Price and Jaffe (Reference Price and Jaffe2023) argue that acknowledging failure humanises the past. It can also serve as a lens through which to reflect on archaeological reasoning. Here, we turn to the Roman world, and the frontier of northern Britain in particular, to consider how intentionality, distributed agency and moral judgement intersect with the recognition of failure in the past—and with the failures of archaeologists themselves.

Failure is a product of moral judgement, in the past and from the present. Price and Jaffe note that scholars have been more enthusiastic about big-F failures, such as societal collapse, than the small-f failures that characterise daily life. This observation extends to studies of the Roman world where big-F failure is a recurrent theme—and one also deeply implicated in moral judgement: from Gibbon's (Reference Gibbon1776–88) analysis of the failures leading to the empire's ‘decline and fall’ through to Scheidel's (Reference Scheidel2019: 16) contention that the failure of the centralised and inflexible Roman state was a necessary step for instigating the “fragmentation and competition” that formed “an important precondition or source of European development”. Likewise, analysis of cultural change in the Roman world invokes notions of both failure and moral judgement. Is the persistence of ‘Iron Age’ roundhouse architecture in Roman-period Britain a successful, resilient tradition or the failure of the indigenous population to ‘romanise’? A century ago, Haverfield (Reference Haverfield1912: 11–12) characterised the incomplete ‘Romanisation’ of Britain and other parts of the empire as a failure to be explained with reference to a racialised hierarchy. More recently, this narrative has been reversed, reassigning the failure to Rome rather than its subjects; for example, Hingley and Hartis (Reference Hingley, Hartis, Hekster and Kaizer2011: 85) see the construction of Hadrian's Wall as evidence of the state's failure to integrate provincial populations. Clearly, when Rome constitutes a powerful ancestral archetype for Western society, the analysis of Roman failure can never be entirely objective; as a mirror to the present, the failures of the Roman state may be seen to reflect our own societal failures. But in addition to big-F failures, the West's unwarranted sense of familiarity with the Roman past also has implications for the archaeological recognition of small-f failures: if ‘the Romans’ were like ‘us’, then surely it is straightforward to impute Roman intentionality and comprehend ancient perceptions of everyday failures?

A brief consideration of one of Rome's best-known frontiers illustrates how the combination of moral judgement about large-F failures and overfamiliarity in relation to small-f failures may play out in archaeological reasoning. The need for any physical frontiers to delimit the empire has been seen to reflect a wider strategic failure to complete world conquest (Mann Reference Mann and Temporini1974: 508–12). Setting aside the question of whether indigenous populations would recognise frontier-building as a ‘failure’, such a characterisation depends on correctly discerning the intentionality of the Roman state and the changing contexts in which decisions were made and implemented. With almost no contemporaneous textual sources specifying the strategic objectives of Roman frontiers, we must read intentionality from their material remains. The largest, most complex and well-studied Roman structure in Britain is Hadrian's Wall; we have extensive knowledge of its design, construction, modification and use (Breeze Reference Breeze2019) offering unparalleled possibilities for discerning intentionality and agency. For example, the regularity of the structure along its 120km course and its ambivalence towards local conditions (e.g. gateways opening onto cliff edges) points to the existence of a high-level masterplan, perhaps even the hand of Hadrian himself (Breeze Reference Breeze2009). Whatever the plan's intentions, the fabric of the frontier reveals many changes were made during the construction process, some localised (e.g. moving a turret to maximise visibility) and many systematic and wide-ranging (not least the decision to add forts to the line of the curtain wall, in some cases requiring the demolition of structures already started). These changes speak to the distributed agency required to implement the blueprint and the contingent intentions of those involved. Many of the changes have been read, in one way or other, in terms of failure. Comparison of the rigid blueprint with the Wall as built has been seen to reveal failures of planning and the need to correct for an idealised model (Graafstal Reference Graafstal2018: 87). For example, after construction had begun a decision was made to switch to a narrower curtain wall, perhaps reflecting an overestimation of the quantity of material required or an underestimation of the construction timetable. However, such changes might equally be understood in terms of the flexibility of the state and the army to adapt. For instance, was the decision to leave short sections of the ditch to the north of the Wall ‘unfinished’ evidence of localised small-f failures or a recognition that the ditch was unnecessary in some places and could be omitted without compromising the Wall's broader intended purpose? That the term ‘unfinished’ implies an end goal does not automatically imply a failure on the ground. Similarly, the curtain wall was built on unusually shallow foundations and several stretches collapsed and were rebuilt. Was this a failure of design and/or implementation—poor planning and shoddy construction? We must assume the intention of building a wall is that the structure will remain standing rather than collapse, but over what time span? If the objective were to complete the curtain wall as quickly as possible, a structural collapse a century later (as at Wallsend, Bidwell Reference Bidwell2018) would hardly constitute a failure of the immediate objectives. A keenness to identify the builders’ small-f failures may betray our overfamiliarity with the Roman past and misplaced confidence in assuming their intentions.

Elsewhere along the frontier, other examples illuminate the wider intersection between the failures of the past and of the present. For example, a causeway built across the Vallum ditch at Benwell is pierced by a culvert more than 1m above the base of the ditch; this discovery perplexed the excavators who suggested that it might be explained if construction had already begun before the need for a drain to avoid “an excessive accumulation of stagnant water” was recognised and acted upon (Birley et al. Reference Birley, Brewis and Charlton1934: 178–9). Was this a failure to understand the implications of the causeway's construction and to plan accordingly? Or does this reveal an unwarranted assumption that the Vallum ditch should be kept dry? Hingley (Reference Hingley2022: 222–3) has recently argued that the elevated position of the culvert was intentional, serving to maintain the water level to the west of the causeway for symbolic purposes.

A similar intersection of past and present failures comes from the Wall's eastern terminus. Here, the prevailing interpretation is that the slightly later addition of a stretch of curtain wall from present-day Newcastle upon Tyne to Wallsend points to a potential inadequacy of the original plan—perhaps a recognition of the need to provide extra protection of the eastern flank of the bridge over the River Tyne. In contrast, Graafstal (Reference Graafstal2021) has recently reviewed the evidence and argued for the possibility that the Wall had originally been intended to cross over the Tyne and run along the southern riverbank all the way to the coast and the fort at South Shields. In other words, rather than an extension of the Wall intended to resolve an inadequacy of the original plan, the section from Newcastle to Wallsend may represent a reduction in its planned length—perhaps a result of a lack of manpower, a need to hasten completion (both arguably failures of planning), a recognition that a wall south of the river was an unnecessary component (a lesson learned), or because the threat in this area had diminished (adaptation to a new reality). Whether or not we accept this revised model of the Wall's original plan, the reinterpretation highlights how assumptions about intentionality can be viewed differently when the evidence is systematically assessed independently of interpretations developed decades ago based on a smaller dataset. The study of Hadrian's Wall is marked by periodic declarations that the major questions have been resolved and only minor details remain to be slotted into place (e.g. Richmond Reference Richmond1950: 43; Wheeler Reference Wheeler1961: 159). Such bold statements of success betray a failure to recognise the contingent nature of interpretation and the potential of new discoveries to transform interpretations based on fragmentary data. It is only through recognition of such failures that archaeology progresses. Our own failures are therefore as worthy of study as are the failures of the Roman past.