After the Revolution: A Review of 3D Modelling as a Tool for Stone Artefact Analysis

3.1. Data collection

The final dataset contains 213 papers where 3D modelling is used as either an analytical, archival, or pedagogical tool (n = 152, 71.3%) or solely as an illustrative technique (n = 61, 28.6%). The papers in the dataset were initially identified by searching Google Scholar and Web of Science databases with the search terms: “stone art*fact”, “lithic”, “3D model*ing”, “scanning”, and “photogrammetry”. Additional papers were found by reviewing the publishing history of identified authors and by searching for documents that cited or that were cited within identified papers. Full bibliographic data for identified papers was then exported from Web of Science to import into R. Partial data for identified papers not listed on Web of Science was manually added (see Supplementary Information for details).

Papers on both flaked stone artefacts and non-flaked stone artefacts or tools were included in the dataset. Studies that incorporated flaked glass were also added (e.g. Dogandžić et al. 2020; Perston et al. 2022). The dataset includes both archaeological and experimental studies. Artefacts identified exclusively as “art” objects were excluded (e.g. Grosman et al. 2017). Papers where the models were created by 3D scanners, photogrammetry, and various forms of computer tomography scanning were all included. Papers where models were created by 3D microscopy (e.g. Hiscock et al. 2016; Morales and Vergès 2014) were excluded as these studies typically only partially model artefacts, and in terms of their content and intellectual history owe more to microscopy papers than they do 3D modelling papers (see Stemp, Watson and Evans 2015 for further discussion).

Books, book chapters, theses, and conference proceedings, as well as publications in languages other than English were also excluded as they are not consistently listed in the academic databases necessary for bibliometric analysis. However it is important to recognise that 3D lithic analysis research has been published in Chinese (e.g. Zhou and Guan 2017), Czech (e.g. Kaňáková, Šmerda and Nosek 2016), French (e.g. Martin-Moya et al. 2020), German (e.g. Dietrich 2021), Italian (e.g. Caricola et al. 2018b; Cristiani et al. 2018), Japanese (e.g. Noguchi 2019), Russian (e.g. Chistyakov and Kovalev 2019; Kolobova et al. 2020b; Shalagina et al. 2020) and Spanish (e.g. Cebrià et al. 2014; Duque Martínez and de Francisco Rodríguez 2015; Morales et al. 2013; Soto et al. 2018) and undoubtedly other languages that were not identified.

Data cleaning was carried out, primarily to identify and consolidate multiple names that represented a single author. Each paper was manually given a set of keywords based on the following categories (see Supplementary Information for further details):

• Thematic – Keywords in this category were designed to capture the general conceptual focus of papers.

• Analysis – This category included keywords relating to the methodological approach that a paper utilised.

• Industry – Keywords in this category relate to the industry, or techno-complex of the artefacts being analysed. These terms were sourced directly from the paper.

• Artefact – This category always included the generic keywords: “core”; “flake”; “retouched flake”; “point”; “biface”; “other artefact/tool type” where relevant. Author designated terms were also included.

Papers could have multiple keywords in each category. Additional information about whether the study included an experimental component, study region, modelling methodology, and analytical software used was also recorded for each paper.

3.2. Data analysis and visualisation

Bibliometric analysis is a quantitative, systematic, and transparent method of synthesising a body of research (Aria and Cuccurullo 2017; Cobo et al. 2011). There are two main outputs of bibliometric studies, performance analysis and scientific mapping (Cobo et al. 2011). Performance analysis aims to measure the impact on an academic field of scientific actors, i.e. researchers, institutions, countries, journals, or papers (Cobo et al. 2011). Science mapping is a way of spatially representing the networks between different scientific actors or keywords to map the social, intellectual, and conceptual structure of a research field (Aria and Cuccurullo 2017; Cobo et al. 2011).

For this study, only papers where 3D models of stone artefacts were used for analytical, archival, or pedagogical purposes were analysed. The overall aim was to identify the historic development of 3D lithic analysis and summarise the current state of the field. To that end, key authors, journals, and papers were identified. A collaboration network of all authors who have published ≥3 articles was created to explore the social structure of the field (Glänzel and Schubert 2005) and a historiograph of the 25 highest locally cited papers was created to explore the intellectual development of the field (Aria and Cuccurullo 2017). Co-word analysis of the manually added keywords was used to map patterns between the types of artefacts being analysed, the methodologies being used to analyse them, and the research questions they were being used to answer (Cobo et al. 2011). Trends in modelling method, software use, and study region were also analysed.

Bibliometric and other quantitative analyses were all carried out in R v.4.2.1 (R Core Team 2022), using the packages bibliometrix v.4.0.0 (Aria and Cuccurullo 2017), igraph v.1.3.1 (Csárdi and Nepusz 2006), qgraph v.1.9.2 (Epskamp et al. 2012) and tidyverse v.1.3.2 (Wickham et al. 2019).

4.1. Sources, authors, and institutions

Bibliometric analysis of the analytical dataset showed clear patterns of publishing and collaboration. 3D lithic analysis studies have primarily been published in archaeological science journals and journals focused on Pleistocene-age archaeology (see Figure 2). Per Bradford’s Law, there are three “core” journals, seven “zone 2” journals and thirty-two “zone 3” journals that have each published approximately a third of all papers (Bradford 1934). Regional and more humanities-orientated archaeological journals are less well-represented and are only found in “zone 3” journals.

Figure 2

A collaboration network of all researchers who have published ≥ 3 articles (n = 57) identified 12 clusters of authors (see Figure 3). Clusters either map to institutions (Clusters 1, 2, 7, 8, 10, and 12) or multi-institutional research projects (Clusters 3, 5, 6, and 11). The first five clusters are centred on the 10 most productive authors (see Figure 4). The prominence of researchers affiliated with Israeli and Australian institutions reflects the early adoption of 3D modelling by researchers at the Hebrew University in Jerusalem (Cluster 1) and the University of Queensland (Cluster 2).

Figure 3

Figure 4

Researchers affiliated with the Hebrew University in Jerusalem have been pioneers in the development of software packages for analysis of 3D models (Grosman et al. 2014, 2022; Herzlinger and Grosman 2018; Valletta, Dag and Grosman 2021). University of Queensland affiliated researchers have published experimental quantitative studies, generally focusing on testing new reduction indices (Clarkson 2013; Clarkson and Hiscock 2011; Muller and Clarkson 2014, 2016; Shipton and Clarkson 2015a). Both groups have published studies on experimental artefacts (e.g. Clarkson 2013; Clarkson and Hiscock 2011; Grosman et al. 2011; Grosman, Smikt and Smilansky 2008) and on Acheulean bifaces (e.g. Grosman, Goldsmith and Smilansky 2011; Herzlinger, Goren-Inbar and Grosman 2017; Muller, Shipton and Clarkson 2022; Shipton and Clarkson 2015a, 2015b). The influence of these key authors and institutions can be seen in a direct historic citation network of the field (Figure 5). Key early papers in the development of the field from these institutions include Grosman, Smikt and Smilansky (2008), Clarkson and Hiscock (2011), and Grosman et al. (2011). Early, influential work done outside of these institutions, namely Bretzke and Conard (2012), Lin et al. (2010) and Shott and Trail (2010), is also visible in the historiograph. More broadly, the historiograph maps the impact on the development of the field of papers that have tested modelling workflows and accuracy (Grosman, Smikt and Smilansky 2008; Lin et al. 2010; Porter, Roussel and Soressi 2016; Shott and Trail 2010), trialled new methodological approaches (Clarkson 2013; Clarkson and Hiscock 2011; Herzlinger, Goren-Inbar and Grosman 2017; Morales, Lorenzo and Vergès 2015; Muller and Clarkson 2014), developed analytical software (Archer et al. 2015; Herzlinger and Grosman 2018) or have an experimental component (Caruana et al. 2014; García-Medrano et al. 2019; Grosman et al. 2011; Herzlinger, Goren-Inbar and Grosman 2017; Shipton and Clarkson 2015a, 2015b).

Figure 5

4.2. Modelling, software, and accessibility

Laser and structured light scanning have been the most common modelling methods throughout the field’s history. Photogrammetry has become more common since 2016 (see Figure 6), reflecting the impact of Porter, Roussel and Soressi’s (2016) methodology paper. Studies have found the quality of the resulting models to be comparable, with the primary difference between methods coming down to the cost, time, and skill required to produce a high quality model (Magnani 2014; Porter, Roussel and Soressi 2016; Slizewski and Semal 2009). While not represented in Figure 6, there has been a record number of papers published in 2022 where CT scanning was used, with an emphasis on efficiently modelling large numbers of artefacts (Falcucci et al. 2022; Falcucci and Peresani 2022; Göldner, Karakostis and Falcucci 2022).

Figure 6

A broad range of software has been utilised for 3D lithic analysis (see Figure 7). Early papers tended to not report on what software was used, reflected in the high number of “unreported”. Published papers continue to use a mix of proprietary software, such as Avizo (e.g. Li, Kuman and Li 2015; Viallet 2019; Weiss 2015; Wiśniewski et al. 2019) and GeoMagic Design (e.g. Caruana and Herries 2021; Feizi, Vahdati Nasab and Wynn 2020; Lin et al. 2019), custom-designed packages for existing proprietary software, such as GLiMR (e.g. Davis et al. 2015; Davis, Bean and Nyers 2017) and Artifact3-D (e.g. Grosman et al. 2014, 2022), custom freeware such as AGMT3-D (e.g. Herzlinger and Grosman 2018) and open source software such as QGIS (e.g. Paixão et al. 2021; Zangrossi et al. 2019; Zupancich et al. 2019) and CloudCompare (e.g. Benito-Calvo et al. 2018a, 2018b; Cristiani et al. 2021; Dietrich and Haibt 2020; Proffitt et al. 2021). More recently there has been a shift towards using the coding language R, most commonly for geometric morphometric studies (e.g. Archer et al. 2015; Selden, Dockall and Shafer 2018; Weiss et al. 2018; Wiśniewski et al. 2020). The recent and ongoing development of the R package Lithics3D (Pop 2019) will hopefully expand the range of analyses able to be conducted directly in R.

Figure 7

4.3. Regional trends

The geographic distribution of 3D lithic analysis of archaeological material has been heavily concentrated on European assemblages (see Figure 8). West Asian, North American, and Southern and Eastern Africa assemblages all have ≥ 10 studies. Much of the early published research was on bifaces (e.g. Clarkson 2013; Grosman et al. 2011; Grosman, Goldsmith and Smilansky 2011; Grosman, Smikt and Smilansky 2008; Shipton et al. 2013; Sumner and Riddle 2008). The early development of indices for 3D analysis of bifaces provided a methodological template for later researchers to build upon and, consequently, studies have been concentrated on regions which have Acheulean assemblages. Bifaces are not the only artefact that have had a strong influence on where studies have been carried out. The prominence of North America as a study region reflects extensive research on ancestral First Nations American points (e.g. Davis et al. 2015; Gingerich et al. 2014; Sholts et al. 2012).

Figure 8

4.4. Co-word analysis: artefacts, methods, and research themes

A network analysis was performed to understand the relationship between keywords. Based on results of the analysis of the 150 most common keywords across the four categories (see Table 1), five main clusters were identified (see Table 2 and Figure 9).

Figure 9

The centrality of techno-morphological studies of Acheulean bifaces to the field is reflected in the prominence of keywords relating to these studies in Cluster 1 (e.g. “biface”, “Acheulean”, “handaxe”, “large cutting tool”, “Lower Palaeolithic” etc.). Analytically these studies explore handaxe symmetry (e.g. Li et al. 2018; Li, Kuman and Li 2016; Presnyakova et al. 2018; Sánchez-Yustos et al. 2017; Shipton, Clarkson and Cobden 2019), and volume loss (e.g. García-Medrano et al. 2019, 2020a; García-Medrano, Despriée and Moncel 2022; Li et al. 2018; Li, Kuman and Li 2015; Shipton and Clarkson 2015a, 2015b) to address questions on hominid cognition (e.g. Li et al. 2017; Li, Kuman and Li 2016; Shipton 2016; Shipton and White 2020), with a large number of studies using knapping skill as a proxy (e.g. Caruana 2020, 2022; Caruana and Herries 2021; Herzlinger, Wynn and Goren-Inbar 2017).

Keywords linked to geometric morphometrics are also prominent in Cluster 1 due to the widespread use of 3D landmark geometric morphometrics as a tool for analysing bifaces (e.g. García-Medrano et al. 2020a, 2020b; Herzlinger and Goren-Inbar 2019; Presnyakova et al. 2018; Shipton et al. 2013). Keywords relating to geometric morphometrics are also prominent in Cluster 3, e.g. “2D gmm” and “elliptical Fourier analysis”, which are strongly linked to North American point studies, though it should be noted that North American point studies use a mixture of landmark morphometrics (e.g. Davis et al. 2015; Davis, Bean and Nyers 2017; Selden, Dockall and Dubied 2020; Selden, Dockall and Shafer 2018) and elliptical Fourier analysis (e.g. Ahmed, Carter and Ferris 2014; Gingerich et al. 2014; Sholts et al. 2012, 2017). Geometric morphometrics has been the most widely used analytical approach in 3D lithic analysis, and has been successfully applied to both flakes and retouched flakes (e.g. Archer et al. 2018; Chacón et al. 2016; Delpiano, Gennai and Peresani 2021; Weiss et al. 2018). Its use is conspicuously absent in studies on cores, where it has proven to have limited analytical application (see Porter, Roussel and Soressi 2019).

Almost a third of all 3D lithic analysis studies include experimental stone artefacts, often to assess a methodology before applying it to an archaeological assemblage. Experimental reduction studies frequently take advantage of the fact that an experimental artefact can be modelled at different stages of reduction (e.g. Clarkson 2013; Grosman et al. 2011; Lombao et al. 2020; Morales, Lorenzo and Vergès 2015). Unsurprisingly, the keywords “experimental” and “reduction strategy” co-occur in Cluster 2. Most noticeably, experimental 3D lithic studies have been used to evaluate older lithic reduction indices and to develop and trial new indices to quantify mass and volume loss (e.g. Clarkson and Hiscock 2011; Lombao et al. 2019, 2020; Maloney 2020; Morales, Lorenzo and Vergès 2015; Ranhorn et al. 2019; Shipton and Clarkson 2015a, 2015b) or fracture mechanics (e.g. Archer et al. 2018; Dogandžić et al. 2020; McPherron et al. 2020; Muller and Clarkson 2014, 2016). A growing body of 3D refitting studies have also been key in the study of reduction strategies (e.g. Delpiano et al. 2019; Delpiano and Peresani 2017; Delpiano, Peresani and Pastoors 2017; Weiss 2015).

Functional studies, defined as papers investigating the physical traces of use, are a relatively new and niche development in 3D lithic analysis. Keywords associated with these studies cluster tightly together (Cluster 5) but are relatively unconnected to the rest of the field. GIS software and CloudCompare are commonly used in these studies to create digital elevation maps to identify macroscopic use-wear on a worked surface (e.g. Arroyo and de la Torre 2020; Benito-Calvo et al. 2015; Caricola et al. 2018a; Luncz et al. 2016). There are three key differences between 3D functional studies and other 3D lithic technology papers. The first is that the use of 3D modelling is primarily used to supplement traditional approaches to functional analysis (e.g. Caruana et al. 2014; Zupancich and Cristiani 2020), rather than supplanting it entirely as is generally the case with techno-morphological studies. Second is that these papers almost exclusively analyse non-flaked stone artefacts or non-artefactual tools (e.g. Cristiani et al. 2021; Dietrich and Haibt 2020; Paixão et al. 2021). The third is the strong link between functional studies and primate archaeology (e.g. Benito-Calvo et al. 2015; Haslam et al. 2013; Luncz et al. 2016; Proffitt et al. 2021). Almost all functional papers include an experimental component (with primate archaeology studies being the notable exception).

Artefact and industry keywords are split across multiple clusters. As discussed above, Cluster 1 is closely associated with Acheulean bifaces. Keywords present in Cluster 3 strongly reflect point studies from both North America (see above) and central Europe (Kaňáková et al. 2022; Kaňáková, Bátora and Nosek 2019, 2020; Nosek and Kaňáková 2021; Weiss et al. 2018). In Cluster 2, keywords are associated with industries with highly standardised reduction strategies, such as blade production (e.g. Bretzke and Conard 2012; Clarkson 2013; Porter, Roussel and Soressi 2019; Valletta, Dag and Grosman 2021; Valletta and Grosman 2021), discoidal cores (e.g. Clarkson 2013; Lombao et al. 2020; Malinsky-Buller, Grosman and Marder 2011; Ranhorn et al. 2019) and other types of prepared cores (e.g. Clarkson 2013; Li et al. 2017; Ranhorn et al. 2019).