The role of colored accessory bracts in the reproductive biology of Lavandula stoechas

Ecology, 78(2), 1997, pp. 494-504 C) 1997 by he Ecological Socie y o Ame ica THE ROLE OF COLORED ACCESSORY BRACTS IN THE REPRODUCTIVE BIOLOGY OF LA VANDULA STOECHAS JAVIER HERRERA Depa amen o de Biologla Vege al y Ecolog a, Uni e sidad de Se illa, Apa ado 1095, E-41080 Se illa, Spain Abs ac . This s udy in es iga es he ole played by showy s uc u es in he ep o- duc ion o La andula s oechas L. (Lamiaceae), a sel -compa ible Medi e anean sh ub wi h compac , head-like in lo escences ha a e e mina ed by a u o conspicuous pink b ac s. Pollina o isi a ion a es o plan s wi h b ac s co ela ed posi i ely wi h ha o pai ed, ea ed sh ubs wi h all hei b ac s emo ed and, o e all, b ac less and con ol plan s had equi alen isi a ion a es. In expe imen ally spli sh ubs, howe e , a s ong e ec on pollina o choice was e iden , and dis an isi o app oaches we e mos o en o ha side keeping he a ac i e o gans. Fecundi y was no dep essed in plan s wi h educed isual displays, and nei he was he amoun o pollen dispe sed pe lowe o he size o s igma ic loads. The e we e, howe e , mo e pollen ubes, on a e age, a he lowe mos pa o he s yle in con ols (10.8 ubes/s yle) han in b ac less sh ubs (8.3 ubes/s yle). Small, non- signi ican declines in a e age seed size, ge minabili y, and seedling mass we e de ec ed ollowing b ac emo al. Da a suppo he no ion ha b ac s a e no essen ial o seed p oduc ion in ma u e, ela i ely dense La andula popula ions whe e bee pollina o s a e o en in good supply, bu may be all-impo an i plan densi y is ex emely low, o du ing popula ion es ablishmen . The ole o an ampli ied display in inc easing a e age pollina ion dis ance and p ogeny quali y in his species is discussed. Key wo ds: ecundi y; lo al display; La andula; Lamiaceae; Medi e anean; ou c ossing; pol- lina o a ac ion. INTRODUCTION Many aspec s o animal-pollina ed plan ep oduc- ion a e in luenced by he size o he lo al display, including isi a ion a es and ecundi y (e.g., Willson and P ice 1977, Lack 1982, Go i 1983, C uzan e al. 1988, Galen 1989), ou c ossing le els (Gebe 1985, Robe son 1992, Klinkhame e al. 1994), success as a male (Bell 1985, S an on e al. 1986, C uzan e al. 1988, Campbell e al. 1991, De lin e al. 1992, among o he s), and pollen discoun ing (Ha de and Ba e 1995). Whene e la ge displays a ise om he p esen a ion o many lowe s, consequences may also include adjus - men o a ying esou ce le els and he oppo uni y o selec i e emb yo abo ion (Su he land 1986; see S e- phenson 1981, and Caspe and Niesenbaum 1993 o e iews). Ei he by making use o lowe polymo phisms (So- lomon 1987, Galen and S an on 1989, Campbell e al. 1991, S an on e al. 1991, o example) o by manip- ula ing a ac i e lowe pa s (Bell 1985, Young and S an on 1990), mos s udies on he e ec s o inc eased lo al displays on plan ep oduc ion conside he in- di idual lowe as he a ac i e uni (see, howe e , C uzan e al. 1988, Ande sson 1991, Ande sson and Widen 1993). Ne e heless, in many angiospe m g oups he lowe s a e educed in size and agg ega ed in o head-like in lo escences ha ac as uni s o a - Manusc ip ecei ed 13 July 1995; e ised 2 May 1996; accep ed 7 May 1996; inal e sion ecei ed 30 May 1996. ac ion. Lea ing aside he examples in which in lo- escence a ac i eness elies on lowe s la ge han a - e age ( o example, in he Composi ae), conspicuous seconda y s uc u es associa ed wi h lowe s bu no composed o lowe s hemsel es a e exceedingly com- mon in se e al plan amilies (e.g., he colo ed b ac s o many A aceae, B omeliaceae, Eupho biaceae, and Nyc aginaceae; Heywood 1978). Non lo al showy s uc u es a e aci ly accep ed o imp o e pollina ion success by enhancing isual dis- play. Howe e , empi ical da a on his issue a e sca ce (bu see Mendez and Obeso 1992), which is su p ising since manipula ing such s uc u es is a easie han inducing changes in e e y lowe o an indi idual. Fu - he mo e, a majo di e ence be ween lo al and non- lo al s uc u es is ha he la e can o en be ela i ely ee om he s ong phylogene ic cons ain expe i- enced by mos quan i a i e lo al ai s (e.g., Webb 1984). The e o e, because non lo al de ices a e likely o be e olu iona ily mo e e sa ile, such o gans may p o ide a use ul ool o he s udy o he ela ionship be ween lo al display and plan ep oduc ion. In ad- di ion, ancilla y s uc u es inc ease showiness wi hou inc easing game ophy e numbe s (as opposed o p o- ducing la ge numbe s o lowe s), and he e ec o display on isi a ion a es, pheno ypic gende , and ou - c ossing le els can be es ed wi hou he con ounding e ec s o di ec sexual alloca ion. The pu pose o his s udy was o in es iga e he p e- sumed ad an age o ha ing non lo al, colo ed s uc- 494 This con en downloaded om 193.147.173.147 on Tue, 11 Jul 2017 14:45:49 UTC All use subjec o h p://abou .js o .o g/ e ms Ma ch 1997 FLORAL DISPLAY IN LAVANDULA 495 PLATE 1. A medium-sized, 1 m high La andula sh ub. u es (b ac s) associa ed wi h he lowe s o La andula s oechas, a Medi e anean sh ub. The ai is ela i ely ee om phylogene ic cons ain since, among he se - en Eu opean species o La andula, only L. s oechas and he ela ed L. i idis ha e such b ac s (Guinea 1972). Al hough he e is na u al a ia ion o he s ud- ied cha ac e is ic (in he in ensi y o colo , shape, and size o b ac s), i is o en ela i ely di icul o quan i y, so my app oach consis ed o manipula i e expe imen s o simula e he appea ance o "mu an s" (Wase 1983) comple ely lacking b ac s. Pollina o s can pe cei e hese o gans as e idenced du ing pilo , b ac - emo al expe imen s a he begin- ning o his s udy, when I obse ed ha bees o en app oached he disca ded b ac s on he g ound. The e- o e, he gene ic ques ion is no he a he ob ious "A e b ac s a ac i e o pollina o s?" bu "How much, and in which way does his ai in luence plan ep oduc- ion?" The ollowing speci ic ques ions a e add essed: (1) do b ac s inc ease plan isi a ion a e?; (2) a e b ac s c i ical o he numbe o pollen g ains deposi ed on he s igma and/o he numbe o pollen ubes in he s yle?; (3) is he a e o pollen expo a ec ed by he p esence o b ac s?; (4) do b ac s inc ease ecundi y?; and (5) is he e any di e ence in igo among he p og- enies o no mal and b ac less plan s? METHODS S udy plan La andula s oechas L. (Lamiaceae; La andula, he e- a e ) is an a oma ic, Medi e anean xe ophy ic sh ub up o 1 m high, which is e y common in well-inso- la ed, d y si es in sou he n Spain. F om sea le el o moun ain anges up o 1000 m abo e sea le el, pop- ula ions occu as agg ega ions o sh ubs in pa ches o a ying a ea. Regene a ion occu s only om seed (He - e a 1987). Rep oduc i e cha ac e is ics p e iously s udied in his species include head a chi ec u e (De- esa e al. 1985), b eeding sys em (Mu ioz and De esa 1987), lowe and ui biology (He e a 1991, 1993), and insec lowe choice (Du ield e al. 1993). Flowe s ha e a small (5 mm long), da k-pu ple ubula co olla inse ed in o a ubula calyx, and a e agg ega ed in heads. These a e composed o igh ly packed dichasia (i.e., g oups o lowe s on a highly educed b anch) a ached o a cen al, common axis (Fig. 1). Each head encloses some 100 buds, al hough he e a e seldom mo e han 20 lowe s open simul aneously. Each head is e mina ed by a u o showy pink b ac s (Fig. 1), each measu ing 20-30 X 5-10 mm. The e a e ypically ou pe head, and since an a e age indi idual bea s 20-50 heads, he o al numbe o b ac s on a lowe ing sh ub is la ge and hei isual impac is (a leas o humans) conside able. La andula lowe s sec e e minu e (0.5-1.5 ALL) amoun s o suga - ich (30-60%) nec a and ha e an a e age pollen: o ule a io o 1000. Al hough lowe s se ui ollowing expe imen al pollina ion wi h pollen om he same indi idual (and a e hus conside ed sel - compa ible; Mu ioz and De esa 1987), au ogamy (i.e., wi hin- lowe pollina ion) a ely occu s in na u e be- cause o ma ked p o and y. Because many newly open, pollen-shedding lowe s cus oma ily coexis wi h olde , emale-phase lowe s in any sh ub, howe e , wi hin- plan pollina ion (i.e., gei onogamy) is p obably e y common. S udy si es The s udy was conduc ed du ing he win e and sp ing o 1992, 1993, 1994, and 1995 a wo sou he n Spanish locali ies (Do iana and Aznalcaza ) sepa a ed FIG. 1. A head o La andula s oechas showing he cha - ac e is ic g oup o s e ile, colo ed b ac s a he op. This con en downloaded om 193.147.173.147 on Tue, 11 Jul 2017 14:45:49 UTC All use subjec o h p://abou .js o .o g/ e ms 496 JAVIER HERRERA Ecology, Vol. 78, No. 2 a EZ* wnEZE bC [J *2 FIG. 2. The h ee s a egies used in his s udy o spa ial in e spe sion o ea men s (a, b, c), wi h squa es ep esen ing sh ubs, and con ol/ ea ed expe imen al uni s deno ed by open/shaded symbols. The a ying a eas o symbols ep esen a ia ions in size (head numbe ) among uni s. Ho izon al lines delimi expe imen al blocks. by 50 km. The Do iana si e is in he Do ana Biological Rese e, a coas al a ea nea he gul o Cadiz whe e ege a ion is domina ed by sc ub g owing on old s a- bilized sand dunes (see Ri as-Ma inez e al. 1980 o a desc ip ion). A his si e, La andula plan s g ow in- e mingled wi h many o he sh ub species. The Az- nalcaza si e is u he inland bu a abou he same ele a ion as he coas al si e. This popula ion is a nea ly pu e s and o La andula inhabi ing a woodland clea - ing. The subs a e is made up o g a el, and ege a ion is la gely domina ed by Pinus pinea woodland wi h an unde s o y o xe ophy ic sc ub. The s udy popula ions we e selec ed because o hei con as ing spa ial s uc u es, sh ub size dis ibu ions, and pollina o ac i i y egimes. The popula ion a Az- nalcaza is composed o some 400 indi iduals dis ib- u ed o e an a ea o 20 X 40 m. Mos plan s bea 20- 50 heads when in ull bloom (medium-sized sh ubs, he ea e ), al hough la ge plan s (50-200 heads) a e also common. This is he commones popula ion s uc- u e o La andula h oughou sou he n Spain (pe sonal obse a ion). In con as , he Doiana popula ion is made up o small (<20 heads) and medium-sized plan s e enly sp ead o e many hec a es. Si es also di e in he abundance and di e si y o he lowe - isi ing in- sec a ay: a Do iana, Apis melli e a accoun s o nea - ly 90% o isi s (He e a 1988), whe eas a Aznalcaza a a ied assemblage o soli a y bees is ound in ad- di ion o honey bees. Fu he mo e, a he coas al si e he e a e o en mild o s ong b eezes, which nega i ely a ec pollina o ac i i y. Whole-plan b ac emo al expe imen s To es he hypo hesis ha b ac s acili a e pollina- ion in La andula I selec ed a sample o plan s, e- mo ed all he b ac s om sh ubs chosen a andom, and hen compa ed i ness- ela ed a iables (e.g., seed se ) among indi iduals wi h and wi hou b ac s. Be- cause he buds om which heads will de elop appea synch onously in La andula, mos b ac s in a sh ub could be emo ed a a single da e when lowe ing was abou o begin. Addi ionally, howe e , I echecked each ea ed plan weekly o elimina e b ac s om he ew esidual head buds ha could ha e appea ed anew. This "whole-plan app oach" was used o s udy he e ec s o b ac s on insec isi a ion a es, ecundi y, s igma ic pollen loads, and he numbe o pollen ubes in he s yle. Insec isi a ion a es.-Ten adjacen plan s a he Do iana popula ion we e selec ed a he beginning o he 1992 lowe ing season. Indi iduals we e pai ed, and heads clipped as necessa y om one o he indi iduals in each pai in o de o ob ain equal numbe s o heads on bo h plan s (Fig. 2a). This numbe anged om 36 o 60 heads. A plan wi hin each pai was hen selec ed a andom, and all colo ed b ac s we e clipped. Pai ed plan s we e close enough (0.5-1 m) o allow simul a- neous obse a ion o hei isi o s ollowing a schedule in which I obse ed one pai o 4 min, hen mo ed o he nex pai o ano he 4 min, e c., un il all i e pai s had been moni o ed. Fou such uns we e pe o med on each o wo da es, a 1000, 1200, 1400, and 1600. In all, his accoun s o 32 min o obse a ion pe pai o sh ubs. The same p ocedu e was used a he Az- nalcaza popula ion wi h ano he se o pai ed sh ubs in which he lo al display anged be ween 11 and 33 heads/plan . All censusing was ca ied ou in wa m, clea wea he , which is ypical o he blooming pe iod o La andula in sou he n Spain. Fecundi y.-In 1992 and 1994 he ecundi y o plan s wi h and wi hou b ac s was compa ed a he Do iana si e (Fig. 2b). I was unable o use he same indi iduals in 1992 and 1994 because o high plan mo ali y, bu plan samples we e compa able in ha he sh ubs inhabi ed he same a ea and we e simila in size ( o he 1992 se , 18 heads/plan on a e age, ange 5-70, N = 30; o he 1994 se , 19 heads/plan , ange 5-36, N = 20). The me hod (and iming) used o b ac emo al was iden ical o ha desc ibed p e iously (see abo e Whole-plan b ac emo al expe imen s). The main componen o ecundi y s udied was seed se . In La andula he o a y has ou o ules, and seed se pe lowe is hus i ually a ca ego ical a iable (ze o, one, wo, h ee, o ou seeds may de elop). To ge a measu e o ecundi y s a is ically mo e powe ul han pe - lowe seed se , I examined he p opo ion o o ules wi hin a head ha de elop in o iable seed (i.e., seed se pe head). Fu he mo e, since he loss o low- e s (caused by bee les in he genus Melige hes, Ni i- dulidae) can also be a majo de e minan o ecundi y in La andula (He e a 1993), p eda o ac i i y was also in es iga ed. A di ec measu e o p eda ion in ensi y o heads is di icul o ob ain because wil ing co ollas a e d opped om hei calyces wi hin 7-10 d om an hesis. Howe e , an es ima e can be ob ained a he ui ing s age om he p opo ion o calyces con aining insec d oppings, cu icles, o chewed lowe pa s (mos ly he o a y). This con en downloaded om 193.147.173.147 on Tue, 11 Jul 2017 14:45:49 UTC All use subjec o h p://abou .js o .o g/ e ms Ma ch 1997 FLORAL DISPLAY IN LAVANDULA 497 Heads a e in e nally he e ogeneous because hey a e in ac agg ega ions o small, 5-7 lowe ed uni s (di- chasia). Flowe s opening i s in each dichasium a e mos likely o p oduce ui s, whe eas la e opening ones a e much mo e esou ce limi ed (He e a 1991). When sampling ui s o assess seed se /head, I used only cen al (i.e., i s -opening) lowe s, so ecundi y es ima es epo ed below a e uppe bounds. This was pa icula ly con enien du ing he 1994 season, when plan ecundi y d opped o ex emely low le els be- cause o d ough . When sampling heads o p eda ion, on he o he hand, I used la e al (i.e, la e opening) lowe s. Since hese a e p eyed upon mo e in ensely han ea ly-opening ones (He e a 1993), p eda ion a es epo ed a e also maximum es ima es. On 4-5 heads pe plan , I sampled all (10-15) i s -o de calyces o iable seeds along wi h 10 second-o de calyces o p eda ion. Sampling di e en se s o lowe s o check ecundi y o p eda ion helps o a oid edundancies in he esul ing da a se and o make hese es ima es el- a i ely independen . Fu he mo e, using he same se o lowe s o es ima e ecundi y and p eda ion would ha e o ced me o deal wi h ye ano he sou ce o a i- a ion (namely wi hin-head he e ogenei y) which was o no in e es o he s udy. Pollen loads and he numbe o pollen ubes in na - u ally pollina ed lowe s.-One way in which b ac s migh enhance pollina ion is by inc easing he numbe o imes a lowe is isi ed, hus esul ing in la ge s igma ic pollen loads and/o g ea e pollen ube num- be pe lowe . To es his, 20 small sh ubs abou o lowe we e ma ked a Do iana du ing he 1994 season and e e y colo ed b ac emo ed om 10 o hese plan s. Sh ubs we e sp ead o e an a ea o 900 m2 and in e mingled wi h many nonexpe imen al indi iduals (see Fig. 2b). A weekly in e als on h ee di e en da es, d y co ollas (and he s yles he ein) ha e- mained a ached o he calyx we e haphaza dly col- lec ed om a numbe o heads on each sh ub and ixed in o maline: ace ic acid: alcohol (5:5:90). Ma e ial om each sh ub was ho oughly mixed (pollen was i mly adhe ed o s igmas, so g ain exchange be ween s igmas was unlikely). Then a sample o 10 s yles was chosen a andom. A e clea ing wi h ho , 8 mol/L NaOH o 30 min, s yles we e insed, s ained in a d op o decolo ized aniline-blue, and obse ed unde UV ligh (Ma in 1959) o coun he numbe o pollen g ains adhe ed o he s igma and o pollen ubes a he base o he s yle. Spli -plan b ac emo al Using sh ubs as expe imen al uni s o pe o m b ac emo al allows one o di ec ly assess he ai 's sig- ni icance o i ness. Howe e , ex ensi e be ween-plan a ia ion exis ed o e e y s udied a iable, which migh gene a e inconclusi e esul s i he hypo hesized e ec was ela i ely weak. To allow some con ol o be ween-plan a ia ion, I also used hal -sh ubs as ex- pe imen al uni s in a sepa a e se o expe imen s. In hese, I selec ed a g oup o plan s, hen spli each plan in o wo equal-sized hal es (i.e., wi h he same numbe o heads on each side), and compa ed esponse a i- ables wi hin and among plan s ollowing b ac emo al om one andomly chosen hal (see Fig. 2c). Plan spli ing was done by clipping heads so as o c ea e a 10-20 cm wide s ip de oid o lowe s ac oss he sh ub. In his way, I s udied he po en ial e ec s o b ac s on pollina o choice, numbe o heads isi ed, amoun o pollen expo ed by indi idual lowe s, as well as seed and seedling cha ac e is ics. As in he whole-plan ap- p oach, b ac s we e emo ed om heads be o e he onse o he blooming season. A disad an age o using spli plan s as expe imen al uni s is ha ea men e ec s a e no di ec ly in e - p e able in e ms o i ness (since hal -plan s a e no eal en i ies bu a i ices), and one has o conjec u e abou how and i esul s om hal -sh ubs loca ed side by side would ex apola e o ac ual sh ubs. A leas ega ding in e plan dis ances and i s bea ing on pol- lina o a ac ion, his is no a se ious d awback because ue indi iduals a e mos o en side by side in La an- dula popula ions, and pollina o s a e hus likely o dj s- play simila ligh pa e ns, ega dless o he ac ha hey ace a pai o eal sh ubs o a pai o expe imen- ally p oduced hal -sh ubs. The e could be some doub , howe e , on whe he hal -sh ubs sha e esou ces: i he e exis ed compensa o y e ec s wi hin a spli sh ub, b ac emo al e ec s in ( o example) seed size migh be obscu ed. The a ailable e idence, howe e , indi- ca es ha such e ec s a e unlikely o occu . Fo ex- ample, he low ui se ha cha ac e izes la e-opening lowe s wi hin a head can be a i icially inc eased by emo ing a subse o lowe s om he head (Mu ioz and De esa 1987), bu i canno be by emo ing o he heads on he sh ub (He e a 1991). This suppo s he idea ha La andula heads beha e like independen e- sou ce alloca ion uni s, and ha compensa o y e ec s wi hin spli sh ubs a e no likely. Pollina o choice.-Du ing he 1993 season, 10 plan s pe popula ion we e selec ed and spli as de- sc ibed abo e. The esul ing hal es we e sequen ially obse ed du ing 30-min pe iods o a o al o 120 min pe plan (i.e., a o al o 1200 min o obse a ion/pop- ula ion) o eco d he numbe o insec s ha , when app oaching a spli sh ub o he i s ime, selec ed he side wi h b ac s o he side wi hou b ac s. This was in ended o emphasize he e ec o b ac s on insec choice a a dis ance (i.e., a en ion was no paid o shi s om he i s -chosen hal -plan o he neighbo ing one). The numbe s o heads on hal -sh ubs anged be ween 10 and 60 a Do iana (mean ? 1 SE, 32.0 ? 2, N = 20), and be ween 14 and 55 a Aznalcaza (34.7 ? 4, N = 20). Obse a ions we e ca ied ou om mo ning o a e noon on di e en da es a each o he popula- ions. Numbe o heads isi ed.-Inc eased conspicuous- This con en downloaded om 193.147.173.147 on Tue, 11 Jul 2017 14:45:49 UTC All use subjec o h p://abou .js o .o g/ e ms 498 JAVIER HERRERA Ecology, Vol. 78, No. 2 ness b ough abou by b ac s migh help sh ubs o e ain isi o s o a longe pe iod o ime, which migh esul in mo e lowe s o heads being isi ed by each polli- na o . This migh in u n a ec he emale and/o he male componen o ep oduc ion (by inc easing ei he s igma ic pollen loads and/o pollen expo ). The num- be o heads isi ed by pollina o s was in es iga ed a Aznalcaza by selec ing h ee la ge sh ubs, spli ing each in o hal es as desc ibed abo e, and hen emo ing b ac s om one hal . When an insec a i ed, I no ed how many heads i isi ed be o e lea ing ha hal . To accoun o pollina o -speci ic e ec s, sepa a e eco ds we e kep o Apis melli e a and soli a y bees. O e all, 188 insec s we e acked. Pollen dispe sal.-B ac s could ha e no appa en e - ec on ecundi y, and ye ha e a posi i e e ec on he male componen o ep oduc ion. To check his pos- sibili y, 405 lowe buds abou o open we e ma ked ea ly in he mo ning on Ma ch 1995 a he Do iana popula ion om a o al o 20 p e iously spli sh ubs (da a we e pa ially missing o one o he sh ubs, whe eas ano he plan happened o be male s e ile, so he inal sample size was only 18 indi iduals). O hese buds, 61 we e on heads bagged wi h 0.2-mm mesh ha excluded all kind o pollina o s, whe eas he emaining (on heads wi h o wi hou b ac s) we e exposed o in- sec s. Flowe s we e le o open and be isi ed no mally du ing hei 1s d o an hesis, a he end o which a sample o co ollas was picked up wi h g ea ca e o a oid any pollen losses and indi idually kep in num- be ed ials. The emaining lowe s we e le exposed o pollina o s du ing he 2nd d o an hesis and hen collec ed. All ou an he s o lowe s a e open a he ime o an hesis in La andula, and pollen has los mos iabili y a e 48 h o elease om an he s (Mu loz and De esa 1987). The amoun o pollen emaining in he an he s in all h ee g oups o lowe s (bagged, exposed o 1 d, and exposed o 2 d) was es ima ed by adding 0.2 mL o a wa e -de e gen -sa anine solu ion o each ial, ca e- ully squashing he an he s wi h o ceps, and aking wo 0.5-[iL aliquo s o he solu ion. The numbe o g ains in each aliquo was hen coun ed a 40X unde he mic oscope. Because he numbe s in he i s and he second ex ac ion o each lowe we e highly co - ela ed ( = 0.960, N = 405, P << 0.001), hese we e a e aged and, a e co ec ing o he dilu ion ac o , he esul ing igu e was used as an es ima e o he pol- len ha each lowe had ailed o dispe se. Wea he condi ions we e simila du ing he 2 d spanned by he expe imen , wi h ligh b eezes, wa m empe a u es (daily maxima - 250C), and mode a e o high bee ac- i i y. Seed and seedling cha ac e is ics.-To in es iga e he e ec o b ac s on seed cha ac e is ics, se en sh ubs a Do iana we e spli as p e iously desc ibed a he onse o he 1993 blooming season. By he ime he ui s we e ipe (ea ly summe ) heads om each hal - CD 10 3 CD 9 Z 8- w -5 CD 3 1 3 5 7 9 11 13 15 17 19 2 4 6 8 10 12 14 16 18 20 PLANT IDENTIFICATION NUMBER FIG. 3. Visi a ion a es o pai ed La andula sh ubs (con- ol/no b ac s) ma ched by size. Each ba is he mean numbe o pollina o s a i ing a a sh ub pe 4-min census (N = 8 o each plan ). Open ba s ep esen con ols, and shaded ba s ep esen ea ed sh ubs. Ve ical lines span 2 SE. Plan s 1- 10 a e om Do iana, 11-20 a e om Aznalcaza . plan we e ha es ed, aken o he labo a o y, and s o ed a ambien empe a u e. The nex au umn, seeds om ea ly-opening lowe s (i.e., hose expe iencing less abo ion) we e picked, and hose un illed o show- ing signs o p eda ion disca ded. A o al o 526 an- domly chosen seeds we e indi idually weighed o he nea es 0.01 mg o es o he e ec o b ac emo al on seed size. This expe imen was epea ed in 1995, his ime using 18 di e en sh ubs. Do iana plan s we e a he small, hus making i di - icul o collec la ge numbe s o seeds. The e o e, he e ec o b ac emo al on ge mina ion a e and seed- ling size was s udied in i e la ge sh ubs om Aznal- caza . Th ee-hund ed seeds om each hal -plan we e sown in po s con aining mois pea inside an unhea ed glasshouse ( h ee po s pe hal -plan , 100 seeds pe po ), and he numbe o seedlings appea ing in each po eco ded 2 wk la e . Seedlings we e hinned o 20 pe po a his ime, and hose emaining we e egula ly wa e ed o he nex 30 d. A his poin I ha es ed he seedlings and de e mined hei esh masses o he nea - es millig am. Da a analyses Expe imen al uni s in his s udy a e ei he whole sh ubs o hal -sh ubs (see Fig. 2) and, acco dingly, a ia es o he s udied a iables a e a e ages o ei he plan s o hal -plan s. Fo example, when looking a isi a ion a es o sh ubs wi h o wi hou b ac s, he numbe s o insec s pe 4-min census o a gi en plan we e a e aged and he esul ing igu e used. I wha was being add essed was he e ec on ( o example) pollen expo , he numbe o g ains le in se e al low- e s wi hin a hal -plan we e a e aged and he mean used o ep esen his hal -plan 's a e o ailu e in pol- len expo . In gene al, means dis ibu ed no mally and no ans o ma ions we e needed, wi h he excep ion o pe cen age seed se , which was a csine squa e- oo ans o med. When app op ia e (i.e., he designs in Fig. This con en downloaded om 193.147.173.147 on Tue, 11 Jul 2017 14:45:49 UTC All use subjec o h p://abou .js o .o g/ e ms Ma ch 1997 FLORAL DISPLAY IN LAVANDULA 499 TABLE 1. (A) Rela ionship o colo ed b ac s and ecundi y in La andula. The analysis o co a iance (B) used he a - e aged seed se o i e heads pe sh ub as he esponse a iable, and a e age p eda ion in ensi y as a con inuous co a ia e. "T ea men " e e s o b ac emo al. The ex- pe imen al design co esponds o ha shown in Fig. 2b. A) Seed se o sh ubs (%) Wi h b ac s Wi hou b ac s Yea (X ?+ SE) (X ?+ SE) 1992 49 ?5 55 ?5 1994 21 ?4 14 ?4 B) Yea Sou ce d MS F P 1992 T ea men 1 277.5 0.78 0.39 P eda ion 1 15.6 0.04 0.84 E o 27 355.4 1994 T ea men 1 82.9 0.69 0.42 P eda ion 1 546.6 4.55 0.04 E o 17 120.0 N = 10 in all cases excep o 1992 (wi h b ac s), whe e N = 14. 2a, c), pai s o expe imen al uni s we e ea ed as an- domized blocks. Analyses we e pe o med wi h p o- cedu es in SAS (1990) and SYSTAT (Wilkinson 1986) s a is ical packages. RESULTS Whole-plan expe imen s: insec isi a ion a es A he spa se popula ion sh ubs ecei ed om 1 o 3 insec s/census, whe eas a he denses si e be ween 2 and 8 insec s/census we e eco ded (Fig. 3). The a es o isi a ion o pai s o con ol and ea ed sh ubs co ela ed posi i ely ( , = 0.650, N = 10, P < 0.05), indica ing ha among-plan a ia ions a e due (a leas in pa ) o spa ial loca ion and head numbe . A andomized block analysis o plan means e ealed a signi ican e ec o "Block" (i.e., sh ub pai ; MS = 0.561, d = 9, F = 3.64, P = 0.03), while i ailed o de ec any e ec due o b ac s (MS = 0.299, d = 1, F = 1.95, P = 0.20; e o MS = 0.154, d = 9). Because o high among-plan a ia ion in isi a ion a es, how- e e , his analysis is p one o Type II e o . Wi h he obse ed a iance in isi a ion a es and N = 10, he powe o a es o di e ences among ea men means is only 0.67 (i.e., a ue di e ence would ail o be de ec ed a P < 0.05 in 33% o es s). O e all, b ac less sh ubs ecei ed 2.9 ? 0.6 pollina o s/census, as op- posed o 3.8 ? 0.8 ecei ed by con ols (i.e., a 23% dec ease). Fecundi y A e age o e all seed se was :50% in 1992, and as low as 15% in 1994. D ough du ing 1994 may (in addi ion o plan iden i y) accoun o he di e ence (Table 1). In 1992 seed se was highe in ea ed sh ubs han in con ols, and he e e se occu ed in 1994. uj 00 ? 12 - 0 0 z 0 U3i 0 0 0 00 0 0 ? DD Z 8 - uj 0 0 0 CL 6 - 4 _ L 0 50 100 150 NO. POLLEN GRAINS ON STIGMA FIG. 4. The ela ionship be ween pollen load on he s ig- ma and he numbe o ubes in he s yle in La andula. Each do is he a e age alue o a sh ub wi h (0) o wi hou (0) b ac s. The eg ession line o his ela ionship ( ubes = 6.8 ? 0.04 s igma ic load, 2 = 0.27, N = 20) is also shown. Howe e , in no yea we e ea men e ec s signi ican (Table 1). The co a ia e p o ed signi ican only du ing 1994, implying ha ecundi y and lowe p eda ion co - ela ed (nega i ely) du ing ha yea . Pollen loads and numbe o ubes as a ec ed by b ac s O e all, he e we e, on a e age, 74 ? 5 pollen g ains pe s igma (median 52, minimum 0, maximum 400) and 9.5 ? 0.5 pollen ubes pe s yle (median 10, min- imum 0, maximum 24) in he s udied sample. The a - e age pollen g ain: pollen ube a io was 7.8, and ha o pollen ubes o o ules 2.4. O 200 lowe s analyzed, only six had no pollen on hei s igmas. Plan means o he size o s igma ic pollen loads anged om 16.8 o 141.5 g ains, and he numbe o TABLE 2. (A) Rela ionship o La andula b ac s and he numbe o pollen ubes ha each he lowe mos pa o he s yle. The analysis o co a iance (B) used he a e age numbe o ubes in 10 na u ally pollina ed lowe s pe sh ub as he esponse a iable, and he a e age numbe o pollen g ains on he s igma as a con inuous co a ia e. "T ea - men " e e s o b ac emo al. The expe imen al design co esponds o ha shown in Fig. 2b. A) Sh ubs Wi h b ac s Wi hou b ac s Va iable (X ? 1 SE) (X ? 1 SE) Pollen g ains on s igma 73.2 ? 12 74.6 ? 12 Pollen ubes in s yle 10.8 + 1 8.3 I 1 B) Sou ce d MS F P T ea men 1 32.7 9.23 0.007 Pollen load 1 35.8 10.090 0.006 E o 17 3.6 N = 10 in all cases excep o pollen g ains on s igma (wi hou b ac s), whe e N = 16. This con en downloaded om 193.147.173.147 on Tue, 11 Jul 2017 14:45:49 UTC All use subjec o h p://abou .js o .o g/ e ms 500 JAVIER HERRERA Ecology, Vol. 78, No. 2 TABLE 3. (A) Pollina o beha io as a ec ed by colo ed b ac s and (B) analysis o a iance. The app oaching a e is he a e age numbe o insec s, which, when a i ing o a spli (b ac /no b ac s) sh ub om a dis ance, chose ei he he con ol o he ea ed hal . The expe imen al design co esponds o ha shown in Fig. 2c, wi h each spli sh ub ep esen ing a block. A) App oaching a e (no. insec s/30 min) T ea men X + 1 SE N Con ols 4.04 0 0.3 20 No b ac s 2.38 + 0.3 20 B) Sou ce d MS F P T ea men 1 1.88 15.50 0.001 Block 19 0.19 1.53 0.18 E o 19 0.12 pollen ubes in he s yle a ied be ween 4.9 and 13.2. I is wo hwhile no ing ha pollen ube numbe was less a iable among plan s (c = 27%) han s igma ic pollen loads (c = 50%). Thus, pollen deposi ion on he s igma was hea ily a ec ed- by ac o s ela ed o plan iden i y, whe eas he po en ial o o ule e il- iza ion was ela i ely ee om cons ain by indi id- ual-speci ic ai s (see Discussion). A signi ican , posi i e linea ela ionship be ween pollen ube numbe and s igma ic pollen load was de- ec ed (Fig. 4; F = 6.69, d = 1, 18, P = 0.02). Because sh ubs wi hou b ac s appea mos ly below he line o eg ession, his ela ionship seems o be a ec ed by b ac emo al. This is con i med by analysis o co- a iance (Table 2). A e age pollen ube numbe de- pends no only on he numbe o pollen g ains on he s igma, bu also on he p esence o absence o b ac s. On a e age, b ac emo al dec eased he numbe o pollen ubes in he s yle by 23% (Table 2). Spli -plan expe imen s: pollina o selec ion o plan hal es Plan hal es wi h b ac s did be e a ge ing dis an ( om any o he sh ub) pollina o app oaches han hal es wi hou b ac s (Table 3), since b ac emo al was associa ed wi h a 41% dec ease in a e age ap- p oaching a e. Fu he mo e, and because plan iden- i ies (i.e., ac o "Block") did no explain a signi ican pa o a ia ion (Table 3), ea men e ec s occu ed wi h independence o sh ub size and a bo h he dense and he spa se s udy si es. Numbe o heads isi ed Apis melli e a p obed 5.7 ? 0.9 heads (N = 30) on con ols, s. 5.5 + 0.8 (N = 30) on ea ed hal -sh ubs wi hou b ac s (T = 0.14, d = 58, P = 0.8). Fo soli a y bees, means we e, espec i ely, 3.1 ? 0.4 (N = 53) and 3.7 ? 0.5 (N = 55; T = 0.85, d 106, P = 0.4). Depa u e decisions we e he e o e independen o he p esence o b ac s in bo h majo pollina o g oups. B ac s and pollen dispe sal O he 4000 pollen g ains p oduced by each La an- dula lowe (as es ima ed in he bud s age), 57% we e dispe sed du ing he 1s d o an hesis, whe eas, on a - e age, 17% emained in he an he s a he end o he 2nd d (Table 4). The e we e no signi ican ea men e ec s on he amoun o pollen le undispe sed. A highly signi ican e ec was de ec ed o he "Block" ac o , howe e , implying ha ce ain sh ubs dispe sed pollen much mo e e icien ly han o he s. The signi i- can Block X Time in e ac ion e eals ha indi idual e iciencies in dispe sing pollen may a y om one day o he nex . Seed and seedling cha ac e is ics The masses o seeds p oduced in hal -sh ubs wi hou b ac s we e, on a e age, smalle han hose o con ols in bo h 1993 and 1995 (Table 5). Howe e , he di e - ences we e minimal (-0.02 mg), and he e we e no signi ican ea men e ec s o any o he yea s (s a- is ical powe is 0.95 and 0.83, espec i ely, o 1992 and 1995). Pe cen age seed ge mina ion was also s a- is ically undis inguishable in ea ed and con ol hal - plan s (powe equals 0.95). G eenhouse-g own seedlings (30 d old) om ea ed hal -sh ubs we e signi ican ly smalle han hose o con ols in wo ou o i e cases (Fig. 5). The e we e, howe e , no o e all signi ican ea men (Ms = 300, d = 1, F = 1.9, P = 0.24) o block (Ms = 323, d = 4, F = 2.1, P = 0.25; e o MS = 155, d = 4) e ec s. Again, his analysis is p one o Type II e o , wi h TABLE 4. (A) B ac s and pollen expo in La andula sh ubs spli in o hal es (b ac s/no b ac s). The a e age numbe s o pollen g ains emaining in he an he s o 10 lowe s pe hal -sh ub a he end o he i s and second day o an hesis we e subjec o a epea ed-measu es analysis (B), wi h spli sh ubs ep esen ing blocks (see Fig. 2c). A) Undispe sed pollen g ains pe lowe on hal -sh ubs Day o Wi h b ac s Wi hou b ac s an hesis (X ? 1 SE) (X + 1 SE) Fi s 1763 ? 245 1698 ? 231 Second 725 + 75 662 ? 60 B) Repea ed-measu es analysis Sou ce d MS F P Be ween subjec s T ea men 1 73578 0.36 0.55 Block 17 1121702 5.51 <0.001 E o 17 203668 Wi hin subjec s Time 1 19333151 129.3 <0.001 Time X T ea men 1 14 0.0 0.9 Time X Block 17 797423 5.3 <0.001 E o (Time) 17 149462 Con ol lowe s no exposed o pollina o s e ained 3275 ? 254 g ains. N = 18 in all cases. This con en downloaded om 193.147.173.147 on Tue, 11 Jul 2017 14:45:49 UTC All use subjec o h p://abou .js o .o g/ e ms Ma ch 1997 FLORAL DISPLAY IN LAVANDULA 501 TABLE 5. (A) Cha ac e is ics o he seeds and seedlings om La andula hal -sh ubs. In he ANOVAs (B), "T ea men " e e s o b ac emo al, and blocks a e spli sh ubs. A) Side wi h b ac s Side wi hou b ac s Va iable Yea X ? SE N X+ SE N Seed mass (mg) 1993 0.75 ? 0.06 7 0.73 + 0.05 7 1995 0.77 ? 0.02 18 0.75 + 0.02 18 Ge mina ion (seedlings pe po ) 1993 23.5 + 5.6 5 22.8 + 6.9 5 B) Dependen a iable Yea Sou ce d MS F P Seed mass 1993 T ea men 1 0.2 0.7 0.67 Block 6 409.7 53.4 <0.001 E o 6 7.7 1995 T ea men 1 21.1 1.6 0.22 Block 17 209.1 15.8 <0.001 E o 17 13.3 Ge mina ion 1993 T ea men 1 1.1 0.1 0.87 Block 4 360.7 9.3 0.03 E o 4 38.9 s a is ical powe being only 0.72. O e all seedling mass was 86.2 ? 7.9 mg o con ols, and 75.2 ? 5.8 mg o ea ed hal -sh ubs. DISCUSSION Visi a ion a es Mean insec isi a ion a e dec eased by 23% in b ac less sh ubs as compa ed o con ols, so he answe o he i s ques ion add essed (do La andula b ac s enhance isi a ion?) mus be yes. This is in ag eemen wi h p e ious epo s on he ela ionship be ween plan ad e isemen and pollina o isi a ion (Willson and P ice 1977, Augspu ge 1980, Thomson 1981, Go i 1983, Wase 1983, Schemske and Ag en 1995). Ne - e heless, di e ences in his s udy we e no g ea E~~~~~ U) 100 50- Cl) 1 2 3 4 5 PLANT IDENTIFICATION NUMBER FIG. 5. Mean seedling sizes om spli (b ac s/no b ac s) La andula sh ubs. Open ba s a e con ol, and shaded ba s a e ea ed hal -sh ubs. Ve ical lines span 2 SE. Means signi i- can ly di e en a P < 0.05 a e ma ked wi h an as e isk (S uden 's es ). Plan s 1-5 a e om Aznalcaza . enough o each s a is ical signi icance, which could be conside ed jus an unwelcome a e ma h o ex ensi e among-plan a iance and subsequen low s a is ical powe o es s o , al e na i ely, migh be e idence ha he (o he wise ha dly ques ionable) posi i e e ec o display on pollina o isi a ion need no be s aigh - o wa d. I can be no iceably obscu ed i , o example, insec s emembe he loca ions o speci ic sh ubs om days p e ious o he expe imen and hei beha io be- comes hus in pa independen o display (Janzen 1971, Hein ich 1976, Eickwo and Ginsbe g 1980, Thomson 1988). Addi ionally, and simply because o nea es -neighbo o aging (Le in and Ke s e 1969, Wadding on 1983), con ol sh ubs wi h b ac s p obably "passed" isi o s o i s b ac less coun e pa s, which may also ha e esul ed in o a enua ed di e ences. The o e whelming e ec o b ac s on pollina o choice demons a ed in spli -sh ub expe imen s, on he o he hand, sugges s ha hese o gans may be ex emely use- ul o sh ubs ha need o be app oached om a dis ance (e.g., du ing ea ly popula ion es ablishmen s ages). Pollina ion in ensi y and pollen- ube numbe s B ac s a e no c i ical o he numbe o pollen g ains adhe ed o he s igma, bu hey a e o he numbe o ubes in he s yle. The e a e no many epo s on he size o mic ogame ophy e popula ions wi hin pis ils o na u ally pollina ed lowe s, bu conside able in e plan a ia ion has been shown on occasions (Le in 1990, Niesenbaum 1994). In La andula, a iance in mean s igma ic loads was compa able o ha o pollina o isi a ion a es (c = 50 and 63%, espec i ely), whe eas ube numbe pe s yle was mo e cons an (c = 27%). The p oxima e cause o his ela i e lack o a ia ion is p obably ha he iny ansmi ing issues (0.2-0.3 mm in diame e ) inside s yles we e unable o This con en downloaded om 193.147.173.147 on Tue, 11 Jul 2017 14:45:49 UTC All use subjec o h p://abou .js o .o g/ e ms 502 JAVIER HERRERA Ecology, Vol. 78, No. 2 accommoda e as numbe s o ubes, and s yles con- ained as many ubes as hey could hold. Judging om mic oscopic obse a ions, he e was i ually no ee space in s yles con aining 20-24 ubes, which p obably ep esen s an uppe limi o his species. A p esen I can only specula e on why s yles om b ac less sh ubs ha e ewe ubes han con ols in spi e o equi alen s igma ic pollen loads. A possible expla- na ion is ha b ac emo al dis up ed a hypo he ically op imal c ossing dis ance (Wase and P ice 1983, 1991, Dudash 1990). Pa ial suppo o his idea comes om pollina o -choice expe imen s, in which con ol hal - sh ubs ecei ed dis an pollina o app oaches mo e o - en han hei b ac less coun e pa s. I dis an pollen a i es less equen ly o b ac less sh ubs and he g ow h o ubes om spa ially close neighbo s is some- how discou aged, he e would, on a e age, be ewe ubes in b ac less sh ubs. E en a weak ejec ion o close-neighbo ubes migh educe mean ube numbe in his way, bu he exis ence o such a mechanism emains o be demons a ed. Male unc ion Pollen p oduced pe lowe is in La andula ela i ely scan (4000 g ains, 83% o which was dispe sed a he end o he 2nd d o an hesis). Fu he mo e, pollina o s p obed he same numbe o heads on con ol and b ac - less hal -sh ubs, and his made changes in pe - lowe isi a ion a es unlikely (a p econdi ion o inc eased pollen expo ; S an on e al. 1986, C uzan e al. 1988, Young and S an on 1990). Acco dingly, he a e age a e o pollen expo was unal e ed by b ac emo al and he hypo hesized e ec o inc eased display on he success as a male mus be ejec ed (ques ion 3). Ob- iously, o e dis ances as sho as plan diame e (0.3- 0.6 m) co ollas p o ide enough ad e isemen , and heads a e no missed by pollina o s e en i hey lack b ac s. Fecundi y and p ogeny quali y B ac emo al had no signi ican e ec s on he e- cundi y o indi idual La andula sh ubs. Li le o no change in seed p oduc ion ollowing expe imen al ma- nipula ions o ad e isemen in ensi y a e no uncom- mon (Wase 1979, Ande sson 1991, Mendez and Obeso 1992) and, a leas in La andula, his p obably s ems om ui p oduc ion being mo e esou ce han pollen limi ed. Fo example, he abo ion a e o la e-opening lowe s can be educed by emo ing de eloping o a ies on he same head (Muiioz and De esa 1987), and a - i icially e ilized sh ubs inc ease hei ecundi ies el- a i e o un e ilized con ols (J. He e a, unpublished da a). Howe e , his is no o say ha b ac s a e gen- e ally useless in his espec , only ha sh ubs in he s udied popula ions we e o some ex en o e ad e - ised. Wi hin-sh ub di e ences among b ac less and con- ol hal -plan s ega ding he igo o he p ogeny we e signi ican (and in he expec ed di ec ion i b ac s in- c eased seed quali y; ques ion 5) only in wo ou o i e cases. O e all, he e ec on p ogeny quali y was nonsigni ican , bu hese kinds o epe cussions can be ha d o demons a e in La andula because o ex ensi e a ia ion be ween sh ubs and ensuing high a e o Type II e o . Sowing he seeds in he ield ins ead o he g eenhouse may ampli y di e ences (e.g., Dudash 1990, Mon al o 1994), bu i is ob ious ha a gene ic app oach would help o ge a be e knowledge o his issue. The e o e, he las ques ion add essed a he begin- ning o his s udy (i.e., do b ac s con ibu e o inc eased p ogeny igo ?) canno be unambiguously answe ed wi h he a ailable da a. Since he e we e ewe pollen ubes wi hin s yles o b ac less sh ubs, howe e , and because ube numbe has a numbe o implica ions on p ogeny de elopmen (e.g., h ough mic ogame ophy e compe i ion, Hill and Lo d 1986, Snow 1986, Mulcahy e al. 1983; and/o h ough emale choice a ei he p e- o pos zygo ic s ages, S ephenson and Be in 1983, Ma shall and Folsom 1991, Snow 1994, o example), he e is ci cums an ial e idence ha inc eased ad e - isemen p o ided by b ac s can make a con ibu ion o i ness e en in ma u e, ela i ely dense popula ions wi h mo e han adequa e pollina o se ice. ACKNOWLEDGMENTS This s udy bene i ed om con e sa ions wi h many people abou he ole o b ac s in La andula. Thanks a e pa icula ly due o An onio Rod iguez-Sie a o his en husias ic help du ing ield wo k, and o Ramona O ega o d awing Fig. 1. The au ho g ea ly acknowledges C. M. He e a, P. G. L. Klinkhame , N. M. Wase , and an anonymous e iewe o cons uc i e commen s on a p e ious e sion o he manu- sc ip . The Es aci6n Biol6gica de Do iana p o ided pe mis- sion o wo k in he Rese e, and he Spanish Jun a de An- daluc a ga e inancial suppo h ough g an 4086 o he P o- g ama de Ayuda a los G upos de In es igaci6n. LITERATURE CITED Ande sson, S. 1991. Flo al display and pollina ion success in Achillea p a mica (As e aceae). Hola c ic Ecology 14: 186-191. Ande sson, S., and B. Widen. 1993. Pollina o -media ed se- lec ion on lo al ai s in a syn he ic popula ion o Senecio in eg i olius (As e aceae). Oikos 66:72-79. Augspu ge , C. K. 1980. Mass- lowe ing o a opical sh ub (Hyban hus p uni olius): in luence on pollina o a ac ion and mo emen . E olu ion 34:475-488. Bell, G. 1985. On he unc ion o lowe s. P oceedings o he Royal Socie y o London, Se ies B, Biological Sciences 223:224-265. Campbell, D. R., N. M. Wase , M. V. P ice, E. Lynch, and R. J. Mi chell. 1991. Componen s o pheno ypic selec- ion-pollen expo and lowe co olla wid h in Ipomopsis agg ega e. E olu ion 45:1458-1467. Caspe , B. B., and R. A. Niesenbaum. 1993. Pollen e sus esou ce limi a ion o seed p oduc ion: a econside a ion. Cu en Science 65:210-214. C uzan, M. B., P. R. Neal, and M. F Willson. 1988. Flo al display in Phyla incisa-consequences o male and emale ep oduc i e success. E olu ion 42:505-515. De esa, J. A., J. A oyo, and J. He e a. 1985. Con ibuci6n This con en downloaded om 193.147.173.147 on Tue, 11 Jul 2017 14:45:49 UTC All use subjec o h p://abou .js o .o g/ e ms