Gould, Trends as Changes in Variance: A New Slant on Progress and Directionality in Evolution

320 JOURNAL OF PALEONTOLOGY, V. 62, NO. 3. 1988

FIGURE 1 -- Increase in size of means and extreme values within a clade as a function of origin near a lower limit in size. From Stanley (1973).

coelacanths and monoplacophorans are pushed into the deep seas (for we must have losers as well as winners), while Aysheaia rises from sea to land to yield the modern Onychophora.

Only one major interpretive (as opposed to methodological) reform has graced the study of trends in our century: the preference for orthoselection over various forms of near mysticism and special pleading, and the resulting emphasis upon opportunism and fluctuations in rate. But the fundamental, topological conviction had held firm-that trends are anagenetic sequences expressed as entities exhibiting net movement in a specified direction.

A DIFFERENT TOPOLOGY

If macroevolution has no independent status and must be viewed, in uniformitarian perspective, as an extrapolation into geological vastness of the apparatus used to interpret ecological moments in modern populations, then the anagenetic view must prevail. Shifts in gene frequencies within populations dominate the ecological moment, and geology surely provides enough scope to make a trend by extending these effects into deep time. A different view of trends therefore requires the replacement of this extrapolationist vision-a model that, among its other doleful effects, implies a subservient status for paleontology among evolutionary disciplines (as guardians of a descriptive pageant fully explained by principles from the more prestigious world of the observable now).

But if we view species as stable entities for most of their geological existence, not as temporary names for transient states in the great and continuous flux of life, then we must interpret trends differently. If species are not generally changing, then macroevolution cannot be an extrapolation of anagenesis within populations. Macroevolution-including the central phenomenon of trends -must be conceptualized as the differential success of species. If microevolution results from the sorting of organisms within populations, then macroevolution occurs by sorting of species within clades (Vrba and Gould, 1986). Nature operates an entity making and breaking machine in the processes of speciation and extinction. If differential birth and death produce anagenesis in populations, then differential speciation and extinction forge trends within clades. To understand trends, we must adopt what Eldredge (1979) called a taxic, rather than a transformational, view of macroevolution.

But why should a differential production and death of species produce a trend in anything other than number of entities? Why should the differential success of species yield a directional movement of traits?

Consider an example of a common situation, first generalized by Stanley in his seminal article (1973) on reinterpreting Cope's rule. Suppose that the founding species of a clade originates near one boundary of its potential range (near the lower limit of potential size, or simply near shore for a marine Bauplan), and that the number of descendant species within the clade then increases steadily and substantially. Suppose that the modal class never changes-that is, the most common size or geographic position of later species remains at the value of the clade's founding member. Yet the location of the founder at an edge of the potential range virtually guarantees that new species will be differentially added in the direction of greater available space-larger body sizes, or deeper water in our examples above. Stanley (1973) both recognized this principle as a generality and empirically documented the right-skewed nature of histograms for body size within clades. His simulation (Figure 1, from Stanley, 1973) illustrates my central point both graphically and dramatically.

In such cases, many of our usual measures will yield an apparent trend anagenetically expressed. If we make a deep epistemological error and focus on extreme values through time as though they were "things" in themselves rather than tails of a range in variation, we might interpret the basic phenomenon-increase in number of species, with asymmetry about a starting point near one end of a constrained range-as a directional trend (see Figure 1). If we took a more "sophisticated" view, and represented the clade by a measure of central tendency, we might still fall into the same error. Mean values will also increase (though less dramatically than extremes) towards the open end of the range, simply because distant excursions from the original value can proceed in only one direction. If we increase our sophistication and calculate a median instead of a mean (recognizing the old principle that one Rockefeller may unfairly balance a thousand paupers in determining a mean income), we may still note an increase through time because the open end of the range provides more space for new items in general, not only a larger vista for distant excursions from the founding value. Thus, we face the peculiar situation of a clade with an unchanging modal value through time, but a measured increase anagenetically depicted in extreme values, means, and medians.

Should we view such a situation as a "trend" towards larger sizes or deeper waters within the clade? I can imagine certain questions that would legitimately engender such an interpretation. But surely, the unchanging modal value makes a powerful case for an overriding stability of the most common class of species, and this fact should be acknowledged in any interpretation. More importantly, whatever myopic measure we choose to employ, the central phenomenon of this clade's history is an increase in variance mediated by the entity-making machine, not an anagenetic march anywhere. Our common anagenetic

142