'...where they pass their unenterprising existence...': change over time in the Mesolithic of Ireland as shown in radiocarbon-dated activity

Abstract

This paper presents the first detailed audit and analysis of radiocarbon-dated Mesolithic activity in Ireland. This provides a data set of 119 definite or possible sites and nearly 450 dates. This data set is reviewed to understand how changes in the character of archaeological activity and other biasing factors influence the distribution of radiocarbon dates across space and time. We present very different outcomes than previously published reviews, especially regarding the claimed impact of the 8200 cal BP climatic downturn. This is an outcome of the more reliable data set presented here. Some of the patterns identified are a product of sampling strategies, such as a peak in Earlier Mesolithic activity attributable to multiple dates from Mount Sandel. Others relate to restructuring of activity by hunter-gatherers leading to significant changes to the archaeological visibility of that activity at the change from Earlier to Later Mesolithic. Increasing numbers of radiocarbon dates at the end of the period, alongside the changing character of the archaeological evidence, may indicate low levels of population growth.

Introduction

During the Mesolithic, hunter-gatherer communities settled on the island of Ireland, a large island in the Atlantic northwest of Europe. A cut marked bear patella from the Alice and Gwendoline Cave, Co. Clare, suggests that hunter-gatherers also visited the island in the final parts of the Upper Palaeolithic (Dowd and Carden 2016), but the island appears to have been abandoned during the Younger Dryas, before a new phase of settlement in the early Holocene. Ireland has been an island since prior to any human activity in the area, and the Mesolithic settlement of Ireland was established by sea-crossings.

Because of its island status, Ireland's environment and ecology was different to other parts of Europe, and was impoverished in terms of the numbers and diversity of terrestrial species (plants and animals) present (see Warren 2022 for review). At least some aspects of the Irish Mesolithic have often been understood in the context of this limited ecological diversity, with the development of a stone tool technology which distinguishes Ireland from its European neighbours often considered a kind of adaptation to it. This sense of change is often collapsed into a binary periodisation: an Early/Earlier and Late/Later Mesolithic, with the change primarily defined by changing lithic technology (Woodman 2015). More recently, Woodman proposed that the lithic traditions should be divided into four 'facies' and highlighted ongoing change within stone tool technology within at least some of these phases, but these refined models have not proved influential yet. Beyond this, and although it lasted for approximately four thousand years, the Mesolithic of Ireland is often considered to see relatively limited change over time.

This understanding of hunter-gatherers as comparatively unchanging over time has also been commonplace beyond Ireland (Sassaman and Holly 2011). It reflects much broader societal prejudices. Understood as being part of nature, or as closely adapted to their environments, hunter-gatherers are considered historically static and contrasted to the historically dynamic farmers that shaped and transformed the world to their ends. Our title, for example, uses part of a quote from R.A.S. Macalister, where he compares 'progressive' farmers to 'unenterprising' food gatherers:

Food gatherers are inevitably squeezed out after the arts of food producing have by any means been established in the regions where they pass their unenterprising existence. Thereafter they cannot compete with their more progressive neighbours in the battle for life; sooner or later they must emigrate or perish.

(Macalister 1928, 41)

In Ireland, until recently, this has been exacerbated by a small data set of Mesolithic sites and radiocarbon dates from which to identify changing historical trajectories. This paper presents a new analysis of the substantial data which is now available for the period and a new understanding of change over time in the Mesolithic in Ireland.

Over recent years the application of a variety of modelling tools to radiocarbon dates and/or site densities has been used to identify patterns of change overtime for all archaeological time periods in Ireland and beyond. These so-called 'big data' approaches rely on the increasingly large data sets available and the computational power to model them. In Ireland, models using either Summed Probability Distributions (Armit et al. 2014; Mc-Laughlin et al. 2016) or Kernel Density Estimates (McLaughlin et al. 2018; Hannah and McLaughlin 2019) of radiocarbon dates, have been applied to a wide variety of prehistoric and historic periods. McLaughlin has recently reviewed all archaeological periods in Ireland to provide a potential population model (McLaughlin 2020). Big data models are now common-place accounts of the dynamics and population of Mesolithic Europe (Jørgensen et al. 2020; Solheim 2020; McLaughlin et al. 2021). Such accounts often identify relationships between climate change and prehistoric demography, most notably in discussions of the 8200 cal BP¹ event—the most significant period of climatic decline in the Holocene—although there is some inconsistency in the supposed societal impacts identified in association with this change (see discussion in Warren 2020). In Ireland, this event which started c. 8200 BP, lasted for c. 150–200 years and seems to have led to the establishment of cooler and drier conditions.

Three previous attempts have been made to analyse the frequency of radiocarbon dates during the Irish Mesolithic and have been accompanied by other statements about moments of population change. In 2009 Riede published a summed probability analysis of the very small number of dates available at that time (Riede et al. 2009; Riede 2009). He argued that the 8200 cal BP event led to population decline and cultural isolation in Ireland. Griffiths and Robinson (2018) used Bayesian modelling to assess the potential impact of the 8200 cal BP event and argued that it was not associated with population decline in Ireland. The data set available to both studies was limited, and on this basis Warren (2020) argues that neither set of conclusions are robust.

McLaughlin's recent overall model of Irish population dynamics based on Kernel Density Estimate models includes the Mesolithic in a robust, large data set (McLaughlin 2020). He suggests that the period be understood in two main phases: a long period of 'steady state' from c. 9000 BC to 5200 BC, following which, population grew rapidly until the appearance of the Neolithic c. 3800 BC. The population growth of this period is argued to be connected to a northwest European 'edge-effect' associated with the expansion of Neolithic agriculture into Europe at this time. In commentary on McLaughlin's paper, Cassidy suggests that the increased population might date back to the transition from the Earlier to Later Mesolithic, but it is not clear what her evidence for this claim is (Cassidy 2020a). Baillie proposes a possible association at 4350 BC between a severe climate event identified in oak growth rings and a slight decline in radiocarbon dates in McLaughlin's model (Baillie and Brown 2020). Finally, and although not directly driven by radiocarbon evidence, recent genomic analysis of Irish Mesolithic individuals identifies three important aspects of demography: a significant population bottleneck at some stage within the Mesolithic; that overall population may have been low; and that there was no evidence for inter-breeding with Mesolithic communities beyond Ireland (Cassidy 2020b; Cassidy et al. 2020). It is important to note that the genetic data can only speak to breeding populations, not total population levels. These varied arguments for prehistoric population change are discussed in detail below.

It is useful to highlight some distinctions between the analysis presented here and that of McLaughlin (2020). In part, these are about the review of the data and the definitions used —especially towards the end of the period (see below)—but some of our assumptions are also different. To operationalise his analysis, McLaughlin (2020) made the explicit assumption that radiocarbon dates were a proxy for population—his 'null hypothesis'. Although there are potential problems with the effect of varying archaeological visibility and sampling on the frequency of radiocarbon dates, McLaughlin argued that the very large data set (over 9000 dates) meant that the overall patterns should remain robust. Our analysis does not start from this null hypothesis. We assume that changing frequencies of radiocarbon dates may provide information about prehistoric population levels, but that we must first interpret them in their local archaeological context, with careful consideration of potential issues around visibility and sampling. In part, this reflects the significantly smaller overall data set available for the Mesolithic, but also recognises the issues in dealing with hunter-gatherer archaeology in Ireland, which is frequently characterised by small, ephemeral sites where issues of sample bias are significant. As we discuss below, in this context it is highly significant that the changes associated with the Earlier-Later Mesolithic transition are manifested in changing patterns of archaeological visibility.

This paper provides the first systematic review, analysis, and extended discussion of the corpus of radiocarbon dates from Ireland. We first outline the methodology used for the audit and models, before reviewing some overall issues with the data set. We then review the radiocarbon record, before presenting a narrative of change over time. It is important to emphasise at the outset that this analysis would not be possible without the remarkable quality of the Irish data set, and the ease of access of most of it. Large parts of this data set are the product of excavations in advance of development in recent years, and this work, and the scale of data recovered, has begun to transform our understandings of the Mesolithic. Our evidence leads us to strongly reject Waddell's recent statement that for the Mesolithic: '...the extraordinary expansion of archaeological work in the commercial sector over a decade ago (c. 1997–2007) has confirmed the island-wide presence of hunter-foragers but has not appreciably advanced our knowledge of the period.' (Waddell 2020, 54) and Woodman's comment that this contribution 'is not as great as might have been expected' (2015, 79).

Methodology (audit)

Analysis commenced with data provided by Chapple's regularly maintained date lists (Chapple 2021). These were initially reviewed with all dates prior to c. 3750 cal BC included. This marks the end of the Early Neolithic 1 as defined by Whitehouse et al. (2014), a period for which there is still some potential for ongoing Mesolithic activity (Smyth et al. 2020). Pre-Holocene dates on a cut-marked bear patella from the Alice and Gwendoline Cave (Dowd and Carden 2016) are excluded from this analysis.

An initial 651 dates were returned, including many which were not archaeological (i.e., cave fauna or peat dates). These were removed from analysis. A more detailed grey literature review followed, supplemented by liaison with colleagues across the archaeological profession and the provision of dates from unpublished excavations. Dates were quality checked for accuracy, including consultation with laboratories where there were inconsistences in publications. Dates were classified by reference to their cultural association as being either 'Mesolithic', 'possibly Mesolithic', not having any artefacts associated ('no artefacts'), or 'indeterminate' (after c. 5300 uncal BP, and with ambiguous cultural associations). For example, dates from domestic cattle from Ferriter's Cove (Woodman et al. 1999) were classified as 'indeterminate' whereas distinctive and characteristic Later Mesolithic activity on the site was classified as 'Mesolithic'.

Our definition of 'indeterminate' recognises that the data available contains considerable ambiguities in classifying sites near the Mesolithic-Neolithic transition. Without accessing primary site archives and materials it is not possible to identify whether some sites are best categorised as 'Mesolithic' or 'Neolithic' in this period, or indeed whether these terms are meaningful. Our analysis in this paper therefore does not broach discussion about the processes of the Mesolithic-Neolithic transition in Ireland. A more focused re-examination of this period is ongoing by Warren.

In an attempt to better understand the archaeological context of the data, dates were also classified according to whether they could be more closely associated with an 'event' (i.e., a hearth or pit fill), a 'non-event' (occupation soil or spread), or an 'artefact'. Whilst there was a degree of subjectivity in making these definitions, and a degree of overlap between categories, this exercise divided the data into three usefully independent subsets. Our goal of investigating the impact of changing behaviour and site visibility could then be met through comparing these subsets with each other and the overall signal, as well as through inter-regional patterns. Information on material dated was also recorded, as was any associated palaeodietary information.

Full details of the methodology are provided in Appendices 1 and 2. The full data set used in analysis is provided in Appendix 3, with the exception of four sites (Derragh, Doolin, Fanore 1, Fanore 2) where information was provided in advance of publication and agreed not to be released into the public domain.

Methodology (model)

The data and metadata described above were used to generate time series models of relative activity. This was done using a variety of tools, most notably the method for compound Kernel Density Estimation (KDE) developed by McLaughlin (McLaughlin 2019; Hannah and McLaughlin 2019; McLaughlin et al. 2021). The KDE technique is similar in principle to the more familiar summed probability distributions (SPDs), but the method has the advantage of being an effective low-pass filter, thus suppressing calibration curve effects and enabling analysis of growth rates, which can be estimated via numerical differentiation of the KDE. The uncertainty of the KDE models is expressed as a confidence envelope and was calculated through repeated recalibration and 'Monte Carlo' resampling of the radiocarbon dates. The bandwidth of the KDE models was set at 100 years unless detailed otherwise. Multiple dates occurring at the same site within 100 years were thinned using a version of the hierarchical cluster analysis method of Bevan and Crema (2018). Dates were calibrated using the IntCal20 terrestrial and marine curves as appropriate (Reimer et al. 2020; Heaton et al. 2020). Custom calibration curves for mixed terrestrial/marine consumers were calculated and employed for the calibration of these samples, using mixing ratios estimated via the palaeodietary carbon-13 isotope data originally published alongside each date. Our methods do not apply outlier models to charcoal dates because it is unlikely that built-in wood age can significantly influence radiocarbon density estimates in an Irish context (Hannah 2021). We also produced SPDs and histograms of the data for the purposes of comparison. The dependence of inter-site distance with time was estimated using the site location data associated with each date in the database. Computer scripts that reproduce the analysis shown below are presented as Appendix 2.

Overall influences on the data set

The changing frequency of radiocarbon-dated Mesolithic activity over time is in part a record of prehistoric activity, but is also influenced by significant biases and filters, not least given large-scale palaeogeographic change within the Mesolithic. We emphasise these here before outlining them in detail in the narrative below. One key factor is changing sea levels. Put simply, sea level rise since the onset of the Mesolithic means that in many parts of Ireland, areas that might have been settled by hunter-gatherers are now submerged (Edwards and Craven 2017; Warren and Westley 2020; Westley and Woodman 2020). Because of differential glacial isostasy, or the rebound of the earth's crust following the removal of an ice mass, in parts of the north and east of Ireland shorelines from the later parts of the Mesolithic survive, but this is not the case in the south. Given that these coasts would have been a preferred location for settlement (for review, see Warren and Westley 2020), this loss is highly significant in influencing our understanding of changing site frequencies over time. On a priori grounds, we might expect to find more sites in the later parts of the Mesolithic because these shorelines are preserved and earlier ones are not. Indeed, the overall reduction in land mass of the island of Ireland over time would also lead to increased site and radiocarbon date frequency in the modern landscape even if activity levels and past population remained constant, simply because this activity took place within a smaller area.

Archaeological sites are subject to processes of destruction from the moment they are created and again, on a priori grounds, we should therefore expect more sites from more recent periods and less from older ones. This has been demonstrated empirically by Surovell et al. (2009) who proposed a correction factor to apply to density models of archaeological data. We have not applied this to our data as it unlikely to be applicable to a region like Ireland where archaeological features are found in various sedimentary contexts each with different taphonomies (cf. Ward and Larcombe 2021), but for this reason alone we can anticipate that we might find fewer early sites than later ones.

As will be detailed below, the distribution of radiocarbon-dated activity in Ireland is uneven, with a small number of sites having a very large number of dates. This is compensated for in the model, to an extent, through the application of cluster analysis. However, in any situation where a large number of dates have been obtained from a site, it becomes much more likely that those dates will identify multiple phases of activity. It is difficult to remove this effect from the models without reducing their sensitivity. For example, in Fig. 1 the spike in all dates and phases c. 7700 BC is created by the very well-dated site of Mount Sandel (Bayliss and Woodman 2009; Woodman 1985). Given that most well-dated sites are in the Later Mesolithic, this has important, if subtle, impacts on our models, as discussed below.

Many of these issues are controlled in part by considering the first-order derivative of the KDE models, which allows us to identify periods of significant growth or decline (Fig. 2). In doing so we are directly modelling the relative changes in the KDE at each given instant, rather than its absolute level.

Fig. 1.

Kernel Density Estimate (KDE) models indicating the relative abundance of radiocarbon dates and radiocarbon-dated site phases. Kernel bandwidth is 100 years.

Finally, although many approaches to demography seek to establish correlations with climate and other forcing events, assessing synchronicity between different proxies remains challenging, even before an assessment of causation can be advanced. Radiocarbon dates themselves may contain a dating lag through the potential use of older material on a site, and the 'old wood' effect, and this lag can be exacerbated by how the KDE modelling process applies its Gaussian model of sampling error symmetrically across each element in the model. Consequently, in the analysis that follows, synchronicity is only interpreted broadly, and we have deliberately avoided being too specific in analytical approaches to correlation.

Fig. 2.

Top panel: KDE model of site phases with statistically-significant periods of growth and/or decline shaded. Lower panel: dynamic growth rate model in annualized geometric growth rate (%).

Radiocarbon dating the Mesolithic in Ireland

Overall numbers of dates

The first conclusion of our analysis is that there are considerably more Mesolithic dates than might be supposed, with 119 definite or possible sites and nearly 450 dates (Table 1). The 119 figure includes all sites with any kind of Mesolithic date (possible or certain). Eight sites have both Mesolithic and possibly Mesolithic dates from different phases, giving a total of 127 Mesolithic and possibly Mesolithic sites. Fifty-five of these sites are unequivocally Mesolithic and 26 fall in the indeterminate category at the end of the period, many of which are likely to be Mesolithic. The 'possibly Mesolithic' and 'no artefact' sites are best considered in terms of their relationship to the overall distribution of dates (Fig. 3).

Table 1.

Overall cultural classification of radiocarbon dates. See text for discussion of total site numbers.

Possibly Mesolithic

Dates identified as possibly Mesolithic broadly match the overall distribution of Mesolithic activity except for a sharp increase in the centuries prior to 6000 BC followed by an equally sharp decline. Although not as marked, this pattern is also present in the Mesolithic data set. This growth in 'possibly Mesolithic' dates includes five from Gortore 1b typically modelled by the algorithm as three distinct site phases. Mesolithic activity here is argued by the excavator to be disturbed and mixed with Neolithic material (O'Donoghue 2011). However, the prominent spike is not an artefact of the relatively high frequency of dates from this one site, as it also occurs (albeit somewhat less dramatically) when the site is excluded from the analysis (see Appendix 2). We therefore consider the 'possibly Mesolithic' and 'Mesolithic' sites together in our models.

No artefacts

A total of 22 sites presented Mesolithic dates but without any associated artefacts (Table 2). Caution is clearly needed in interpreting this material, as there is a possibility that some of it may not be of anthropogenic origin or directly associated with its contexts. However, in some instances, hearths and larger quantities of charcoal are unlikely to be residual, and in all cases the original excavators consider these dates reliable. The features identified in this manner include pits (ten or eleven), hearths (three or four), structural features (three), post/stake holes (three) and spreads/fills (two). The 'no artefact' sites are only present from c. 6800 BC. They then grow in frequency in the mid-sixth millennium BC, declining again before becoming more frequent again in mid-fifth millennium BC. This is not wholly in keeping with the behaviour of dates with Mesolithic cultural association, but the differences are minor and the sample sizes are small. Again, in the models below we incorporate the 'no artefact' sites, highlighting differences where appropriate.

Fig. 3.

KDE models of the relative frequency of archaeological radiocarbon dates from samples with different classes of artefactual association.

Two scenarios can be evoked to explain the comparatively late start within the Mesolithic for sites without artefacts:

• -. The first is that sites of this kind were present throughout the Mesolithic but are only recognised after c. 6800 BC, possibly because earlier features were heavily transformed by taphonomy and hard to recognise. Whilst plausible, this interpretation is weakened by the relatively clear archaeological features that have been identified in the field, plus oscillations in the density model following 6000 BC. It is hard to see how a purely taphonomic filter would explain such patterning.

• -. The second explanation is that this reflects changing behaviour within the Mesolithic, and the generation of sites with little artefactual record is a new aspect of behaviour. This might be related to the broader changes in stone tool production, use and discard ongoing at this time (see below for discussion). This is our preferred explanation.

Table 2.

'No artefact' dates.

Discussion

Our data set of 119 radiocarbon-dated sites and c. 450 dates is almost 50% larger than the 300-325 dates identified by Woodman (2015, 120), probably reflecting the addition of more recent excavations. McLaughlin (2020) identifies c. 550 dates from the Mesolithic (6% of a total of 9076). Given slight differences in methodology, the difference between our sample sizes is probably not meaningful. Including those dated on typological grounds, Woodman (2015) argued that there are c. 130 Early Mesolithic and c. 250 Later Mesolithic sites known in Ireland.

Radiocarbon dates in Mesolithic archaeology in Ireland: what was dated, when and where?

How many dates per site and what was dated? Both the total numbers of dates and the number of dated sites show the transformative impact of the boom in commercial archaeology in the early 2000s, with 59.7% of all sites radiocarbon dated to the Mesolithic having been excavated in 2000–09, most of these being commercial excavations (Table 3). The lesser numbers of dates and sites in 2010–19 reflects declining archaeological activity following the economic crash of 2008 and, possibly, some more recently excavated sites not having fully completed reporting. The significant drop in the number of sites excavated in the 1980s should be noted. In total, 58% of sites dated to the Mesolithic were excavated in advance of development, clearly refuting the statements of Woodman and Waddell cited above that commercial archaeology has made little contribution to the period beyond expanding the geographical range of Mesolithic activity. Such sites dominate our data set.

Table 3.

Date of excavation of sites with Mesolithic radiocarbon dates. Sites excavated in, for example, the 1970s and later redated are classified by the date of their original excavation. Sites excavated across two decades are classified as the decade when most excavation took place.

The 447 dates from 119 sites means that on average each site has 3.7 dates, but this obscures an uneven distribution (Table 4). The majority of sites (75/119 or 63%) have only one date and a further twenty have only two or three. Many of these are commercially excavated sites. This demonstrates the transformative impact of recent developer funded work, but also the limitations posed by finite post-excavation budgets, as complex multi-period sites must often be dated with a comparatively small number of expensive radiocarbon samples. Most significantly, eleven sites have ten or more dates (Table 5), accounting for 56.1% of all dates, with one site alone (Derragh) having 72 dates, 16.1% of the total for the whole island.

Charcoal of varied kinds dominates the material dated, with increasing use of single entity samples over time, although it is not always possible to closely assess the species dated, even from recent excavations. Overall, approximately 39.6% of the 447 dates are on short lived charcoal or charred plant materials, 14.3% on wood, 11% on bone, and 10.1% on shell with the remainder on bulk charcoal (7.6%), charcoal (9.4%) or unknown materials (0.4%).

Table 4.

Number of dates per site.

Table 5.

Sites with 10 or more radiocarbon dates. Period is an approximate range to indicate the duration over which these dates fall and does not indicate continuous activity.

An assessment was made of the relationship between the date obtained and the context dated (Table 6). Approximately even numbers of 'event' (e.g., a hearth) and 'non-event' (e.g., an occupation soil) context dates are available. Given that organic artefacts are often considered to be comparatively rare in the Mesolithic it is surprising that 45 (10.1% of total) dates are directly on artefacts. These are predominantly fish traps from Clowanstown (Mossop 2009; Mossop and Mossop 2009) and North Wall Quay (McQuade and O'Donnell 2007), wooden stakes from Derragh (Breathnach, pers. comm.) and bone points from Cutts and other River Bann sites (Woodman 2015; David et al. 2015). The 45 artefact dates come from nine sites, but 38 of them come from three of these (Clowanstown, Derragh and Cutts). This high proportion of directly dated artefacts is unexpected.

Table 6.

Contextual association of the dates.

Where are the dated Mesolithic sites? The geographical distribution of dated sites is not even (Table 7, Fig. 4). At county level, although on average each county has four dated Mesolithic or possibly Mesolithic sites, and fourteen dates, five counties (Cavan, Fermanagh, Laois, Monaghan and Roscommon) have no Mesolithic dates or sites identified in this review; six have only one site, and three have only one date. In contrast, Antrim has fifteen sites and 48 dates, Meath fifteen sites and 28 dates, and Tyrone thirteen sites and fifteen dates. There is considerable variation in the average number of dates per site, with six counties only having one date per site. Paradoxically, the counties with the highest numbers of dates per site (Clare, Longford and Mayo) are counties which have little evidence for Mesolithic activity apart from recent, well-dated sites. This clearly shows how research traditions influence our data set.

At the level of provinces, the disparity becomes even more stark, with Leinster and Ulster having a higher proportion of dates and sites than we would expect based on their land surface, and Connacht very significantly under-represented (Table 8).

Comparison of the Mesolithic dates with post 5000 uncal BP dates from the Chapple database confirms the uneven distribution (Table 9).² The distribution of numbers of dates and sites by county is significantly different for the Mesolithic than for later periods. As previously argued (McLaughlin 2020) the distribution of radiocarbon-dated Mesolithic archaeology at county level across Ireland is not the same as for all other periods of Irish archaeology. In some regions, especially in Ulster, more Mesolithic sites and dates are available than would be expected from the overall national sample. The high proportion of sites and dates in Ulster is likely to be a product of historical research bias and the preservation of Mesolithic shorelines in this region. Tyrone is unusual, having a much higher proportion of Mesolithic sites than might be expected, but also somewhat fewer dates. This reflects the methodological context of the fieldwork

Table 7.

Distribution of dates by county.

discussed above, as many of the Tyrone sites were discovered during recent road schemes, with most sites known by single dates; the influence of archaeological practice on our data set is highly significant.

Fig. 4.

Overall distribution of radiocarbon dated Mesolithic sites in Ireland (left panel) and per time period (right panels).

Counties with significantly lower than expected numbers of Mesolithic radiocarbon dates are mainly in Connacht. This is not just because of an absence of archaeological work in the region; the numbers of Mesolithic radiocarbon dates and dated sites in this area are lower than what we would expect given the numbers of sites and dates of other periods. The county with the fewest Mesolithic sites and dates relative to expectation is Tipperary, where a substantial amount of archaeological research has been undertaken. This under-representation may reflect the difficulty of identifying sites in flint-poor regions of Ireland, as for example in Mayo where quartz is the predominant raw material (Warren 2009), but it remains possible that Mesolithic activity in these areas was less intensive. In Meath the very significant lower proportion of Mesolithic radiocarbon dates, but not sites, likely reflects very substantial dating programmes on archaeological sites of later date in this region, both through development and research excavations.

Table 8.

Distribution of dates by provinces.

Table 9.

Comparison of county distribution of Mesolithic and post Mesolithic radiocarbon dates. Counties where Mesolithic sites or dates are ≥ 5% more common than expected are highlighted in grey and in bold. Counties where Mesolithic sites or dates are ≥ 5% less frequent highlighted in grey and in italics.

Variation in the growth models for the different provinces (Fig. 5) is significantly impacted by sample sizes and individual well-dated sites, such as Fanore 1 and 2, Doolin and Ferriter's Cove for Munster, all mainly falling into the fifth millennium BC. Regional variation is discussed below, but caution is needed in its interpretation.

Summary

In summary, a surprisingly high number of radiocarbon dates are available from Mesolithic sites in Ireland. These include a distinctive group of features which are not associated with any artefacts, and may be a new development in behaviour during the period. A significant proportion of the dates are directly on artefacts. Many sites have very few dates, so we must be careful not to allow the small number of sites with many dates to skew our interpretations. The geographical distribution of Mesolithic radiocarbon dates is distinctive compared to other archaeological periods in Ireland, with an especially notable absence in Connacht. Recent archaeological work in advance of development has made a significant contribution to this data set. The data set is also clearly not a direct reflection of past population but is substantially filtered and constructed by archaeological activity in the present, including excavations and other research projects.

Change over time

Our data audit and density models provide the following indications of change over time in the frequency and distribution of radiocarbon-dated activity in Ireland, when the data are modelled using individual site phases. The overall growth model is presented in Fig. 2. Fig. 6 shows changing average distance between dated Mesolithic activity over time, and Fig. 7 combines the overall date frequency, growth rates and distance, comparing these to the nGRIP climate proxy. We divide these models into four broad periods:

• -. Prior to 7700 BC

• -. 7700 to 7000 BC

• -. 7000 to 5500 BC

• -. 5500 to 4000 BC

Prior to 7700 BC

Dates falling in this period either have very large date ranges (for example Lough Boora UB-2268 8980±360 BP; Ryan 1980) or are only 'possibly Mesolithic', including the possible platform at Ballyoran (Tierney et al. 2013) and the cut-marked bear vertebrae from the Catacombs, Co. Clare (Dowd and Carden 2016). Apparent change over time is not meaningful. It is not clear that permanent settlement of the island was established at this time.

Fig. 5.

KDE models of the relative frequency of Mesolithic radiocarbon dates in the four provinces of Ireland. Statistically significant periods of growth and decline are highlighted.

7700 to 7000 BC

The model shows a distinctive period of growth in the centuries surrounding 7700 BC, rapidly followed by a period of decline following 7500 BC. This correlates to the first clear evidence for the establishment of settlement in Mesolithic Ireland at Mount Sandel Upper (Woodman 1985), henceforth simply Mount Sandel. Superficially, this might be considered an attempt at establishing settlement on the island followed by collapse, an interpretation in line with recent genetic evidence of a very substantial population bottle neck at some period within the Mesolithic (Cassidy et al. 2020; Cassidy 2020b). However, the growth and decline seen in this model is much more likely to be an artefact of the very extensive number of dates available for Mount Sandel in the context of a very small overall number of sites dating to the earlier half of the eighth millennium BC.

Fig. 6.

Dynamic model of settlement density derived from the average distance between all sites in a 100-year rolling time window, stepped through the dataset at 5-year intervals, and smoothed using a locally-weighted scatterplot (LOESS) regression model.

Some of the sites falling within this period are either dated to very low resolution, or are not clearly Mesolithic. Mount Sandel Lower is represented by two dates, both with errors >200 years; Castleroe has one date from bulked charred material (Woodman 1985). There are also possible Mesolithic dates from two sites: wood from basal sands at Newferry, with a few artefacts and possible placed stones (Zone 9; UB-487 8190±120 BP; Woodman 1977); and a possible, but very doubtful, date on brushwood from a crannog at Lough Gara cited by Fredengren (8160±50 BP, no code provided; Fredengren 2002).

Setting these dates aside, the key sites dated to this period are Mount Sandel, Hermitage and Lough Boora. Bayesian modelling of the activity at Mount Sandel suggests that the three main activities of huts, little pits and big pits all fall within the period c. 7750/7670–7560/7510 BC (Bayliss and Woodman 2009). Two dates from a cremation burial Pit A at Hermitage (c. 7500–7200 cal BC); ten dates on a mixture of bulked and single entity samples from a lake edge settlement site at Lough Boora (c. 7600–7200 cal BC); and one date from an axe manufacture site at Doolin (c. 7340–7075 BC; most of the dates from this site are later) are likely to fall after the c. 7600 BC peak, which is represented only by Mount Sandel and Castleroe. Rather than genuinely reflecting changes in prehistoric behaviour, we therefore think it much more likely that this peak is an artefact of the large number of dates from Mount Sandel and the consequent identification of multiple phases of activity at this site. Few sites are reliably dated to this period.

Fig. 7.

KDE model for archaeological site phases compared to its first derivative, the dynamic growth rate, settlement density, and the nGRIP δ18O palaeotemperature proxy. Note the so-called 8.2kyr BP event occurring on the BC scale around 6250 cal BC.

7000 to 5500 BC

The period c. 7000 BC to 5500 BC shows initial stability in the numbers of dated phases, before a slight decline prior to 6500 BC and very significant increase following c. 6400 BC, continuing to c. 6000 BC when a slight decline takes place to c. 5500 BC. The period c. 7000–6500 BC also sees a significant fall in the average distance between contemporary phases, and the period following 6500 BC sees a clear increase in the average distance between phases. The period c. 6500–6000 BC therefore sees a slightly paradoxical increase in both the number of dated phases and increased dispersal of these phases. That this is broadly contemporary with the 8200 cal BP event makes it imperative to better understand the processes at play. This period also sees significant technological shifts, including the much-discussed transition from Early/Earlier to Late/Later Mesolithic which includes the shift from the use of microliths to a diverse lithic industry including the use of broader blades and flakes (Woodman 2015; Warren 2022).

There is a slightly higher proportion of 'possibly Mesolithic' sites in this period, including pits disturbed by later activity (Gortore 1b; O'Donoghue 2011), a possible structure (Burnchurch; see Appendix 3), and dates from Woodman's archive for which we have little accompanying detail (Cairncastle Road; see Appendix 3). As noted above, it is difficult to understand why these possibly Mesolithic dates would cluster in this period, and it is not considered likely to determine the overall pattern of activity. The mid-seventh millennium BC sees a slight rise in the frequency of direct dates on artefacts, all of which are bone points from the River Bann. The period immediately prior to 6000 BC also sees an increase in the number of 'events' dated, although some of these are only 'possibly Mesolithic'. There is little to no change in the frequency of 'non-events'.

It is possible that the decline in the inter-site distance indicates a period of clustering of settlement c. 6750–6500 BC, followed by an increase in the archaeological visibility of activity in Ireland including both more sites being present over time and some evidence that these are dispersed more widely. The increased numbers of dates over time is most marked in Ulster and Leinster, and is not apparent in Connacht (Fig. 4), presumably because of the small number of samples here. That so many of the dates for this period are from one broad geographical region makes the increased inter-site distance even more surprising. The changing frequency and dispersion of dates does not appear to be driven by significant changes in the type of activity being dated, although the impact of the bone point dating programme may influence our data and this is the broad period when no-artefact sites appear in our data. These changes are likely to have resulted from reorganisation of the relationships that linked communities across space.

Whilst these significant changes are broadly synchronous with the significant climatic deterioration and change of the 8200 cal BP event, the modelling processes mean that precise synchrony is difficult to establish. Taken at face value it seems most likely that the processes that changed the archaeological visibility of radiocarbon-dated activity were initiated prior to the climatic change, although they may have been amplified by it. In any case, the period is marked by increased frequency and dispersion of activity. It is possible that the slight decline in activity following c. 5900 BC reflects changes at the end of this period of climate change, but this is far from certain.

It is important to consider the changing radiocarbon data in the context of the changing character of the archaeological record, notably changes in stone tool technology at this time. Woodman (2015) argues that microliths were falling out of use by c. 7000 BC (the Creagh facies of the Earlier Mesolithic), and that a shift to hard hammer production of larger blades and flakes took place by c. 6800–6600 BC (the Cushendum facies of the Later Mesolithic) with the classic distinctive retouched forms and uniplanar cores of the (flint-based) Later Mesolithic present from c. 6000 BC (the Newferry facies of the Later Mesolithic).

The changes in technology are best understood in the context of changing constraints on mobility and the changed social relationships enabled by the removal of these constraints (Costa and Marchand 2006; Costa and Sternke 2007; Costa and Sternke 2009). Earlier Mesolithic technologies required high quality raw materials which were spatially limited, constraining social groups either through the need to structure direct mobility to sources or to maintain long-distance exchange networks to access these materials. These technologies were also tightly structured in terms of core reduction and working. The dynamic changes that established Later Mesolithic technologies through the seventh millennium BC included and enabled the greater exploitation of a wider range of raw materials. The more flexible Later Mesolithic technologies removed the constraints on mobility.

It appears that this important process of change is also manifest in our radiocarbon data, with the removal of constraints on mobility presumably indicated by the increased inter-phase distances and increased numbers of sites being identified, including some which are not characterised by any artefacts. This confirms earlier hypotheses that the technical change should be associated with broader processes of transformation in Mesolithic society.

The later part of the seventh millennium BC therefore sees significant changes in the nature of the archaeological record of the Mesolithic in Ireland. Changes in technology, which had been ongoing since c. 7000 BC led to a significant reduction in constraints on mobility caused by a reliance of high-quality raw materials. The development of more flexible approaches to working stone, including the exploitation of a wider range of raw materials, seems to have been associated with, or possibly caused or enabled, increased dispersal and frequency of activity at this time. These changes were already underway when Ireland saw significant climate change at 6200 BC, which saw a move to cooler, drier conditions. These do not appear to have immediately changed the trajectory of dispersal. The 8200 cal BP event may have increased the tempo of change, but it is difficult to provide any meaningful detail that might provide a causal link between changes to climate and any shift in subsistence or settlement.

Finally, it is also difficult to understand how to reconcile our evidence of increased dispersion and activity with previous suggestions that the 8200 cal BP event led to significant decline in the population of Ireland (Riede 2009; Riede et al. 2009). It is possible that the increased dispersion and activity is a result of new mobility strategies in the context of population change, but demonstrating this would require a much more detailed argument; it is not our initial interpretation of this data.

5500 to 4000 BC

The period from c. 5500 BC to the end of the Mesolithic sees a significant increase in the numbers of phases dated, with notable periods of increase in the centuries prior to 5000 BC and c. 4700–4500 BC. The centuries following 5000 BC also see a significant increase in the dispersal of dated phases across the landscape, which then declines slightly throughout the fifth millennium BC. This period corresponds to the significant growth following 5300 BC previously identified by McLaughlin (2020) and argued to be paralleled elsewhere in northwest Europe and an 'edge effect' caused by the arrival of the Neolithic in the region. Understanding this period first requires careful consideration of biases in our data.

It is important to note that some sites with multiple radiocarbon dates lie in this period (Table 5) including the most frequently dated site, Derragh. As noted in our discussion of Mount Sandel, this one site slightly increases the number of phases of activity in the density model around 5000 BC. However, given the larger data set available for this period, which includes many more sites, the effect is not particularly significant. The same dynamic was present in McLaughlin's (2020) models which did not include dates from Derragh and running our models without Derragh shows that it is not significantly altering overall dynamics (see also notes in Appendix 2). Dates directly on artefacts increase significantly from c. 5600 BC to 5000 BC, declining thereafter and these come almost exclusively from the extensive dating programmes on fish traps at Clowanstown (Mossop 2009; Mossop and Mossop 2009) and stakes at Derragh. But artefact dates are only a small proportion of the whole, and the trend to increase is also apparent for 'non-event' and 'event' dates, the latter seeing an especially marked increase.

The impact of sea level change may be important in considering this period (see above). Many of the most important Later Mesolithic coastal sites date to the very latest parts of this period. Ferriter's Cove (Woodman et al. 1999), Fanore (Lynch 2017), Rockmarshall (Mitchell 1947; Mitchell 1949) and Belderrig (Warren 2009), for example, all include activity c. 5000-4000 BC, but it is hard to assess whether this is a new form of settlement, or an outcome of the loss of earlier coastlines. The midden and associated activity at Baylett, in contrast, dates from c. 5700 BC to 4300 BC (not necessarily continuously), suggesting that coastal occupation was a feature of earlier periods as well, and that shoreline preservation may be shaping our data to some extent.

These two biases may account for some of the increase in the visibility of activity, but would not directly explain changes in the character of that activity. Evidence in this period is diverse, with a wide range of dated activity, including pits, hearths, structural features, middens and artefacts. However, it is hard to assess whether this diversity is a genuine historical development or a product of our increased sample size: the small number of sites in the eighth millennium BC, for example, are also strikingly varied in character. Two types of activity do appear to be new in this period and their appearance is difficult to explain as a product of bias: i) the construction of lake edge platforms; and ii) the treatment of human bone. Although not formally part of this review, it is interesting to note that Moynagh points may also be a very late Mesolithic innovation.

Aside from one questionable early date from a possible platform crannog at Lough Gara discussed above, artificial lake edge platforms appear to reliably date to this period. These include Derragh (c. 5500/5300 BC–c. 4000 BC); Clowanstown (fish traps and artefacts dated to c. 5300–4700 BC, earliest direct dates for platform c. 4300–4000 BC) and Moynagh Lough (4300–4000 BC; Mossop 2009; Mossop and Mossop 2009; Bradley 1991, 2001). Whilst activity on lake edges is known from the earliest parts of the Irish Mesolithic (i.e., Lough Boora) these ways of occupying and modifying lake edge environments through the construction of platforms (Blinkhorn and Little 2018) appear to be innovations at this time.

This period also sees a reappearance of human remains in the archaeological record. These were known from Earlier Mesolithic contexts (Killuragh and Hermitage), but with the exception of one poorly understood femur from the Cutts at c. 6000 BC, were then absent until the mid to late fifth millennium BC (Ferriter's cove, Rockmarshall and Killuragh). It is difficult to understand the absence of human remains for c. 1500 years as simply the product of bias. This may indicate changing treatment of human bone in the Mesolithic, both when they stopped appearing in the archaeological record and when they reappeared. The situation in Britain is broadly comparable. Human remains are more common early in the Mesolithic, then become less frequent, and only reappear in the Neolithic, with the exception of human remains from the very late Mesolithic shell middens on Oronsay (Conneller 2021; Meiklejohn et al. 2011).

This period therefore may see both an overall increase in the levels of archaeologically visible activity, and new forms of behaviour. This may correlate with population increase at this time, as argued by McLaughlin (2020). McLaughlin posits that this period of growth in the Late Mesolithic is a common feature of the margins of Europe and can be understood as an 'edge effect' associated with the first appearance of Neolithic populations and practices in the wider western European region. This transformation was associated with loss of habitat and prey, and therefore the fragmentation and dispersal of existing hunter-gatherer groups. Whilst this process eventually led to the extinguishment of Mesolithic cultures, over the shorter term such pressures may have been met with creative resilience, including economic and technological innovation, and migration.

Indeed, for many researchers, considering Ireland as part of a regional 'edge effect' would run contrary to the dominant models which propose that Ireland was culturally, as well as genetically isolated in the Mesolithic, cut off from contact with the Mesolithic communities of both the island of Britain and mainland Europe (for a review, see Warren 2015). These arguments have mainly been based on the differences in stone tool technology, and the absence of clear evidence for imports of distinctive materials across the Irish Sea. They have received a fillip from the recent demonstration that Irish populations were genetically distinctive and showed no evidence of inter-breeding with other groups. However, they do not adequately account for a wide range of specific shared cultural practices that link the Irish and European Mesolithic periods, including specific forms of treatment of the human dead, forms of jewellery, and important aspects of stone tool technologies (for discussion see Warren 2022). It is likely that, although at a low level and in a form that did not involve breeding relationships, Ireland was in contact with the rest of Europe and can be considered in terms of potential edge effects, whatever these might be.

Discussion

Our new data set allows a reconsideration of the Irish Mesolithic from perspectives that were not available before. We have identified three periods of enhanced archaeological visibility and some associated falls. The first, at c. 7700 BC, was most likely an outcome of unavoidable bias caused by the large number of dates from the important early site of Mount Sandel. The second, from c. 6500 BC onwards, is associated with significant changes in mobility, technology, and the character of the archaeological record. And finally, the later parts of the period see significantly increased visibility and some innovations. Here we review this against two related themes: firstly, population levels and dynamics; and secondly, other historical dynamics of change.

Population

The resurgence in popularity of using radiocarbon and other data as proxies for past population levels has led to increased awareness of the distinctions between hunter-gatherer demography as witnessed archaeologically and ethnographically (Tallavaara and Jørgensen 2021). Ethnographic data suggests that hunter-gatherer populations can expand at a rate of 1–3% per year, but over the long term, archaeological data suggests much more static population levels, implying that either there were oscillations in population levels that were too rapid to detect using archaeological proxies, or that there were processes that operated over successive generations that moderated population levels. These processes could reflect social or environmental controls, or a mixture of both. If we can trace demographic change in the Irish archaeological record, it will be from these long-term perspectives at coarse resolution.

At first glance the spike and fall of dates associated with the establishment of Earlier Mesolithic settlement could be taken as an indicator of population collapse, but these data are heavily biased by multiple dates from Mount Sandel. Our data do not provide archaeological support for any bottleneck early in the Mesolithic that could have contributed to the genomic signals of this detected in later Mesolithic individuals (Cassidy 2020b; Cassidy et al. 2020).

We think the overall transformation of technology and mobility seen in the seventh millennium BC has fundamentally altered the ways in which sites were formed and might be identified. Given this dynamism, it is very hard to determine how the complex suite of changes can be directly mapped to gross changes in population level. It is however important to stress that previous claims that population fell in association with the 8200 cal BP event are hard to sustain with our new data, which, if anything, evince a time of growth. Baillie and Brown's (2020) claim that there may be a population effect of the 4350 BC climate decline (see also Tipping 2010, for discussion of this period) cannot easily be assessed from our data given the ways in which we have classified 'indeterminate' sites.

The situation in the final part of the Mesolithic is different. The changes seen from c. 5300 BC onwards in our models are the most likely to relate to population increase. This is compatible with the genetic evidence that suggests that, although the population was isolated, it avoided inbreeding during this time. We would however still advise against using our KDE models as a direct proxy for the whole population, especially given the biases introduced by sea-level change.

Understood in broad comparative context, increasing population density is often considered to be associated with a range of possible changes to the social dynamics of hunter-gatherer populations (Kelly 2013). In influential models, population increase is a key driver of developing inequality, sometimes considered as the development of 'complex' hunter-gatherers, through mechanisms such as the development of exclusive use of resources, intensification of exploitation of those resources, and storage. While these models have not been without their critics, they provide a framework to consider the innovations of the later parts of the Mesolithic. This might be especially marked in Ireland, which was a comparatively impoverished environment compared to many parts of Europe, meaning that we might expect increased pressure for limited resources to lead to a marked archaeological signal. Does the record for this period match the expectations of increased inequality? At best, such an association is not compelling.

Elsewhere in Europe the appearance of Mesolithic cemeteries is often used to indicate the presence of 'complex' hunter-gatherers, with cemeteries seen as indicators of a territorial claim and increased emphasis on ancestry (for review, see Warren 2017). The increased evidence for the manipulation of the human dead in the later parts of the Irish Mesolithic could indicate an increased interest in ancestry and, through this mechanism, ways of claiming a relationship to place. But it is important to stress that Ireland's best candidate for a cemetery (Hermitage) is much earlier within the Mesolithic (Little et al. 2017). The creation of distinctive places at lake edges might also be interpreted as claims to locations–either for settlement or other purposes–but this is far from certain (Blinkhorn and Little 2018). Although very poorly understood, the character of finely ground Moynagh points could be considered in terms of status objects. However, these tenuous indicators must be set alongside an archaeological record for the period which is dominated by mobility and a preference for forms of mobile architecture (tents) that enable the maintenance of sharing and egalitarian relationships (Warren 2022). Whilst aspects of innovation and change in the later parts of the Irish Mesolithic can be read in the context of population change and the possible social trajectories frequently associated with this in comparative context, the evidence is far from compelling. Many alternative explanations are possible.

This argument from absence can be interpreted in three ways:

• -. That there was no population increase, and that the changes in archaeological visibility are a product of the changing archaeological record with no demographic association.

• -. That long-term population change that took place over this period was not sufficient to drive very significant social change and inequality.

• -. That social organisation evolved in a way that enabled local populations to actively resist the development of inegalitarian relationships that might have been associated with population rise (see Warren 2022).

We do not consider the first explanation likely, and believe that a combination of the second and third may be appropriate. Our conclusion is therefore that, with the proxies available to us, it is likely that, from c. 5300 BC, long-term population growth took place in hunter-gatherer Ireland, although possibly not at very marked levels. It was into this dynamic context that the new material practices, objects and people of the Neolithic would arrive. Although we have not attempted to analyse this in detail it is important to note that McLaughlin (2020) identifies that in his data set there is on average 0.1 dates per year for the duration of the Mesolithic, but 1.7 dates per year for the Early Neolithic. The change in our evidence at the time of arrival of agriculture is highly significant.

Historical dynamics

The available radiocarbon data speak clearly of the dynamic and innovative character of the Mesolithic in Ireland, and provide a valuable context for some long-standing interpretations of the period.

During the very earliest, pre-7700 BC phases of the Irish Mesolithic, our analysis is limited by difficult-to-interpret sites and noisy, imprecise data. There is little meaningful dynamic present in this material, but it may indicate sporadic pioneer activity. Settlement proper is still first evidenced at Mount Sandel. Sites immediately following this are diverse but limited in number, and it is hard to understand change over the following millennium.

In contrast, the transition from Earlier to Later Mesolithic can now be understood as a set of changes to stone tool technology, and broader patterns of behaviour, which may not all have taken place at the same time, and which played out over a period of centuries. The changes in stone tool technology have been argued to transform constraints on mobility by removing the need to access spatially restricted high quality raw material sources needed for the production of regular blades (Costa et al. 2005; Costa and Sternke 2007; Costa and Sternke 2009). Consequently, we could predict a changing visibility and spatial distribution of archaeological activity. Indeed, this is what is found, including the first appearance of Mesolithic sites and features without associated artefacts. Whilst these changes may in part reflect an adaptation to the distinctive Irish ecology, and a reduction in the use of projectile hunting, they are also an adaptation to the limited range of high quality lithic raw materials available in Ireland. At the broadest of levels, this dynamism is matched elsewhere in Europe in the seventh millennium BC. Gronenborn (1999) describes this as a 'time of crisis' in Central Europe; Costa and Marchand (2006) show parallels between Ireland and France in changing lithic technologies at this time period; and Conneller (2021) argues that the period following 7000 BC sees significant regionalisation in Britain. The hunter-gatherers of Ireland were changing and dynamic, just like their neighbours.

It is not currently possible to assess the overarching causes of this shared dynamism. Although often emphasised as a driver of change, we have not been able to isolate the 8200 cal BP event as causing the origin of this process in Ireland. Problems with identifying synchronicity are important, but many aspects of this process of change were established long before the period of cooler, drier conditions. The event may have served to amplify ongoing dynamics. Although we do not have appropriate data to substantiate the claim, it is, for example, possible that the cool and dry conditions further encouraged routine mobility, which in turn has increased the archaeological visibility of activity. A stronger social-climatic relationship appears to exist at the end of the period of cool, dry conditions. This may indicate a further process of reorganisation of settlement at this time, perhaps a reduction in mobility leading to reduction in visibility. That said, it would be easy to over-interpret the spatiotemporal data here and the question would greatly benefit from more detailed work on the sites themselves, to establish seasonality and occupation history.

Following 5300 BC, there is increased archaeological activity visible in terms of radiocarbon-dated phases, changes to the distribution of these sites, and the appearance of some innovations. As argued above, of all periods within the Mesolithic, this is the one that is most likely to have seen population change, though potentially not at very high rates compared to later in prehistory. These changes have a broader European context, albeit one that we do not fully understand, which might relate to the appearance of agricultural communities in the region. There are implications here for our understanding of contact and connection between different regions at this time which need to be considered alongside the genetic evidence for isolation in terms of breeding networks.

This population growth does not appear to have led to the development of indicators of potential inequality, often assumed to relate to population packing, although some aspects of the data might indicate potential for such relationships to develop. It is possible that social strategies were employed to reduce this tendency to inequality, notably a continued emphasis on open living, shared spaces, and communality (Warren 2022). The maintenance of equality may have been an important social achievement by Irish hunter-gatherers.

Conclusion

This paper is the most detailed analysis of the corpus of Mesolithic radiocarbon dates from Ireland and conclusively demonstrates that the period was characterised by significant dynamics of change, poorly captured in the traditional binary division of the period. The key findings can be summarised as follows:

• -. There are substantial numbers of radiocarbon-dated Mesolithic sites.

• -. The distribution of these sites is unlike other periods of Irish archaeology and is geographically very varied, with notable areas with few dates, especially in the west and north-west.

• -. Many sites have very few Mesolithic dates, while a small number have a very large number of dates. This sometimes provides some bias to models of change over time and highlights the value of dating multiple features on sites.

• -. A significant proportion of these sites were excavated in advance of commercial or infrastructural development.

• -. A surprisingly high proportion of dated Mesolithic activity is in the form of direct dates on artefacts.

• -. Three main periods of increased growth in the frequency of Mesolithic radiocarbon dates are observed: a peak associated with first settlement; a peak associated with the dynamic changes of the seventh millennium BC; and finally, growth following c. 5300 BC.

• -. We consider the latter peak to have the most likely relationship to long-term population growth, but this may not have been substantial in scale.

• -. Previous arguments that the 8200 cal BP event saw population decline cannot be sustained, but it may have amplified ongoing processes of change.

• -. Dynamic change and innovation is present throughout the Mesolithic, but not at an even pace. It is often best understood in a European context, but includes distinctive Irish developments. As yet, we do not fully understand the drivers for these dynamics.

• -. We are not able to identify any population bottlenecks, which might match genetic data.

• -. Because of our approach to the categorisation of dates and the availability of data we do not make any direct comment on the Mesolithic-Neolithic transition.

Our analysis provides a new perspective on the Mesolithic of Ireland. This reflects the very high standards of excavation, analysis and availability of data in Irish archaeology in general. Much of the analysis for this exercise was carried out during lockdown in the 2020 pandemic, and it is salutary that so many reports from commercial excavations were easily accessible online. Colleagues across sectors were wholly supportive of requests for information, reports from many years ago, and in several cases, the provision of confidential information in advance of publication. There is much that Irish archaeology should be proud of in the data set available to researchers and the fieldwork that generated this material. We close by reiterating that we strongly disagree with the comments of Woodman and Waddell that archaeological mitigation in advance of development has done little for the Mesolithic in Ireland. On the contrary, when studied using the appropriate analytical tools, it provides a data set that helps transform how we think of the period.