Abstract. -Large-scale evolutionary trends may result from driving forces or from passive diffusion in bounded spaces. Such trends are persistent directional changes in higher taxa spanning significant periods of geological time; examples include the frequently cited long-term trends in size, complexity, and fitness in life as a whole, as well as trends in lesser supraspecific taxa and trends in space. In a driven trend, the distribution mean increases on account ofa force (which may manifest itself as a bias in the direction of change) that acts on lineages throughout the space in which diversification occurs. In a passive system, no pervasive force or bias exists, but the mean increases because change in one direction is blocked by a boundary, or other inhomogeneity, in some limited region of the space. Two tests have been used to distinguish these trend mechanisms: (1) the test based on the behavior ofthe minimum; and (2) the ancestor-descendant test, based on comparisons in a random sample of ancestor-descendant pairs that lie far from any possible lower bound. For skewed distributions, a third test is introduced here: (3) the subclade test, based on the mean skewness of a sample of subclades drawn from the tail of a terminal distribution. With certain restrictions, a system is driven if the minimum increases, if increases significantly outnumber decreases among ancestor-descendant pairs, and if the mean skew of subclades is significantly positive. A passive mechanism is more difficult to demonstrate but is the more likely mechanism if decreases outnumber increases and if the mean skew of subclades is negative. Unlike the other tests, the subclade test requires no detailed phylogeny or paleontological time series, but only terminal (e.g., modern) distributions. Monte Carlo simulations ofthe diversification ofa clade are used to show how the subclade test works. In the empirical cases examined, the three tests gave concordant results, suggesting first, that they work, and second, that the passive and driven mechanisms may correspond to natural categories of causes of large-scale trends.