were of no chronological value because there were no differences. Intuitive recognition of differences of a more subtle sort (Uhle 1907) was easily dismissed (e.g., Kroeber 1909). However, once the record was conceived in quantitative terms, as of varying frequencies, not differing kinds, and combined with historical types (or seriation), major changes took place that forever altered the course of U.S. archaeology.

One of these changes was the use of stratigraphy. Now not only could it provide chronology where before it had little to say in the United States, it could also be used to guide excavation and generate chronologically valuable data. This led rather quickly to “metric stratigraphy” and “arbitrary levels,” approaches that have been much maligned lately because the early cautions about their use by nels nelson (1918) have been ignored.

More important was the development of seriation, the first truly archaeological method and the first serious quantification undertaken by archaeologists. Although sir w. m. flinders petrie is considered the “father of seriation” (e.g., Heizer 1958), his conception of using kind to derive chronology was an extension of the strong European tradition of “kind tells time” begun by jens j. a. worsaae and christian j. thomsen and was essentialist in character (Jensen 1975). The U.S. approach, by contrast, relied on quantifying variability. Its origin can be traced to Alfred L. Kroeber and his students and colleagues working the American Southwest (e.g., Kroeber 1916; Nelson 1918; Spier 1917), and it took the form now generally known as “frequency seriation” (Cowgill 1972; Dunnell 1970, 2000). Kroeber (1916) first used the method in its most elemental form to chronologically order otherwise undated and undatable surface ceramic assemblages in the vicinity of Zuni Pueblo. By working backward from modern, recently remembered but abandoned, and completely unknown ruins with varying degrees of preservation, he noticed that the frequencies of different pottery “types” changed in an orderly fashion, with modern types decreasing and types present only anciently increasing with apparent age. Characteristically, he appreciated the significance of sampling and sample size and recognized that superposition could be used to test the order. His student Leslie Spier, working nearby, did just that, introducing most of the modern elements to the method, including normalizing samples with the use of percentages and recognizing the unimodal (“normal”) character of frequency distributions (1917). Although further refinements were made, particularly with respect to conditions of application (Ford in Phillips, Ford and Griffin 1951), its use in this form spread throughout the Americas. By the 1950s chronologies nearly everywhere were based on seriation (e.g., Evans and Meggars 1957; Ford 1938, 1949; Mayer-Oakes 1955).

All three of the methodological innovations central to culture-history entered into seriation. First, the types had to be conceived as tools of measurement, not recovered “ideas” in the minds of makers or simply arbitrary conventions for description (Ford 1954, contra Spaulding 1953). Furthermore, as Holmes had dimly envisioned, the attributes used to create types had to be what came to be called “stylistic” or “historical” rather than functional or haphazard confections. Critically, culture-historians employed an empirical test for the historicity of their types—the test of historical significance (Krieger 1944), which required temporal and spatial contiguity. No proposed type could be established unless these conditions were met, and only these kinds of types displayed the unimodal distributions required by the method. Treating types as tools stemmed from seeing the past not as a string of differences but as continuously changing frequencies (materialism). The only way to describe variability (as opposed to difference) was through quantification.

Superposition, an important tool in its own right that was also tied to these three innovations, played a minor role in large-scale culture-history and in seriation. In the latter it was used to determine the direction of time through a seriation (seriation provided only an order, not a direction) to test the results of seriation when some of the seriated assemblages were superimposed and to help select attributes and define types. Correlations (i.e., extension of stratigraphy beyond a single column) also depended