The Changing Face of Neolithic and Bronze Age Ireland: A Big Data Approach to the Settlement and Burial Records

In this paper we present new insights into Irish Neolithic and Early Bronze Age prehistoric settlement patterns that have been discovered during the last two decades of intensive, developer-funded fieldwork, and compare these results to the arguably more familiar monumental megalithic landscape. The overall aim is to identify points in time and space where settlement patterns were at their most dynamic, in order to examine both demographic and cultural changes and the ways in which the archaeological record challenges notions of population influxes and booms and busts. We focus on a comparison between (1) Mesolithic–Neolithic transitional sites (4300 cal BC–4000 cal BC), (2) Neolithic sites (c. 4000 cal BC–2500 cal BC), and (3) sites of the Chalcolithic and opening centuries of the Early Bronze Age (c. 2500 cal BC–1900 cal BC).

Population dynamics has recently re-emerged as a subject of active debate in prehistoric studies, especially in parts of the world where large amounts of radiocarbon data are available (e.g. Peros et al. 2010; Shennan et al. 2013; Kelly et al. 2013; Crombé and Robinson 2014). The patterns of social and economic change that these studies appear to indicate—particularly oscillating patterns of activity—conflict with traditional views of the steady, slow pace of change in prehistory. Human history from the start of the Neolithic onwards—at least in Europe—has been viewed conventionally as a gradual but continual increase in the scale and complexity of social organisation (e.g. Childe 1925). By contrast, meta-analyses of large radiocarbon datasets and of archaeological sites using Bayesian chronological approaches (e.g. Whittle et al. 2011; Stevens and Fuller 2012; Feeser and Furholt 2013; Armit et al. 2013; Wicks et al. 2014; Whitehouse et al. 2014) appear to indicate a past that was highly punctuated and dynamic. This applies both to particular phenomena and to entire societies, with the latter emerging, flourishing in certain regions, and then undergoing decline, collapse or a radical re-structuring as in decline-and-fall accounts of classical civilisations. These meta-analytical approaches provide a useful means of directly comparing human cultural dynamics with climatic and landscape palaeoecological data (e.g. Whitehouse et al. 2014; Wicks and Mithen 2014; Woodbridge et al. 2014) and thus have great potential for exploring human environmental impacts over the Holocene, and the creation of the cultural landscape.

Criticisms of these approaches as applied to the study of prehistory—particularly the use of cumulative probability distributions of calibrated radiocarbon dates as proxies for human populations—are centred on taphonomic considerations, and the fact that the shape of the calibration curve strongly influences the results (cf. Contreras and Meadows 2014). Archaeological materials are an imperfect record, with highly complex social factors underlying the initial deposition of the material drawn on for radiocarbon dating, whereupon millennia of taphonomic processes eventually result in the ‘what’ and ‘where’ of that which survives (Schiffer 1987). Nor are data subsequently encountered in a neutral manner by archaeologists: many biases exist as the result of research traditions and the practicalities of undertaking fieldwork. Many syntheses of radiocarbon datasets have failed to fully confront these factors, limiting the applicability of these approaches and leading to criticisms within the archaeological literature (e.g. Bamforth and Grund 2012; Contreras and Meadows 2014; Crombé and Robinson 2014; Sheridan and Pétrequin 2014). Here, we present methods whereby the imperfections and limitations of the archaeological record can be better contextualised, understood and taken into account when using meta-analytical approaches centred on radiocarbon dates.

Ireland (Fig. 1), a large European island immediately west of Britain, is divided into two territories; the northeastern quarter being Northern Ireland and part of the United Kingdom, the remainder being the Republic of Ireland. A large number of developer-led excavations took place in the island—especially in the Republic—between the late 1990s and 2008, the timeframe of the most recent economic boom. A large number of prehistoric settlements of all kinds were discovered by this work (Armit et al. 2013; Ginn 2011; Whitehouse et al. 2014). In the wider context of Neolithic Europe, Ireland has a special place, being situated at the endpoint of a series of cultural, economic and demographic transformations which are among the most intensively studied in world prehistory (e.g. Barker 1985; Gronenborn 2003; Rowley-Conwy 2004, 2011; Whittle 1996; Whittle et al. 2011). Despite the newly-discovered wealth of settlement evidence, Ireland’s archaeological record has not yet featured prominently in many of these accounts, except as auxiliary to British research (e.g. Thomas 2004; Stevens and Fuller 2012). This means that the connections between Ireland and its wider world have been relatively poorly explored, except perhaps within narratives concerned with technology and trade (e.g. Cooney and Mandal 1998; Cooney 2000b). Geographically, Ireland, whilst sharing many ecological similarities with Britain, also has a number of distinct biogeographic characteristics such that its records can and should be examined on their own merits (e.g. Mitchell 2005; Montgomery et al. 2014). These distinctions are primarily expressed in an impoverished flora and, particularly, fauna, presenting a rather more limited range of ecological niches for the first human colonisers in the early Holocene (only limited evidence for a Palaeolithic human presence has been found in Ireland, see Dowd and Carden 2016), and impacting on the process of neolithisation. As well as presenting no large game animals other than wild boar (red deer, roe deer and aurochs all being absent), a reduced fauna is likely to have resulted in a very different Mesolithic landscape structure due to reduced browsing pressure (Warren et al. 2014). The paucity of large game might have made Neolithic domesticated cattle, sheep, goat and pig all the more attractive to Irish hunter-gatherers, assuming some indigenous uptake of farming (Woodman 1985). Furthermore, the introduction of domesticated grazers into Ireland would certainly have had a range of ecological consequences, subjecting areas to grazing at levels not previously attained, perhaps ultimately leading to reduced ecosystem function and ecosystem resilience.

Syntheses of development-led excavations have proven challenging, due to the huge volume of raw data, most of which remains unpublished. However, chronology can be a powerful tool to provide an overview of the data (Shennan and Edinborough 2006). The potential of radiocarbon chronology to revolutionize the understanding of prehistory has been anticipated for some time in Ireland (e.g. Woodman 2000). However, as the majority of radiocarbon dates from Irish sites have only become available in the last decade (Fig. 2), the potential of development-led excavations in Ireland to enhance our understanding of the past is only beginning to become apparent.

Here, we consider a database of radiocarbon dates that we have collated using published dates and date lists (e.g. Brindley 2007; Chapple 2008); unpublished site reports, the National Roads Authority’s archaeological database; and direct consultation with private sector archaeologists. The Neolithic period (c. 4000–2500 cal BC) is represented by 1150 radiocarbon dates from 177 sites, with a further 739 dates recorded from 254 Chalcolithic and Early Bronze Age sites (2500–1900 cal BC). These numbers comfortably exceed the minimum 500 radiocarbon dates recommended for this type of analysis (Williams 2012). We have recorded details about material sampled for each date, its context, and various physical and archaeological parameters of each site, including geographical location and typological status, storing this information in a relational database. A compendium of published radiocarbon data and a list of sites is appended to this paper as electronic supplementary data. For analysis, the radiocarbon database was contained in a set of data tables, structured to include a description of material dated and its archaeological context. Each record of the database also contained its archaeological context, including information about the phasing of the site, and the cultural material associated with each date. This database could be related to a list of sites that also contained geographic co-ordinates and the general category of each site, such as ‘court tomb’ or ‘Neolithic house’. Meta-categories were created to allow for all types of settlements, enclosures and burials to be dealt with either separately or together, depending on the research question. In this way, the database could be queried to provide a list of dates for a given phase at a single site, or for entire groups of different sites, or any combination thereof. The date lists were calibrated and summed using custom software written in R.

Location information was taken from the Sites and Monuments Record of the Republic of Ireland and that of Northern Ireland (hereafter collectively referred to as the SMR). We include in our database the details of excavated but as yet undated sites yielding unambiguously Neolithic material. Research on undated Chalcolithic and Early Bronze Age sites is at a less advanced stage, so these were not included in the database except for wedge tombs, cist burials and burnt mounds, details of which could be found in the SMR.

To investigate spatio-temporal patterns, the database was deployed in a Geographic Information System (GIS) and additional utilities were developed specifically for space–time analyses, using GRASS 6.4.1 for GIS operations, MySQL 5.5 for data queries and R 2.14 for radiocarbon calibration (the routine used is supplied as supplementary data) and statistical analyses. Scripts written in these environments were used to select data, plot maps and provide summary statistics. Radiocarbon dates were calibrated within these environments, using the Intcal13 or Marine13 calibration curves as appropriate (Reimer et al. 2013). Summed radiocarbon probability distributions were calculated separately for burial and non-burial contexts, and from charcoal and non-charcoal samples, thereby allowing for old wood samples to be excluded from the analysis if required. For the summed probability modelling, multiple dates occurring within the same phase at a given site were included in the analysis to avoid the risk of rejecting potentially interesting dates that had been erroneously attributed to a given phase—indeed the strength of summed modelling is that no assumptions about the phasing of the data are made. A comparison between the sum of all dates, and those instances where only one date per phase is considered (see supplementary data) indicates that there is little significant difference between the results. We do, however, follow an overall chronological framework for the Neolithic derived through Bayesian approaches already published elsewhere (Carlin 2011b; Cooney et al. 2011; Schulting 2014, 2015; Schulting et al. 2012; Whitehouse et al. 2014). We follow archaeological convention and present Bayesian modelled phase boundaries in italics.

This approach allows for comparisons between multiple data contexts and varying qualities. As a first step in exploring the significance of short-term fluctuations in Irish radiocarbon data, the decay model proposed by Surovell et al. (2009) was used. Surovell and his colleagues derived this model by regressing the Bryson et al. (2006) database of radiocarbon dates from volcanic eruptions against a power function in the form:where N_t is the number of radiocarbon dates for the time period t, and a, b and c are correlation coefficients (see Surovell et al. 2009, pp. 1716–1717). The function can therefore be used to model a ‘taphonomic correction’ to summed probability distributions.

GIS was used to create maps of Ireland where the geographic location of each radiocarbon-dated sample could be visualized according to its context, the type of material dated and the probability that it lies within a given calibrated timeframe. Given the difficulties associated with representing detailed contextual information on such a map, sites were divided broadly into houses, enclosures, burnt mounds, burials and ‘other’ categories. The ‘house’ category includes rectangular, circular and irregular structures; the ‘other’ category consists mainly of ‘pit-and-spread complexes’. Although individual pits are found on many sites, pit-and-spread complexes are those sites where pits and spreads of cultural material are the defining features (Smyth 2012). A set of summary statistics was produced for each category, taking the site as the unit of analysis—for example, multiple rectangular houses occurring together were counted as one site. Multiple phases occurring at one site were treated separately. A succession of maps, plotted using narrow temporal bandwidths (of the order of decades or even years) and viewed as animations, was used to visualize the changing character of the archaeological record. Static maps were also generated for longer timespans, allowing for the comparison of newly-discovered excavated sites and extant, unexcavated and undated sites whose details are known from the SMR.

For analyses of site elevation, GIS was used to store a digital elevation model of Ireland, built using Shuttle Radar Telemetry Mission (SRTM) data at 3 arcsecond resolution. A surface interpolated from this model was sampled at the point coordinates of the various sites, and the result was fed back into the database as a statistic that could be analyzed in tandem with other site attributes. Sites near errant reflections in the SRTM data were identified as outliers and excluded from elevation analysis. An average elevation statistic was calculated for most site categories, except where small numbers and topographic heterogeneity resulted in no clear central tendency. A random sample of 1500 points in the digital elevation model has a median elevation of 86 m and an interquartile range of 55 m–142 m.

What follows is an overview of the evidence for archaeological sites in Ireland, 4300–1900 cal BC. For convenience, the Neolithic period is divided into transitional, Early, Early/Middle, Middle and Late periods, following conventions already established (McClatchie et al. 2014; Whitehouse et al. 2014). We also explore the presentation of spatial–temporal trends in the data without recourse to categorisation by time period, via animated mapping and presentation of overall trends using summary statistics and summed radiocarbon probabilities.

In 1989, the absolute chronology of Ireland’s prehistory was based on 109 radiocarbon dates from 61 sites (O’Kelly 1989, pp. 347–352). In this study, the figure stands at 4679 dates from 1536 sites. These numbers are biased towards the Neolithic period, and the true figures are probably much higher (e.g. Armit et al. 2013). The amount of raw archaeological data available for Ireland has substantially increased, thanks to development-led excavation. The challenge of how to plumb these data for knowledge of the human past can be met by the development of appropriate information systems and analytical tools, such as the methods described in this paper and elsewhere (e.g. Surovell et al. 2009; Shennan et al. 2013; Stevens and Fuller 2012; Armit et al. 2013, Kerr and McCormick 2013). Many of these methods are applicable to other regions and countries working through a similar backlog of excavated sites (Bradley 2012). We have concentrated upon chronological and geographical dimensions in this paper, although a more in-depth study of prehistoric life could also integrate scientific approaches, such as palaeoecology, lipid analysis and other analyses of artefactual material, stable isotope analysis and other biomolecular studies on ancient animal and human bones. Ancient DNA analysis will be particularly useful in testing hypotheses concerning incoming populations, as the pioneering work by Cassidy et al. (2016) has shown. In the developer-led sector in Ireland, none of these techniques has been extensively applied, although this situation is slowly changing with respect to palaeoecology, stable isotopes and lipid analyses (e.g. Murphy et al. 2010; Stuijts et al. 2010; Whitehouse et al. 2014; Schulting et al. 2012; Schulting et al. 2016; Smyth and Evershed 2015).

Despite the wealth of data, gaps and discontinuities in the archaeological record remain, although some of these gaps are as informative as they are problematic. The Irish Neolithic gets rather more enigmatic with time, with the Early Neolithic period being relatively visible archaeologically, and attracting a high level of research interest. The recent discovery that there is only scant evidence for recognisably ‘Neolithic’ activity in Ireland before c. 3750 cal BC has important ramifications, and many of the ideas postulated in this paper depend on this revised chronology. As discussed above, the old wood effect is particularly problematic for Early Neolithic Ireland, and has confused the dating of the earliest Neolithic in Ireland. McSparron’s (2008) demonstration of the short period of use for rectangular houses is therefore of key importance; so too is the increasing awareness of the episodic nature of other types of site (McSparron 2003, 2008; Cooney et al. 2011; Schulting et al. 2012; Whitehouse et al. 2014).

Modern-day development pressures have revealed clusters of Neolithic houses around the northern and eastern coasts—foci of settlement in the Early Neolithic. These newly-discovered clusters coincide with funerary landscapes, and these places seem geographically poised to be likely points of contact with Britain—perhaps reflecting the original points of entry for people bringing with them the new plants, animals and ways of thinking which together constituted the Neolithic. Due to an absence of data, and perhaps a reluctance to broach politically-loaded debates (Stout 1996), Irish archaeology has been generally unable to meaningfully investigate the prehistoric dimensions of Ireland’s cultural and demographic relationships with Britain and elsewhere. We have taken this first step of exploring prehistoric demographics in Ireland with a view to future exploration of both its position on the periphery of Europe, and its links to its neighbours.

The Early to Middle Neolithic transition (Middle Neolithic I) is a key time period in both Ireland and Britain (cf. Bishop 2015; Stevens and Fuller 2012). Evidence for settlement is frequently found, but has received less attention both from researchers and from those working in development-led archaeology. Hence fewer sites have been dated, and understanding of the burial traditions of the period is correspondingly hazy. The pit-and-spread complex has emerged as an important site-type warranting much greater research emphasis both in the field and in the laboratory. Notwithstanding these issues, by considering the large number of sites that date to the period, Ireland supported 10–20 generations of significant numbers of people, from the manifestation of Neolithic sites after 3750 cal BC, until 3500–3400 cal BC, when the archaeological record undergoes significant changes.

Despite archaeological monitoring at thousands of sites in Ireland, and the numerous prehistoric settlements that have been found as a result, secure, well-dated evidence of Middle Neolithic II settlement during the period 3400–3000 cal BC is still to be discovered. This challenges our view of overall lifestyles during the era of the developed passage tomb. We can only speculate about the nature of settlement during this period, as there is so little evidence, and neither pastoralism nor arable agriculture offers a model for the economy of the period that fits the available archaeological and environmental data. There are no directly dated cereals from Middle Neolithic II, so evidence of a cereal economy, if there was one, is absent (McClatchie et al. 2014). Investigation of the faunal remains from the period is also hampered by a lack of suitable material. The limited stable isotope evidence on human remains for this period does not suggest a significant return to marine resources (Ditchfield 2014; Schulting et al. 2012, 2016), but nor does it exclude a decline in the use of cereals or other plants in favour of animals. Analyses of lipids absorbed in Irish pottery indicate dairying was of considerable importance throughout the Neolithic (Cramp et al. 2014; Smyth and Evershed 2015; but note that fewer vessels have been analysed from the later period). This is not incompatible with a mobile or semi-nomadic mixed pastoral economy for Middle Neolithic Ireland, although it should be noted that there is no contrast between Early, Middle and Late periods in this regard. Pastoralism, which seems to offer one potential explanation for the lack of cereals, has been applied to the Middle Neolithic in Britain (Stevens and Fuller 2012; although see also Bishop 2015). However, in Ireland, the commensurate impact on the landscape is not seen in palaeoecological records—many pollen diagrams from this period indicate a phase of less extensive use of the landscape than would be expected under pastoralism, with evidence for forest regeneration in many parts of Ireland, especially westerly areas (O’Connell and Molloy 2001; Whitehouse et al. 2014). The palaeoclimatic context of this period is also interesting, given that the end of the Holocene Thermal Maximum led to a sustained period of lower temperatures and wetter conditions in Ireland and elsewhere in northern Europe (Steig 1999; Roland et al. 2015).

The sudden increase in activity in the Early Neolithic was echoed, to a degree, at the start of the Late Neolithic. New pottery forms and site types—Grooved Ware and henges—were introduced from Britain, although this episode is only one of many hints of reciprocal influence between Ireland and Scotland from the Middle Neolithic period onwards (Schulting et al. 2008; Sheridan 2004).The chronological relationship between Britain and Ireland, in terms of the timing of these influences, is still uncertain and resolving this issue will be a key research question for the immediate future. Furthermore, despite recent progress in ancient DNA, the small sample of Neolithic and Bronze Age individuals analysed by Cassidy et al. (2016) does not inform about potential Late Neolithic and/or Chalcolithic population movements to Ireland—in short, much more work on this front is needed. It remains tempting to postulate that this work, when combined with well-dated archaeological data from Ireland and northern Britain, will reveal clear evidence that there was continual movement of people between the two islands—a process probably as widespread in prehistory as it was during recorded history (e.g. Anderson 1982) and indeed in the present day.

In Ireland, now that burial chronology is becoming much better understood (Brindley 2007; Schulting et al. 2008), the whole of the third millennium BC can be seen to have been characterized by numerous episodes of rapid change in material culture, ceremonial behaviour and burial tradition. Set against this, however, is the steady rate of pit deposition, and settlement types in many ways continue to resemble those of the Early–Middle Neolithic transition. However, new site-types such as stone and timber circles, henges and burnt mounds were also manifest, evidencing wide-ranging successive and contemporary beliefs and practices, some imported and others insular developments. In the near future, we hope that better-resolved observations of this key timeframe can be made through the analysis of archaeological datasets.

The boom-and-bust model has been successful in illustrating that change in prehistory occurred—at least on occasion—at a much greater tempo than has been traditionally acknowledged, and indeed various patterns in the Irish data suggest periods of rapid change in the rates of creation of certain site-types and deposition of materials. It seems plausible that some of these changes were related to demographic events, albeit viewed through a heavy fog of bias. Indeed, whilst it is clear that there are phases in some aspects of the archaeological record, there may also be simultaneous booms, which express themselves differently—for example the shift of focus towards burial in the archaeological record in Ireland around 3400 cal BC. Therefore, without contextualizing archaeological data, such boom-and-bust interpretations of summed probability distributions are an over-simplification, and misleading.

An unknown, despite advances in radiocarbon dating and Bayesian modelling, is the way in which these events were witnessed by the contemporary human population. Changes that seem relatively rapid in the archaeological record may have taken several human generations to fully express themselves, and any awareness of such change must have been made in the context of on-going adaptations to a ‘real world’, where environmental, economic and demographic interests compete, often in a state of dynamic equilibrium but occasionally breaking down altogether.

These glimpses of a more dynamic past are a recent development in prehistoric archaeology, and have been made possible by the availability of increasingly large datasets, as well as by new approaches to their analysis (Shennan 2013). Development-led archaeology in Ireland, sampling the landscape in a rather less biased fashion than is usually the case, has achieved much in redressing the traditional illusiveness of prehistoric settlement evidence, and it is this accomplishment that we wish to celebrate. As the rate of excavation work on the island now increases in line with its economic recovery we may expect frequent new discoveries, some of which will serve to fill the gaps in knowledge that still exist at the time of writing, especially if systematic programmes of radiocarbon dating are instigated, together with a renewed appreciation of the importance of careful sample selection. The combination of contextual and chronological archaeological data (e.g. Figure 9) illustrates just how dynamic past societies were, and for the first time we can more clearly identify the points in time and space where the changes were at their greatest. In Ireland, the points at the start and end of the Middle Neolithic period (c. 3400 and 3000–2900 cal BC) are perhaps the most intriguing of all, and these deserve some of the research attention usually lavished upon the Early Neolithic.