That larger areas will typically host more diverse ecological assemblages than small ones has been regarded as one of the few fundamental ‘laws’ in ecology. Yet, area may affect not only species diversity, but also the trophic structure of the local ecological assemblage. In this context, recent theory on trophic island biogeography offers two clear-cut predictions: that the slope of the species–area relationship should increase with trophic rank, and that food chain length (i.e. the number of trophic levels) should increase with area. These predictions have rarely been verified in terrestrial systems. To offer a stringent test of key theory, we focused on local food chains consisting of trophic specialists: plants, lepidopteran herbivores, and their primary and secondary parasitoids. For each of these four trophic levels, we surveyed species richness across a set of 20 off-shore continental islands spanning a hundred-fold range in size. We then tested three specific hypotheses: that species richness is affected by island size, that the slope of the species–area curve is related to trophic rank, and that such differences in slope translate into variation in food chain length with island size. Consistent with these predictions, estimates of the species–area slope steepened from plants through herbivores and primary parasitoids to secondary parasitoids. As a result of the elevated sensitivity of top consumers to island size, food chain length decreased from large to small islands. Since island size did not detectably affect the ratio between generalists and specialists among either herbivores (polyphages vs oligophages) or parasitoids (idiobionts vs koinobionts), the patterns observed seemed more reflective of changes in the overall number of nodes and levels in local food webs than of changes in their linking structure. Overall, our results support the trophic-level hypothesis of island biogeography. Per extension, they suggest that landscape modification may imperil food web integrity and vital biotic interactions.