Most investigations of insect gas exchange patterns and the hypotheses proposed to account for their evolution have been based either on small-scale, manipulative experiments, or comparisons of a few closely related species. Despite their potential utility, no explicit, phylogeny-based, broad-scale comparative studies of the evolution of gas exchange in insects have been undertaken. This may be due partly to the preponderance of information for the endopterygotes, and its scarcity for the apterygotes and exopterygotes. Here we undertake such a broad-scale study. Information on gas exchange patterns for the large majority of insects examined to date (eight orders, 99 species) is compiled, and new information on 19 exemplar species from a further ten orders, not previously represented in the literature (Archaeognatha, Zygentoma, Ephemeroptera, Odonata, Mantodea, Mantophasmatodea, Phasmatodea, Dermaptera, Neuroptera, Trichoptera), is provided. These data are then used in a formal, phylogeny-based parsimony analysis of the evolution of gas exchange patterns at the order level. Cyclic gas exchange is likely to be the ancestral gas exchange pattern at rest (recognizing that active individuals typically show continuous gas exchange), and discontinuous gas exchange probably originated independently a minimum of five times in the Insecta.