Fig 1.
Olduvai Gorge, Tanzania raw material sources.
a.) Naibor Soit is the source of quartzite from Olduvai Gorge. b) Modern exposure of basement basalts occur at locality KK and cobbles of basalt are available for procurement in the adjacent river bed. Basalt cobbles such as these are eroded from the volcanic highlands to the East of Olduvai Gorge.
Fig 2.
Oldowan production behaviors A-D.
a) The primary prerequisite knowledge for flake production is the identification of an appropriate platform. Oldowan Behavior A (OBA) is the identification of an acutely angled edge on the cobble. b) Oldowan Behavior B (OBB) requires placement of a platform above a scar made from a previously removed flake (such as an OBA flake, though the process assumes no particular order in behaviors). c) Oldowan Behavior C (OBC) is bifacial flaking, using the noncortical surface produced from previous flake scars as the platform for the flake. d) Oldowan Behavior D (OBD) requires an obliquely struck flake relative to the cobble’s edge (represented by the dotted white line). Striking in such a direction removes a flake along the perimeter of the core instead of removing a flake toward the center of the core (as in OBC flakes).
Table 1.
Measurements taken on whole flakes and corresponding definitions.
Fig 3.
Trimming the classification tree.
Each misclassification graph is constructed via random substitution of samples with empirically known classes into the trained algorithm. The X value represents the number of terminal nodes (ends of branches). The smallest value of X for the smallest misclassification value of Y is considered the most economical tree. The classification tree constructed in Fig 4 represents the most economic trees for both raw materials.
Fig 4.
Behavioral classification for Oldowan artifacts produced on Olduvai Gorge basalt and quartzite raw materials.
At each junction of the classification tree, if a given measurement is equal to or larger than the listed value, then the classification moves up the tree; if a given measurement is less than the listed value at the junction, then the classification moves down the tree until an ultimate behavioral classification is reached. For Olduvai Gorge quartzite and basalt, dorsal and platform cortex account for a large amount of the variation leading to classification. This relationship may or may not continue for other raw materials utilized at different Oldowan sites and is expected to vary in other technologies.
Table 2.
Probability of correct classification for individual flakes (see Fig 4).
Table 3.
Average number of flakes removed per core and average number of production behaviors per core for each raw material.
Table 4.
Distribution of assemblage variation for each measurement and raw material for experimentally produced flakes.
(SL: Size-Standardized Length; SW: Size-Standardized Width; ST: Size- Standardized Thickness; SMAX: Size-Standardized Maximum Dimension; SBT: Size-Standardized Bulbar Thickness; SAREA: Size-Standardized Area).
Fig 5.
Experimental vs. archaeological distribution of Oldowan stone tool manufacturing behaviors and associated classification error margins.
The first distribution (“Experimental Expectation”) represents the distribution of behaviors empirically determined from the controlled experimental replications. If all of the flakes produced at a site were produced in a least effort manner and were still present at the site when it was excavated, the archaeological distribution is expected to be similar to that of the experimental distribution. The second distribution is archaeological and classified via the classification algorithm Fig 4. However, each flake classified via this algorithm has a misclassification statistic associated with it (Table 2). Therefore, the third distribution (“Max OBD and OBB”) and the fourth distribution (“Max OBC and OBA”) demonstrate the potential extremes of the archaeological distribution given these misclassification rates. Finally, the last distribution (“Range of Variation”) combines the second, third, and fourth distributions into one combined statistic. For each bar on the graph, the dark colors represent the potential range of representation for that production behavior, given the potential misclassifications outlined in Table 2. The line inside the colored box represents the actual archaeological value, given classification via Fig 4. If no line is present in the colored box, then either the upper or lower extreme is the actual classified value. (Chi-square results: Expected vs. DK: Yates’ chi-square = 8.58, df = 1, Yates’ p = <0.01).
Table 5.
Archaeological samples of cores per site compared to the experimentally determined expectations of flakes per core (see Table 4) and the actual number of archaeological flakes recovered from the site.
Fig 6.
Schematic images of average basalt flakes from DK, Olduvai Gorge falling within and outside of expected variation.
These images represent idealized, schematic images of average measurements for Olduvai Gorge archaeological flakes from the site of DK that are made from basalt. The first column shows the ventral and profile views of the average DK flakes that fall within experimentally expected variation. The second column shows the ventral and profile views of the average DK flakes that fall outside of experimentally expected variation. For consistency and ease of comparison, the maximum dimension was placed at a 45-degree angle from the length measurement. All measurements are still size standardized, but are translated into inch units, such that if the average size-standardized length measurement is 1.0, this becomes 1.0 inch in this image. Similarly, average platform area is depicted as a square, such that platform width and thickness are equal and both are the square root of average platform area. Significant differences between outside-of-range and within-range assemblages are dotted lines, p-values are listed.
Fig 7.
Schematic images of average quartzite flakes from DK, Olduvai Gorge falling within and outside of expected variation.
These images represent idealized, schematic drawings of average measurements for Olduvai Gorge archaeological flakes from the site of DK that are made from quartzite. The first column shows the ventral and profile views of the average DK flakes that fall within experimentally expected variation. The second column shows the ventral and profile views of the average DK flakes that fall outside of experimentally expected variation. For consistency and ease of comparison, the maximum dimension was placed at a 45-degree angle from the length measurement. All measurements are still size standardized, but are translated into “inch” units, such that if the average size-standardized length measurement is 1.0, this becomes 1.0 inch in this image. Similarly, average platform area is depicted as a square, such that platform width and thickness are equal and both are the square root of average platform area. Significant differences between outside-of-range and within-range assemblages are dotted lines, p-values are listed.