Evaluating Transformations in Small Metal Finds Following the Black Death

Abstract

THIS PAPER SEEKS to evaluate transformations in portable material culture following the Black Death in England (1348–1349), specifically through an analysis of small metal finds data recorded with the Portable Antiquities Scheme (PAS). It will discuss the use of Geographic Information Systems and other computational methods in archaeological research, and apply this to compare and contrast PAS data against that from excavated urban sites. The importance of PAS data will then be highlighted as a chronological and spatially wide-ranging resource for understanding socio-economic change in portable material culture throughout the Middle Ages, focussing on the significant period of demographic change in the 14th century. It is suggested that an improvement in living standards is reflected in the variety of portable objects that have been recovered, and case studies of certain artefact types, with specific emphasis on dress accessories, will be used to demonstrate this.

Résumé

Évaluation des transformations concernant les petits objets métalliques dans la période suivant la peste noire par Eljas Oksanen et Michael Lewis

Ce papier tente d’évaluer les transformations concernant la culture matérielle des petits objets dans la période suivant la peste noire en Angleterre (1348–1349), spécifiquement par le biais d’une analyse des données des petits objets métalliques du Portable Antiquities Scheme britannique (programme pour les petits objets archéologiques, PAS). Il évoquera le recours aux systèmes d’information géographique (SIG) et à d’autres méthodes informatiques pour la recherche archéologique, et les utilisera pour mettre en lumière les similitudes et les différences entre les données PAS et celles de sites urbains fouillés. L’importance des données PAS sera alors soulignée en tant que ressource de grande portée chronologique et spatiale permettant de comprendre les changements socioéconomiques qui sont intervenus dans la culture matérielle des petits objets tout au long du Moyen-Âge, en se concentrant sur la période significative de changement démographique au 14^e siècle. Il est suggéré que la diversité des petits objets retrouvés reflète une amélioration des conditions de vie, et les études de cas de certains types d’artefact, en insistant particulièrement sur les accessoires vestimentaires, permettront de le mettre en évidence.

Zussamenfassung

Eine Bewertung der nach dem Schwarzen Tod eingetretenen Veränderungen bei kleinen Metallfunden von Eljas Oksanen und Michael Lewis

Dieser Beitrag will die Veränderungen bewerten, die bei beweglichen materiellen Kulturgütern nach dem Schwarzen Tod in England (1348-1349) festzustellen sind, insbesondere durch eine Analyse von Daten über kleine Metallfunde, die im Rahmen des Portable Antiquities Scheme (PAS) erfasst wurden. Die Rolle Geographischer Informationssysteme und anderer Berechnungsmethoden in der archäologischen Forschung wird erörtert, und diese werden eingesetzt, um PAS-Daten mit den Daten von Ausgrabungen urbaner Stätten zu vergleichen. Die Bedeutung der PAS-Daten als chronologische und räumlich weitreichende Ressource für das Verständnis des sozioökonomischen Wandels in der beweglichen materiellen Kultur während des gesamten Mittelalters wird hervorgehoben. Der Schwerpunkt liegt dabei auf jener bedeutenden Zeit demographischer Veränderungen im 14. Jahrhundert. Die Ansicht liegt nahe, dass sich eine Verbesserung des Lebensstandards in der Vielfalt der geborgenen beweglichen Objekte widerspiegelt, was anhand von Fallstudien zu bestimmten Artefakttypen, insbesondere Kleidungszubehör, aufgezeigt wird.

Riassunto

Valutazione dei cambiamenti dei piccoli reperti metallici dopo la peste nera di Eljas Oksanen e Michael Lewis

Lo scopo di questo studio è valutare i cambiamenti della cultura materiale di oggetti metallici portatili dopo la peste nera in Inghilterra (1348-1349) e, in modo specifico, mediante l’analisi dei dati di piccoli reperti metallici registrati nell’ambito del progetto Portable Antiquities Scheme (PAS) del British Museum e del museo del Galles. Si esamina l’uso dei sistemi informativi geografici e di altri metodi informatici nella ricerca archeologica e lo si applica per confrontare e contrapporre i dati PAS con quelli dei siti urbani oggetto di scavi. Si sottolinea l’importanza dei dati PAS quale risorsa cronologica e spaziale di ampio respiro per la comprensione del cambiamento socioeconomico nella cultura del materiale portatile durante tutto il Medioevo concentrandosi sul periodo del XIV secolo, significativo per il cambiamento demografico. Si sostiene che la varietà di oggetti portatili reperiti rispecchia un miglioramento del tenore di vita e per dimostrarlo si usano studi analitici di alcuni tipi di manufatti, con particolare rilievo per gli accessori di abbigliamento.

The Portable Antiquities Scheme ‘records archaeological finds discovered by the public to advance knowledge, further interest in the past and to tell the stories of past communities’ (PAS Vision 2025, PAS 2021). Its database contains over 1.5 million items from across England and Wales, with approximately 231,000 dating to the medieval period (c 1066–1540).³ The largest in proportion is its numismatic data, representing 39% of the dataset with nearly 90,000 records of coins, jettons and tokens. It also contains at least 500 other object types, including buckles, strap-ends, harness pendants, etc. This diversity of finds, as well as their wide geographic distribution, is among the strengths of the dataset.

This paper seeks to apply the small-finds data on a large spatial and temporal scale to study demographic, economic, social and cultural change across the medieval period, focussing on the 14th century and beyond. Termed the ‘Great Transition’ by Bruce Campbell (2016), the factors impacting society were complex and interlinked, ranging from the human to agricultural and climatic; the net result is well understood to have been a significant decline in population (Nightingale 2005). Among these, the Black Death (the bubonic pandemic, Yersinia pestis) that arrived in Britain in 1348—with several periods of widespread recurrence—had the sharpest and most obvious demographic impact. Although the scale of population loss from ‘the plague’ in England has been contested, it is now generally accepted that it killed 40–60% (varying by region) of the country’s inhabitants (Aberth 2001; Broadberry et al 2015). This understanding frees the focus to move towards a more finely-grained quantification of the effect of the Black Death, and its afterlives, as they emerge from the wider flow of technological and material culture history. Recent examples of such investigations include domestic pottery use and deposition (C Lewis 2016, 2020), church and cathedral building (Buringh et al 2020), and zooarchaeological studies of animal husbandry and dietary patterns (Thomas 2005; Thomas et al 2013, 2018). Given its wide geographical coverage across England and Wales, finds recorded through the PAS have the facility of producing similar large-scale assessments of change in medieval material culture.

PAS DATA AND THE BLACK DEATH

Large-scale diachronic change in the PAS finds data supports a maximalist interpretation of population loss during the 14th century. As will be discussed more in detail below (see Fig 3, see p. 166) the total number of small finds recovered and recorded dramatically declines from the 14th to the 15th centuries, although for some ‘object types’ this decrease is less marked, and for a minority of them there is actually an increase. As will be explored with particular reference to dress accessories (buckles, strap-ends, badges, etc.), this lends itself to the thesis of a smaller but on a per-capita basis wealthier population (recently considered by Broadberry et al 2015). Consequently, the demand created for an increased variety of metal-work goods can be read in the small-finds data (see also Standley 2013, 3–4 for the chronological and historical context for interpreting medieval dress accessories).

Fig 1 Findspots of PAS recorded finds (black areas) against places where metal-detecting is prohibited or restricted (grey areas) in East Anglia, eastern England. Some older finds may have been recorded with low spatial accuracy or recovered before an area became protected. Data: PAS, Robbins 2014.

Importantly, the nature of small-finds discoveries, mostly being casual non-stratified losses from rural areas, contrasts with excavated material predominantly found in urban areas through development-led archaeology. When analysed together, these datasets have the potential to complement each other. In England, public finds (i.e. those recorded by the PAS) are most numerous in the eastern and southern parts of the country, which is where metal-detecting is most prevalent because of their cultivated landscapes (e.g. Cooper and Green 2017; Lewis 2021, 35–37). The significant exception is London, where thousands of finds from all periods have been recovered by the public (particularly mudlarks) from the River Thames’ foreshore. How the former objects came to be in the ground varies, but most have been lost by chance, though ritual or purposeful burial is likely for some, including coin hoards (Andrews 2019) and (perhaps) certain religious items (Lewis 2022). Although many factors undoubtedly shape deposition and recovery patterns at local levels, there is an overarching relationship between finds recovery numbers and historical population density, which extends to reflect historical economic activity (Bevan 2012; Oksanen and Lewis 2020). Consequently, PAS data provide a snapshot of medieval rural life, although within a landscape that has significantly changed over time.

Although anyone may engage in archaeology, most public finds are made by metal detectorists on agricultural land in the countryside (Robbins 2012; Lewis 2021, 37). Invariably, these are principally metalwork and lack immediate stratigraphic context. Since most finds are recovered from cultivated ground (Daubney 2016), it is ploughing (not metal-detecting in itself) that is responsible for dislodging finds from an archaeological context and bringing them nearer to the surface. Furthermore, they are dominated by copper-alloy or lead items, although medieval coinage is typically silver. It is an essential point that detectorists normally set their machines to discriminate against iron (Robbins 2012, 99–100), so ferrous items are rarely located. Certainly, these finds go through a different series of processes than those that are professionally excavated, which introduce their own biases into the data (Robbins 2013). In terms of interpreting the medieval data, this presents both challenges and opportunities.

Concerning PAS data, ground-breaking work in contextualising archaeological small finds within the historic landscape and examining recovery biases has been carried out by Adam Daubney (2016) and Katherine Robbins (2013; see also Cooper and Green 2017). These show that the spatial biases affecting PAS data (or any sort of archaeological data, in fact), highlighted here through a case example of mapping archaeological finds in East Anglia, on the eastern coast of England (Fig 1), make much more sense when viewed against urban areas and places where metal-detecting is not permitted (such as on protected sites) or where it is restricted (see Robbins 2014). A variety of historical and modern processes further influence deposition, recovery and recording activity in ways that can be surprising or not immediately apparent. An example is the relationship developed between local Finds Liaison Officers (FLOs)—archaeologists working for the PAS, liaising with the public and recording their finds—and members of the metal-detecting community in their area; it may be noted that more established contacts typically result in a richer and more substantial body of local finds (Robbins 2013).

The PAS database, like large cultural heritage databases in general (see Newman 2011), is organic, with its structure, potential and biases emerging from a confluence of institutional, research and data-management processes. Likewise, though significant, the research attention that the PAS data has enjoyed has been chronologically uneven. Most previous projects have investigated its potential from multiple perspectives for the prehistoric, Roman and early medieval periods, but the medieval and post-medieval finds data, despite comprising nearly half (c 23% and c 20%, respectively), have not been the subject of similar endeavours to the same extent (however, see Oksanen and Lewis 2015, 2016, 2020; and below). This is particularly true for non-numismatic items, though they have been studied as individual types (e.g. Lewis 2014 on mirror cases; Standley 2016 on spindle whorls).

The capacity of the PAS data to shed light on research problems of wide importance for the Middle Ages is nonetheless highly promising. The potential for deploying crowd-sourced data in archaeology to detect population shifts has already been demonstrated by an analysis of ceramics sherds recovered from the thousands of test-pit excavations conducted by the Currently Occupied Rural Settlements project (C Lewis 2016, 2020). It maps not only the shrinking nature of many rural settlements but also the bourgeoning of others, presumably as the relative proportion of populace living in towns and market settlements increased in the later Middle Ages (Palliser 2000). The interpretation that lower volumes of recovered pottery sherds correspond to demographic change may be challenged by noting the adoption of costlier copper cooking pots by peasant households in late-medieval England (Dyer 2013, 25), although this too fits with the model of a smaller but wealthier population, which this paper seeks to explore below.

A major strength of the PAS database is its combination of broad spatial and temporal coverage within one unified dataset, enabling powerful large-scale diachronic analyses of the material culture that it contains. This has been recognised by studies focussing on major chronological periods, such as Roman Britain (Brindle 2014; Cool and Baxter 2016) and early medieval England (Richards et al 2009), and the Middle Ages (Oksanen and Lewis 2020; for medieval artefacts see also Standley 2013, Cassels 2013 and Webley 2022). It has been exploited by major research projects such as the ERC-funded project English Landscapes and Identity (EngLald), which focussed on gathering and analysing archaeological and heritage ‘big data’ from the Middle Bronze Age to the Early Middle Ages (c 1500 bc–ad 1086; Cooper and Green 2016, 2017; Gosden et al 2021; Green et al 2017). The study of numismatics and coin economy through the PAS data has also attracted sustained attention (e.g. Walton 2012; Moorhead 2013; Walton and Moorhead 2016; Bland et al 2020; see Brown and Moorhead 2021 for the Roman period; see Andrews 2019; Kelleher 2012, 2013, 2017 for the Middle Ages). These studies have developed possibilities for using detector finds data recorded by the PAS and other archaeologically recovered material as a proxy for demographic and commercial developments.

Since the number of finds in the PAS data varies greatly between different areas, distribution patterns based on their absolute numbers can be misleading. Often, they relate to the intensity of modern-day metal-detecting rather than historical processes leading to object deposition. Although the immediate stratigraphic context of ‘stray’ finds is typically lost (unless they are found in situ), valuable information is recovered from studying their relationships with the broader historical landscapes within which they were once located. Therefore, quantitative spatial techniques are critical for situating finds in context and comparing cross-regional variance (see Standley 2013). When handled statistically as sets of hundreds or thousands of finds, it is possible to detect such processes as the growth of towns and the spread of market events during the Middle Ages. Powerful statistical and Geographic Information Systems (GIS) methods such as spatial data smoothing can be employed to bring out particular features in the distribution patterns. The maturation of these techniques and approaches not only advances quantitative research but complements and supports qualitative approaches into the experience of living in the past (cf Wheatley and Gillings 2002; Campbell and Bartley 2006; Conolly and Lake 2006; Bevan and Lake 2013; Gillings et al 2020).

A density-ratio or relative risk surface analysis divides a mathematical kernel density surface of a subset of events (e.g. findspots of certain object types) by the surface of the full dataset to establish spatial/regional proportionality. This reveals whether the subset is over- or under-represented against the mean (Bithell 1990; Bevan 2012; see O’Sullivan & Unwin 2010 on kernel density estimates). A relative risk surface laid over a map of East Anglia—chosen here owing to its particular density of metal-detected finds—locates PAS data with other historical and landscape information (Fig 2). Here, as an example find type, are contrasted the relative distribution of strap-ends (discussed further below) dated to either the 13th or the 15th century, reflecting their diachronic and regional variations. Only strap-ends that could be dated to a minimum date range of 200 years were considered. From this were extracted all strap-ends where the centre point of their date range fell into either the 13th or 15th century, creating two non-overlapping temporal datasets. Represented in the map is the regional decline (colder colours) or increase (warmer ones) in recovered strap-ends between these chronologically distinct subsets of the PAS; removed from the analysis are statistically less significant areas of sparse finds density. At a county level, it is clear that the number of finds declined in most areas of Norfolk in the later Middle Ages, and by contrast rose in Suffolk. These regional differences appear connected to demographic change: county-level estimates indicate that the population declined across the 14th century in Norfolk—from c 485,000 in 1290 to c 175,000 in 1377—much more significantly than in Suffolk—c 225,000 to c 115,000 (Broadberry et al 2015, 23).

Fig 2 Contextualising recorded 13th- and 15th-century PAS strap-ends against Deserted Medieval Villages. Higher values or warmer colours in the choropleth relative risk surface indicate an increase in finds from the earlier to the later period, with lower values and cooler colours representing a decrease. Data: Historic England, PAS.

Furthermore, there is a significantly greater density of deserted medieval villages (DMV) in Norfolk as recorded in Historic England’s National Heritage List for England, also understood to be part of the region’s overall demographic transformation across the later Middle Ages. The similarity between broad regions of DMVs and declining finds numbers is striking and unlikely to be random. An overall greater negative impact in Norfolk than in Suffolk was likewise observed in the number of excavated ceramics dating to the central and late medieval period (C Lewis 2016, 795; 2020, 41–43). As will be explored below, it is proposed that a per capita rise in the number of small metal objects owned in the later Middle Ages is highlighted by the PAS data. It appears that the relatively lower level of population decline in Suffolk was insufficient to mask the overall rise in metalwork material culture. However, in many areas of Norfolk, being a region hard hit particularly by 14th- and 15th-century demographic change, the result was an overall decline in recovered items.

The current authors have established the potential of the medieval PAS data to elucidate long-term change at different spatial scales of investigation (Oksanen and Lewis 2015, 2016, 2020). It captures data on material culture that add to the body of documentary, surveyed and excavated evidence. Although studies have only started to examine PAS data in this depth, they offer the opportunity for an enhanced understanding of the post-14th-century development of the countryside and those processes which (if not necessarily initiated) were enhanced or shaped by the Black Death and its long-term after-effects.

DATING AND CHRONOLOGY

In the rest of this paper, long-term temporal patterns in the PAS dataset will be examined: first, by considering all PAS medieval data (though divided between coins and non-coin artefacts), followed by an analysis of their material, and then by object type, with specific reference to dress accessories (notably buckles and strap-ends), for reasons that will become clear. Finally, surface treatments applied to PAS will be explored including these terms.

The hypothesis presented here follows a premise that the long-term impact of events such as the Black Death can be observed, both quantitatively and qualitatively, in material culture recorded by the PAS. Unsurprisingly, in terms of the kinds of large-scale statistical analyses highlighted above and employed here, it is not easy to find archaeological evidence that is directly associated with the initial arrival of the plague in England in 1348 and its immediate aftermath. However, in terms of understanding the Black Death’s longer-term impact on the production and use of metalwork, and therefore on subsequent demographic, social, economic and cultural trends that the plague influenced, the PAS data does have the potential to provide unique evidence and perspectives.

As has been noted, most PAS finds are recovered from the plough zone, where they are divorced from a precise archaeological context (Lewis 2016a, 131). As such, and no matter whether they are found through metal-detecting or more traditional methods of archaeological intervention, these finds are typically dated by their morphology, style and decoration. Unlike coins whose manufacture can be dated quite precisely, medieval artefacts have much broader chronologies and are often only dated to within a century or two, if that. This is especially problematic for archaeologists trying to appreciate, understand and interpret the chronology and development of artefact types. But when data consist of hundreds or thousands of finds, it is possible to start examining broad probabilistic patterns. Thus, while objects are recorded by likely date of manufacture rather than the unknowable deposition date (albeit most PAS medieval finds are likely to be casual losses rather than deliberately deposited and, as such, reflect items in current use), for nearly all non-numismatic finds the necessarily loose dating timeframe means this is unlikely to complicate the aggregate picture.

A way of approaching this is through aoristic analysis. This computational method (Ratcliffe and McCullagh 1998) calculates an averaged, weighted distribution of imprecisely dated events across a pre-determined number of temporal periods known as ‘bins’. Coupled with probability simulation (such as via the Monte Carlo method), it is highly suitable for mapping large-scale and long-term changes in archaeological data, especially large numbers of imprecisely dated metal-detected finds (Crema 2012; Orton et al 2017; Oksanen and Lewis 2020). This is further refined by applying beta-distribution to the simulated data (here with the parameters alpha = beta = 2; see Cool and Baxter 2016) in order to model recording behaviour. In terms of probability distribution this establishes a moderate concave curve around the centre of an individual object’s assigned ‘timeline’ and emphasises this central portion of the assigned date range rather than either end. The hypothesis is that most imprecisely dated objects are somewhat more likely to be from the middle portion of their proposed timeline than the very edges.

The results of this analysis show that the PAS data broadly reflects historical production, circulation and deposition patterns (Fig 3). Medieval coin finds, for example, show a rise in their numbers from c 1150 until c 1300. They then decline numerically until c 1400, where they level off until the 16th century, before increasing dramatically. This pattern is collaborated by numismatic studies on the number of coins in circulation during the Middle Ages, notably by Martin Allen (2001, 2005, 2012, 2015), including the peak in currency in circulation in the first decades of the 14th century. The numismatic evidence provides a useful benchmark for assessing the non-coin data at both the national level and more locally (Oksanen and Lewis 2020, 124–34).

Fig 3 PAS artefacts and coins. Temporal distribution of medieval artefacts (grey, broad period ‘medieval’ n = 141,962), artefacts made of copper-alloy (green, broad period ‘medieval’ n = 106,480), and coins on the PAS database (black, broad period ‘medieval’ n = 80,524). Coin data is estimated with beta distribution of a = b = 1, whereas artefact data is estimated with a = b = 2. Data: PAS.

The trend for non-coin items in the PAS database is broadly similar to that of the coins, though these (non-coins) become more numerous a bit later (after c 1200). Also, their relative numbers increase more gradually.⁴ What is then notable is a sharp decline in non-numismatic items into the 15th century, presumably linked to the Black Death and the resulting social and economic change. The aforementioned study by Carenza Lewis (2016, 794–5) noticed a similar pattern in the decline of ceramics. This is important since another finds category shows a similar pattern to the PAS data, despite having different taphonomic and recovery mechanisms. As with coins, PAS object-finds level-off after c 1400, with a significant increase at the end of the Middle Ages: the drop-off in the early modern period is due to the PAS being necessarily selective in recording more modern finds, that is, those dated to after 1540 (PAS 2016). The important point here is that PAS data, in broad terms, seems to reflect the material culture in use, including the relative amounts of metal objects being manufactured, used and deposited.

Excavated evidence

In order to fully appreciate and utilise PAS public finds data, it is necessary to contextualise it against the evidence from medieval sites excavated archaeologically (i.e. through professional fieldwork), though this is somewhat easier said than done for reasons that will be clarified shortly. Here, as a sample, material from sites in London (Egan and Pritchard 2002), Norwich (Margeson 1993) and York (Ottaway and Rogers 2002) has been used. These places have been chosen for being major urban centres in the Middle Ages that contrast with the predominantly rural distribution of PAS-recorded finds, but which also have some geographical spread. Assessed as well, because of its importance in terms of the quantity and variation of medieval finds, is the eroded site of Meols, Cheshire (Griffiths et al 2007), which also offers a location in western Britain. Obviously many other productive/urban sites could be explored against the PAS database (see Cassels 2013, 84–115; Standley 2013, 17–24 for a discussion of some other assemblages, and Smith 2009 for some rural sites), but, importantly, all the chosen sites are especially rich in metal finds.

Cassels (2013, 5–7) highlighted the advantages of exploring archaeologically excavated urban (over rural) sites—including their contextual evidence for dating finds, namely, that these were places of manufacture and that they had a wide social demographic—but, in doing so, somewhat side-lined the PAS evidence (Cassels 2013, 7). Indeed, archaeologists (like Cassels 2013) often view finds from within excavated archaeological contexts as having superior value to those from the surface/plough zone, but it is clear from what follows that the picture is more complicated. Interpreting the excavated evidence is not without challenges, especially as sites are not explored or recorded using a single system or a set of procedures (Standley 2013, 9). Finds are often only dated by context (rather than also against other excavated parallels) and residual material within dated layers can be common. In short, this evidence (like the metal-detected finds data) is not always easy to evaluate or interpret.

London

The evidence from London represents a wide variety of finds and comes mostly from urban redevelopment in the 1970s and early 1980s. Most finds were recovered from sites adjacent to the River Thames, both from, ‘dumped fills, mainly of highly organic refuse, deposited for land reclamation, and from more mixed gravel and silt foreshores that accumulated against the successive medieval revetments’ (Egan and Pritchard 2002, vii). The main evidence is represented by 17 sites: also included are some ‘inland sites’, but here the preservation of finds in soils away from the waterfront is much poorer—so much so that Geoff Egan and Frances Pritchard (2002, 1) claimed that, ‘but for the waterfront assemblages, the surviving material culture from London … would give little indication that this was the most important city of the realm and the thriving market for the consumption of goods of all kinds that documentary sources attest’. Not only does this highlight the value of the foreshore sites for preserving material culture but also the complexities of interpreting London’s archaeology, not least as some sites contained the aforementioned ‘dumped fills’ from elsewhere, deposited in the city for land reclamation.

Norwich

Between 1971 and 1978, almost 40 sites across Norwich were excavated as part of the ‘Norwich Survey’, which also gleaned evidence from watching briefs and other interventions. An objective of this work was to ‘excavate and research sites threatened by development’ (Margeson 1993, 1); as in London, it was development control that led to archaeological works. Those investigated varied considerably in size and nature (ibid, 241–3), some being high status (e.g. sites 150 N and 169 N) or ecclesiastical (e.g. sites 36 N, 102 N, 157 N, 203 N) in nature. This, compared with London, gives value to the dataset. Indeed, the finds retrieved from Norwich, ‘provide rich and varied evidence of lifestyles and occupations’ in the medieval period, but, as Sue Margeson warned, this must be interpreted carefully given the nature of the materials used, their deposition, and the medieval attitudes towards reuse (ibid, 1). Compared to London, where the richest sites are close to the foreshore, the numbers of surviving finds from Norwich are not significant in themselves; this is unfortunate (for us) given the diversity of the sites themselves.

York

During the 1970s and 1980s, excavations in York recovered large assemblages of medieval artefacts from four major sites—16–22 Coppergate, 46–54 Fishergate, Bedern Foundry (‘the Foundry’,) and the College of the Vicars Choral of York Minster at Bedern (‘Bedern’)—as well as a number of smaller excavations (Ottaway and Rogers 2002, 2673). Some 6,000 stratified medieval (both pre- and post-Conquest) items were found, of which a third were made of non-ferrous metals. As in Norwich, these sites vary in nature and include evidence of periods of occupation before and beyond the Middle Ages. For example, 16–22 Coppergate, famous for its remains of Viking-Age Jorvik, lies on a spur of land between the rivers Ouse and Foss, where the increased density of population, the nature of deposition of materials and the proximity of this site to the river provided conditions conducive to the survival of organic remains in the Viking Age (ibid, 2674–9; Hall 1994, 27–29); therefore, as in London, watery contexts provide for the best chances of artefact survival (both organic and inorganic) and subsequent recovery. In contrast, excavations at Bedern (both sites) are further away from York’s rivers. Here, there are extant medieval remains of the College of the Vicars Choral, established in 1252, and an area of medieval and post-medieval industrial activity adjacent to the college precinct, including a complex of workshops (Ottaway and Rogers 2002, 2685–8).

Although these sites were excavated several decades ago, they represent some of the best published evidence for medieval urban settlements in England.

Material composition

Besides understanding the nature of metal small finds through time, PAS data also offer opportunities for better appreciating their materiality, including the significance of their metallic composition. As noted earlier, the vast majority (97%) of medieval coins in the database are made of silver. Among non-coin objects, the temporal distribution of copper-alloy artefacts closely mirrors the overall PAS distribution of medieval finds (Fig 3). This is as expected, given the recovery bias against ferrous metals; hence, copper-alloy items account for about 74% of all medieval non-coin artefacts.

As outlined above, the PAS data show a gradual rise in all artefact data from c 1200 until the late 14th century, then a decline and levelling off until c 1500, with a gradual increase thereafter well into the post-medieval/early modern period. Most of these objects were made of copper alloys. To put this abundance into some context, excavations across York (for example) recovered about 2,000 non-ferrous metal objects, representing about 40% of the total (Ottaway and Rogers 2002, 2673). This varied from site to site and through time, but it was still a relatively low percentage of the total; at Fishergate, 33% of finds were copper alloy, whereas at Coppergate this was just 9%. The broad temporal distribution of copper-alloy finds was not assessed in the excavation reports, but in the case of 4,250 iron finds from Coppergate there is an increase during the late 12th to the early-mid-13th century, and then a decline, which is similar for Fishergate (some 740 iron finds), though the increase in numbers extends to the mid-14th century because of how contexts are grouped (ibid, 2976–81). In London, there is a similar increase in the numbers of copper-alloy dress accessories from about 1270, which spikes in the second half of the 14th century, and then declines thereafter. This seems to broadly mimic what can be seen in the PAS data, though the initial peak is later, as is the decline in the 14th century. As such, it highlights this period (across all the sites investigated) as a moment of interest.

For other common metals, the picture is somewhat different. The PAS dataset shows an increase in the use of lead in recorded objects (14% of all medieval non-coin artefacts), with peaks at the end of the 12th century and also in c 1600 (Fig 4). It is important to note, however, that while detectorists recover lead, lead finds are not always readily identified as ‘recordable finds’ (i.e. of archaeological interest), and therefore might not be shown to the PAS; in short, finders are most likely to report precious metal finds (where there is a legal obligation to do so under the Treasure Act 1996) and those made of copper alloy, compared to those made of lead and other materials.

Fig 4 Temporal distribution of finds recorded in the PAS database as lead (grey, broad period ‘medieval’ n = 12,772), lead alloy (black, broad period ‘medieval’ n = 6,588), and silver (blue, broad period ‘medieval’ n = 3,416). Data: PAS.

Fig 5 Categories of broad period ‘medieval’ object types (n = 222,486) recorded in the PAS database. Data: PAS.

In the excavated urban assemblages, taking York as an example, lead-alloy finds were not particularly common in Coppergate (1% of all metal finds), rising to something more akin to what is seen in the PAS data at Fishergate (13%) (Ottaway and Rogers 2002, 2700). At the excavations in Bedern (both sites), lead-alloy items were most numerous with over 200 finds, representing 8% of all metal finds. The majority date from at least the mid-14th century, with most being from the 15th century or later; these include buckles, window cames and weights (ibid, 2699–700, 3091–2 and 3122–4). Here then, the archaeological data provide a mixed picture on lead use and its survival.

It is apparent from Geoff Egan and Frances Pritchard’s (2002, 18) work on medieval dress accessories in London that, ‘very few … of the excavated strap-fittings made of non-ferrous white metals [i.e. lead/pewter] pre-date the London Girdlers’ Guild’s charter of 1321, which sought to ban the use of what were regarded as inferior metals’; note that 171 of the 181 medieval tin/lead buckles from London are dated to after 1400 (ibid, 21). This was re-iterated in a charter of 1344 (ibid, 18). These restrictions on the use of lead were eased by 1391 and probably lacked much (if any) enforcement at all by the early 15th century. Indeed, it is market forces that probably best explain the rise in the use of lead towards the end of the medieval period. Egan and Pritchard (2002, 19) made the reasonable assertion that the main customers for pewter were probably a wide spectrum of town inhabitants, not just the urban elites, and this was facilitated by the increased availability of lead. It is therefore interesting that the rise in leaden objects in the PAS data from the 12th to the 13th centuries is not obviously witnessed in London or York, well before the 14th-century restrictions on its use. It might be that the PAS data (since it is a countrywide, and predominantly rural, dataset) serves as a better guide to the actual use of lead in the Middle Ages than the other urban assemblages studied. It is nonetheless noteworthy that many of the lead objects with a 13th-century date are seal matrices (hence somewhat skewing the data); this seems to be less common in the archaeologically excavated urban assemblages.

Turning to the other ‘white metals’, silver and alloys thereof (but excluding coinage) are the third most common artefact metal in the PAS dataset, though only a fraction (1.5%) of the total. Such items are rare among archaeological (i.e. professionally excavated) finds, even though the overall trend (in the PAS dataset) sees the rising use of silver throughout the Middle Ages, with increases (as with lead) dated to around the end of the 12th century and the end of the 15th century (Fig 4).

Sumptuary laws (notably in 1363) sought to control the use of ‘noble metals’ so that they could only be used by the upper echelons; their use in London was regulated by Goldsmith’s Company in the later Middle Ages (Egan and Pritchard 2002, 21; see Smith 2009, 314–6 for other restrictions on dress). It is generally thought that these laws were ineffective, and any impact of regulation and enforcement is not obvious in the PAS data. It seems that ‘white metals’ (silver, but also lead and tin) became more widely used relative to copper alloys.

It is clear that the metals used in the Middle Ages, as well as the apparent increased preference for ‘white metals’ at the expense of copper alloys (though the latter remain numerically superior), suggest a shift of some importance. The PAS-recorded copper-alloy finds best reflect what is seen in the aforementioned archaeological record for pottery: a reduction in numbers, reflecting a demographic decline. The steady (though relative) increase of other metal finds, particularly silver and lead, through the Middle Ages and particularly at the end of the period suggests a changing attitude in favour of these materials—which might be to do with taste but could also reflect the relative abundance of these metals and the ease with which they can be used in manufacture. Cassels (2013, 125–43) sought to identify manufacturing workshops through the evidence of the material culture of urban centres—see also Smith (2009, 317) who likewise implies the production of dress accessories was primarily an urban phenomenon—though this is likely to be just a small representation of the whole. Working metals like lead and materials such as bone, horn and wood suggest that medieval England had a wide and varied manufacturing basis that would have served a wide social stratum.

The prominence of lead in the PAS dataset for the later Middle Ages is intriguing, since many types of crafted leaden objects are less likely to survive in ploughed land than those made of other metals, including copper alloys and silver. Even so, this is likely to represent just a fraction of the lead used for dress accessories and other items during the Middle Ages. Griffiths et al (2007, 78) noted at Meols that, ‘the significant numbers [of dress accessories]’ made of lead/tin, which regularly fail to survive in adverse soil conditions’, and make the site of particular interest. Similarly, in London, there is a notable level of survival of items in these metals in the later medieval period, compared to excavated sites in Norwich, York and elsewhere. Cassels (2013, 52) believed that this abundance of lead alloys reflected the reality of what was being used in later London. However, this situation is probably best explained by the anaerobic conditions found on the Thames foreshore, where objects (especially organic, but also metals like lead and tin alloys) survive well in the mud. Furthermore, the PAS dataset reflects a similar abundance of lead objects; this is not apparent in urban places outside London. As such, these data (from London and through the PAS) might be more representative of the national picture than revealed through urban excavations in Norwich and York, where the evidence is likely to be more partial. In this regard, the PAS data helps fill some of the gaps, especially in the vicinity of medieval urban sites.

Object types and categories

An alternative way of interrogating PAS medieval finds data is to explore how objects transform over time. For ease of analysis, medieval items can be divided into larger ‘object type categories’ broadly based on function (here items with at least 50 individual records are considered): commerce, domestic, personal care, dress, literacy, religion, transport, tools and warfare, with coins further separated from other ‘commerce’ objects due to their numbers (Fig 5). Inevitably, some object types fit into these categories better than others, while others may be grouped into more than one, but they have been tested (following Leahy and Lewis 2018, 5) and appear robust. The advantage is that this consolidates some 200 object types—ranging in recorded numbers from 50 to several thousand—into a smaller set of categories that can be more easily investigated.

Numerically (excluding coins), the two largest categories are ‘dress’ (27% of total) and ‘domestic’ (15%), and interestingly they follow distinctive different trends, so they are worth exploring together in detail. With dress accessories, it is apparent that the number of finds increases significantly from the late 12th century onwards. Then, in the 15th century, the finds noticeably dip, stabilising somewhat mid-century, and then gradually increase in number in the 16th century (Fig 6); unsurprisingly, this closely reflects the patterns for copper-alloy artefacts, as 98% of medieval dress accessories in the PAS database are made of this material. Just over half of these ‘dress’ objects are buckles, and about 12% are strap-ends. Even when separated from the rest of the dress accessories, they have a very similar chronological profile (Fig 7). An aoristic analysis places the aggregate peak for finds numbers in the third quarter of the 14th century, after the arrival of the plague.

Fig 6 Temporal distribution of finds recorded in the PAS database as dress (grey, broad period ‘medieval’ n = 61,335), and domestic (black, broad period ‘medieval’ n = 33,884). Data: PAS.

Fig 7 Temporal distribution of buckles (grey, including frames and plates, broad period ‘medieval’ n = 32,369), and strap-ends (black, n = 7,273) recorded in the PAS database. Data: PAS.

By contrast, the number of finds in the ‘domestic’ category (also mostly made of copper alloy, at 67% of the total) gradually increased through the Middle Ages, but with much less variation and only a small dip around 1400 (Fig 6). Reconciling this with the significant decrease in dress accessories at the time of the Black Death is difficult. Possibly the relative increase of ‘domestic’ objects per head of population from the late 14th century might be related to the increase in living standards (Broadberry et al 2015; Platt 2001, 126–37) also witnessed in peasant households in later Medieval England (Dyer 2013), whereas the PAS data for dress accessories better reflect demographic change. This could be because personal adornment may be more responsive to such transformations, therefore highlighting the value of this object category in the dataset to better understand change and continuity during the medieval period. Since dress accessories dominate the PAS data, they may also reflect more general trends, although it is apparent that changes in the amount and types of metalwork in circulation were uneven and impacted items with different functions in distinct ways. It has been hypothesised (based on 125 objects from seven sites in County Durham, Lincolnshire and Yorkshire) that dress accessories may reflect ‘resistant identities’ amongst the medieval peasantry (Smith 2009). However, new information brought by the study of over 60,000 PAS-recorded medieval dress accessories suggests that the story is likely to be more complicated.

After the 14th century, the overall composition of PAS finds becomes more complex and heterogeneous. There are many reasons for this, not least the fact that new object types came into use, including some types of pilgrim souvenirs. However, when focussing on dress accessories, the trend in the PAS data is clear. While it is apparent that the absolute numbers of copper-alloy dress objects decline after the 14th century (following the general patterns of the material composition of objects observed above), that is not the case for lead-, silver- and tin-alloy items. As highlighted before, these steadily increase in number, and even more significantly in terms of relative quantity. This national trend within the PAS dataset is confirmed by excavated assemblages from London, where copper-alloy is the most common metal type used for items like strap-ends; from the second half of the 14th century, lead and tin become increasingly common (Egan and Pritchard 2002, 23). While this is less evident in Norwich and York, as noted above, objects made of these metals survive better in the London foreshore conditions, and therefore these (the excavated evidence from London, together with the PAS medieval data) appear to reflect actual manufacture and deposition patterns more closely.

Buckles

The single largest category of medieval ‘dress’ objects in the PAS dataset are buckles (including their frames and plates, which can be found separately); here, copper-alloy examples dominate with 98% of the total. The sample of lead- or lead-alloy buckles is small (at 115 finds), but mostly date to the later Middle Ages. As has already been noted, it is perhaps the case that ‘white metal’ finds (including items of medieval dress) might be under-represented in the PAS dataset; many detectorists are perhaps less likely to recognise, retain or bring in for recording objects that are fragmentary or severely damaged, unless they are finely made or outstanding.

These trends are broadly similar to those Egan and Pritchard (2002, 21) observed in London. Here, ‘copper-alloy and lead were used for buckle frames (for example) throughout the period’ (until about 1450), but pewter (lead/tin) came into use for this purpose from about 1270 to 1350, and thereafter proportionally increased. In the urban areas of Norwich and York, the picture is somewhat different; this is perhaps expected given what has already been learnt about the distinct nature of the archaeology of London’s foreshore sites, and how the PAS data (perhaps surprisingly) mimic it, albeit in very general terms.

While high numbers of copper-alloy buckles occur in both medieval and post-medieval contexts, most Norwich examples were made of ferrous metals (Margeson 1993, 24–34). In York, buckles of iron, copper-alloy and (to a lesser extent) lead alloy were found at all the sites, though the sample was small, with the lead/tin examples being at least early 15th century in date (Ottaway and Rogers 2002, 2886–7). In York, as elsewhere, the metals used varied depending on buckle type. For example, D-shaped frames were the most common form amongst iron buckles (ibid, 2891), suggesting that the metal used could be dictated by function; indeed, some (perhaps many) buckles (especially the iron ones) could have been for horse harnesses, etc., rather than for dress.

From this perspective, there can be seen in London—and the PAS dataset more generally—a wider range and greater variety of buckles being made from about the 13th to the early 14th century than before (Fig 8). This is true also for Norwich, where a variety in copper-alloy buckles and buckle-plates increases from about the 13th century (Margeson 1993, 24–33). Buckles with integral plates make an appearance in York contexts dating to the 13th and 14th centuries (Ottaway and Rogers 2002, 288) and rectangular, square and trapezoidal ones generally appear to be of 14th-century origin (ibid, 2891). In London, about a 100 years later, in the 15th century, some higher-quality composite forms were produced in both copper alloy and other alloys. Similarly, in Norwich, ‘15th and 16th-century buckles have more elaborate frames, and new types emerge…’ (Margeson 1993, 25). Together, these mark a transition in buckle types, towards more complex and decorative forms, with more being made of lead alloys.

Fig 8 Sample range of buckle types found in the PAS database (not to scale). Images: PAS.

Strap-ends

Strap-ends, another large category of medieval dress accessories, were designed to protect the ends of leather and textile belts or straps, and (in the case of woven braid and some other materials) to ease threading. Found in London are a wide variety of strap-ends from throughout the Middle Ages, though examples from the middle of the 12th century to the late 13th century are not common (Egan and Pritchard 2002, 126). Most are copper alloy; ‘strap-ends made from lead-tin or tin do not appear in deposits before the late 14th century, and are in forms exclusive to these metals’ (ibid, 23, 124). Forked spacers appear on London strap-ends from c 1270 to 1350 onwards; they are also used in buckle plates from the 14th century (ibid, 23–24). Ornate strap-ends date from the early 15th century and are predominantly made of lead alloys. In Norwich, so-called ‘simple’ strap-ends (though few were excavated, and all were copper alloy) date from about the end of the 13th century, although it is assumed that ‘they must have been very common throughout the medieval period’ (Margeson 1993, 34–36). Those consisting of two or three plates (with the latter there is a ‘spacer’ plate) appear to be later, thus from the 15th century, with even more elaborate examples coming from post-medieval contexts (ibid). Likewise, in York, strap-ends tend to be copper-alloy, with examples formed of one piece of metal being most commonly folded and (less commonly) lengthways, as well as those constructed of two sheets riveted together (Ottaway and Rogers 2002, 2900–2). Here, the study of contexts from which strap-ends derive suggests that ‘the simple one-piece strap-ends represent the earliest type’, though the earliest context dates to the mid- to late 13th century, with two-piece examples being roughly contemporary (ibid, 2902). Interestingly, while strap-ends with forked spacer plates seem to be from late 13th- to early 14th-century contexts in London, those in York appear later, in 15th- and 16th-century contexts (ibid); perhaps these types had a longer use in this part of the country, or they are residual. Therefore, the story of strap-ends is broadly similar to that of buckles, with simple types appearing earlier in the period and more complex varieties being later in date (Fig 9).

Fig 9 Sample range of strap-end types found in the PAS database (not to scale). Images: PAS.

What is seen in the PAS dataset is a gradual increase in the use of white metals over copper alloys. It is important to note that this move is relative since most dress accessories are still made from copper alloy. Even so, since lead alloys are less likely to survive (especially in the plough zone) and also easier to recycle, what survives is likely to be just a small proportion of the once-extant total. It is also apparent that the diversity in the form of dress accessories (here observed in strap-ends and buckle elements) increases towards the end of the period, especially so after the Black Death. This perhaps suggests that objects were being made for a new clientele; this will be examined further below when considering the surface treatment and decoration applied to various dress accessories.

Surface treatment and decoration

Concurrent with the appetite for more complex ‘white metal’ dress accessories over those of copper alloy, the PAS dataset suggests a decrease in the use of gilding as well as a simplification (or speeding up) in the application of some decorative techniques, although the picture is complex and hard to fully appreciate. This is interesting since it seems to reflect changes in manufacturing processes—as craftworking moves more towards mass-production (e.g. Bourgarit and Thomas 2012, 3069; Sawicki 2017, 25–30)—that enable craftspeople to make ornate and diverse but (importantly) cheaper wares.

Applying gold (gilding), silver (silvering) and tin (tinning) to metal objects are common surface treatments used in the Middle Ages. It was observed in London (Egan and Pritchard 2002, 27) that gilding occurs mainly on objects of purer copper, which suggests that these objects were of higher quality and crafted with more care; indeed, Theophilus, writing in the 12th century, advocated lead-free copper alloy to ease the application of gold coating (Hawthorne and Smith 1979, 145), again highlighting an additional degree of craftsmanship required in the gilding of dress accessories and other items. The technique uses gold amalgam (a solution of gold dissolved in mercury), whereby the solution is spread over a base metal and then gently heated to evaporate the mercury; clearly, the technique involves extra time and resources to make the finished product. Gilding (in particular) and silvering (to a lesser extent) are mostly used to embellish objects, whereas tinning (as will be explained below) has other practical uses.

Likewise, there seems to be a change in the way that objects, including dress accessories, were decorated with tools in the later Middle Ages. A logical explanation seems to be that there was a relative decline in the use of some metal-working techniques, which caused the production of dress accessories to be more time intensive to make and therefore more expensive. Invariably, if a strap-end, for example, was not gilded and/or not decorated, or decorated in a simplified (less resource-intensive) way, it would take less time to make and therefore could be sold at a lower price. Throughout the Middle Ages, objects were made with varying degrees of craftsmanship for a range of customers. Hence, the picture may not be as simple as it seems, but around the aftermath of the Black Death there is a noticeable change in how many objects were being produced, suggesting that the market for these goods had changed as well, perhaps driven by economics.

Surface Treatment

It is useful to consider the surface treatment applied to dress accessories recovered from urban excavations. In London, for example, tin coating is most common. This is often used for practical purposes, as it, was a standard protection for iron items against rusting, though it might also be used to give a decorative effect (Egan and Pritchard 2002, 27). In Norwich, many iron buckles and some other personal fittings—all from contexts dating to at least the 14th century—also had non-ferrous coating (Margeson 1993, 32–34); otherwise finds of iron dress accessories were rare. Here, it is also useful to consider the site of Meols—a ‘lost’ settlement washed away by the sea. As with items in the PAS dataset, objects found here were dated ‘based largely on the implications (sometimes limited to general stylistic inference) of comparanda from dated sequences elsewhere’ (Griffiths et al 2007, 77), rather than any discrete archaeological context. Several medieval dress accessories made of iron were recovered as well, some plated with non-ferrous metals, all buckle parts, though few were dated precisely due to unstratified locations (ibid, 188–9). Given that iron is normally coated for practical reasons, it remains constant through the Middle Ages, and an obvious decline in its use is not seen. The ‘flushing’ or ‘flashing’ of an iron surface with tin has been described as a ‘simple process’ (Jope 1956, 37), whereby a flux (such as resin) is applied to the metal, which is sprinkled with tin filings and then heated. Even so, this is clearly an additional task in the production process, requiring extra resources and time to execute.

The use of gilding on dress accessories has a notable peak in London between about 1150 and 1260 and a later one in the early 15th century (Egan and Pritchard 2002, 27; Webley 2022). This contrasts with the other urban sites studied, where few discussed items were gilded and, likewise, silver coating is rare. In Norwich, for example, none of the copper-alloy brooches or strap-ends appear to have surface treatment, and only one (of 14th-century date) of the buckle plates has it (Margeson 1993, 15–16, 24–31, 34–37). The situation appears to be similar in York, where none of the buckles and strap-ends appear to have been gilded (Ottaway and Rogers 2002, 2886–902). Likewise, in Meols, only a few copper-alloy items were gilded, though those that could be precisely dated were of the late 12th century (Griffiths et al 2007, 78), which is consistent with London finds.

Notwithstanding issues of precisely dating unstratified finds, this peak in the use of gilding falls within the more broadly defined high point in gilded dress accessory production in the PAS dataset between 1200 and 1400 (Fig 10). Thereafter, a levelling off is evident in the 15th century, followed by a small increase into the 16th; again, this is later than in London. The differences between the PAS and London data are not necessarily easy to explain, but the significant and sharp rise and drop off on either side of these dates in the PAS dataset indicate that gilded dress accessories are being recorded within a defined chronology, probably following established conventions, that may (or may not) reflect reality. Perhaps more significant, however, is the relationship between copper-alloy and gilded dress accessories in the PAS database, as both (giving a broad surface appearance of gold) declined proportionately towards the end of the 14th century, whereas the use of silver and lead increased. This again provides evidence of a preference away from that colour and towards favouring white’ metals, as observed in the material composition of the dress accessories studied.

Fig 10 Temporal distribution of gilded (grey, broad period ‘medieval’ n = 4,171), and surface decoration (black, broad period ‘medieval’ n = 2,141) upon dress accessories recorded in the PAS database. Data: PAS.

Surface Decoration

Similar trends can be observed in the decoration of objects in the PAS dataset, though it is important to note that the techniques used vary considerably—with, for example, some objects being carefully engraved with complicated patterns and inscriptions, and others being scratched or stamped with simple lines or marks. However, in very general terms, and here examining medieval dress accessories specifically, there is a sharp drop-off in these techniques being applied to metalwork (i.e. incising, engraving, chasing or inscribing) from c 1375 to 1425, and then a levelling off in their application through the rest of the Middle Ages. As implied above, this is not a straightforward narrative. Whilst it might seem logical to suppose that a reduction in these techniques led to objects becoming less ornate, the reality is that some manufacturing processes, such as the casting of tin badges in stone moulds, allowed for very complex designs to be produced more efficiently.

In London, Egan and Pritchard (2002, 29) noted that secondary decoration in the form of tooling occurs on many copper-alloy objects, but (as might be expected) only on a very few of the iron and lead/tin ones. Since these objects are more likely to be functional than dress accessories; it is also difficult to add incised decoration to iron objects and easy to add decoration (using a mould) to lead-alloy items. As noted above, this decoration varies considerably and is hard to interpret without in-depth analysis. Egan and Pritchard (2002) note that this secondary decoration is most commonly found on buckle frames and plates, as well as strap-ends, so these are worth examining in a little more depth.

In Norwich and York, a notable increase in the use of decoration on buckles and strap-ends—in terms of both moulding and applied work—is seen from about 1300. In general, whilst all buckle forms are common to all metals, those of copper alloy are more likely to have decorative embellishment, ‘perhaps indicating greater use of this metal for buckles on clothing or in other visible roles’ (Ottaway and Rogers 2002, 2886). In Norwich, copper-alloy buckles were usually cast, and decorated with various techniques (Margeson 1993, 24). Though this decoration varies considerably, it is apparent that embellishment of single-looped buckles and buckle plates is more common on those of 12th- and 13th-century dates, compared to those of the 14th and 15th centuries (ibid, 25–28), thereby matching what is seen in the PAS dataset. In terms of double-looped, circular and other buckle types, decoration is less apparent, though still more evident on earlier than later examples (ibid, 25–31). Only a small number of non-ferrous annular buckles were identified in York. All were plain, apart from two having pins with decorative mouldings: one was from a late 14th-century context, whereas the other was residual. In contrast, oval buckles from excavations in York are of varying shapes and styles of decoration (Ottaway and Rogers 2002, 2887). The earliest was found in a mid-13th century context, but they date across the medieval period, though the dataset was small.

Besides having a functional role (to protect the ends of leather and textile belts or straps), strap ends were also used for embellishment (Margeson 1993, 34); leather belts, in particular, can also have stamped designs and be further decorated with mounts, etc. It is likely that simple strap-ends of cut-sheet copper alloy, with little decoration or none at all, were common in the Middle Ages; only one excavated in Norwich was decorated, with a simple rocker-arm ornament (i.e. engraved zig-zag lines that are slightly curved). Those of two or three plates, or more elaborate designs, appear to be of 14th- and 15th-century dates, with the latter having minimal decoration besides their ornate knops (ibid, 34–36). As in Norwich, the simplest forms of strap-end from York are produced from one or two pieces of sheet copper alloy, which are less commonly decorated (Ottaway and Rogers 2002, 2900–1). Much more common are those made of three pieces—a pair of plates separated by a spacer—which are more embellished than the one- or two-plate strap-ends, and usually have a decorative knop (ibid). In general, the later, more decorative examples date to the 14th and 15th centuries (ibid, 2900–2), but, importantly, this decoration is not usually sophisticated.

Punched or stamped decoration (where a shape or mark is hammered onto metal) was used to embellish dress accessories from the late 12th century to the late 14th century, whereas engraving techniques—such as the aforementioned rocker-arm (or ‘wriggle-work’, where a sharp tool was employed in a zig-zag on metal to create broken lines of engraving)—is used on items from at least the early 13th to the early 15th centuries (Egan and Pritchard 2002, 30–32). It is the latter that appears most commonly on urban sites in Norwich and York, and it is significant that these methods (which can be less time-consuming) lend themselves to the processes involved in mass production. Indeed, whilst this is less clear among the PAS-recorded dress accessories (perhaps because of the abundance of buckles, strap-ends and the like, which are undecorated), these simple engraving techniques are the most common form of decoration on copper-alloy dress accessories from the urban areas under investigation. The data are thus showing the use of new decorative techniques later in the Middle Ages, with these perhaps being more efficient for production in terms of labour and material resources.

In summary, by the end of the medieval period, there was more variety in the form of dress accessories (including buckles and strap-ends), albeit with a decline in the quality of craftsmanship from the later 14th and early 15th centuries for many examples, compared to earlier. This is evidenced in the decline of gilding on copper-alloy items, as well as their hand-worked decoration being simplified. In the case of lead-alloy items, the manufacturing process (i.e. casting) allowed for more elaborate forms to be made (though mass-produced). Hence, there also seems to be a rise in the tinning of items, probably reflecting a growing preference for a white-metal look towards the end of the period.

CONCLUSIONS

This paper has confirmed that large-scale patterns related to the deposition of material culture (and, by inference, production and consumption) in the PAS data follow current thinking on demographic change as one moves from the central to the Later Middle Ages (Broadberry et al 2015). The application of computational methodologies allows the exploration of important variances in these trends within subsets of the PAS data, such as object type, material composition, or surface treatment and decoration. This provides important—and still under-exploited—information for understanding medieval material culture, offering potential new insights into wider changes in society; further areas of research stemming from this may include the consideration of shortages in manufacturing and craft specialisation because of the Black Death. Indeed, the PAS data helps contextualise trends in the excavated data, but with a dataset that is significantly greater in volume and geographically wide-ranging. Although some have already argued for a ‘dress accessory revolution’ in the Later Middle Ages driven by new methods of production and changes in consumer consumption (Cassels 2013, 157–65), the demographic impact of the Black Death—apparent as a contributing factor to impacts on the form and design of dress accessories recorded with the PAS data—has hitherto been somewhat understated. In absolute terms, it led to a decline in crafted metalwork. But long-term socio-economic change is also reflected in transformations in the object morphology seen in the PAS data (and elsewhere), indicating that dress accessories had increased in availability and variety.

In terms of metals used for dress accessories, there seems to have been a relative increase in the favouring of white metals over copper-alloy and/or gilded ones. More expensive materials such as silver were then emulated by the peasantry using cheaper metals, such as tin or lead. The use of lead alloys also lent itself to more complex forms at affordable prices, as evidenced by pilgrim souvenirs and other badges. It is probable that leaden items were even more common than the surviving data suggest, given that items made of these metals could be readily recycled and some are more vulnerable than others to survival in the ground. However, whilst dress accessories might be more easily crafted, it is also apparent that they are often more complex in form. The PAS data (alongside the excavated data) show objects with greater variety, if probably less intensively crafted. The use of gilding and engraved/incised decoration declines in favour of simpler decorative techniques, and the use of moulds for producing lead-alloy items means that more skill was needed to make the mould, rather than the many items produced from it. The general picture that emerges is a wider variety of metals being used for a greater variety of dress accessories—and perhaps other objects as well. These vary in quality and material, though many were easier to make, implying a larger and more varied customer base.

PAS small finds data confirm what was known (or at least suspected) about long-term changes in portable metalwork from local (predominately urban) assemblages, such as in London, can be generalised across larger portions of the country. It may be appreciated that the approach outlined here is a fairly rough-grained way of assessing the PAS data across England. There are undoubtedly significant regional and local variations that cannot be dealt with in this paper (see Standley 2013, 1–2, 9–26, 114; Cassels 2013, 146–57, 193–4, who explore regional diversity in medieval material culture). Indeed, in terms of deposition and subsequent finds recovery rates, different parts of the country responded in various ways to the demographic shocks of the 14th century, as was discussed (see above) in the case of East Anglia. Moreover, it seems there is a greater link between the PAS data and excavated London finds when compared to excavated archaeology from elsewhere. This suggests that London (certainly in the later Middle Ages) is not as unique as it first might seem in terms of finds types and fashions, and the PAS data is more typical of the whole.

These interpretations hold broader relevance for the study of medieval material culture. First, as previously thought when looking at padlocks, mounts for furnishings and candleholders (Lewis 2016b), it appears that medieval material culture was to a large degree homogenous and shared across all of society, although invariably the upper echelons would have owned better-made objects created from more costly materials (see Cassels 2013, 2–3; Standley 2013, 26, who explore the social significance of later medieval dress). Second, and backed up by more case-study evidence from urban excavations, changes in the character and availability of small metal finds can be seen through the later medieval society at the national level, as evidenced by the PAS data, indicating a dramatic shift in the availability and consumption patterns of these objects. It is hypothesised that this is linked to the overall rise in living standards in society.

It must be appreciated that the metal-detected data adds to the excavated information, and in any geographically or temporally large-scale investigation these must be explored together. This approach assists in identifying and mitigating biases inherent in both types of evidence. In particular, from the perspective of dating PAS finds, the possibility of circular arguments is encountered: the PAS evidence matches the excavated evidence only because the PAS evidence is based on typologies and chronologies derived from the latter. Establishing the independence of the PAS data consequently remains a challenge that is acknowledged here, and it points the way towards future research.

Finally, the approaches and methodologies deployed here can be easily transported to broader international (especially European) contexts. While the PAS has been a pioneer in public finds recording, in recent years similar schemes have been established or are in the process of being developed in several countries, including the Czech Republic, Denmark, Estonia, Finland, Flanders and the Netherlands. Beyond even national-scale investigations into long-term material cultural change, these, and other pan-European cultural heritage efforts (e.g., ARIADNEplus, see Richards and Niccolucci 2019), provide a foundation for a new transnational assessment of our shared past.

ACKNOWLEDGEMENTS

The authors would like to thank Roberta Gilchrist, Kevin Leahy and Rob Webley for their helpful comments on the draft text, our proof-reader Albion Butters, as well as the anonymous reviewers and Ben Jervis, who helped with aspects of the later draft. Likewise, the authors are grateful to have had the opportunity to present aspects of this research at the International Medieval Congress (Leeds) and the Society of Medieval Archaeology’s ‘Black Death’ conference (York), both in 2019, as this assisted in the formulation of the ideas presented here. Any outstanding errors remain ours alone. Eljas Oksanen’s time on this project was funded through the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement no. 896044.

Notes

3 Portable Antiquities Scheme (1997—ongoing), <www.finds.org.uk/database>. Data for statistical calculations downloaded 9 July 2021.

4 In this simulation, the actual number of finds per temporal bin falls within the area indicated by the coloured envelope’s upper and lower boundaries at 95% probability. For large data samples, as in Fig 3, there is less aggregate variation in the overall pattern of probability distribution and this region tends to be narrower; it is wider for smaller samples, as in Figs 4 and 10. Note that the coloured probability envelopes for artefacts are wider than for coins, since (as a consequence of their less certain dating) the former cannot be as tightly defined.