Compartments in insect-plant associations and their consequences for community structure

1. Compartmentation has been less explored than other forms of community structure. We assessed compartmentation of associations between insects and plants on a regional scale, and analysed some of its causes and consequences. The data set used was the host records of fruit flies (Diptera; Tephritidae) that breed in flowerheads of plants of the tribe Vernonieae (Asteraceae) in the Espinhaco Mountain range, Minas Gerais, Brazil. This data set was obtained with a consistent sampling protocol and is taxonomically fully resolved. 2. The binary association matrix had a total of 35 insect and 81 plant species. Most of the insects were specialized on plants of a single subtribe, genus or species group. Correspondence analysis showed that the association matrix is divided in six well-delimited compartments of insects specialized on subtribes or genera of plants. 3. Host dissimilarity among insects and insect dissimilarity among plants were expressed as Jaccard distances. Tests with a multi-response permutation procedure (MRPP) showed that both kinds of dissimilarities were higher among compartments than within them. 4. Monte Carlo randomizations were used to compare matrix parameters with values expected in the absence of compartments. In 4000 runs, the number of insect species that shared at least one host plant (ecological neighbours) was smaller than expected. Nevertheless, mean host similarity among insects, and the proportion of exclusive host plants used by each insect species did not differ from null model predictions. Host similarity of insects with their nearest neighbours in niche space was higher than expected by the null model. On the other hand, host similarity with farthest neighbours was lower than expected. 5. The observed compartmentation of insect/plant associations can be ascribed to the marked specialization of flowerhead tephritids, and allows the reduction of diffuse competition among insects. However, compartmentation did not decrease overall niche overlap among insects because reduction in number of neighbours is offset by increased overlap with species in the same compartment. Therefore, the pattern in this system cannot be derived from resource partitioning alone.