Abstract
Between 1276 and 1318, English magnates unsuccessfully attempted to establish a lordship in the Irish kingdom of Thomond, southwestern Ireland, by exploiting a dynastic feud dividing the then-ruling lineage, the Uí Bhriain. The conflict coincided with a series of extreme events that beset western Europe in the late 13th and early 14th centuries, such as the beginning of the Little Ice Age and the Great European Famine of 1315–1322. The goal of this work was to evaluate to the extent to which economic degradation at the turn of the 14th century affected the outcome of the war. The hypothesis that such degradation affected the war’s outcome was tested using agent-based modeling, which involved the virtual reconstruction of Late Medieval Thomond to study past conditions by proxy. This article describes the historical research carried out to elaborate the conceptual model, the implementation of the model as a computer simulation, and the experiments carried out to virtually explore the Uí Bhriain Civil War. A quantitative analysis of the experimental results revealed some correlation between late 13th century economic degradation and the fortunes of belligerent factions in the wars of 1276–1318, although the effect was not sufficiently strong to have been a crucial factor in the outcome of the conflict.
Introduction
“In the early history of the English arms in Ireland” wrote 19th-century antiquarian Thomas Westropp, “one episode stands, in a great measure, isolated from other events, with strong personalities on both sides of the combat and noteworthy by its very failure.” (Westropp, 1890, p. 284). In the second half of the 13th century, the Irish kingdom of Thomond was engaged in a major civil war. Following the death of king Conchobhair Ó Briain in 1268, two pretenders, Brian Ruaidh and Tadhg, fought for political supremacy. Although de facto independent, the Uí Bhriain of Thomond were nominal subjects of the English Crown, and the escalation of hostilities became a focus of instability in a colony already plagued by recurring warfare. Keen to establish a lordship of his own, English baron Thomas de Clare obtained a grant for the lands of Thomond and attempted to pacify the local Irish aristocracy. Following a modus operandi employed to great effect by other English magnates in Ireland, he and his successors exploited the division of the Uí Bhriain, backing one candidate in the civil war in exchange for political support and dependency. After nearly forty years of intermittent hostility, however, the de Clares had little to show for their efforts. Thomas’ son Richard died in battle against the descendants of Tadhg Ó Briain in 1318, after which the Irish lineage consolidated its power, eventually taking over de Clares’ lands. For Westropp, it was one of the “decisive battles of our [the British] empire,” one of the first defeats in a string of setbacks that would curtail the reach of English authority for the next two hundred years (Westropp, 1890, p. 284).
Westropp’s narrative may have overlooked an important element: economic degradation. According to 13th and 14th century sources, the English and Gaels had to contend with alarming environmental conditions. Irish annals contain many mentions of severe snowfall and overall bad weather (Mac Airt, 1944, p. 370) (Hennessy, 1871, p. 490) ((AI 1226.7. Mac Airt, 1944, p. 390) (Mac Carthy, 1893, p. 432). The population of Ireland was beset with famine (Mac Carthy, 1893, p. 334), food scarcity, (Freeman, 1944, p. 156; Hennessy, 1871, p. 466) and hardship (Freeman, 1944, p. 240). Disease and animal pestilence also abounded, with the cattle population—a crucial cog in the Irish economic machinery—taking a particularly heavy hit (Hennessy, 1871, p. 526; Mac Carthy, 1893, p. 436). Although these troubles were not caused directly by human action, they seem to have been aggravated by relentless warfare, specifically the devastation wrought between 1315 and 1318, when a Scottish army led by Edward Bruce put the English Lordship of Ireland to the torch (Duffy, 2002; McNamee, 1997). “Not, since the world’s beginning and the expulsion of the Fomorians from Ireland, a better deed was done to the men of Ireland” state the Annals of Connacht in the year of the Scottish prince’s death. “For three and a half-years, theft, famine and death of people were brought upon the whole of Ireland, and people undoubtedly ate each other” (Freeman, 1944, p. 252).
Recent research based on paleoclimatic proxies has confirmed that Northern Europe indeed experienced a drastic shift in weather conditions at the turn of the 14th century, with possibly dire consequences for agriculture and cattle husbandry. The period is generally seen as the beginning of a long-term climate regime known as the Little Ice Age. The decade of 1310, the tipping point in the wars of Thomond, appears to have been particularly momentous. A sharp increase in sea-surface temperatures in the 1300s, known as the Dantean Anomaly, (Brown, 2001, p. 251) was responsible for bouts of unstable and stormy weather. This development, in turn, provoked back-to-back harvest failures in 1315 and 1316, with follow-up events occurring in 1321, 1324, 1328, and 1331 (Campbell, 2016, p. 192). The temporal proximity of these extreme events compromised the chance for recovery, culminating in a period of scarcity and hunger known as the Great Famine of 1315–1322 (Kershaw, 1973). To Slavin (2018, p. 495), it was “the single harshest subsistence crisis in Europe of the last two millennia.” In Ireland and Britain, in particular, conditions were aggravated by large-scale military operations in Wales (1277–1283, 1316), Gascony (1294–1303), and Scotland (1296–1328) (Prestwich, 1997, pp. 170–202; 376–401; 469–517). These expeditions were partly supplied by purveyance, an often predatory practice in which the English Crown demanded provisions from its vassals. Reliance on Irish victuals and funds for campaigns compromised the grain surplus of Anglo-Irish manors and left magnates with meager reserves with which to mitigate the environmental disaster (Lyons, 1984, p. 9).
The temporal proximity between these extreme events and the Uí Bhriain Civil War in Thomond begs an important question: could the outcome of the conflict and the ultimate end of de Clares’ lordship have been caused at least partially by economic degradation? This hypothesis is not new. As early as 1984, Lyons (1984, p. 330) argued that the “worsening climatic conditions of the early years of the fourteenth century” had a significant impact on the cereal-based sectors of the Irish economy. Lyons, however, was not concerned primarily with the case of Thomond, and the local specificity of Ireland’s lordships and polities makes generalization dangerous.
Would it be possible that the de Clares’ ultimate failure in establishing a lordship in the region is connected, to one degree or another, to these unfavorable conditions? Or, at the very least, that the economic degradation caused by the early Little Ice Age, the Great European Famine, and wars being fought by the English Crown had some observable effect on this regional conflict in Thomond?
This article summarizes the findings of research conducted in the attempt to shed new light on the Uí Bhriain Civil War, focusing specifically on the interaction between economic degradation and military logistics. For this purpose, the ThomondSim agent-based computational model was designed and employed (Marino Carvalho, 2022).
The next section of this paper introduces the reader to the methodology of the study, including the fundamentals of agent-based modeling (henceforth ABM). Based on fresh readings of contemporary Irish and English sources and archaeological surveys, a model of the economy and logistics of 13th–14th century Thomond is presented in the third section. The 4th section describes how this conceptual model was implemented and validated as a computer simulation. The fifth section describes the experiment designed to test the study hypothesis and the statistical analysis of its results. In the 6th and final sections, the findings are presented and discussed in light of current scholarship, with the noting of limitations of the research design and offering of guidelines for future research on the topic.
Research Methodology
ABM is an in silico 1 research method for the analysis of relationships between agents and their environments. It consists of the creation of a virtual microworld for the study of the real world by proxy. Data are obtained by running experiments to test how the constituent elements of the simulation react to different starting conditions. The results are then subjected to statistical analysis. ABM can be performed with just about any programming language, but specific environments dedicated to the technique exist. This project was conducted in NetLogo 6.2.2., a framework for ABM development that boasts a sizeable sample model library, extensions, and reasonably wide adoption among humanities and social science scholars (Wilensky, 1999).
Due to its capacity to process large quantities of data and calculate several outcomes in a short span of time, ABM can be useful for the investigation of causality. By comparing the outcomes to available data and existing interpretations from historiography, researchers can assess how the hypotheses fare against one another. The working hypothesis for this project was that hazards brought on by the Little Ice Age and other external events had noticeable impacts on the Uí Bhriain Civil War.
Thomond is a compelling object for analysis using an in silico approach because, although many authors have attempted to gauge the effects of the Little Ice Age in Ireland, many of the sources needed to measure the most obvious impacts and responses (e.g., detailed manorial accounts, records of commercial transactions, and reliable proxies for population estimates) are lacking for this region. To overcome these hurdles and make the most of the ABM method, I relied on an abstract probabilistic model of the main types of extreme event and resulting impacts based on considerations from the existing literature. 2
Validation and Analog Model Development
The validation of an agent-based model based on historical sources is epistemically tricky, especially when, as in this project, it incorporates theoretical causal mechanisms. Hypothetically, a conclusion can be induced by the selection of initial parameters that deliberately generate the desired results. Moreover, even when a model is designed in a responsible and transparent manner, ABM results can never be taken as evidence of the same epistemic level as data extracted from sources. They cannot, by themselves, “prove” anything. They can, however, help researchers in at least three important areas: theory building and refinement (by mapping the consequences of certain theoretical premises), competing hypothesis testing, and data exploration (e.g., by assessing whether a given outcome is possible under the initial parameters) (Romanowska et al., 2021, p. 349).
To ensure that a model is epistemically valid, careful testing in the conceptual planning and coding stages is required. Although many guidelines for such testing exist, (Popper & Pichler, 2015; Sargent, 2011) many of these techniques are suboptimal for historians, who are typically not well versed in programming. As such, modelers risk getting stuck between a rock and a hard place: either arriving at a computationally sound simulation that is unintelligible to historians or designing a historically sound model that cannot properly be implemented as an agent-based model.
Since ThomondSim was designed with the aim of contributing to traditional historiographical discussion, I decided to err on the side of intelligibility when developing the model. To validate the simulation, I designed an analog version of ThomondSim that closely mirrors the model’s code, but can be manipulated easily without prior knowledge of coding or model building in the fashion of a tabletop wargame. Archaeologists and historians from two research groups based at the University of São Paulo—Arqueologia Interativa e Simulações Eletrônicas (ARISE) and Laboratório de Estudos Medievais (LEME-USP)—tested and validated the model. This parallel project, the development of which has been outlined in another paper, (Marino Carvalho) led to the release of a board game named The Triumphs of Turlough (Marino Carvalho, 2022).
A Conceptual Model of 13th–14th Century Thomond
Agents and Environment
ThomondSim’s environment is a representation of the physical and political landscape of Thomond between 1276 and 1318. Situated at the edge of the English Lordship of Ireland, Thomond (Thuaidhmumhain in Irish) nominally fell under the authority of the English Crown. It was an overkingdom (mór túath) whose king (rí, pl. ríthe) exerted overlordship over a number of petty kingdoms (túatha). Although Thomond was de facto independent, its kings were tributaries of the de Clares, whose power base was the manor of Bunratty, located near the Shannon estuary (Figure 1). During of wars of 1276–1318, the ruling lineage, the Uí Bhriain, split in two major factions: Clann Taidhg and Clann Briain Ruadh. The de Clares did not have an ironclad alliance with either party, although they tended to side with Clann Briain Ruadh during most of the conflict. Political map of Thomond during the wars of 1276–1318.
The Irish pre-modern political landscape was distinguished by its fragmentation. Ríthe ruled over territories vastly inferior in size to the realms of most kings elsewhere in Western Europe. At the size time, they enjoyed significant regional autonomy, and ensured the compliance of their subjects—who were also “kings” in their own right—according to a quasi-imperial framework of rights and obligations. Due to the peculiarities of their succession system, Irish lineages were prone to dynastic feuds, and developed mechanisms to pacify would-be pretenders. One such mechanism was the institution of the tánaiste (lit. “the second”), a title given as consolation to would-be pretenders to kingship to dissuade them from rebellion. Tánaistí received vast grants of land, but were far removed from a ruling lineage’s ceann hit (“head place”) (Gibson, 2012, Chapter 9: The World of the Caithréim Thoirdhealbhaigh, The Opposed Confederacies of the Ua Briain Civil War, para 13). In this way, they were both empowered by the king and kept at arm’s reach.
From the late 12th century onward, ríthe shared the political landscape with English magnates who attempted—with varying success—to establish lordships in Ireland according to English traditions of governance. Yet, far from the centers of royal power, operating with limited resources, and under constant military threat, many magnates adopted the modi operandi of ríthe to deal with their local subjects—for example, by fielding Irish soldiers or relying on traditional Irish forms of taxation. These barons also attempted to cement their influence by meddling in civil wars like the Uí Bhriain conflict of 1276–1318.
Types of Agents in the Simulation and Their Main Characteristics.
To “rule” in 13th and 14th century Ireland involved controlling people more than holding on to territory. To reflect this particularity, ThomondSim features no hard border between polities. A túath—or an English lordship, in the case of Bunratty—consists simply of the set of settlements over which secular overlords had certain rights. Three types of settlement exist in the model. Monasteries represent parochial or otherwise important churches, as well as their immediate lands and holdings. Altogether, ThomondSim has 63 monasteries, selected from an original list of 75 influential churches from previous surveys of the ecclesiastical landscape of Thomond and simplified to fit the model’s scale (McInerney, 2014; Nugent, 2005). Castles are the two fortifications operated by the English during the conflict: Bunratty (the manorial capital) and Quin (a second stronghold built between 1278 and 1280). Finally, longphoirt (sg. longport) represent fortified secular residences that served as ceann áiteanna for the ríthe. Due to the drastic changes in Thomond’s political landscape that occurred in the Late Medieval and Early Modern periods (Gibson, 2012, pp. 221–264), many of these settlements left no archaeological evidence. To deal with this limitation, I erred on the side of the available evidence and represented only the longphoirt for which we have a reasonable idea of overall location, even if only by proximity to other sites connected to the ruling lineage.
Rather than integrating these settlements in a spatial environment, I designed ThomondSim’s environment as a network. While this choice might seem at first counterintuitive given the availability of spatial data, it is actually more convenient for both ThomondSim and The Triumphs of Turlough in a number of ways. First, modeling the environment as a network makes it relatively straightforward to implement least cost algorithms, formulas that calculate the easiest—that is, less “costly”—path between two nodes. Second, a network model was a better fit for the project given its hybrid nature, as it is much more easily converted into an analog simulation, where the calculations need to be simple enough to be undertaken by players in no more than a couple of minutes. Finally, although a network environment is considerably more abstract than the spatial environment of other historical ABMs that dealing with movement and navigation (Craenen et al., 2012; O’Brien & Bergh, 2016), these questions are merely tangential to my research question.
The network environment comprises a set of 165 nodes connected by two types of link: routes (representing pathways) and tracts (open territories with no route of military importance). Settlements are programmed to spawn at nodes corresponding to their historical locations. These nodes also have two variables—attacker-modifier and defender-modifier—that affect the success of military engagements based on the presence of woodlands and peat mires. To reconstruct the pre-modern vegetation, I compared records from the 18th-century Books of Survey and Distribution (Simington, 1967), Ireland’s inventory of ancient and long-established woodlands (Perrin & Daly, 2010), data on peat-rich soils from Teagasc’s (n/d) Irish Soil Information System, and textual mentions of woods in the CT. Routes of military importance were reconstructed based on archaeological data, Early Modern maps, mentions in the Irish annals, and descriptions of itineraries in the CT. The methodology and results of this process have been published separately (Marino Carvalho, 2021).
Military Logistics
Although the picture of a Gaelic Ireland characterized by bartering, “non-market transfers,” and “reciprocal acts” (Kerr et al., 2013, pp. 70–71) pitted against English commercialism does not fully capture the nuances of the Irish economy, sources indicate that important differences existed between the Irish and English economic organizations and, specifically, military logistics. Before addressing how ThomondSim incorporates these discrepancies, a discussion of the available evidence is warranted.
The English administration had a cash-based, highly bureaucratized fiscal hierarchy centered at the exchequer. The lordship of Ireland had its own exchequer, although its records were (after 1293, at least) audited at the exchequer in England (Connolly, 1998, p. ix). This fiscal machinery was employed to manage the collection of taxes, tributes, and loot, as well as the payment of wages and purchase of military provisions. An example of these expenses can be found in the pipe rolls, which state that the castle of Bunratty boasted a garrison of 5 men-at-arms, 12 hobelarii (light mounted infantrymen), and 78 foot soldiers in 1318. The retention of this small army for 38 days of service cost £35 12s 6d, with daily wages of 1s per man-at-arm, 4d per hobelarius, and 1.5d per footman (Public Records Office of Ireland, 1911, p. 21). Full-fledged campaigns would have significantly larger costs. Army size in Ireland during this period ranged from a few hundred to around 1000 soldiers, (Frame, 1998, p. 291; Harbison, 1989, p. 106; Simms, 1979, note 49) who required provisions in addition to their wages (See e.g., Connolly, 1998, pp. 112, 125).
The Irish, on the other hand, commonly relied on cows, rather than metal currency, for tributes, payments, and commercial exchanges. Regarding taxation, ríthe collected regular rent (cís) from their subjects (Simms, 1987, p. 130). Because they lacked a centralized financial institution like the English exchequer, many exactions, particularly during wartime, were collected in kind. This modus operandi was known as coinnmheadh 7 buannacht, anglicized as “coyne & livery” and meaning the direct billeting and quartering of soldiers with the population, during which supplies, foodstuffs, and cattle were taken.(Simms, 1987, p. 131) Another important contribution was the cóisir or cuid-oidhche (“cosher” or “cuddy” in English), an obligation derived from the traditional custom of providing a winter feast for the rí, but which had, by the post-Invasion period, morphed into the right to seize food from subjects.(Simms, 1978) The Suim Cíosa Uí Bríain (“O’Brien’s rental”), a 15th century list of dues owed to the kings of Thomond, indicates that some territories were required to provide maintenance (bonna agus bairr) to support galloglass and gilladaibh (Suim Cíosa Uí Briain Hardiman, 1826, pp. 37–38)—a term that can be translated as “servant” or “youth of an age to bear arms” (eDil, 2019) and which McInerney (2015, p. 30) believes implied a military role of some sort. Galloglass were not yet active in Thomond during the wars of 1276–1318, but the obligations enumerated in the rental closely match the exactions described in the annals for this period. In the Suim Tigernais Meic na Mara, an early 14th century list of tributes owed to the Meic Conmara of Clann Chúillen, the lands from which food contributions at overlords’ request are often referred to as “free territories” (bfearanaib saora) or “free quarters” (cceatrumnaib saora) (Hardiman, 1826, p. 43). As Simms tentatively noted, the implication is that cóisir provided ríthe with the means to levy resources from lands that were otherwise exempt from taxation (Simms, 1978, p. 80). This impression finds an echo in the frequent denunciations of predatory exactions against monasteries, which otherwise enjoyed the right of sanctuary (Columcille, 1953, p. 56; O’Grady, 1929, p. 35; Plummer(Org.), 1914, p. 17).
Traditional Irish mechanisms for the recruiting and provisioning soldiers were not exclusive to ríthe. The repeated condemnation of the billeting of kerns in Anglo-Irish parliamentary records shows that the practice was frowned upon by the Crown (Berry, 1907, pp. 328, 376, 446). However, the very need to condemn it suggests that it was widespread. Other records confirm this impression. The CT indeed mentions the “battle-armed kern of Butler” marching alongside Muirchertach Ó Briain of Clann Taidhg (O’Grady, 1929, p. 61). An account of wages paid to an army led by the deputy justiciar William de Burgh in 1308 indicates that 55% of this army was under the control of commanders with Irish names (Connolly, 1982). That these soldiers were maintained via coyne & livery is hinted at in the Míorbuile Senain, a description of marvelous acts by St. Senan of Inis Cathaigh that allegedly took place during the Uí Bhriain civil war. The documents credit Richard de Clare’s death in Dysert O’Dea to divine retribution after his armies had plundered lands in Corcabhaiscinn under the protection of the saint (Plummer(Org.), 1914, p. 12). The expedition in question probably never happened, but the existence of the episode suggests that English magnates in Ireland billeted their armies with churches, notwithstanding how objectionable such a practice might have been in England (See Prestwich, 1972, pp. 18–19).
The hybrid position of the de Clares is also reflected in Thomas de Clare’s 1287 inquisition post-mortem (Sweetman, 1879, p. 459, No. 202). Although the magnate held lands in other parts of Ireland, Bunratty alone accounted for nearly half of the total valuation of his estates (£171 4s 6d of £356 1s 4d), of which most of the expected earnings (£121 11s) were rents owed by the Uí Bhriain. As was customary with inquisitions post-mortem, these numbers were probably underestimated (See Altschul, 2019, p. 194; Hartland, 2001, p. 248), but the more generous figure in the pipe rolls reiterates the relative importance of the Uí Bhriain’s contribution. Two parts of Bunratty manor alone rendered £406 11s 2 ¾ d in 1298-99, of which £249 15s 8d came from the rent of the Irish kings (Public Records Office of Ireland, 1906, p. 39). Given the cashless nature of the Irish economy, these tributes were likely paid in livestock. The pipe rolls recorded several examples of Irish chiefs offering reparations in cattle (Public Records Office of Ireland, 1904, p. 74, 1905, p. 27), and Séan Mac Craith, author of the CT, uses the word cáin (“tribute”) to denote a cattle herd being driven by kerns towards Bunratty (O’Grady, 1929, p. 14). This partial reliance on Gaelic practices of tribute-taking was typical of English lordships located on the Frontier, parts of which would have not been manorialized at all, but left instead in the hands of local Irish chiefs in exchange for a yearly stipend. As Keith Waters (2004, p. 162) remarks, forcing the Irish to pay rent in this fashion was as close to a “conquest” as the English were capable of achieving in the far fringes of the Lordship of Ireland.
Model Implementation in ThomondSim
The military logistics model was implemented in ThomondSim in the following manner. Four principal types of resource are provided in the model: pounds, provisions, cattle, and army strength (i.e., number of soldiers). Pounds are a means of purchasing provisions and soldiers that can be obtained exclusively by magnates. Provisions represent victuals such as grain and wine, alluded to in English financial records, and are spent by mobilized armies on the march at the rate of one provision to one army strength. If an agent starts a period without the appropriate provision quantity, it will lose army strength until it is forced to demobilize. Cattle combine the functions of pounds and provisions, although only Irish agents can exchange them directly for army strength. To further accentuate Gaelic Ireland’s economic reliance on cattle, ríthe generate income in cattle instead of pounds. Ríthe also have limited access to provisions. All resources except pounds were implemented on a numeric scale ranging from 1 to 10. In addition, two global variables, provision cost and cattle cost, determine how many pounds magnates must spend to acquire provisions and cattle, respectively. With the exception of army strength, which accurately reflects documented army sizes at a 1:100 scale, the measures are abstract and meant simply to “give the model some room to roam,” to provide “some numerical space out there for the model to explore” (Agar, 2003, paras 4.15, 4.18). Given the limited information provided by sources, reliance on concrete figures was not possible.
Cattle reserves increase at the rate of one per period, unless they are at maximum value (10). Clann Taidhg and Clann Briain Ruadh agents also receive a cattle income from every ríthe who supports their side in the war, a stance determined by the loyalty variable. They can also coerce hostile agents to pay tribute by defeating them militarily. This action creates a hostage directed link between vanquished and vanquisher that compels hostage givers to pay tribute until the link is terminated by the loss of a campaign. Magnates, on the other hand, earn incomes in pounds equal to the global variables Bunratty’s income (earnings from Thomond tenants) and external income (the combined income of all other manors held by the de Clares). These variables are initially set to 7 and £18, respectively, according to the ratio declared for Thomas de Clare’s 1287 inquisition post-mortem (Sweetman, 1879, p. 459, No. 202). Finally, a campaign win by a pretender supported by de Clare generates a hostage link to the magnate, compelling the payment of cattle tributes to the English agent.
In addition to obtaining regular income resources, magnates and ríthe can obtain cattle by raiding enemy settlements and save resources by billeting armies in settlements, as per the coyne & livery practice. They may also use coyne & livery for recruitment, spending their longport’s integrity to acquire army strength at the rate of 1:1.
Model Design
Schedule
Each period of the model represents one year in the war. The model has three phases (Figure 2). The pre-maintenance and maintenance phases represent the occurrence of hazards and gradual economic recovery, respectively. The likelihood of triggering hazards is based on the initial scenario, and some hazards affect the probability of subsequent events in addition to the economy (Figure 3). Due to the lack of detailed evidence for the effects of these calamities in Thomond, I kept the hazard sub-model deliberately vague; it simulates, in broad strokes, the most obvious consequences of economic shocks based on considerations from the historiography, (Campbell, 2008; Newfield, 2009; Slavin, 2018) but it is not meant to be a realistic reconstruction of local nuances. Values for the intensity and/or extent of each hazard’s impact were adjusted so that each shock had a significant effect on the course of the simulation, but was not enough, in itself, to provoke an unmitigated disaster. Because this process involved a great deal of guesswork, I incorporated these parameters as sliders in ThomondSim’s interface to allow other researchers to test their own combinations. I also conducted a one-factor-at-a-time sensitivity analysis to measure how changes in the fixed initial parameters
5
affected outcomes of relevance to the study. The data generated for these tests has been made available at the model’s page (Marino Carvalho, 2022). ThomondSim’s phases and major procedures. Effects of hazards in ThomondSim.
The expedition phase operates in a loop, with each iteration representing one day in the campaign. The loop ends when a pretender (Clann Taidhg or Clann Briain Ruadh) is exiled because it has been eliminated in combat or runs out of means to feed their warriors. Ríthe and magnates did not have standing armies in the modern sense of the word, and could not raise soldiers on a whim. To reflect this constraint, an agent can only mobilize if: (1) it is a currently exiled pretender, invading Thomond with the help of an external (agent); (2) it was called to arms by one of the factions; (3) it receives news of the war; or (4) it is scripted to do so based on the initial conditions of the scenario, which are designed to represent specific moments in the war before a major offensive. When an agent mobilizes, it can recruit soldiers, move, fight other armies, and raid settlements. Mobilized armies need to spend provisions or cattle at the end of each iteration, lest they suffer attrition from desertion.
Due to the relatively small sizes of 13th–14th century armies in Ireland, commanders often sought strength in numbers before significant engagements. To make the most of their superiority, larger armies attempted to neutralize individual armies before they could combine their forces. To illustrate this dynamic, battles in ThomondSim are divided into three procedures: enter combat, representing the moment at which opposing forces decide to engage; fight, during which casualties are calculated until one army is eliminated; and disengage, with which an attacked army may refuse combat. Simple formulas are used for battle calculations
Experiment Design
Scenarios
The experiment comprised four scenarios based on two parameters: toggle hazards and year. Toggle hazards is a categorical variable that determines whether extreme events occur.
The year determines the starting date of the simulation. This parameter has bearing on the loyalty of factions, balance of forces in the war, and probability of specific hazards. The model allows seven starting years: 1276, 1277, 1281, 1284, 1315, 1317, and 1318. Although data collection should ideally account for every possible date variation, I limited the experiment to two year states (1276 and 1317) to avoid the production of a dataset that was too large, complex, and unyielding. Although these dates do not account for every possible model state, they assign very different probabilities to extreme events. The year 1276 features recurring heavy rainfall episodes, which should place heavier burdens on Irish agents' settlement integrity and cattle; 1317 features fewer rainfall events, but more harvest failures and famines, which may disproportionally affect English finances and recruiting capabilities.
All other model parameters were kept the same throughout the experiments. Each scenario was run 100 times, for a total of 400 runs.
Output Variables
Output variables were the measurements taken after each run. Some of them account for possible effects on the economy, and others track changes in the political landscape. The economic reporters were as follows: final integrity, which is the sum of the integrity of all settlements by the end of the run. The highest possible value for this reporter is 730, as 73 settlements exist in the model. devastation, which is the sum of damage to settlement integrity from all sources throughout the run. The comparison of this reporter with final integrity provides a measure of the resilience of the environment. High values for both reporters reflect settlements’ replenishment to compensate most of the damage inflicted. mean global cattle, which is the mean of the sum of cattle owned by all agents (ríthe and magnates) at each step of the simulation. This reporter reflects the effects of warfare and extreme events on the cattle population. I chose to use the mean rather than the sum because, like integrity, cattle replenish over time, and gross outputs are not reliable indicators of the damage to the population throughout the run. total pounds, which is the total number of pounds obtained by a magnate in a run, regardless of whether or how it was spent. mean global pounds: mean of pounds owned by a magnate at each step of the simulation. max global pounds, which is the maximum value of pounds obtained by a magnate during the simulation. This indicator, like mean global pounds, provides insight on the magnate’s capacity to maintain a financial reserve. Higher values reflect larger reserves, and values exceeding 25 (the total predicted English income in years without disasters) indicate a surplus.
The political reporters were: ruling faction, which is the name of the faction (Clann Taidhg or Clann Briain Ruadh) left standing at the end of the war. CTg wins, which is the number of expeditions won by Clann Taidhg. CBR wins, which is the number of expeditions won by Clann Briain Ruadh. years under English rule, which is the number of expeditions won by a faction allied with the English (usually Clann Briain Ruadh, but sometimes Clann Taidhg, as the de Clares start the 1317 scenario unaligned).
Run Duration
Each run was set to end when one of the belligerent factions (Clann Taidhg or Clann Briain Ruadh) ceases to be able to amass support. This political capital was represented by the numerical variable febas (Old Irish for “excellence”), which decreases with military setbacks and increases with victories and shows of prosperity (e.g., starting a period with a full complement of cattle). After some informal experimentation, I stipulated a starting febas of 20 for ThomondSim’s pretender ríthe, which roughly resulted in an average run duration of 30 ticks. This length was sufficient to allow variations in the output variables to fall into discernible patterns. Furthermore, I included a safeguard to end any run after its 50th tick to prevent the possible (although infrequent) situation in which the war would devolve into a perpetual “tie breaker,” with neither pretender winning enough consecutive campaigns to fully deplete its opponent’s febas.
Methods of Analysis
To test if military performance was affected by extreme events, I conducted Kruskal–Wallis tests 5 to compare the rank totals of all economic and political output reporters between runs with hazards and control runs in which hazards had been disabled. Four independent groups (normal and control runs for each starting date) were compared. The procedure revealed significant differences between groups, and I performed a post hoc Dunn test to identify which pairs of groups exhibited discrepancies. Finally, I examined Spearman rank correlations (due to non-normal data distribution) between all political and economic output reporters to better understand the effects of economic degradation over the agents and environment.
Results and Discussion
The chosen scenarios disproportionally favored Clann Taidhg. Of the 400 runs executed on BehaviorSpace, Clann Briain Ruadh was the faction left standing in only 80 (20%) runs. In-person playtests of The Triumphs of Turlough yielded comparable results, despite players’ use of far more complex and varied strategies than included in ThomondSim’s algorithms. Interestingly, factions sometimes lost the war despite winning more campaigns during a run. This outcome merits further investigation, but is likely related to the manner in which febas is calculated. Losing a campaign by fleeing has a more forgiving febas penalty (−1) than does being defeated in battle (−2), a discrepancy included in the model to account for the historical fact that defeated pretenders very often chose to flee. Thus, an agent that opts to escape might last longer in the war than does one that repeatedly stands its ground. In addition, agents can amass febas reserves by defeating enemies during the enter combat and fight procedures. An agent that wins battles during an expedition may theoretically end a round with a positive change in febas even if it is ultimately defeated.
Differences in Outcome Reporters Between the Four Experimental Groups (1276 With Hazards, 1276 Without Hazards, 1317 With Hazards, 1317 Without Hazards).
Playthroughs of The Triumphs of Turlough had different outcomes. Although the analog model does not allow for the same kind of testing undertaken with ThomondSim, playtesters reported that winning became much harder as more hazards were drawn. This was particularly true for players controlling the English faction, as price increases following harvest failures or famines could quickly deplete their pound reserves. These results highlighted the possibility that economic degradation affected the course of the war despite its disputable direct short-term connections to hazards.
Spearman’s Rank Correlation Test Results.
To return to my original question, is it likely that the economic degradation caused by the early Little Ice Age, the Great European Famine, and wars being fought by the English Crown had some observable effect on this regional conflict in Thomond? The ThomondSim experiments and The Triumphs of Turlough playtesting sessions suggest that the economic impacts of the extreme events examined were not sufficient to affect the outcome of the wars of 1276–1318. Counterfactual scenarios in which no extreme event occur produced outcomes that were comparable to the historically based scenarios. Specifically, although markers of economic degradation seem to have had some impacts on the performance of the major factions in the conflict, these impacts seem to have been of secondary importance at best. Clann Taidhg, the Uí Bhriain faction that historically won the dynastic conflict, was victorious in 80% of the simulation runs. These results corroborate the outcomes of matches of The Triumphs of Turlough. Almost all playtesting sessions resulted in the ultimate victory of Clann Taidhg, and players who controlled the English faction found it very difficult to keep a friendly pretender in power. On more general terms, these findings attest to the potential value of ABM as a heuristic tool for evaluating causal hypotheses. Specifically, ThomondSim suggests that causal models for disasters observed elsewhere in 13th-14th centuries Western Europe cannot necessarily be extrapolated to Ireland.
Campbell and Ludlow recently offered an alternative explanation for the impacts of extreme events, arguing that they affected not the campaigns themselves, but the factors that led to warfare.(Campbell & Ludlow, 2020) According to the authors, scarcity periods in the 14th century—caused or worsened by environmental factors——may have contributed to a “low-equilibrium trap characterized by endemic violence.” (Campbell & Ludlow, 2020, p. 180) Although my model, by its very design, cannot be used to test this hypothesis, given the assumption that war will always happen at the beginning of each period, Campbell and Ludlow’s causal mechanism lacks documental backing. The authors use a single medieval source to illustrate the connections between hazards, scarcity, and violence, and this source relates to a different period of environmental instability—the Oort Minimum—that occurred three centuries before the Little Ice Age (Campbell & Ludlow, 2020, p. 162). This factor in itself is not sufficient to invalidate their hypothesis, as medieval annalists’ failure to notice connections between phenomena—or to express them in our terms—does not mean that no connection existed. However, the authors do not address the many competing dynastic and political casus belli mentioned in medieval sources, which suggest that the major conflicts that occurred in 13th and 14th century Ireland were not waged merely—or even chiefly—due to demographic crises or competition for resources. 5 Aware of the limitations of their hypothesis, the authors acknowledged that “linkages may have been subtle and indirect,” consisting, for example, of triggers to “launch opportunistic attacks on weakened rivals or to question and challenge the fitness of dynasties to rule.” 6 Even with extreme events playing this more modest role, however, their relative importance remains contentious, as political events could expose one’s flank to attack as much and as often as economic upheaval. The Uí Bhriain Civil War in Thomond offers many examples of such events, ranging from instances when the de Clares were temporarily recalled to Dublin or England, leaving their Irish allies on their own, (O’Grady, 1929, pp. 78, 93) to outrageous ceasefires enforced by external political actors and easily broken by disgruntled Clann Taidhg and Clann Briain Ruadh ríthe (Mac Airt, 1944, p. 378; O’Grady, 1929, pp. 16, 23). I do not deny that the mechanism proposed by Campbell and Ludlow may indeed have been operating in some Irish conflicts that occurred during this period, but more data and interpretive work are needed to determine which conflicts these were. Furthermore, we may have to live with the fact that such data do not exist for many such conflicts.
My reservations about Campbell and Ludlow’s hypothesis notwithstanding, I see no reason to doubt their overarching conclusion that extreme events in the 13th and 14th centuries helped to consolidate “a low-level equilibrium of sparse population, economic underdevelopment, […] poverty and low resilience to natural hazards” by the 15th century (Campbell & Ludlow, 2020, p. 160). The most important effects of economic and environmental hazards on the Uí Bhriain Civil War may not have been felt at the tactical and operational levels, but as the narrowing of the window of opportunity in which expansionistic projects such as the de Clares’ were feasible. The English push into Thomond in the late 1270s happened at the very end of a period of expansion of the Lordship of Ireland, in which the colony was turning out a profit for the Crown and “financial and commercial dislocation remained limited and localized.” (Campbell, 2008, p. 898) By 1321, with the death of the last male de Clare of Thomas’ line, the situation was very different. Hartland noted that English lords were still willing to become involved in the defense of their Irish lands, but only insofar as the lands represented significant shares of their income sources (Hartland, 2006, p. 20). This was certainly the case for Richard de Clare, who perished, sword in hand, protecting his interests in Thomond (O’Grady, 1929, pp. 144–145). It was not, however, the case for the Anglo-Irish lineages that inherited what remained of his lands after the death of his son, for whom the prospect of carving out a new lordship was simply not worth the effort (Waters, 2004, pp. 165–166).
Examination of the two types of historical process to which this hypothesis alludes—the minutiae of baronial politics and long-term transformations in the English colony’s economic landscape—falls outside the scope of this project. ABM may not be the appropriate method with which to study them. As Braudel once famously argued, not all varieties of historical change can be perceived with the same lens (Braudel, 1958, p. 739). Some require magnification; others, that we take a step back. Model-based history, of which this project is an example, may be more attuned to the durée of cliometrics, with its cycles of “ten, twenty or fifty years.” (Braudel, 1958, p. 727). The sudden deaths, changes of heart, and serendipity of histoire événementielle are swept under its fascination with rules and mechanics, repetitions, and regularities. The centuries-long developments of the longue durée may be too complex to be predicted responsibly, at least within the confines of a single research project. ABM may one day enable the overcoming of these limitations, as it has in history’s sister disciplines of anthropology and archaeology. However, we should not let our acknowledgement of its shortcomings turn to complacency. As Braudel admonished, “a clear conscience of this plurality of social time is indispensable to a common methodology of the human sciences” (Braudel, 1958, p. 726). Historians who turn to modeling with the specific hope of building bridges with other fields, whichever they may be, would do well to heed these words.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research project was funded by the São Paulo Research Foundation (FAPESP), grant numbers 2017/14023-3, 2018/18458-7.
