The Kingston assemblage

A total of 425 Romano-British finds have been recovered from the River Thames at Kingston. The majority (approximately 70%) of archaeological finds are directly associated with the construction of the current Kingston Bridge between 1825 and 1828, which replaced earlier timber-framed structures dating back to at least the twelfth century a.d.Footnote ³⁷ Direct records of finds and the circumstances of their recovery are sparse; however, Biden’s The History and Antiquities of the Ancient and Royal Town of Kingston upon Thames provides an anecdotal account of the objects recovered during these bridge works.Footnote ³⁸

The remainder of the finds are either broadly recorded to the local area, such as ‘Kingston’ or ‘near-Kingston’ (approximately 25%), or attributed to a specific, local stretch of river like ‘north-bank, downstream of bridge’ (approximately 5%).

Perhaps the most immediately noticeable aspect of the Kingston assemblage is the conspicuous lack of military equipment. While antiquarian accounts do refer to Roman weapons, these claims require re-appraisal. For instance, a bronze sword and an iron axe-head recovered in 1842 (Fig. 5), both published as Roman by Biden,Footnote ³⁹ have since been re-identified as being of Middle Bronze AgeFootnote ⁴⁰ (c. 1550–1150 b.c.) and early medieval (a.d. 900–1100) date, respectively.Footnote ⁴¹ Additional weapons described by Biden as Roman, including an iron dagger, several spearheads or ‘lances’ (see Fig. 5), and a military standard, are now untraceable.Footnote ⁴² As no illustrations were provided, a modern typological reassessment of the suggested Roman date is impossible.

Fig. 5.

Illustrated weapons recovered in 1842 during the construction of Kingston Bridge, all erroneously identified as Roman: MBA sword, early medieval/Viking axe head, Middle Saxon spearheads. (After Biden Reference Biden1852).

Fig. 5 Long description

The illustration shows a collection of historical weapons. On the left is a sword with a long, narrow blade and two holes near the hilt. Below the sword is an axe head with a curved blade and a detailed design. To the right are three spearheads, each with distinct shapes and textures. The spearheads vary in size and design, with pointed tips and textured surfaces.

It is important to note that British antiquaries in the late eighteenth and nineteenth centuries held Roman objects in particular esteem and frequently identified prehistoric and later medieval objects as Roman.Footnote ⁴³ Therefore, while is possible that the remaining weaponry recovered from Kingston was Roman, this cannot be confirmed and it is entirely plausible that these ‘Roman’ weapons, such as the iron dagger identified by Biden, also represent early medieval or later prehistoric finds. Even if they were positively identified as Roman, weaponry would represent a minor component of the overall assemblage (1.4%).

The Kingston assemblage is dominated by domestic, civilian material with no distinct military connections, much of which is broadly dated. This includes significant quantities of pottery (Fig. 6), predominantly utilitarian, consisting of locally produced coarse wares such as storage jars, bowls and drinking vessels which make up 81% of the Kingston pottery sub-assemblage and 8.5% of the Kingston assemblage, alongside a smaller but notable quantity of Samian vessels, representing 19% of the pottery and 2% of total finds. It is important to note that the recorded quantity of ceramic material is likely a significant underrepresentation of the archaeological reality, as nineteenth-century recovery practices frequently overlooked or discarded such finds in favour of more prestigious and valuable metal finds; due to its distinctive visual qualities, Samian ware is more likely to be recognised and retained than coarser wares.Footnote ⁴⁴

Fig. 6.

Archaeological finds recovered from the River Thames at Kingston, by object category (%). (Clifford Reference Clifford2026).

Fig. 6 Long description

The bar graph illustrates the distribution of archaeological finds recovered from the River Thames at Kingston, categorized by object type. The x-axis represents the percentage from 0 percent to 100 percent, while the y-axis lists the categories: clothing, coin, personal adornment, pottery, tool, weapon, pottery (Samian), domestic object and fishing equipment. Clothing occupies the largest portion, followed by smaller segments for other categories such as coin, personal adornment and pottery. The graph visually emphasizes the predominance of clothing in the assemblage.

The assemblage also includes other small finds, such as ornamental dress accessories including brooches, alongside objects such as styli, fishing hooks and keys, which are broadly consistent with material patterning expected from accidental loss or the discard of waste observed in other Roman riverine settlement contexts, although deliberate deposition of individual objects can never be excluded.Footnote ⁴⁵ Significant quantities of ceramic building material have also been noted in the context of riverine finds from Kingston, but specific quantities have not been recorded and therefore cannot be directly quantified in this analysis.Footnote ⁴⁶

The extensive numismatic assemblage recovered from Kingston was interpreted as a dispersed hoard (IARCH-CA989C), recovered during excavations at Eden Street approximately 250 m from the modern bridge (Fig. 4). The hoard is described as ‘a group of over 350 coins, mostly of fourth-century date, excavated from alluvial silt and gravel in an old side-channel of the Thames at Kingston’.Footnote ⁴⁷

The hoard is broadly dated to the later third and fourth centuries a.d., the bulk of it comprising Nummi from the House of Constantine (310–365); the latest individual coin is attributed to the House of Theodosius between 388 and 402 and the earliest coins are late third-century coins of Claudius II (268–270).Footnote ⁴⁸ The interpretation as a hoard implies a specific, likely singular depositional event, fundamentally different from attritional loss or repeated small-scale offerings which might be expected from daily activity around a crossing. It is therefore important to scrutinise this initial interpretation further.

When initially discovered in 1989, the Kingston finds were described as a dispersed assemblage.Footnote ⁴⁹ It is difficult to comment on this initial ascription, as the excavation has not yet been published in full, leaving the relative location of individual finds, and therefore the spatial basis by which these coins were interpreted as a ‘hoard’ rather than an accumulation of individual finds, entirely obscured.

It is therefore feasible that the Kingston ‘hoard’ represents either a single depositional event, or a series of individual depositions which have accumulated as a result of hydrological action in the fluvial environment. In the context of a single depositional event, dispersal of a hoard deposited in a bag or similar vessel could be the result of fluvial action.

Alternatively, it could reflect an event such as the sparsio, recorded by various late Roman authors including Claudian and later regulated by Justinian,Footnote ⁵⁰ which consisted of the ritual throwing or scattering of coins (missilia) during religious festivals or civic ceremonies, by which the local elite would demonstrate their wealth and generosity.Footnote ⁵¹

Another pertinent note regarding the Kingston ‘hoard’ is the quality and condition of the coins themselves. The excavator’s initial report included the comment ‘Many of them are forgeries’,Footnote ⁵² referring explicitly to the potential hoard. They also comment that an undetermined quantity of the coins were filed or otherwise defaced prior to deposition.

Defaced or otherwise ‘mutilated’ coins in Roman contexts are typically linked directly to votive practices, with modified coins frequently being interpreted as having been ritually killed or sacrificed before being permanently deposited.Footnote ⁵³ While various authors have provided suggestions as to how mutilated coins might be interpreted through a strictly utilitarian perspective, such as chop-marking representing a means of checking purity,Footnote ⁵⁴ the severity of modification would be immediately apparent, rendering these coins unusable as currency in day-to-day life, effectively precluding any monetary function and reinforcing the interpretation of the hoard as an intentional, votive deposit.

The use of forgeries of counterfeit coins in ritual contexts is similarly well attested in provincial Roman settings, though the presence of these coins within temples, sanctuaries and within votive assemblages has elicited a range of interpretations from Romanists.Footnote ⁵⁵ On one hand, the presence of forgeries in votive contexts may represent a form of substitution or tokenisation of currency as a votive; where the financial loss to the depositor is minimal, while the counterfeit retains its potency as a ritual object, its value is detached from the intrinsic value of the coin itself.

On the other hand, forgeries within votive contexts may simply reflect their general prevalence at the time of deposition. In this light, the presence of forgeries amongst the late third-century coins of the Kingston ‘hoard’ is entirely unsurprising, given the epidemic of coin copying under way in Britain during this period.Footnote ⁵⁶ Unfortunately, while the Kingston ‘hoard’ remains unpublished, all interpretation is almost entirely speculative.

While they fall outside of the period considered ‘Romano-British’, a notable, though now untraceable, hoard of ‘at least 10’ gold tremisses of Justin I (a.d. 518–527) was dredged during bridge works in 1849, perhaps hinting at continuing ritual deposition.Footnote ⁵⁷

In addition to the bronze coin hoard, a limited quantity of nine individual coin finds have been recovered from the Thames at Kingston. In contrast to the later gold hoard, the vast majority (96%, 359) of coins recovered from the river at Kingston, including the previously mentioned individual finds, are low-value, bronze issues from the later Roman occupation, chronologically skewed to the late third and early–mid-fourth centuries.

Spatially, the distribution of the Kingston assemblage appears highly concentrated around the inferred site of the Roman crossing, with the Kingston ‘hoard’ recovered from a side channel approximately 250m from the crossing point (Fig. 4). However, any interpretation of this pattern is severely complicated by numerous methodological issues, including taphonomic processes and recovery biases. Most of the assemblage was recovered during nineteenth-century dredging operations, a process which has become notorious in archaeological scholarship for its imprecise and non-systematic recording practices.Footnote ⁵⁸

In the case of Kingston, this issue is further compounded by river dynamics around bridge foundations and the dynamics of dredging itself. Bridge foundations such as those of Kingston Bridge act as significant obstructions to the flow of water and sediment, causing high-velocity flow to scour the riverbed immediately surrounding bridge piers, dislodging sediment and any embedded archaeological material.Footnote ⁵⁹ This scoured sediment and displaced archaeological material is transported in suspension, alongside any other suspended material from upstream, and deposited in slower-moving water immediately downstream, forming shoals in extremely close proximity to the bridge itself.Footnote ⁶⁰

These shoals therefore effectively act as ‘nets’ capturing sediment, debris and archaeological material into a single stretch of riverbed. Additionally, the intensive dredging required to remove these shoals for the purpose of navigation essentially targets a concentration of finds, which may not accurately reflect the full archaeological record and the original patterning of past depositional activity.

It is also important to note that the foundations of a Roman bridge, if present, would have the same hydrological impact on the Roman river system. The foundations would act as anchor points for the advection or transportation of archaeological material deposited upstream; therefore, while archaeological concentrations at bridge site may reflect genuine archaeological patterning, they may not directly reflect the specific location where past depositional activity itself took place. In effect, the bridge may have functioned as a depositional locale itself, but it also likely would have functioned as a net for material deposited further upstream.

Previous work has generally favoured a non-votive, incidental perspective towards the archaeological material recovered from Kingston. Early work in the 1800s connected the discovery of ‘Roman’ weaponry at Kingston with a local Roman-period cemetery, which itself was then broadly associated with Caesar’s campaigns in Britain and military crossing of the Thames.Footnote ⁶¹ Later scholarship drew on the significant quantities of pottery and ceramic building material recovered via dredging tentatively to suggest that a significant portion of the Roman settlement on the northern bank was destroyed by hydrological change and riverine erosion over a period of centuries.Footnote ⁶² This explanation was posited to account for the large quantities of material recovered across the wider Kingston stretch of the Thames, alongside its alluvial deposits and gravels downstream, suggesting they were the product of riverside middens and other terrestrial deposits gradually washed into the modern river channel.

The cumulative impact of these processes means it is now methodologically impossible to adequately distinguish between material that was deliberately deposited directly into the river and the material which has been redeposited from elsewhere. While the absence of later medieval finds may appear to suggest that erosion is unlikely to be the primary mechanism of deposition, it is highly likely that the Kingston assemblage contains material from both sources, and both now form a heterogenous ‘riverine’ assemblage.

The Goring–Streatley assemblage

A minimum of 99 individual Roman finds has been documented from the Goring-Streatley crossing. This recorded total, however, likely represents a significant underrepresentation caused, as discussed above, by biases in late eighteenth- and early nineteenth-century recovery practices.Footnote ⁷²

The Goring–Streatley assemblage is not only smaller but also compositionally distinct from that at Kingston. The complete absence of weaponry and military objects is again particularly notable, as is the lack of ‘military-adjacent’ objects such as highly ‘militarised’ brooch types or horse harness fittings typically associated with a military identity.Footnote ⁷³

The only potentially military associated find is one of two brooches recovered from the crossing, previously identified as a trumpet brooch, a type sometimes linked to Roman military identities, particularly in a British provincial setting.Footnote ⁷⁴ However, a re-examination of the museum’s accession records shows that the object held under the ‘Goring-Thames’ provenance does not conform to the typological description of a trumpet brooch, or the physical description of the Goring brooch. It appears likely that mislabelling has resulted in confusion with a genuine trumpet brooch from another unrelated local site, and the second brooch in the assemblage may represent the original Goring–Streatley find. Due to this uncertainty regarding the actual provenance of the find, it cannot be reliably leaned upon as evidence of a potential military connection, which further emphasises the overall non-martial character of the remaining Goring–Streatley finds.

Instead, the assemblage is dominated by domestic, ‘civilian’ archaeological material with no clear military associations, primarily pottery and coinage. The ceramic finds consist entirely of coarse wares, with high-status finer wares such as Samian absent, and eight complete vessels recorded. The dominant pottery types recovered at the Goring–Streatley crossing point are grey ware and black-burnished ware, though specific types and forms are not frequently recorded. This lack of detail is symptomatic of a wider issue within British river archaeology, wherein pottery, particularly sherds, was not generally retained or examined by a specialist prior to being redeposited back into the river or repurposed as ballast. Consequently, while the Goring–Streatley assemblage appears to be particularly notable for its numerous intact ceramic vessels, which represent over 90% of the pottery recovered from the crossing, this likely reflects a significant recovery bias, where sherds of fragmented vessels may simply not have been recorded. However, a recent appraisal of Thames Conservancy Dredging records indicates that this stretch of the Thames was dredged relatively frequently in relation to the surrounding area, which would appear to suggest that if significant quantities of sherds were present, they would have been encountered at least in small quantities.Footnote ⁷⁵ The dominance of intact vessels, therefore, appears to be a genuine, if potentially exaggerated, archaeological pattern.

Coins make up the bulk of the assemblage, accounting for approximately 88% of the finds recovered from the crossing (Fig. 9). There is no contextual evidence to suggest that these coins predominantly derive from a single large deposit or hoard, as at Kingston. The available information regarding provenance indicates scatterings of coins recovered at various locations along the Goring–Streatley stretch of river, although specific findspots are not typically recorded. This pattern would appear to be more consistent with the gradual accumulation of individual or small group losses or deposits over a period of time. The chronological profile of the recovered coins, with a small quantity of first- and second-century silver issues (7.95%, 7) alongside a majority of later Roman, low-value bronze issues (92%, 81), appears to suggest a general longevity throughout the Roman period.

Fig. 9.

Archaeological finds recovered from the River Thames at Goring, by object category (%). (Clifford Reference Clifford2026).

Fig. 9 Long description

The bar graph illustrates the percentage distribution of archaeological finds by category. The x-axis represents the percentage from 0 percent to 100 percent. The y-axis lists the categories: coin, domestic object, personal adornment and pottery. The graph shows that coins make up the majority of the finds, occupying most of the bar. Domestic objects, personal adornments and pottery are represented by smaller sections at the end of the bar, indicating a much lower percentage compared to coins.

The Goring–Streatley assemblage presents an entirely different set of interpretative issues when compared to the assemblage from Kingston, centred around the processes influencing deposition at a rural, likely seasonal ford. The notion of accidental loss is a plausible, though partial, explanation for material deposited specifically at crossing points, particularly fords. An individual could feasibly lose a coin or drop an intact vessel while wading through the water, but in these circumstances the hypothetical resulting patterning of random, chance losses does not match the specific character of the material recovered.

As Richard Bradley has famously argued for later prehistoric assemblages, a highly specific composition of particular object types and significant quantities of finds are strong indicators of structured, intentional practice, rather than the gradual accumulation of accidental losses.Footnote ⁷⁶ While Bradley’s original argument focuses on deposits of prestige metalwork, the same underlying logic applies. The clear preference for two specific finds categories, coins and intact ceramic vessels, appears to indicate a selectivity to deposition at Goring–Streatley.

Another key issue is the consideration of taphonomic formation processes, which address the question of whether the recovery location reflects the original point of deposition, particularly in relation to the potential for post-depositional transportation from upstream. Hydrologically, a shallow, gravelly ford in a lowland river system, such as that at Goring–Streatley, acts as a low energy ‘trap’ or functional ‘riffle’.Footnote ⁷⁷ As water flows over the shallow gravel bed of the crossing, the velocity of flow changes, causing objects to be rolled along the bed or transported in suspension to drop out and become lodged amongst the gravel and fine sediments.Footnote ⁷⁸ This process leads to the ford becoming a natural accumulation point which primarily traps objects deposited at or very near the crossing.Footnote ⁷⁹

This contrasts with the greater potential energy obstruction represented by a bridge pier, which can accumulate material transported from a much larger upstream area through obstructing flow and creating a depression and vortex which wraps around the bridge piers.Footnote ⁸⁰ The presence of numerous intact ceramic vessels recovered from the Goring–Streatley stretch suggests the material was likely deposited directly at, or very near, the crossing, as it is highly unlikely that fragile ceramic vessels would have survived transportation in high-energy scour or flood conditions.

The key changes under Roman rule

Throughout later prehistory and the early medieval period, the communities of the Middle Thames Valley engaged with the river through a long-standing tradition of intentionally depositing prestigious metal objects, particularly weapons.Footnote ⁸¹ The onset of the Roman occupation marked a profound and abrupt departure from this long-established practice. The large, bladed weapon deposits which characterise earlier and later assemblages effectively vanish from the river, replaced by an entirely new suite of material composed almost entirely of pottery and coinage. It is this influx of new materials, rather than an intensification of existing practices and behaviours, which accounts for the overall rise in the rate of deposition in the Romano-British period (Fig. 10).

Fig. 10.

Relative proportions of Middle Thames finds per chronological sub-period by object category. (Clifford Reference Clifford2026).

Fig. 10 Long description

The chart displays stacked bars representing different object categories across various chronological sub-periods. Each bar is divided into segments showing the percentage distribution of categories: weapons, tools, ornaments, ingots/currency bars, pottery, coins and other items. The sub-periods include labels such as Unknown (SW) with 67 items, LBA (12th-11th) with 150 items, EIA (9th-7th) with 108 items and others. The x-axis is labeled with percentages ranging from 0.00 percent to 100.00 percent. Each category is represented by a distinct color, allowing for comparison across different time frames. The chart provides a visual representation of the relative proportions of these categories over time.

The widespread adoption of coinage in Roman Britain, particularly the proliferation of low-value bronze coins, provided a new, accessible medium for depositional activity. It is plausible that the deposition of coins was seen as a new, ‘Roman’ alteration of existing depositional practice, a more abstract and less overtly ‘native’ practice than the deposition of weaponry, suitable for multiple scales of activity. The contrasting assemblages from the case studies at Kingston-upon-Thames and the Goring–Streatley crossing exemplify this; the large hoard from Kingston suggests a significant single event, while the attritional patterning of small, isolated deposits at the Goring–Streatley ford suggests a more dispersed pattern of repeated, small-scale deposits.

This compositional shift was also notably accompanied by a spatial one. The broad concentrations of finds seen in later prehistory are replaced by tight clusters of material almost exclusively associated with specific nodes in the Roman landscape.Footnote ⁸² The greatest concentrations are found at urban centres that also served as major crossing points on the arterial road network, such as Kingston and Staines. Smaller concentrations are also found at rural crossing points, further highlighting the link between depositional hotspots and the highly managed, settled landscape of Roman Britain.

Negotiating flow: hydrology, taphonomy and experience

An often overlooked aspect of riverine deposition is the physicality of the river itself, particularly the dynamic nature of flow. The hydrological characteristics of a river change significantly along its course; this is demonstrated particularly clearly by the vastly different physical environments at Goring–Streatley and Kingston, where a low-energy, relatively shallow meander sharply contrasts with a high-energy, deep and channelled straight. This changing physicality naturally impacts the nature of human–river interaction, both in terms of the direct, lived experience engaging with the river and the taphonomic processes which influence the survival, transportation and recovery of deposited objects.

The sensory and physical experience of crossing a deep, potentially fast-flowing and dark fording point is often alluded to in archaeological discussions regarding rivers and movement, with a general sense of focus on the hazard and danger of such a crossing in contrast to a bridge.Footnote ⁸³ In archaeological explorations of deposition at crossing sites, this experience is typically emphasised in terms of providing psychological impetus for small, propitiatory ‘votive’ offerings intended to ensure safe passage or provide a means of thanks for a successful one.Footnote ⁸⁴ Similar ideas are explored in analyses of larger prehistoric and early medieval deposits, which are often framed as highly emotive, ritualised reactions to increasing hazardous hydrological conditions amongst riverine settlements and communities, such as in direct response to intensive flooding.Footnote ⁸⁵

While a model centred on accidental loss is plausible for isolated finds, it is insufficient adequately to explain the quantity and specific patterning of the assemblages from either site. The different fluvial profiles and hydrological processes at either site further support an intentional, structured depositional narrative. At Kingston, the greater rate of flow in the lower–Middle Thames, which would only be accentuated locally by the inferred bridge piers, would allow objects to become trapped within accumulated fine sediment and debris.

However, turning attention to the Kingston hoard, hydrological forces would be insufficient to transport a heavy, contained deposit of this nature from upstream; even models accounting for the highest potential rate of flow in a deeper, faster river system do not permit the transportation of a weighty, intact vessel or such a large quantity of individual, accumulated coins.Footnote ⁸⁶

Conversely, at the rural Goring–Streatley crossing, the natural form of the ford itself dictates local hydrological conditions. The Goring–Streatley ford functions as a semi-artificial riffle, a shallow section of the river with a coarse, solid gravel bed where the river’s flow, despite surface turbulence, has significantly reduced capacity to transport heavy objects in suspension or through rolling them across the riverbed.Footnote ⁸⁷ Relatively dense objects, such as coins, dropped into such a hydrological environment would quickly lose any downstream momentum imparted by the rate of flow and become lodged within the interstitial spaces of the gravels, effectively becoming ‘trapped’ in situ.

The accumulation of individual coins here is therefore better explained through repeated deliberate or non-deliberate action at this specific location, rather than through fluvial transportation from upstream. So, at both Kingston’s urban bridge site and Goring–Streatley’s rural ford, the available contextual information appears to indicate a structured, likely intentional pattern of depositional activity, despite the contrasting hydrological conditions at either location.

The question of retrieval

While the specific hydrological conditions at river crossings such as Kingston Bridge and the Goring–Streatley ford provide an explanation as to why river finds might artificially concentrate at these points, the sheer quantity of valuable material deposits raises another fundamental question: why were they not retrieved? It is important to note that the problem of retrieval in watery contexts is not a new issue, but has been a recurrent problem first addressed in some of the earliest archaeological literature exploring river finds. For example, G.F. Lawrence in 1929 noted that the apparent failure to recover valuable objects from rivers is puzzling:

Lawrence’s observation, although predating any systematic study of river finds, directly confronts an often unstated but nevertheless underlying assumption prevalent in archaeological study. Most contemporary interpretations of river finds, whether viewing them as intentional votive offerings or accidental, chance losses, operate on the premise that once deposited, these objects were physically impossible to retrieve.Footnote ⁸⁹

This assumption is difficult to sustain, with historical accounts providing clear counter-examples. For instance, in seventeenth-century Denmark, military accounts describe civilians hiding silverware and other valuables in watery contexts, including lakes, bogs and rivers, during times of war, with the explicit intention of recovery.Footnote ⁹⁰ Though functional rather than votive, these retrievable watery hoards clearly demonstrate the feasibility of recovering objects from riverine contexts. This principle is even more applicable in the specific context of a river ford.

The physical characteristics of a ford make the suggestion of irretrievability particularly tenuous. Fords are shallow, relatively stable and accessible sections of a river, meaning the recovery of a dropped or intentionally deposited object would be comparatively easy. The locations where finds are most concentrated are also locations where physical retrieval should have been easiest, suggesting that the barriers to retrieval were not physical, but perhaps also mental and/or social.

The notion that a prevailing cultural taboo prevented objects deposited in certain contexts, particularly watery places, from being recovered has a strong basis within the Graeco-Roman world.

The role of water in Roman religious beliefs is complex and multifaceted,Footnote ⁹¹ but the potential for watery places, including springs, bogs, lakes and rivers, to be active, sacred spaces is widely acknowledged.Footnote ⁹² Once an object entered such a space, it plausibly ceased to be worldly, terrestrial property and became res sacra, with ownership immediately and irrevocably transferred to the divine. This transfer was rooted in the sacred place itself, a concept perhaps most clearly embodied by the Graeco-Roman temenos, with the example of animals wandering into the precinct of Zeus Temenites, which were instantly consecrated by and immediately became legal property of the god.Footnote ⁹³ Julius Caesar provides another account of this taboo in his Commentarii de Bello Gallico.Footnote ⁹⁴ While focused on cult sanctuaries rather than rivers, he notes that piles of metalwork were left in plain sight as retrieval was strictly forbidden.

Roman literary sources demonstrate this retrieval taboo applying to watery loci sacri. For example, Strabo describes vast quantities of material deposited in both sacred lakes and a local temple, emphasising ‘it was the lakes, most of all, that afforded the treasures their inviolability’.Footnote ⁹⁵ For these deposits, the prohibition against their retrieval was clear: ‘numerous people had dedicated them and no one dared lay hands on them’.Footnote ⁹⁶

The strength of this culturally specific taboo is ironically highlighted by its breach. After the Romans conquered the region, the lakes were sold to private citizens, who promptly dredged them to recover the valuables, demonstrating the barrier was a social one, not a matter of practicality.

Changing legal and social norms

The fundamental restructuring of depositional practice observed across the Middle Thames in the Roman period should not be attributed to any singular phenomenon, but rather to a complex interplay of evolving social norms, new material cultures and a changed relationship with the landscape. Perhaps the most dramatic shift in the overall depositional profile of the Roman Middle Thames is the near-total disappearance of weaponry deposits from the Thames. This reflects a wider trend across the Roman province of Britannia, where the previously well-attested tradition of depositing martial equipment in watery places declines significantly during the Roman period, with the sole exception of frontier, militarised areas.Footnote ⁹⁷

This was likely not the result of any specific legal prohibition or governed control regarding civilian weapon ownership. While commonly misunderstood, Roman laws in this area, such as the often cited Lex Julia de vi publica, were primarily concerned with preventing organised sedition and armed provincial insurrections through criminalising the stockpiling of arms, but did not extend to curtail personal weapon ownership.Footnote ⁹⁸ Widespread disarmament of the provincial population would have been both impractical to enforce and counter-productive within a landscape where banditry and predatory threats to livestock remained a persistent reality.Footnote ⁹⁹

Rather, the change is perhaps best understood as an expression of ‘soft control’ exerted by the Roman state. Soft control refers to an indirect and often intangible modality of influence which blends Foucault’s conception of an inherently diffuse, disciplinary authority with a distinctly ‘hardened’ variety of Nye’s ‘soft power’, informally shaping social norms and behaviours.Footnote ¹⁰⁰ Drawing on Foucauldian ideas relating to informal social control, it depends upon the internalisation of authority and the ‘quiet’ regulation of conduct at the wider socio-cultural level; while in line with Nye’s framework, it also draws on the unconscious cultivation of consent, whereby the values of the governing state are semi-independently reproduced amongst the population without the state taking an active, agentive role.Footnote ¹⁰¹ In this sense, Roman soft control naturalised ‘hard’ authority within everyday life. Yet whereas Nye’s model of soft power relies solely on the principal of cultural attraction, the efficacy of Roman soft control was ultimately underwritten by the latent coercive threat of ‘hard power’ and military force.

The habitual, highly visible presence of a professional, heavily armed and active military, which functionally held a monopoly on legitimised violence, would have rendered any overt or public acts of deposition of weaponry or militaristic paraphernalia a socially and politically inadvisable action.Footnote ¹⁰² The issue of civilian, non-military weapon ownership at the individual level is rarely addressed in Roman literary sources; however, a notable example appears in the fictional Satyricon (late first century a.d.):

This example vividly illustrates the distinction between the formal legality of civilian weapon ownership and the social reality enforced by the Roman military. Here, a civilian is directly challenged precisely because he is carrying a sword, and the initial challenge ‘of what regiment and company are you?’ reveals an underlying social assumption, that the open carrying of a weapon, particularly a sword, was the sole prerogative of the Roman military. The subsequent confiscation of the weapon is a direct example of soft control; it is an informal assertion of the military’s authority over weapons and martial symbols, enforced with an implicit threat of violence. While fictional, this example demonstrates how the ever-present authority of the Roman military could effectively police the public display of weapons without requiring a formal, legal statute, providing a powerful social deterrent against the expression of martial ideas and associations that characterised prehistoric depositional traditions.

The Middle Thames assemblages illustrate this social dynamic, charting a widespread shift that culminates in the effective abandonment of riverine weapon deposition. The assemblage from Goring–Streatley, with its complete absence of weaponry or any militaria, exemplifies a fully realised shift towards a non-martial depositional grammar in a dispersed, rural landscape. This is particularly notable given that weapon ownership in rural settings was likely far more socially acceptable, given the persistent need for personal protection, defending livestock from predators or actively taking part in hunting.Footnote ¹⁰⁴ The absence of weapons finds from the Goring–Streatley crossing does not imply a disarmed local populace, which would have been neither practical nor realistically possible for the Roman state;Footnote ¹⁰⁵ rather it suggests a shift in social norms, where the intentional deposition of weaponry was no longer considered an acceptable or publicly permissible act, even in a setting where such objects would have been commonplace.

The small and problematic group of weapons and militaria from Kingston, however, offers a more multifaceted perspective. While most do appear to be prehistoric or medieval objects, misidentified by their nineteenth-century finders, if the five iron spearheads and dagger (all sadly not illustrated) are Romano-British, they would represent a significant change in the underlying logic governing depositional practice in the region, contrasting with the highly structured mass depositions characteristic of later prehistory and the early medieval period.Footnote ¹⁰⁶

The absence of swords is particularly noteworthy, as they are markedly present in every other period.Footnote ¹⁰⁷ In Roman Britain, swords are quintessential martial objects directly associated with a military identity;Footnote ¹⁰⁸ their absence suggests the Kingston weapons may not represent a martial assemblage. This interpretation finds some support in the ambiguous nature of the items present. Spearheads, for instance, are well attested in both military settings and non-military contexts like boar and deer hunts.Footnote ¹⁰⁹ However, the line between military and hunting equipment was often blurred; even swords and shields could be considered legitimate hunting tools in the correct context.Footnote ¹¹⁰ Secondly, and more convincingly, the dagger suggests a purely civilian interpretation of the weapons may be overly presumptive. Despite their potential use in both military and civilian contexts, daggers and hunting knives (cultri) are conspicuously absent from non-military iconography, including hunting scenes.Footnote ¹¹¹ This absence lends the dagger a more distinct military association with a distinct link to a martial identity, akin to that of the sword, and in stark contrast to the more polyvalent spear.

As discussed earlier, within a Roman provincial context, soldiers held a monopoly on legitimate, public displays of violence and the expression of a martial identity. At an individual scale, military equipment was not only functional, but it also provided a physical medium through which to broadcast a complex, multi-layered identity, inextricably tied to the formal structures and power of the Roman state and military.Footnote ¹¹²

The scarcity of large-scale martial deposits from the Middle Thames stands in stark contrast to the abundant Roman weaponry recovered from frontier zones across north-west Europe, such as Germany’s lower Rhine and Slovenia’s Ljubljanica river.Footnote ¹¹³ The absence of militaria, particularly weaponry, within British rivers appears to be unique within the Western European provinces, with weapon finds from British rivers in highly militarised frontier zones also extremely rare in comparison to the continent.Footnote ¹¹⁴ It seems likely, therefore, that the conspicuous absence of weapons and militaria reflects the civilian character of the Middle Thames region overall and southern Britain more broadly, a point which is further supported by the general prevalence of low-value bronze coins at both case-study sites, given high-value coinage in Britain is predominantly recovered from military installations.Footnote ¹¹⁵

The civilian nature of riverine deposition across the Middle Thames Valley does not necessarily indicate a population devoid of military experience. The recruitment of local auxiliaries and the eventual settlement of legionary veterans across southern Britain suggests that martial knowledge and a military identity would be widespread amongst the population.Footnote ¹¹⁶ The Roman military apparatus, through the imposition of its own internal culture and values or disciplina, effectively blurred the military and civilian identities of its members.Footnote ¹¹⁷

Therefore, while weapons certainly were present in provincial settings, particularly amongst rural populations, the expression of their relationship with weaponry and other militaria or martial equipment would have been heavily influenced by a military ‘subculture’, extending far beyond the occasional presence of armed soldiers.

The distinct absence of weaponry in the Roman period, alongside the abrupt resurgence of martial deposition across the Middle Thames in the early medieval period, suggests that for the wider population, new material in the form of pottery and coinage acted as behavioural proxies, filling the same functional role that martial objects previously occupied, within a new socio-cultural framework brought about through Roman occupation.Footnote ¹¹⁸

The not-quite-right thing in the right place

While the use of coins as substitutes for weapons and militaria might initially appear to be a perplexing choice, coins acting as martial proxies are largely unsurprising within a provincial Roman setting. Coins are strongly associated with a military identity; in Britain this relationship is especially prominent between the first and third centuries,Footnote ¹¹⁹ and the deposition of coins in ‘watery’ contexts is a well-attested practice across the Roman Empire, notably including vast quantities of coins deposited at fording points and bridge sites.Footnote ¹²⁰

The significant rise in the rate of coins deposited within the river during the Roman period cannot be attributed solely to the ‘Romanising’ monetisation of the province and the corresponding availability of coinage. It appears to represent the distinct adoption of an intentional ritualised practice especially in the later Roman period, a notion supported by the concurrent spike in the number of coin hoards, both terrestrial and riverine, from the same Late Romano-British period (fourth–fifth centuries a.d.). The widespread adoption of coin deposition in the Middle Thames therefore appears to represent the direct exchange of one long-standing votive tradition for another, more explicitly ‘Roman’, behaviour.

A closer examination of the coin assemblage suggests that this practice was not monolithic; rather it may have accommodated a range of depositional ideas, some of which appear to demonstrate a continuity of pre-Roman practices. On the one hand, the majority of coins recovered are low-value bronze issues. Their proliferation suggests a potential universalisation of deposition, where the act becomes an accessible and ubiquitous practice in stark contrast to earlier Bronze Age and Iron Age traditions which focused on high-status, high-value objects.Footnote ¹²¹

On the other hand, the presence of gold and silver coins alongside intentionally modified bronze coins suggests a more complex amalgam of ‘new’ and ‘old’ ideas. The deposition of significant quantities of high-value coins appears to echo the pre-Roman emphasis on high-value, socially and culturally inaccessible deposition of prestigious objects. Similarly, the intentional modification of coins, including bending, filing and cutting, mirrors the intentional destruction or mutilation of weaponry and metalwork seen in river finds dated to the late Bronze Age and Iron Age.Footnote ¹²² This act of conscious, intentional modification suggests that, while the material ‘vocabulary’ of what constitutes a martial object changed, the conceptual ‘syntax’ governing the appropriate treatment of such objects was transposed onto new objects. The presence of modified coins at Kingston, but not Goring, highlights the potential for local variations in these practices, demonstrating that deposition represents a negotiation between imported Roman ideas and long-standing local traditions, rather than a singular monolithic tradition.

The sudden influx of Roman pottery presents a greater interpretative challenge. Pottery is ubiquitous in the domestic assemblages of the later prehistoric, Roman and early medieval periods, yet its presence in the Middle Thames is a uniquely Roman phenomenon. The ‘traditional’ Romanist interpretation leans towards a functional explanation, with riverside settlements such as those at Kingston and Goring utilising the river as a convenient means of rubbish disposal, in the same manner identified at Roman fort sites.Footnote ¹²³

Using a river as a convenient means of waste disposal is not a uniquely Roman idea; it represents a commonplace practice throughout prehistory and the post-Roman period, as demonstrated by the riverside midden at Dorestad Wic in the NetherlandsFootnote ¹²⁴ or the systematic disposal of waste attested by archaeological excavations of London’s medieval waterfront.Footnote ¹²⁵

This ‘traditional’ perspective is not without merit; it is entirely plausible that the disposal of waste was the primary driver behind Roman deposition in the Middle Thames. While it may initially appear logical to expect a relatively consistent, low-level presence of ceramic materials from all periods, this is unrealistic when the physicality of refuse assemblages and the hydrological environment presented by an active river system are considered together. The unique characteristics of Roman pottery, with broadly consistent fabrics and high-temperature firing, provide it with far greater durability when compared with later prehistoric and early medieval material, which demonstrates far greater variation in quality and is often porous and friable, rarely surviving even in terrestrial settings.Footnote ¹²⁶

Furthermore, while a midden site in a terrestrial setting would be expected to produce a broad spectrum of refuse material, such as animal bone or mollusc shells, the inherent selectivity of riverine collecting practices would invariably exclude such material. Therefore, the apparent selectivity for coins and pottery observed across the Roman Middle Thames can just as plausibly reflect ‘cherry-picking’ from riverine middens as it does genuinely selective, structured deposition.

However, the interpretation of the Middle Thames assemblage in terms of structured, deliberate deposition is supported by the condition of the finds. The discovery of intact or near-intact vessels is inconsistent with casual discard. At Goring, for example, numerous intact vessels were recovered, alongside at least one vessel with holes deliberately made in its fabric prior to deposition. This may have served a practical purpose, such as suspension or use as a timer, but the act of intentional modification, alongside the potential for display, situates the vessel against a wider backdrop of Romano-British ritual behaviour seen in other watery contexts, such as wells.Footnote ¹²⁷

The intentional modification of objects, including pottery and coins, as well as the intentional selection of particular objects for deposition, indicates a highly structured practice, governed by a largely consistent underlying logic. While remarkably consistent from a long-term perspective,Footnote ¹²⁸ this underlying logic does appear to have been somewhat adaptable. For example, the depositional locales demonstrate a continued focus on specific places, typically crossing points, now re-framed by Roman infrastructure and settlement, while the adaptability of the practice is apparent in the changing material corpus.

Changes in depositional norms through time are not unprecedented in the archaeological record. For instance, in the southern Netherlands, Fontijn highlighted gradual changes between Bronze Age and Iron Age depositional traditions, suggesting that any changes to long-term established behaviours were likely imperceptible to the participants, who would believe they were maintaining an identical practice.Footnote ¹²⁹ However, the Roman Middle Thames appears to present a far more abrupt shift, with Roman social-cultural influence causing the sudden and widespread adoption of the ‘not-quite-right thing’, as depositional practice is adapted in line with the exclusive Roman ownership of a martial identity. The equally abrupt return to weapon deposition in the early medieval period suggests this change was not only conscious, but importantly, did not erase the underlying association between rivers, martial equipment and militaria among the wider population.

The persistence of these ideas, despite the apparent effectiveness of Roman soft control in curtailing practice, may lie in the unique physical characteristics of the river as a depositional context. Unlike other watery contexts, rivers are dynamic and characterised by constant movement and flow. In a flood, scour would periodically unearth and move older deposits, while in shallow areas, prehistoric weaponry and other deposits may have been visible from the surface,Footnote ¹³⁰ providing a tangible and active experiential link to prehistoric ideas, despite their temporary prohibition.