OIKOS 94: 433–441. Copenhagen 2001

Effects of epiphytic lichens on host preference of the vascularepiphyte Tillandsia usneoides

Ragan M. Callaway, Kurt O. Reinhart, Shirley C. Tucker and Steven C. Pennings

Callaway, R. M., Reinhart, K. O., Tucker, S. C. and Pennings, S. C. 2001. Effects ofepiphytic lichens on host preference of the vascular epiphyte Tillandsia usneoides. –Oikos 94: 433–441.

We investigated the potential for nonvascular epiphytic species (primarily lichens) toaffect the quality of different host tree species for the vascular epiphyte Tillandsiausneoides in the southeastern USA. Different host tree species had substantiallydifferent abundances of Tillandsia, and these abundances were correlated with thecomposition of nonvascular epiphyte communities. In greenhouse experimentsTillandsia grew significantly faster on the branches of Quercus �irginiana (a specieswith very high natural abundances of Tillandsia) when the dominant lichen on Q.�irginiana was left intact than when the lichen was removed from the branches. Inlaboratory experiments, extracts from Cryptothecia rubrocincta, a lichen that was 10times more common on poor host species for Tillandsia than on good host species,reduced Tillandsia seedling survival and growth in comparison to extracts from otherspecies and rainwater. In field experiments, lichens increased the proportion ofTillandsia seeds and vegetative strands that adhered to the trunk of Ilex opaca (apoor Tillandsia host), but lichens did not affect propagule adherence to Q. �irginiana.Our results are by no means exhaustive of the possibilities, but they suggest that thestructure and diversity of vascular and nonvascular epiphytic communities that growin different tree species may not be simply the product of host tree characteristics, butmay also be influenced by interactions among the epiphytes themselves.

R. M. Callaway and K. O. Reinhart, Di�. of Biological Sciences, Uni�. of Montana,Missoula, MT 59812, USA (callaway@selway.umt.edu). – S. C. Tucker, E�olution,Ecology, and Marine Biology, Uni�. of California, Santa Barbara, CA 93106, USA. –S. C. Pennings, Uni�. of Georgia Marine Inst., Sapelo Island, GA 31327, USA.

Direct interactions among species may be altered bystrong indirect interactions involving additional species.Indirect interactions involving consumers have beenwell documented, but we know less about indirectinteractions that occur within plant communities. Mostresearch to date has shown that within a community ofinteracting plants, direct pair-wise interactions can bemodified by other species if they disproportionatelysuppress one of the competitors in the direct pair-wiseinteractions (Holt 1977, Connell 1991, Stone andRoberts 1991, Miller 1994, Wooton 1994). In otherwords, if species A, B, and C form a competitive ‘‘web’’with A as the best competitor and C as the worst,species A may indirectly facilitate species C by sup-

pressing species B. This competitive web is the classicform of indirect interaction, but competing plants mayalso interact indirectly through ‘‘diffuse’’ effects. Dif-fuse effects are the cumulative individual effects ofmany different species acting together on a single spe-cies at the same time (Davidson 1980, Wilson andKeddy 1986a, b, Vandermeer 1990). Despite the poten-tial importance of indirect interactions for understand-ing community organization, there have been fewexperimental investigations of indirect effects amongplants. Those experiments that have been conductedprovide a powerful conceptual framework for how indi-rect interactions may affect species abundances, coexis-tence and diversity in plant communities (Wilson and

Accepted 20 March 2001

Copyright © OIKOS 2001ISSN 0030-1299Printed in Ireland – all rights reserved

OIKOS 94:3 (2001) 433

Keddy 1986a, Miller 1994, Levine 1999, Callaway andPennings 2000).

In previous experiments, we examined the direct ef-fects of different host tree species on two vascularepiphyte species in a coastal maritime forest in thesoutheastern United States (Callaway et al. unpubl.).All epiphytic species are by definition facilitated by thetree species that provide them with habitat, but wefound that the vascular epiphytes, Tillandsia usneoidesL. (Spanish moss) and Polypodium polypodioides L.(resurrection fern), were much more common on somehost species than on others. The growth rates ofTillandsia strands that were experimentally transplantedonto these host tree species were higher on the hostspecies on which they occurred most frequently innature, suggesting that host tree may have importantdirect effects on their epiphytes. However, during theseexperiments we also observed substantial differences inthe composition of epiphytic lichen and nonvascularplant communities between good and bad hosts forTillandsia. Others have quantified differences in lichenspecies on different host tree or shrub species (Olsen1917, Billings and Drew 1938, Longan et al. 1999),different interactions among epiphytic and rock-inhab-iting lichens (Armesto and Contreras 1981, Armstrong1982, Woolhouse et al. 1985, Dale and John 1999), andnegative effects of lichens on vascular plant (Lawrey1986, 1995), raising the possibility that host trees mightalso affect Tillandsia indirectly through the lichen com-munities that develop on their branches. If epiphyticlichen communities that develop on preferred host treespecies differ in their effects on Tillandsia, or othervascular epiphytes, this raises the possibility of yet-unidentified types of indirect interaction webs; those inwhich direct facilitative interactions (host tree effects onTillandsia and lichens which depend on the habitatprovided by trees for their existence) lead to eitherindirect facilitative interactions (if some lichens benefitTillandsia), or to indirect negative interactions (if somelichens harm Tillandsia).

We hypothesized that the direct positive effect of treehosts on lichens (providing habitat) could result ineither positive or negative indirect effects of the treehosts on vascular epiphytes. Vascular epiphytes live inlow-nutrient environments (Benzing 1974, 1980), andepiphytic lichens could facilitate plant epiphytes bycontributing nitrogen and other nutrients to thethroughfall from canopies (Schlesinger and Marks1977, Callaway and Nadkarni 1991, Knops et al. 1991,1996). In addition, lichens might facilitate vascularepiphytes by providing rough, adherent substrate onwhich colonization by vegetative or reproductivepropagules is enhanced. Alternatively, lichens mightnegatively affect vascular epiphytes through the pro-duction of harmful secondary metabolites (Lawrey1986, 1995). We investigated the potential roles thatepiphytic lichens might play in the overall effect of host

trees on vascular epiphytes by 1) quantifying differencesin lichen community composition in 10 different hosttree species, 2) correlating lichen community composi-tion with the abundance of Tillandsia usneoides indifferent host tree species, 3) experimenting with theeffects of selected lichen species on the growth ofTillandsia usneoides, and 4) experimenting with theeffects of lichens on the ability of T. usneoides to attachto host tree species.

Methods

Study site

All work was done on the southern end of SapeloIsland, Georgia, USA (31°25� N, 81°16� W), within theSapelo Island National Estuarine Research Reserve andan adjacent area administered by the Department ofNatural Resources as the ‘‘Natural Area’’. Sapelo Is-land (ca 7000 ha) is a Pleistocene barrier island withsandy soils. Average annual rainfall is ca 130 cm, and isconcentrated in the late summer (July–September). Theclimate is subtropical, with hot humid summers andmild winters. Average low temperature in January is ca4.5°C (40°F), and hard freezes are rare.

Tillandsia host preference

We quantified host-epiphyte associations by surveying9–27 haphazardly selected individuals between 20 and40 cm DBH (diameter at breast height) of each of eighttree species in a �20-km2 area on the south end ofSapelo Island (also see Callaway et al. unpubl.). Sam-pling was constrained by the natural distribution pat-terns of host trees, but we interspersed samples ofdifferent tree species as much as possible to avoid anypossible bias due to local microclimate differences.Each tree was assigned an index of abundance of eachepiphyte species on a scale from 0 to 10 by visuallydividing the canopy into 10ths, much like standardsurveys used to rank dwarf mistletoe infection(Hawksworth and Wiens 1972). A rank of 10 meantthat the epiphyte occurred in all 10 subsections of thecanopy, a rank of 1 meant that only one section wasoccupied, even by as little as one individual epiphyte,and a rank of 0 meant that no epiphyte was visible.

We also examined host quality by transplantingTillandsia and Polypodium onto seven host species:Celtis lae�igata, Quercus �irginiana, Juniperus �irgini-ana, Liquidambar styraciflua, Quercus nigra, Ilex opaca,and Magnolia grandifolia (also see Callaway et al. un-publ.). Tillandsia clumps were collected from a singledead Q. �irginiana and separated into single strands ofroughly equal lengths (�25 cm). Single strands weretransplanted onto each of 15 individuals of each host

434 OIKOS 94:3 (2001)

tree species. Before transplanting we measured the totallength and number of nodes of each strand. Tillandsiastrands were fastened to branches 3–7 m above theground using plastic straps. All transplants were out-planted in late December 1997 and harvested in earlyAugust 1998. Total strand length and node numberwere re-measured for Tillandsia and host effects ongrowth were analyzed using one-way ANOVA.

Lichen communities

Lichens were quantified on lower branches that rangedfrom 2 to 4 cm in diameter on the same eight host treespecies sampled for Tillandsia abundance. Lichens andnonvascular epiphytes were identified to species whenpossible or to genus. On each of 15–18 individuals foreach host species, we randomly chose one branch be-tween 3 and 3.5-cm diameter for sampling lichens. Oneach branch we sampled 30–40 cm along the length ofbranch. On the upper surfaces of these branches werecorded the proportional cover of all lichen species,bryophytes and ‘‘green algae’’ (see page 9). Cover wasquantified by placing line transects along the uppersurface of the branch and parallel to the branch, andrecording the percentage of each transect occupied byeach nonvascular taxon (line-transect method). Usinghost species to group our samples (plots) into classes,we used these data to develop a species by samplesmatrix and analyzed this matrix with detrended corre-spondence analyses (DCA; Hill 1979) in the multivari-ate package PC-ORD (McCune 1997).

Growth experiments with Parmotrema

The thallose lichen species, Parmotrema tinctorum andP. rigidum, were much more common on hosts withabundant Tillandsia than on hosts with low levels ofTillandsia. Therefore, we experimented with the effectsof these Parmotrema species (the species were oftenmixed on the branches and therefore not separated inour experiments) on Tillandsia growth using branchescut from Q. �irginiana that were covered (�90%) withParmotrema and relocated to a greenhouse. Branchlengths ranged from 20 to 46 cm and diameters rangedfrom 2.4 to 7.0 cm. Branches were arranged into 13pairs based on similar lengths and diameters, and alllichens were manually removed from one randomlychosen branch of each pair.

Branches were placed in an open-air greenhouse(plastic roof but no walls) and attached to a woodenframe so that they extended upwards at a 75-degreeangle. Tillandsia was collected from a single dead Q.�irginiana in January 1998. Individual strands wereselected that were undamaged and healthy in appear-ance, lacked algae on the leaf surfaces, and had at least

seven nodes. Three Tillandsia strands were attachedapproximately 5 cm apart to each branch using plasticcable ties. Before starting the treatments we measuredthe length of each Tillandsia strand from the basalportion of the stem to the last node (apex).

The branches and Tillandsia were watered three timesper week with rainwater beginning on 28 January 1998.We sprayed 500 ml of rainwater onto the surface ofeach branch so that water flowed over the branches(with or without their cover of Parmotrema) and ontothe Tillandsia strands. From 29 May through 19 June1998 (the end of the experiment) watering was donewith distilled water because little rainwater was avail-able. At the end of the experiment the length of eachplant was measured from the basal portion of the stemto the last node (apex). Measures of new growth for thethree Tillandsia strands on each branch were averagedand a paired t-test was used to compare the mean newgrowth on each of the 13 branches with and withoutlichens.

Lichen extracts and the growth and survival ofTillandsia

Parmotrema tinctorum and P. rigidum (in a mixture asabove) and Pyxine caesiopruinosa were collected fromthe branches of Quercus �irginiana. These lichen specieswere common on host tree species with abundantTillandsia. Cryptothecia rubrocincta was collected fromthe trunks of Ilex opaca and ‘‘green algae’’ from thebranches of Magnolia grandiflora. ‘‘Green algae’’ con-sisted of a complex mixture of taxa that could not beseparated for experiments nor in field measurements ofbranch cover, but microscopic analyses indicated thatour category ‘‘green algae’’ was dominated by severalspecies of coccoid green algae and included one leafyliverwort species. ‘‘Green algae’’ were common on treespecies with sparse Tillandsia. We used extracts fromthese four taxa and rainwater in experiments withnewly germinated Tillandsia seedlings. Lichen and algalextracts were produced by scraping lichens and algaefrom branches in the field and drying them at 25°C.Care was taken to damage the epiphyte tissue as littleas possible and to avoid collecting tree bark. Every twoweeks we would filter 10 ml of distilled water through 1g dry weight of lichen or algae to produce new batchesof extract for watering Tillandsia seedlings. Tillandsiaseeds were initially germinated on filter paper in Petridishes using distilled water. After seedlings were 2–4mm long, they were randomly sorted into five sets andplaced in new Petri dishes. We used five Petri dishes foreach set and placed 20 seedlings into each Petri dish.Petri dishes containing Tillandsia seedlings were placedin a growth chamber with a 14/10 hour day/night cycle,at 30°/20°C, and watered daily. After 80 d survivingseedlings were counted and measured for total leaflength.

OIKOS 94:3 (2001) 435

Lichen strand and seed adherence

In the field, we observed far more small Tillandsiastrands caught on lichen-covered bark of smooth-barked tree species than on bare bark surfaces of thesame host species. Therefore, we conducted trials withTillandsia seeds and vegetative strands in which wemeasured adherence on lichen-covered bark and barkwithout lichens on Ilex opaca (15–20 cm DBH) andQuercus �irginiana (25–30 cm DBH), the host specieswith the smoothest and roughest bark, respectively. Weheld five-seed clusters against the trunks at DBH, re-leased them, and counted how often they adhered. Foreach host tree species we chose 10 individual trees withvery little lichen cover on the trunks and 10 individualswith large amounts of lichens on the trunks. We thenreleased seed clusters 10 times for each tree from ran-dom points around the circumference of the trunk, butalways at DBH. The number of clusters that stuck tothe tree for more than 10 s was recorded. We alsomeasured the ability of Tillandsia fragments to adhereto these same hosts by pressing a 25-cm strand ofTillandsia against tree trunks at DBH and measuringthe length of the strand that remained attached to thetree for at least 10 s after release of the strand.

Results

Tillandsia host preference

Tillandsia was highly abundant on Celtis and Q. �irgini-ana, moderately abundant on Juniperus and Liq-uidambar, and uncommon on the other species (Fig. 1).

Field experiments indicated that vascular epiphytesgrew significantly faster on that host species that har-bored the most epiphytes (Fig. 1). Tillandsia growthrates were higher on Celtis, Q. �irginiana, Juniperus,Liquidambar, and Q. nigra than on Ilex and Magnolia.For one host species, Ilex, Tillandsia actually decreasedin size. Based on lichen abundance and growth rates oflichens in experimental treatments, we subjectively di-vided the eight tree species into three groups, ‘‘good’’Tilandsia hosts, intermediate hosts, and ‘‘bad’’ hosts.

Nonvascular communities

Lichens constituted over 85% of the vegetative cover onthe branch surfaces of the 8 other tree species that westudied (Table 1). Mosses comprised less than 1% of thecover and ‘‘green algae’’ (see above) 13% of the coveron branch surfaces. The hepatic genus, Frullania, wasrare but occurred on branches of all three groups ofhost tree species. Only two lichen species in our sample,a sterile white crustose species and Graphis afzelii, wererestricted to one group of host species; the intermediatehosts and the poor hosts, respectively. Both of thesespecies were rare. However, the centroids and 95% C.I.sfor samples in the DCA demonstrate that the overallcomposition of nonvascular communities differed sub-stantially among three groups of host trees organizedon the basis of Tillandsia abundance (Table 1, Fig. 2).The group of tree species that harbored large amountsof Tillandsia (Q. �irginiana, Juniperus, and Celtis) alsohad significantly higher abundances of the large folioseParmotrema species and Pyxine caesiopruinosa and sig-nificantly lower abundances of Rinodina applanata,Cryptothecia rubrocincta, Gyrostomum scyphuliferum,Trypethelium spp., and green algae than the group ofspecies that had low or no amounts of Tillandsia. Theonly lichen species that was more abundant on thegroup of trees with intermediate levels of Tillandsia (Q.nigra, Acer, and Liquidambar) was Graphina spp., but itwas 2–5 times more common on tree species in theintermediate group than on those in the other twogroups. Bare bark constituted about half of the brancharea sample for all three host groups and did not differamong the groups.

Growth experiments with Parmotrema and lichenextracts

The growth of Tillandsia on Q. �irginiana branch seg-ments from which Parmotrema had been removed was19.8% less than growth of Tillandsia on branches withthe natural abundance of Parmotrema present (Fig. 3).Tillandsia seedlings that were watered with extractsfrom Cryptothecia rubrocincta, a species common onpoor Tillandsia hosts, had significantly lower growth

Fig. 1. Ranked abundance and new growth of Tillandsiausneoides on host tree species on Sapelo Island, Georgia. Errorbars show one standard error and different letters within adependent variable indicate significant differences at P�0.05,post-ANOVA Tukey HSD tests. Fgrowth=7.58; df=7,118;P�0.001. Frank=24.1; df=7,156; P�0.001. Host specieshave been ranked subjectively into three groups for compari-sons with nonvascular epiphyte communities and experiments.

436 OIKOS 94:3 (2001)

Table 1. Cover of nonvascular epiphytes (�one standard error) on tree species that vary in the natural abundance of thevascular epiphyte, Tillandsia usneoides, in their canopies. The ‘‘good’’ hosts were Celtis lae�igata, Quercus �irginiana, andJuniperus �irginiana ; intermediate hosts were Quercus nigra, Acer rubrum, and Liquidambar styraciflua ; and ‘‘bad’’ hosts wereMagnolia grandifolia and Ilex opaca (holly). Bold type emphasizes significantly higher abundances.

Bad hostsGood hosts Intermediate hosts(n=30)(n=44) (n=46)

Parmotrema spp.1 22.3�2.9a 3.7�2.9b 0.9�0.4b

Pyxine caesiopruinosa 6.9�1.6a 1.5�0.6b3.3�0.8b

Phaeographis lobata 3.3�0.8a 1.6�0.4a 0b

Pertusaria amara 0.7�0.5a 0a0.6�0.3a

Moss2 0.3�0.3a0.7�0.4a 0.2�0.1a

Graphina spp. 4.9�1.0c 9.0�1.7b28.5�4.0a

Unknown3 0a0a 0.6�0.6a

Frullania spp.4 1.6�0.5a 0.5�0.2a1.4�0.4a

‘‘Green algae’’5 20.1�5.0a2.7�1.1b 1.4�0.9b

Trypethelium spp. 0.7�0.4b 6.4�1.8a1.9�0.7b

Rinodina applanata 0.9�0.3b 2.0�0.6a 4.8�1.6a

Cryptothecia rubrocincta 0.5�0.3b 5.6�1.4a2.8�0.8a

Gyrostomum scyphuliferum 0.1�0.1b 1.9�0.7a 3.2�0.9a

Graphis afzelii 0a 0a 1.0�0.7a

Bare bark 51.6�9.8a 46.7�11.0a50.8�6.7a

1 primarily Parmotrema tinctorum and P. rigidum, but included several other species in the genus that were difficult todistinguish. 2 included several similar species. 3 unknown lichen species consisting of sterile white crust. 4 we were unable toidentify the species. 5 ‘‘green algae’’ was dominated by several coccoid species and included one leafy liverwort species.

and survival than did those watered with extracts fromParmotrema, Pyxine caesiopruinosa, ‘‘green algae’’, orwith rainwater (Fig. 4).

Lichen strand and seed adherence

Ilex opaca trunks with abundant lichen cover caught farmore vegetative and reproductive propagules ofTillandsia than trunks that had no lichen cover (Fig. 5).No Tillandsia seeds adhered to Ilex trunks withoutlichens, but 85% of all seeds released on lichen-coveredtrunks successfully adhered. Similarly, adherence ofvegetative strands of Tillandsia was four times higheron Ilex trunks with lichens than on bare trunks. Ourobservation during the experiment was that Tillandsiapropagules consistently caught directly on the lichensthat were on the trunks of Ilex. In contrast to theresults for Ilex, there was no significant effect of lichencover for seed or vegetative adherence on the trunks ofthe rough-barked Quercus �irginiana.

Discussion

Our results show that lichen and nonvascular epiphytecommunities differed substantially in compositionamong tree species on Sapelo Island, and suggest thatdifferent epiphytic lichen species occurring on differenthost tree species have the potential to indirectly feed-back in both positive and negative ways to the distribu-tion and abundance of the vascular epiphyte Tillandsiausneoides. The potential for indirect positive feedbackswas suggested by: 1) tree species that harbored largeamounts of Tillandsia also hosted abundant Par-

motrema, and 2) in greenhouse experiments Par-motrema stimulated the growth of Tillandsia fragments.The possibility of an indirect negative feedback wassuggested by: 1) tree species that harbored little or noTillandsia also had abundant amounts of the lichenspecies Rinodina applanata, Cryptothecia rubrocincta,

Fig. 2. Detrended correspondence analysis (DCA) of nonvas-cular epiphyte communities sampled on eight different hosttree species (pine hosts had no lichens). Closed circles repre-sent good host tree species for the vascular epiphyte Tillandsiausneoides, open circles represent intermediate quality host spe-cies, and closed triangles represent poor host species. The largesymbols represent the mean for each group on the x- andy-axes and are shown with 95% confidence limits in eachdirection. The names of the four nonvascular epiphytes used inthe extract experiment (Fig. 3) are shown at their x- and y-axiscoordinates. The x-axis is negatively correlated with the meanTillandsia rank on each host species, r= −0.75; P�0.01.Eigenvalues were 0.79 for the x-axis and 0.48 for the y-axis.MDS was used for comparison and showed highly similarpatterns.

OIKOS 94:3 (2001) 437

Fig. 3. New growth of Tillandsia usneoides tillers attached toexcised branches of Quercus �irginiana in the greenhouse eitherwith the lichen Parmotrema intact or removed, n=13 pertreatment. Error bars show one standard error. t-test, t=2.98;df=1,12; P=0.21.

ana. The strong preferences of lichen species for differ-ent host tree species in our study may be quite local anddepend on particular environments. For example,Graphis afzelii was highly restricted in distribution onSapelo Island, but Tucker (1979) found that in southLouisiana, G. afzelii occurred on at least nine tree hostspecies, primarily species that were smooth-barked.

Our results suggest that tree species may indirectlyaffect vascular epiphytes in their canopies via theireffects on the nonvascular epiphytes in the same

Fig. 4. Total leaf length and survival of Tillandsia usneoidesseedlings watered with extracts from Cryptothecia rubrocinctaand ‘‘green algae’’ (common on ‘‘bad’’ Tillandsia host treespecies), a mixture of Parmotrema tinctorum and P. rigidum,and Pyxine caesiopruinosa (latter two common on ‘‘good’’Tillandsia host tree species), n=5 groups of 20 seedlings pertreatment. Error bars show one standard error and differentletters indicate significant differences at P�0.05, post-ANOVA Tukey HSD tests. One-way ANOVA, F=12.2; df=4,29; P�0.001.

Gyrostomum scyphuliferum, Trypethelium spp., and‘‘green algae’’, and 2) extracts from Cryptothecia ru-brocincta reduced the survival and growth rates ofTillandsia seedlings in the laboratory. Allelopathic ef-fects of lichens have been reported for other species(Rundel 1978, Lawrey 1995). Host tree species them-selves have strong direct effects on vascular epiphytesthrough bark traits, microclimate effects, or throughfallchemistry (Frei and Dodson 1972, Benzing 1974,Schlesinger and Marks 1977, Hietz and Hietz-Seifert1995, Kernan and Fowler 1995, Talley et al. 1996, Hietzand Briones 1998), but our results suggest that differentnonvascular species also have the potential to affecthost preference for vascular epiphytes.

The direct effects of host tree species on epiphyticlichen communities may be mediated through differ-ences in the ways that canopies transmit light (Kershaw1985, Lucking 1999). For example, the occurrence ofCryptothecia in Louisiana, USA is limited by its prefer-ence for shade (Tucker 1979), which fits our findingthat this species was common on Magnolia and Ilex,species with exceptionally dark interior canopies. Rela-tively few species of lichens appear to be able to growon trunks of evergreen dicot trees, especially in deepforest, and there are only perhaps five or six that growwell on Magnolia grandiflora trunks. Kershaw (1985)also noted that a cover of green algae also is typical ofbranches and trunks in low light in deep forests. Par-motrema species, on the other hand, tolerate high levelsof light well and are usually found on exposed trunksand branches, on the edges of forest or in exposedsituations such as the more open canopies of Q. �irgini-

438 OIKOS 94:3 (2001)

Fig. 5. Adherence of Tillandsia usneoides seeds and vegetativestrands to the trunks of Ilex opaca (smooth barked and ‘‘bad’’Tillandsia host) and Quercus �irginiana (rough barked and‘‘good’’ Tillandsia host) either with or without lichens. Errorbars represent one standard error.

cular epiphytes. Other experiments on the direct effectsof host tree species on Tillandsia indicated that strongspecies-specific relationships were highly correlated withthe water-holding capacity of the host tree’s bark(Callaway et al. unpubl.), which is also unlikely to behighly affected by nonvascular epiphytes.

Indirect interactions involving consumers are welldocumented (Paine 1966, Lubchenco 1978, Kerfoot andSih 1987, Wooton 1994, Pennings and Callaway 1996),but progress toward understanding indirect interactionsamong competitors has been primarily theoretical untilrecently (MacArthur 1972, Pianka 1974, Buss and Jack-son 1979, Case 1991, Stone and Roberts 1991, Millerand Travis 1996). However, several recent studies alsopoint to the importance of indirect interactions amongcompeting plants. Miller (1994) quantified direct andindirect effects among five old-field plant species andconsistently found important indirect positive effects.Levine (1999) experimentally demonstrated that Carexnudata, a riparian sedge, had a strong indirect effect onthe liverwort Conocephalum conicum by reducing theabundance of Mimulus guttatus, a strong competitor ofConocephalum. Other experimental manipulations ofwetland plant communities suggest that indirect diffuseeffects are important, but depend on community char-acteristics (Wilson and Keddy 1986a). Callaway (1994)and Callaway and Pennings (2000) demonstrated thatthe saltmarsh shrub Arthrocnemum subterminale di-rectly competed with the winter annual species Spergu-laria marina. However, when the combined effects ofArthrocnemum and the much stronger competitor Mo-nanthechloe littoralis were tested, the effect of Arthroc-nemum was facilitative, ameliorating the negative effectof Monanthechloe on Spergularia. Li and Wilson (1998)found that intraspecific interactions among Symphori-carpos occidentalis shrubs was competitive unlessgrasses were also present. Apparently the combinedeffects of several Symphoricarpos individuals increasedtheir individual ability to compete with, or withstandcompetition from, the grasses.

Most of these previous studies of indirect interactionsamong species within a trophic level boil down concep-tually to demonstrating that ‘‘an enemy’s enemy is afriend’’, but our results are somewhat different. In thecase of Parmotrema, the beneficial direct effects of Q.�irginiana on Tillandsia may be enhanced indirectly byQ. �irginiana also facilitating Parmotrema, thereforecreating a chain of positive effects. To our knowledgethere have been no previous studies that suggest thesekinds of linked positive interactions among autotrophicorganisms, but there is correlative and experimentalevidence for facilitation among epiphytic and rock-in-habiting lichens (Woolhouse et al. 1985; Stone 1989).The presence of Cryptothecia rubrocincta, however, mayadd indirectly to direct negative effects of poor hosttree species (e.g. Ilex and Magnolia) or counteract anydirect positive effects that these host species may have.

canopies, but there are several caveats. First, our hy-pothesis that trees have indirect effects on vascularepiphytes mediated through their direct effects on non-vascular epiphytes rests on the untested assumptionthat the trees in fact have direct effects on the nonvas-cular species. The associations that we documentedamong tree and lichen species were correlative, and wehave not demonstrated that direct biological interac-tions drive the development of different epiphytic lichencommunities. We consider such direct effects, probablymediated through different bark traits, to be the mostlikely explanation for the existence of different lichencommunities on different tree species; however, it isconceivable that host-specific lichen communities mightinstead be the product of different microclimates inwhich different tree species occur. We examined hosttree species that were intermixed spatially as much aspossible (Callaway et al. unpubl.), but host trees werenot fully interspersed. Second, although our experi-ments with branches cut from Q. �irginiana (establish-ing an indirect positive feedback) represent reasonablyrealistic conditions for interactions among lichens andTillandsia, our experiments with lichen and algal ex-tracts did not mimic field conditions and extracts mayhave been more or less potent than natural solutions inthe field. Abruptly rewetting lichens as we did to pro-duce our extracts may have damaged membranes andcreated unrealistic extracts, but abrupt rewetting eventsappear to be quite common at our field site. Third, wehave not evaluated the relative importance of the directeffects of hosts on Tillandsia versus the effects of thelichens that grow on the hosts. For example, pines werevery poor hosts for Tillandsia (Callaway et al. unpubl.)and we found virtually no nonvascular epiphytes on thepines, indicating that the poor suitability of pines ashosts for vascular epiphytes is not mediated by nonvas-

OIKOS 94:3 (2001) 439

In sum, our experiments here and those in Callaway etal. (unpubl.) suggest that a large array of direct andindirect interactions occur among host trees and epi-phytes. The wide variety of outcomes that could beproduced by different combination of interactions mayplay important roles in structuring and maintainingdiversity in epiphyte communities.

Acknowledgements – Visits by Ragan Callaway to SapeloIsland were supported by the Univ. of Georgia Marine Insti-tute Visiting Scientist Program. We thank Marshal Darley foranalyzing our green slime. This is contribution number 879from the Univ. of Georgia Marine Institute. We gratefullyacknowledge support for this research to Ragan Callawayfrom the Andrew W. Mellon Foundation.

ReferencesArmesto, J. J. and Contreras, L. C. 1981. Saxicolous lichen

communities: non-equilibrium systems. – Am. Nat. 118:597–603.

Armstrong, R. A. 1982. Competition between three saxicolousspecies of Parmelia (lichens). – New Phytol. 90: 67–72.

Benzing, D. H. 1974. The nutritional status of Encyclia tem-pense and Tillandsia circinata on Taxodium ascendens andthe availability of nutrients to epiphytes on this host insouth Florida. – Bull. Torrey Bot. Club 101: 191–197.

Benzing, D. H. 1980. The biology of the bromeliads. – MadRiver Press, Eureka, CA.

Billings, W. D. and Drew, W. B. 1938. Bark factors affectingthe distribution on corticolus bryophytic communities. –Am. Midl. Nat. 20: 302–330.

Buss, L. W. and Jackson, J. B. C. 1979. Competitive networks:nontranstitive competitive relationships in cryptic coralreef environments. – Am. Nat. 113: 223–234.

Callaway, R. M. 1994. Facilitative and interfering effects ofArthrocnemum subterminale on winter annuals in Califor-nia salt marsh. – Ecology 75: 681–686.

Callaway, R. M. and Nadkarni, N. M. 1991. Seasonal patternsof nutrient deposition under Quercus douglasii canopies. –Plant Soil 137: 209–222.

Callaway, R. M. and Pennings, S. C. 2000. Facilitation maybuffer competitive effects: indirect and diffuse interactionsamong salt marsh plants. – Am. Nat. 156: 416–424.

Case, T. J. 1991. Invasion resistance, species build-up andcommunity collapse in metapopulation models with inter-species competition. – Biol. J. Linn. Soc. 42: 239–266.

Connell, J. R. 1991. Apparent versus ‘‘real’’ competition inplants. – In: Grace, J. B. and Tilman, D. (eds), Perspec-tives on plant competition. Academic Press, pp. 9–26.

Dale, M. R. T. and John, E. A. 1999. Neighbour diversity inlichen-dominated communities. – J. Veg. Sci. 10: 571–578.

Davidson, D. W. 1980. Some consequences of diffuse competi-tion in a desert ant community. – Am. Nat. 116: 92–105.

Frei, J. K. and Dodson, C. H. 1972. The chemical effect ofcertain bark substrates on the germination and earlygrowth of epiphytic orchids. – Bull. Torrey Bot. Club 99:301–307.

Hawksworth, F. G. and Wiens, D. 1972. Biology and classifi-cation of dwarf mistletoes (Arceuthobium). – USDAForest Service Agricultural Handbook 401.

Hietz, P. and Hietz-Seifert, U. 1995. Structure and ecology ofepiphyte communities of a cloud forest in central Veracruz,Mexico. – J. Veg. Sci. 6: 719–728.

Hietz, P. and Briones, O. 1998. Correlation between waterrelations and within-canopy distribution of epiphytic fernsin a Mexican cloud forest. – Oecologia 114: 305–316.

Hill, M. O. 1979. DECORANA – A FORTRAN program fordetrended correspondence analysis. – Cornell Univ.,Ithaca, NY.

Holt, R. D. 1977. Predation, apparent competition, and thestructure of prey communities. – Theor. Popul. Biol. 12:197–229.

Kernan, C. and Fowler, N. 1995. Differential substrate use byepiphytes in Corcovado National Park, Costa Rica: asource of guild structure. – J. Ecol. 83: 65–73.

Kerfoot, W. C. and Sih, A. 1987. Predation. Direct andindirect impacts on aquatic communities. – Univ. Press ofNew England.

Kershaw, K. A. 1985. Physiological ecology of lichens. –Cambridge Univ. Press.

Knops, J. M. H., Nash III, T. H., Boucher, V. L. andSchlesinger, W. L. 1991. Mineral cycling and epiphyticlichens: implications at the ecosystem level. – Lichenologist23: 309–321.

Knops, J. M. H., Nash III, T. H. and Schlesinger, W. L. 1996.The influence of epiphytic lichens on the nutrient cycling ofan oak woodland. – Ecol. Monogr. 66: 159–179.

Lawrey, J. D. 1986. Biological role of lichen substances. –Lichenologist 89: 111–122.

Lawrey, J. D. 1995. Lichen allelopathy: a review. – In: Allelo-pathy. Inderjit, Dakshini, K. M. M. and Einhellig, F. A.(eds), American Chemical Society, Washington, DC, pp.26–38.

Levine, J. M. 1999. Indirect facilitation: evidence and predic-tions from a riparian community. – Ecology 80: 1762–1769.

Li, X. and Wilson, S. D. 1998. Facilitation among woodyplants establishing in an old field. – Ecology 79: 2694–2705.

Longan, A., Gaya, E. and Gomez-Bolea, A. 1999. Post-firecolonization of a Mediterranean forest stand by epiphyticlichens. – Lichenologist 31: 389–395.

Lubchenco, J. 1978. Plant species diversity in a marine inter-tidal community: the importance of herbivore food prefer-ence and algal competitive abilities. – Am. Nat. 112:23–39.

Lucking, R. 1999. Ecology of foliicolous lichens at the ‘Botar-rama’ trail (Costa Rica), a neotropical rainforest. IV. Spe-cies associations, their salient features and theirdependence on environmental variables. – Lichenologist31: 269–289.

MacArthur, R. H. 1972. Geographical ecology. – Harper andRow.

McCune, 1997. PC-ORD. Multivariate analysis of ecologicaldata, Ver. 3.0. – MjM Software Design, Gleneden Beach,OR.

Miller, T. E. 1994. Direct and indirect species interactions inan early old-field plant community. – Am. Nat. 143:1007–1025.

Miller, T. E. and Travis, J. 1996. The evolutionary role ofindirect effects in communities. – Ecology 77: 1329–1335.

Olsen, C. 1917. Studies on the succession and ecology ofepiphytic bryophytes on the bark of common trees inDenmark. – Bot. Tidskr. 34: 313–342 (in Danish withEnglish summary).

Paine, R. T. 1966. Food web complexity and species diversity.– Am. Nat. 100: 65–75.

Pennings, S. C. and Callaway, R. M. 1996. Impact of aparasitic plant on the structure and dynamics of salt marshvegetation. – Ecology 77: 1410–1419.

Pianka, E. R. 1974. Niche overlap and diffuse competition. –Proc. Natl. Acad. Sci. USA 71: 2141–2145.

Rundel, P. W. 1978. The ecological role of secondary lichensubstances. – Biochem. Syst. Ecol. 6: 157–170.

Schlesinger, W. H. and Marks, P. L. 1977. Mineral cycling andthe niche of Spanish moss, Tillandsia usneoides L. – Am. J.Bot. 64: 1254–1262.

Stone, D. F. 1989. Epiphyte succession on Quercus garryanabranches in the Willamette Valley of western Oregon. –Bryologist 92: 81–94.

440 OIKOS 94:3 (2001)

Stone, L. and Roberts, A. 1991. Conditions for a species togain advantage from the presence of competitors. – Ecol-ogy 72: 1964–1972.

Talley, S. M., Lawton, R. O. and Setzer, W. N. 1996.Host preferences of Rhus radicans (Anacardiaceae) in asouthern deciduous hardwood forest. – Ecology 77: 1271–1276.

Tucker, S. C. 1979. New or noteworthy records of lichensfrom Louisiana. – Bryologist 82: 125–140.

Vandermeer, J. 1990. Indirect and diffuse interactions: compli-cated cycles in populations embedded in a large commu-nity. – J. Theor. Biol. 142: 429–442.

Wilson, S. D. and Keddy, P. A. 1986a. Species competitiveability and position along a natural stress/disturbance gra-dient. – Ecology 67: 1236–1242.

Wilson, S. D. and Keddy, P. A. 1986b. Measuring diffusecompetition along an environmental gradient: results froma shoreline plant community. – Am. Nat. 127: 862–869.

Woolhouse, M. E. J., Harmsen, R. and Fahrig, L. 1985. Onsuccession in a saxicolous lichen community. – Lichenolo-gist 17: 167–172.

Wooton, J. T. 1994. The nature and consequences of indirecteffects in ecological communities. – Annu. Rev. Ecol. Syst.25: 443–466.

OIKOS 94:3 (2001) 441