My post lead Martin to observe:
This sounds like complete bollocks to me. If pre-scientific humans had been technological optimisers, then all arrowheads globally would fall into a small number of standardised task-specific types. Culture does not change through time in a manner like that of biology, because cultural selection is largely irrational. Most of what people do is white noise from an evolutionary perspective. Null mutations.
Today I paint my face red for the ritual dance. Twenty years from now my daughter will paint hers blue. It is deeply significant from a cultural point of view, but completely irrelevant from an evolutionary one. Today I will make arrowheads 15% longer than those I made last year. I have no idea what the practical consequences of this decision, if any, will be.
Kambiz responded by, among other things, pointing out the difference between functional and symbolic traits and this is actually a crucial distinction for evolutionary archaeologists. To begin with, evolutionary archaeologist, in keeping with contemporary evolutionary theory, consider behavior to be subject to selection (see West-Eberhard’s Developmental Plasticity and Evolution for examples) and human behavior is no different. In this view, material culture is seen as an expression of phenotypic variability and change is seen as affecting changes in trait frequencies (much like some argue that evolution is a change in gene frequencies) rather than transforming the traits themselves. In some ways this is analogous to the debate over group selection (or maybe the debate about whether higher level taxonomic groups are “real”), evolutionary archaeologists argue that in focusing on higher level groups such as tribes or chiefdoms we are being unnecessarily typological and looking at the wrong level for explanations of change.
Whereas selection in organisms is defined based on differential reproductive success, for material traits the concept of replicative success is used. In those instances where the replicative success of the trait affects the the reproductive success of the bearer the trait is considered a functional trait, otherwise it is considered stylistic. A good example of the latter is the example of Martin’s that I quoted above (obviously, Martin’s example could be impacted by sexual selection, which I will ignore for the purpose of this discussion). Stochastic traits are, as Martin points out, stochastic and, it has been argued, can be modeled by, say, Markov processes. There has also been some linkage of stylistic traits with neutral theory, such as what you see in Kimura. Lyman et al would agree with Martin concerning his statements on optimality. Certainly, O’Brien has expressed disagreement with optimal foraging theory, and in one paper he says:
The drive towards measuring fitness values of any trait often is based erroneously on an assumption of optimality – over the long term only the fittest survive. Overlooked is that fact that “it is sufficient to be superior and not at all necessary to be perfect” (Mayr 1982:589). One might suggest that it is sufficient simply to be adequate.
With that in mind, let us take another look at the paper in question. As Kambiz pointed out, in the post linked to above, Lyman et al studies over 1,000 projectile points from three widely separated sites in North America (Verkamp Shelter, Gatecliff Shelter, and Mummy Cave). The time range spans some 9,000 years. Based on measures of a number of attributes, Lyman et al calculated the corrected coefficient of variation and compared them over time via t-tests. What they were interested in was comparing the variability in dart points (used with atlatls) to points used on arrows – specifically as the bow and arrow was being introduced and in the periods following that introduction:
We are interested in the narrow question of how variation relates to innovation, in this case, the magnitude of variation in attributes of projectile points before, during, and after the appearance of the bow and arrow.
Here an interesting article, published 2005 in American Antiquity, by Michael Schiffer comes into play (The Devil is in the Details: The Cascade Model of Invention Processes, 70(3):485-502). In that paper Schiffer looked at the the generation of material novelty. This is the important part:
In a nutshell, the cascade model posits that, during a CTS’s [complex technological system - afarensis] development, emergent performance problems – recognized by people as shortcomings in that technologies constituent interactions [emphasis in the original - afarensis] – stimulate sequential spurts of innovation. As adopted inventions solve one problem, people encounter new and often unanticipated performance problems, which stimulate more inventive spurts, and so on.
Schiffer goes on to talk about cascades in largely Darwinian terms (and even discusses the introduction of the bow and arrow as one example). The point here is that, as the bow was introduced, new projectile points were needed – what works on the atlatl does not work on an arrow. Consequently, knappers started experimenting (i.e. generating variability) with individual attributes that make up the points (length, width, thickness, hafting type, etc.) to come up with a workable arrow point and selection then winnowed out those trait states that ineffective. In point of fact this is what Lyman et al saw in their data:
The data indicate that the magnitude of variation in attributes of projectile points responds in predictable ways to changes in the selective environment. The change in selective environment comprised the shift from the atlatl and dart to the bow and arrow as the dominant weapon delivery system. Variation in dart-point attributes tends to be stable or decrease in early parts of all sequences. Total variation (ΣCCVs) increases coincident with the appearance of the arrow, as dart points are redesigned (likely through trial and error) to perform effectively with the new weapon system.
The change in selective environment resulted in considerable variation in arrow points. The CCVs of many attributes of arrow points are relatively large in the early strata in which arrow points first appear. We found indications of our prediction that high variation in arrow points would decrease as less-effective variants were selected against though our results are not universal on this likely because the recent portions of two evolutionary sequences (Verkamp, Gatecliff) are missing. Given that the overall patterns of variation we predicted are represented among the point assemblages we consider, we suspect that those patterns may be typical of the evolution of projectile points in general and perhaps other manifestations of material culture when innovative technologies are introduced.
Okay, so, some traits survived out of the hodge-podge of variability, while others did not. What governs this? Why were some traits selected over others? There has been some research on the issue. For example, width and thickness impact penetration (no obscene jokes in the comments, please), while width and length influence cutting edge and wound severity, and length and weight influence flight dynamics. Finally, hafting has an impact as well.