| [Thread Prev] | [Thread Next] | [Thread Index] | [Date Prev] | [Date Next] | [Date Index] |
"Shaw effect": I believe using this in MacLeod's sense would be inappropriate, although he is right that there is an obvious connection between quantitative biostratigraphy in general, confidence intervals in particular, and the Signor-Lipps effect. Classical graphic correlation in the tradition of Shaw is concerned with correlating appearance datums, not with establishing the error on any one datum in any one section. Furthermore, it deals with _first_ appearances in addition to last appearances, unlike the whole Signor-Lipps problem. There is no need for nomenclatural debates; the terminology should be clear to everyone. "Barren interval": I and others have argued endlessly that the "gap" in the Hell Creek section is trivial, regardless of whether it can be modeled using simple confidence intervals (would someone please try?) or needs to be explained by invoking some change in taphonomic modes (as Doug implies). One meter is nothing in a rapidly deposited, "predominantly fluviatile" terrestrial formation where most of the vertebrates in the upper part of the section occur in deeply incised channels (Archibald 1982). As for the "thousands of fossils" (reference?), this must refer to either 1) non-dinosaurian vertebrates, or 2) small, isolated dinosaur elements such as centimeter-sized teeth that occur as reworked material _above_ the boundary. What we really want to know about is the occurrence of articulated, clearly non-reworked dinosaurs, of which there are surely far fewer than "thousands" in the Cretaceous as a whole, much less in the Hell Creek. MacLeod seems to be convinced that the 95% confidence interval would fall within the uppermost meter of the formation; let's see him (or someone) actually do some calculations. Maybe the Berkeley or Milwaukee people can provide some data. Ignoring Signor-Lipps: biostratigraphy - like all empirical sciences - requires either parsimony or null hypothesis testing. The Signor-Lipps effect is just an name for one aspect of the relevant null hypothesis, i.e., that observed taxonomic ranges are shorter than real ranges. Ignoring Signor-Lipps and then equating real and observed ranges is _not_ null hypothesis testing; there is simply no way to falsify such a claim. As for parsimony, a catastrophic extinction would predict a large number of observed non-overlaps (disjunctions) between taxa above and below the boundary. In fact, a relatively huge number of disjunctions for mammals are strongly implied by published data (if you want some numbers, I'll provide them). The only parsimonious explanation is that such an extinction event did occur.
Partial index: