A Roman floor could show a hunt, a gathering of dolphins or a pattern of intersecting lines. From the doorway the image might look almost like painting, yet close up it resolved into thousands of small pieces pressed together in mortar. These pieces, called tesserae, were generally made from stone or pottery, with glass used for some colours and details. The cut and placement of each piece gave the image its lines, shading and colour.
Before the picture could be laid, builders had to construct a floor capable of carrying it. Workers first needed a stable base, then a plan suited to the dimensions of the room, and finally a way to arrange the tesserae in mortar before it set. A large monochrome pattern and a finely shaded scene shared important construction methods. They differed considerably in how much material, labour and artistic control they required.
Two ancient authors, the architect Vitruvius and Pliny the Elder, provide part of our evidence. Neither wrote a complete guide to Roman mosaic workshops. For the work itself, archaeologists also study surviving pavements and their bedding layers, removable picture panels, pigments and traces found during conservation. Together, the written and physical evidence helps reconstruct both the work of making a mosaic and the circumstances in which it was made.
What Vitruvius and Pliny Actually Tell Us
Writing late in the first century BC, Vitruvius set out instructions for flooring in his treatise On Architecture. The passage in Book 7, chapter 1 begins with the ground below a pavement and follows the building layers laid on top. His instructions aimed to prevent settling, cracks and damp. The chapter discusses different pavements, among them stone slabs and surfaces laid with tesserae. It does not tell us how an artist cut coloured stone or transferred an intricate image from a sketch.
Vitruvius’s specifications must be read as prescriptions, rather than as a record of every floor Roman builders laid. He recommends particular mixtures and minimum depths for some layers, with additional precautions when a floor rests on timber. Those instructions reveal one Roman author’s idea of sound construction. Excavated floors are often different. We cannot assume that all ancient builders followed his measurements or used the same method in every province.
Pliny, writing in the first century AD, was interested more in what a mosaic could depict. In Natural History 36.184, he praises Sosus of Pergamon for an “unswept room” floor whose scattered scraps of food looked real. He also describes doves drinking around a vessel, with one bird’s shadow falling across the water. Pliny was recording an artist’s reputation and the visual effects achieved with coloured pieces. He was not describing the people who prepared the mortar or laid the tesserae.
A Roman dove mosaic from Hadrian’s Villa at Tivoli, now housed in the Capitoline Museums, shows a subject similar to the one Pliny celebrated. It is often interpreted as a later version of Sosus’s Pergamene work. It is not the original pavement itself, and the resemblance cannot make it so. Pliny’s decision to name Sosus also does not mean he was the only accomplished mosaicist in antiquity. Pliny preserves a particular artist’s reputation, not a complete account of the craft.

Building the Floor Beneath the Picture
A durable mosaic needed a firm foundation. Vitruvius begins by telling builders to examine and compact the soil. On solid ground, they might place a layer of fairly large stones, known as the statumen, before adding aggregate mixed with mortar. Timber-supported floors demanded special treatment. Wood could shift as it dried or carried weight, and movement in the base could crack an otherwise carefully made surface.
Next came the rudus, a coarse layer of rubble or broken stone set in lime mortar and thoroughly compacted. In one prescribed arrangement, Vitruvius gives it a thickness of about three-quarters of a Roman foot. Above that he describes the finer nucleus, made with crushed pottery and lime and at least six Roman inches deep. These measurements apply to the construction he is describing; they are not dimensions that should be imposed on every excavated floor. Even modern summaries sometimes confuse the mixtures or the ratios of lime to aggregate, and the upper setting bed should not simply be equated with the nucleus.
The top mortar immediately below the tesserae had to be soft enough to receive them. Builders could complete the heavier supporting layers earlier, then prepare small working areas of bedding as the mosaicists advanced. The cubes went into fresh mortar as work advanced; they were not glued onto a finished slab. Once the bedding hardened, it gripped the tesserae and helped bear the repeated pressure of people crossing the room.
Research on surviving floors has shown just how varied these arrangements were. A 2019 non-invasive investigation of floors at the Roman villa of Pisões in Portugal used ground-penetrating radar and photogrammetry to examine their structure and condition without lifting them. A separate study of mortars from Gerasa in Jordan, published in 2024 detected differences in the materials and construction of floors from different periods. These results can be compared with Vitruvius’s instructions without treating his method as universal.
Cross-sections of Roman mosaic floors often show a neat succession of stone, coarse mortar, finer mortar, bedding and surface tesserae. The familiar five-layer section is a useful illustration, but excavated floors do not always follow it. Excavators also find repairs, levelling deposits and other variations required by the building beneath the decoration.

Where the Tesserae Came From
For most floors, builders needed great quantities of stone that was hard enough to endure foot traffic. Pale limestone, darker stone and different marbles provided common colours. Pieces cut from ceramic tile or broken pottery could add reddish and orange tones; glass was useful for colours that local stone could not easily supply, especially in small pictures. Material choice involved cost and availability as well as appearance. Covering even an ordinary room meant obtaining thousands of pieces of reasonably consistent size and colour.
The Getty Museum’s survey of Roman mosaics notes that mosaics usually relied on local stone for most of the pavement, with imported marbles and glass used where particular colours were wanted. How much a workshop imported depended on quarries, trade connections and its patron’s resources. Black-and-white floors could demand considerable skill without expensive foreign stone. Equally, a handful of brightly coloured glass tesserae does not mean the whole pavement was made from luxury materials.
Stone and ceramic were broken or cut down, then trimmed into pieces that would fit the design. The regular courses of a border favoured similarly sized cubes, whereas an eye, a curved outline or a subtle passage of colour could call for smaller, uneven pieces. Tesserae sometimes preserve marks made during cutting, but no Roman instruction survives prescribing one tool, workbench arrangement or output for all mosaic makers. Hammers, chisels and cutting supports are reasonable reconstructions based on the craft; Vitruvius does not set out that working routine for us.
The successive floors at Fishbourne Roman Palace near Chichester offer a particularly useful comparison. Its earlier geometric mosaics and later, more elaborate work reveal changes in materials and technical range. The second-century Cupid on a Dolphin floor uses red-gloss Samian pottery among its stone pieces. Because Samian pottery can be dated independently, these pieces may help establish when the floor was laid. It does not follow that every part of the floor was built from broken or reused vessels.
Conservation at Fishbourne revealed another unexpected part of the building’s history. When the Cupid on a Dolphin pavement was lifted, an older geometric mosaic lay beneath it. A conservation record indexed by the Smithsonian describes the discovery and the design preserved on the earlier surface. At least one room, then, received a new decorated pavement over an earlier one. This room had been redecorated with a new pavement laid over an older one.

Planning a Design to Fit the Room
The room determined much of the design. A border had to meet the walls and doorways, and central pictures needed an orientation that made sense to those entering, walking past or reclining nearby. A corridor suited a repeating pattern that could extend along its length; a dining room might have a central scene meant to be seen from the diners’ positions, with surrounding motifs that worked from several directions. Their plans had to account for the room’s size, entrances and intended uses.
Planning probably began with measurements and guidelines, including the main axes of the room and the areas set aside for borders and central scenes. Straightedges, string and compasses could establish lines and curves. Maintaining an even meander, an interlacing band or a braided guilloche pattern across a large floor required careful spacing. Archaeology seldom records the exact order in which a particular workshop marked those guides. The likely use of these tools does not establish the exact sequence followed in any particular workshop.
The Getty’s 2016 collection catalogue suggests that mosaicists probably referred to sketches or drawings, as other ancient craftspeople did. Motifs found at widely separated sites make some exchange of visual models likely. They do not demonstrate the existence of a standard pattern book from which every workshop copied its designs. Artists might have shared training, travelled between commissions, followed familiar decorative conventions or adapted a drawing to a room in front of them.
A British Museum panel offers more direct evidence for one preparatory method. Researchers examined an opus vermiculatum panel in a terracotta tray held at the British Museum. Their 2010 technical study identified traces of Egyptian blue, hematite and carbon-based pigments on the mortar beneath its tesserae. They interpret these remains as a coloured preparatory drawing. The finding reveals something of the artist’s sequence on this particular object. It cannot establish that all Roman mosaic workshops drew their designs onto mortar in exactly the same fashion.
Laying Tesserae: Broad Fields and Fine Detail
For a comparatively simple floor, the maker could mark out the border and then work across the areas inside it, placing rows of tesserae close together. Consistent spacing helped give the surface its coherence even if no two cubes were precisely alike. Modern mosaic specialists call the lines of successive pieces andamento. Their course matters to the picture: a row curving around a face or an animal can clarify its outline as effectively as a change of colour.
Archaeologists commonly use opus tessellatum for mosaics made with tesserae of ordinary paving size. The term opus vermiculatum describes finer pictorial work, especially associated with Hellenistic and Roman mosaics, in which much smaller pieces follow the contours of a figure. The two techniques could appear together in a single pavement. A broad field of regular rows could surround a finely shaded central image in the same room.
The Alexander Mosaic, excavated in Pompeii’s House of the Faun, shows what skilled workers achieved with very small tesserae. Alexander the Great rides against the Persian king Darius III in a composition generally associated with the Battle of Issus, although scholars have debated which battle is depicted and whether the design reproduced a lost Greek painting. Tiny tesserae create the expressions on faces, the surfaces of armour and the movement of horses. The Smarthistory account of the mosaic gives a figure of roughly 1.5 million pieces. It suggests the immense labour involved, not an exact tally recorded by the ancient craftspeople.
A mosaicist chose tesserae according to what they needed to achieve. Larger cubes filled a plain ground efficiently; much smaller ones could give a horse its eye or make shadow pass gradually into light. Differences in colour and contrast mattered too. A dark border might separate a figure from the background, while adjacent stones of warmer and cooler shades gave the impression of rounded forms. From a distance, the scene forms a coherent image; close up, the arrangement of individual stones becomes apparent.

How Workers Installed a Mosaic on Site
Most mosaics formed part of the building in which they were laid. Workers prepared the supporting layers, spread fresh bedding over an area they could manage and set the design in sections. They had to complete each stretch before its mortar became too firm, while ensuring that its lines met the adjoining parts. Poor alignment at a join could remain visible as long as the floor lasted. The division of labour probably differed with the room and the complexity of its design.
Modern specialists distinguish a direct approach, with each piece pressed into bedding face-up, from an indirect approach, with tesserae first arranged face-down and later transferred. The two approaches help explain possible methods, although the evidence for ancient practice is less complete than modern terminology might suggest. Portable images made in trays survive archaeologically. That evidence does not prove that Roman mosaicists routinely used paper-backed transfers, or that complete floors were installed through a single standard indirect system.
The Baths of Neptune at Ostia provide a different example of accomplished floor design. In this public bathing complex, dolphins, marine gods and other sea creatures spread across great black-and-white surfaces. The archaeological park assigns these mosaics to the reign of Hadrian, AD 117–138. Instead of modelling every subject with many colours, the workshop relied on strong outlines, scale and the arrangement of imagery between connected rooms. The outlines and placement of the figures show what skilled designers achieved with two contrasting colours.
When the pieces were in place, the surface still needed attention. Vitruvius describes levelling and smoothing pavements so that uneven edges did not protrude underfoot. His treatment covers floors generally; it does not establish that every figurative mosaic was polished mechanically to the same finish. Surviving pavements show differing degrees of wear, uneven repairs and changes in their mortar, because the surface continued to change through use and repair.

Portable Picture Panels and the Older Greek Tradition
Some luxury mosaics included a small picture made separately from the surrounding pavement. Known as emblemata, these panels could be constructed in stone or terracotta trays with especially fine tesserae, then carried to a building and installed within a prepared space. This practice was associated particularly with Hellenistic and early Roman work. The Getty catalogue discusses such portable pictures and the way they were incorporated into larger floor compositions.
The British Museum panel examined in 2010 remains in its terracotta support. Pigment traces beneath its tesserae offer rare material evidence for a preparation stage in its manufacture. They do not, however, reveal where its workshop was located. A technical resemblance to mosaics made elsewhere does not prove that this panel came from Pergamon, Alexandria or any other named centre. That question remains unresolved.
Greek mosaic artists were making sophisticated pictures long before Roman workshops adopted cut tesserae on a large scale. The stag hunt from Pella, in ancient Macedonia, dates to approximately the late fourth century BC. Hunters and hounds pursue the animal in a design bearing the name Gnosis. The floor was made from pebbles, not as a Roman opus vermiculatum panel. Its framed central image offers a comparison with later luxury compositions, but it should not be described as an early surviving example of a Roman prefabricated tray.
The pebble work at Pella belongs to an older Greek and Hellenistic tradition from which later Roman mosaic techniques developed. Greek and Hellenistic makers explored ways to produce pictorial effects on floors, including the gradual use of deliberately cut pieces rather than naturally rounded pebbles. Roman patrons inherited that experience and expanded the kinds of commissions available. A workshop might make a finely worked portable image, cover a set of rooms with geometric designs or bring both approaches together in one floor.

Who Did the Work?
A completed pavement rarely reveals how many people contributed to it. Someone had to obtain stone from quarries or suppliers, cut it to usable sizes and transport it. Others prepared lime, built the supporting layers, measured the room, planned the composition and placed the pieces. Some of these jobs called for an artist’s judgement, others for construction expertise or long hours of physical work. A complex floor almost certainly involved several people, although excavated pavements rarely allow us to count the members of a particular workshop.
Late Roman documents refer to different kinds of decorative specialists, including painters and people associated with mosaics. Their labels cannot be used to impose one fixed hierarchy on every Roman workshop over several centuries. The pictor imaginarius, for instance, was connected with the painting of figures or images; the title alone does not prove its holder always laid tesserae. The names of those who cut stone, mixed mortar and set tesserae are rarely preserved. Owners and depicted subjects are often easier to name than the makers.
Some pavements were installed during renovations rather than during the first construction of a building. A patron might replace the decoration in one room while leaving other rooms in use. Fishbourne’s changing pavements, including floors laid one above another, preserve just such a history of renovation. They do not represent a single huge commission completed in one year. Estimates of labour or quantities of tesserae must therefore be based on the individual floors that have been excavated, not simply on the total area of a villa or palace.
What Archaeologists Can Learn from the Floors
A mosaic’s layers help establish when part of a building was constructed or altered. A pavement covering a wall or earlier floor must have been installed afterward. Pottery and coins securely sealed in its foundation may establish the earliest possible date for that installation. Objects found in the rubble above it can sometimes help date its abandonment or destruction, but their presence does not mean the pavement was laid in the same year. Dating a floor therefore depends on establishing how those objects relate to its construction and later use.
The older pavement beneath Fishbourne’s Cupid on a Dolphin makes the sequence of redecorations visible. The use of Samian pottery as tesserae offers another clue, but even well-dated ceramics could be stored or reused long after they were made. A sherd therefore does not deliver an exact date for the floor on its own. Its evidence has to be checked against building phases, decorative styles and the other finds recovered from the room.
Road construction at Lod, in modern Israel, uncovered a large mosaic floor in 1996. Its panels depict animals, birds, fish and ships. The archaeologist Rina Talgam explains that the pavement probably belonged to a prosperous house and dates from the late third or early fourth century AD. Much of the associated walling has disappeared, so the original purpose of the room is uncertain. There is no dedicatory inscription naming its patron. Nor can the subjects depicted on the floor alone establish that person’s religion or occupation.
Mortar and pigment analysis can reveal details that are invisible in the finished picture. Microscopy and chemical testing can identify pigments, aggregates and the materials binding mortar together. Sometimes those ingredients distinguish an older recipe from a later building campaign. The Gerasa mortar study examined Roman, late Roman and early Islamic material spanning around 650 years and found both continuing practices and changes in production. It offers evidence for that region and sequence, not a formula that can be applied indiscriminately to every Roman mosaic from Britain to Syria.
Removing a mosaic to conserve or display it can uncover earlier work, as happened at Fishbourne. It can also disturb the evidence connecting the surface to the structure below. Conservator Thomas Roby explains why conservators now prefer, where possible, to stabilize pavements where they were found. A photograph records the decoration; plans and conservation records preserve the bedding, repairs and architectural relationships that help explain its history.
Why Roman Mosaics Look Different Across the Empire
The techniques of setting tesserae spread across the Roman world, while choices of stone, colour and subject varied by region. In Britain, builders could use local chalk, limestone, shale and ceramics, while workshops developed their own favourite arrangements. Black-and-white pavements were especially popular in central Italy from the late Republican period onward and remained common for centuries. As the Getty catalogue emphasizes, shared knowledge of tesserae and mortar did not prevent provinces and local workshops from developing recognizable decorative preferences.
In the eastern Mediterranean, workshops drew on long-standing Hellenistic traditions without producing one uniform style. Antioch is known for richly coloured scenes from myth. At Lod, by contrast, the composition and some of the images have parallels with North African and other western Mediterranean mosaics. Talgam argues that these connections help explain the pavement better than simply treating it as another example of nearby Antioch’s work. An elaborate floor in Palestine cannot by itself identify Alexandria, Antioch or any other city as the workshop that supplied its design.
At more than seventeen metres long, the Lod pavement has room for large marine animals and ships arranged in successive sections. Its effect depends on subject, strong framing and contrasting colour, as well as the skill with which individual tesserae were set. The groups of images could be encountered in sequence as a visitor moved through the room. The Getty’s discussion of Roman floor decoration treats that relationship between a pavement and movement through a building as an essential part of understanding Roman mosaic design.
The contrasts are visible in the surviving pavements. Ostia’s sea creatures occupy expansive black-and-white designs; the Alexander Mosaic achieves unusually fine pictorial modelling; Fishbourne records changing tastes and rebuilding; and Lod fills broad panels with animals and ships. Each floor still depended on a sound base, appropriate materials and careful planning. The room’s function, local materials, artistic inheritance, the patron’s resources and the craftspeople’s decisions shaped each result.

Why So Many Roman Mosaic Floors Survive
Stone and ceramic endure in ways that timber, cloth and many painted surfaces do not. Even where water, tree roots or rebuilding damaged a Roman pavement, tesserae embedded in mortar could remain beneath later deposits. That durability is one reason archaeologists know much more about Roman floor decoration than about ordinary textiles, furniture and wall hangings. The durable materials of the pavements explain part of that imbalance in the archaeological record.
Survival has not always meant preservation of context. Many floors were detached for museums before their original buildings had been properly recorded. Such panels may still display extraordinary craftsmanship, but say less about the people who entered, dined or worked in the rooms they once covered. When a mosaic remains in its setting, thresholds, repairs, bedding and neighbouring spaces help recover its purpose. To understand how that floor was made and used, archaeologists need its foundations and setting as well as its decoration.









Discussion
Comments