Pairs of deep grooves cross exposed limestone at many places in Malta and Gozo. Some run alongside one another, some intersect, and others survive only in disconnected lengths. They are usually called cart ruts, a name that already suggests an explanation. Their formation, purpose and age have nevertheless remained subjects of archaeological debate, with proposals ranging from transport wear to deliberately cut agricultural furrows.
This is a landscape mystery with substantial physical evidence. The important questions concern what can be inferred from that evidence: whether a groove’s present shape records its original use, whether nearby archaeological remains date it, and whether a measurement from one site can stand for tracks across the islands. Original investigations have made progress without supplying one demonstrated age and function for every surviving rut.
A scattered landscape, rather than one monument
Misraħ Għar il-Kbir provides a striking example, illustrated above, but the phenomenon extends beyond that locality. Huw S. Groucutt’s 2022 study in the Journal of Archaeological Science: Reports describes field visits there and at its eastern group, San Pawl tat-Tarġa near Naxxar, Binġemma and Xemxija. Those visits supplied landscape context for analysis of measurements previously published by other researchers.
The study is therefore neither a newly excavated cart nor a complete fresh survey of every track. Its quantitative work compiles existing measurements, principally from the investigations of Ventura and Tanti and the catalogue edited by Magro Conti and Saliba. Different measurements have different sample sizes because the earlier records do not contain every dimension for every rut. That qualification matters when reading its averages.
Earlier mapping also had limits. The University of Malta record for H. S. Gracie’s 1954 paper reproduces his explanation that time prevented a whole-island examination, with relatively few observations in the low-lying southeast. On the published maps, closely packed parallel tracks could be represented by one line, and gaps through cultivated land were not joined by conjecture. A surviving map line should consequently not be treated as an uninterrupted road surveyed end to end.
What the dimensions actually show
Groucutt found fairly clustered dimensions, together with real exceptions. In the compiled dataset, individual grooves had an average top width of about 26 centimetres and a median of 23 centimetres. Average depth was about 20 centimetres, with a median of 19. These are summaries of recorded examples, not specifications followed by every track maker. Much wider and deeper examples occur.
One particularly useful correction concerns the word “gauge”. The often-repeated resemblance between Maltese cart ruts and modern railway gauge compares measurements taken differently. Railway gauge is measured between the inner faces of the rails. The approximately 140-centimetre measurement commonly quoted for Maltese ruts is generally taken between the midpoints of the two grooves. Those are not equivalent reference points.
In the study’s dataset, mean midpoint-to-midpoint distance was about 139 centimetres, with a median of 140. Mean distance between the inner edges was about 111 centimetres, with a median of approximately 113. The numerical similarity to railway gauge therefore does not demonstrate a historical connection with railways. Measuring between groove midpoints may still be useful for estimating the spacing of vehicle wheels; the point is to state which quantity is being compared.
The midpoint measurement was relatively consistent compared with the other width measurements. Within individual pairs at Naxxar, recorded widths also tended to vary less than widths between different pairs. Such regularity supports investigation of a reasonably consistent device or practice. It does not, on its own, identify the vehicle, cargo or century: similar equipment can be used for a long time, and later wear can alter the dimensions.
Could ordinary carts wear the rock?
Derek Mottershead, Alastair Pearson and Martin Schaefer’s 2008 Antiquity paper approaches the question through geomorphology. Its published abstract proposes two-wheeled carts, approximately 1.40 metres between wheels, carrying moderate loads. Wet-weather traffic could erode limestone progressively; once the grooves reached the vehicle’s ground clearance, traffic would need another route. This offers a physical reason for multiple nearby tracks rather than requiring a purpose-built railway.
That is a proposed formation model, not the recovery of a cart from a rut or a direct date for its first use. Groucutt’s later field observations supply further reasons to take transport seriously. At the Naxxar Gap, tracks curve uphill along the gentlest available passage through an escarpment. At Misraħ Għar il-Kbir, some pass across a high point, making a simple water-collection explanation less persuasive for those examples.
An agricultural alternative
Claudia Sagona’s 2004 Oxford Journal of Archaeology paper challenges the transport interpretation for clusters of roughly parallel, deep grooves. She proposes deliberately made field furrows associated with efforts to extend arable land. Cut rock, retained soil and managed runoff would form parts of agricultural rehabilitation, and the exposed grooves would be traces of places where that effort eventually failed.
Her proposed Temple Period setting rests on an argument about agricultural knowledge, social organisation and economic pressure. It is an interpretation, rather than a direct scientific measurement of the grooves’ age. Groucutt subsequently questions how an agricultural-drainage explanation fits some steep sites, where rapid soil loss would be a problem. This disagreement is concrete: it asks whether the actual slope, drainage and arrangement of tracks suit the activity proposed for them.
Tool marks, dissolution and later change
A separate question is whether people deliberately cut the grooves before using them. In The origin of Maltese cart-ruts: cut by wheels or tools?, Mottershead examines published photographs interpreted as tooling evidence. He argues that weathering can expose natural joints and other small forms resembling workmanship, while the curvature of some supposed pick marks is difficult to reconcile with an efficient cutting action.
He also reports comparisons with exposed quarry surfaces where deliberate cutting is certain, but where similar marks were not found. His criticism concerns the strength of the particular tooling interpretation; it does not establish that people never cut or improved any track. The broader methodological point is that an apparently shaped rock surface can result from more than one process, so a proposed tool mark needs independent support.
Groucutt likewise records clear signs of limestone dissolution, including hollows within groove bases and water moving through tracks after rain. At some points, ruts descend into natural depressions that complicate an image of carts travelling directly over the present exposed surface. Earlier soil cover, mechanical abrasion, repeated traffic and later dissolution may have contributed in different proportions. Present-day depth need not be identical to the depth during use.
The 2022 analysis explores relationships between width, depth and cross-sectional shape, but describes these as exploratory. The correlations are weak and do not separate competing histories conclusively. A deeper or more V-shaped rut cannot simply be declared older from those plots. The study calls instead for more continuous three-dimensional measurements and close examination of tracks within their landscape settings.
The dating question remains distinct
A useful distinction is between relative sequence and calendar age. Where archaeological features intersect, their relationship may show which came first; identifying the age of one feature can then constrain another. A nearby settlement, by contrast, does not automatically demonstrate that its inhabitants made every adjacent groove. Neither impressive depth nor a resemblance to an ancient transport system supplies a calendar date by itself.
Joel Grima’s 2026 preprint abstract argues, from a wider survey and review of dating evidence, that some ruts are Roman and that a broader Classical setting is supported for many. It states that firm evidence for a prehistoric pair is lacking while not excluding that possibility. These are the author’s stated conclusions; the abstract does not justify turning every Maltese rut into an individually dated Roman road.
A later journal publication record exists for that study. Its full final text was unavailable during this review, so this account does not reproduce detailed final-version statistics or claim to have checked its individual archaeological examples. The older formation and measurement studies remain useful, while the newer chronological argument requires that clear source limit.
The best-supported general interpretation in the studies examined here is human use, especially transport, followed by modification of the limestone. The remaining research problems concern particular routes, equipment, workloads, environmental changes and dates. Those questions can be tested through better recording and archaeological relationships. They do not require treating an unresolved date as evidence of a supernatural origin.
Approximate map area. The pin marks a rounded reference for the main Misraħ Għar il-Kbir group, at 35.852°N, 14.397°E, based on the field locality recorded in Groucutt’s methods section. It represents one documented example within a Malta-wide case. It does not locate every track on Malta and Gozo, an exact original cart position or a single event of construction.
Source note: Groucutt’s complete nine-page 2022 paper, including its measurement definitions and tables, was read; its figure showing the measurement points was inspected. This original summary of that study is credited to Huw S. Groucutt under CC BY 4.0. Mottershead’s six-page tooling discussion, selected argument and chronology sections of Sagona’s paper, Gracie’s repository abstract, the 2008 abstract and Grima’s preprint abstract were also examined. No journal figure is reproduced. Related reading: Antikythera mechanism and Other Mysteries. Checked 5 November 2026.