Cladistics of Bactris (Palmae)

Marie Selby Botanical Gardens Inc.
CLADISTICS OF BACTRIS (PALMAE): SURVEY OF CHARACTERS AND REFUTATION OF BURRET'S

CLASSIFICATION
Author(s): Roger W. Sanders
Source: Selbyana, Vol. 12 (1991), pp. 105-133
Published by: Marie Selby Botanical Gardens Inc.
Stable URL: http://www.jstor.org/stable/41759779
Accessed: 28-01-2016 05:32 UTC
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/41759779?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
Marie Selby Botanical Gardens Inc. is collaborating with JSTOR to digitize, preserve and extend access to Selbyana.
http://www.jstor.org
This content downloaded from 165.123.34.86 on Thu, 28 Jan 2016 05:32:38 UTC
All use subject to JSTOR Terms and Conditions
12:105-133
Selbyana
CLADISTICS OF BACTRIS (PALMAE):

SURVEY OF CHARACTERS AND REFUTATION OF
BURRET'S CLASSIFICATION
Roger W. Sanders
FairchildTropicalGarden,11935 Old CutlerRoad, Miami,Florida33156; and
BiologicalSciences,FloridaInternational Miami,Florida33199
University,
Abstract. In 1934,M. Burret published ofBactris

a classification inwhich hesegregated Guilielma
andPyrenoglyphis andinwhich herecognized twosubgenera, foursections, andtwosubsections within
s. str.Hissupport
Bactris , which
ofGuilielma includes theediblespecies centered around B. gasipaes,
has
sustaineda long-standingcontroversy.Herein, is tested
thisclassification byapplying parsimony-based
cladistic
analyses to49 representative otu's,scored
species-level for106characters, which werepolarized
bytheoutgroup genera, AstrocaryumandDesmoncus. Bactrisis monophyletic onlyifneither Guilielma
norPyrenoglyphis Burret's
is segregated. subgenus andsection Bactris areparaphyletic becausehisother
taxaarenested within them.Thecladisticanalysissuggeststhere arefourmajorclades,noneofwhich
corresponds toBurret's
directly groups.Threeofthefour arestronglysupported byautapomorphies. Because
theinterrelationshipsamongthemajorcladesarebasedona few,homoplasious synapomorphies,segre-
ofanycladeisunwarranted.
gation Severalotherwell-marked constituent someofwhich
clades, correspond
toBurret'sgroups,arenestedwithinthemajorclades.Theanalysis reveals thatBurret weightedtooheavily
characters
certain andincorporated toofewofthecladisticallymostreliable Thereliable
characters. char-
arecomponents
acters ofdiverseorgansbutareconcentrated in a suiteassociated withthefruits;they
include:
presence ofan ocrea,petiolarspinesinthree ranks,structure ofleaflet ofrachis
apices,structure
bracts
andrachillae, ofthereduced
arrangements structure
cincinni, ofstaminate petalsandstamens, shape
andindûment ofpistillate
corollas,
presenceofa staminodial ring, fruitshape,epicarp color,mesocarp
composition,endocarp andthestructure
shape, andattachment ofendocarp fibers.Theremaining characters
areconsistent onlyat lowertaxonomie levelsanddeserve moredetailed analysis infuture monographic
studies.
BactrisJacq. ex Scop., the largestgenusof in GuilielmaMart,werecompletelyexcluded

Americanpalmswithover 250 describedspe- fromGuilielmaKarst.Epithetspublishedunder
cies,urgentlyneedstaxonomierevisionat both bothhomonyms have been appliedindiscrimi-
the specificand infragenericlevels (Clement, natelytothecultivated plants,andtherehasbeen
1988;Uhi & Dransfield, 1987). Generalcollec- littleregardforthecorrectnameswhentreated
torshavetendedto avoid speciesofBactrisbe- as combinations in Guilielmavs. Bactris(Mora-
cause notonlyare theyusuallyspiny,butthey Urpi& Clement,1981;Clement,1988).
also are bulkylike manyotherpalms.Hence, Anyrecenttaxonomist, ecologist,or agrono-
materialwithwhichto dis- mistwhomustdeal withtheseintractable
thereis insufficient prob-
variationalpatterns.
cernintraspecific Nomen- lemsinBactrishashadtoconsulta compendium
claturaltypesof numerousspeciesnamed by byBurret(1933-1934).Thisworkresulted from
Wallace,Barbosa Rodrigues,and Burrethave theneed forthenumerousspecimensthathad
been destroyed.Furthermore, comprehensive accumulated in Europeanherbaria bythe1920's
monographic studieshaveneverbeenpublished. tobe identified. In thedecadepreceding theSec-
The combination of thesethreefactorshas led ond World War, Burretproposedincreasing
to taxonomiechaos. numbers ofnewspeciesin an attempt to sortout
The taxonomieconfusionis mostcriticalin thevariation hefound.Perhapsheperceived im-
theeconomicmembersofBactris,taxacentered pendingpoliticalupheavalin Europe;thismay
aroundB. gasipaesH.B.K. The cultivated plants explainwhyhe did notmakea long-term com-
and theirwildand semi-wild relativesarewide- mitment to completing a comprehensive mono-
spreadthroughout thelowlandtropics.Ninelo- graphwithcomparative descriptionsand keys.
cal variantsfromwidelyseparatedpartsofthis To providethetaxonomie contextforhis45 new
rangehave been namedas species.To further species,he insteadpublisheda synopsis(Burret,
complicatematters, Martius(1823-1850),who 1933-1934).Itcomprised: 1)thedivisionofBac-
has been followedby severalmorerecentau- trisintoseverallargetaxa(Figure 1),2) descrip-
thors,segregatedthesetaxa as GuilielmaMar- tionsvalidating38 of his species,3) literature
tius.Karsten(1856) laterhomonymously pub- citationsto the144 previously publishednames
lishedGuilielmaKarst,non Mart,forspecies thathe accepted,and 4) specimencitationsto
witha staminodial ring.Hence,speciesincluded both. In the absenceof more comprehensive
105
106 SELBYANA [Volume12
Figure1. Phylogenetic ofclassification

interpretation ofBactris s.l.published
byBurret
(1933-1934). For
thetypelineofBactris,
autonyms,ratherthanBurreťsnames,havebeenused,inaccordance withtheicbn
(Greuter,
1988).AcronymsforBurreťs
groupsarefrom Table 1. Indicatedforeachofthegroupsis thetotal
numberofspeciesthatwereaccepted
byBurret orhavebeenpublished since1934.Thebarsandcharacters
indicate
themajordiagnosticcharacters
usedbyBurrettodistinguishthegroups.
work,ithasbecomethestandardtaxonomieref- amylocarpus"] wasfurther dividedintotwosub-

erenceon Bactris. sectionswhichare not illustrated but are dis-
Burretdealtwiththedispositionof the seg- cussedbelow).Allofthese,exceptthemonotypic
regategenerathathad beenproposedbyearlier sect.Aiphanoides Drude (B. caryotifolia Mart.),
workers and withthegenericintegrity ofBactris comprisednumerous,similarspecies. Burret
itself(Figure 1). He agreedwithMartiusthat presented Bactriss. str.,subgenusBactris("Eu-
Guilielmashouldbe recognized as a distinctge- bactris")andsect.Bactris("Acmophyllum"), the
nus based on the characters of massivestems largestgroupsat theirrespective levels,as the
and flattened fibersadnateto theendocarp.To coregroupsconforming to a typologicalconcept
deal withthoseBactrisspecieshavinga stami- of Bactris.These weresimplycontrasted with
nodial ring,whichKarstenhad placed in the thealternate groups,whichweredefined byone
homonymousGuilielmaKarst.,Burrettrans- ora fewatypicalcharacters. Unfortunately, Bur-
ferredtheminto the segregategenusPyreno- retdid notexplicitly characterize hisconceptof
glyphisKarst.Karsten(1856, 1857) had origi- thetypicalbactrid.
nallyerected Pyrenoglyphisas a monotypic genus Figure 1 summarizesBurreťs groups as
forBactrismajorJacq.which,according to Kar- branchesof thephylogenetic treethatcan rea-
sten(1856; see also footnoteto AppendixIII), sonablybe said to be implicitin hisformaltax-
hadbotha staminodial ringandan intracalycine onomy.The advantageoftransforming histax-
ring.Burretexpandedthe circumscription of onomyintoa hypothesis of relationshipsin a
Pyrenoglyphis itbythestaminodial
, defining ring phylogeneticcontextis thattheimportant ele-
only.Perhapshe hopedthiswouldendthecon- mentsof his classification can be evaluatedby
fusioncreatedby the publicationof Guilielma phylogenetic analyses.Indeed,theimpetus toap-
Karst,nonMart. praisetherelationshipssuggested byBurret come
Burretretainedothersegregates in Bactriss. fromthegermplasmexplorers and agronomic
str.,whichhe subdividedbycombining ideasof plantbreeders whodesirea resolution ofthetax-
Spruce(1871), Trail(1877), Drude(1882), and onomiecontroversy overthespecific andgeneric
BarbosaRodrigues(1903). He recognizedfour relationships ofB. gasipaes.Particularly, "What
sections distributedamong two subgenera are thelimits,taxonomieposition,closestrela-
(Figure 1; sect.Amylocarpus Barb.Rod. ["Eu- tives,and originof theedibleBactrisspp. and
1991] SANDERS: CLADISTICS OF BACTRIS 107
Table1. Species
andspecimensexamined
toobtaindataforthisstudy.
Theotu acronyms areusedelsewhere
thisarticle.
throughout Herbarium
(herb)acronymsfollow
Holmgren etal. (1981).
Name otu Specimens herb
Astrocaryum G. F. W.Meyer
(out- astr
group1)
alatumH. F. Loomis Hubbuch& Nemenyi 54,associatedliving ftg
ftg78424
Nemenyi2,assoc.livingftg87165 ftg
aureum & H. Wendl.
Griseb. & Budhoo1759
Sanders ftg
mexicanum
Liebm. ftg60415
Fantz3461,assoc.living ftg
Hubbuch& Nemenyi 56,assoc.livingftg ftg
59650
L. H. Bailey
standleyanum Hubbuch& Nemenyi 55,assoc.livingftg ftg
87157
Desmoncus
Mart,(outgroup
2) desm
Liebm.& Mart.
chinantlensis Coons1844 ftg
Kellerman s.n.9 Aug1940 bh
Moore& Bossard 6361 bh
Moore& Hartshorn 10121 bh
L. H. Bailey
isthmius 16728
Bartlett bh
Mart.
longifolius Schunke 6927 bh
Mart.
macroacanthos Amarai etal. 706 ny
Liesner& González 5694 bh
Mori& Bolten 8100 bh,ny
orthacanthosMart. Stahels.n.Aug1947 bh
polyacanthosMart. Croat21619 bh
pumilusTrail Davidse27712 ny
BactrisJacq.exScop.
GuilielmaGroup guil
dahlgrenianaGlassman dah Clement501-CR-88 ftg
H.B.K.
gasipaes gas 1173
Davis& Marshall ny
Dunlap1948 bh
& Foster
Foster 2118 bh
Gentry 47512
& Cuadros mo
Kayap370 bh
Read1413 bh,ftg
macana(Mart.)Pittier mac Foster 1731
& Foster bh
Pyrenoglyphis
Group pyre
Mart.
bifida bif Plowman etal. 12418 ny
Schunke 8374 bh
Mart.
concinna cnc Croat19329 bh,mo
Gentryetal. 52256 ny
Pranceetal. 16725 ny
& H. Wendl.
Griseb.
cruegeriana cru Beard131 bh
7
Seifriz bh
Barb.Rod.
gastoniana gst Mooreetal. 10340 bh
Drude
gaviona gav WesselsBoer1585 bh
majorJacq.complex maj Bailey& Baileys.n.Feb1921 bh
Bemaletal. 1209 ftg
HullH-1 ftg
Moore& Putz10511 bh
Sanders 1753 ftg
Sanders 1766 ftg
Sanders & Watson 3 ftg
incl.augustineaL. H. Bailey Bailey437 bh
incl.balanoidea H.
(Oerst.) Moore6543 bh
Wendl.
L. H. Bailey
incl.superior Bailey162 bh
marajaMart. mar Moore& Palmtak 10359 bh
oligocarpaBarb.Rod. oli Balicketal. 937 bh
Barb.Rod.
ottostapfeana Ott Fantz3464 ftg
Table 1. Continued.
BactrisGroup bact
balanophora
Spruce bal Berry 1468 bh
Schultes3938 bh
Schultesetal. 18304 bh
campestris
Poepp. cam Davidseetal. 16915 ny
Mori& Cardoso 17287 ny
caudataH. Wendi,
exBurret eau Langlois 1 bh
Moore6633 bh
Moore& Cordoba 6700 bh
coloniata
L. H. Bailey ein Bernaletal. 1099 ftg
coloradonis
L. H. Bailey clr Bernaletal. 1086 ftg
Bernaletal. 1097 ftg
corossilla
Karst,
complex cor Steyermark etal. 101415 bh
incl.duplexH. E. Moore Allen3357 bh
elegans Barb.Rod. ele deGranville 2591 bh
Mooreetal. 10331 bh
(L.) H. E. Moore
guineensis gui Bailey253 bh
Essig70000711-1 bh
Gentry & Cuadros 47621 mo
hondurensis
Standi. hon Beach1368 bh
Hammer 39,86 ftg
Holdridge 5123 bh
Holm& Iltis920,923 bh
Mooreetal. 10127 bh
jamaicanaL. H. Bailey jam Bailey216 bh
Evans175 ftg
Read1692 bh,us
incl.plumeriana
Mart,(com- Jiménez s.n.1 Sep1960,assoc.living
ftg ftg
binedwithB.jamaicanadueto 60647
missingdataandnearlyidentical Sanders etal. 1712 ftg
inknown
states char.) Watson 1284 ftg
longisetaH. Wendl.
exBurret Ion Holdridge 5118 bh
Moore6575 bh
mexicanaMart. mex Sanders 1767 ftg
militaris
H. E. Moore mil Hammer 67 ftg
montícola
Barb.Rod.complex mon Moore& Palmtak 10315 bh
Mori& Cardoso 17734 ny
Steyermark & Liesner 127352 mo
incl.actinoneura
Drude Krukoff8127 ny
incl.diviscupula
L. H. Bailey Johnston 1552 bh
incl.sigmoideaBurret Bernaletal. 1096 ftg
orariaL. H. Bailey ora Allen2580 bh
pilosaKarst. pil Bernaletal. 807 ny
Steyermark 99952 bh
Burret
porschiana por Hammer 84 ftg
Holdridge 5121 bh
savannarum
Britt. sav Bailey156 bh
Sanders & Budhoo1746 ftg
setosaMart. set Bailey459 bh
dosSantos& Mattos 33168 bh
setulosa
Karst,
complex stl Mooreetal. 9837 us
Steyermark 99128 bh
Steyermark & Agostini91399 bh
Steyermark & Liesner 118954 mo
incl.cuesaGriseb.
& H. Wendl. Bailey108 bh
Baileys.n.26 Feb1946 bh
Bailey& Baileys.n.13Mar1921 bh
Beard107 bh
Delan& Swabey 13142 bh
Barb.Rod.
aff.turbinocarpa tre Bernaletal. 1103 ftg
Table 1. Continued.
Aiphanoides
Group aiph
Mart.
caryotifolia car Bailey328 bh
27028
Lueîzelburg bh
AmylocarpusGroup amyl
aubletiana
Trail aub Mori& Boom14731 ny
geonomoidesDrude geo Berry& Uhi1538 bh
deGranville3775 bh
Egier& Irwin 46414 bh
hirta
Mart. hir Schultes
& Cabrera 12806 bh
pedinataMart. pec Costa680075 bh
deGranville2272 bh
Mart,complex
simplicifrons sim Asunción 196 mo
Dominguez 75 mo
Prance& Huber28496 ny
Vásquez & Jaramillo 569 mo
incl.arenaria
Barb.Rod. Mooreetal 9542 bh
incl.leutzelburgii
Burret Steyermark & Bunting 102440 bh
incl.trinitensis
(L. H. Bailey) Popenoe s.n.23 Nov1978 ftg
Glassman Sanders& Budhoo1447 ftg
Tomlinson 9-7-62B ftg
tenuis
Wallace ten Berry591 bh
Croat20722 mo
Schultes
& Black46-219 bh
Schultes
& Black46-314 bh
turbinata
Spruce tur Steyermark 90198 mo
Steyermark etal. 130208 mo
PirangaGroup pira
Mart.
acanthocarpa aca Balicketal. 924 bh,ny
Sperlingetal. 5802 ny
Barb.Rod.
acanthocarpoides acd Mooreetal. 10321 bh
alleniana
L. H. Bailey all Allen1804 mo
Allen2951 bh
Croat12462,22303 mo
Croat& Porter 16404 mo
barronis
L. H. Bailey bar Allen9,2538 bh
Bernaietal. 1019 ftg
Croat6541,10297 mo
Duke5612 mo
Juncosa1859 mo
constanciae
Barb.Rod. ens Sanders1811,1816 ftg
Oerst.
glandulosa gla Córdoba 131 bh
humilis Burret
(Wallace) hum Boom4129 ny
Costa670040 bh
Steyermark 88407 bh
Steyermark & Gibson95769 bh
Wessels
rhaphidacantha Boer rha Liesner& González 11252 ny
WesselsBoer1224 bh
theirprogenitors?"(Mora-Urpi & Clement,1981; monophyletic statusofBactriss.l. itself,

a more
Clement& Mora-Urpi,1987; Clement,1988; usefulapproachis totestthemonophyletic status
Clement, pers.comm.;Mora-Urpi, pers.comm.). and relationshipsofall ofBurret'sgroups,such
The underlying goal of thispaper is to use thata samplingofall majorlinesin Bactriss.l.
cladisticmethodology to ascertain:1) thephy- shouldbe included.To do so requiresusingnot
logeneticbasisfortherelationshipsofB. gasipaes onlythosecharacters uponwhichBurretrelied,
and 2) whether itssegregationas thegenusGui- butalso a widerangeofadditionalonesforcon-
lielmaMart,is consistent withcladisticresults. structinga broadlybased,parsimonious cladistic
Becausethiswouldinvolvetheexamining ofthe hypothesis. Thispaperrepresents thefirst phase
TABLE 2. Data matrix used in this study, with OTUs (see TABLE
1 and APPENDIX I for abbreviations, specimens studied, and
representation of taxa) scored for 106 characters, given
sequentially starting with character 1 in first column on
left (see APPENDIX II for detailed definitions).
19911 SANDERS: CLADISTICS OF BACTRIS 111
TABLE 2 (continued). Designations are: 0 = primitive; 1 (2 or

3) = derived (or multistate ordered); exceptions are the
unordered characters 1, 45, 46, 65, 75, and 76. ? = the
relevant structure missing from available specimens or
variable within the OTU. X = the precursor state unknown.
of recognizing all majormonophyletic lineages AppendixII). The rejectedcharacters and fea-

within Bactriss.l.andofdetecting important sys- turesare summarized in AppendixIII.
tematiccharacters in thegenus.It is hopedthat Cladograms wereconstructed bytheWagner-
thispaper,by providing a phylogenetic context treealgorithm(Farris,1970; Kluge & Farris,
forthe investigation of smaller,monophyletic 1969)usingthesoftware paup (Swofford, 1985)
groups,willbe thecatalystforthefuture mono- on a microcomputer. Becausethedata set was
graphicstudiesin Bactris. large,searchesforsuccessively shorter treeswere
conducted byalternating betweenlocalandglob-
al branch-swapping (swap= alt). Forsubsetsof
Material and Methods usuallynine or fewer otu's (seebelow),theshort-
est treewas foundwiththebranch-and-bound
All oftherelevantinfrageneric groups,segre- option(bandb).Optimization was by theminf
gategeneraandoutgroups (seebelow)weresam- (minimumtree length),farris (median-state
pled with representativespecies (Table 1, value,maximizing reversalsoverparallelisms),
AppendixI). Morphological datawereobserved and DELTRAN (delayedtransformation, maxi-
directlyfrom157specimens (Table 1) thatwere mizingparallelisms overreversals) options;root-
selectedto provideas completeas possiblecom- ingwas bytheancestor option.
parativedata forvegetative, floral(bothstami- To determine theoutgroup, a preliminary cla-
nateand pistillate), and fruiting structures. To disticanalysiswas conductedfortheentireBac-
assureidentifications oftheBactrisspecies,spec- tridinae J.D. Hooker.Data foreachgenuswere
imenswithannotations bypalmspecialists were extracted fromtheliterature (particularly Uhi &
givenpriority, andall specimens werecompared Dransfield, 1987)andfromlivingandherbarium
tothespecies'protologues. Thus,139specimens materialat ftg. The character statesin thenon-
representing 6 1 speciesofBactrisweretreated as spinyCocoeae werecompiledfora generalized
operational taxonomie units(otu's) intheinitial outgroup.In all resulting cladograms, Bactris,
stagesofthisstudy.Based on personalfieldex- Desmoncus, andAstrocaryum together formed a
perience,on thecomments ofotherfieldworkers clade strongly supportedby the following syn-
(e.g.,WesselsBoer,1965;Croat,1978;Galeano apomorphies:stems(primitively) caespitose,
& Bernal,1987; De Nevers,1988),and on the pistillate sepalsconnate,pistillate petalsstrongly
formation ofmonophyletic groupsoftwotofour fusedintocup or tube,mesocarpnotmucilagi-
highly similarspeciesin theinitialanalyses,cer- nous,endocarpporessupramedial. However,the
tain potentiallysynonymous taxa were com- interrelationships amongthethreegenerawere
binedas singleotu's. Thisyielded49 otu's within equivocal.The autapomorphies forDesmoncus ,
Bactrisforthefinalanalyses(Tables 1, 2). lianehabitand cirrus,and theprobableautapo-
Characters wereobtainedat magnifications up morphiesforAstrocaryum, staminateflowers
to 30 X fromdirectobservation of dried,fresh, congestedintoa terminalrachillaportionand
or liquidpreserved materialor frommanipula- sunkenin pits,suggest thatthesetwogeneraare
tion and dissectionof rehydrated fragments. eachmonophyletic. Therefore, Astrocaryum and
Originally,246 features ofgrossmorphology were Desmoncuswereeachtreated as otu's (Table 1),
surveyed.As an increasingly largersampleof taggedas outgroups, and analyzedtogether with
specimenswas recorded, features thatwereun- Bactrisotu's in theparsimony analyses.
stablewithinotu's or invariantamongotu's As advocatedby Farris(1969), a character's
weredeleted.Gap codingwas usedto transform cladisticreliability is consideredto be directly
descriptive and quantitative characters intodis- proportional tothecharacter's consistency.Thus,
cretestates.Dependingon the apparentrela- thecharacters wereweighted basedon theirunit
tionships ofthestates,characters werecodedas consistenciesin an initial analysis without
binary(simplelinear),linearordinalmultistate, weighting. To facilitateinput,the consistency
additivebinarymultistate, or branchedmulti- values wereroundedto the nearesttenthand
state(nonadditive binaryormixedordinal/non- multiplied by 10 (AppendixII). The onlyexcep-
additive)following Pimentel andRiggins (1987). tion was character96 (adnation/freedom and
The stateoccurring in bothoutgroup generawas structure of endocarpfibers), witha C value of
judged to be ancestral.Charactersthatvaried 0.7.Itwasgivena weight of10becausethechange
betweenor withintheoutgroup generawereleft fromprimitive (endocarppalewithfibers black-
unordered.The charactersinitiallyappearing ish,strongly flattened and adnateto theendo-
stablewithinotu's wereusedin an initialpaup carp)toderived(endocarpblack,withfibers dark
analysis.Those witha unitconsistency (C) less to pale, tereteand freefromendocarp)is very
than0. 1weredeleted.Theultimate datasetcom- complex, anda reversal is considered tobe highly
prised57 linearand 22 branchedcharacters, unlikely.
yielding 106 cladistic characters(Table 2, Theinitialunweighted andweighted runswere
Figure2. Branch-and-bound (bandb) ofproximal

analysis and/ormajor cladeswithin
Bactris.
Monophyletic
cladesof9 orfewer otu'swereanalyzed toobtaintheshortest foreachcladeand,inturn,
tree(s) werereplaced
byhypothetical ancestors(htu's)forsubsequentiterations
untiltheseproximal cladeswereobtained.
Givenis
theConsistency Index(ci)ofthemostparsimonious treeorequally
parsimonious trees
obtainedineachanalysis.
In all cases,equallyparsimonioustreesresult
from missing data.In casesofmultiplebandbtrees,thestrict
consensus treeofeachcladeprovides thetopologydepictedinFigure3.
conductedwiththe initialset of 63 otu's (61 resultedfrombandb,the strictconsensustree

Bactrisspp. + 2 outgroupotu's). paup 2.4 ac- was inserted in thecompositecladogram.
commodatesonly50 otu's; also, themulpars To checktheextentof totalhomoplasy,the
optionis greatly slowedbymissingdata.There- structureofthecompositecladogramwas spec-
fore,the approachchosenwas to: 1) obtaina ified,analyzedwiththedata setoftheultimate
preliminary cladogrambased on a majorsubset 49 otu's plus thehtu of Bactriss.l.,and opti-
of otu's; 2) identify well supportedmonophy- mizedby bothFARRis and deltran. To search
leticgroups;3) use bandbto analyzeeach sub- for shortertrees,branchesof the composite
clade of nineor fewerotu's (theapproximate cladogramwere swappedgloballyand locally
maximumotu's accommodatedby bandb) to (swap = alt).
findall possibleshortest treesforeach;4) replace
eachbya hypothetical ancestor(htu) inthedata
Results and Discussion
set; 5) use bandb to analyzethenextmostin-
clusivecladeina repeatedfashionuntilallclades Topologies,Clades,and Synapomorphies
couldbe runin a singlebandbanalysis;and 6)
reconstruct thetotalcladogramas a composite The initial,weightedcladogramyieldedsev-
ofcladograms ofthesubclades. eraldistalmonophyletic groupsincluding amyl
To getan initialstructure ofthewholegenus, (acronyms in smalluppercase correspond to the
a subsetof37 preliminary otu's, whichhadcom- speciesand groupsin Table 1),pira,guil, and
plete(ortheleastmissing) dataorthatweresim- a groupofspeciesin pyrewithspicateinflores-
ilarto otu's withmoremissingdata,werean- cences.Further nestingof thesewas as follows:
alyzed.Upon therecognition of subclades,the amyl withinthe nonfibrousclade (marked
remaining preliminary otu's wereincludedin predominantly bythelossofendocarpfibers, 963
thebandbanalyses.As indicatedabove,several [numberand superscriptfollow notationin
initialotu's werecombinedat thisstageto yield AppendixII and Figure 3]), pira withinthe
theultimate 49 otu's. To scorecharacters ofthe EXSERTED-ANTHER clade (namedfortheunique
htu's,htu matrices wereoutputusingboththe synapomorphy, 61), guil withjam to formthe
MiNFand FARRis options.If thetwo optimiza- NON-OCREATE clade (namedfortheloss of the
tionsdisagreed,the htu was assignedmissing welldevelopedocreaein adultleaves,72),and
dataforthecharacter. Ifmorethanonetopology thespicatespecieswithinpyre(Figure 2). pyre
114 SELBY ANA [Volume12
was nestedalongwithseveralotu's withinthe thebandbof proximalclades.This is expected

ATROPURPUREOUS CLADE(namedforthe syna- becausemanycharacters wereconsistent among
pomorphy of purple-black epicarps,88), which relatedspeciesbuthomoplasiousin thecontext
wasfurther nestedwithin thetuberculateclade ofwidercomparisons. Homoplasiesweremasked
(namedforthetuberculate endocarpsurface, 97). by the optimizedhtu's as the bandb analyses
Together,the nonfibrous clade, EXSERT- progressed butemerged intheWagner- treeanal-
ED-ANTHER CLADE,CAM + SAV (CAMPESTRISysisofall 49 otu's.
clade), and cau (caudata clade) forma more Onlyone treewas foundthatwas shorter (by
inclusiveclade(latibracteate clade) unitedby one step,<0.1% different, Figure 4). This dif-
transversely elongaterachisbracts(char.26). In feredonlyby theplacingof caudata withthe
all BANDB analysesof thisinclusiveclade, the CAMPESTRIS clade (synapomorphies: 74°, pistil
EXSERTED-ANTHER and CAUDATA CLADESwere cylindric, C = 0.14; and 75°,stylebranchesfree,
unitedby largestaminatereceptacles (char.49) C = 0.25). In the contextof the wholegenus,
as a clade,whichin turnformeda trichotomy char. 49 (large staminatereceptacles),which
withtheothertwoclades(Figure 2). unitesexserted-antherand caudata, has a C
The BANDB analysisof mostof theseclades ofonly0.17. Hence,neither positionofcaudata
resultedinonlyonemostparsimonious topology is verystrongly supported, and the strictcon-
witha highconsistency index(Figure2). In cases sensustree,whichplaces caudata as a clade
withtwo or moreequallyparsimonious topol- equivalent toexserted-antherandcampestris,
ogies,theequivocality developedbecausemis- shouldbe acceptedfortentative taxonomiecon-
singdatacouldbe assignedtwoor morefeasible siderations (Figure 5).
statesbypaup,notbecausehomoplasiouschar- Thebasicinternal structure ofBactris(Figure
actersconflicted.The highnumberofequivalent 3, Table 3) can be viewedas consisting offour,
topologies forexserted-anther+ caudata re- roughly equivalentclades.Each of theseis dis-
sultedfromthelack of staminatecharacters in tinguishable byitsautapomorphies (i.e.,synapo-
severalcloselyrelatedmembersoftheB. acan- morphiesof theconstituent otu's), butthein-
thocarpa complex.Thus,theoverallstructure of ternodesseparating thefourcladescontainfew,
all cladesis wellsupported. mostlyhomoplasious, synapomorphies.
The BANDB analysisofthehtu's oftheproxi- ofthefour,thebalanophora clade,
The first
mal clades(tuberculate, exserted-anther+ comprisesthe singlespecies B. balanophora,
CAUDATA, NONFIBROUS, CAMPESTRIS, NON-OCRE-whichBurretplacedin sect.Bactris.This is not
ATE,and BALANOPHORA [i.e.,Otu bal]) resulted surprising; all autapomorphies, exceptthebasal-
in a singlemostparsimonious treewitha high ly broadenedendocarp(102), occurin parallel
consistency index(Figure 2). The detailedcom- elsewhere inthegenus.However,itsretention of
positecladogram (Figure 3) is builtup fromthis theplesiomorphies- flattened,adnateendocarp
basicstructure bytheinsertion ofthebandbtree fibers (96°)andlargestemdiameter (l2)- should
or strictconsensustreeof bandb treesof the haveledBurret, whoemphasized theseso strong-
constituent clades.Becauseweights andmultiple lyin Guilielma, to placethetwolineagesinclose
statesmodify thelengthofa givencharacter, the proximity. It is surprising thatBurretdid not
totallength ofthecladogram wouldbe 606 steps noticetheendocarpcharacter becausehe cited
ifitwerecompletely consistent (i.e.,ifci = 1.0). Spruce'smaterial, whichis probably thematerial
However,the lengthwas 1,502 stepsand the fromwhichSpruce(1871) describedthemature
consistency wasmuchlower(ci = 0.447) thanin drupes.Thissuggests thatBurretdid notconsis-
Figure3. Composite cladogram forall sampled otu'sinBactris, showing detailsofcladesthatconstitute

Figure2. SeeTable 1 andAppendices I andII foracronyms andnumber codesofcharacters. Cladesalong
themainaxisofthecladogram areindicated byangled brackets.Curved bracketsmarkthebasalnodeofeach
oftheadditional cladesmentioned inTable 3: A = antillean,B = guilielma(guil),C = nonfibrous, D=
amylocarpus (amyl), E = exserted-anther, F = piranga(pira),G = campestris, H = caudata,I = militaris,
J= SETULOSA, K = COROSSILLA, L = TURBINOCARPA, M = CONST ANCIAE, N = BIDENTATE, O = LONG-PROPHYLL,
P = tubiflorou s, Q = pilosa.Dotsindicate or
unique parallel synapomorphies. Unless thechanged character
stateis indicated
bya superscript,thechange indicate
is tothestate1. Circles reversalstostate0 (unless other
state Theunitconsistency
isindicated). values(C) within Bactrisareindicated 1)underlined
as follows: numbers
areuniqueoccurrences within Bactrisoruniquereversals from theancestralsynapomorphy forBactris (C =
1.00);2) numbers not underlinedare moderately homoplasious (C = 0.5 to0.33);3) not
characters listedhave
C = 0.25to0.1.Patristic distanceportrayed includesdistances contributedbyunlisted characters,i.e.,those
withC < 0.33.Correction inproof:35 onK is a reversal.
116 SELBY ANA [Volume12
Figures4, 5. 4. Singleshorter
treeobtainedbyswapping branchesofcladogram in Figure3. Onlythe
branch
thatdiffers
from 5. Strict
Figure3 is illustrated. treeforFigures3 and4.
consensus
tently examinenon-guilielmas forthecharacter. uniquesynapomorphies (97, tuberculateendo-

The BALANOPHORA CLADEis thesistergroupto carp;106,asymmetry ofendocarppores).It en-
theremaining threeclades,whichareunitedonly compassesa widerangeofspeciesfromrelatively
byfourhomoplasiouscharacters. plesiomorphic ones to thosethatare themost
TheNON-OCREATE clade, comprising guil and divergent in thegenus.
theantilleanspeciesof sect.Bactris,is strongly
supported. Thesetwogroupssharefourunique DispositionofBurreťsGroups
synapomorphies (72,ocreaelostfromadultleaves;
9, petiolarspinesin threeranks;48, staminate To providethecontextforevaluating Burreťs
petalsrounded; and94,mesocarp thickly starchy) groups,one shouldconsiderthedispositionof
andseveralhomoplasious ones(Figure3,Table thefourmajorcladesalreadydiscussed.Theclas-
3). On theotherhand,theonlyapomorphy dis- sificatoryphilosophy adoptedhereis thatfora
tinguishing guil fromthismoreinclusiveclade taxonto be accepted,it mustbe monophyletic
is thesuppression ofpeduncular spines(char.24, orat leastnotbe showntobe non-monophyletic
C = 0.17). Althoughtheantilleanspecieshave (i.e., para-or polyphyletic) whenthe phyloge-
developednearlyfreeendocarpfibers in parallel neticstatusofthetaxonis known.Becausenone
withtherestofgenus,theystillsharewithguil of themajorcladescorresponds directlyto the
theseunusualsynapomorphies. The non-ocre- segregate generaand together forma monophy-
ate clade isthesistergrouptothetworemaining leticgroup,themostconservative viewis to ac-
cladesof Bactris , whichare unitedonlyby the ceptthetraditional limitsof Bactris , whichin-
twopartially homoplasiouscharacters, stemdi- cludesall fourclades.The alternative proposal
ametermoderate(l1) and endocarpfiberscom- oftreating thecladesas fourgenera(oroftreating
pletelyfreeand terete(962). theclade latibracteate + tuberculate as a
The LATiBRACTEATE clade (Table 3) contains thirdgenus)is undesirable. As notedearlier,the
theNONFIBROUS, EXSERTED-ANTHER, CAMPESTRIS,synapomorphies ofNON-OCREATE + latibracte-
and caudata clades. It is the least well sup- ate + TUBERCULATE and of LATIBRACTEATE +
portedofthefourmajorcladesbecausethethree tuberculate are few and homoplasious.In
defining synapomorphies sufferfromhomoplasy practicalterms,themoreplesiomorphic mem-
(26,reversing ina fewoTu's,C = 0.33; 75,stigma bersofeachcladeresembleone anotherand are
recessed, C = 0.22; 98,outermesocarpfibers few, "typical"Bactrisspecies.Thus,a morereason-
C = 0.20). able approachwouldbe to treateach cladeas a
The TUBERCULATE clade constitutes thelast subgenusor sectionofBactris.
majorlineagein Bactris.Its memberssharetwo Figure 6 illustrates that guil, pyre,aiph,
Table 3. Summary ofknown monophyletic groups Table3. Continued.

inBactris s.l.,their delimiting characters, andtheir
possible hierarchical levelsas suggested bythecla- GuiLiELMA clade(Sectional level)= Guilielma sensu
disticstructure ofBactris whentreated as a single Burret:
genus. in
Numbers squarebrackets arethechar- Peduncular spines strongly reduced orlacking [24];
acters(Appendix II); unlessindicated bya super- endocarp fibers stout, flattened andadnate[96°].
script,thenumber indicatesthe(first) apomorphic1LATiBRACTEATE clade (Subgeneric level):
state,i.e.,7 = 71.Eachcladeis characterized by Stems < 6 cm in diam.[1];inflorescence rachis-bracts
thesynapomorphies ofitsmembers. Forcladesin
Figure3 (i.e.,nodeson thecladogram) thatare (excluding muero orawn)usually transversely ob-
notnamed andcharacterized, thesynapomorphies longtolinear [26];endocarp fibers thin, terete and
aregivenundertheconstituent clades.Forex- freeor lacking [962];outermesocarp fibers few
ample, char.63 is listedas an apomorphy ofthe [98].
TUBIFLOROUS, PILOSA, andPYRENOGLYPHIS CLADES, nonfibrous clade (Sectional level):
butitarisesinthenodebelowthetubiflorous Staminate receptacle usually completely adnateto
clade. Likewise, char.7 is givenforBactris, al- petals[50°];stigmatic surface usually marginal
though itis a synapomorphy ofbothBactris and onlobes[76°*];endocarp fibers lacking [963].
Desmoncus. Reversals to apparently primitive several plesiomorphic clades at subsectional
statesthatserveas synapomorphies areindicated level,including B. hondurensis andB. colora-
by Character statesthatcontrast a cladewith doniscomplexes andunsampled ones.
equivalent cladesarenottypically listed eitherbe- AMYLOCARPus clade (Subsectional level)= Sect.
causetheyaresimply implied plesiomorphies or Amylocarpus sensuBurret:
because, insomecases,thecladeisvariable, hav- Stems<1.3 cmindiam.[Io];peduncular bract
ingbothapo-andplesiomorphies (refer toTable apexacute[19];rachillae few[28],rigid[33];
2,Appendices II andIII,andFigure3).However, triads contiguous from baseto nearapexof
a fewplesiomorphies thatareoflimited distri- rachilla [40°*,41°*,42];pistillate corollacy-
bution and,hence, contribute to theclade'sdis- lindric [66];fruit pea-sized [85].
tinctiveness anddiagnosis within Bactris arelisted exserted- antherclade (Sectional
andindicated "±" indicates more level):
byunderlining. Staminate receptacle fillinghalfofflower [49],not
orless. adnatetopetals[50];anthers small
completely
[60],exserted between petals[61];pistillate co-
Bactris(Generic level) rollacylindric [66].
Atleastjuvenile (andusually adult) leaveswith a mem- fewplesiomorphic clades at subsectional level,
branous ocreathatbecomes fibrous andoften shred- including B. mexicana complex and,at least,
dedatmaturity, leaving anabscissionlike scaracross theunsampled B. oligoclada complex.
theadaxialfaceofthepetiole(i.e.,leafbase)[7]; piranga clade = Sect.Piran-
rachillae covered withbulbous ormoniliform hairs (Subsectional level)
in three ga sensu Burret:
[35];triads ormoreranks[39],whichare Rachillae usually>40 [30],filamentous com-
usually longitudinally asymmetrically placedonthe pared to other spp. [34]; triads ± contiguous,
rachilla[43],female-sterile triads(i.e.,staminate diads
ormonads) usually distributed among andbelowthe often separated byonlyoneortwonodesof
female-fertile triads staminate diadsor monads[401*];stamens
[402,41]. free[54],usually central on receptacle [53];
BALANOPHORA clade (Subgeneiic level): pistillate corolla lepidote tosetose[71].
Stemsusually >6 cmindiam.[I2];spinesfascicled campestris clade (Sectional
[10];rachillae lanate [36];staminate petals connate level):
[47];filaments free[54];pistillate Spinesstrongly flattened [2],broad[32],andpale
calyxannulate [6];staminate sepalsstrongly connate [46°];sta-
[63],truncate [64];pistillate corolla truncate [68], minate receptacle notcompletely adnate topet-
lanate [72];stigma discoid [77];fruit mesocarp oily als[50];stamens free
ovoid[102],baseexpanded clearly [54]; pistillatecalyx
[95];endocarp [104], annulate [63];fruit pea-sized [85].
stout,
fibers flattened and adnate [96°].
NON-OCREATE clade (Subgeneric caudata clade (Sectional level):
level): Petioles hollowupondrying [8];staminate recep-
Stemsusually >6 cmindiam.[I2];ocrealacking on taclefilling halfofflower adnate
adultleaves[72];petiolar/rachis spines in three [49], completely
ranks[9]; leaflet bidentate with topetals[50°],these strongly connate [47].(Pos-
apexunequally a constituent ofeither theexserted- anther
subapically terminating midrib [15°*];staminate sibly
flowers shorter thanpistillate rounded orcampestris clade.)
[44],petals
andnotcompletely adnatetoreceptacle [48,50]; tuberculateclade (Subgeneric level):
mesocarp richin mealystarch andoil [94,95]. Stemsusually<6 cmindiam.[1];endocarp fibers
(Alsoleafsegments numerous, linear, in 3 to 6 thin,free, terete withbasalattachment forming
ranks andclustered.) tubercles on theendocarp [962,97]; twoofthe
ANTiLLEAN clade (Sectional threeendocarp poresdisplaced together on one
level): sideofendocarp
Endocarp fibers few, proximally flattened andpar- [106].
tiallyadnate, distally terete andspreading/de- militaris clade (Sectional level):
clined[96]. Leafblade4 ormoretimes longer thanbroad[12];
Table3. Continued. Table3. Continued.
peduncular spines stronglyreduced orlacking [24]; cylindric [66],cuspidate orlobed[ 68°];infruit,

pistillatecalyx annulate [63];epicarp orange [88°]; calyxaboutequaltocorolla[78].
endocarp pitted [103]. pilosaclade (Subsectional level):
SETULOSA CLADE (Sectional level): Rachillae glabrous to lepidote [35°*];pistillate
Stemusually >6 cmindiam.[I2*];spines slightly calyx urceolate [62];pistillate corolla lepidote
broadened nearmiddle[3],fascicled [10];an- [71],cupulate [66°],truncate [68];infruit, ca-
thers oblong-linear [58°*];pistillate corollapa- lyxaboutequaltocorolla[78].
telliform [67]andfimbriate [70];epicarp orange PYRENOGLYPHIS clade (Subsectional level)= Pyr-
[88°]. enoglyphis sensuBurret:
COROSSILLA clade (Sectional level): Rachillae glabrous to lepidote [35°*], alveolate
Spinesfascicled [10]; peduncular bractdensely [38];pistillate calyxurceolate [62];pistillate
covered withflexible setiform spinules [22];ra- corolla lepidote [71],cupulate orcylindric [66°
or*],truncate [68],internally withbasally ad-
chillaeglabrous orlepidote [35°*];anthers ob-
long-linear [58°*];epicarp yellowish [88°];fruit natestaminodial ring [73];infruit, calyx about
apexrostrate [92]. halfthelength ofcorolla [78°*]; endocarp with
fertile andsterile pores atsameleveldistally
TURBiNOCARPA clade (Sectional level): [100°*]. (Allconstituent complexes sampled.
Anthers oblong-linear [58°*];pistillate calyxdel- Thiscladeexhibits thefollowing divergent
tate-lobed [65];fruit ellipsoid [81°*, 82];epicarp character trends, culminating inthemost apo-
ferruginous-lepidote [87]andpaleyellow-brown morphic taxain Bactris : Inflorescence in-
[88°]; fruit apex rostrate [92];mesocarp con- creasingly reduced inrachis length andnum-
of
sisting juicesacsattached toendocarp fibers ber of rachillae (paniculate - digitate-
[93]; endocarp withfertile and sterilepores spicate) [25, 27-29]; rachillae increasingly rig-
offsetdistally [100°*]andbaseexpanded [104]. id,stout andshort, andwith increasingly close
constANciAE clade (Sectional level): triads either bylossorrefeminization offe-
Anthers oblong-linear [58°*];epicarp rosypurple male-sterile triads[32,33,40°*];increasing
[89];mesocarp consisting ofjuicesacsattached connation ofstaminate petals, andflattening
toendocarp fibers [93],theseveryfine, numer- andadnation offilament bases[47,55,56°*,
ous,andbasally attached tominute tubercles or 57°*]; increasing elongation andattenuation of
papillaeon theendocarp [972];endocarp with fruitandendocarp (obovoid-globose ellip-
fertileandsterile poresat sameleveldistally soid- fusiform) [82,101°*,104].)
[100°*].(Alsoepicarp splitting intoacicular corky
tubercles; character notaddedtoanalysis, which
wasnearly completed whenauthentic material
becameavailable.) AMYL, and pira (exceptfortheinclusionofcns)
ATROPURPUREOUS clade (Sectional are each monophyletic. However,it also shows
level):
Spines usually broadened slightly togreatly inmid- thatBurretsimplydumpedtheremaining, non-
dle[3];anthers oblong-linear [58°*]; epicarp dark divergent taxa into bact (sect. Bactris ), which in
purple-brown orblue-black [88];mesocarp con- realityis a paraphyletic assemblageofthemore
sistingofjuicesacsattached toendocarp fibers generalized species.The monophyly of there-
[93],theseveryfine, numerous, andbasally at- maining groupsis insufficient to supporthistax-
tachedto minute tubercles or papillaeon the onomiescheme.
endocarp [972].(Someconstituent clades,e.g., GuilielmasensuBurret(guil) does notform
B. tomentosa complex, notsampled in study; a clade or fromtheremainder
theseeither maybeequaltooraretobeincluded of Bactrisuntil entitydivergent
inthefollowing subsectional levelclades.) it is groupedwiththeantillean
speciesofbact to formthenon-ocreateclade.
BIDENTATE clade (Subsectional level): That is, segregating guil makesBactriss. str.
Shaftofspinespale[6];leaflet apexwithsub- paraphyletic and guil cannotbe treatedas a ge-
apicallyterminating midrib[15°*],usually
equallybidentate, in nus without all equivalent cladesbeingso treat-
except set; pedunculared. This would Bactris into eightor
spinesstrongly reduced orlacking [24];sta- fragment
minate petalsstrongly connate [47]. morepoorlydefinedgenera.Treatingguil as a
subgenus, as wasdonebyDrude(1887)andMac-
LONG-PROPHYLL clade (Subsectional level): Bride(1960),wouldlikewiseundulyinflatethe
Spinestapering from base [3°*];prophyll over
halfoflength ofpeduncular bract[17],this numberof subgeneraand would obscurethe
densely covered withsetosespinules [22];in strongsupportforthenon-ocreate clade. If,
fruit,calyx about equal to corolla [78]. however, thenon-ocreateclade is treatedas a
TUBIFLOROUS clade (Subsectional subgenus, thenguil maybe treatedas a section
level):
Rachillae15 or fewer [27];pistillate calyxur- (Table 3).
ceolate[62];pistillate corollalepidote [71], The sameholdstrueforPyrenoglyphis sensu
Burret(pyre).However,it is evenmoredeeply
Figure6. DispositionofBiuret's(1933-1934)groupsandillustration in fruit

oftrends in the
characters
contextofthephylogenyofBactris.Summarized monophyletic
groups areindicated Dottedline
bytriangles.
encompasses inbact(sect.Bactris
taxaincluded which
sensuBurret), isparaphyletic.
Drawingsofwhole spherical
arefruits
structures removed
withepicarps withdistalendontop,nottoscale.Partial
insideview,oriented
illustrations
areofenlarged
endocarp andfibers;
surfaces indicates
stippling mesocarp;
starchy pinnatestructures
arejuicesacsattached Outlines
to fibers. areendocarpsin polarviewshowing radialposition
relative of
germination poretoleft.
poreswithfertile Colorsareoftheepicarp. ofcharacters
attainment
Notesequential
tobuildupthe"pyrenoglyphis"-typefruit.
nestedwithinthetuberculate clade thanguil clade; thismakesBurreťssubgenuspolyphy-

is withinthenon-ocreateclade. Indeed,pyre leticand completely untenable as a taxon,amyl
is divergent andapomorphic. However,the"ge- and pira are bothseveralnodesfromthebase
neric"character of thestaminodialring(73) is ofthelatibracteate clade, whichsuggests that
simplya single,clear-cutsynapomorphy thatis theyshouldbe recognized at thesubsectional or
one amongseveralthathave developedat dif- serieslevel(Table 3), ratherthanthesectional
ferentstagesintheevolutionofthetuberculate levelas was donebyBurret.
clade. Indeed,the striking similaritybetween Furthermore, Burretpartitioned sect.Amylo-
certainpyrespp.,e.g.,B. maraja, and closely carpusintotwosubsections (notshowninFigure
relatedspecieslackingthestaminodial ring,e.g., 1). Subsect.Isochlamys(Spruce)Burretsensu
B. montícola ,willbe noticedbyanyonewhotries Burrethas thepistillatecalyxequallingtheco-
toidentify collections containingboth.Thus,rec- rolla. It corresponds directlywiththe Bactris
ognizing pyreat eitherthegenericor subgeneric simplicifrons complex(AppendixI) and is rep-
level (Drude, 1887; MacBride,1960) without resentedon thecladogramby theclade sim +
treatingequivalentclades similarlycontradicts ten (Figure 3). SubsectionIsochlamys , thus,is
thephylogenetic evidence,pyrewouldbestbe monophyletic. However,subsect.Brachycalyx
treatedas a subsection orserieswithinthetuber- Burret,whichis distinguished by annulate,in-
culate clade (Table 3). conspicuous calyces,is leftas thepara-
pistillate
Therearealso problemswitheachofBurreťs phyletic remainder ofamyl. Thesetaxa should
subgroups in Bactriss. str.SubgenusAmylocar- notbe maintainedas equivalentgroupsas was
pussensuBurret comprises amylandpira.These donebyBurret.
twogroupsare monophyletic, but theydo not In subg.BactrissensuBurret, themonotypic
forma moreinclusivemonophyletic group.Each sect.Aiphanoides(i.e., B. caryotifolia ; aiph) is
arisesonseparatebranches ofthelatibracteate deeplynestedwithinthe tuberculate clade
Figure7. Evolution ofcharacter

states
ofleaf-segmentapicesinBactrissuggestedbyparsimony analysis.
Having oblique apiceswiththeprimary
veinterminating (o,char.151)isa synapomorphy
apically forthespecies
ofBactris.Having twostronglyunequalapicalteeth
withtheprimary veinterminating (u,char.15°)
subapically
isa synapomorphy forthenon-ocreateclade.Having twoshort equalteethwith theprimary
veinterminating
subapically(e,char.15°)aroseintheancestorofthebidentate clade (orcladedistaltoset,which often
has
unequal teeth).Thepraemorse apex(p,char.161)inaiphis notfundamentally different
totherestofBactris
,
butis simply a modification
oftheequallybidentapexthrough distalbroadening.
amongspeciesofbact. To recognize repeatedat increasingly

thisspecies higherlevelsof gener-
as the sistergroupof all thetypicalspeciesof ality.Thus,I agreewithNeff(1986)thatall char-
Bactrishidesits close relationship withtheB.actersthatone proposesto observeand report
guineensis complexandthesequentialevolution shouldbe viewedas hypotheses thatare either
of itsdistinctive leafform(Figure 7). Becauseacceptedorrejectedas theanalysesproceedand
bact is paraphyletic and does not reflect rela-
as otherworkers tryto use thosecharacters. In
tionships thiscontext,AppendicesII and III detailthe
as theyexist,sectionBactrisshouldnot
be maintained withBurreťscircumscription. selection,use, and problemsof each character.
The modusoperandi ofsystematics inthetime
The discussionherefocuseson generalizations
of Burret(1933-1934) was overallsimilarity. madeinthestudyand on thetransformations of
Whenviewedfromthisperspective, groupsof characters
histerminal thatare important in the
pheneticgroupsare usuallyconsistent withthe
taxonomy ofBactris.
distributions of characters,attestingto his per-
Mostofthe106 characters weredefined qual-
ceptivity.However,whenevaluatedbycladistic Because33 characters
itatively. lentthemselves
criteria,Burreťsclassification failsand should
to quantification,
theyweredefined to assessthe
notbe used. variationmoreprecisely. Onlysix ofthese(12,
18, 42, 43, 52, and 83: e.g.,length-width ratio
and
Stability Evolution of of peduncular bracts and number of stamens)
TaxonomieCharacters brokeintoclearly discretestates.Theotherquan-
titativelydefinedcharactersyieldedmeasure-
The processofselecting,
Generalevaluation. ob- mentswithslightoverlaporwithoverlapinonly
characters
and analyzing
serving, is cyclic,being one or a fewotu's. This variationalpatternis
1991] SANDERS: CLADISTICS OF BACTRIS 12 1
discussedforninecharacters (1, 11, 27-31, 62, thecladogram, yetfruit shapereversed fivetimes

and 63) in AppendixII, buttheremaining eigh- independently of endocarpshape. Thus, what
teen(17, 25, 32, 44, 46, 47, 49, 59, 60, 78-82, initially appearedtobe redundant characters were
84, 85, 92, and 101)also exhibitthispattern. In onlypartially so.
thisstudy, thesampleofspecimens perotu was Indeed,beingable to pinpointand compare
small,such thatthe hypothesized gaps and/or cladesin whichthecharacters aredependent vs.
modes mustbe verifiedin subsequentmono- onesin whichtheyareindependent maylead to
graphic studiesinwhichmeaningful statisticscan importantinsightsinto the morphogenesis,
be obtained. adaptiveecology, andevolutionoftherespective
Character analysisinBactrisillustrates thatthe taxa. In thecase of fruitand endocarpshapes,
hierarchical levelatwhichoneevaluatesanduses bothcharacters changeconcurrently to theapo-
a character is important. Acrossa largeand spe- morphicstatein theancestorofall Bactrisspp.
cies-richgenus,thereare twomajorsourcesof exceptthebalanophora clade. Boththenre-
homoplasy:1) trueparallelisms and reversals in verseconcurrently to theplesiomorph in species
characters withsmallgenotypic/phenotypic dif- (or cladeswithfewspecies)of whichthesister
ferences betweenstates;and 2) convergences, in speciesretaintheapomorph(1: in cln, butnot
whichthesubtledistinctions betweencharacters lon; 2: in maj [+ othermoredistalpyre],but
are notreadilyperceived.Howeverin a small, notOtt or cru; 3: in gas + dah, butnotmac).
homogeneous group,such as the distalmono- Additionally, in cases 2 and 3, the fruitshape
phyleticgroupsanalyzedby bandb,numerous divergesto state812 in the nextclosestsister
parallelisms/reversals andtheoccurrence ofcon- species(2: inmarandinpil; 3: injam).However,
vergent characters are less likely. Comparisons fruit shapereverses to 8 Iowithout a concomitant
ofhomoplasiesat different levelsdrawone'sat- changein endocarpshapein twospecieswhile
tentionto thecharacters whichshouldbe rein- thesisterspeciesretainsbothapomorphs(4: in
terpreted and/orreanalyzed.For example,the cam,butnotsav; 5: in aub, butnotsim+ ten).
occurrence of bifidleafletapices (char. 15°) on Byknowing thesesister-group relationships,one
two distantinternalclades suggeststhatreex- can make two-waycomparisonsof differences
aminationof thischaracter shouldresultin its withinsistergroupsand amongdependentvs.
redefinition as two distinct,nonhomologous independent classes(i.e., cases 1-3 vs. 4-5) to
characters (Figure 7). answerthe following questions:What are the
A similarproblemencountered in thisstudy commonand dissimilar offruit
features and en-
was thatsomecharacters, suchas thoseassoci- docarpdevelopment? Howdo fruit andendocarp
atedwithpistiland stigmastructure, are inter- ontogenies becomedecoupled? Arethereanyfea-
pretedwithdifficulty fromrehydrated material. turesin theontogenies to suggestthattheindi-
As newcollections permitstudyoffreshorpick- vidual charactersare not homologousamong
led material, a moreaccurateassessment ofthe cases?Whatecologicalparameters aresimilaror
distribution ofthesecharacters amongtaxacan dissimilaramongcases? Can the reproductive
be made. successofeachspeciesin itsrespective environ-
It is generally agreedthatdifferent characters mentbe relatedto itsparticular combination of
thatare developmentally or functionally corre- statesofthesetwocharacters? Becausethereare
lated,and presumably expressthesamegenetic morethanone case perclass,congruent devel-
information, shouldbe treatedas a singlecla- opmentalor ecologicalpatterns wouldstrength-
distic character.The same effectwould be en one's conclusions.
achievedby dividingthe normalweightof a Sequentialevolution. The cladisticanalysisre-
character by thenumber ofcharacters in the cor- veals a striking case of sequentialevolutionof
relatedcomplex.This preventsgivingundue structures thatareassociatedwithderivedmor-
weight to anyonecharacter. However,theprob- phologies.Forexample,ifonethinksofPyreno-
lemis to determine whether twoor morechar- glyphisas a specialized,distinctgenusor sub-
actersare actuallycorrelated, i.e., does one al- genus,thenthereis thetendency also to thinkof
waysoccurwheretheotherdoes?This maybe itsderivedfruit typeas a uniquecharacter com-
hardto see in a largedata set,suchas thisone. plexin whichtheconstituent characters evolved
It is probably bettertorunan initialanalysisand together in a correlated wayonlyin thisgroup.
followthetwocharacters, whichis easilydone However,thisis clearlynotthecase (Figures3,
in PAUPwiththechglist option.For example, 6). The characters ofthegroupevolvedstepwise
fruit shape(81) and endocarp shape (101) would in a seriesof ancestors leadingup to pyre.That
seemto be underthesamedevelopmental con- is,increasingly closersistergroupsretainincreas-
straints to forma highlycorrelated set of char- ingnumbersof the "pyrenoglyphis"-like char-
acters.Theychangedtogether at fournodes of acters.Thus,theseriesofancestorsofpyreare
notexpectedto showall thebasic pyrecharac- rhombic;stamenssix,emerged aroundthemar-

ters,justas thefossil(andperhapsancestral) an- ginofthereceptable andpossiblyslightly adnate
giosperms thatpaleobotanists hopeto findneed to thepetals;and filaments flexuously foldedin
notpossessall ofthespecializations ofmodern bud, thread-like, holdingthe ovate-oblong an-
angiosperms (Doyle & Donoghue,1986). therswithin thecorolladuring andafter anthesis.
Charactertrendsand reliablecharacters. Many A fewcharacterchangesare unique (e.g., the
vegetative characters are so highlyplasticor in- non-ocreateclade withroundedcuculiatepet-
consistent as to be unsuitableforcladisticanal- als; the EXSERTED-ANTHER clade withanthers
ysis(Appendix3). However,certainones were exserted betweenthepetals;and cnc withcom-
completely ornearlyconsistent. In particular,the pletelyconnate,membranous petalsand 12 sta-
loss of the ocreaein the adultleaves,strongly mens).Otherchanges(e.g.,flowerlength, calyx
unequallybidentateleaflets,and three-rankedlobe shape,stamenpositionand adnation,fila-
petiolarspinesdefinesthenon-ocreate clade; mentstructure, andanthershape)aremoderately
the slightlybroadened spines pervade the homoplasious butappearto facilitate thedelim-
ATROPURPUREOUS clade; and theequallybiden- itingofsomemajorandmanyminorclades.The
tateleaflets definethebidentateclade (Figures remaining characters (calyxlobe posture,mod-
3, 7, Table 3). Othercharacters, e.g.,fascicled eratepetalconnation, and receptacle shapeand
spines,extremely broadened andflattened spines, adnation)are cladistically unreliable,but they
and pale-coloredspines,are usefulin restricted mayassistin defining speciesor minorcladesin
groupsbutarehomoplasiousoverthewholege- future studies.
nus. Of specialconcernis Burret'scharacter of Likewise, floralcharacters
pistillate (otherthan
massivestems(l2). Eventhoughoutgroup com- thestaminodial ring,whichcan be seenin fruit)
parisonfailsto polarizeit, parsimony analysis playedlittlepartin Burret's scheme.The results
suggests thatit is plesiomorphic relativeto both heresuggestthatthe ancestralpistillateflower
moderately thickand arundinaceous stemsand had a cupulate,cuspidatecalyx,and a shortly
thattheselattertwoevolvedinparallelinBactris urceolate,cuspidatecorollathatinternally was
and Desmoncus. beset with6 triangular basallyadnate stami-
Amonggeneral inflorescencefeatures,themost nodes.Most of thechangesobservedare unre-
important sourcesof characters are size differ- liableforcladisticanalysisexceptat thelowest
encesand relativelengthsand numbersof ra- levels.As notedabove,thismaybe due in part
chillae.Armature and indûment oftheinflores- to difficultyofinterpreting thecharacters drawn
cenceare oftentoo variableto be used. Burret fromrehydrated material.Observations ofpick-
distinguished subg.Amylocarpus bytherachillae led immature flowerssuggest thatsomecharac-
withcontiguous triads;amylbyrachillae1 to 8, ters,suchas theconnation ofstylebranches, may
a fewcmlong,andwithtriadsnearlytothedistal be underrather laxdevelopmental controls. Cer-
end;and pira byrachillae20 to 100,ofnormal tain of the characters (calyxlobing,calyxand
length, andwithtriadsonlyintheproximalhalf. corolla truncation, corolla indûment),though
Theseinflorescence characters do helptodelimit homoplasious, mayhelptodefinemajorandmi-
amyl and pira as monophyletic groupsbutare norclades.Curiously, thepresenceoftriangular
onlypartially reliablein cladisticterms.Thatis, staminodes is extremely plasticin Bactris,vary-
thetrendtowardspicateinflorescences and con- ingintraspecifically, as wellas amongflowers of
tiguoustriadsis accomplishedin variousways a singleplant.Thisextreme plasticity was prob-
in severallineages,occurringin at least the ablythecase in ancestralspecies.However,the
NONFIBROUS, EXSERTED-ANTHER, TUBIFLOROUS, mutationforconnate,truncate staminodes(the
and pyre CLADES.The B. tomentosacomplex staminodialring)becamefixedin theancestor
(AppendixI), ifdistinctfromthetubiflorous ofpyre,theplasticity waslost,andthecharacter
clade, represents anotherparalleloccurrence of becamecompletely stable.
thetrend.Thus,Burret weighted thesetooheavi- Fruitsprovidethemostreliablecharacters at
lyandreliedonthemtoformmajorgroupswhile the genuslevel and are important in defining
excluding othermorereliablecharacters. majorcladesand majorevolutionary trendsin
The staminateflowers are extremely ephem- thegenus(Figure 6); whyBurretdid notprofit
eralinmostspeciesofBactris.Becausetheywere frommoreofthesecharacters is unknown. The
notavailableonmanyofthespecimens thatBur- resultsshowthattheancestraltypecan be char-
retstudied,he did notuse themin his classifi- acterizedas: elongatewitha red,orange,or yel-
cation.However, parsimony analysissuggests that lowishepicarp;mesocarpthinly starchy; anden-
theyare important and thatthe ancestralsta- docarp pale, coveredwithblackish,flattened
minateflower canbe characterized as follows:as adnatefibersthatradiateand anastomose, with
longas or longerthanthepistillate; calyxlobes threegermination poreslatitudinally equidistant
triangular; petals shortlyconnate,erect,and butlongitudinally offset (sterilepairproximalto
fertileone).In thebalanophora clade, theen- theepicarpis a lightgreen,thenfleshcolored,

docarpbecamedilatedproximally. In theances- thengradually maturesto a richrosypurple.Be-
torto theremainder oftheBactris,theendocarp causeBurretincludedcns in pira (AppendixII,
becameoblate(fruit turbinate)withtheporesnot char.40), its colorationwas coded as a unique
offset.In theancestor ofthecladelatibracteate character distinctfromthatfoundin the atro-
+ TUBERCULATE, the endocarpbecame black- purpureous taxa.Rathersurprisingly, theresults
enedand thefibers becamecompletely freeand showthatatropurpureous epicarpsevolvedonly
terete.Somewhat inparallel,theantillean species once.Moreover,trc and cns aresuccessivesis-
withinthe non-ocreate clade developed a tergroupsto theatropurpureousclade; i.e.,
darkenedendocarpand partially free/terete en- theyare intermediate and forma bridgeto the
docarpfibers. Itis notclearwhether thetubercles orange-colored ancestors.Thus, the stramin-
evolvedde novoin theancestorofthetubercu- eous,rose-purple, andpurple-black epicarpsmay
LATECLADE withtheLATIBRACTEATE CLADE hav- forma singletransition seriesthatclarifies the
ingretained thepreviouscondition, as suggested evolutionary developmentof the purple-black
bytheresults.Ifchar.97 (presenceand formof epicarps(Figure 6).
thetubercles) had beencodedas a modification Becausemostfruitcharacters arecladistically
of 961 (fibersproximally flattened and adnate, reliable,futurestudiesand classifications must
distallytereteand free)ratherthana modifica- relyon them.Especiallyusefulwouldbe devel-
tionof 962 (fiberscompletely tereteand free), opmentalanalysesto betterdetermine theho-
thenthetubercles wouldhave arisenin thean- mologiesof the tuberclesat the bases of the
cestorof both clades and subsequently would endocarpfibersand oftheallometric growth of
havebeenlostintheancestorofthelatibracte- theendocarpand epicarp.
ate clade.
A further changewithinthe latibracteate Conclusions
clade is thecompletelossoftheendocarpfibers
fromthenonfibrousclade. Perhapsdevelop- Becausetheabove characters are likelyto be
mentally correlatedwith thesmall stature ofamyl in
significant future classifications,collectors
is thereduction in fruitsize,a character usedby shouldtakespecialcareto gathermaterialthat
Burret. showsthesecharacters. In additionto following
In theancestor ofthetuberculate clade, the the standardtechniquesfor collectingpalms
two sterilegermination poresbecamelongitu- (Dransfield, 1986),theyshouldattempt toobtain
dinallydistantfromthefertile one. This is very thefollowing: 1) immature, unopenedinflores-
pronounced insomelineages, e.g.,thebidentate cences(as wellas wholeopenedones)thatmay
clade. In theancestorofthecladeatropurpur- be takenbyshavingthesubtending sheathaway
eous + CNS+ trč, themesocarpchangedfrom oppositethestemfromtheinflorescence, pulling
starchy to theformof cellularjuice-containingawaythewholeleafand leafbase,and wedging
sacs attachedto theendocarpfibers.In thean- undertheinflorescence untilit snapsoffat the
cestorofthecladeatropurpureous+ cns,the pointofattachment; 2) maturefruits, nearlyma-
endocarp fibersbecame very fine and numerous, ture or
fruits, (ifthese are not available) old en-
andthetubercles becamereducedto papillae.In docarpsfromthe ground;3) the leaf spearto
theancestorof pyre,the endocarpporesonce showtheyoungocreaoratleasta petioleshowing
again became latitudinally offset.Withinthis theocreascar;and 4) a sectionof stemlacking
clade,therehasbeena markedapomorphic trend sheaths.
towardelongation and attenuation oftheendo- Notall characters werereliablein thisprelim-
carpand wholefruit. inarycharacter evaluationandcladisticanalysis,
The cladisticresultssuggesta historyof epi- and thesehomoplasiouscharacters mayneedto
carpcolorthatwasnotpreviously realized.Most be eliminated fromfuture analysesofthewhole
specieshave eithera darkpurplishblack fruit genus.However,because homoplasyin these
(presumably anthocyanin-based) or one thatis characters is less pronouncedin smaller,more
orange-redto yellow (presumablycarotene- homogeneous clades(Figure2), theyshouldnot
based).Ofthetwo,thelatteris clearlyplesiomor- be discardedwithoutmoredetailednumerical
phous.trč is uniquein havinga stramineous- anddevelopmental analysesatlowerhierarchical
yellow epicarp covered with ferruginous levelswheretheymayshowabsolutegapsormay
trichomes. Attheoutsetofthisanalysis, thiscol- be reinterpreted as distinctcharacters.
orationwas consideredan odd modification of Given thatBurret' s classification shouldbe
theplesiomorph in whichitwasincluded,cns is abandoned,whatshouldreplaceit?For several
also uniquein havingan epicarpthatearlyin reasons,I believeit is premature to providea
development becomesfissuredinto numerous newformaltaxonomy basedon thepresentcla-
aciculatecorkyprojections. Soon afteranthesis, disticanalysis.First,thecladisticsampling is still
incomplete, and thepositionsof severalclades BarbosaRodrigues,J. 1903. Sertum palmaram

at the sectional,subsectional, and lowerlevels Brasiliensium. 2 vols.VeuveMonnom, Brussels.
remainunresolved. Second,thelimitsofconstit- Burret,M. 1933-1934.Bactris undverwandte Pal-
uentspecieswill remainuncertainuntilintra- mengattungen. Repert. Spec.Nov.RegniVeg.34:
specific variation is evaluated, such that 167-253.
discontinuitiescan be thebasisofspecieslimits Clement, C. R. 1988.Domestication ofthepejibaye
palm (Bactris gasipaes): past and present. Ad-
and synonymy. to
Third, lendsupport to clades vancesEcon.Bot.6: 155-174.
basedon homoplasious synapomorphies, explo- and J.E. Mora-Urpí.1987. Pejibaye palm
rationsfornovelcharacters and developmental (Bactrisgasipaes, Arecaceae): multi-usepotential
studiesofknowncharacters areneeded.For ex- forthelowland humid tropics. Econ.Bot.41: 302-
ample,Tomlinson(1961) noteda numberofin- 311.
terestingvegetativeanatomicalcharactersin Croat,T. B. 1978. FloraofBarroColorado Island.
Bactris,suchas thedistribution offiberbundles Stanford University Press, Stanford. 943pp.
intheleaflamina,thatdeserveattention infuture Dahlgren,B. E. 1959. IndexofAmerican palms:
or cladisticwork.Fourth,current plates.FieldMus.Nat.Hist.,Bot.Ser.14:pl. 1-
monographic 412.
or agronomicarticlesrarely De Nevers,
ecological,floristic, G. C. 1988.Bactris diviscupulaandBac-
need to referto an infrageneric classification. trisfuscospina reexamined. Ann.Missouri Bot.
Thus,a newinternal classification
shouldbebased Gard.75: 1151-1152.
on future monographic and cladisticstudies. Doyle,J.A.andM.J.Donoghue.1986.Seedplant
However,severalofthecladesin Table 3 do phylogeny andtheorigin ofangiosperms: an ex-
have availablenames at the hierarchical level perimental cladisticapproach. Bot.Rev.(Lancas-
thatI havesuggested. Ifone is compelledto use ter)52:321-431.
infragenericnames,theearliestoftheseavailable Dransfield, J. 1986. A guideto collecting palms.
namesmaybe used forthesecladesin compli- Ann.Missouri Bot.Gard.73: 166-176.
O. 1882. Palmae.Pp. 253-583inC. F. P.
ance withthe International Code of Botanical Drude, VonMartius,ed.,FloraBrasiliensis 3(2).
Nomenclature(Greuter,1988): non-ocreate . 1887. Palmae.Pp. 1-93inA. Englerand
clade = Subg. Guilielma(Mart.)Drude (type: K. Prantl,eds.,Die Natürlichen Pflanzenfami-
B. speciosa(Mart.)Karst.= B. gasipaes);lati- lien2(3).
bracteate clade = Subg. Trichobcictris Oerst. Farris,J.S. 1969.A successive approximations ap-
(type:B. glandulosa);nonfibrousclade = Sect. proach tocharacter weighting. Syst.Zool.18:374-
Isochlamys Spruce(typeheredesignated: B. ne- 385.
grensis Spruce = B. simplicifrons); amyl = Sub- . 1970.Methods forcomputing Wagner trees.
sect. Isochlamys (Spruce) Burret; exsert- Syst.Zool.19:83-92.
G. and R. Bernal. 1987. Palmasdelde-
ed-antherclade = Sect.Macrophyllum Drude Galeano,
partamento de Antioquia. Universidad Nacional
(typeheredesignated: B. acanthocarpa); tuber- deColombia, Bogota. 221 pp.
culate clade = Subg.Bactris(type:B. minor Greuter,W.,ed. 1988. International CodeofBo-
Jacq.= B. guineensis); atropurpureousclade tanicalNomenclature (Reg.Veg.118).KoeltzSci-
= Sect.Bactris. entificBooks,Königstein. 328pp.
Holmgren, P. K.,W.Keuken, andE. K. Schofield.
Acknowledgments 1981.IndexHerbariorum: partI, theherbaria of
theworld, ed.7 (Reg.Veg.106).Bohn,Scheltema
I thankN. W. Uhi,J.Dransfield, andJ.Mora- & Holkema, Utrecht. 452pp.
Urpiforencouraging me in thisproject,and A. Karsten,H. 1856. Plantaecolumbianae. Linnaea
28:241-282,387-413.
J.Hendersonand L. R. Noblickforhelpfuldis- . 1857("1856")-UeberdieBewurzelung der
cussions.The curatorsof BH, FTG, MO, NY, Palmen. Linnaea28:601-608.
and US are thankedfortheloan of specimens. Kluge,A. G. and J.S. Farris. 1969. Quantitative
T. A. Zanoni,J.-J.de Granville,R. Bernal,G. phyleticsandtheevolution ofanurans. Syst.Zool.
Galeano,C. R. Clement, V. Patiño,andthestaff 18:1-32.
oftheUniversity oftheWestIndies,St. Augus- MacBride, J.F. 1960.FloraofPeru.FieldMus.Nat.
tine,expeditedmycollecting in thefieldand/or Hist.,Bot.Ser.13:321-418.
obtainingresearchmaterial.J. B. Fisher,P. F. Martius,C. F. P. von. 1823-1850.Historia natura-
Stevens,N, W. Uhi,andT. W. Waltersreviewed lispalmaram. 3 vols.Munich.
draftsofthemanuscript. Mora-Urpí, J.E. andC. R. Clement.1981.Aspec-
tostaxonómicos relativosal pejibaye(Bactrisgas-
ipaesH.B.K.).Revista Biol.Trop.29: 139-142.
Literature Cited Neff,N. A. 1986.A rational basisfora priori char-
acterweighting. Syst.Zool.35: 110-123.
Bailey,L. H. 1947. Indigenous palmsofTrinidad Pimentel, R. A. and R. Riggins.1987. Thenature
andTobago.Gentes Herb.7: 352-445. ofcladisticdata.Cladistics 3: 201-209.
Sattler,R. 1988.Homeosis inplants.Amer. J.Bot. BALANOPHORA CLADE

75: 1606-1617. 1. balanophora complex: B. balanophora.
Spruce, R. 1871.PalmaeAmazonicae. J.Linn.Soc., NON-OCREATE CLADE
Bot.11:65-183.
Swofford, D. L. 1985. Paup:phylogenetic analysis ANTILLEAN CLADE
using parsimony, version2.4.Illinois
Natural His- 2. plumeriana complex: B. cubensis Burret, B.
tory Survey, Urbana. jamaicana, B. plumeriana.
Tomlinson, P. B. 1961. Palmae.Vol. II in C. R. CLADE
Metcalfe,ed.,Anatomy ofthemonocotyledons. GUILIELMA
Clarendon Press,Oxford. 453pp. 3. gasipaes complex: B. caribea Karst., B. ciliata
Trail, J. W. H. 1877. Descriptions ofnew species (R.&P.) Mart., B. dahlgreniana ( =Guilielma
ofpalmscollected inthevalleyoftheAmazon in microcarpa Huber, nonB. microcarpa Spruce),
north Brazilin 1874.J.Bot.15:75-81. B. gasipaes, B. insignis (Mart.)Baili.,B. ma-
Uhl,N. W.and J.Dransfield.1987. Generapal- cana,B. speciosa(Mart.)Karst.,B. utilis
marum: ofpalmsbasedonthework
a classification (Oerst.) Hemsl.
ofHaroldE. Moore,Jr.L. H. BaileyHortoriumLATIBRACTEATE CLADE
andInternational PalmSociety, Lawrence, Kan-
sas.610pp. NONFIBROUS CLADE
WesselsBoer,J.G. 1965.Theindigenous palms of 4. ptariana complex: B. ptariana Steyerm.- no
Suriname. Leiden.172pp.
E. J.Brill, Otu,nearbaseofclade(apomorphies 26,45,
63,963,99;distinct inhaving 15°*and10).
5. coloradonis complex: B. coloradonis, B. dia-
AppendixI. Representativenessof nueura Burret, B. porschiana.
Taxa Sampled 6. hondurensis complex: B.hondurensis, B.pau-
la L. H. Bailey, B.pubescens Burret, B. stand-
Belowarelistedallthespeciesreferable toBac- leyanaBurret, B. wendlandiana Burret.
triss.l. thatare knownto me. Specimensofthe 7. chaetospatha complex: B. armata Barb.Rod.,
data sample(Table 2) wereidentified as ele- B. chaetospatha Mart.- no otu, apparently
on cladehon + amyl;verysimilar to hon,
mentsofthespeciesinboldface(otu's, Table 1); butpistillate calyx urceolate (62) andlonger
thespeciesin italicshad no specimensidentified thancorolla.
withthem,and hence,werenot sampled.To
confirm thattheunsampledspeciesare notthe AMYLOCARPUS CLADE
solemembers ofanysubgeneric orsectionallevel 8. fioccosa complex: B.fioccosa Spruce- nootu,
cladesthatwouldhave been discoveredby the nearbaseofamyl(plesiom. 36°,68°,7Io,72°;
cladisticanalysis,the unsampledspecieswere apom.19,21,33,40°*,42,63;leafsegments
associatedwiththeotu's intwoways.1)Allspe- several, sigmoid).
9. aubletiana complex: B. aubletiana, B. kuhl-
cieswerephenetically grouped intoinformal spe- mannii Burret.
cies complexesbased on a perusaloftheproto- 10. simplicifrons complex:B. acanthocnemis
logues,published platesoftypes(Dahlgren, 1959) Mart.,B. amoenaBurret, B. arenaria, B.
and commentsby Burret(1933-1934),Wessels brevifolia Spruce, B. carolinensis Spruce, B.
Boer(1965), Croat(1978), Galeano and Bernal dakamana (L. H. Bailey)Glassman, B. esse-
(1987), and De Nevers(1988). 2) The proto- quiboensis (L. H. Bailey)Glassman, B. glea-
loguesof theunsampledspeciesweresearched sonii(L. H. Bailey)Glassman, B. huberiana
forapomorphies, whichwerecomparedto the Burret (=Amylocarpus angustifolius Huber),
of the otu's to check the B. inermis Trail, B. killipiiBurret, B. leutzel-
apomorphies phenetic
associations. Thereis a growing consensus among burgii, B. maguirei (L. H. Bailey)Steyerm.,
B. microspatha Barb.Rod.,B.paucisecta Bur-
recentworkers thatsomeofthesecomplexes are B.
ret,B. schultesii (L. H. Bailey)Glassman,
reallyonlypolymorphic speciesthatcomprise simplex Burret, B.simplicifrons, B.soropanae
numeroussynonyms. Steyerm., B. stahelii(L. H. Bailey) Glassman,
On theotherhand,thisprovisional assignment B. tenuis, B. tenuissimus Barb.Rod.,B. trin-
ofspeciesdoesnotimplymonophyly ofthecom- itensis, B. uaupensis (Trail)Spruce, B. ulei
plexes. Rather, the complexes were associated Burret, B. xanthocarpa Barb. Rod.
withthe clades (presentedhierarchically as in 11. hirtacomplex: B. atroxBurret, B. ericetina
Barb.Rod.,B. geonomoides, B. hirta,B. hop-
Table 3) by the sharedoccurrenceof otu's. B. huebneri B. integri/olia
pii Burret, Burret,
Complexeslackingotu's wereassociatedwith Wallace, B. mollis Dammer, B.pulchra Drude.
knowncladesbyappropriate apomorphies drawn 12. pedinata complex: B.formosa Barb.Rod.,B.
fromtheliterature. These apomorphiesare in- hylophila Spruce, B. lakoiBurret, B. lepto-
dicatedbelow,as are additionalcharacters that chaetaBurret, B. multiramosa Burret, B. mi-
help distinguishthe unsampled complexes. crocarpa Spruce, B. pectinata, B. setipinnata
Character notationsare as in Table 3. Barb.Rod.,B. syagroides Barb.Rod.& Trail,
B. turbinata,B. unaensis Barb.Rod.;perhaps ATROPURPUREOUS CLADE

included inthehirta complex. BIDENTATE CLADE
13. mitiscomplex: B. cuspidata Mart.,B. mitis 27. setosacomplex: B. cuyabaensis Barb.Rod.,
Mart.-nootu,nearcladegeo + hir+ pec
+ tur (apom.19,33,35°*,40°*,42,63;leaf B.escragnollei Burret, B.lindmaniana Lindm.,
segments few,wide,sigmoid). B. setosa.
28. chloracantha complex:B. chloracantha
EXSERTED-ANTHER CLADE Mart.-no otu,nearset (plesiom. 62°,63°;
14. mexicana complex: B. acuminata Liebm., B. apom.6, 15°*,24,88;butlacks812).
aureodrupa L. H. Bailey, B. dasychaeta Bur- 29. piritu complex: B. piritu (Karst.)H. Wendi.
ret,B.graciliorBurret, B.mexicana, B.tricho- (=Guilielma Karst.)-nootu,nearset
piritu
phyllaBurret. (apom.15°*,812,88,93,972).
15. oligoclada complex: B. oligoclada Burret- no 30. guineensis complex: B. anisitsii Barb.Rod.,
otu,oncladepira+ mex(apom.60,61). B. bidentula Spruce, B. glaucescens Drude, B.
PIRANGA CLADE guineensis (= B.minor Jacq.),B.hórrida Oerst.,
B. nigrispina Barb.Rod.,B. oraría, B. palus-
16. acanthocarpa complex: B. acanthocarpa, B. trisBarb.Rod.,B.polyclada Burret, B. rotun-
acanthocarpoides, B. aculeiferaDrude, B.bar- da Stokes, B. tucum Burret.
ronis,B. bradeiBurret, B. devia H. E. Moore, 31. caryotifolia complex: B. caryotifolia.
B. exscapa(Barb.Rod.)Barb.Rod.,B.fragae
Lindm., B.humilis, B.interrupte-pinnataBarb. LONG-PROPHYLL CLADE
Rod.,B. macrocalyx Burret, B. microcalyx
32. fusca B.fuscaOerst., B. longiseta.
Burret, B. mindellii Barb.Rod.,B. pinnati- complex:
sectaBurret, B. rhaphidacantha, B. taruma- 33. coloniata complex: B. coloniata.
nensisBarb.Rod. 34. vulgaris complex: B. glazioviana Drude,B.
17. cuescocomplex: B. cuescoEngel, B. trailiana vulgarisBarb. Rod.- no otu's, near
Barb.Rod.-no otu; separated fromacan- LONG-PROPHYLL CLADE (plesiom. 2°,30or ' 6°,
thocarpa complex bysimple leavesonly.Ac- 62°;apom.63,81,88,93,972;distinguished
cording toA. J.Henderson (pers.comm.), in by25,27).
theacanthocarpa complex leavesvaryintra- 35. longifrons complex: B. littoralis
Barb.Rod.,
from tosimple. B. longifrons Mart.,B. maraya-acu Barb.
populationally pinnate Rod.-nootu's,nearlong-prophyll clade
18. glandulosa complex: B. alleniana, B. glan-
dulosa(=B. bifidaOerst.),B. macrotricha (asinvulgaris complex butwith27°and812).
B. oerstediana Trail. 36. granatensis complex: B. granatensis (Karst.)
Burret, H. Wendi. (=Guilielma granatensis Karst.)-
CAMPESTRIS CLADE nootu,nearlong-prophyll clade(plesiom.
19. campestris complex: B. campestris, B. lan- 3°or', 62°,63°,73°[as 6 distinct acuminate
ceolataBurret, B. leptocarpa ImThurn, B. sa- staminodial apom.6, 88,93).
teeth];
vannarum.
CAUDATA CLADE tubiflorous clade
20. caudatacomplex: B. caudata. 37. montícola complex: B. actinoneura, B. chae-
tochlamys Burret,B. diviscupula, B. elatior
TUBERCULATE CLADE Wallace?, B. erostrata Burret,B.fuscospina L.
MILITARIS CLADE H. Bailey, B. granariuscarpa Barb.Rod.,B.
21. militariscomplex: B. militaris. gymnospatha Burret,B. kamarupa Steyerm.,
B.leptospadix Burret, B. leptotrichaBurret,B.
SETULOSA CLADE longicuspis Burret, B. longisecta Burret, B.
22. setulosacomplex: B. circularisL. H. Bailey, macrocarpa Wallace?, B. microspadix Burret,
B. cuesa,B. cuvaroKarst.,B. setulosa, B. B. montícola, B.paucijuga Barb.Rod.,B.pe-
sworderiana Becc. nicillataBarb.Rod.,B. platyacantha Burret,
COROSSILLA CLADE B.sanctae-paulae Engel?, B.sigmoidea, B.syl-
vaticaBarb.Rod.,B. trichospatha Trail,B.
21. corossillacomplex: B. corossilla,
B. duplex. umbraticola Barb.Rod.,B. umbrosa Barb.
24. ripariacomplex: B. coccínea Barb.Rod.(= Rod.,B. vexans Burret.
Guilielma mattogrossensis Barb.Rod.,nonB. 38. elegans complex: B. elegans, B. elegantissima
Barb. B.
mattogrossensis Rod), ferruginea Burret.
Burret,B. inundata Mart.,B. riparia Mart.- 39. tomentosa complex: B. capillacea Drude,B.
nootu,nearcorossillaorsetulosaclade pickeliiBurret,B. tomentosa Mart.-nootu's,
(apom.97;butepicarps orange because lacks nearlong-prophyll ortubiflorous clade
88). (pleisom. 62°;apom.3,6,27,71,88,93,972;
TURBINOCARPA CLADE distinquished by28,29,32,33,433,leaf-seg-
25. turbinocarpa complex: B. turbinocarpa. ments in3 to 12sigmoid pairs).
40. eumorpha complex: B. arundinacea (Trail)
CONSTANCIAE CLADE Burret, B. bellaBurret, B. bijugata Burret,B.
26. constanciae complex: B. constanciae. capinensis Huber,B. chlorocarpa Burret, B.
eumorpha Trail,B. incommoda Trail,B.ju- NAMES KNOWN FROM INADEQUATE MATERIAL (unlikely
ruensis Trail,B. pirangaTrail,B. pulchella torepresent a major, unrecognized clade):
Burret- nootu's,possibly included intheto- H. E. Moore,B.
B. baculifera Karw.,B. baileyana
mentosa complex butdiffers inhaving only3 bergantina Steyerm., B. duidaeSteyerm., B.falcata
to4 pairsofleafsegments, 60or 29°or l,pos- J.R. Johnston, B.faucium Mart.,B. gracilisBarb.
sibly812. Rod., B. kalbreyeri Burret, B.platyspina(Barb.Rod.)
41. sphaerocarpa complex: B. angusti/olia Dam- B. venezuelensis
B. longipes described as Burret, Steyerm.
mer, Poeppig? (fruits
darkpurple, butBurret [1933-1934] placed it
nearB. hirta), B.sphaerocarpa Trail-nootu, AppendixII. Characters Used
possibly included inthetomentosa complex
butdiffers in havingleavesbifidor with2 Unlessa character is indicatedas beingunor-
pairsofsegments, 6°,71°,andpossibly 812. it is with 0 as thepresumedan-
42. megistocarpa B. megistocarpa Bur- dered, ordered,
complex: cestral state. As advocated byPimentel and Rig-
ret-no otu, neartubiflorousclade (ple-
siom.74°,8Io;apom.27,62,88;leafsegments gins(1987), multistate characters, withinwhich
numerous, linear). thetransition ofthestatesis evident,arecoded
43. fissifrons complex: B.fissifronsMart. - nootu, in twoways.A linearseries,in whichone is the
neartubiflorous clade (plesiom. 71°,74°; apomorphic precursor statetotheonefollowing,
apom.27-29,62,66; leafsegments few,sig- is codedas a simplemultistate orderedcharacter
moid). withthestatesindicatedby sequentialintegers.
CLADE A branched(i.e.,partially orderedcharacter) se-
PILOSA
ries, in whichtwostatesor twolinearseriesof
44. pilosacomplex: B. hirsuta Burret, B. pilosa, stateshad is codedas a com-
B. setiflora Burret. independent origins,
45. macroacantha B. acanthospatha plex oftwoormorebinaryororderedmultistate
complex:
Drude,B. conjluens Lind.& H. Wendi., B. variables.Characters35 and 36, in whichthe
macroacantha Mart.-nootu's,nearpilosa transition ofthestatesis unclear,werecodedas
ortubiflorous clade (plesiom. 30or'; apom. a complexofadditivebinaryvariables(following
6, 27,62,69,71,81,88;rachillae hirsute). Doyle & Donoghue,1986). That is, it was un-
knownwhether state00 or 10 was theprecursor
PYRENOGLYPHIS CLADE stateof a thirdstate,whichwas coded as XI.
46. maraja complex: B. brongnartii Mart., B. bur- Weights, as determined fromC valuesin an ini-
retiiGlassman ( =Pyrenoglyphis microcarpatialanalysis,aregivenin squarebrackets. C val-
Burret nonB. microcarpa Spruce), B. leucan- ues obtainedfromthe
thaLindm.& H. Wendl., B. maraja, B. pal- composite cladogram
lidispina Mart.,B. piscatorum Drude,B. ri- (Figure 3) aregivenin curvedbrackets.
vularisBarb.Rod.,B. strictacantha Burret, B. Stemdiameter
tenera (Karst.) H. Wendl. (=Guilielma tenera < 13 mm;1 = 15-60mm;2 = usually
1.0 = usually
Karst.). >60 mm: unordered. [5]{0.33}
47. cruegeriana B.
complex: cruegeriana, B.
ImThurn. Thestates arenotabsolutely discrete,thecharacter
megalocarpa andsystematic valuemaybecom-
48. ottostapfeana complex: B. ottostapfeana. exhibits homoplasy,
49. majorcomplex: B. albonotata L. H. Bailey, promised. However, thenondiscreteness maybevari-
B. augustinea, B. balanoidea, B. beata L. H. ation induced by environmental stress.
B. broadwayi L. H. Bailey, B. cateri L. Thecharacter couldnotbepolarized because Astro-
Bailey, possesses state2 andDesmoncus is variable,
H. Bailey, B. chaetorhachis Mart., B. chapa- caryum
densis Barb.Rod.,B. demerarana L. H. Bai- having bothstates 0 and1.
ley,B. ellipsoidalis B.
L. H. Bailey, infestaArchitecture and color of general armature
Mart., B. major, B. mattogrossensis Barb.Rod. (spinesofstem, leafbases,petioles, leafrachises,and
(nonGuilielma mattogrossensis Barb.Rod.), peduncular bracts; excluding weakprickles orseta-like
B. minaxMiq.,B. obovoidea L. H. Bailey, B. spinules, especially onsofter tissues)
ovataOerst.,B. planifolia L. H. Bailey,B. = orob-
socialisMart.,B. superior, B. swabeyi L. H. 2. 0 cross-section terete,elliptic,
triangular,
long; 1= linear (i.e.,spine flattened).
strongly [3]{0.33}
Bailey. appearto be discrete. Somespeciesof
50. bicuspidata complex: B. bicuspidata Spruce, The states have state 1 butthisis considereda par-
B. curuena Drude,B. exaltata Barb.Rod., B. Astrocaryum ,
B. pyrenoglyphoides A. D. Hawkes allel derivation within the outgroup.
gaviona,
(=Pyrenoglyphis hoppiiBurret), B. nemorosa3. 0 = in longitudinal shape,tapering frombaseto
Barb. Rod. apex;1 = slightly broadened nearmiddle; 2 = strongly
51. concinna complex: B. concinna. broadened near middle. [2]{0.22}
52. bifida complex: B. bifida. Generally, strongly broadened spines arealsostrong-
53. oligocarpa complex: B.aristata Mart.?, B.gas- lyflattened, andundoubtedly thecorrelation is par-
toniana, B. oligocarpa. tiallyconstrained developmentally. However, thereis
an incomplete correlation, as tapering spinesmaybe partial ocreae, andnumerous adultleaveslacking oc-

either nearly terete orstrongly flattened. Thus,char- reae.Becauseocreaearelacking from bothadultand
acters 2 and3 weremaintained tostudy thedegree to juvenileleavesinAstrocaryum, as wellas Aiphanes,
which theymight beuncoupled. Acrocomia, andGastrococos, absence intheadultleaf
State2 isdistinctive; thespines aremany times wider only(state2) is interpreted as a derivation ofstate1.
inthemiddle thanatthebase.State1 is easilyover- Furthermore, herbarium specimens canbe evaluated
lookedonspecimens becausethedistinction between forstate2 becausethesheath degrades toan auricle-
itandstate 0 issubtle (middle ofshaft onlyabouttwice likeflaponlyon eachmargin (nota scaracrossthe
thewidth ofitsbase). adaxialface)ofthepetiole oftheadultleaves.
4. 0 = pulvinus stramineous to white;1 = blackto 8. 0 = petiole firm orwoody; 1 = ± hollow, crushable,
grayordarktomedium brown. [1]{0.17} andusually collapsing inward upondrying. [3]{0.25}
Spinesareradially spread oreventually reflexed by 9-10. 00 = spinesdistributed ± evenly overtheab-
theswelling ofa pulvinus, a moundoftissuein the axialsurface oftheleafbase,petiole, oralsoleafrachis;
axilofeach.Hence,on immature growth thespines 10= arranged intwomarginal andonemedialrank,
arenotyetspreadandthepulviniarehidden. The whichbecomeespecially discrete on thepetiole and
normal condition istohavepalepulvini forthemature rachis; 01 = arranged intransverse linesorfascicles.
lifeoftheorgan. Onaging organs of a few specimens,[10,3] {1.00,0.25}
thepulvini mayeventually darken, although somelight- Among collections ofgas,somecultivated
colored onespersist. pheno-
Bailey(1947)mayhavefailed to types have91modified either byhaving onlytwomar-
appreciate theprocesses ofpulvinus and
exposure age- ginalranks orbyhaving
colorplasticity whenhedescribed different onlythemedialrank.
dependent
Trinidadian populations ofB. majoras eight newspe- Leaf-blade shape
cies,five"withspines white-bulbous atthebase"and 11-0 =
to 1.5 (averaging
theothers "withspines. . . notthickened andwhite at times slightly longer up to 1.4)
longer than rachis; 1 = at least1.5to several
thebase."Itisimportant, therefore, thatfully exposed timeslonger.
pulvini areexamined onmature, butnotaging [5]{0.33}
organs. The smallamount
= ofoverlapin measurement of
5-6. 00 entire shaftofspineblackor verydark ranges wasseparated by means, midpoints andmodes.
brown tochestnut brown orgray;10 = basalregion of Onlytheotu ten,which wasscored as state1, ranged
shaft (lowerV5-V3) stramineous; 01 = middleregion from1.45to 1.55.It couldbe codedas unknown or
stramineous; =
11 baseandmiddle stramineous, though as either butthetopology ofFigure3 wouldnot
state,
apexis usually dark.[5,3] {1.00,0.33} be changed. Thecharacter is partially correlated with
Thesetwocharacters areonlypartially correlated.bifid leaves,char.13.
Likewise, 61is partially correlated with32.Plantsof 12. 0 = about1-3
camandsavwithgray spines aredistinctive, butvari- 4 ormoretimes. timeslonger thanwide;1 = about
ationamongorgans andcollections ofthesespecies In [10] {0.50}
verifies thatgray spines intergrade intochestnut brown withthe otu's sampled,
bifidleaves
thiswaspartially
in
correlated
andnearly blackones.Species with state11areusually bifid (char. 13); the species withlong
consistent forthestramineous color.Bactris montícolaascend leaves,theveinsdiverge at a smallangleand
s.l.maybe an exception. Thespeciesusually is char- natifidrather thanspreading outward as in thepin-
acterized state 11 but of B. relatives.
by , plants actinoneura,
which wasplacedinsynonymy byWessels Boer( 1965), 13. 0 = pinnatifid; 1= simple (i.e.,bifid)orirregularly
arecharacterized bystate00. Thus,thesecharacterspinnatifid proximally witha largeterminal bifidseg-
willrequiremorecareful analysesof variation in mentcomprising at leastthedistalhalfoftheblade.
monographic studies. [5]{0.20}
Theswitch from pinnatifid to bifidmaybe devel-
Leafbase,petioleand leaf rachis opmentally andgenetically simple.Theapomorphic
7. 0 = ocreaelacking; 1 = ocreaedeveloped on all stateacrossthewholegenusis clearly homoplasious;
leaves;2 = ocreaedeveloped onjuvenileleaves(of itoccurs inseveral diverse lineages andseparates close-
seedlings andsucker shoots) only.[10]{1.00} lyrelatedspecies.Variability ofthecharacter within a
An ocreais a distaltubular extension oftheleaf species hasbeensuggested forcertain taxa(conspecific
sheath abovethepoint where thepetiole diverges from status ofB. hirtaandB. pedinatacomplexes, andof
thesheath. It is a synapomorphy ofDesmoncus and simandten; WesselsBoer,1965;A. J.Henderson,
Bactris. InDesmoncus, theocreaearepersistent, green pers.comm.;L. R. Noblick, pers.comm.). Thisvari-
andleaflike, andhence, aregenerally noticed andde- ability isclearlydocumented forB. duplex (Allen 3357,
scribed intheliterature. InBactris they areephemeral,bh).Thisneedsmuchmorecareful attention infuture
membranous, andoften shred intoa fibrous mass;and monographic studies.
they arerarely included inspecimens ofBactris. Nev-
ertheless, evidence ofan ocrea(state1) is available Leaf-bladeindûment
evenonoldleaves, ifthecorrect portion ofthepetiole 14. 0 = adaxialnerves glabrous; 1 = tomentose. [10]
isincluded; thedeteriorated ocrealeavesa scar(usually {1.00}
as an inverted "U" or"V") acrosstheentire adaxial Taxathataresupposedly distinguished bythepres-
faceofthepetiole. Within single caespitose clumps of ence,density, andlength ofhairson thenerves and
species ofthenon-ocreate clade canbefound a few laminar tissues weregenerally so variable thatindû-
juvenile leaveswithocreae, fewer subadult leaveswith ment characters wereuseless cladistically. However, of
these, thisonecharacter didappeartobeconsistent in Inflorescence axis

distinguishing ele.Itmaybeuseful ina posteriori plac- 24. 0 = peduncle armed withspines orstout spinules;
ingtheB. tomentosa complex nearele. 1 = unarmed orwitha fewweakspinules. [3]{0.17}
Thestates areusually clearcut.A fewspecies (hon,
Structureof leaf-segment apex mon,andset) arevariable, having theintermediate
15. 0 = midrib (primary vein)ofsegment bisects the condition insomecollections.
distal portion ofthesegment intotwoequalorunequal 25. 0 = rachis withwelldeveloped internodes, about
teeth, andit either terminates subapically (Astrocar - xkormoreofthe ofthepeduncle; 1 = rachis
or extends to Desmoncus length
and
yum Bactris) verytip( ), abbreviated with poorly developed internodes; lessthan
notforming margin ofeither sideofsegment; 1= mid- V7ofthelength ofthepeduncle. [3] {0.25}
ribcontinues as inner margin ofthelonger sideofthe =
segment, usually terminating apically(thus,apexis 26. 0 shapeofrachisbracts (especially
oroblong,
proximal
withsides
oblique). [3]{0.50} ones)deltate, ovate,lanceolate,
toapicaltermination ofnerve; =
1 tranversely
Instate1, onesideofthesegment is always shorter tapering
anddecurrent ontothemidrib. Eithertheouter margin oblong, terminal nerve mayextend as mucronate, cus-
ofthelonger sideis decurrent ontoanawnlike exten- pidate,subulate, orcaudate tip,butisweakandephem-
sionofthemidrib ortogether themidrib andthelonger eral;2 = transversely linear(i.e.,annulate flapsub-
sideterminate inan acutepoint.Sometimes in state tending rachis branch), terminal veinusually extending
1, theshorter sideofthesegment maypullawayfrom onlyas a muero orweakawn.[3]{0.33}
themidrib, two
forming unequal teeth;but the midrib
stillcanbeseenforming theinner margin ofthelonger Rachillae
side.Instate0,neither sideofthesegment isdecurrent27-31.00000= numbering about15to40; 10000=
ontothemidrib, andeachusually forms a tooth. Ifone about5 to 15;11000= 3 to5; 11100= 1or2; 00010
toothis muchlonger, thecondition maysuperficially = about40 to80;00011 = about100ormore.[ 3, 5,
resemble theapomorphy. I neverobserved anytaxon 10,2, 10]{0.13,0.33,0.50,0.33,1.00}
withanintermediate condition. Inthecontext ofthewholegenus, thestates arenot
Theoccurrences ofstate0 inAstrocaryum andBac- sharply delineated and thereis marked homoplasy.
trisprobably arenothomologous. Furthermore, the However, thediscontinuities areclearer among closely
conditions ofstate0 (equalteethin thebidentate related taxa.Scoring wascomplicated insomespecies
clade and veryunequalteethin thenon-ocreate bywidevariation. Forinstance, B. simplicifrons s. str.
clade; Figure7, Table 3) might be codedbetter as hasa single rachilla, butbroadly delimited ithas1 to
twoseparate modifications ofstate1. 4 rachillae. It was scoredas state11000,themore
=
16. 0 narrowed andentire orbidentate; =
1 broad- plesiomorphous of thetwopossibilities, on theas-
sumption thatthisis theancestral condition.
enedandpraemorse. [10]{1.00}(SeeFigure7.)
32. 0 = at anthesis, longer than6 cm;1 = 2-6 cm.
Prophyll [10]{0.50}
17. 0 = V2orlessas longas peduncular bract;1 = 3/5 33-34.00 = at=anthesis, firm,either straight,curved,
tonearly as longas peduncular bract.[3]{0.33} orflexuous; 10 rigidly straight; 01= filamentous and
weak.[10,10]{0.50,1.00}
Peduncularbract Rachilla diameter couldnotbe usedas anindexof
18. 0 = 5-15timeslonger thanwide;1 = >25 times thesecharacters; thatis,diameter is onlyimperfectly
correlated with them. The entirerange ofdiameters in
longer thanwide.[10]{1.00} is onlyfrom about0.3to3.0mm,andthere is
Bactris
19. 0 = apexabruptly constricted anddistally pro- often overlap indiameters between states 00 and01.
longed =
intoa beak;1 ± acuteandnotprolonged into
a beak[10]{1.00} 35-36.00 = glabrous, lepidote, orlepidote-mealy; 10
= scurfy ortomentose by meansofbulbous ormonil-
20. 0 = abaxialindûment onlyscurfy, tomentose, la- iform hairs;X1 = densely, thickly lanatebymeansof
nate,ormixture thereof; 1 = predominantly lepidote curly, slender hairs.[2,5] {0.50,0.50}
orlepidote-setose. [10]{0} = tuberculate or aciculate; 10 =
State1isuniqueinB. actinoneura andB. montícola37-38.00 1= surface a raised
s.st.Whenthese arecombined withB. diviscupula and lacunose; 0 alveolate (with thin, reticulum).
B. sigmoidea, theotu monbecomes variable andis [10,10]{1.00,0.50} issomewhat
thecharacter as Typically, therachilla inBactris grooved
codedas Thus,paupcalculated andridged between intervening nodesbecause vascular
invariant. andfiber bundles makeupmostoftheinternal struc-
21-22.00 = armature moderately densely composed ture.Thepedicelscarsandremnants areusually ele-
ofspines andspinules; 10= lacking; 01= very densely vatedandsometimes prolonged. Therachilla bracts
composed ofsoft,flexible setiform spinules. [10,3] aresometimes persistent, indurate andsharp. Thewhole
{1.00,0.33} appearance isoneofanirregular andsomewhat spirally
23. 0 = baseandmiddleofshaft ofspines/spinules twisted arrangement ofalternating grooves andpoints
blackor darkbrown(or lacking); 1 = lightgrayor alongthelength oftherachilla.
stramineous. [2]{0.17} Bothapomorphies appearto be independent re-
Thisis onlypartially correlated withchar.61,as it sponses tohaving pistillate flowers onadjacent nodes.
occurs inseveral otu'slacking thatcharacter state. Lacunose rachillae (state10)aremorecylindric than
tuberculate ones(state00) anddo notrevealtheun- 45. 0 = calyx rotate, i.e.,lobesspreading; 1= cupulate

derlying vascular traces as clearly. Eachflower is ac- or campanulate, i.e.,lobeserector ascending: unor-
tuallyimpressed intoa lacunaor depression ofthe dered.[1]{0.10}
rachilla, whichis reminiscent ofthestructure ofthe 46. 0 = calyxshallowly deltate-lobed orcuspidate, se-
(staminate) spikelet rachis inTripsacum L. andrelated palsconnate morethan% oflength from base; 1=
grasses.Unlikethetuberculate rachillae, thealveolate calyxmoderately triangular- oracuminate-lobed, se-
ones(state01) havethegrooves limited tothelength palsconnate aboutxhto Vi'2 = deeplysubulate- or
oftheinsertion ofthepistillate flowers, andthetissue linear-lobed, sepalsconnate onlyxkorless:unordered.
between (part axis and part remnants of rachilla bracts) [3]{0.33}
israisedintothinridges; thus, a regular geometric pat-
ternis formed. 47. 0 = onaverage, petals connate for0.1-0.4oftheir
= = length; 1 = onaverage, connate 0.45-0.6oflength; 2
39. 0 phyllotactic spiralsymmetrical; 1 spiral = connate, apices sealedbythinmembra-
and nodeslaterally offset completely
fromadaxial noussuture,
asymmetric not atanthesis.
midline (orsymmetrical ifwholeinflorescence is re- [3]{0.33} possibly opening normally
ducedtoa single thickened spike). [10]{constant with-
inBactris, forwhich itis a synapomorphy} 48. 0 = petalsrhombic, trullate ortriangular acumi-
nate,flatorsharply inflexed apically;1= ovate, round-
Arrangement of triads(reduced cincinnus oftwo ed,incurved andsomewhat cuculiate from thebase.
lateral
staminate andonecentral pistillateflower) [10]{1.00}
40. 0 = on contiguous nodes;1 = sometimes oneor Thetwocomponents, petalshapeandpetalposture,
twodiads/monads ofstaminate flowers tri- arecompletely correlated andwerecombined as one
separating
ads,many ofwhich areoncontiguous nodes;2 = nearly character.
alwaysseparated byfewto severalnodesofdiads/ 49. 0 = receptacle extending intoflower nomorethan
monads. [4] {0.40} lhthelength ofthepetals;1 = extending aboutVithe
Thiswasinitially codedas binary fortwostates (0 length ofthepetals.[5]{0.17}
and2). However, severaltaxain pirathatwerere- 50. 0 = petalsadnatetofulllength ofthereceptacle,
ported intheliterature topossessstate0 actually had so there is no clearseparation between thetwo;1 =
a fewscattered diads/monads amongthetriads, and notcompletely adnate, sothere isa clearseparation of
thecharacter wasdividedintothreestates. It is not receptacle andpetaltissues abovetheinsertion ofthe
clearwhether state1preceded state2 intheevolution petals.[1]{0.14}
oftheancestor ofBactris orwhether state1 is simply
a modification ofstate2 intheancestor
=
ofpira.State 51. 0 petals coriaceous; 1= membranous. [10]{ 1.00}
0 (contiguous nodes)inBactris wascodedthesameas Androecium
contiguous nodesintheoutgroups toseeifparsimony =
analysis wouldsuggest thatamylandpiraareprimi- 52. 0 stamens 6; 1 = 12.[10]{1.00}
tivein Bactris. the
However, analysis confirmed the 53. 0 = stamens marginal onthereceptacle; 1= ± cen-
opposite. Therefore, developmental studies wouldbe tral,clearly separate from thepetals.[5]{0.33}
useful toevaluate thehomology amongBactris, Des- 54-55.00 = stamens barely adnatetopetalsforonly
moncus, andAstrocaryum. a shortdistance abovereceptacle; 10= completely free
In positioning in
cns pira,Burret (1933-1934) im- from petals; 01 = strongly adnate topetals[3,3] {0.20,
pliesthat thespecies ischaracterized bystate 0,a report 0.33}
thatwasrepeated byWessels Boer(1965).However, Thestatesareusually clearcut.Characters 53 and
thematerial ofcnsavailable clearly possessed state2. 54 arepartially correlated, butnotalltaxahaving 541
41.0 = lacking nodesofdiads/monads ofstaminate alsohave53
flowers proximal tonodesoftriads; 1 = nodesbearing 56. 0 = filaments triangular-attenuate; 1= filamentous
diads/monads proximal totriads. [5]{1.00} orvermiform. [3] {0.50}
Although thisispartially correlated withchar.40,it Becausestate0 is found onlyintheoutgroups and
wasrecognized tohelpdistinguish pirafrom amyl,in within derived lineages oftheatropurpureous clade,
whichstate1 is lacking. Developmental studies are developmental studies areneededtoevaluate homol-
neededtoredefine thesetwointerrelated characters. ogyofthecharacter inthedisparate groups.
42. 0 = triads onlyintheproximal %orlessofrachilla 57. 0 = filament erect andunfolded (exceptforinflexed
length; 1 = extending thefulllength oftherachilla or insertion ontoanther); 1 = tortuous orfolded (some-
so.
nearly [10]{1.00} timesbecoming straight butpersistently marked by
43. 0 = onrachilla, triads distichous orhelically sub- foldlines).[3]{0.25}
distichous, distaldiads/monads thesame,or several 58. 0 = anther rectangular-linear oroblong-linear; 1
ranked; 1 = triads and(whenpresent) diads/monads= ovateorovate-oblong. [2] {0.33}
3 to 5 ranked; 2 = triads3 to 5 ranked butdiads/
monadsabout5 to 7 ranked; 3 = triadsanddiads/ 59-60. 00 = anther aboutV5-V4 ofcorollalength; 10
= usually aboutxh(toapproaching Va)ofcorolla length;
monads about7 to 15ranked. [10]{1.00} 01 = aboutV6or lessofcorollalength. [3,5] {0.25,
Staminateflowers- general structure and 0.50}
PERIANTH 61. 0 = anthers included during allstages; 1= exserted
44. 0 = staminate flowers longer thanpistillate; 1= between corollalobesat timeofpollenreleaseand
shorter. [3] {0.33} persisting thereas flowers deteriorate.[5]{1.00}
1991] SANDERS: CLADISTICS OF BACTRIS 13 1
Pistillateflowers-perianth and staminodes thereason whythecharacter ishomoplastic abovethe

62-63.00 = calyxbroadly or
cylindric cupulate, in- levelof species-complexes.
cluding anycusps, aboutV3-% as longas corolla;10= 76. 0 = stigmatic surface marginal; 1= covering entire
urceolate (narrowed at mouth) ornarrowly cylindric,surface: unordered. [3]{0.17}
including anycusps, aboutlkas longas orlonger than 77. 0 = stigma 3-lobed; 1 = unlobed. [2]{0.17}
corolla;01= annulate orflattened, including anycusps,
aboutVio-1/» as longas corolla.[3,2] {0.25,0.14} Cupule(accrescent calyxandcorollainfruit)
Becausethecalyxandcorollaareoften proximally =
pressed against therachilla anddilated, theflower was 78. 0 corolla about2 timesormoretheheight of
removed from therachilla formeasurement. Short ca- the calyx; 1 = about 1Vi times or lessthe height of the
lyceshaveonlya smalldistalportion visible from the calyx.[3]{0.33}
side.Intra- andinterspecific variation needsfurther79-80.00 = relative sizeofcalyxandcorolla remain-
study toverify gapcoding. ingunchanged between anthesis andfruiting; 10 = ca-
64-65.00 = calyxsinuate orshortly cuspidate, with lyxenlarges (relative to corolla)1.5-1.8times;20 =
sinuses large(relative tolobesorcusps)androunded; calyx enlarges 2-4 or more times; 01 = calyxsizerel-
=
10 truncate, completely without sinuses orcusps;01 ative to corolla reduced to 0.7 or less oforiginal pro-
= lobed,withsinusessmall,narrow andacute:65, portional size.[5,3] {0.33,0.14}
unordered. [10,10]{0.50,0.50} Externalcharactersof fruit
66-67.00 = corollacupulate (broadly cylindric) or =
broadly urceolate; =
10 narrowly urceolate ornarrowly81-82.00 shape(excluding rostrum) globose, ovoid,
aboutequalling
obovoid from baseto
cylindric; 01 = broadly campanulate-patelliform. [1, to 1.2timestransverseapex,length
(i.e.,
10= distinctly de-
10]{0.10,1.00} diameter);
pressed, i.e., oblate, broadly obovoid, orbroadly ob-
68-69.00= corolla ± lobedorevenly sinuate, iflobed, conic(length 0.9-0.8timesdiameter); 20 = strongly
thelobesbroadly deltate andthesinuses "V"-shaped depressed, transversely ellipsoid oroblongoid inside
oracute;10= truncate, completely without sinuses or view(length 0.75times diameter orless);01= ellipsoid
cusps;01 = cuspidate, thesinuses broadandrounded. (length 1.25-1.4times diameter); 02 = narrowly ellip-
[5,1] {0.33,0.11} soid(length 1.5-2.0times diameter). [3,10]{0.15,0.50}
70. 0 = corollamargin notfimbriate; 1 = fimbriate.83. 0 = spherical intopview,alltransverse diameters
[3]{0.33} aboutequal;1 = elliptical oroblong intopview,dis-
71-72.00= corolla externally glabrous; 10= lepidote tinctly laterally compressed. [10]{1.00}
tosetoseorscabrous; 01 = lanate.[2,10]{0.17,0.50} 84-85.00 = length + diameter, including rostrum =
73. 0 = staminodes absent orpresent onlyas 6 distinct21-35 mm;10= length + diameter = 40-80mm;01
toothlike appendages oncorolla;1 = connate, forming= length + diameter = 20 mmorless.[3,3] {0.17,
a cupandproximally adnate to base of corolla. [10] 0.25}
{1.00} Thiswasusedas an indexoffruit size.Takensep-
No intermediates areknown. Becausethepresencearately, both length anddiameter lacked sufficientbreaks
ofdistinct toothlike staminodes ishighly variable with- inmeasurement ranges genus-wide tobecodedas dis-
inspecies, itwasnottreated as anadditional state. cretestates, whereas thecombined measure yielded
gaps.
Gynoecium 86-87.00 = glabrous; 10= setoseorscabrous; 01 =
= lepidote or scurfy; 02 = thinly tomentose or lanate. [5,
74. 0 pistil ± cylindric, narrowly ovoidoroblongoid;
1 = ± globose orbroadly ovate,often proximally tu- 2] {0.40,0.22}
mid.[1]{0.14} 88-90.000= mature epicarp red,orange, bright yel-
Asindicated inthediscussion section, allfeatures of low,orstramineous; 100= darkpurplish brown, vi-
thegynoecium (74-77)needtobeexamined from fresh olet-black, orblue-black; 010= deeprosypurple; 001
orliquid-preserved material forthepatterns ofvari- = red-orange withrapidnecrosis toblack.[10,10,10]
ationtobeclarified. Thedatainthisstudy suggest that {1.00,1.00,1.00}
thesecharacters willhavegreatest utility at thesec- Forcomments about010,seediscussion section.
tionallevelorbelow. Certain taxaoftheB. hirta complex wereoriginally
75. 0 = stylebranches freeor onlybasallyconnate, described as having blackfruits (char.90'); thiswas
erect or 1 = branches com- consistent with thelabel data on thelisted sample stud-
entirelyexposed, recurved; ied here. own field observations
pletelyconnate alongmargins, either remaining ternate However, my (made
orrounding outintoa blunttriangle orcircleseated after thepaupanalysis wascompleted) suggested that
onapicalsurface ofstyle, truncate acrosstop,central thiscondition results from precocious necrosis ofthe
convergence recessed intostyle(see char.77); 2 = epicarp. Thistendency maybegenetically based;but,
branches connateand completely sub- because the collection time is critical to scoring, the
completely from thetopology ofthecladogram
merged andrecessed intocenter ofstyle, indicated only statewasdeleted
bytrilete markon truncate apexofstyle: unordered.(Figure3).
[1]{0.22} 91. 0 = epicarp weakly striateorsmooth; 1= distinctly
Examination ofdeveloping flowers intrc{Bernal et striate.[2] {0.20}
al. 1103,ftg)suggested tomethatpostgenital fusion A logicalbasisforthischaracter couldbethepres-
ofstyle branches caneasilybemodified. Thismaybe enceorabsence ofstoutbundles belowtheepicarp in
theouter mesocarp (char.99),buttheresults showthat 103.0 = endocarp surface smooth, tuberculate, or

thesetwocharacters werenotcorrelated. weakly pitted onlyatbase;1 = sharply pittedmedially
92. 0 = apexrounded, mammillate orapiculate (i.e., anddistally. [3]{0.25}
anytypical apicalprojection lessthan15%ofthelength 104.0 = endocarp basenormal, notexpanded; 1=
ofbodyoffruit); 1 = clearly rostrate (typical rostrum or
expanded elongate asevidenced by more widely sep-
longer than15%ofbodyoffruit). [2]{0.25} aratedfibers ortubercles. [10]{0.25}
Theassumption is madethatthisrepresents an al-
Mesocarpand endocarp lometric difference, i.e.,therateorduration ofelon-
93-95.000= mesocarp consisting ofthinly mealy or gation oftheproximal halfoftheendocarp relative to
starchy pulp;100= consisting ofthin- walled juicesacs thedistalhalf.Becausetheresults suggest thatit is
thatareattached toendocarp fibers;010= consistingcorrelated with char.101,developmental studies ofthe
ofthickor densestarch; 001 = containing copious fruit mayclarify therelationship andsignificance of
amounts ofoil thatis easilypressed outoffresh or thesecharacters.
rehydrated material. [10,10,5] {1.00,1.00,0.33} 105-106.00 = endocarp poresnotequidistant (inpo-
Thediscreteness ofstates 000and010needsfurtherlarview),notseparated by equalangles;10 = pores
analysis using fresh or liquid-preserved material. about60°apart, equidistant; =
01 sterile poresstrongly
96-97. 00= endocarp fibersmoderate innumber, stout, displaced toonesideofendocarp. [10,10]{1.00,1.00}
flattened,apressed, andadnate toendocarp surface; 10 Theextremes ofstates 00 and01 arereadily distin-
= moderate to fewin number, firm andwiry, proxi- guishedsubjectively although detailedquantitative
mally somewhat flattened,appressed, adnate andoften characterization andanalysis areneededto evaluate
sunken ingrooves, distally terete,diverging from en- theactualgapamong thethree states.
docarpsurface anddeclined toward endocarp base;20
= moderate to fewin number, firm andwiry, com-
pletelyterete,freefrom endocarp from pointofinser- AppendixIII. Rejected Characters
tion,diverging anddeclined through mesocarp; 21 =
moderate innumber, terete, firm andwiry tofineex- Thesecharacters wereinitially includedinsur-
ceptatpoint ofattachment where abruptly swollen into but wereeitherwidelyvariablewithina spe-
tubercular mound, radiating oronlydistally vey
declined; cies or
22= extremely fine mass species-complex, wereunknownin too
numerous, terete, very (entire taxa were uniform
almostcottony), attached byminute tubercles orpa- many (marked by
pillae,
radiating oronlydistally declined; 30= varying throughout thegenus(written in bold),or had a
from fiberslacking to veryfew,andifpresent, then unitconsistency (C) less than0.1 in theinitial
terete,free,anddeclined. [10,10] {0.50,1.00}(See PAUPanalysis(written in italic).
Figure6.)
Inmostcases,scoring forstates 21and22wasclear Habitand stem. 1. Number ofstems percluster. 2.
cut.Thefewtransitional taxa(e.g.,cor) werescored *Rhizome lengths relative todiameters as indicator of
as state21,themostplesiomorphous ofthetwostates. cluster densities. 3. *Number andsizeofbasaladven-
Quantification ofnumbers is needed. titiousroots.4. Stemheight. 5. Internode: a. *length,
= b. Ozonation orbanding, c. *armature, d. *indument.
98. 0 outermesocarp fibers (periclinal andjustin- 6. *Width ofnoderelative tointernode below.
ternaltoepicarp) relatively numerous, covering most
ofthecircumference ofmesocarp; 1 = relatively few, General features of armature. 7. Modal length
leaving noticeable sections ofareasofmesocarp un- of:a. longest spines, b. moderate spines, c. shortest
covered. [2]{0.20} spines.8.Orientation ofspines. 9.Presence ofpulvinus-
Thedifferences arerather subjective; noquantitativelikeswelling at insertion. 10.Indûment of:a. base,b.
basiswasformulated forthisstudy butis neededfor shaft. 11. Colorofapex.
future analyses. Generalfeaturesofleaves. 12.Number incrown.
99. 0 = outer mesocarp fibers regularly anastomosing13.Orientation within crown.
overwholesurface ofmesocarp; =
1 parallel orwith
a fewanastomoses atpolesofmesocarp. [2]{0.13} Leafsheath. 14.Total length. 15.Apicalfreepor-
100. 0 = endocarp vertically obliquewithfertile pore tion:a. "length, b. *texture, c. *shape.16.Overall: a.
clearlymore distalthan thesterilepores; 1 = not oblique, striations,b. armature, c. structure of trichomes, d.
fertile
andsterile poresataboutthesamevertical po- indûment color.
sition.[5]{0.20} Petiole. 17.Dimensions: a. length, b. diameter. 18.
101.0 = endocarp notdepressed, length equalling or Adaxial features. 19.Distribution ofspinules. 20. In-
greaterthandiameter; 1= depressed, broadly obconic, dûment: a. structure oftrichomes, b. color.
oroblate, diameter greater thanlength. [5]{0.25}
Although apparently connected withchar.81,itis Leaf rachis. 21. Length. 22. Presence ofterminal
inpartnotcongruent withit.Furthermore, themea- filamentous extension. 23.Adaxial indûment: a. struc-
surement ofendocarp length isnotcomplicated tureoftrichomes,
bythe structure b. color.24. Abaxialindûment: a.
presence of a rostrum as in the case of char. 81. of trichomes, b. color. 25. Distribution of:a.
b.
102. 0 = greatestdiameter ofendocarp medial ordistal spines, spinules.
(hence, endocarp isglobose, obconic, ellipsoid, etc.);1 Blade. 26.Dimensions: a. length, b.width, c. length-
= proximal (hence, endocarp ovoid).[10]{1.00} widthratiowhen lessthan3. 27.Proportion ofmargins
thatareparallel. 28. Widestpointalonglength. 29. Staminateflowers. 72. Calyxlength. 73. Corolla

Shapeat base.30.Angleofterminal bifurcation. 31. length. 74.Corolla indûment: a. structureoftrichomes,
Number ofprimary nerves. 32.Texture. 33.Indûment,b. distribution. 75.Filament length. 76.Anther inser-
adaxialsurface oflamina: a. structure oftrichomes, b. tion.77.Shapeofanther base.78.Pistillode: a. pres-
color.34.Indûment, adaxialsurface ofnerves: a. struc- ence,b. length, c. shape.
tureoftrichomes, b. color.35.Indûment, abaxialsur-
faceoflamina: a. structure oftrichomes, b. color.36. Pistillateflowers. 79. Calyxlength. 80. Presence
abaxial surface of nerves: a. structure oftri- ofannular flange inside calyx and adnate toitsbase.1
Indûment, 81.Corolla 82. Staminodes: a. b. de-
chomes, b. color.37. Distribution ofspines.38. Dis- length. presence,
tribution ofspinules. 39.Number ofsecondary nerves greeofconnation/adnation ofbasalportion, c. shape
nerve. 40.Structure oftransverse nerves. whenapicalportion is freeanddistinct, d. length of
perprimary adnateportion relative tocorollalength. 83.Pistilin-
Bladesegments (leaflets). 41.Modalnumber. 42. dûment type.
Number ofplanesintowhich segments areoriented rel-
ativetorachis. 43.Spacing. 44.Structure ofinsertion.Fruitingcupule. 84. Calyxshape.85. Corolla:a.
45. Outline shapeofapex.46. Curvilinear course: a. shape,b. splitting, c. color,d. striations.
basalregion, b. middleregion, c. apicalregion. 47. Fruit. 86.Indûment color.87.Apicalregion sunken
Marginal bands:a. presence, b. general structure, c. ornot.88.Lackofmucilage inmesocarp. 89.Endocarp
width, d. color. length(nogapsforcoding). 90.Structure ofendocarp
Terminalvs.nonterminal segments fibers
radiating from fertile pore. 91. Endosperm ho-
(characteriza-
tionofeachdescriptor foreachofsegment types).48. mogeneity.
Dimensions: a. length, b. width, c. length-width ratio.
49. Length adnateto rachis.50. Widestpointalong
length. 51.Proportion ofmargins thatareparallel. 52.
Number ofprimary nerves. 53.Lateral connation.
Inflorescence (general). 54.*Modeofemergence 1Karsten (1856,1857)founded themonotypic ge-
from leafsheath. nusPyrenoglyphis on B. majoranddiagnosed it by
thepresence ofa staminodial-like ringontheinternal
Prophyll. 55.Width. 56.*Shapeofapex.57.Pres- faceofthepistillate calyx, in addition to thestami-
enceofarmature. 58. Indûment: a. structure oftri- nodialringin thecorolla(thecharacter he usedto
chomes, b. color. defineGuilielma Karst.).Burret(1933-1934)inti-
Peduncularbract. 59. Length. 60. Protraction of matedthatKarsten misinterpreted orerred inhisob-
surface: a. color, b.texture.servation becauseBurret himself wasunabletolocate
apex.61. Adaxial (internal) onanyofthematerial ofB. major
62.Abaxial(external) surface color. anysuchstructure
thathe studied. WithBurret's comment in mind,I
Peduncle. 63. Orientation. 64. Length. 65. Cross- beganexamining thematerial ofB. major. To mysur-
sectional shape.66.Diameter. 67.Indûment: a. struc- prise,thefirst specimen I examined (HullH-l, ftg),
tureoftrichomes, b. color,c. distribution. whichoriginated fromnearthetypelocality, clearly
possessedthisintracalycine ring; however, specimens
Rachis. 68.Length. thatoriginated elsewhere werenearlyidentical but
Rachisbracts. 69.Length of:a. largest, b.^smallest. lackedthering.Thisappearsto be a geographically
localized
character within B. majorthatis ofnotax-
Rachillae. 70. Armature distribution. 71. Indû- onomie consequence although itmayrepresent a case
ment: a. distribution, b. color. ofhomoeosis (Sattler, 1988).
Noteaddedin proof-changesto AppendixI: Recentstudysuggests thatBactrisferruginea should

be movedfromtheripariacomplex(24) to thelongifrons complex(35) and thatB. pickeliishould
be movedfromthetomentosa complex(39) to theeumorphacomplex(40). A studyby Wessels
Boer(Palmasindígenasde Venezuela.Pittieria17: 1-332. 1988)cameto myattention in whichhe
knownspeciesin synonymy, as follows:B. bergantina= B. setulosa,B.
placedseveralinadequately
- B. corossilla.
duidae= B. corosilla,B. falcata= B. setulosa,B. venezuelensis

Cladistics of Bactris (Palmae)

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cladistics of Bactris (Palmae)

Uploaded by

Copyright:

Available Formats

Marie Selby Botanical Gardens Inc.

CLADISTICS OF BACTRIS (PALMAE): SURVEY OF CHARACTERS AND REFUTATION OF BURRET'S

You may need to log in to JSTOR to access the linked references.

CLADISTICS OF BACTRIS (PALMAE):

Abstract. In 1934,M. Burret published ofBactris

BactrisJacq. ex Scop., the largestgenusof in GuilielmaMart,werecompletelyexcluded

Figure1. Phylogenetic ofclassification

work,ithasbecomethestandardtaxonomieref- amylocarpus"] wasfurther dividedintotwosub-

theirprogenitors?"(Mora-Urpi & Clement,1981; monophyletic statusofBactriss.l. itself,

TABLE 2 (continued). Designations are: 0 = primitive; 1 (2 or

of recognizing all majormonophyletic lineages AppendixII). The rejectedcharacters and fea-

Figure2. Branch-and-bound (bandb) ofproximal

conductedwiththe initialset of 63 otu's (61 resultedfrombandb,the strictconsensustree

was nestedalongwithseveralotu's withinthe thebandbof proximalclades.This is expected

Figure3. Composite cladogram forall sampled otu'sinBactris, showing detailsofcladesthatconstitute

tently examinenon-guilielmas forthecharacter. uniquesynapomorphies (97, tuberculateendo-

Table 3. Summary ofknown monophyletic groups Table3. Continued.

Table3. Continued. Table3. Continued.

peduncular spines stronglyreduced orlacking [24]; cylindric [66],cuspidate orlobed[ 68°];infruit,

Figure6. DispositionofBiuret's(1933-1934)groupsandillustration in fruit

nestedwithinthetuberculate clade thanguil clade; thismakesBurreťssubgenuspolyphy-

Figure7. Evolution ofcharacter

amongspeciesofbact. To recognize repeatedat increasingly

discussedforninecharacters (1, 11, 27-31, 62, thecladogram, yetfruit shapereversed fivetimes

notexpectedto showall thebasic pyrecharac- rhombic;stamenssix,emerged aroundthemar-

fertileone).In thebalanophora clade, theen- theepicarpis a lightgreen,thenfleshcolored,

incomplete, and thepositionsof severalclades BarbosaRodrigues,J. 1903. Sertum palmaram

Sattler,R. 1988.Homeosis inplants.Amer. J.Bot. BALANOPHORA CLADE

B. turbinata,B. unaensis Barb.Rod.;perhaps ATROPURPUREOUS CLADE

an incomplete correlation, as tapering spinesmaybe partial ocreae, andnumerous adultleaveslacking oc-

these, thisonecharacter didappeartobeconsistent in Inflorescence axis

tuberculate ones(state00) anddo notrevealtheun- 45. 0 = calyx rotate, i.e.,lobesspreading; 1= cupulate

Pistillateflowers-perianth and staminodes thereason whythecharacter ishomoplastic abovethe

theouter mesocarp (char.99),buttheresults showthat 103.0 = endocarp surface smooth, tuberculate, or

thatareparallel. 28. Widestpointalonglength. 29. Staminateflowers. 72. Calyxlength. 73. Corolla

Noteaddedin proof-changesto AppendixI: Recentstudysuggests thatBactrisferruginea should

You might also like