Fifty years ago, Ehrlich and Raven proposed that insect herbivores have driven much of plant speciation, particularly at tropical latitudes. There have been no explicit tests of their hypotheses. Indeed there were no proposed mechanisms either at the time or since by which herbivores might generate new plant species. Here we outline two main classes of mechanisms, prezygotic and postzygotic, with a number of scenarios in each by which herbivore-driven changes in host plant secondary chemistry might lead to new plant lineage production. The former apply mainly to a sympatric model of speciation while the latter apply to a parapatric or allopatric model. Our review suggests that the steps of each mechanism are known to occur individually in many different systems, but no scenario has been thoroughly investigated in any one system. Nevertheless, studies of Dalechampia and its herbivores and pollinators, and patterns of defense tradeoffs in trees on different soil types in the Peruvian Amazon provide evidence consistent with the original hypotheses of Ehrlich and Raven. For herbivores to drive sympatric speciation, our findings suggest that interactions with both their herbivores and their pollinators should be considered. In contrast, herbivores may drive speciation allopatrically without any influence by pollinators. Finally, there is evidence that these mechanisms are more likely to occur at low latitudes and thus more likely to produce new species in the tropics. The mechanisms we outline provide a predictive framework for further study of the general role that herbivores play in diversification of their host plants.

Many insect herbivores build shelters on plants, which are then colonized by other arthropod species. To understand the impacts of such ecosystem engineering on associated species, the contributions of ecosystem engineer and host-plant identities must be understood. We investigated these contingencies at the patch scale using two species of leaf-tying caterpillars, which vary in size and tie construction mode, on eight species of oak (Quercus) trees, which vary in leaf size and leaf chemistry. We created three types of artificial leaf ties by clipping together pairs of adjacent leaves using metal hair clips. We left the first type of leaf tie empty while adding individuals of the leaf-tying caterpillars of either Pseudotelphusa quercinigracella or Psilocorsis cryptolechiella to the other two. We also created a control treatment of untied leaves by affixing clips to single leaves. Leaf ties increased occupancy in the early season and arthropod alpha diversity throughout the experiment, on average fourfold. Furthermore, the presence of leaf ties increased arthropod species density on average three times and abundance 10-35 times, depending on the plant species. The mean phenolic content of the leaves of each oak species was positively correlated with the leaf-tie effect on abundance and negatively correlated with the leaf-tie effect on species diversity. Species diversity, but not abundance, was affected by the identity of the tie-maker. Arthropod species composition differed between untied leaves and artificial leaf ties, and between ties made by the two leaf-tier species. Our results demonstrate that the presence of leaf ties adds to habitat diversity within the oak-herbivore system, not only by creating a new kind of microhabitat (the leaf tie) within trees, but also by exacerbating differences among the eight oak species in apparent habitat quality. The identity of the leaf-tying caterpillar adds to this heterogeneity by creating leaf ties of different size, thus influencing subsequent colonization by other leaf-tying caterpillars of different sizes.

Almost 40 years ago, Terry L. Erwin published a seemingly audacious proposition: There may be as many as 30 million species of insects in the world. Here, we translate Erwin's verbal argument into a diversity‐ratio model—the Erwin Equation of Biodiversity—and discuss how it has inspired other biodiversity estimates. We categorize, describe the assumptions for, and summarize the most commonly used methods for calculating estimates of global biodiversity. Subsequent diversity‐ratio extrapolations have incorporated parameters representing empirical insect specialization ratios, and how insect specialization changes at different spatial scales. Other approaches include macroecological diversity models and diversity curves. For many insect groups with poorly known taxonomies, diversity estimates are based on the opinions of taxonomic experts. We illustrate our current understanding of insect diversity by focusing on the six most speciose insect orders worldwide. For each order, we compiled estimates of the (a) maximum estimated number of species, (b) minimum estimated number of species, and (c) number of currently described species. By integrating these approaches and considering new information, we believe an estimate of 5.5 million species of insects in the world is much too low. New molecular methodologies (e.g., metabarcoding and NGS studies) are revealing daunting numbers of cryptic and previously undescribed species, at the same time increasing our precision but also uncertainty about present estimates. Not until technologies advance and sampling become more comprehensive, especially of tropical biotas, will we be able to make robust estimates of the total number of insect species on Earth.Abstract in Spanish is available with online material.

1. Shelter-building is widespread in the animal world and such shelters often influence the success of their builders. Shelters built by caterpillars influence the likelihood of attacks by natural enemies, but how particular shelter traits influence caterpillar survival is not known. Furthermore, the differential effects of certain shelter traits on some natural enemies, such as predators, may lead to 'enemy-free space' for other natural enemies (parasitoids). The parasitoid enemy-free space hypothesis has not been directly tested for shelter-building caterpillars. 2. To understand how shelter traits influence caterpillar survival, shelter traits, predation and parasitism were measured simultaneously for 24 caterpillar morphospecies (1,465 caterpillars) in a tropical dry forest and analysed in a phylogenetic context. 3. Shelter type, shelter openness and whether shelters accumulated frass had different amounts of phylogenetic signal, with frass accumulation displaying the most and shelter openness the least. 4. All three traits affected the frequency with which caterpillar species experienced predation. Predation was elevated in two shelter types (leaf folds and leaf rolls) compared to cut-and-fold shelters. Combinations of shelter openness and frass accumulation also affected predation, with closed frass-free shelters having the lowest predation and closed frass-filled shelters having the highest. 5. Parasitism was not affected by shelter traits but was strongly correlated with evolutionary history and negatively correlated with predation. 6. These results confirm a trade-off between predation and parasitism and demonstrate that predation can be more frequent than parasitism. Different shelter types result in different amounts of predation. These defensive shelter traits and their effectiveness also vary phylogenetically. Together, our results suggest that predation and parasitism determine the success of shelter-building caterpillars, and that success is a function of the specific shelter they construct. More generally, our results demonstrate the importance of considering the effects of defensive traits on both predators and parasitoids when investigating interactions between herbivores and natural enemies.