Coevolutionary studies on plants and plant-feeding insects have significantly improved our understanding of the role of niche shifts in the generation of new species. Evolving plant lineages essentially constitute moving islands and archipelagoes in resource space, and host shifts by insects are usually preceded by colonizations of novel resources. Critical to hypotheses concerning ecological speciation is what happens immediately before and after colonization attempts: if an available plant is too similar to the current host(s), it simply will be incorporated into the existing diet, but if it is too different, it will not be colonized in the first place. It thus seems that the probability of speciation is maximized when alternative hosts are at an ‘intermediate’ distance in resource space. In this review, I wish to highlight the possibility that resource similarity and, thus, the definition of ‘intermediate’, are subjective concepts that depend on the herbivore lineage's tolerance to dietary variation. This subjectivity of similarity means that changes in tolerance can either decrease or increase speciation probabilities depending on the distribution of plants in resource space: insect lineages with narrow tolerances are likely to speciate by ‘island-hopping’ on young, species-rich plant groups, whereas more generalized lineages could speciate by shifting among resource archipelagoes formed by higher plant taxa. Repeated and convergent origins of traits known to broaden or to restrict host-plant use in multiple different insect groups provide opportunities for studying how tolerance and resource heterogeneity may interact to determine speciation rates.