In 1970 Robert Mutch formally proposed the following hypothesis: Fire-dependent plant communities burn more readily than non-fire-dependent communities because natural selection has favored characteristics that make them more (Mutch 1970:1047). Although the paper in which the hypothesis was presented is widely cited in the ecological literature, there has been little effort at critical evaluation. The recent article by Buckley in this journal (Buckley 1983), while purporting to test Mutch's hypothesis, in fact does not and addresses two specific hypotheses of his own. Buckley's study involved dunefield vegetation in central Australia. The vegetation consists of dense grassland dominated by Triodia basedowii in the swales and an open shrub cover on the dune crests. The first specific hypothesis that Buckley tested was that fire is essential for the seedling regeneration of Triodia. His experiments showed that it was not. The second hypothesis was that shrub species found in the grasslands (subject to frequent and intense fires) survive fire better than or resprout more vigorously than shrub species characteristic of the dune ridges (burned by infrequent, low-intensity fires). The results showed that this was the case. The implication that fires are then responsible for excluding dunecrest shrubs from the swale grasslands is weakened somewhat by his statement that the dune shrubs regenerate from seed. It appears that fire may play a role in maintaining this vegetation gradient, but unfortunately Buckley does not separate the role of fire from that of the obvious edaphic gradient. In the absence of fire I suspect that a vegetation gradient would still remain. Mutch's hypothesis, however, deals rather specifically with flammability rather than reproduction or recovery after fire. Mutch discussed mainly chemical aspects of fuels, but the argument is easily extended to other traits that influence flammability (Rundel 1981). It seems apparent that communities that burn readily or frequently are more flammable than communities that do not. A lightning strike (or a dropped cigarette) in a grassland is much more likely to cause a fire than one in an adjacent mesic forest. The problem I see in Mutch's hypothesis is in attributing this to natural selection for increased flammability. Although the hypothesis is presented at the level of community evolution, the evolution of increased flammability by individual species seems to be the main thrust of his paper. Most fire-dependent communities are in fact made up of one or a few species responsible for the flammability (e.g., grasses, pines, or eucalypts) plus many species well adapted to the burning regime but lending little to fuel or flammability. The apparent evolution of numerous characteristics to resist fire or recover after fire by species in communities that burn frequently has lead to the suggestion that these species have also evolved characteristics that increase their flammability. The traits to resist or recover are numerous, but even some of these may have evolved for reasons other than recovery after fire (Gill 1981). I would contend that many characteristics of species that increase their flammability are merely incidental or secondary results of selection for other traits that increase individual fitness: e.g., herbivore deterrence, water-use efficiency, or nutrient retention; and not the result of selection for flammability. Along similar lines, Newman (1978) argued that although allelopathy may operate in nature, allelopathic interactions are best considered fortuitous outcomes of other plant characteristics. For example, the following scenario is quite straightforward. A plant with higher levels of a phenolic compound in its leaves suffers less herbivore damage and refore produces more offspring than its neighbors. In time the entire population has higher phenolic contents. The phenolics in the leaves cause them to decompose more slowly and the higher fuel accumulation leads to hotter and/or more frequent fires. There is now selective pressure on individuals (with the higher levels of phenolics) for improved mechanisms to resist fire or recover after the fire. In this example there is selection on individuals for increased herbivore resistance and fire resistance, but no selection for increased flammability. It is more difficult to imagine how increased flammability of an individual would increase its fitness. It would seem to require a simultaneous increase in resistance or recovery ability at the very least. High flammability may also be due to traits that are selectively neutral. For example, grasses by their general growth form tend to produce well-aerated litter. If a grass grows as scattered individuals in the understory of a forest it does little to increase flammability. In dense swards, however, grasses can form an excellent fuel. Although Williamson and Black (1981) suggest otherwise, there is probably little selective pressure on deposition patterns of trees. Differences between pine and oak leaf litter would seem to be an unavoidable consequence of differences between oak leaves and pine needles. This is not to say that trees whose leaves form heavy fuels would not have more