Tropical Forest Community Ecology

Tropical Tritrophic Interactions 285

above (e.g., as outlined by Irschicket al. 2005)
should obviously consider this alternative. Each
investigative approach outlined below allows for
thorough tests of these and other alternative
hypotheses.

Natural history

While ecologists have made great advances in
recent decades by focusing on experimental
approaches and utilizing cutting edge molecu-
lar techniques, it is still natural history and
correlational data that form the basis of our
most important theories and new hypotheses.
Perhaps the greatest contemporary tropical
tritrophic dataset is the Janzen–Hallwachs plant–
caterpillar–parasitoid dataset, which provides
basic natural history data for lepidopteran host
plant affiliations and the parasitoid fauna that
they support (Janzen and Hallwachs 2002).These
data provide the raw materials for a thorough
approach to interesting questions in community
ecology and evolution and have already been used
to guide experimental (Sittenfeldet al. 2002) and
molecular (Hebertet al. 2004) approaches to test-
ing complex hypotheses. Similar databases are
being developed throughout the tropics (e.g., Dyer
and Gentry 2002, Novotnyet al. 2002). These
databases have been used to test and gener-
ate numerous hypotheses, including subsets of
the general hypotheses outlined above (see Lill
et al. 2002, Barbosa and Caldas 2004, Janzen
et al. 2005, Novotny and Basset 2005, Singer and
Stireman 2005, Stiremanet al. 2005).

Phylogeneti capproa ches

Advances in molecular systematics and compara-
tive methods have provided a relatively new set of
tools for ecologists to examine classic ecological
questions. Farrellet al. (1992) present a use-
ful outline for using phylogenies to test for tight
coevolution between plants and herbivores. There
are other research foci within tropical tritrophic
interactions that would benefit from a phyloge-
netic approach. For example, Heilet al. (2004)
used a phylogenetic approach to demonstrate
that herbivore-induced extrafloral nectar inAcacia

myrmecophytes is a plesiomorphic state while

constitutive flow of nectar is derived. Thus, plant

rewards became more readily available for spe-

cialized ants that kill herbivores, indicating that

a tritrophic interaction has driven recent coevolu-

tionary relationships between plants and insects;

this result is directly relevant to Hypotheses 2 and

3 above.

The phylogenetic approach is clearly necessary

for testing hypotheses about specialization at any

trophic level (e.g., Hypothesis 1 above) by utilizing

phylogenetically controlled comparisons between

specialist and generalist consumers and exam-

ining phylogenetic trends within taxa towards

narrower or broader diet breadths (outlined by

Irschicket al. 2005). This approach should be

combined with a concerted effort to document

ecological and functional specializations. First,

tropical ecologists must establish the actual diet

breadth of different species within a clade, despite

the large amount of descriptive work involved.

Second, a better integration of field and labora-

toryobservationswithinthesametaxonwillallow

for rigorous tests of how resource use is related to

performance, thus differentiating between ecolog-

ical and functional specialization (Irschicket al.

2005). For example, experiments could reveal

the relative performance of specialized consumers

when placed on more generalized diets or when

exposed to alternative food items.

Large-scale, long-term experiments

The temporal and spatial scales of many exper-

iments in the tropics are generally very small

(Schmitzet al.2000,HalajandWise2001).Larger

and longer experiments may cast light on the gen-

eralityof thecopiousstudiesdoneinsmallplotsfor

1 year or less. Eventually, meta-analyses will pro-

vide direct quantitative comparison. Since some

large-scale experiments are not possible, mensura-

tive experiments, such as the formation of islands

or fragments free of vertebrate predators (Crooks

and Soule 1999, Terborghet al. 2001), provide

a viable alternative. Experiments should also be

integrated in cohesive research programs that uti-

lize or acknowledge models, correlational data,

observational data, and phylogeny.