattern and rate of ¯ower production were number of ¯owers on a plant. We studied moth behavior similar for the two years. Flower density did not differ during ¯ower visitation by observing focal ¯owers after between 1995 (22.5 Ϯ 4.8 ¯owers per plant per census dark between 2000 and 2300. When a moth arrived at (n ϭ 14 wk) and 1996 (25.2 Ϯ 5.7 ¯owers per plant a ¯ower, we recorded the sequence of its behaviors and per census, n ϭ 13 wk) ¯owering seasons (means Ϯ 1 duration of each behavior to the nearest second. Ob- SE)(t ϭ 0.37, df ϭ 24, P ϭ 0.72). If we extrapolate servations of ¯ower visitation behavior ceased when our phenology results through August, each plant in moths either left a ¯ower or remained in a nectaring the population produced an average of 3138 Ϯ 545 position for Ͼ1 h. We recorded 240 individual behav- (range 291±10 861) and 3370 Ϯ 522 (196±8867) ¯ow- iors by 94 moths during 22.8 h of observation. ers per season in 1995 and 1996, respectively. Senita ¯owers opened shortly after sunset (ϳ2000 Moth life stages and survivorship Mountain Standard Time [MST]) and remained open On six nights for 30 plants, we counted the total for 6±12 h. In 1995, which had a cool spring, ¯owers number of ¯owers per plant and the number of eggs frequently remained open after sunrise (ϳ0500) until on accessible ¯owers, in order to determine (1) distri- 0900. However, during the warmer spring of 1996 ¯ow- bution of eggs among ¯owers in the cactus population ers were closed by 0400 in the morning, Ͼ1 h before and (2) proportion of ¯owers per plant receiving an sunrise. In 1995 and 1996, respectively, pollinator ex- egg. We marked two cohorts (A and B)ofU. virescens clusion experiments indicated that 75 and 90% of fruit eggs by tagging ¯owers containing an egg to study set resulted from nocturnal pollination by Upiga vi- survivorship, life stage associations with L. schottii, rescens; the remaining 10±25% resulted from polli- and agents of mortality. Our studies of the two cohorts nation by diurnal co-pollinators (Fleming and Holland September 1999 POLLINATING SEED-CONSUMER MUTUALISM 2077

FIG. 1. Flower phenology (no. ¯owers per plant [mean Ϯ 1 SE]) of Lophocereus schottii in 1995 (⅜) and 1996 (Ⅺ). Flower censuses occurred from the ®rst week of April through early July. The same 20 plants were censused every week in both years.

1998). Hand-pollination experiments indicated that ¯ower, all ¯owers measured for 10 plants lacked nectar fruit set was apparently resource limited in 1995 and while all ¯owers for the other 10 plants contained nec- 1996, with Ͼ50% of fruit aborting by day six (Fleming tar. In other words, individual cacti had ¯owers that and Holland 1998). either contained nectar or lacked nectar. Bagged ¯owers lacking nectar suggested that not all plants within the Nectar production and seed set cactus population produced nectar. Alternatively, the Nectar production was variable among plants within polymorphism in the population could be for reab- the cactus population. Both bagged and unbagged ¯ow- sorption of nectar prior to ¯ower closing. Nonetheless, ers produced nectar and lacked nectar (Fig. 2). For neither the presence nor volume of nectar served as a bagged ¯owers that contained nectar, ¯owers most fre- reliable resource to attract co-pollinators. quently produced 0.50±0.74 ␮L of nectar (Fig. 2). Of Fruit matured in ϳ30 d and contained small black the 20 plants where nectar was measured for Ͼ1 bagged seeds weighing 2.7 Ϯ 0.07 mg (mean Ϯ 1 SE). In each

FIG. 2. Relative frequency of nectar volume for bagged (n ϭ 56 ¯owers across 29 plants) and unbagged (n ϭ 54 ¯owers across 29 plants) ¯owers within the cactus population. 2078 J. NATHANIEL HOLLAND AND THEODORE H. FLEMING Ecology, Vol. 80, No. 6

FIG. 3. Activity period of Upiga virescens. Proportion (mean Ϯ 1 SE) of ¯owers occupied by moths (Ⅺ) every hour for each of six nights and the relative frequency of observed moths that were mating (⅜, n ϭ 15 moth matings) during the night. fruit, an average of 182 Ϯ 11 seeds (range 84±340 oviposition (1.5 Ϯ 0.96 min). Female moths actively seeds) set. Of those seeds in each fruit, 97.5 Ϯ 0.54% collected pollen by lowering and rubbing their abdo- germinated. mens among anthers. Females actively pollinated ¯ow- ers by climbing onto the stigma in a head down position Life history of Upiga virescens and rubbing their abdomen onto the stigma, thereby Activity period.ÐDuring the day, moths rested on transferring pollen from abdomen to stigma. Female the 5±10 cm long spines of senita branches. Moths moths have a ``pollen brush'' on their abdomens (for began to visit ¯owers when they opened at 2000 (Fig. illustrations, see Fleming and Holland [1998]) that fa- 3). On six nights, proportion of ¯owers occupied by cilitates pollen collection, transport, and delivery to the moths ranged from Ͻ10% to 45%. Occupancy peaked stigma. When (presumably) consuming nectar, moths at 2000 and 2100 and declined steadily thereafter, ap- crawled deep into the corolla and remained there from proaching zero at 0100 (Fig. 3). Proportion of ¯owers 30stoϾ1 h. Ovipositing moths bent and rubbed their occupied by moths varied signi®cantly during the night abdomen on the ¯ower, and then deposited an egg on

(F5,25 ϭ 21.52, P Ͻ 0.0001; Fig. 3). Peak proportion the upper surface of a tip or among anthers. When of ¯owers occupied by moths was not correlated with ¯owers closed, eggs were located inside wilting co- number of open ¯owers per plant on any of six nights rollas. (r2 Ͻ 0.10 and P Ͼ 0.05 for each night). Plants pro- Eggs.ÐEggs were uniformly distributed across ¯ow- ducing a greater number of ¯owers did not necessarily ers on ®ve of six nights of observation (Table 1). Fewer attract a proportionately greater number of moths. ¯owers contained zero eggs than expected, many more Adults.ÐPrimary events during the adult life stage ¯owers contained one egg than expected, and fewer of U. virescens were ¯ower visitation and mating. Rel- ¯owers contained two or three eggs than expected (Ta- ative frequency of mating increased as the proportion ble 1). Thus, moths avoided ovipositing more than one of ¯owers occupied by moths decreased through the egg per ¯ower. Of ¯owers receiving an egg, 94% had night (Fig. 3). All mating events were observed on only one egg while 6% had Ͼ1 egg. On two nights of senita cactus spines with the peak occurring between egg censusing at Polilla Flats, 22.4 Ϯ 4.6% and 19.7 2300 and 2400. Flowers were visited mostly by female Ϯ 3.4% (mean Ϯ 1 SE) of ¯owers received an egg (n moths: of 50 moths collected from ¯owers, only two ϭ 30 plants); on four nights of egg censusing at Seri were males. Sex ratio of adults collected at night on Flats, 43.7 Ϯ 3.1, 39.4 Ϯ 3.2, 37.1 Ϯ 2.9, and 33.2 Ϯ senita plants but not in ¯owers did not differ signi®- 3.7% of all ¯owers contained eggs (n ϭ 30 plants). cantly from 1:1 (n ϭ 119 adult moths; ␹2 ϭ 2.72, P Ͼ Larvae.ÐLarvae hatched from eggs within 3 d of 0.10). Flower visitation consisted of one or more of oviposition and began to enter fruit as early as 3 d. ®ve behaviors: pollen collection (1.5 Ϯ 0.38 min [mean After hatching, larvae moved through the wilting co- Ϯ 1 SE]), active pollination (0.45 Ϯ 0.10 min), nec- rolla toward the ovary; larvae entered fruit by boring taring (14.6 Ϯ 3.21 min), resting (7.9 Ϯ 2.26 min), and a hole through the base of the corolla into the top of September 1999 POLLINATING SEED-CONSUMER MUTUALISM 2079

TABLE 1. Observed (O) and expected (E) distribution of moth (Upiga virescens) eggs on senita cactus (Lophocereus schottii) ¯owers, and estimated dispersion patterns (s2/␮).

0 Eggs 1 Egg 2 Eggs 3 Eggs Date² Site OE OE OE OE ␹2 df Ps2/␮ 960603 Polilla 111 115.3 40 31.9 1 4.4 0 Ͻ1.0 5.28 1 Ͻ0.025 0.77 960604 Polilla 95 93.9 26 27.6 4 4.0 1 Ͻ1.0 0.16 1 Ͼ0.650 0.93 960604 Seri 267 292.4 180 131.9 9 29.7 3 4.5 35.1 2 Ͻ0.001 0.72 960624 Seri 413 446.4 249 185.9 12 38.7 3 5.4 43.8 2 Ͻ0.001 0.73 960625 Seri 412 435.4 201 157.4 11 28.4 1 3.4 26.1 2 Ͻ0.001 0.74 960626 Seri 326 339.0 133 109.4 9 17.6 0 1.9 11.9 2 Ͻ0.003 0.80 Note: When expected values were Ͻ1.0, the tails of the distribution were summed, and ␹2 calculated based on three terms (0, 1, Ն2 eggs) instead of four (0, 1, 2, 3 eggs). ² Dates are given as yymmdd; i.e., 960603 indicates 3 June 1996. young fruit. Although more than one egg was observed pupated, not all adults emerged from fruit because fruit on ¯owers and two larvae occasionally occurred in co- harden during the pupation period. Although adults pu- rollas, only one larva entered a fruit to consume de- pated and emerged from fruit in the laboratory, no pu- veloping seeds and fruit wall tissue. Upon arriving at pal cases were observed in aborted fruit collected in the bottom of fruit, larvae made an exit hole at the the ®eld (n ϭ 783 fruits from 30 plants), and only one point of fruit-to-areole attachment. They then bored a of 783 aborted fruits contained a larva. Thus, U. vi- hole into the areole and consumed tissue beneath the rescens does not pupate in aborted fruit nor do larvae areole, creating a small gallery in the cactus branch. Of commonly occur in aborted fruit in the ®eld. 1375 reproductive areoles examined, 21% contained lar- Survivorship.ÐLess than 20% of U. virescens larvae val galleries and 2.3% of areoles had Ͼ1 gallery, in- survived to 6 d of age; survival was relatively constant dicating that larval galleries did not render areoles in- from 6 d to pupation (Fig. 4). Four factors contributed capable of further ¯ower or fruit production (Table 2). to preadult survivorship in U. virescens: (1) low pro- Creation of an exit hole in a fruit always resulted in portion of eggs hatching, (2) corolla-induced mortality, abscission of the fruit. Thus, while larvae consumed (3) resource-limited fruit abortion, and (4) wasp par- only a fraction of developing seeds in fruit, larval- asitism. Low survival of eggs due to predation and/or induced fruit abscission ultimately resulted in total seed low fertility accounted for the greatest mortality for mortality for fruit that contained a larva. All ¯owers both cohorts (Fig. 4). For cohorts A and B, respectively, containing an egg did not necessarily abscise because only 33 and 26% of eggs produced larvae. Since eggs survivorship of eggs and early instar larvae was low were not oviposited directly into ¯ower ovaries as usu- (see Survivorship, below). Thus, not all fruit of ¯owers ally occurs with yucca moths (Riley 1892, Aker and pollinated and oviposited on contained a larva. Of 589 Udovic 1981; some yucca moths oviposit into ¯oral ¯owers containing eggs, 67% set fruit and 21% of fruit pedicels and into [see Davis 1967]), eggs were abscised due to larvae, providing the cactus with a 4.8 exposed to abiotic (e.g., desiccation) and biotic (e.g., bene®t-to-cost ratio for interacting with Upiga virescens. predation) agents of mortality. Pupae.ÐGalleries created by larvae in cactus Survival during days two through six was a function branches were used as pupation chambers. To deter- of both corolla-induced mortality and fruit abortion. mine if pupation occurred in aborted fruit, fruit con- Larvae need to reach the base of corollas before day taining a larva were brought into the laboratory (n ϭ six, because corollas wilt and harden after ¯owers 10 fruits). Although all larvae remained in the fruit and close, providing an unfavorable environment for lar- vae. As indicated by the increase in proportion of dead

TABLE 2. Larval/pupal environment of the moth Upiga vi- larvae in corollas after day ®ve of fruit maturation (Fig. rescens on the senita cactus, Lophocereus schottii, in the 4), larvae that fail to exit corollas before day six die, Sonoran Desert. likely due to their inability to eat through the base of hardened corollas. For cohorts A and B, respectively, Branches per Areoles per % Areoles % Galleries 35 and 17% of larvae died within corollas. In addition, Statistic cactus branch with gallery with wasp L. schottii ¯owers not setting fruit were aborted during Mean 40 Ϯ 3 172 Ϯ 11 21.4 Ϯ 6.6 16.6 Ϯ 2.4 the ®rst 6 d after ¯owering (Fig. 5). Consequently, even Range 16±89 111±213 4.5±64.2 5.6±25.3 if larvae hatch from eggs, a proportion of larvae (de-

Notes: Data in ®rst row are means Ϯ 1 SE. Sample sizes: termined by percentage fruit set) nevertheless dies due number of branches per cactus (n ϭ 30 plants), number of to fruit abortion. For cohorts A and B, respectively, 42 areoles per 35 cm of branch (n ϭ 8 plants), percentage of and 67% of ¯owers with an egg set fruit. Thus, of larvae areoles with larval gallery per cactus branch (n ϭ 8 branches), and percentage of larval galleries per branch with parasitic hatching from eggs, 58 and 33% died due to resource- wasp (n ϭ 8 branches). limited fruit abortion. Low egg hatching, corolla 2080 J. NATHANIEL HOLLAND AND THEODORE H. FLEMING Ecology, Vol. 80, No. 6

FIG. 4. Age-speci®c survivorship (mean Ϯ 1 SD of population proportion) of two Upiga virescens cohorts, A (n ϭ 229 individuals) and B (n ϭ 589 individuals), and the proportion of dead larvae observed (n ϭ 49 larvae) in corollas. deaths, and limited fruit set accounted for mortality up the spruce budworm (Morris et al. 1958) and winter to day six for cohorts A and B (Fig. 4). moth (Varley et al. 1973), the senita moth has a rela- Survival of U. virescens remained relatively constant tively high survival rate. from day six until pupation, with ϳ20% of eggs pro- ducing larvae that reach pupation. Of larvae reaching Other ¯ower visitors pupation in cohort B, 12% were parasitized by an en- When ¯owers remained open after sunrise in 1995, doparasitic wasp. Furthermore, of the 21% of areoles they were visited and pollinated by halictid bees (Au- containing a gallery, 17% of larval galleries contained gochlorella sp. and Agapostemon sp.). The primary a wasp pupa or pupal case instead of a U. virescens nocturnal ¯ower visitor was U. virescens, but occa- pupa or pupal case (Table 2). Thus, 12±17% of larvae sionally ¯owers were visited by other , in- that survive to pupation are killed by a parasitic wasp. cluding: Neuroptera (Hemerobiidae), Orthoptera (Gryl- Adult moths were killed by spiders that spin webs lacrididae), Coleoptera (Alleculidae, Cerambycidae, among cactus branches. In summary, ϳ17% of a cohort Cleridae, Tenebrionidae), (Pyralidae), and of U. virescens survived through pupation. In com- Araneae. Flower visitation by organisms other than U. parison to the 1% survival to the adult life stage for virescens was rare: of 19 h and 7660 separate obser-

FIG. 5. Cumulative percentage abortion of senita fruit during the 6 d of fruit abortion fol- lowing ¯owering in two separate studies (Ⅺ, n ϭ 113 fruits; ⅜, n ϭ 197 fruits). After day 6, fruits abscise but do not abort. Larvae may be- gin to enter fruit at day 3, but seed consumption by larvae did not occur before the larval age of 5d. September 1999 POLLINATING SEED-CONSUMER MUTUALISM 2081 vations of ¯owers, only 1.2% of observations yielded populations specialize on one another, (2) interactions insects other than U. virescens. These arthropods be- are mutualistic and apparently coevolved, (3) polli- haved in a manner that would not facilitate pollination, nators oviposit eggs into ¯owers, and (4) pollinators but instead were ¯ower/fruit herbivores and pollen con- actively pollinate ¯owers. However, one substantial sumers. The only other Lepidoptera visiting ¯owers, difference occurs between the senita mutualism and the Cactobrosis fernaldialis (Pyralidae), was uncommon. ®g and yucca mutualisms that relates to the community Adults were not observed probing for nectar, but in- context in which the interaction occurs: absence or stead were only observed sitting on ¯ower petals. Lar- presence of co-pollinators. Co-pollinators are absent vae of C. fernaldialis utilized ϳ3% of senita fruits as from ®g and yucca mutualisms, where fruit set is de- a food resource, but there was a temporal separation pendent upon one or two species of pollinators (Bron- from U. virescens in fruit occupancy. Larvae of C. stein 1987, James et al. 1993, Addicott and Tyre 1995). fernaldialis did not enter fruit prior to day 14, 2±4 d In contrast, co-pollinators are still present, but uncom- after senita larvae entered cactus branches. Larvae of mon, in the senita mutualism and are not completely C. fernaldialis bored into the fruit from the outside and excluded by senita cacti. then consumed all seeds before emerging from the fruit Presence of co-pollinators is a similarity between the to pupate elsewhere. senita mutualism and interactions between Lithophrag- ma and Greya moths. In years or populations where DISCUSSION co-pollinators are abundant, Lithophragma and Greya This study reports a mutualism between the senita interactions are commensal because Greya produce no cactus and senita moth, an active pollinator whose net negative or positive effect on seeds matured in Lith- progeny consume seeds of ¯owers pollinated by the ophragma (Thompson and Pellmyr 1992, Pellmyr and adult. Female senita moths are the third known polli- Thompson 1996). Conversely, when co-pollinators are nator to oviposit an egg onto ¯owers, actively collect absent, rare, or unpredictable, Lithophragma and Greya pollen, and actively pollinate ¯owers during ¯ower vis- interactions may be mutualistic because effects of Gre- itation. Like other Lepidoptera (Thompson and Pellmyr ya on seed set are not masked by co-pollinators 1991, Dempster 1992, Pettersson 1992b), female senita (Thompson and Pellmyr 1992). Whereas the outcome moths avoided ovipositing eggs in ¯owers that already and specialization of Lithophragma±Greya interactions contained an egg (Table 1). Eggs hatched within 3 d depend on the year, population, and community context of ¯ower closing and larvae crawled down the wilting (i.e., co-pollinators) of interactions (Thompson and corolla and bored into the top of fruit, where they con- Pellmyr 1992, Pellmyr and Thompson 1996), the pres- sumed fruit before entering the cactus branch to pupate. ence of co-pollinators has not altered the outcome of Not all fruits pollinated by moths were destroyed by senita cactus and senita moth interactions for different their larvae, because egg and larval survivorship to 6 populations or years of study. This study and another d was low: only a fraction of eggs produced larvae that at Bahia Kino (Fleming and Holland 1998) found senita survived to become seed consumers. Bene®t-to-cost cactus and senita moth interactions to be mutualistic, ratio for Lophocereus schottii interacting with senita and a third study (J. N. Holland and T. H. Fleming, moths was 4.8 (ϳ21% of fruit destroyed by larvae), unpublished data) conducted in south-central Arizona and is consistent with the fact that 21% of areoles con- in 1997 found senita interactions to be mutualistic in tained a larval gallery (Table 2) and that only 20% of three different populations. Thus, in all three studies Upiga virescens survive to day six (Fig. 4). Bene®ts senita moth and senita cactus interactions were mu- of U. virescens to pollination and fruit set in L. schottii tualistic and unaltered by the presence of co-pollina- are thus Ͼ3±4ϫ the costs of seed mortality induced by tors. Presence of effective co-pollinators in commu- larvae, which is comparable with the yucca mutualism nities may prevent specialization of Lithophragma pop- (Addicott 1986, Powell 1992). ulations on pollination by Greya moths. In contrast, presence of effective co-pollinators has not prevented Comparison with other pollination±seed-consumption specialization between L. schottii and U. virescens, be- interactions cause, unlike co-pollinators of Lithophragma (Thomp- The senita and senita-moth mutualism is the sixth son and Pellmyr 1992), co-pollinators of senita cacti known case of interaction of pollination with seed con- generally have been rare and/or temporally unreliable. sumption; other cases include ®gs and ®g wasps, yucca and yucca moths, Trollius europaeus and Chiastocheta Effects of life history traits on maintenance of ¯ies, Lithophragma and Greya moths, and Silene and mutualism Hadena moths. If our ®ndings of the senita mutualism Necessary traits for the evolution of obligate polli- generalize to most years and populations, then senita nation mutualism between yucca and yucca moths in- cactus and senita moth interactions are most ecologi- cluded (1) local host speci®city of pollinators, (2) ovi- cally and evolutionarily similar to the pollinating seed- position into ¯owers, (3) partial seed consumption, and consuming interactions of yucca and ®g mutualisms. (4) behavior that facilitates pollination (Pellmyr et al. In all three mutualisms, (1) both plant and pollinator 1996). The senita mutualism is similar to the yucca 2082 J. NATHANIEL HOLLAND AND THEODORE H. FLEMING Ecology, Vol. 80, No. 6 mutualism in that it exhibits traits necessary for the yucca moth interactions. Furthermore, resource-limited ecological and evolutionary maintenance of an obligate fruit production, a common life history trait of senita, pollination mutualism. Upiga virescens is host speci®c. yucca, and some ®g plants (Bronstein 1988, Pellmyr Our observations of ¯owers and fruits of other cactus and Huth 1994, Herre 1996, Huth and Pellmyr 1997, species have not revealed senita moths or larvae; in Fleming and Holland 1998; J. N. Holland and T. H. contrast, the other pyralid (Cactobrosis fernaldialis) Fleming, unpublished data), is important to survival of that parasitizes senita fruit is not locally host speci®c. larvae but its ecological importance to the mutualistic Its larvae occur in buds and fruit of organ pipe, saguaro, outcome remains unknown and warrants further inves- and senita cacti (Steenbergh and Lowe 1977; J. N. Hol- tigation. land, personal observations). In addition to host spec- i®city, senita moths oviposit into ¯owers and actively On selective abortion and active pollination in pollinate ¯owers, and senita larvae only consume part obligate pollination mutualisms of the seed crop pollinated by adult moths. While U. In the yucca mutualism, selective abortion of fruits virescens is an obligate mutualist, L. schottii is at the containing large egg/larval loads should occur when interface between an obligate and a facultative mu- resources limit fruit set. Selective abortion in yucca tualist because pollination and seed set sometimes re- simultaneously reduces costs to yucca plants and main- sult from co-pollinators. tains moth populations, since other yucca fruits with Number of moth generations per ¯owering season low egg/larva numbers are retained. Selective abortion should re¯ect duration of ¯owering season in special- in yucca appears to be a mechanism inhibiting over- ized pollination mutualisms (Waser et al. 1996). In the exploitation by yucca moths (Pellmyr and Huth 1994, yucca mutualism, ¯owering season ranges from 2±3 d Richter and Weis 1995, Huth and Pellmyr 1997). How- to 3±6 wk with moths having just one generation per ever, senita fruits only contain one larva each; no con- ¯owering season. Senita cacti ¯ower for up to 20 wk tinuum of larval infestation occurs (i.e., high to low and senita moths have several generations per ¯owering egg/larva load per fruit) as in yucca fruit (Pellmyr and season. Yucca and senita mutualisms thus differ in the Huth 1994). Thus, criteria for selection of fruit to abort duration of ¯owering and number of moth generations would have to be presence or absence of a larva in a per season, but they are similar in that within each fruit. If such criteria were the basis for selective abor- mutualism the number of moth generations re¯ects the tion, and senita cacti selectively aborted fruit with a duration of the ¯owering season. The strong association larva, then population interactions would be destabi- between all life stages of U. virescens and senita cacti lized since selectively aborting senita fruit containing should allow for moth population dynamics and ¯ow- a larva would result in moth population extinction. In ering season to be synchronized (Waser et al. 1996). addition, resource-limited fruit abortion occurs up to 6 Along with plant speci®city, mating on plants, ovi- d after ¯ower closing (Fig. 5), and not during the sub- position on ¯owers, partial seed consumption, and pol- sequent 20 d of fruit maturation. During these 6 d, lination behavior (Pellmyr and Thompson 1992, Pell- larvae have just begun to enter fruit and seed con- myr et al. 1996), survivorship is another important life sumption has not yet occurred. Hence, selective abor- history trait that contributes to the maintenance of ob- tion by senita cacti would have to be based on some ligate pollination mutualisms. Because a one-to-one re- factor other than seed consumption. This appears to be lationship occurs between senita larval survival and the case in the yucca mutualism, where yucca fruit the cost to the plant (i.e., fruit/seed mortality), an in- abort prior to seed consumption by larvae (Pellmyr and crease in the survivorship of larvae could alter the net Huth 1994). For the yucca mutualism, this is a paradox outcome of senita cactus and senita moth interactions. because yucca differentially abort fruit with large num- For example, increasing larval survival to day 10 by bers of eggs/larvae but the mechanism for selective 30% (i.e., increased egg hatching and/or reduced co- fruit abortion is not based upon larval seed consump- rolla and fruit abortion deaths) would shift the outcome tion (Pellmyr and Huth 1994). of interactions from mutualism to commensalism or Selective abortion of fruit as a mechanism to reduce predation for that particular cohort of moths. For senita overexploitation in senita moths likely does not occur, and senita moth interactions, it is the high bene®t to but selective abortion of fruit based on pollination qual- plants from pollination by moths and the low survi- ity and quantity is plausible since this provides a strong vorship of moth larvae that maintains the high bene®t- selective pressure for the evolution of active pollination to-cost ratio for the plant. The generality of low sur- (Pellmyr and Huth 1994, Pellmyr et al. 1996). Flowers vivorship as a necessary life history trait in maintaining with greater pollination quality and quantity would be the mutualistic outcome of interactions between seed- preferentially retained by plants where resources limit consuming pollinators and plants is unknown, since fruit production, thus increasing progeny survival of data are lacking on the survivorship of other similar moths that actively pollinate. The three known active mutualists. Nonetheless, we predict that survival is low pollinators, ®g wasps, yucca moths, and senita moths, in egg and early instars of yucca moths, and that in- have specialized morphological and behavioral traits creased survival could alter the outcome of yucca and that increase the effectiveness of pollen collection, September 1999 POLLINATING SEED-CONSUMER MUTUALISM 2083 transport, and delivery to the stigma. Active pollina- Davis, D. R., O. Pellmyr, and J. N. Thompson. 1992. Biology tion, and associated morphological and behavioral and systematics of Greya Busck and Tetragma, new genus (Lepidoptera: ). Smithsonian Contributions to traits, can be selected for in a pollinator population Zoology 524:1±88. when a small number of co-pollinators are present and Dempster, J. P. 1992. Evidence of an oviposition-deterring the behavior does not require any counteradaptation on pheromone in the orange-tip butter¯y, Anthocharis car- the part of the plant (Pellmyr and Huth 1994, Pellmyr damines (L). Ecological Entomology 17:83±85. et al. 1996). Active pollination in senita moths supports Fleming, T. H., and J. N. Holland. 1998. The evolution of obligate pollination mutualisms: senita cactus and senita such a proposition since co-pollinators are present. The moth. Oecologia 114:368±375. senita mutualism appears to be the ®rst documented Fleming, T. H., M. D. Tuttle, and M. A. Horner. 1996. Pol- case of the evolution of active pollination in the pres- lination biology and relative importance of nocturnal and ence of co-pollinators. Reduction in and lack of nectar diurnal pollinators in three species of Sonoran Desert co- production in the senita cactus (Fig. 2) allow for spe- lumnar cacti. Southwestern Naturalist 41:357±369. Gibson, A. C., and K. E. Horak. 1978. Systematic anatomy cialization on senita moths by discouraging co-polli- and phylogeny of Mexican columnar cacti. Annals of the nators that visit ¯owers for nectar rewards, and reduced Missouri Botanical Garden 65:999±1057. nectar production conserves energy for use in fruit pro- Herre, E. A. 1996. An overview of studies on a community duction in populations where fruit set is resource-lim- of Panamanian ®gs. Journal of Biogeography 23:593±607. Huth, C. J., and O. Pellmyr. 1997. Non-random fruit reten- ited (Pellmyr et al. 1996). tion in Yucca ®lamentosa: consequences for an obligate mutualism. Oikos 78:576±584. ACKNOWLEDGMENTS James, C. A., M. T. Hoffman, D. C. Lightfoot, G. S. Gorbes, We thank the Mexican SEDESOL for permission to conduct and W. G. Whitford. 1993. Pollination study of Yucca research in Mexico and several landowners for allowing us elata: an experimental study of a mutualistic association. to work on their property. We are grateful for excellent as- Oecologia 93:512±517. sistance provided by J. Bier, K. Conway, C. Maley, and K. Janzen, D. H. 1979. How to be a ®g. Annual Review of Pearson. R. Burgess, J. Donahue, R. Leuschner, and M. Mc- Ecology and Systematics 10:13±51. Intosh assisted in identi®cation of insects. We also thank J. Kjellberg, F., P.-H. Gouyon, M. Ibrahim, M. Raymond, and F. Addicott, W. S. Armbruster, O. Pellmyr, J. N. Thompson, G. Valdeyron. 1987. The stability of the symbiosis be- and an anonymous reviewer for discussion and comments on tween dioecious ®gs and their pollinators: a study of Ficus this paper. Research was supported in part by NSF grant DEB- carica L. and Blastophaga psenes L. Evolution 41:693± 9421673. 704. Lindsay, G. 1963. The genus Lophocereus. Cactus and Suc- LITERATURE CITED culent Journal 35:176±192. Addicott, J. F. 1986. Variation in the costs and bene®ts of Machado, C. A., E. A. Herre, S. McCafferty, and E. Ber- mutualism: the interaction between and yucca mingham. 1996. Molecular phylogenies of ®g pollinating moths. Oecologia 70:486±494. and non-pollinating wasps and the implications for the or- Addicott, J. F., J. Bronstein, and F.Kjellberg. 1990. Evolution igin and evolution of the ®g±®g wasp mutualism. Journal of mutualistic life-cycles: yucca moths and ®g wasps. Pages of Biogeography 23:531±542. 143±161 in F. Gilbert, editor. life cycles: genetics, Miles, N. J. 1983. Variation and host speci®city in the yucca evolution, and co-ordination. Springer-Verlag, New York, moth, Tegeticula yuccasella (Incurviidae): a morphometric New York, USA. approach. Journal of the Lepidopterists' Society 37:207± Addicott, J. F., and A. J. Tyre. 1995. Cheating in an obligate 216. mutualism: how often do yucca moths bene®t yuccas? Oi- Morris, R. F., C. A. Miller, D. O. Greenbank, and D. G. Mott. kos 72:382±294. 1958. The population dynamics of the spruce budworm in Aker, C. L., and D. Udovic. 1981. Oviposition and polli- eastern Canada. Proceedings of the Tenth International nation behavior of the yucca moth, Tegeticula maculata Congress of Entomology 4:137±149. (Lepidoptera: Prodoxidae), and its relation to the repro- Munroe, E. 1972. In R. B. Dominick et al. The moths of ductive biology of Yucca whipplei (Agavaceae). Oecologia America north of Mexico. Fascicle 13.1B, Pyraloidea (in 49:96±101. part). Anstett, M. C., M. Hossaert-McKey, and F. Kjellberg. 1997. Parker, K. C. 1989. Height structure and reproductive char- Figs and ®g pollinators: evolutionary con¯icts in a coe- acteristics of senita, Lophocereus schottii (Cactaceae), in volved mutualism. Trends in Ecology and Evolution 12: southern Arizona. Southwestern Naturalist 34:392±401. 94±99. Pellmyr, O. 1989. The cost of mutualism: interactions be- Bronstein, J. L. 1987. Maintenance of species-speci®city in tween Trollius europaeus and its pollinating parasites. Oec- a neotropical ®g-pollinator wasp mutualism. Oikos 48:39± ologia 78:53±59. 46. . 1992. The phylogeny of a mutualism: evolution and . 1988. Limits to fruit production in a monoecious coadaptation between Trollius and its seed-parasitic pol- ®g: consequences of an obligate mutualism. Ecology 69: linators. Biological Journal of the Linnean Society 47:337± 207±214. 365. . 1992. Seed predators as mutualists: ecology and . 1997. Pollinating seed eaters: why is active polli- evolution of the ®g/pollinator interaction. Pages 1±44 in nation so rare? Ecology 78:1655±1660. E. A. Bernays, editor. Insect±plant interactions. Volume IV. Pellmyr, O., and C. J. Huth. 1994. Evolutionary stability of CRC Press, Boca Raton, Florida, USA. mutualism between yuccas and yucca moths. Nature 372: . 1994. Conditional outcomes in mutualistic inter- 257±260. actions. Trends in Ecology and Evolution 9:214±217. Pellmyr, O., and J. N. Thompson. 1992. Multiple occur- Davis, D. R. 1967. A revision of the moths of the subfamily rences of mutualism in the yucca moth lineage. Proceedings Prodoxinae (Lepidoptera: Incurvariidae). United States of the National Academy of Sciences, USA 89:2927±2929. Natural History Museum Bulletin 255:1±170. Pellmyr, O., and J. N. Thompson. 1996. Sources of variation 2084 J. NATHANIEL HOLLAND AND THEODORE H. FLEMING Ecology, Vol. 80, No. 6

in pollinator contribution within a guild: the effects of plant Shreve, F., and I. L. Wiggins. 1964. Vegetation and ¯ora of and pollinator factors. Oecologia 107:595±604. the Sonoran Desert. Volumes 1 and 2. Stanford University Pellmyr, O., J. N. Thompson, J. M. Brown, and R. G. Har- Press, Stanford, California, USA. rison. 1996. Evolution of pollination and mutualism in the Steenbergh, W. F., and C. H. Lowe. 1977. Ecology of the yucca moth lineage. American Naturalist 148:827±847. saguaro. II. Reproduction, germination, establishment, Pettersson, M. W. 1991a. Flower herbivory and seed pre- growth, and survival of the young plant. National Park dation in Silene vulgaris (Caryophyllaceae): effects of pol- Service Scienti®c Monograph Series 8:1±242. lination and phenology. Holarctic Ecology 14:45±50. Thompson, J. N. 1997. Evaluating the dynamics of coevo- . 1991b. Pollination by a guild of ¯uctuating moth lution among geographically structured populations. Ecol- populations: option for unspecialization in Silene vulgaris. ogy 78:1619±1623. Journal of Ecology 79:591±604. Thompson, J. N., and O. Pellmyr. 1991. Evolution of ovi- . 1992a. Taking a chance on moths: oviposition by position behavior and host preference in Lepidoptera. An- Delia ¯avifrons (Diptera: Anthomyiidae) on the ¯owers of nual Review of Entomology 36:65±89. bladder campion, Silene vulgaris (Caryophyllaceae). Eco- Thompson, J. N., and O. Pellmyr. 1992. Mutualism with logical Entomology 17:57±62. pollinating seed parasites amid co-pollinators: constraints . 1992b. Density-dependent egg dispersion in ¯owers on specialization. Ecology 73:1780±1791. of Silene vulgaris by the seed predator Hadena confusa Varley, C. G., G. R. Gradwell, and M. P.Hassell. 1973. Insect (). Ecological Entomology 17:244±248. population ecology: an analytical approach. University of Polis, G. A., R. D. Holt, B. A. Menge, and K. O. Winemiller. 1996. Time, space, and life-history: in¯uences on food California Press, Berkeley, California, USA. webs. Pages 435±460 in G. A. Polis and K. O. Winemiller, Waser, N. M., L. Chittka, M. V. Price, N. M. Williams, and editors. Food webs: integration of patterns and dynamics. J. Ollerton. 1996. Generalization in pollination systems, Chapman and Hall, New York, New York, USA. and why it matters. Ecology 77:1043±1060. Powell, J. A. 1992. Interrelationships of yuccas and yucca West, S. A., E. A. Herre, D. M. Windsor, and P. R. S. Green. moths. Trends in Ecology and Evolution 7:10±15. 1996. The ecology and evolution of the New World non- Richter, K. S., and A. E. Weis. 1995. Differential abortion pollinating ®g wasp communities. Journal of Biogeography in the yucca. Nature 376:557±558. 23:447±458. Riley, C. V. 1892. The yucca moth and yucca pollination. Wiebes, J. T. 1979. Co-evolution of ®gs and their insect Third Annual Report of the Missouri Botanical Garden:99± pollinators. Annual Review of Ecology and Systematics 10: 159. 1±12.