INTERSPECIFIC VARIATION IN FRUIT SHAPE: ALLOMETRY, PHYLOGENY, A N D ADAPTATION T O DISPERSAL AGENTS' 4
. Investigations on fruit and fruiting characteristics of animal-dispersed. fleshyfruited plants havc been generally interpreted in terms of adaptations by plants to dispersal
agents. Most often, these studies did not formulate non-adaptive, alternative null hypotheses. a n d the potential inllucncc of phylogenctic effects o n observed patterns was not
assessed either. This paper presents an analysis of interspecific variation in fruit shape (as
assessed bq length and width) among vertebrate-dispersed plants of the Iberian Peninsula.
Tests of predictions from both adaptive (to dispersal agents) a n d null (based o n allomctry)
hypotheses arc presented and the inllucncc of phylogenctic effects is also accounted for.
Interspccilic variation in fruit shape was unrelated to sced dispersal mode ("bird dispersed"
vs. "bird plus m a m m a l dispersed"), and predictions from adaptive hypotheses were not
supported. Variation in fruit shape did not depart significantly from that predicted by the
allomctr)-based, null hypothesis. Deviations of individual species from the predicted allometric relationship were unrelated to dispersal mode. and originated from genus- and
species-specific variation in fruit shape. There exists a considerable influence of phylogeny
o n fruit shape variation, a n d near14 half of total variance is attributable to variation among
genera within families. After accounting for phylogenctic effects. the null hypothesis still
held within taxonomic categories above the species level. Observed constancq in the relative
variation of fruit length a n d width, despite variation in dispersal mode and morphological
type. is interpreted in terms of shared morphogcnetic and physical constraints independent
of dispersal. T h e implications of substantial phqlogcnctic effects on interspecific patterns
of variation in fruit and fruiting traits arc discussed.
INTRODUCTION
Owing to the potentiallq cocvolutionar) nature of
t h e ~ relationship
r
with seed vectors (Snow 197 1 . McKey
1975. Howc and Estabrook 1977. Janzcn 1983). the
li-u~tand fruiting characteristics of endozoochorously
dispersed plants have frequentlq been examined from
an adaptationist perspecti\e. Variation in traits like
fruiting phenology, fruit s i x and color. nutritional
c o n i p o s ~ t i o nof the pulp, and type of fruiting displaq.
among others. has been often studied in recent years
(e.g.. Stiles 1980. Herrera 1982. 1987. Willson and
Thompson 1982. Janson 1983. Johnson et al. 1985.
Wheelwright and Janson 1985. Piper 1986. Dcbussche
et al. 1987. Dcbussche and Isenmann 1989. Willson
and Whelan 1990). These investigations generally attempted t o interpret observed intcrspccilic patterns in
terms of adaptations of plants to their animal sced
dispersers. Irrespective of their success in verifying
adaptationist expectations. most previous analyses
aiming to seek adaptive explanations for observed \ ariation in dispersal-related plant features havc faced at
I htanuscr~pt
recel\ed 6 hla) I99 I . re\ ~sed10 October 199 1 .
acicp1c.d 7 5 No~ember199 1
least two potentiallq important methodological problems.
Firstlq. studies testing adaptationist hqpothcscs related to variation in fruit and fruiting traits generally
havc not formulated altcrnati\c. non-adaptive hypotheses that could plaq the role o f "null hypotheses"
(scnsu Strong 1980). This fact must probablq be attributed, in part. to the influence of adaptationist
traditions in recent evolutionary ecologq (Gould and
Lewontin 1979. Herrera 1986). but also to the formidable practical difficulties involved in formulating
biologicallq reasonable null hypotheses when complex
traits, like most fruit a n d fruiting features, arc in\ olvcd.
And secondlq, most investigations were conducted in
interspecific contexts a n d , therefore. used species a s
the "units" for analyses and comparisons. Species.
however. are not statistically independent entities, as
they are all related among themselves t o a variable
degree and belong t o a c o m m o n , hierarchically organized phylogenq (Felsenstein 1985). Similarit) between species maq be d u e t o parallel and convergent
e\olutionary change in response t o similar selective
pressures (the implicit assumption in most studies using species 3s the units for analyses). but also t o comm o n inheritance (Felsenstein 1985. Pagel and H a r \ ey
October 1992
INTERSPEC IFIC VARIA TlON IN FRUIT S H W E
19880). The use of species as statistically independent
units tends t o overestimate degrees of freedom in statistical analyses a n d , therefore, t o o \ eremphasize the
extent of convergence or parallel evolutionary change
(Pagel and H a r \ e y 19880. Burt 1989). Spurious interspccilic pattcrns. o r "taxonomic artifacts." may thus
emerge simply a s a consequence of "phqlogenetic effects" ifone does not account for the non-independence
of the taxonomic entiries used in the anallses. T h e
influence of phqlogenetic effects on interspecific patterns of fruit and fruiting traits has received little attention to date. but even the few preliminary studies
available suggest that it may often be substantial (Herrera 1986. 1987. Gorchov 1990). For other reproductive plant traits not related t o dispersal, the importance
of phylogenetic effects as explanations for interspecific
patterns has been demonstrated whenever specific attention has been paid t o them (Hodgson and MacKey
1986. Kochmcr a n d Handcl 1986. M a r e r 1989. 1990.
Stratton 1989).
1 present in this paper a n analysis of interspccific
\.ariation In a single f r u ~ ttrait, namclq fruit shape.
(Throughout. 1 use "fruits" in its ecological, not botanical sense, to denote "functional fruits." i . ~ . pack.
ager made up o f seeds plus accessory tissues used a s
food b> vertebrate disperscrs, irrespective of their a n a t o n i ~ c a ol r i g ~ n .Rather
)
surprisingly. this character has
n e \ c r been examined in detail before despite its evident
sirnpl~citqand potentla1 ecological signilicance (see If1.porilc,.c~>c
u t ~ dprc~/~cfiotic
below). It is a c o m m o n observation that, with a few noticeable exceptions from
P1pc7r).the ripe fleshy
tropical habitats (c.g.. C~'c~cro~piu,
fruits of animal-dispersed plants typically arc roughly
spherical or ellipsoidal in shape. This usually holds not
onl! for "true" fruits originating Srom a single ovary
like berries o r drupes. but also for fruits of varied a n atomical origins (consider-. e.g.. .Illtlipcr~c.5strobili. E'iC.L/\ syconia, and fir.agarra pseudocarps). This general
uniformity in fruit shapc would perhaps suggest the
existence of selective pressures from the animals that
cat them and disperse the enclosed seeds. leading to
convergence. Variations occurring within the general
spherical-ellipsoidal fruit template might also be exp l a ~ n e da s adaptations to disperscrs. Alternatively. patterns of variation in fruit shape might reflect allometric
constraints a n d / o r phqlogenetic effects.
Fruit shapc has some advantages in relation to other.
more complex fruit traits. for conducting a rigorous
examination of adaptive hypotheses. Although an accurate description of fruit shapc would obviously rcquire further measurements. it may conveniently be
s u m m a r i ~ e dby its two dominant linear dimensions.
namely fruit length and fruit width (transversal diameter). These variables describe the fruit's "aspect
ratio." a n d roughly define its shapc from side \ i c w
(throughout this paper "fruit shapc" will be used in
this restricted context). Furthermore. one straightforward null hypothesis may be formulated based on al-
1833
Iornetrlc considerations. R1y objective here is to test
both adaptive and null hqpotheses related t o variation
in fruit shapc among animal-dispersed plants from the
Iberian Peninsula. dissecting the relative importance
of allometry. phylogcnctic effects. and adaptation to
disperscrs, as explanations for observed pattcrns of
variation. This is the first investigation of thcsc characteristics conducted on a dispersal-related trait of
fleshy-fruited plants.
i l l Z i r l l l ~ ~ p o t h c t r-From
r.
s ~ m p l es c a l ~ n gc o n s ~ d eratlons and assumlng that f r u ~ td e n s ~ t q remains
roughlq constant (see I p ~ ~ l i c c ~ h rofll~l)otl~c~tc'c
lit~
In t h ~ s
s c c t ~ o n ) .\ arlation of fruit length ( L ) and f r u ~ th ~ d t h
(I$? w ~ t hfresh fluit mass ( \ I ) would be descnbcd bq
the power equations
L
=
.
A , .
and
Lf
=
A,, \I
.
(1
wrherc X I and A,, are constants. T h e o r d ~ n a r qa l l o m e t r ~ c
equations arc o b t a ~ n e dbq t a k ~ n glogar~thms:
log L
log
=
=
('/?)log\I
(I1?)log\I
+ log hi.
+ log h ,.
(2)
Sol\lng for the log 11 term and rearranging. the expected a l l o m c t r ~ cr e l a t ~ o n s h ~between
p
L and I? becomes
log I f .
=
log L
+ log h,,
-
log k,.
(3)
As log I\, and log X , are constants, the r c l a t ~ o nmaq be
wrrlttcn as either
log I f - = log L
log L
=
log I t '
+a
-
u,
(4)
(5)
where a stands for l ~ g ( h , , ~ A , ) .
Allometry thus provides a simple null hypothesis
against which to test observed v a r ~ a t i o nin fruit shapc.
If variation in shape depends mainly o n allometric
constraints ("null hqpothesis" hereafter), the prediction may be advanced that the slope of the regression
line between log fruit length and log fruit width should
equal unity (from Eqs. 4 and 5 ) . Departures from this
p r c d i c t ~ o nwould reveal shape variation that cannot be
accounted for by allometry alone.
(21 :t(/upti\,c l ~ ~ . ~ ) o r h c-Gape
~ s c ~ . ~width
.
of vertebrate
dispersal agents sets a fairly rigid upper limit t o the
si/c of fruits that can be grasped and swallowed. Among
comparatively small \crtebrates that swallow fruits
whole. like most frugivorous birds, interspecific correlations between gape width a n d mean size of ingested
fruits have been often reported (Wheelwright 1985.
Jordan0 1987h, Debusschc and lscnmann 1989. Lambcrt 1989). and sire is also oftcn a n important determinant of fruit choice bq frugivorous birds in experimental conditions (McPhcrson 1988). Furthermore.
interspccific differences in the composition of the dis-
HEKKER4
perser assemblages of bird-dispersed plants are related
to \.ariation in liuit s i ~ (Kantak
e
1979. Pratt and Stiles
1985. Lanibert 1989). a n d differences between species
In r c l a t ~ v esced dispersal success arc sometimes related
to \ anation In f r u ~ st i x (Herrcra 1984. Piper 1986).
Fruits are usuall! swallowed lengthwise by birds.
hence frult width. rather than f r u ~ length,
t
is probably
the d ~ n l e n s i o nniost directly influencing whether a f r u ~ t
can be ingested by a'given disperser o r not (Wheel\\right 1985). In some bird-dispersed species, mean
l i i ~ i twidth of individual plants is correlated with fruit
crop renioval rates (Piper 1986. J o r d a n o 1 9 8 7 ~Her.
rera 1988. Obeso 1988). One would thus expect that
selecti\e pressures for keeping fruit sire within the
"s\+allowablc" range, if any. would be stronger on fruit
w ~ d t hthan on fruit length. At this regard. Obeso (1988)
lound that i n d i v ~ d u a lvariation in fruit width. but not
in fruit length. was related to variation In fruit crop
rcrno\ al sate In a southern Spanish bird-d~spersedplant.
It ma! then be predicted that, a m o n g species dispersed
I>! small- to niediurn-si/ed frugivorous birds that swallo\\ the f r i ~ ~whole.
ts
fruit width should tend t o lncreasc
proportionallq slower than fruit length with increasing
Ssuit mass (i.e.. c o m p a r a t ~ v c l ylarge fruits should tend
to be proportionall! niore elongated than srnall ones).
T h ~ sp rediction would be supported if the regression
\lope In Eq. 4 werc s~gnilicantlysmaller than unity.
In contrast. fruit si/e is probably less important in
setting I ~ n i i t sto ingestion by comparatively large Xertcbratcs that have wide gapes. like niost non-flying
mammalian f r u g i ~ores (Debusschc and Iscnmann 1989.
Herrera 1989). In this case. selective pressures on fruit
si/e. if a n ) . would be expected to have similar effects
on li.u~twidth and k u i t length. T h e prediction ma!
t h ~ be
~ \a d \ anced that groups ofplants having d ~ s p e r s a l
body si/es and fruit handling
agents ~ v ~ contrasting
th
capabil~ties(e.g.. birds vs. mammals) would exhibit
ditYercnt fruit length-width relationships. yielding contrasting regression slopes for Eq. 4.
131. I ~ ~ ~ ~ l r c , u o/ ~/ 'illii~n~' i ~ o t l r -The
c ~ , c ~ ~null
. and adapt ~ \ eh >potheses described above are designed specificall! for Iberian plants and frugivorcs and. although
the! will probably hold for other plant-frugivorc sqsterns as well. thcq should not be uncr-iticallq applied
without re-e\aluating their assumptions. At least two
ofthese may not hold elsewhere. Among Iberian plants.
water content of most fruit specles falls within relati\el! narrow limits (Hcrrera 1987: Fig. 6). thus supporting the assumption ofrough constancy in fruit densit) r e q u ~ r e dby the null hypothesis. In tropical habitats.
in contrast. liuit densit! maq be niore variable (Snow
1'37 1. hlcKc) 1 Y 75). Another assumption that will not
hold uni\ersall! I S that gape width offrugivorous birds
sets a rigid limit t o the si/e offruits that can be ingested.
In the Neotropics. lor instance, a t least two major groups
of Srugi\orous birds (tanagers and e r n b e r i ~ i dfinches)
handle fruits b) crushing them in the bill (Levey 1987).
k3ecausc these birds d o not swallow fruits whole. the)
are much Ie5s gape I~rnitedIn the si/cs o f fruits the!
Ecolog?. Vol 73. N o 5
can take. a n d thc a d a p t ~ x ehqpotheses mould need t o be r e - e ~ a l u a t c dfor- those dispersal s!sterns where these b ~ r d spartlapate. I used the comprehensive data base for fleshy liuits
lion1 the Iberian Peninsula described in detail by Hcrrera (1 987). with minor additions (six species). A total
of 117 species. in 35 families and 6 4 genera. were included in the analyses. repr-esenting nearly 95Oh offarnilies. 90'Yo of genera. and 60°h of species with fleshy
fruits that occur on the Iberian Peninsula. Most taxa
not included in the sample were either narrow cndeniics o r belonged t o species complexes. including microspecies in the genera Rocel and Rir/?irs,w h ~ c haccounted
for 55(?/,1
of "niiss~ng" species (Herrera 1987). The species sample was dominated by shrubs (53.0% of spccies). Trees represented 24.8%. herbs 15.4'%~,
and woody
vines (>.8(%1.
Taxonomically. the sample was dominated
b! the farnilies Rosaceac (27 species). c'aprifoliaccac
( 13). L~liaccae( 10). Rharnnaceae (7). and Solanaceae
(7). Information o n geographical provenance and sarnpling methods, and a list of specles. may be found in
Herrera ( 1987).
T h e information available for each species lncludcd
length and width (measured with calipers t o the nearest
0.05 rnni) frorn at least 30 fully mature, i n d i v ~ d u a l
fruits collected from several (usually 5-10) plants of a
single population. In a fe\v cases (three specics). onl!
average values for fruit length and width were available. and these species were dropped frorn some analqses requiring detailed information from individual
Sruits.
In the Iberian Peninsula. seed dispersal of fleshqfruited plants is performed by birds and terrestrial
mammals (Hel-rera 1987. 1989). Bird dispersal is mainly
accomplished by small- (body mass 12-1 8 g) t o medium- (80-100 g) si/ed species in the genera T~rrc/ir.\.
all of which swal.SlY~,ia,I<ritlrel(.ir\, and I'lioc~rrrc,~ir~r.~,
low the kuits whole (Guitian 1984. Hcrrera 1984. 1985.
J o r d a n o 1984, 198711. 1988. Fuentes 1990). There are
n o instances of avian dispersers handling fruits by
crushing them in the bill prior to swallowing ("mashers"
sensu Leve? 1987). With minor exceptions (e.g.. part~cipationof rabbits. O r ~ ~ c ~ / o l a (, q. ~~/rrsl i c ~ ~in
/ l lthe
r ~ , dispersal of some species; R . c'. Soriguer. pi.rsor1c11 c.ottimammalian dispersal is performed by
ttr~/rri(~c~~iori).
carnivorous m a m m a l s in the families c'anidae ( 1 '~rlpcs.
C'urli.\). llrsidae (C~r.s!i.\).Xlustelidae (.\lilt.t~s..\lc,lc,).
and V i ~ e r r i d a e((;c3rlc7trcl) (Herrera 1989. and referenced therein). Species in the sample were categorized
Ihr the ana1)scs into two broad groups according t o its
known seed dispersal agents, nanicl> "bird dispersed"
(.\' = 73 species) and " b ~ r dplus m a m m a l dispersed"
( 4 4 species). Assignment o f s p e c ~ c to
s groups was mainI? based on published information (Herrcra 1989. and
October 1992
INTERSPECIFIC V.4RIATION I N FRLIIT SH.4PE
rcfcrenccs therein). but 1 also ilsed my own unpubI15hed records for some specles. Exclusi\e mammalian
d ~ s p c r \ a lhas been not proxen to date for any Iberian
flesh?-li-uited plant. If it at all occurs. it will ccrtainl?
In\ ol\ e a negligible proportion of species.
Kc:qrc,.\.\ii~/ia/~al~..cc.
Testing the predicti6ns formulated a b o \ e requires
the evaluation of regression coefficients. Three statistical techniques arc usually employed to estimate rcgrcsslon slopes: least squares regression (LSR hereafter:
the familiar Model I linear regression method): reduced
major axis regression ( R M A R : also sometimes termed
"geometric mean regression." or "standard major axis
regression"): and major axis regression ( M A R ) . T h e
three methods arc dcri\ cd from the s a m e general structural r e l a t ~ o nmodel by making different assumptions
about the nature of crror variabilitl in the variables
i n \ o l \ e d ( K u h r y and hfarcus 1977. Seim and S z t h e r
1983. R a l n e r 1985). T h e accuracy o f t h e slope estimate
pro\ ided b l a particular method will depend on the
cxtcnt to which its assumptions hold In the actual data.
and inappropriate selection of the method maq lead t o
erroneous inferences (Pagcl and Harvey 1 9 8 8 ~ ) Al.
though LSR has been the most widely used method in
allonietrl and ecolog~calstudies (e.g.. Peters 1983). it
is also probabll the least appropriate. as their assumptions will on14 rare11 be met. The relevant assumptions of the three niethods. using Eq. 3 above for
notational rckrence. are brietlq summari/ed below. For
further details. see Kuhry and Marcus ( 1 977). Sokal
and Rohlf (198 1). Scim and S z t h e r (1983). Kabner
( 1 985). and McXrdle ( 1988).
Kcgrcss~onslopc estimates from LSR assume that
log I f , but not log L. 1s measured w ~ t herror, and a
particular causation d ~ r c c t i o n is also i n \ o l \ e d (the
"predictor" o r "independent" variable is used to infer
the \-aluc of the "criterion" or "dependent" one). In
contrast. both R M A R and MAR methods allow for
error in both \ ariates. and d o not distinguish predictor
from crlterlon variables. In KMAK. the regression slope
I S e s t ~ n ~ a t eodn the assuniptlon that the ratlo ol the error ~ a r ~ a n c ol
e s the
1v.o barlables equals the ratlo of
t h c ~ ar ctual Larlances and 11 1s coniputed as the ratlo
of the standard d e \ ~ a t ~ ooln log I f to the standard
d c \ l a t ~ o nof log L M 4 K assumes c q u a l ~ t lol the error
\ arlances of the trio \ arlates and the slope r c l a t ~ n g
log
If and log L 1s an cstlniate of the slope of the major
axis of the equal f r e q u e n c ~b ~ ~ a r l a et lel ~ p s eof palred log I 1 and log L ~ a l u e s In ~ n t e r s p e c ~ fconiparlsons
~c
estlniates 01
specles
means depart Irom t h c ~ rtrue \slues because ol nieasurenient crror and s a m p l ~ n gcrror 4 s noted a b o \ e .
the d e t e r n i ~ n a t ~ oof
n the n i a g n ~ t u d eof crror for the
\ a1 lablcs under studq 1s essentlal for a correct select~on
01 the slope d c t e r n i ~ n a t ~ omethod
n
In the case of f r u ~ t
length and \ + ~ d t hmeasurement crror \+111 probablq be
n e g l ~ g ~ b l eand
.
almost all \ a r ~ a b ~ l ~\+111
t q be due to
1835
sampling error. Pagel and Harvel ( 1 9 8 8 ~ proposed
)
that calculation of within-species ~ a r i a n c e sfor each
species can pro\ ide estimates of error \ ariances in interspecilic coniparati\e studies, thus helping to select
the most appropriate regression method for intcrspccific coniparisons (see also Seim and Srrther 1983).
This method w ~ l lbe used here.
Itilrc~.\r~i~~i/~c
~.c~rcclrior~
Although the main emphasis of the present study is
on interspecific patterns. intraspecific variation in L
and I l ' will also be examined because of its relevance
to the interpretation of interspccilic variation in fruit
shape. We would tend t o recogni/c a given interspccilic
pattern a s a truly adaptive one when it is inconsistent
with the prevailing patterns ofvariation within species.
For this reason, the null hlpothcsis based on allornctry
(which, obviouslq, applies also intraspccilically) w ~ l be
l
separately tested for individual species. In this wa?.
supplementary information will be obtained for intcrprcting the results of interspecific analqses. It is ~ n i portant t o note. however. that ~ n c l u d i n gintraspecific
analyses hcrc should not be taken a s a n indication that
the same logic about selective forces on fruit shape (see
It~frod~rc~fio/~:
Il~.potl~i~cc,r
c~ri(//~rc~c/ic.tio/zs
and .-ldupf~\.(,
applies equall! inter- and intraspecificall!.
/11po1/1i~~i~~)
Intraspecilic analqscs arc merell used for determining
whether w~thin-speciesvariation In liuit shape conforms or not to expectations d e r i \ r d Srom allonietrq.
In intraspecific studies. sampling \.ariabil~tlcould
be estimated by obtaining. for ~ n d i idual
\
species. within-population variances of L and I f ' for a suficicntlq
large number of populations. I d o not ha\ e these data.
and estimates of within-species sampling variabilitq
that could help to decide on the most suitable slope
estimation niethod are thus not possible. For this reason. the three slope estimation methods will simultaneousll be used for each species. If results are consistent. confidence ma! be placed o n conclusions even in
absence of precise information on the applicabilitq of
the underlying assumptions.
Irifrt 5 / 1 c ~I//( I U I rtrtlot~utld pli) loqrr~~fl(
r f f ~15 c
4 \ a r ~ e t ! ol methods h a \ e been d e s c r ~ b c dIn recent
qears t o assess the Influence of p h b l o g e n ~on lnterspec ~ f i cp atterns (e g . Felsenste~n1985 Pagel and H a r ~ e q
1988h Bell 1989 Burt 1989 Grafen 1989 G ~ t t l e n i a n
and Kot 1990) Sonic o f t h e m requlre a d c t a ~ l e dk n o n Iedge of the ph\logcnl of the specles I n \ o l \ e d (or at
least a p l a u s ~ b l ephblogenet~ch \ p o t h e s ~ s ) .and t h e ~ r
a p p l ~ c a t ~ oto
n a t a x o n o m ~ c a l l \ \erq heterogeneous
sample such a s the one used hcrc 1s ~ n i p r a c t ~ c aMethl
o d s that Infer p h l l o g e n ~from the taxononilc h ~ e r a r c h l
and are based on the use of nested a n a l ~ s l sof ~ a r l a n c c
and co\arlancc are more appropriate In the present
Instance and w1I1 be those used here (follo\+~ng
Pagel
and Har\ e! 1988h and partlcularl\ Hell 1989 \s here
niethodolog~cald e t a ~ l sand j u s t ~ f i c a t ~ omna ? be lound)
October 1992
INTERSPECIFIC VARIATIOh Ih FRL IT SHAPE 183'
null hqpothesis also h o l d s Sor Lariation in fruit s h a p e
taking place a m o n g higher t a x o n o n i i c entities. F o r t h i s
a n a l l s i s . L a n d IT. o f individual fruits were log transf o r m e d . a n d each indi\.idual fruit w a s classilied according t o species. genus. a n d faniilq. I n d i \ idual fruits
rcprescnted replicates within species ( e r r o r t e r m ) . species were ncsted within genera. a n d genera within Samilics. N u n i b e r o f species in t h e d a t a set w a s t o o s m a l l
for e x p a n d i n g t h c t a x o n o n i i c hierarchy b e y o n d tlie
faniil) Icvel.
In tlie species set s t u d i e d , d i f k r e n c e s a n i o n g gencra
w i t h i n families a c c o ~ l n t e dfor m o s t o f t h e Lariation in
fruit d i m e n s i o n s ( T a b l e 2 ) ( 4 8 . 9 a n d 35.00:ir o f total
variance
Ibr log [ { . a n d log L, respecti\-el?). Variation
0.5 1.0
1.5
a n i o n g families ( 1 9 . 1 a n d 20.0ibi)). a n d a n i o n g species
Log,,(FRUIT LENGTH)
within genera ( 2 3 . 6 a n d 27.4%). accounted Ibr a smaller
'
I I.
Variation oflog,,,(meanfruit \\idth) \\it11 log,,,(mean p r o p o r t i o n o f total variance. T h e largest covariance
l i u ~ tlength) (both measured in millimetres) in a sample of c o m p o n e n t also occurred a t t h e g e n u s level. After acI I ' species of Iberian animal-dispersed. flesh>-fruited plant c o u n t i n g f o r t a x o n o m i c effects. t h e o\,erall. i n t r ~ n s i c
\\as fitted uslng the major asis
5pccit.s. Regression line (----)
est~mationniethod (equation: log T i . = 0.9873 log 1. 0.0190). regression s l o p e w a s 0 . 9 8 9 . n h i c h d i d n o t differ sig.Sho\\n also are the 9j0/(lbi\ariate equal-frequcnc) ellipse and. nificantll f r o m 1 ( T a b l e 2 ) . Intrinsic regressions a t c v t;\r rcli.rcnce. the line 1 = .\ (both - --).
c r l t a x o n o n i i c le\ el likewise yielded slopes n o t differing f r o m 1. T h e null h ) p o t h e s i s is t h u s also s u p p o r t e d
aftcr accounting Ibr potential plig logenetic c f k c t s d e K L I A R . t h e similarit! In c r r o r variabilities associated r-i\cd fi-on1 t h e taxononiical heterogeneit!, o f t h e s a m n i t h log L a n d log I f ' pro\,ides m o r e justilication Ibr ple a n d tlie statistical n o n - i n d e p e n d e n c e o f species d u e
ilslng M A R . a n d t h i s n i e t h o d will b e used in t h e re- t o s h a r e d ancestrq.
mainder- o f t h i s p a p e r for e s t i m a t i n g regression slopes.
In a d d i t i o n t o p r o v i d i n g a c o n ~ c n i c n tnull h q p o t h M e a n I I . a n d 1. w e r e c o n i p u t c d fbr each spccies. a n d esis. allornetr) ma! also b e used a s a subtraction crit h e rcgrcsslon e q u a t i o n \ \ a s o b t a i n e d a l i e r logarithmic terion. T h e scatter o f species a r o u n d t h e log If.- log L
tr-an.;lhrniation o f species m e a n s (Fig. 1). T h e e q u a t i o n rcgrcssion in Fig. 1 rcllects tlie inlluencc o f Sactors o t h e r
o b t a ~ n e dfor tlic functional relation between log 1 I . a n d t h a n a l l o n i e t r ) . including dispersal n i o d e a n d phhlolog I. ( m e a s u r e d in niillinictres) a n i o n g species w a s log genetic e f k c t s . E x a m i n a t i o n o f t h e regression residuals
I f . = 0 . 9 8 7 3 log L - 0 . 0 1 9 0 . T h e 95'4ir c o n S ~ d e n c c niaq t h u s pro\.ide further insight i n t o t h e potential ini n t e r \ a l for t h c slope (0.9034-1.0789) e n c o m p a s s e s fluence o f ecological a n d ph!logenetic factors o n liuit
ilnit?. h e n c e interspecific variation o f Sruit s h a p e is in s h a p e a l i e r statisticall! r c m o v i n g pure11 allonietric efa c c o r d a n c e u,ith t h e prediction f r o m t h c null h l p o t h - fccts. T h e influence o f dispersal n i o d e (bird dispersal
\ s. bird plus nianinial dispersal) a n d ph!logenetic
efcscs.
Ncstcd a n a l q s ~ so f c o v a r i a n c e ( A N C O V A ) w a s uscd fects ( a s inferred Ifoni t a x o n o n i i c a f i l i a t i o n ) o n ret o d e t e r m i n e if t h i s pattern c o n t i n u e s t o hold after gression residuals w a s e x a n l i n e d siniultaneouslq using
a c c o u n t ~ n gfor t h c c o n s ~ d c r a b l et a x o n o n i i c heteroge- a r a n d o m eff'ects nested A N O V A ( T a b l e 3). T h e effect
nelt) represented in t h c s a m p l e . a n d t o assess if t h e o f dispersal m o d e o n regression residuals w a s n o t sig-
TABLE2. Ncsted ;inal?s~sof\arlanc(: and co\arlancc of log f r u ~ twidth ( H I and log fruit length (L) for lbcrlan \ertebratcd~spersedplants.*
Mcan s q ~ ~ a r e s
Le\el
dl
Log II
Log L
Fani~l!
Cicnus
Cpec~es
Errort
Total
35
27
5l
2427
2540
1.08259
0.68343
0.14395
0.00238
0.02734
I .06 197
0.64982
0.16333
0.0071 1
0.02766
Var~ancecomponents
Mcan
product
Log I I (00)
0.96453
0.58744
0.12549
0.00160
0.02359
0.00537 (19.1)
0.01378 (48.9)
0.00665 (23.6)
0.00738 (8.4)
0.028 18
Log 1
(%I)
0.00553 (70.0)
0.01245 (45.0)
0.00757 (27.4)
0.00210 (7.6)
0.02765
Cox dr1ancc
0.0051 2
0.01 180
0.00582
0.00160
0.02434
I n t r ~ n s ~major
c
ax~s
slope
/I (95(Vo CLS)
1.016 (0.853. 1.21 1 )
0.945 (0.7 18. 1.238)
1.082 (0.828. 1.424)
0.989 (0.96 1. 1 .0 18)
* Anal!scs of \ arlance and co\ariance were perfbrrned using procedure NESTED in S.AS (SAS 1988). Intrins~cregression
slopcs at the \ariou, nesting le\els \\ere computed uslng lheir respcctl\e \.ariancc and co\ariance components and standard
formulac for mrqor asls regression In Sokal and Rohlf (198 1 : 594-599)
t I n d ~ \dual l'ru~tsw i t h ~ nspec~es.
1838
CARLOS M HERRERA
TABLE
3. Random etfects nested ANOVA for the erect of
d~spersalmodel ("bird" \ s . "bird plus mammal") and phqlogen) on the residuals ol' the fitted major axis regression
o f log f i u ~ width
t
on log k u ~ length
t
shown in Fig. 1 .
Al\rOV,\ table
i'ar~ancc
df
source*
Ll~spersalmode
F'am~l!
Cicnus
Error
S~gn~ficance
of the
Mean
square
k'
P
Variance
accounted
for
((hof
total)
0.002836 0.36 .55
0.00
0.007890 1.13 .38
6.75
76 0.006077 2.26 ,0063 40.39
51 0.003082
52.86
model: k' = 2.56. df = 65.5 1. P = ,00034
1
38
* D~spcrsalmode \+.asused as the main cfict. and famil>
and genus wcrc hierarchicall) nested \\ithin each le\el of dispersal mode. The error term corrcsnonds to variation anionrspccles h ~ t h l ngenera The model has t~tteduslng procedure
C J L M ( ~ q 111
~ sum
e of squares) and proponlons of barlance
here obtalned uslng procedure VARCOhlP (SAS 1988)
niticant. Phklogeny. in contrast. did ha\.c a significant
influence. and accounted for 47. I0/o of variance in residuals. Vanation of residuals among families within
dispersal modes was not significant. while \-ariation
among genera w ~ t h i nfanlilies was significant. Phylogenctlc eft'ects on r e s ~ d u a l swere thus alniost cntirelq
d u e to \.ariation among genera within fanlilies (40.4°/n
of residuals \.ariancc). It may be concluded that the
degree of departure of individual species from the pred ~ c t c dlog It-log L regression line is attributable t o
ph! logenetic effects alone. being unrelated t o dispersal
mode. Departures froni allonietrq are associated with
genus-specific. intrafamilial variation in fruit shape indepcndent of dispersal mode.
T o further document the conclusion that interspec ~ f i cLar~ationin fruit shape is unrelated t o dispersal
mode. 1 examined in detail the patterns of variation
in f r u ~ dimensions
t
occurring w ~ t h i nthe Rosaceae and
the Caprifoliaceac. the two families contributing the
largest number of species t o the s a n ~ p l e .In thc Caprifoliaceac. l'ruit consuniption by manimals has not
been recorded t o date for any of the 13 species in my
sample. and apparentlq all species are exclusivelq bird
dispersed. In the Rosaceae. in contrast. mammalian
dispersal occurs in at least 74 of the 77 species in the
sample. Regressions for the log 12-log I_ plots ofspecies
~ c o l o g >~. o '3.
l
NO
5
means (not shown) were coniputed separate11 for each
f a n i ~ l )(Table 4). D e s p ~ t et h e ~ rcontrasting d ~ s p e r s a l
modes. the two f a n i ~ l ~ \sere
es
s ~ n i i l a rIn h a \ ~ n gInterspec~ficrcgresslon slopes not d e p a r t ~ n gs~gn~ficantl!.
froni 1
Bi\.ar~ateregression analyses are routine11 used In
ecological research to elucidate relationships between
variables. Most often. the objective of such analqses is
just t o determine the nature and statistical significance
of the relationship between Lariables. the particular
values of regression parameters (slope and intercept)
having little or n o rele\ance to results. In other cases.
howe\ er. specific hypotheses concerning the \.slues of
regression parameters are the lbcus ofanal>,ses.In thcsc
instances. accurac? of parameter estiniates is essential.
and the r e l ~ a b ~ l of
~ t ?results w ~ l ldepend on whether
the ass,,ptlons
rcqulred b) the particular e s t ~ n l a t l o n
method used actual11 hold In the data D e t a ~ l c ddescriptions of the assumptions in\.olvcd in the \.arious
regression estimation niethods have been presented.
among others. by Seini and S z t h c r (1983), Rayner
( 1 985). McArdle ( 1 988). and Pagel and Harvey (1 988h).
These studies should niakc clear to e\.eryone that the
routinely used. simple least squares regression (LSR)
is an inappropriate cstiniation niethod for most ccological applications requiring accurate estimates o f r e gression parameters. T h e present study illustrates well
the risks of using a n Inappropriate regression method.
as its niain conclusions would have been re\rcrsed had
I used ordinary LSR estimation. Using LSR. the resulting slope for the regression of mean log TZ'on niean
log L across species (Fig. I ) is 0.888 (9j0/0 confidencc
inter\.al = 0.808-0.968). In addition. i n t r ~ n s i cregrcssion slopes at the various ta-ionomic le\.els in the nested ANCOVA (Table 2 ) likewise become significantly
smaller than unit? when LSR estimation is used. By
undercstiniating regression slopes. uncritical application of LSR would have erroncouslq led to rcject~on
of the null hypothesis and acceptance of one of the
adaptive hypotheses (predicting that fruits become niore
elongated with increasing mass).
Predictions based on dispersal-related adaptive hypotheses arc not supported by the present stud?. Instead. interspecilic variation in fruit shape (as described
by length and width) of Iberian vertebrate-dispersed
plants does not depart significantly from that predicted
TABLE
4. Summar? of interspccilic regressions (major axis est~mationmethod) of log fruit midth on log fruit length for the
tho maln families in the sample. .V = number of species.
Regress~orlparameters
Rosaccae
( 1 = 27)
Slope
(95%confidence limits)
Intercept
(9jn/nconlidencc limits)
-0.034
( 0 . 7 21 . L O . 124)
Capr~follaceae
( Y = 13)
(
0.149
1.235. +0.340)
October 1992
INTERSPEC'IF'IC' VARIATIOK IK FRUIT S H 4 P E
1839
TABLE
5 D~lferenccsIn tru~tsve (mean -r I SD) bctbcen the tno groups ol spcclcs recognlled In the sample on the bas~s
ol seed d15pcrsal mode \ = number ol speclcs
Fru~ttralt
B~rddispersed
( V = 73)
8.1 -r 2.9
Lcngth (mni)
7.4 -r 2.3
M'ldth (mm)
319 439
Fresh mass (mg)
* Statistleal tests conduCted on log-transformed data.
*
Bird plus mammal
dispersed
(.V = 44)
1.'
I'
12.4 i 6.3
11.7 6.1
1589 i 2994
32.86
38.66
39.96
a ,0001
b? a sirnplc null hypothcsis based o n allornetry. Dc\ iations of i n d i v ~ d u a lspccies from the predicted all o n i e t r ~ crelationship (regression residuals) arc also unrelated to dispersal mode. a n d are best explained bq
the existence of genus- a n d speclcs-specific variation
in fruit shape. h e s t e d analysis of covariance further
shows that. after accounting for the inlluence of common descent o n observed interspecific patterns (as ink r r e d from taxonomical affiliation). the null hypothesis still holds within taxonomic categories abo\.e the
spccles le\ el.
N o evidcncc was found that species groups. or Ihmilics (comparison Rosaceae vs. Caprifoliaceae). of Iberian plants differing in seed dispersal niodes had contrasting patterns of intcrspccilic variation in f r u ~shapc.
t
The two dispersal modes r c c o g n i z d in the species saniplc ("birds" vs. " b ~ r d splus mammals." B and BPM
specics groups hereafter). involve partly overlapping
a r r a l s ofdispersers. It might thus be argued that they
arc not distinct cnough to provide an adequate b a s ~ s
for testing the adaptive predictions forniulated here.
and that here lies the reason for the failure to find
supporting evidence. Nevertheless. species in the B and
EPM groups d i l k r significantly in average fruit length.
~ i d t h and
.
fresh mass (Table 5 ) . UPM specics having
largcr fruits than B spccies. Previous investigations have
also documented additional differences between BPM
and L3 lberlan species in other fruiting traits. including
color and nutritional composition of the pulp (Herrera
1989: see also Debussche and lsenniann 1989). T h e
two groups o f s p c c ~ e sconsidered here are thus difkrent
cnough in other fruit features a s to deny the possibilitq
that thcq could also exhibit differences in fruit shape.
d i f i r e n c e s between the two
F ~ ~ r t h c r r n o r significant
c.
groups In nican f r u ~ tsi/e are consistent with the as5urnptions that led to the adaptive hypotheses tested
herc (scc Itlrrod~rc~riorl:
/Il.porhc~rc7.s~nd.t~rc~c/ic/cotlr
and
11l~/~Or/l~~.\~~s).
l(/~i/?rl\~o
1t is striking that the allonietr~cprediction holds uniforml) in thc spccies set examined. given the broad
xarict? o f morphological fruit types represented. Out
of 117 spcc~cs.36 (39.3'/0) produce bcrries and 38
(11 .0°41) produce drupes. while fruits from the reniaining 13 species (19.7%) include a variety of morphological t > p c s (e.g.. syconla of F i c w , strobili of .Iloic()( ~ i ( . \ , arillate secds of I:'lcoti~~tt~irs,
and a number of
structural types found in the Rosaccae). An analysis of
*
Dilference*
+c ,0001
+c .OOO 1
~ a r i a n c eof the log If'-log L regression residuals of
indi\.idual species. using fruit type as a three level categoriring \. ariable (bcrries. drupes. and "others"). does
not rc\.cal significant heterogeneity aniong the three
speclesgroups ( F = 1.77. dl'= ? . I 1 = .18).Homogeneity
in the pattern of relative variation o f n i e a n f r u ~ length
t
and width in face ofcontrastinganatomical o r ~ g i n (and
s
thus. presumably. dc\.cloprnental pathways) may be
interpreted in ternis of shared constraints. Exaniination o f t h c factors influencing fruit shapc w i t h ~ nspecies
ma? help to identify these constraints.
With regard to shape. large fruits are simply scaledup \.crsions of small oncs. both within and aniong specles. This suggests that the rnorphogenetic and physical
constraints responsible for intraspecific patterns ofshapc
\ariation may also apply to interspecific oncs. Within
species. fruit volunie at rlpeness largclq depends o n the
volunie of lleshy tissue. which. in turn. depends o n cell
number and ccll volume (Coombe 1976. Esau 1977).
T h e nuniber and volume ofcells in fruit llesh at ripcness depend on the number at anthesis and the ratc
and duration of cell di\.ision and cell expansion thereafter. Active ccll division in the llesh is generally liniited to a short period after anthcsis. and ensuing cell
enlargement I S the process niost dircctlq detcrniining
linal fruit volume (Bollard 1970. Coombe 1976. Staudt
et al. 1986). Before cellular expansion takes place. fruit
shape depends closely o n the shape of the ovary (and!
or ancillary structures in\.olvcd) and the orientation of
ccll division planes. T h e final shape. however. will be
largely determined bq internal hydrostatic pressure and
mechanical stresses o n the fruit surface (generated by
intense solute and water accumulation during cellular
enlargement) (Considine and Brown 198 I). Physical
models demonstrate that surface stresses are strongly
shape dependent. and that they are m i n i m i ~ e din fruits
with a Icngth,'width ratio of u n ~ t q(Considine and Brown
198 1 . Considine 1982). Independently of fruit mass
and anatomical structure. the combined action of internal hqdrostatic pressures and surface stresses operating on a relatively adjurtable niass of expanding
cells will therefore lead to the observed convergence
o n Icngth,'width ratios close t o unit?. In interspecific
coniparisons. deviations from this prevailing pattern
will reflect species-specific \.ariation in ovary ( a n d ' o r
assoc~atedstructures) shape. organization of cellular
di\.ision plancs. and duration o f t h e cell di\.ision phase.
1840
CARLOS M. HERRERA
-
Ecologh. Vol. 73. KO. 5
Second e d l t ~ o nJohn
.
n'lley R Sons. Kem York. Nem 't'ork.
I ISA.
Considlne. J . A. 1982. Phbslcal aspects of kult growth: cuticular ti-acture and fracture patterns In relation to fruit
structure in l r r r c ~ , r n i f i ~Journal
a.
of Horticultural S c ~ e n c e
57:79-9 1 .
C'onsidine. J.. and K. Brown. 1981. Phgs~calaspects o f l i u ~ t
growth: theoretical anal!s~s of distribution of surface growth
forces in fruit in relation to cracking and splitting. Plant
Ph1slology 68:37 1-376.
C'oombe. B. G. 1976. The development of fleshy fruits.
Annual Reviem of Plant Physiolog1 27:507-528.
Debussche. M.. J. Cortez. and I . Rimbault. 1987 Variation
In Heshy fruit composition in the Mediterranean region: the
importance of r i p e n ~ n gseason. life-form. fruit type and
geographical distribution. O ~ k o s49:244-252.
Debussche. M.. and P. Isenniann. 1989. Flesh1 k u ~ chart
acters and the choices of bird and mammal seed dispersers
in a Mediterranean reglon. Oikos 56:327-338.
Esau. K. 1977 Anatom! of seed plants. Second edition.
John M'ile) & Sons. New York. Nem York. IJS4.
Felsenste~n. J. 1985. Ph1logcnles and the comparative
method. American Naturalist 125: 1-1 5.
Fuentes. M. 1990. Relaciones entre p a p r o s 1 liutos en un
niatorral del norte de Esparia: variacihn estaclonal !d ~ f e r encias con otras areas geograficas. Ardeola 37:53-66.
Glttlcman. J. L.. and M. Kot. 1990. .Adaptation: statlstles
and a null model for estimating ph!logenetic effects. S l s teniatic Zoolog! 39.227-24 1 .
Gorehov. D. L. 1990. Pattern. adaptation. and constralnt
in frultlng sknchrony mlthin vertebrate-dispersed mood1
plants. Oikos 58: 169-180.
Ciould. S. J.. and R. C. Lewontin. 1979. The spandrels of
San Marco and the Panglossian paradlgni: a critique of the
adaptatlonist programme. Proceedings o f t h e Royal Soclcty
of London. B. Biological S c i e ~ c e 205:58
s
1-598.
Grafen. A. 1989. The phklogenetic regression. Ph~losophical
Transact~onsof the Royal Society. London. B 326:l 19157.
Gultlan. J . 1984. Ecologia de la c o m u n ~ d a dde passeriformes
d e un bosque montano de la Cordillera C a n t a b r ~ c aocciFor Srultful diseuss~onson allometr?. regression. and comdental. Dissertation. University of Santiago. Santiago. Spain.
paratlx e niethods. I an1 most grateful to Pedro Jordano and
Herrera. C'. M. 1982. Seasonal variation in the quality of
Luis Lope/-Soria. Pedro Jordano. Doug Levey. Douglas
fruits and diffuse coevolution betbeen plants a n d avian
Schemske. and an a n o n l m o u s revlewer provided helpful
dispersers. Ecologl 63:773-785.
comments and suggestions on an earlier version o f t h e manu,
1984. A stud! of avian liug~vores.bird-dispersed
serlpt. hl) l n ~ t l a ladaptatlonlst expectations with regard to
plants. and thelr Interaction In Mediterranean scrublands.
f r u ~ st hape mere first Inspired by studles on seed dispersal b1
Ecological Monographs 54: 1-23.
carnivorous mammals (supported In part b1 the Consejo Su. 1985. Habitat+onsumer interactions in fruglvorous
perior de Invest~gacionesC'lentificas. Financlacion Basal 1986b ~ r d s Pages
.
341-365 rn M . L. Cody. editor. Habitat selec1087). and later nurtured b! sonie erroneous preliminar)
tlon In b ~ r d sAcademic
.
Press. Nem York. Nem York. IJSA.
anal1ses. mhlch. by uncrlt~call! a p p l y ~ n gLSR estimation
. 1986. Vertebrate-dispersed plants: mhy the! don't
methods. provided befitting underestimates of regression cobehave the may the! should. Pages 5-18 rtl A. Estrada and
eficlents. I a m grateful to Manolo Carrion for h ~ ind~spenss
T . H. Fleming. editors. Fruglvores and seed d~spersal.Dr.
able help d u r ~ n gstudies o n niammallan frugivon and seed
LV. Junk. Dordrecht. The Netherlands.
dispersal. L'v'hile preparing thls paper. I mas supported by grant
. 1987. Vertebrate-dispersed plants of the lbcrlan
PB87-0457 from the Dlrcccihn General de Investigaeion
Peninsula: a stud1 of fruit characterist~cs.Ecological MonoC'lentifica ). Ti-cnlca. Mlnlsterio d e Educaclon y Clencia.
graphs 57:305-33 1 .
. 1988. The fruiting ecologl of Os~~rrc
y~tudr.ll,arrrtu:
individual varlatlon and evolutionan potential. Ecolog1
69:233-249.
Bell. G . 1989. A comparative method. American Naturalist
. 1989. Frugivon and seed dispersal b! carnivorous
133:553-571.
mammals. and associated fruit characteristics. in undisBollard. E. G. 1970. The ph1siology and nutrition of deturbed Mediterranean habitats. Oikos 55:250-262.
\eloping l i u ~ t s Pages
.
3 8 7 4 2 5 it1 A. C'. Hulnie. editor The
Hodgson. S . G.. and J . L. M. MacKey. 1986. T h e ecological
biochemistr? of fruits and their products. Volume 1. 4 e specialil-ation of dicot!ledonous families mithln a local floademic Press. London. England.
ra: sonie factors constraining optiniil-ation of seed si/e and
Burt. 4 . 1989. C'oniparatlve niethods using ph1logenetlcall1
their possible evolutlona~? signilicance. New Phktologist
independent contrasts. Oxford Surveks in Evolutionan BI104:497-5 15.
olog) 6133-53.
Home. H . F.. and G. F. Estabrook. 1977. On lntraspecilic
C'onoxcr. M'. J . 1980. Practical nonparametrie statistics.
d~lrin gt h e e a r l i e r s t a g e s o f fruit d e v e l o ~ m e n t ,All
. these
f a c t o r s a r e e x p e c t e d t o b e closel) l n l l u c n c e d b? p h ? l o g e n ) . w h i c h is i n a g r e e m e n t w i t h t h e results f o u n d
h e r e ( T a b l e 3).
N e s t e d a n a l l s e s o f \ a r l a n c e o f log fruit length a n d
\I ~ d t h a \ e r e \ e a l e d a c o n s ~ d c r a b l ep h > l o g e n e t ~ c o r n p o n e n t In t h e \ a r l a t l o n o f t h e s e t r a l t s Nearlq h a l f o f
t h c l r t o t a l h a r l a n c c is attributable t o \ a n a t i o n a n l o n g
g e n e r a w i t h i n families. w h i l e v a r i a t i o n a m o n g s p c c i e s
\ \ i t h i n g e n e r a . a n d a n i o n g families. is less i m p o r t a n t
( T a b l e 9 ) . T h e s e ligures suggest t h a t . i n t h e s a n i p l e
e \ a m i n e d . d l \ c r s l l i c a t l o n o f fruit s i z e h a s main11 o c c u r r c d a t t h e g e n c r l c lchel. F u r t h e r s t u d l c s a r e n e e d e d
o n o t h e r 5pccles a s s e m b l a g e s a n d o t h e r fruit f e a t u r e s
b c f o r c t h e gcnerallt) a n d l m p l l c a t ~ o n so f t h e s e results
c a n b e p r o p e r l ) a s s e s s e d . b u t t h e ~ n f l u c n c co f p h q l o g e n )
o n lnterspecllic x a r i a t l o n In fruit t r a l t s IS p r o b a b l q
greater t h a n o r d l n a r ~ l )acknowledged (Herrera 1986.
1 9 8 7 . Ciorchox 1 9 9 0 . P J o r d a n o . pc,rcotiul t o t t z t t z i r rlr~arrotl)R e c o g n l t l o n o f t h l s fact will h a h c l n i p o r t a n t
consequence5 for cholutionarq lnterprctatlons o f int e r s p c c l h c p a t t e r n s o f o c c u r r e n c e o l frult a n d frultlng
f e a t u r e s In local or regional m u l t i s p e c ~ e sa s s e n i b l a g e s
I f ~ n t c r s p e c l h ch a r l a t l o n In fruit a n d fruitlng c h a r a c terlstlcs o f a n l m a l - d i s p e r s e d s p c c l e s I S eh e n t u a l l ) p r o \ e n t o d c p c n d n i o r e o n a l l o m c t r l c effects a n d t a \ o n o n i l c
atfillation a b o v e t h e s p c c i e s level t h a n o n t h e ecological
conditions o f t h e habitats concerned (including dispersal agents). a d a p t a t i o n i s t I n t e r p r e t a t i o n s o f i n t c r specrfic p a t t e r n s i n frult a n d f r u l t l n g f e a t u r e s will h a \ e
October 1992 INTERSPECIFIC‘ VARIATlOh' I h ' F R U I T SHAPE
competition fbravian dlspcrsers in troplcal trees. American
Naturalist 111:8 17-83?,
Janson. C. H . 1983. Adaptation of liuit morpholog> to dispersal agents in a h'cotroplcal forest. Science 219: 187-189.
Jan/cn. D. H . 1983. Dispersal of seeds by vertebrate guts.
Pages 232-262 r t i D. J. Futuyma and M. Slatkin. editors.
('oc\olution. Sinauer. Sunderland. Massachusetts. USA.
Johnson. R. A.. M . F. Willson. J. N. Thompson. and R. I.
Bertin. 1985. Nutrltlonal values of wild fruits and consumptlon by migrant frugivorous b ~ r d s .Ecolog> 66:819877.
Jordano. P. 1983. Relaclones entre plantas y aves frugivoras
en el matorral mcditerraneo dcl area d c Dotiana. Disscrtation. IJnlversity of Sevilla. Sevilla. Spain.
. 19870. Avlan frult removal: effects of fruit variation. crop slre. and insect damage. Ecolog> 68: 17 1 1-1 723.
. 198711. F r u g i v o r ~ .external morphology and dlgcslive s)stcni in medlterranean sylciid marblers .Yj,lria spp.
1b1s 129: 175-189.
. 1988. Dlet. liuit c h o ~ c cand variation in body condltlon offrugivorous marblcrs in Mediterranean scrubland.
Ardea 76: 193-209.
Kantak. Ci. E. 1979. Observations on some fruit-eating birds
In Mexico. Auk 96: 183-1 86.
Kochmcr. J . P.. and S. h'. Handel. 1986. Constraints and
competition in the evolution of Howcring phcnolog1. Ecological Monographs 56:303-325.
Kuhr!. 13.. and L. F. Marcus. 1977. Bivariate linear models
In b i o m c t n . Skstcmatic Zoology 26:20 1-209.
Lanibcrt. F. 1989. Fig-eating b> birds in a Malaysian low- land rain forest. Journal of Tropical Ecolog> 5 : 4 0 1 4 I 2. LC\ c!. D. J . 1987. Seed sire and fruit-handling techniques of avian liugivorcs. Anicr~canNaturalist 129:47 1-485. I l a / c r . S. J . 1989. Ecological. taxonomic. and Ilfc hlston
correlates of seed mass among Indiana dune angiospernis.
Ecological llonographs 59: 153-1 75.
1990. Sccd mass of Indiana dune genera and families: taxonomic and ecological correlates. Evolutionar1
Ecologk 4:326-357.
IlcArdle. U. H . 1988. The structural relationship: regression
In blolog?. Canadian Journal of Zoolog1 66~2329-2339.
l1cKc.h. D. 1975. The ecology of coevolved seed dispersal
systcms Pages 159-191 / t i L. E. Gilbert and P. H . Raven.
edltors. Coevolution of a n ~ n i a l sand plants. University of
Tcxas Press. Austin. Texas. USA.
IlcPhcrson. J. M. 1988. Preferences of cedar maxmings in
the laborator! for frult species. colour and size: a coniparison mlth lield observations. Animal Behaviour 36:961969
Obcso. J. R. 1988. Exlto de d~spersionde seniillas en una
1841
poblacihn d c 1~cr.I~c~r.i~
i.irlyurrs subsp. r~uttr.uli\(Uoiss.)
He>wood. Lagascalia 15:7 1 1-71 5. Pagcl. M. D.. and P. H . Harve?. 1 9 8 8 ~ . The taxon-level
problem in the cvolutlon of mammalian brain s i x : facts
and artifacts. American Naturalist 132:344-359.
Pagel. M. D.. and P. H. Harvcl. 198811. Recent dcvclopments in the analysis of comparative data. Quarterly Review of Biolog? 6 3 : 4 1 3 4 4 0 .
Peters. R. H . 1983. T h e ecological iniplications of body s i x .
Cambridge Universlt? Press. C'ambr~dgc.England.
Piper. J . K. 1986. Scasonallty of fruit characters and seed
renioval b1 birds. Oikos 46:303-3 10.
Pratt. T . K.. and E. n'. Stiles. 1985. The influence of fruit
s i x and structure on composition of fruglvorc assemblages
in New Guinea. Blotropica 17:3 14-32 1 .
Ralner. J. M. V. 1985. Linear rclatlor~sln biomechanics:
the statistics of scaling functions. Journal of Zoology. London. A 206:4 1 5 4 3 9 .
SAS. 1988. SAS/ST,AT user's guide. Release 6.03 edition.
SAS Institute. Car). North Carolina. IJSA.
Selni. E.. and B. E. Snthcr. 1983. On rcthlnklng a l l o m e t n :
mhich rcgrcsslon model to use? Journal of Theoretical UIology 104:161-168.
Snom. D. M'. 197 1. Evolutionar? aspects of fruit-eating b>
birds. Ibis 113: 194-202.
Sokai. R. R.. and F. J. Rohlf 198 1. Blonietr?. Second edltlon. n ' . H . Freeman. San Francisco. California. IJSA.
Staudt. G.. LV. Schncidcr. and J. Leidel. 1986. Phases of
b e r n gromth in I'irr, ~,itir/i.m.Annals of Botan? 58:789800.
Stiles. E. LV. 1980. Patterns of fruit presentation and seed
dispersal in bird-disseminated moody plants in the eastern
deciduous forest. Amcrlcan Naturalist 116:670-688.
Stratton. D. A. 1989. Longevity of individual flomers in a
Costa Rican cloud forest: ccologlcal correlates and phylogenetlc constraints. Biotropica 21:308-3 18.
Strong. D. R. 1980. h'ull h>pothescs In ccolog>. S1nthi.s~
43:27 1-285.
M'heelmrlght. N. T. 1985. Fruit si/e. gape midth. and the
diets of liu~t-eatingbirds Ecology 66:808-8 18.
n'hcelwright. N. T.. and C. H. Janson. 1985. Colors o f f r u ~ t
displaks of bird-dispersed plants in two tropical forests.
Anier~canNaturalist 126:777-799.
LVillson. M. F.. and J. N. Thompson. 1982. P h e n o l o g ~and
ecology ofcolor In bird-dlspcrscd frults. or mhy some fruits
are red mhen they are "green." CanadIan Journal of Botany
60:701-713.
LVlllson. M. F.. and C'. J M'hclan. 1990. The evolution of
fruit color in flesh>-fruited plants. Amcrlcan Naturalist 136:
790-809.