Ecological Entomology (2000) 25, 460±466

Insect colonisation sequences in bracts of Heliconia

caribaea in Puerto Rico
B A R B A R A A . R I C H A R D S O N and G I L L I A N A . H U L L Department of Biological
Sciences, Napier University, Edinburgh, U.K.

Abstract. 1. It was predicted that insects colonising Heliconia phytotelmata would

exploit the different bract conditions along in¯orescences, which result from
sequential ageing.
2. Flowering of H. caribaea is asynchronous over a 6-month period. A method of
identifying bracts of the same age, regardless of position on the in¯orescence or
in¯orescence age, and from different plants over 2 years, was developed using
comparative ¯owering phenology.
3. Heliconia insect larval communities were remarkably consistent from year to
year in species composition and relative abundance. Occurrence within the plant
population was also similar from year to year, and most species were Heliconia
speci®c.
4. Signi®cant interspeci®c differences were found in bract utilisation, with
populations peaking at different stages of bract development. Ceratopogonid larvae
were the earliest colonisers, followed by psychodids, syrphids, and culicids.
Tipulids occurred much later in the cycle of bract development and ageing. These
patterns were consistent in both years.
5. Patterns of bract utilisation provide strong support for temporal niche
partitioning by variation in oviposition and development time. Communities were
not considered to be structured by predation or pH changes along the bract
sequence.
Key words. Colonisation, ¯owering phenology, Heliconia, insect larvae,
phytotelmata.

Introduction in¯uence colonisation patterns, as may associated increases in

microbiota as an additional nutrient resource. The life-history
The sequence of colonisation of organisms in ephemeral hypothesis derives from the requirement that for organisms to
aquatic habitats may be random, follow assembly rules derived complete their life cycle in an ephemeral patch, they must
from a variety of ecological processes, re¯ect niche separation oviposit only in habitats that are young enough to permit their
based on changing physical or chemical conditions of the patch offspring to complete their development before the patch
over time, or re¯ect life-history adaptations to the short-lived expires. Insects with long larval development periods will
nature of the habitat. Assembly rules may be derived from colonise young patches, while those with short developmental
succession, e.g. processing chains or facilitation (Naeem, periods can colonise either young or old patches.
1990; Heard, 1994; Heard & Richardson, 1995), food web The work reported here examined colonisation sequences in
processes, e.g. predator effects (Naeem, 1988; Chen & Wise, insects inhabiting bracts of Heliconia caribaea as a test of the
1999), or re¯ect competitor or predator avoidance in choice of above hypotheses. It was predicted that insects colonising them
oviposition sites (Seifert & Seifert, 1976). would exploit different bract conditions associated with
Physical and chemical changes in habitats, where the basis ageing.
of the food chain is decaying organic matter, may be indicated Heliconia plants are common components of the under-
by changes in pH as decay progresses. If extreme, this may storey of neotropical wet lowland and mid-elevational forests,
particularly along water courses (B. A. Richardson, pers. obs.),
Correspondence: Barbara Richardson, 165 Braid Road, Edinburgh where light can penetrate. Many species have cymes of ¯owers
EH10 6JE, U.K. E-mail: [email protected] that develop in in¯orescences of brightly coloured, sequen-

460 # 2000 Blackwell Science Ltd

 Heliconia phytotelmata 461

tially aged, ¯uid-®lled bracts. The cyme within each coloured 1075 m a.s.l.). Mean annual rainfall increases with elevation
bract consists of a series of 10±20 ¯owers, each subtended by a from 3537 to 4849 mm (GarcõÂa-Martino et al., 1996), with
much-folded ¯oral bract. Flowers develop under water and, consequential changes in vegetation, described by Beard
when ready to be pollinated, stamens and stigma protrude (1949) as lower montane rain forest (< 600 m), montane rain
above the water. Each ¯ower is short-lived and there is no forest (600±900 m), and dwarf forest on exposed ridges
more than one ¯ower at pollination stage each day (Berry & (> 900 m). This study was carried out in the lowest elevation
Kress, 1991). After ¯owering, abscission of corollas and forest, known locally as tabonuco forest because of the
stamens occurs rapidly and they fold down into the bract ¯uid dominant tree Dacryodes excelsa Vahl. Five locations in the
(Naeem, 1990). Flowers do not necessarily open on successive area of the El Verde Field Station were chosen to give a
days. There is strong inferential evidence that presence of ¯uid representative sample of 15 plants per year. The study was
is a protection against herbivory of ¯owers and seeds (Seifert, carried out in February and March, which is the beginning of
1982). The bract phytotelmata, however, are invaded by the ¯owering season, in 1996 and 1998. The distribution of H.
aquatic insects, mainly dipteran and coleopteran detritivorous caribaea within the forest is not random, as they are gap plants
larvae. These communities have been studied extensively in occurring in clumps along water courses or in tree-fall gaps
different species of Heliconia. They are of two main types: ¯y close to streams. The ®ve sites chosen were along the
(nonmosquito) larvae and beetles (larvae and adults); and Quebrada Sonadora, three of its small tributaries, and a
communities dominated by several species of Culicidae larvae tributary of the Rio Espiritu Santo, to give an adequate
(Seifert & Seifert, 1976, 1979; Seifert, 1982; Machado-Allison coverage of the areas in which Heliconia occurred in the
et al., 1983, 1985). tabonuco forest.
Water in the bracts is replenished by active transport in the
plant, and in¯orescences completely sheltered from rain
typically contain the same amount of water as exposed bracts, Field methods
but exposure to rain can result in periodic dilution of bract ¯uid
(Seifert, 1982). Food resources for the insect communities In¯orescence collection
include nectar and microorganisms, decaying ¯ower parts, and
the bract lining (Seifert, 1982; Naeem, 1990). There is little or Mature in¯orescences of H. caribaea have between ®ve and
no input of detritus from the forest canopy because each bract 16 bracts but because sampling was carried out at the
is sheltered by the one above. beginning of the season most in¯orescences were in the lower
Flowering of Heliconia caribaea Lam. is seasonal in Puerto to mid-range of bract number. Three in¯orescences were
Rico (February±August). Although some insects are phytotelm chosen from widely separated plants (so that they were
generalists, most phytotelm insects are endemic to particular unlikely to be from the same clone) in each of the ®ve
plant species (Seifert, 1982; Machado-Allison et al., 1985). The locations. Each bract was numbered and, to maximise
latter group must therefore also be seasonal and have collection of organisms, ¯uid from each individual bract of
suf®ciently short life cycles to match that of their host plant the in¯orescence was pipetted into a bottle at the sampling site.
species. As individual bracts remain intact for only » 2 months The entire in¯orescence was then cut from the plant, keeping it
(B. A. Richardson, pers. obs.), it is essential that species, for upright, and each individual bract was removed by cutting
which larvae are con®ned to bracts in which they emerged, must through the rachis. Bracts were placed separately into labelled
complete development within this time. The cymes within bags along with their corresponding bottles of ¯uid, for
bracts undergo changes, from sequential ¯owering and nectar transportation to the laboratory, where the ¯uid volume of each
production to decay of ¯ower parts and development of fruits bract was measured. A further 16 plants were labelled and
and seeds, so there will be consequent changes in the amount of monitored every 2 days to determine the rate of bract
decaying organic matter, its physical state, and pH of bract ¯uid. development.
Because plants in Puerto Rican populations come into ¯ower
at different times, bracts cannot be aged and compared directly
by their position in the in¯orescence. Thus it is necessary to Bract ¯uid pH
®nd a method of identifying bracts at the same stage of
development in in¯orescences of different ages on different Four plants were selected that were not used for the
plants. This could be done most readily by identifying a collection of organisms. Fluid was collected from the bracts of
developmental marker in the cymes. Such a method, which these plants in situ. The in¯orescences were then covered with
allows the sequence of insect colonisation of Heliconia polythene bags to prevent entry of rainwater and left for 3 days
caribaea phytotelmata to be compared in separate years, is to allow ¯uid secreted by the plant to replenish the bracts.
reported here. Fluid was then pipetted from individual bracts and taken to the
laboratory for measurement with a pH meter (Corning, New
York, U.S.A.) calibrated at pH 7. For comparison, the pH of
Study area throughfall water, collected into glass containers, was
measured in the same area and at the same time. Four
The study area was in the Luquillo Experimental Forest in the containers were screened with muslin to prevent entry of
mountains of NE Puerto Rico (65°49¢W, 18°19.5¢N, 100± animals and canopy debris, and four were left unscreened.

# 2000 Blackwell Science Ltd, Ecological Entomology, 25, 460±466

462 Barbara A. Richardson and Gillian A. Hull

Fig. 1. Illustration of the method used to determine the equivalent age of bracts in Heliconia caribaea in¯orescences of different ages. The
oldest spike is on the left, the youngest on the right. Flowers in the shaded bracts are at the same stage of development (stage 4). In
standardising the bract age, the shaded bract was renumbered 0. Younger bracts above this were assigned negative values, older bracts below it
were assigned positive values.

Laboratory methods material deposited in the Museum of the University of Puerto

Rico and in the senior author's collection.
Determination of bract age

In¯orescences with bracts numbered in the ®eld from the Statistical analysis
youngest to the oldest bract were renumbered according to the
developmental stage determined by the following method. For Differences in larval abundances for each species between
each bract, the cyme was dissected and the stage of years, and pH at different levels in the in¯orescence, were
development noted for each ¯ower as: (1) developing ovaries, examined using one-way ANOVA. To examine differences
(2) green stamens, (3) stamens dehiscing, or (4) stamens between species and years in colonisation sequences, the bract
collapsed and ¯ower parts decaying. The marker used to age at which the 50% level of abundance of each species was
equate ¯oral development between in¯orescence bracts was reached was calculated from the cumulative frequency curve
the bract in the sequence that was the ®rst to contain a stage 4 for each year. As there were different numbers of bracts in
¯ower; this was designated 0. All younger bracts above this in different age classes, the basis for calculation was the mean
the sequence were given sequential negative numbers, and all number of animals per bract, grouped into sequential pairs of
older bracts sequential positive numbers (Fig. 1). bracts. Signi®cance of differences between bract age at which
50% of the population was reached for each species, and
between years, was examined using two-way ANOVA. The
Collection of organisms actual distribution of larvae among sequentially aged bracts
between years was compared using two-tailed Kolmogorov±
Animals were collected at the same time as bracts were Smirnov tests.
dissected. Each ¯ower, ¯oral bract, and lining of the branch
bract was brushed out with water to dislodge small larvae.
Animals were collected into 80% alcohol, along with those Results
pipetted from bract ¯uid in the ®eld. Larvae and pupae were
identi®ed to morphospecies (Beattie & Oliver, 1994) and Bract and phytotelm developmental changes
counted for each bract. No adult forms were recovered and,
unless otherwise mentioned, references are to larvae. New bracts were found to open on average every 6.4 days
Specimens were sent to taxonomists for veri®cation and (SE = 0.32 days), approximately the same time (6.7 days) as

# 2000 Blackwell Science Ltd, Ecological Entomology, 25, 460±466

 Heliconia phytotelmata 463

on bromeliad phytotelmata, have been observed feeding on

tipulid larvae (B. A. Richardson, unpublished). Other possible
predators were Muscidae larvae but these were not observed as
predators in this study or a study of bromeliad phytotelmata
(Richardson, 1999). The two most abundant species, the
needle-like ceratopogonid and the small transparent acalyp-
terate larvae, were associated with the closely folded ¯oral
bracts and ¯ower parts but were active in the main body of
water at night. Most other species were also nocturnal, and
only a few individuals were observed in the free water of the
bracts during the day.

Fig. 2. Distribution of amounts of water in Heliconia caribaea Colonisation

bracts of different ages. Mean of 15 plants in each of 2 years.
Bars = SE. Bract age increases with bract number, from ±5 to 11. There were signi®cant differences between species in the
location of their peak abundances in the in¯orescence (Table 2,
Fig. 3). The earliest colonisers were ceratopogonids, present in
that found by Machado-Allison et al. (1983) for H. caribaea in even minutely opened bracts; they utilised bracts at the early
Venezuela. The mean pH of the ¯uid in the top two (youngest) ¯owering stages but declined in abundance in bracts containing
bracts was 7.9 (SE = 0.11, n = 7) and signi®cantly different large amounts of decaying material. They were followed by the
from that in the lower (older) bracts, which was 5.2 (SE = 0.27, psychodids. In contrast, tipulids (Limonia sp.) were not present
n = 25) (one-way ANOVA, P < 0.001). The mean pH of until bracts were ¯owering, and their population peaked at later
throughfall water from the canopy in screened and unscreened stages of bract development. Syrphids, the acalypterates, and
containers was 6.9 (SE = 0.09, n = 4) and 6.6 (SE = 0.16, n = 4) Culex antillummagnorum Dyar showed a wide range of bract
respectively, and so unlikely to be responsible for the lower pH utilisation (Fig. 3). There were no signi®cant differences
observed in the older bracts. The ¯uid content of bracts varied, between years in the pattern of colonisation for any of the
depending on changes in the amount of ¯uid secreted by the seven common species (Kolmogorov±Smirnov two-tailed test,
plant as bracts age, and on rainfall and evaporation. Bracts P > 0.05).
before ¯owering contained up to 4 ml of ¯uid but, as ¯owering
commenced (bract 0), amounts of ¯uid increased to a
maximum of 20 ml in some bracts. Towards the end of Discussion
¯owering, the amounts of bract ¯uid declined (Fig. 2).
Heliconia caribaea phytotelm communities showed a striking
similarity in composition, relative abundance and occurrence
Community composition of species, and bract colonisation sequences in different years.
Seifert and Seifert (1976, 1979) and Machado-Allison et al.
Seven morphospecies made up 93±94% of the phytotelm (1983) have shown that these are not top down communities
community and were distributed widely among plants structured by predation and, in general, have low levels of
(Table 1). All were from the dipteran families interaction. This was con®rmed in the present study, where few
Ceratopogonidae, Syrphidae (two species), Culicidae, predators were found. Predatory mosquito larvae [e.g.
Tipulidae, Psychodidae, and an acalypterate taxon (Table 1). Toxorhynchites haemorrhoidalis (Dyar & Knab), observed
The only pupae observed were those of Ceratopogonidae and occasionally within Venezuelan H. caribaea bracts (Machado-
Psychodidae, indicating short life cycles. There was no Allison et al., 1983)] were not found in the Puerto Rican
signi®cant difference in abundance in six of the seven samples, although a Toxorhynchites occurred infrequently in
dominant morphospecies between years (Table 1). The excep- nearby bromeliad phytotelmata (B. A. Richardson, pers. obs.),
tion, Syrphidae, Quichuana sp., had a markedly lower and there were no predatory hydrophilid beetle larvae (Naeem,
abundance and distribution in the second year of the study. 1990). Occasional visiting predators, e.g. staphylinid adults,
Community composition, the total number of species present, earwigs, and spiders waiting on bracts (Seifert & Seifert,
and species distribution were generally similar in each year 1976), were not observed in this study. Hummingbirds have
(Table 1). occasionally been seen apparently sipping ¯uid from the bracts
Unlike Heliconia communities described in South and but if they do take in insect larvae they are not likely to be
Central America (Seifert, 1975; Seifert & Seifert, 1979; selective. Larvae are mostly nocturnal and few are ever
Naeem, 1990), there was a total absence of coleopteran larvae observed in the open part of the bract during the day, but are
and adults, which act as bract scourers. The only predators found within tightly folded ¯oral bracts or within the mass of
found in the system were a few chironomids (Tanypodinae) in decaying ¯ower parts.
two in¯orescences in 1998. These are known piercers or Patterns of distribution in the bract sequences cannot
engulfers (Merritt & Cummins, 1984) and, in a previous study therefore be due to predatory interaction and must be a

# 2000 Blackwell Science Ltd, Ecological Entomology, 25, 460±466

464 Barbara A. Richardson and Gillian A. Hull

Table 1. Occurrence, relative abundance and frequency of the commonest taxa in Heliconia caribaea in¯orescences (15 in each year). Species
are listed in order of decreasing abundance in the ®rst year.

Mean number of individuals Relative Frequency of

per in¯orescence* (6 SE) abundance occurrence (%)²

Taxon 1996 1998 1996 1998 1996 1998

Ceratopogonidae sp. 1 86.2 (6 18.3) 76.8 (6 24.0) 0.36 0.30 100 87

Acalypteratae sp. (family indet.) 55.9 (6 12.9) 84.7 (6 17.9) 0.23 0.33 100 100
Culicidae, Culex antillummagnorum 50.0 (6 14.7) 42.5 (6 9.3) 0.21 0.16 80 80
Syrphidae, Quichuana sp. 15.3 (6 3.7) 3.5 (6 2.1) 0.06 0.01 87 33
Syrphidae, Copestylum sp. 10.9 (6 2.8) 15.6 (6 7.3) 0.05 0.06 66 66
Tipulidae, Limonia sp. 3.4 (6 1.9) 4.6 (6 1.6) 0.01 0.02 47 60
Psychodidae sp. 1 2.7 (6 0.7) 12.6 (6 4.9) 0.01 0.05 73 73
Mean number of individuals per 224.4 (6 30.0) 240.3 (6 42.7)
in¯orescence of the seven taxa above
Mean number of all individuals 238.1 257.7
per in¯orescence
Commonest seven taxa (% of total) 94 93
Total number of taxa recorded 21 22

*Differences between years not signi®cant (one-way ANOVA, P > 0.05), except for Syrphidae, Quichuana sp. (P < 0.01).
²Percentage of 15 in¯orescences in which the taxon was recorded.

Table 2. Sequence of colonisation of Heliconia caribaea 1983) but oxygen levels could be important for the apneustic
in¯orescences. Bract number by which 50% of individuals of each Ceratopogonidae and explain their absence from older bracts.
species had been collected, starting from the youngest. Obtained The high levels of dissolved organic carbon and nitrogen in
from the cumulative frequency curves for each species over the 20± ¯uid of older bracts (Richardson et al., 2000) are indicative of
80 percentile range for each year. Arranged in order of colonisation. anoxic conditions. The six other species living in older bracts
have open respiratory systems, with terminal respiratory
Bract number* siphons or terminal spiracles, and come to the surface for
air, so they can survive as bracts become increasingly anoxic
Taxon 1996 1998 Mean²
due to decay processes. All taxa were detritivores, though
Ceratopogonidae sp. 1 0.0 ±1.5 ±0.7a some Ceratopogonidae (Forcipomyiinae) are known to ingest
Psychodidae sp. 1 1.3 2.3 1.9b fungal spores (Stehr, 1991) and phytotelm ¯uid was observed
Syrphidae, Quichuana sp. 2.1 3.6 2.9bc to contain large numbers of yeast cells. The distribution of
Culicidae, Culex antillummagnorum 3.0 3.2 3.1bc most species was coincident with high water levels in the
Acalypteratae sp. (family indet.) 2.9 3.6 3.3c bracts, but the ®rst colonisers, ceratopogonids, were present
Syrphidae, Copestylum sp. 2.3 5.3 3.8c before the plants began to secrete copious ¯uid into the bracts.
Tipulidae, Limonia sp. 5.1 7.1 6.1d Adult ceratopogonids are minute, and it is possible that they
can enter bracts to oviposit immediately they begin to open.
*Bract age increases with bract number, successive bracts opening at
Differential body size is unlikely to explain the rest of the
6.4 day intervals.
²Values sharing the same postscript letter are not signi®cantly
colonisation sequence as bracts open rapidly once the process
different (P > 0.05, two-way ANOVA, Fisher's LSD = 1.32). has started (B. A. Richardson, pers. obs.). Tipulids, whose
larvae colonise older bracts, have slender abdomens and long
ovipositors, so they could oviposit in newly opened bracts, but
no ®rst-instar larvae were found in young bracts.
re¯ection of oviposition patterns, nutrient or other conditions Oviposition in insects is known to depend on visual,
in the bracts, and active movement and bract selection by some tactile, and chemosensory cues, the last often being plant-
larvae. Syrphids, mosquito larvae, and copepods have been derived chemicals. Field tests on the responses of four
observed moving along the rachis (Seifert, 1982; Naeem, species of mosquito (Lounibos & Machado-Allison, 1993)
1988) and may leave to exploit food sources in younger bracts, showed a high degree of speci®city to individual phytotelm
or if conditions become unfavourable. Whatever the mechan- ¯uids for three species; only one species was a phytotelm
ism, the result is that there is temporal niche separation generalist. In the study reported here, however, which
between some species at their population peaks, and distribu- involved bracts on the same plant, discrimination must be
tion is not random between bracts. even more precise. Bracts are coloured bright yellow and at
It is unlikely that pH differences affect distribution as the ¯owering exude a powerful smell of overripe bananas.
range is well within larval tolerances (Machado-Allison et al., Colour and smell are likely to be distance cues, attracting

# 2000 Blackwell Science Ltd, Ecological Entomology, 25, 460±466

 Heliconia phytotelmata 465

Fig. 3. Sequence of colonisation of Heliconia caribaea phytotelmata by the seven most frequent morphospecies. Proportion of total population
of each taxon in successively opening bracts. Bars = SE. Bract age increases with bract number, from ±5 to 11.

insects to the plants, rather than determinants for precise products, and presence of other animals, that give ®ner
oviposition. It is likely that there are subtle differences in cues, e.g. females of the syrphid ¯y Quichuana aurata
chemistry between bracts, such as sugar content, decay Walker were found to reject bracts if others had previously

# 2000 Blackwell Science Ltd, Ecological Entomology, 25, 460±466

466 Barbara A. Richardson and Gillian A. Hull

laid eggs there, so only newly opened bracts were used for Berry, F. & Kress, W.J. (1991) Heliconia. An Identi®cation Guide.
oviposition (Seifert & Seifert, 1976). Smithsonian Institution Press, Washington.
Heliconia insects exhibit a high degree of speci®city, and Chen, B. & Wise, D.H. (1999) Bottom-up limitation of predaceous
only two of the seven dominant species in this study were arthropods in a detritus-based terrestrial food web. Ecology, 80,
761±772.
found to be phytotelm generalists. Culex antillummagnorum
GarcõÂa-MartinoÂ, A.R., Warner, G.S., Scatena, F.S. & Civco, D.L.
was also found in bromeliads (Richardson et al., 2000), and the
(1996) Rainfall, runoff and elevational relationships in the Luquillo
acalypterate larvae were found occasionally in bromeliads, Mountains of Puerto Rico. Caribbean Journal of Science, 32, 413±
palm leaf bases on the forest ¯oor, and bamboo containers 424.
placed in the forest (Stark, 1997). Speci®city to a particular Heard, S.B. (1994) Processing chain ecology: resource condition and
phytotelm habitat indicates a high level of evolutionary interspeci®c interactions. Journal of Animal Ecology, 63, 451±464.
adaptation. Heard, S.B. & Richardson, J.S. (1995) Shredder±collector facilitation
Wolda (1983) reviewed the evidence for seasonality in in stream detrital food webs: is there enough evidence? Oikos, 72,
tropical insects, which is associated with rainfall changes, 359±366.
rather than daylight and temperature as in temperate regions. Lounibos, L.P. & Machado-Allison, C.E. (1993) Field test of mosquito
Breeding cycles in these phytotelm insects are adapted to the ovipositional cues from Venezuelan phytotelmata. Florida
seasonality of ¯owering in Heliconia, but within this time span Entomologist, 76, 593±599.
Machado-Allison, C.E., Barrera, R.R., Frank, J.H., Delgado, L. &
further precision is required in choice of oviposition and
Gomez-Cova, C. (1985) Mosquito communities in Venezuelan
development sites related to age of bracts and their temporary
phytotelmata. Ecology of Mosquitoes (ed. by L. P. Lounibos, J. R.
nature. Seifert (1982) described differences in the locations of Rey and J. H. Frank), pp. 79±93. Proceedings of a Workshop,
larvae in heliconia bracts of different ages but did not quantify Florida Medical Entomology Laboratory, Vero Beach, Florida.
them. Collectively, these studies and the results presented here Machado-Allison, C.E., Rodriguez, D.J., Barrera, R.R. & Gomez-
provide strong support for temporal niche partitioning by Cova, C. (1983) The insect community associated with in¯or-
variation in ovipositioning and development time among escences of Heliconia caribaea Lamarck in Venezuela.
heliconia insects. Such life-history adaptations to the ephem- Phytotelmata: Terrestrial Plants as Hosts for Aquatic Insect
eral nature of patchy habitats yield predictable sequences in Communities (ed. by J. H. Frank and L. P. Lounibos), pp. 247±
colonisation that ultimately affect community composition and 270. Plexus, Medford, New Jersey.
structure. Merritt, R.W. & Cummins, K.W. (1984) An Introduction to the Aquatic
Insects of North America, 2nd edn. Kendall Hunt, Dubuque, Iowa.
Naeem, S. (1988) Predator±prey interactions and community structure:
Acknowledgements chironomids, mosquitoes and copepods in Heliconia imbricata
(Musaceae). Oecologia, 77, 202±209.
Naeem, S. (1990) Resource heterogeneity and community structure: a
This research was conducted as part of the Luquillo Long
case study in Heliconia imbricata phytotelmata. Oecologia, 84, 29±
Term Ecological Research Project, with additional funds
38.
provided by the USDA Forest Service, Napier University, Richardson, B.A. (1999) The bromeliad microcosm and the assessment
and the Carnegie Trust for the Universities of Scotland. Part of of faunal diversity in a Neotropical forest. Biotropica, 31, 321±336.
this work was submitted by G.H. in partial ful®lment of the Richardson, B.A., Rogers, C. & Richardson, M.J. (2000) Nutrients,
requirements for the B.Sc. (Hons) degree of Napier University. diversity, and community structure of two phytotelm systems in a
We are grateful to the staff of the El Verde Field Station, lower montane forest, Puerto Rico. Ecological Entomology, 25,
University of Puerto Rico, for the use of ®eld and laboratory 348±356.
facilities, Drs G. O'Meara, J. Gelhaus and R. Grimm for Seifert, R.P. (1975) Clumps of Heliconia in¯orescences as ecological
taxonomic advice, Dr P. Klawinski for statistical advice, to islands. Ecology, 56, 1416±1422.
Mike Richardson for ®eld assistance, data management and Seifert, R.P. (1982) Neotropical Heliconia insect communities.
word processing, and to referees for helpful comments. Quarterly Review of Biology, 57, 1±28.
Seifert, R.P. & Seifert, F.H. (1976) Natural history of insects living in
in¯orescences of two species of Heliconia. Journal of the New York
Note added in proof Entomological Society, 84, 233±242.
Seifert, R.P. & Seifert, F.H. (1979) A Heliconia insect community in a
Venezuelan cloud forest. Ecology, 60, 462±467.
Ceratopogonid sp. 1 has been identi®ed by Dr A. Borkent
Stark, R. (1997) A study of the animal communities inhabiting ®ve
(Royal British Columbia Museum) as Forcipomyia
different phytotelm habitats in the Luquillo Experimental Forest,
(Phytohelea) dominicana de Meillon & Wirth, previously Puerto Rico. BSc thesis, Napier University, Edinburgh, UK.
known only from Dominica. Stehr, F.W. (1991) Immature Insects, Vol. 2. Kendall Hunt, Dubuque,
Iowa.
Wolda, H. (1983) Spatial and temporal variation in abundance in
References tropical animals. Tropical Rain Forest: Ecology and Management
(ed. by S. L. Sutton, T. C. Whitmore and A. C. Chadwick), pp. 93±
Beard, J.S. (1949) Natural vegetation of the Windward and Leeward 105. Special Publication no. 2 of the British Ecological Society.
Islands. Oxford Forestry Memoirs, 21, 1±192. Blackwell Scienti®c Publications, Oxford.
Beattie, A.J. & Oliver, I. (1994) Taxonomic minimalism. Trends in
Ecology and Evolution, 12, 488±490. Accepted 25 November 1999

# 2000 Blackwell Science Ltd, Ecological Entomology, 25, 460±466