Theory in Archaeology, Theory in Computing 1: Cross-Fertilisation in CAA and TAG Conferences Between Archaeological Theory and Computer Applications

1.1. Aims and methods

Recent research has discussed publishing trends in archaeology among which the growth of methodological specialties, such as network analysis (Brughmans & Peeples 2017) and GIS (Ullah, Clow & Meling 2023), has been noted. Instead of surveying publications, I will attempt to assess the development of this interconnectedness between conferences, in this case the Computer Applications and Quantitative Methods in Archaeology (CAA) International and Theoretical Archaeology Group (TAG) conferences.

This connection was tangible in Sara Perry and James Taylor’s DigiTAG session in 2016: there was a direct exchange of ideas and sessions between the CAA International and TAG conferences (Perry & Taylor 2018). This event made me think of the earlier occasions in the history of CAA when theory and method had been juxtaposed. In this belated 50^th anniversary article celebrating the 50 years of CAA International, I want to examine whether there was cross-fertilisation between theory and computerised or digital practice at the conferences and how the history of theoretical archaeology is reflected in the history of the theoretical content in CAA conferences. How did the content reflect the changes in archaeological theoretical paradigms? When were computer-aided methods present in TAG and how do the presentations at conferences in this period relate to the paradigm changes?

From the start it was clear that the surviving archive material was not directly comparable. For TAG, there exist the session lists, paper titles and abstract books in an online archive at https://antiquity.ac.uk/TAG, whereas only the published papers are available for the past CAA conferences in the edited proceedings. Sadly, CAA does not have same kind of online archive of the sessions offered as TAG; the past programmes have not been formally archived. The CAA proceedings contain only those items the authors considered worth publishing or had the time to publish. Thus, in practice, in this article I have looked for the theoretical articles within the CAA proceedings up to the 2016 Oslo CAA published in 2018 and the 2017 Atlanta CAA published in 2020, all together 45 sets of proceedings as listed at https://caa-international.org/publications/proceedings/bibliography/, and computing sessions within the TAG conferences up to 2022. The aim of this survey is to establish the historical connection between theory and method up to the 50^th anniversary of CAA International in 2023. Therefore the research period spans the first 50 years of CAA. I review the TAG sessions and paper titles within this period, but the limitations arising from the inaccessible unpublished TAG material result in an asymmetry in the depth of exploration, which favours the CAA. The CAA sessions and papers are discussed on the basis of their conference year, not their publication year which is stated only for bibliographical reference.

Many of the past published CAA papers are archived up to 2011 on the CAA web site at https://proceedings.caaconference.org/, but the original publications were consulted in the Cambridge University Library and the Betty and Gordon Moore Library at Cambridge. Those proceedings that were not available in physical copies were consulted online in the libraries. Specifically, I was looking for the word ‘theory’ or ‘theoretical’ in the article titles in the CAA proceedings and also computer-related vocabulary, such as ‘computer’, ‘GIS’ or ‘network analysis’, in the session and paper titles in the TAG conferences. I also assessed that content of the published CAA papers which reflected on theoretical considerations. The first part of the article scrutinises the theoretical developments in the CAA conferences, analysing the paper titles and exploring the paper contents in relation to the diffusion of first post-processual and interpretative ideas and then critical thought. The second part examines the computer-aided methods at TAG. The final part seeks to contextualise the results of the first two parts and to provide an understanding of how the two conferences interlink.

1.2. Archaeological theory

What do I mean by theory? A theory can mean a tentative explanation of a past phenomenon, but also more widely it refers to theoretical thought in archaeology and its ontological and epistemological basis. My intention is mostly to explore the latter. I am interested in theoretical thinking in computing, although I will mention the occasions when terms have been used differently in the respective conferences.

In order to set this enquiry in its context, one has to set out the general outline of archaeological theoretical thought. The main periods of 1) cultural-historical archaeology, 2) processual archaeology and 3) post-processual archaeology (cf. Trigger 1989; Johnson 1999) can be seen as different paradigms following Kuhn’s (1962) idea. Cultural-historical archaeology was the main line of archaeological thought before the 1960s and even with the development of the so-called ‘New Archaeology’, the early name for the processual archaeology in the 1960s, its influence has never totally disappeared. Cultural-historical archaeology set out to define cultures and describe them using inductive reasoning, where conclusions were based on the testimony of the accumulated evidence and the majority of the cases. The changing fashions in material culture, especially in pottery and metalwork, were considered to provide the key evidence. Processual archaeology can be traced back to the work of Lewis Binford, whose early articles declared archaeology as anthropology and as a science and thus a discipline applying hypothesis testing and hypothetico-deductive reasoning (Binford 1962; 1964). Cultures were not equated with peoples as in cultural history but seen as systems that were made up by different subsystems, such as economic, social, religious and material subsystems that developed within their natural context and provided the framework for developments (cf. Clarke 1968). Mathematics was regarded as a way to find laws, or at least generalisations, in the corpus of excavated material, and thus, statistics and computing provided important methods to look for generalisations. The scientific approach introduced sampling methods that could be treated with statistical methods to base conclusions on material evidence.

By the 1980s the dissatisfaction towards processual archaeology was cumulating, since it was perceived that it was extremely difficult to formulate universal laws. Furthermore, processual archaeology seemed ecologically deterministic and even the Middle Range Theory suggested by Binford (1975; 1977; 1978; 1981; 1983) did not give permanent solutions. This was also a time when structuralist and Marxist thought were influencing the Humanities more widely. Post-processualists saw all research as theory-laden, whereas processual archaeologists distinguished between theory and method. Post-processualists were against positivism and saw science as a social construct and not the only source of knowledge (cf. Johnson 1999: 98–108). Like processual archaeology before it, key works played a crucial role in advocating for post-processual archaeology. This can be seen in the two books by Michael Shanks and Christopher Tilley, Social Theory and Archaeology (Shanks & Tilley 1987b) and especially Re-constructing Archaeology (Shanks & Tilley 1987a; 1992), even if the works of Ian Hodder were also crucial in this period in formulating different ideas how to read the past (especially Hodder 1982a; 1982b; 1986).

After post-processual archaeology, there has been no similar comparable profound paradigm shift that one can put a finger on. However, one can consider general tendencies continuing critical approaches to archaeology (see Gardner, Lake & Sommer 2013) as demonstrated by gender (e.g., Gilchrist 1999; Stig Sørensen 2000; Power 2020) or postcolonial archaeology (e.g., van Dommelen 2011) as building on the currents present in post-processual archaeology. However, there are more processual tendencies, such as neo-evolutionary archaeology (e.g., Prentiss 2021), which indicate that there is theoretical diversity in the approach depending on the subject matter of the research.

Next I will explore the presentation of archaeological theory in the CAA conferences through a systematic chronological survey and critical analysis of the words used in the titles of papers. This review acknowledges the fact that some papers were presented, but not published, or were published with a different focus from the paper given in person. It also draws attention to whether these significant terms carried a theoretical or a narrower meaning.

2.1. The words ‘theory’ or ‘theoretical’ in the paper titles

The early conferences were predominantly about computerised techniques and their archaeological applications. The first mention of the word ‘theory’ was in the title of a paper by Laxton and Litton (1983) ‘Information Theory and Dendrochronology’. The next time the word came up was in 1992 in the article ‘Sites and Monuments Records in England — Theory and Practice’ (Lang & Stead 1992). Perhaps the first time the word was used in a properly theoretical sense was Wheatley’s (1993) paper ‘Going over Old Ground: GIS, Archaeological Theory and the Act of Perception’, presented in the Aarhus 1992 CAA conference. This paper discussed GIS in the post-functionalist framework. Similarly, Lock (1995) discussed theory and contextualisation in his 1994 CAA conference paper, aptly titled as ‘Archaeological Computing, Archaeological Theory, and Moves towards Contextualism’. He tried to place the development of archaeological computing within the general theoretical framework, characterising the post-processual era as emphasising data exploration and visualisation. However, the popular 1995 Leiden conference did not have any papers with ‘theory’ in the title. Nevertheless, Barceló (1996) was discussing fuzzy logics in typology and there were many papers on GIS and predictive modelling.

The 1996 Iași conference was the occasion of two theory papers (Wheatley 2000; Wise 2000). In his paper ‘Spatial Technology and Archaeological Theory Revisited’, David Wheatley explored the upcoming archaeology of place and the importance of perception and its theoretical grounding in GIS analyses. He emphasised that the theoretical basis of this approach) must rest on critical and contextual approaches. Alicia Wise, in her paper ‘Building Theory into GIS-Based Landscape Analysis’ discussed landscape, human, cultural and historical ecology, and phenomenology, and their importance to landscape analysis. All these mid-1990s papers were a response to the GIS criticism within post-processual archaeology.

In the following year that saw the 25^th anniversary conference in Birmingham there was again just one paper with ‘theory’ in the title (Earl 1999). However, Earl’s article was not so theoretical and discussed Danebury’s defensive functions more specifically. Nevertheless, the word ‘interpret’ was coming through (Riley 1999) alongside the expected celebratory historical articles (Laflin 1999a; 1999b; Scollar 1999).

The Barcelona conference in 1998 contained a special theory session, organised by Joan Barceló (the author was present), but none of the papers were published in the proceedings. The 1999 Dublin conference produced two papers with ‘theory’ in the title (Bogdanovic, Vicente & Barceló 2004; González-Pérez 2004). The first article presented the DIASPORA project, an instrument for sharing hypotheses and their testing over Internet, while the latter showcased the information structure of a SIA+ (Archaeological Information System, enhanced version), a software product. These articles are more specific than the general theory articles presented by for example Lock (1995) and Wheatley (2000).

The 2004 conference in Prato again produced two theoretical articles (Barceló 2010; Peeters 2010). Barceló (2010), in his paper ‘Archaeological Theory, Techniques and Technologies: Beyond Quantification and Visualization Methods’ presented very briefly a session he organized rather than theoretical thoughts. Peeters (2010) presented a spatial analysis of Dutch Mesolithic and Early Neolithic sites using percolation theory and thus using the word theory in a more specific sense. However, a paper on the fuzzy logics could also be interpreted as theoretical (Niccolucci & Hermon 2010). There was also one article with the word ontological in its subtitle (Cripps & May 2010), but it discussed data models of cultural heritage systems rather than theory.

The 2005 Tomar conference in Portugal also had two papers with the word ‘ontology’ or ‘ontological’ in their titles (Fernández González, Polo Marquéz & Cerrillo Cuenca 2007; McAuley & Kilfeather 2007), but again in their nature their focus was more on information technology. Then, after a hiatus, a paper with theory in its title from 2007 was published from the Berlin conference. Barceló (2008), in his paper ‘Towards a True Automatic Archaeology: Integrating Technique and Theory’, discussed the possibility of automating the scientific process. In the 2008 Budapest conference Pescarin’s (2011) paper title contained the word ‘theoretic’. There were also two papers with ‘ontology’ in their title (Hiebel & Hanke 2011; Jeffrey et al. 2011), but once again these referred to the data structure rather than theory.

In the Williamsburg 2009 CAA the ‘Walking the Ridgeway Revisited’ paper by Lock and Pouncett (2010) contained the word theoretical in its subtitle. The paper discussed the methodological and theoretical implications of scale dependency in deriving slope and calculating least-cost pathways. Apart from discussing the sizes of the slope calculation windows it explored intentionality, not built into GIS, and presented the concept of mid-distance waypoints. Thus theoretical discussion was observably latent in the early 2010s.

The 2015 Siena conference had one paper (Arnoldus-Huyzendveld, Citter & Pizziolo 2016) with ‘theoretical’ in its subtitle. The 2017 Atlanta CAA conference had two papers which had the word ‘theory’ in the title: Van Havre (2020) wrote about ‘The dynamics of Brazilian rock art landscape: An Agent-Based Modelling approach to theories’ and James (2020) about ‘Context as theory: Towards unification of computer applications and quantitative methods in archaeology’. This last article attempted a wider theoretisation of GIS by using the concept of ‘context’ as a theory of inclusion, as opposed to ‘context’ in computer science and other fields where ‘context’ is a theory of exclusion or only part of a whole. ‘Context’ is multidimensional, with each layer embedded and overlapping in multiple scales, constituting the dimensionality of archaeological research. In James’s paper ‘context’ defines relevancy and theorises GIS as a method that organises this multidimensionality. This multidimensional, multi-scalar perspective of ‘context’ as theory intersects themes and archaeological settings to unify the seemingly fractured character of computer applications and quantitative methods in archaeology.

In CAA the exploration of theorisation has always been GIS-heavy. The initial papers were responses to the post-processual critique but increasingly, contributions have been more freestanding, trying new things. The theory discussion has also partly moved away from the Anglophone region and one has to recognise the contribution of Joan Barceló in this respect.

2.2. Paper content and theory in the processual period

When we look at the proceedings more widely, in the 1974 conference there is a paper by Doran (1974) called ‘What computers can’t do for archaeologists’. In this article he discussed the problems of data representation in his day. The coded programmes of this time could not handle the detailed knowledge gathered by archaeologists. This was seen as a general and major limitation on the use of computers in archaeological interpretation. In the 1976 conference Doran (1976) wrote about ‘Complex process models and the understanding of prehistoric change’. In this paper he also discussed the idea of the ‘right conceptual framework’, by which he meant ‘a relatively coherent collection of concepts, and of facts or beliefs expressed in terms of those concepts’ (Doran 1976: 21). Doran connected this to the ‘New Archaeology’ as constituting such a framework, thus in fact underpinning the whole development of computer applications in archaeology. As potential processes in computing science he presented simulation studies and artificial intelligence.

Angell (1978) wrote about theorising the purpose of megalithic monuments and used a computer to test his mathematical theory on the shadows of the megalithic stones. This was totally in line with the processual framework. In the 1980 conference Iliff (1980) presented a paper on the use of simulation in the understanding of patterning in cemeteries. In this paper she hoped to employ it to improve understanding of the past societies and their workings. However, in Callow’s (1981) paper ‘Have computer, will classify? A Stone Age problem’ resulting from the 1981 London CAA, the questions asked are relatively ontological with the main problem focussed on the choice of meaningful input variables to test hypotheses. This article also pointed out the uncertainty of the appropriateness of the algorithms and the poor state of development of significance tests associated with the analyses. The conference volume also included Haigh’s (1981) article on surface sampling and different systematic sampling techniques.

In the 1982 conference proceedings Ennals and Brough (1982) explored the representation of archaeological knowledge, even if their primary aim was to create an information system a lay person could consult. In his 1983 Bradford conference paper Flude (1983) discussed the problems facing computer archaeology: these were mainly technical and practical but at the end he predicted that in five years computers would be seen as indispensable for archaeology. In the same year, Grimley (1983) discussed the relational data model for databases. This contribution may be seen as a technical paper, but it was the first presentation of the relational data model in the proceedings. Similarly technical was Copeland’s (1983) paper on the information retrieval systems for archaeological data. The extensive application of computerised methods in archaeology can be seen in the 1984 proceedings that contained many papers on databases, monuments records and national archaeological archives, together with reference to the use of home computers (Rains 1984) and computers in museums (Light 1984). Even if these papers cannot be regarded as theoretical, they show how computing was reaching wider audiences.

In the 1985 London conference, George (1986) presented the paper ‘Interrogating fuzzy data’. Richards (1986) wrote on ‘Into the black art: achieving computer literacy in archaeology’ in which he acknowledged the rise of the second generation of computer archaeologists and the emergence of the microcomputer. However, this was in all other respects a practical article about which skills the archaeologists should have, what and how these should be taught and by whom. Ryan (1986), in his treatise on the archaeological toolkit, not only discussed software tools and graphic tools, that he called deterministic, but also thinking tools. By these terms he meant tools that allowed interaction between the user and their data. As an example, he presented the idea of a stratigraphic thinking tool that could present prior and posterior contexts and the genealogical tools created by social anthropologists. In their article Smith and Atkin (1986: 133) referred to the then recent discussions on the roles of computer archaeologists and computer specialists in archaeology.

In the 1988 proceedings, Ryan (1988) published a bibliography of computer applications and quantitative methods. This was organised mainly according to methods and techniques. Elsewhere in the proceedings, in the section ‘Expert systems’, Stutt (1988) presented his article on ‘Expert systems, explanations, arguments and archaeology’ in which he explicitly mentions archaeological reasoning and theory. He states

“Given the prevalence of non-deductive forms of reasoning in archaeology in its attempt to make what Hodder (Hodder 1984) calls ‘leaps of faith’ to the proper interpretation of the data available, it is obvious that archaeological theory will not correspond to the sets of propositions (or equations) with relations of logical or mathematical implication applicable to ‘hard’ sciences such as physics” (Stutt 1988: 264).

According to this approach there will be no general laws in archaeology. Stutt promoted the Argument Support Program for Archaeology (ASPA) that, through its mechanism for modelling several different points of view, could, according to him, be used to model different interpretative theories as sets of facts plus the interpretative transformations and strategies for composing these. Also in this conference, Wansleeben (1988) presented the geographical information systems suitable for archaeological research and their use in site location analysis, site pattern predictions and pattern reconstruction.

2.3. Theoretical content in the post-processual period

In the 25^th anniversary conference in 1997, Scollar (1999) reviewed 25 years of computer applications in archaeology. He explored the fluctuations in the numbers of papers in different categories, such as statistics, databases and geographical information systems (GIS). His verdict was that computing had made archaeological tasks known prior to computers easier, faster or cheaper, but probably rarely had revealed or led to new archaeological knowledge (Scollar 1999: 8). In the same volume, Stewart (1999) inquired if computing provided a greater physical and intellectual access to archaeology. She saw computing’s achievements as including the development of standards in data managements and Lock’s (1995) contextualisation, linking the development of the technology and theory ‘drivers’ (Stewart 1999: 20).

In the same conference, Gillings (1999) presented a critical take on the term virtual reality that raised the question of the motivation of such studies. These potential motivations were theoretical, relating to the embodiment, process and the social nature of space and time, as foregrounded by Barrett (1994) and Tilley (1994). The expansion of the GIS approach was seen through the lens of understanding the shortcomings of person-environmental relations provided by the traditional functionalist and deterministic interpretative frameworks and the push towards the more experimental and embedded notions of Being and Dwelling (Thomas 1996; Ingold 1995). Gillings tried to bridge the gulf between theory and practice by using virtual reality to explore a problem from the outset and defining a number of criteria along the way. The approaches had to be flexible and affordable and also reproducible and communicable to the wider archaeological community and interested parties (Gillings 1999: 248). Gillings (1999: 250–251) further explored the definition of virtual reality (a manufactured intensity; Gray 1995: 343–346) and hyperreality (the more real than the real; Baudrillard 1983: 46; Horrocks and Jevtic 1996: 109). He also discussed the issues of authenticity and faithfulness (Gillings 1999: 252) and noted how the field was under-theorised (Gillings 1999: 253).

Orton (1999) for his part discussed the role of statistics in archaeology and the promise of quantitative methods in the 1970s. He presented the developmental arc in terms of four stages taken by topics in scientific disciplines: early exploration, discovery, consolidation and accommodation. Orton considered the late 1970s and early 1980s as a period of transition from the discovery period to consolidation (Orton 1999: 26) and identified a series of tensions: exploratory versus confirmatory statistics, statistical analysis versus statistical design, classical statistics versus Bayesian statistics, numerical versus graphical presentation and commercial versus home-grown packages together with the all-purpose sampling metaphor versus taphonomy. In addition, there was the question of New Archaeology versus post-processual archaeology in the discipline of archaeology (Orton 1999: 26–27). Orton considered that in the first matter the balanced view was lost when people held the misconceptions that the confirmatory statistics worked only in theory and were only for those who believed in an hypothetico-deductive approach (Orton 1999: 28). Orton (1999: 32) asked what statistics offered to post-processual archaeology and his answer was ‘method’. Even Shanks and Tilley (e.g., 1992: 137–240) used statistical techniques. He also pointed out that theory had already impinged on statistics by creating a need for new forms of analysis, such as social status analysis (Duncan et al. 1988). He also pointed out that the acid test for theory is whether it stands up in the face of the data (Orton 1999: 32).

In the proceedings of the 1998 conference there was a section dedicated to using computers for archaeological explanation, divided into two: statistical analysis and modelling. Even if many of the statistics papers were more methodological (e.g., Westwood, Baxter & Beardah 1999), the modelling papers included many papers on GIS (e.g., van Leusen 1999; Kirkinen 1999), a statement on predictive modelling (Kamermans & Wansleeben 1999) and Gary Lock and Patrick Daly’s (1999) article on ‘Looking at change, continuity and time in GIS’. The last two authors also commented on the temporal variables in GIS (Lock & Daly 1999). In addition, Barceló, Briz and Vila (1999) provided a general introduction to computing archaeology and wanted to change the way we explain the historical process. However, this article was ultimately the introduction to the contributions in the volume rather than a deep consideration of the essence of computer archaeology.

Van Leusen (1999) discussed the argument that the use of GIS inevitably leads to environmental determinism (see also Gaffney & van Leusen 1995). Llobera (1996) had pointed out that the ‘Cartesian space’ was abstract with no observer, no perspective and no history. To address this, different authors introduced concepts to ease the perceived rigidity of the form of GIS then current: uncertainty (e.g., Gillings 1998), the ideal organisation of space and society (Zubrow 1994) and time and change and affordances (Llobera 1996). These offered improvements for visibility analysis whereas there was a suggestion that the least-cost analysis had to be improved (van Leusen 1999: 217). In addition, perception and cognitive landscapes were discussed as ways to escape from ‘objective’ geographical space (van Leusen 1999: 219–220). Van Leusen (1999: 220) called for theoretical justification for the use of least-cost and viewshed analyses.

The many GIS papers do contain theory: for example, Stančič and Kvamme (1999) include the so-called social variables in their study (hillforts and their distances from each other and sea, and visibility) and Vermeulen (1999) formulated hypotheses before different GIS analyses. Naturally, these approaches were more rooted in the processual tradition, as was predictive modelling as a concept (Kamermans & Wansleeben 1999). It is clear that the heated processual versus post-processual GIS debate brought existing theory to the forefront.

The 2000 CAA conference in Ljubljana had a session on future trends in spatial analyses. A slight novelty was the application of neural networks into the study of religious differences in two Slovenian Roman towns (Županek & Mlekuž 2001). The 2003 Vienna conference had two papers on archaeological ontologies: the former (Hermon et al. 2004) discussed fuzzy logics and the other (Missikoff 2004) domain ontologies in a prototype software under the title ‘archaeological processes’. This conference also saw Whitley’s paper on causality in predictive modelling in which he stated that

“Ultimately, the dichotomy between inductive and deductive approaches is not in theoretical orientation, rather it is embodied in our understanding (or failure to understand) that predictive modelling is really a tool useful only for land management, not interpretative archaeology” (Whitley 2004: 236).

The more recent conferences have had overarching themes. The Ljubljana 2000 one had an overarching theoretical theme ‘Computing archaeology for understanding the past’, with a nod to Hodder, as did the 2004 Prato one, i.e., ‘Beyond the artifact—Digital interpretation of the past’. The themes of the following conferences can be similarly interpreted, 2007 (Berlin, ‘Layers of Perception’) and 2008 (Budapest, ‘On the Road to Reconstructing the Past’). This is also a period when many theory-laden papers were given, and these will be discussed below. Of the most recent conferences, the 2019 conference theme in Krakow, ‘Check Object Integrity’, can be interpreted to be ontological as well as computational.

2.4. After the post-processualist critique

In the 2004 Siena conference there were two rather important papers on predictive modelling. Kamermans (2010) discussed ‘The Application of Predictive Modelling in Archaeology: Problems and Possibilities’. He had no doubt that the method had an important role in archaeology, although he acknowledged that many of the problems were not resolved and there had been a significant debate on the subject. One of the core problems was that predictive modelling, especially when performed with the aid of a GIS, was accused of environmental determinism. The way to counteract this emerged as a consensus on incorporating cultural factors in the modelling: these factors should be additional predictors in the process of predictive modelling. This was the argument made by Verhagen et al. (2010). Kamermans (2010) raised as an example the work of Thomas Whitley (2002; 2003; 2004) on the more fundamental aspects of ‘cognitive’ predictive modelling. Much of this work was presented in the CAA conferences, in 2001 in Gotland, 2002 in Heraklion, Crete, and 2003 in Wien.

Cripps (2007) presented in the 2005 Tomar CAA conference his paper ‘Pathways, Perception and the Development of Place: Computational Approaches to Movement and Perception of Landscape in Prehistory’. In this he acknowledged that the recent visibility studies were based on both functionalist explanations and methods and post-processual strands that are based on phenomenology. Even if his approach was mainly methodological and technical, he pointed to the theoretical frameworks.

In the 2006 Fargo proceedings, Posluschny (2007) briefly discussed the character of GIS as a tool in archaeology. He did not consider GIS as a method but as a way to find patterns for interpretation. The Fargo conference also had a session on network analysis (Lock & Pouncett 2007), but their introductory article is more methodological than theoretical. Similarly, Altaweel (2007b) briefly introduced an agent-based modelling and simulation session. However, in his agent-based simulation paper he tried to provide a holistic model that accommodated different theoretical positions besides providing a short history of simulations in archaeology (Altaweel 2007a).

Zapłata (2008), in his CAA 2007 Berlin paper, discussed ways of cultural perception rather philosophically but with few references to the literature. Paliou (2008) on her part made a reference to Bourdieu (1977) and some other social theorists as a background for her autonomous agent approach which is founded on Helbing’s behavioural force model.

In the 2009 CAA, Posluschny (2010) examined the importance of simple GIS tools in interpreting archaeological sites. His analysis was based on the definition of hinterland and cost-surface-based models. The aim was to show how these models can be used for simple interpretations of sites within their landscape. This is a quite traditional processual technique and Posluschny used it in a traditional way.

In 2010 CAA, Dobbins and Gruber presented two papers on reconstructing and contextualising Roman architecture and looking at the lines of sight in a Roman house (Dobbins & Gruber 2013; Gruber & Dobbins 2013). This CAA conference also had sessions on agent-based social simulation (e.g., Barceló et al. 2013; de la Cruz et al. 2013), analytical GIS and issues in least-cost analysis. The last session produced important papers from Irmela Herzog (2013) and Philip Verhagen (2013). The latter called for a better theoretical foundation of movement. There was also an interesting sensory paper from Paliou and Knight (2013) that aimed at multisensory mapping, in this case visibility and audibility. The conference also saw a paper on ontology (Niccolucci et al. 2013) that described an open text documentation system that was based on CIDOC-CRM.

The 2011 CAA conference in Peking had a section on data management, but the papers were more technical in character. The section on conceptual modelling had a few interesting papers: Hug et al. (2012) wrote about epistemology and the understanding of research processes in the Humanities, such as Archaeology. They briefly explained why the existing process models for Humanities were not tracing the research processes sufficiently. They then presented different process models from Information Systems Engineering that allow tracing processes according to different perspectives such as activities, decisions or strategies. They suggested that these process models can be useful to represent research processes in the Humanities.

The articles from sessions on Geographic Information Systems and network analyses were again more technical, but Brughmans, Isaksen and Earl (2012) wrote an article ‘Connecting the Dots: an Introduction to Critical Approaches in Archaeological Network Analysis’ that gave an overview of the critical discussions in Network Science and presented the papers in the homonymous CAA2011 session. The prior critical discussions were not numerous and the authors concluded that many avenues remained unexplored and the already discovered methods and theories wait to be combined in a critical and specifically archaeological framework.

The 2012 Southampton 40^th CAA conference had something rather special. The conference hosted the Personal Histories project and had a panel of Sue Laflin, Phil Barker, Paul Reilly, Nick Ryan, Clive Orton, Irwin Scollar, John Wilcock, Hans Kamermans and Gary Lock remembering the history of CAA. This happening can still be viewed online at https://www.personalhistories.arch.cam.ac.uk/oral-histories/seminars/computer%20applications%20and%20quantative%20methods. Apart from this, the published proceedings pursue a general theme ‘Theoretical Approaches and Context of Archaeological Computing’ that is presented in both of the two volumes of proceedings. These two sections include very diverse articles, most of which are more methodological than theoretical. However, Vis (2013), in his article ‘Boundary Concepts For Studying the Built Environment’, discussed different spatial theories and developed theoretical conceptualisations of boundaries. Similarly, Mlekuž (2013) is highly theoretical in his contribution ‘Everything Flows’ in which he tried to approach movement from a new angle using tools from cellular automata that would complement the optimal path approaches and bring agency into the mix. This article bases its argument on landscape archaeology.

The 2013 Perth conference contained some more theoretical papers in the ‘Computational Modelling and GIS’ and ‘Data Management and Analysis’ sections. Several papers tested hypotheses using simulation as a method (e.g., Romanowska 2015) and Rodier, Kaddouri and Garmi (2015) presented in their paper ‘Systemic Approach and Spatial Organization’ geographical models with urban case studies. All these papers rely on more processual argumentation and can be seen as a sign of the recent and ongoing exploration of scientific methods in archaeological argumentation. In the Data Management section, Martín-Rodilla (2015) presented a theoretically-based characterisation of different archaeological statements in her article ‘An Empirical Approach to the Analysis of Archaeological Discourse’ in order to develop applications in knowledge extraction. Similarly theoretical is ‘Expressing Temporal and Subjective Information about Archaeological Entities’ by Gonzalez-Perez, Martín-Rodilla and Blanco-Rotea (2015).

This period saw movement and perception normalising the post-processual and interpretative input and output. However, predictive modelling still had its proponents and new methods, such as network analysis and agent-based modelling, were bringing theory to the fore.

2.5. Towards the critical digital archaeology

In the 2014 CAA Paris conference there was a Historiography section where Moscati (2015) gave her paper ‘Towards a history of archaeological computing’ in which she counted the then recent efforts to look into the history of archaeological computing: the 20^th issue of the Archeologia and calcolatori journal, the 40^th 2012 CAA conference with the Personal Histories project from Cambridge and the UISPP 2014 in Burgos that contained the session ‘Revolution of the sixties in prehistory and protohistory’. Moscati’s (2015) article listed the early conferences in late 1960s that established the field and finished with the first CAA in 1973. Otherwise, the conference had a theme of ontologies and standards in which, for example, Smith (2015) presented his article ‘the Digital Archaeological Workflow: a case study from Sweden’. In his introductory article, Djindjian (2015) made a plea for ignoring ideologies, paradigms and theories. This is exactly the opposite of what Perry and Taylor (2018) would call for a couple of years later.

The 2015 Siena conference had many sessions with methodological content and the word ‘theory’ appeared in the session title ‘Towards a Theory of Practice in Applied Digital Field Methods’ organised by James Taylor and Nicolò Dell’Unto. The sections of Modelling the archaeological process, Linking data, the multiple spatial analysis sessions and open source and open data and Computers and rock art session also provided content that is relevant to the current discussion. In the last-mentioned session Díaz-Andreu and Mattioli presented ‘Archaeoacoustics of rock art’ to add to the study of less tangible cultural signs. However, Bikoulis (2016) in his ‘Hypothesis testing and validation in archaeological networks’ paper touched upon a theoretical question in network science. It called for more rigorous hypothesis testing, since the adoption of different creation strategies for networks creates hugely different networks. Just because we get patterns does not inherently mean that the network is significant and meaningful, since random processes can also create them. In addition, Arnoldus-Huyzendveld, Citter and Pizziolo (2016) discussed ‘Predictivity – postdictivity: a theoretical framework’ in which they declared that prediction (what, where and how?) and postdiction (why?) are complementary and parts of the same procedure. In their case they looked for the explanation from written sources. Herzog (2016) continued her studies of movement with the article ‘Dispersal versus optimal path calculation’. There were also agent-based modelling and GIS papers (e.g., Chliaoutakis, Sarris & Chalkiadakis 2016).

In the 2016 Oslo CAA conference the themes included ontologies and standards with two papers with ‘ontology’ in the title (Tolle, Wigg-Wolf & Gruber 2018; Vassallo et al. 2018). An important paper is Bruseker et al. (2018) that discusses 20 years of experience in developing of CIDOC-CRM and the name of the article ‘Is that a good concept?’ refers to knowledge modelling and the analysis of conceptual elements in the article. The themes GIS and spatial analysis and complex systems simulation were more methodological in content than theoretical (e.g., Herzog 2018). However, what was important was that TAG had teamed up with the CAA conference to run a successful Digital TAG (digiTAG) session in Oslo, although this did not result in other clear proceedings papers other than Perry and Taylor’s (2018) ‘Theorising the Digital’. This session, organised by Taylor, Perry, Dell’Unto and Berggren, had another instalment in the same year’s TAG, organised by Taylor and Perry. The published article called for more theoretically aware digital archaeology and pointed out that many of the critical GIS articles were not referred to by archaeological practitioners.

In the Atlanta CAA 2017 conference, in addition to the two articles with the word ‘theory’ in title, Higginbottom and Mom wrote about ‘Place: The Physical Embodiment of Collective Information’ in the VR section. This article, mainly an example of a use of software, nevertheless, aimed at using “the interaction and navigation of the spatial, temporal and thematic components of archaeological data to demonstrate how visualisation facilitates deeper insights into archaeological phenomena” (Higginbottom & Mom 2020: 443). However, the most theoretical content was in Nüsslein, Nuninger and Verhagen’s (2020) article ‘To boldly go where no one has gone before’ that aspired to change the image of predictive modelling, characterised as being environmentally deterministic, a-temporal, or even effectively de-humanising the past. However, they pointed out that spatial analysis of settlement patterns had progressed, paying much more attention to the role of socio-cultural factors and the analysis of settlement pattern dynamics. They described their approach to site location analysis and predictive modelling as essentially data driven, yet theoretically informed, and focused primarily on facilitating comparisons between various chrono-cultural contexts.

Unsurprisingly, processual archaeology was a natural environment for computer applications. With post-processual archaeology, GIS and other methodologies were criticised and scholars tried to find solutions for the problem of environmental determinism. The content was often reactive, reflecting upon the introduction of either new interpretative constructs or more commonly new methodology. However, recently, the ‘more scientific’ lines of research, such as agent-based modelling and network science find their framework more in what looks very much like processual archaeology with hypothesis testing and modelling. Nevertheless, critical strands are also presented in the conferences. It is also notable that the more theoretical papers presented in the conferences were not necessarily published in the proceedings: this seems to have happened with both the Barcelona and Oslo theory sessions. Therefore, it is totally possible that the theoretical content of these conferences is underestimated.