The first step in a cognitive analysis of any prehistoric technology is the identification of attributes that carry cognitive implications, and this selection depends to a large degree on the theoretical grounding of the analysis [3,4]. The current analysis is grounded in the discipline of cognitive neuroscience, which is an interdisciplinary approach that combines the methods and concepts of cognitive psychology with the evidence from neuroscience (neuroimaging and clinical neuropsychology). As this approach has generated an immense literature over the past quarter century, it is necessary to narrow the focus. Here we will rely on two sets of studies: first, neuroimaging research focusing on human and non-human tool use, and second, a cognitive model of expert performance.

Dietrich Stout and colleagues [5–9], and Guy Orban and colleagues [10,11] have used functional magnetic resonance imaging (fMRI) to identify brain activation patterns of humans and macaques using tools. Together these studies have identified a complex pattern of activation that includes the frontal, parietal, and temporal lobes of the brain, and which is centered on the inferior portion of the parietal lobes known as the supra-marginal gyrus. This ‘anthropoid object manipulation network’ (AOMN; Wynn’s terminology) includes circuitry governing visually guided reaching, sequential organization of action, and also mirror neurons that fire when action is observed. Stout’s research has confirmed that the AOMN is the primary system underpinning stone knapping, and he and his colleagues have begun to identify additional neural resources that came into play with more advanced knapping techniques [7].

Describing the neural substrate of tool use is important but it does not capture the richness and subtlety of tool use in natural settings. Cognitive psychology provides an appropriate and well-researched model known as ‘expert performance’ or expertise. This is the kind of thinking that experts use in their particular domains of performance [12,13]. Expert cognition has several salient features: rapid problem assessment, virtually error free execution, ability to shift attention and return to task without loss of information, rapid learning of new routines, and flexible responses to problems that arise during performance. Most famously studied in chess masters, this kind of thinking is clearly the basis of expert tool use as well [14–17]. Expertise relies on long-term memory (LTM), especially a huge body of well-learned procedures. As a novice learns procedures he or she consolidates them into coherent ‘chunks’ of information, and attaches a cue to each chunk, such as a word or an image [13–15]. Hearing, seeing, or accessing the cue will give immediate access to the entire chunk of information. The novice accesses a cue in working memory, which is the amount of information one can hold in attention and process in the presence of interference [18–20]. Expertise is thus a cognitive ability that engages both long-term memory and working memory.

In sum, tool use, including stone knapping, is an expert cognitive system that relies on well-learned long-term procedural memories, and the ability to access them quickly and deploy them appropriately. Expert cognition has definite advantages over the effortful problem solving performed in the active attention of working memory [18]. Because the procedures are well-learned they can be activated immediately through a simple visual or even tactile cue. It is a flexible kind of thinking, but it is not inherently innovative [15]. On the neural level this cognitive system is governed by an object manipulation network that initially evolved with anthropoid primates, but which has been expanded and refined over the course of hominin evolution [8,10].

Here we use the reduction sequence reconstructed for cleavers from the Acheulian site of Gesher Benot Ya‘aqov (GBY) to describe the chaîne operatoire used by the knappers. We then use the results of this analysis as the basis for an interpretation of the underlying cognitive strategy.

The detailed analysis of the cleaver component of the GBY assemblages has provided several insights regarding the modes by which the distal morphology and configuration were achieved. There are three different modes for the design of the working edge. The first is the “classical” mode, in which the distal dorsal configuration is achieved by exploiting the scar pattern of the debitage surface of a giant core (Fig 1). The removal of this predetermined flake leaves a remnant of a large negative scar on the distal end of the flake. The intersection between the ventral face and the remnant of the scar on the dorsal face forms the desired working edge morphology. Such production mode is commonly characterized by a slight concavity of the dorsal distal surface forming the working edge, stemming from the morphology of the flake scar.

The second mode consists of cleavers with an unmodified Kombewa surface as their dorsal face. This, in fact, is a minimally modified, and at times completely unmodified, Kombewa flake (Fig 2). The distal working edge is formed by the intersection of the two surfaces, which are the two ventral surfaces of the Kombewa flake. When post detachment modifications do occur, they are commonly restricted to the lateral edges and proximal end, shaping the striking platform area and thinning the bulb of percussion of the tool, similarly to that of the previous mode.

The third mode consists the post detachment shaping of the blank. In this mode invasive retouch is used to modify the dorsal face of the flake (Figs 3 and 4A). The retouch is performed so that the flake scars will cover most of the surface of the tool, leaving the original surface on the distal end delineated (Fig 4B). This delineated surface is always wither flat or convex, but never concave as in the first mode. The distal edge always remains unmodified (Fig 4C) so that the working edge is formed by the intersection of the delineated dorsal and unmodified ventral surfaces (Fig 4D). In many cases the unretouched delineated surface formed a triangular shape (Fig 4E), in which the base is formed by the straight working edge and its opposing vertex, in the middle part of the cleaver, usually at the thickest point.

In the attempt to gain better insight into the technological characteristics of cleaver production we have examined in detail the modification modes of the GBY cleavers. The recorded observations included the blank type on which the tool was modified (flake, Kombewa and Possibly Kombewa) and the mode by which the cleavers dorsal-distal configuration was attained.

Table 1 presents the frequencies of each dorsal-distal end across the different blank types. The observations for each of the tools are presented in the supporting information (S1 Table).

Of the entire studied sample of 168 cleavers, 43 (25.6%) could not be classified to either blank type or modification mode due to continuous weathering. Of the remaining sample, the majority of tools were modified on regular flakes (n = 84, 67.2%). The most common modification mode is that formed by a prior scar on the debitage surface of the core (n = 50, 40%). Naturally, this mode is limited to tools that were modified on regular blanks, as Kombewa (and possibly Kombewa) blanks lack scars of previous removals. The rest of the tools are divided between a group whose distal ends consists of unmodified surface delineated by intensive surface flaking (n = 47, 37.6%) and cleavers lacking any modification of their distal ends (n = 28, 22.4%). Examination of cleavers with unmodified distal ends shows that even though they are produced on all types of blanks, the Kombewa and possibly Kombewa categories consist of almost 93% of the blanks used for this type of distal end configuration. In other words, the probability of having a non-modified distal edge is significantly higher for items produced on Kombewa (or possibly Kombewa) flakes.

Cleavers with distal ends that were modified by delineating a residual surface occur in all types of flakes and in relatively high frequencies. The most common blank of this aspect is a simple flake, followed in descending order by Possibly Kombewa and Kombewa flakes. When considering the origin of the delineated unmodified surface there are two possibilities. The first is that this surface is the original natural flat surface of a basalt slab. The second possibility is that this surface is in fact a remnant of an old ventral face. Our detailed analysis of these items supports the second option due to the slight convexity of this surface on most of the tools in this category. Therefore, the implication of this interpretation is that those blanks classified as simple flakes are in fact Kombewa flakes (or Possibly Kombewa) that were not identified as such due to intensive subsequent modification.

The frequency of Kombewa flakes in the GBY bifacial assemblages has been previously described several times. We have previously reported high values of Kombewa blanks associated with the production of bifaces [21] and lower values in the final report on the lithic assemblages of GBY [25]. We have explained the differences as being the result of two factors: the deterioration and continuous weathering of some of the bifaces on the one hand, and on the other hand a change in our analytical criteria for identifying Kombewa flakes. One criterion that was added over the years is that for an example to be classified as Kombewa flake it had to possess two visible striking platforms. While this criterion rendered the classification of items as Kombewa certain, it also drastically reduced their frequency in the assemblage. Clearly, the results of these analytical differences influenced the frequency of the Possibly Kombewa artifacts. Our current analysis requires only that items present a clear indication of having two ventral faces but not necessarily two striking platforms. Our renewed detailed analysis of the cleavers’ distal ends provides support for the classification of additional items into the Kombewa/Possibly Kombewa categories. If our interpretation is correct, then almost 40% (n = 32) of the cleavers classified as modified on regular flakes were actually modified on flakes produced by the Kombewa technology. This contribution augments the frequency of tools derived of Kombewa technology to a total of 60% (n = 75) of the cleaver assemblage (excluding the indeterminate category). Furthermore, this newly acquired data contributes meaningful new insight to the previously ambiguous nature of the Possible Kombewa category.

With this perspective in hand, it is possible to assess the production procedure of the GBY large-flake cleavers. Several features of the procedure are especially informative:

The procedure consisted of three distinct stages.

This reconstructed procedure has several provocative behavioral components (i.e., behavioral components that are implied by the procedure):

For each alternative there was an established sequence of steps. This was not simply the iterative application of a knapping gesture. One kind of knapping preceded a different kind of knapping in a set sequence.

Summing up the different stages of the reduction process of cleavers we are able now to identify several distinct stages concerning the planning capabilities of the GBY hominins. The first is applicable to the production of both handaxes and cleavers and is expressed in the raw material and percussors acquisition phases of the chaîne opératoire aimed at biface production. The second refers to the core reduction methods aimed at producing the desired large flakes. Here, however, there is a major difference between handaxes and cleavers. The procedure used in handaxe production allows for greater flexibility late in the process. In contrast, the procedure used in cleaver production requires a decision early in the process about the final configuration of the working edge. This decision then entails a number of specific steps leading to the final product. From the reconstruction of each specific core reduction method at GBY it appears that flakes with appropriate morphologies, dimensions and distal characteristics could only be produced by using the Levallois or the Kombewa methods. This means that the decision to produce a cleaver must have been taken as early as the slab fragmentation and core design stage of the chaîne opératoire.

An additional and interesting aspect of our study relates to the dorsal distal modification by delineation of a residual surface. While this appears to be a common modification mode at GBY, it was not a necessary one, as there is a significant number of tools lacking any dorsal distal modification. This observation is even further highlighted given our suggestion that this modification mode was always applied to Kombewa flakes. Clearly, unmodified (or slightly modified) Kombewa flakes provided a similar functionality with regards to the distal morphology that allowed the necessary mass penetration and shearing capacities. Thus, it remains an open question why some Kombewa flakes underwent this additional modification while others did not. We suggest, however, that this selection could be related to stylistic or cultural preferences.

Our results and conclusions shed light and provide additional insights concerning the issue of hominin abilities. It is evident that the GBY hominins, as documented elsewhere in the Acheulian record, were capable of planning, producing and exploiting a wide array of convex surfaces. These abilities are recorded in the production/modification of large predetermined flakes, as well as in the modification of handaxes and cleavers [22–25]. As far as cleavers are concerned, and beyond the general similarities shared with handaxes, controlling the convex surfaces through delineation by retouch was yet another central concept in cleaver modifications.

The exploitation of convex surfaces, so exceptionally well controlled in the Acheulian, is documented in much earlier times (ca. 1.6 Ma) particularly in the production of cores on flakes reported from the Acheulian site of ‘Ubeidiya [29]. The same phenomenon of cores on flakes is well represented in the flint assemblages of GBY and in abundance [25]. A different facet of this mode of exploitation occurs in many Acheulian assemblages where bifaces were transformed into cores by a removal of a large predetermined flake from the convex surface of bifaces (e.g. [30–32]).

The ability to manipulate the convex surfaces independently of the artifacts size (bifaces, large flakes and very small ones) is a remarkable trait of the Acheulian. Furthermore, it is the foundation of many of the Middle Paleolithic cultural entities, and in particularly those associated with the Levallois method, as exploitation of convex surfaces is a fundamental prerequisite of this system. The cognitive abilities that enable designing a core in order to obtain a morphologically predetermined flake, and to exploit and modify convex surfaces form the roots of the evolution of the Levallois method [25].

What, then, are the cognitive implications of these behavioral patterns? Falling back on the theoretical framework summarized briefly earlier, and comparing the GBY knappers to earlier tool users, we suggest the following:

The precise cognitive nature of these concepts is far from clear. ‘Tool’ need not have been a ‘mental template’ in the sense of a stable visual representation imagined ahead of time; indeed, given that the category ‘tool’ clearly subsumed several specific tool types (e.g., cleaver), it could not have been a simple visual target. ‘Tool’ is an abstract category for us–it is a category made up of categories—and must have been for the GBY knappers as well. ‘Cleaver’, on the other hand was a more concrete category that almost certainly did have specific visual and ergonomic components. It existed as a model against which each large flake was evaluated. When modern humans use named tools (e.g., screwdriver) areas of the fusiform gyrus are activated [10] that are not activated when non-human primates use tools. Thus, thinking in terms of tool categories appears to have evolved at some point in hominin evolution. We suspect that this evolutionary development occurred long before the time of GBY, but it is certainly in evidence at GBY. Moreover, ‘tool’ and ‘cleaver’ were shared concepts that must have been acquired through social learning [36,37].

This evidence for cognitive categories (‘tool’ ‘cleaver’ ‘handaxe’) is, we think, very important. Anthropoid technical cognition evolved initially to control sequential procedures, which remain critical components of technical expertise. Categorical knowledge, however, is not a necessary component of anthropoid technical cognition; there is no reason to suppose that chimpanzees possess an abstract category of ‘tool’, or even a concrete concept ‘termite probe.’ Modern technical expertise is replete with categories that are held in LTM in the form of declarative/semantic information. Procedural LTMs are still paramount, but modern experts also rely on well-learned categories. It would appear that this was true for the GBY knappers, as well. The number of categories may have been fewer than one would find with a modern expert, but categories were definitely present in the minds of the GBY knappers. Further, it would seem likely, though this is impossible to know, that the GBY knappers had declarative/semantic labels for these concepts, either in the form of vocal words or perhaps gestures (we favor the former).

In sum, when compared to earlier stone knapping procedures, and the tool-use of non-human primates, the GBY knappers demonstrated significant cognitive developments. Most salient are the clear reliance on cognitive categories, which were socially learned, and significant expansions of long-term memory capacity and working memory capacity. It also appears likely to us that the knappers possessed semantic labels for their concepts. Compared to modern technical expertise, such blacksmithing [14] the GBY knappers demonstrated fewer cognitive categories. However, the archaeological record preserves only a very narrow range of technical activities, especially for the time depth of GBY, and a comparison with modern craft production is almost certainly misleading. What is more significant is the evidence for hominin reliance 780,000 years ago on a cognitive strategy—expert cognition—that continues to be important in the modern world.