construction of an explicit general theory. Consequently, the flaws in the warranting argument (point 3 does not follow from points 1 and 2) went largely unnoticed: the argument sounded plausible. The program spread like wildfire at the expense of culture-history in the 1960s and early 1970s, not because it outperformed culture-history (actually, it did not perform at all) but because its appearance coincided with the rapid expansion of the teaching of archaeology and anthropology in North American universities during this period. The dilemma posed by being scientific and being anthropological (in a 1960s’ sense) was insoluble because anthropology was not science (Dunnell 1982). Then, as now, anthropology was a matter of interpretation, not empirical testing. The persistence of processualism under these conditions is best characterized by what has been called the “Merlin syndrome” (Dunnell 1989). The key to success lay in asking questions in such a way that they could not be tested. This was accomplished by retaining English as the language of observation and common sense and ethnography in the place of theory. The only way questions posed in this fashion can be tested empirically requires a time machine. When common sense does the explaining, the descriptions explained must be common perception.

It was only a matter of time until this flaw was discovered. After all, archaeology was supposed to be a science, and sciences did involve empirical testing. In large measure the life of processualism was prolonged only by the innovative work of Michael Schiffer (1972, 1987), work that came to be known as “formation processes.” Schiffer’s central insight was to distinguish between what the archaeologist saw (archaeological context) and what the archaeologist wanted to talk about (systemic context) and to recognize that a long series of largely unexplored pattern-inducing processes separated the two. Understanding the variability introduced in archaeological remains by nontarget agents is critical for any effort to understand the archaeological record, giving this research value beyond its role in processualism. In that context, however, this understanding simply seemed to delay the inevitable by obfuscating the program’s logical inconsistencies with a haze of empirical busywork.

The shift to functional questions generated other lasting methodological and technical advances. Perhaps one of the most far-reaching changes, already alluded to, is the expansion of the notion of artifact to include the full range of manufactures, their by-products, and material incidentally modified by the human presence (Spaulding 1960b). Culture-historians, while recognizing the human attributes of a wide range of items, really only worried about manufactures—those that embodied style essential to their goals. This old conception is still embedded in terms such as diagnostics. A similar expansion took place at higher levels of organization, so that a greater variety of depositional situations became relevant; even the notion of “site” came under attack on empirical (Thomas 1974) and theoretical (Dunnell 1992b; Dunnell and Dancey 1983) grounds, although in this area CRM procedure has fireproofed the more conservative view.

Whereas these changes can be linked to functionalism, other positive elements of processualism can be linked to CRM, in particular the serious use of sampling. Culture-historians understood the sampling’s significance (e.g., Ford 1949; Kroeber 1916), but the character of the paradigm placed little burden on such matters, and they did not avail themselves of the quantitative tools needed. Gary Vescelius (1960) and Binford (1965) made strong theoretical cases for the usefulness of sampling to structure fieldwork, but it was the large amounts of space forced on archaeologists by cultural-resource surveys that finally made sampling part of the package. Although often characterized by its use of quantitative methods, beyond sampling, processualism’s quantification was window dressing designed to convince the unwary that archaeology was science. Most of the applications were used to create “kinds” that could be claimed to be natural (read cultural) rather than to describe, analyze, or explain variability (Luedtke 1995).

As argued elsewhere (Dunnell 1986a), what the processualist revolution represented was a shift from asking historical questions to asking