Download file in PDF format: TRAC 1993: Theories of Circulation in Roman Houses (pp. 75–98)
Theories Of circulation in Roman houses
Simon P. Ellis
This paper is an attempt to characterise the key elements of Roman
housing as first developed in republican Italy and particularly as spread,
or distributed, through the provinces of the empire – the area under their
political control from the first century to the fifth century AD. The main
instrument of this movement of architecture and culture was the elite of
Italy and the provinces and so this paper will tend to concentrate attention on the richest houses in each settlement Nevertheless in some areas
the number of people who adopted “Roman housing” was larger than in
others and we will be studying housing of what in modern times would
be termed the middle class of tradesmen or landowners.
In another paper in this volume Nevett characterises the housing and
society of the ancient Greeks, which together with Italic and Etruscan
cultures helped to form the social views and provided the architectural
vocabulary of Rome. The house with central courtyard and surrounding
porticoes, which later became the preferred house of the Roman aristocracy, developed in the Hellenistic world of the central and eastern
Mediterranean. The central courtyard provided a convenient place for 76 S. ELLIS
cooking and for collecting water in a large underground cistern. The
various aspects of the surrounding rooms enabled them to catch a breeze
in summer and give shelter in winter. Greek society tended to restrict the
public role of women both in public and at home (Nevett below; Walker
1983). To maintain this cultural tradition the Greeks devised a reception
room, the aV&Qov, to which access was limited. The Greeks also invented
the art of mosaic which was used to form the floor of the richest Hellenistic houses. Many houses with these characteristics can be found at
the sites of Priene, Olynthos, and Delos.
Roman domestic society differed from the Greek in two fundamental
respects. Firstly women, particularly the woman in charge of an
important family or household, had more of a public role (Treggiari
1991). The Roman house therefore did not need a reception room with
access restricted to men. Secondly the Romans had institutional slavery.
Although the Greeks had slaves, slave-owning never reached the scale it
would in the Roman world. The Roman aristocrat had to cope with a
large household that included people with a wide range of social
Whether because of earlier Roman traditions or because of local Greek
and Campanian influences the houses of Pompeii retained a reception
suite near the front entrance to the building. This took the form of the
atrium, a covered court with central pool and the tablinium which opened
off it (Dwyer 1991). Pompeian houses also adopted the peristyle of the
Greeks, setting it further back in the house beyond the atriumltablinium
suite. The additional depth to the house, resulting from the peristyle,
allowed the owner to make some distinctions between access arrangements. Entrance to the peristyle could be restricted to the immediate
family but usually access to the peristyle and to the main reception room
that lay beyond (here termed the triclinium for convenience – see Rebuffat
1970) was open to some guests as a mark of privilege.
The atrium, with some notable exceptions, is not found in provincial
Roman housing. The reason why the room was not adopted in the
provinces is a matter of some dispute. It is possible that by the mid-first
century AD Italians regarded it as old fashioned, or provincials may have
felt it inappropriate. Thebert (1987, 325-326) has argued that there were
no atria in the Roman houses of North Africa because they followed local
architectural traditions. The lack of atria in provincial houses reduced the
number of reception suites available and thus made it harder to control
access to the house.
Provincial Roman houses, where there was a reasonable unencumb-THEORIES OF CmCULA 110N IN ROMAN HOUSES 77
ered space in which to build, were organised around a peristyle of
varying size. Grahame demonstrates in this volume, using the model of
Hillier and Hanson, how the peristyle was central to circulation within
Pompeian houses. In provincial houses without atrium or tablinium the
peristyle had an even more central role in the circulation of visitors.
The centrality of the peristyle meant it was very difficult to restrict
members of the household to one particular area in the house. If the
household was to have any privacy then there had to be some way to restrict access to rooms in which private activity was taking place. This
point applies equally to visitors, the owners’ family, and servants. In the
case of family and servants, familiarity with the house and the habits of
its occupants would soon solve this problem. It was more difficult in the
case of visitors.
I shall argue that the decor and architecture of the house was used by
the owner to prevent visitors from entering inappropriate rooms. If visitors strayed the atmosphere created by the room would immediately tell
them they had taken a wrong turn.
THEORIES OF CIRCULATION
In the last decade an increasing number of theoretical approaches have
been applied to the study of the Roman house. A fundamental part of
each theory has been some assessment of the pattern of circulation
through the house.
Circulation has main! y been assessed on the basis of the distinction
between the public and the private. For this we should read visitors and
the owners’ household, though this definition in itself is often loosely
defined. Analysis has also been undertaken on the distinction between
servant and owner, though work in this area is less common than one
We shall begin by examining two major theories of recent years: those
of Carandini and Wallace-HadrilI.
Carandini (1985) published his theory in his report on the excavation of
the villa of Sette Finestre. The starting point is his adoption of the architectural division of the house into a pars urbana and a pars rustica
(Columella De re rustica 1.6.1). From this he assumes that the former is the
domain of the owner and the latter the domain of the servants. This
translates into the architecture by identifying all rooms which had an agricultural or artisanal function as the domain of the servants, and all rooms
that were used for receptions – dining rooms, bedrooms, and others with 78 S.ELLIS
fine mosaics – as the domain of the owners, family, and friends.
The circulation of the owner through the villa is marked by a blue line
and the progress of the servants is marked on the plans by a red line. Occasionally they appear to meet in a room, and here Carandini accepts that
there was some interplay between the two classes. Thus servants would
enter the dining room to present food, but fundamentally the room was
in the domain of the owner. Likewise the owner would of necessity have
to visit the presses to see how the vintage was progressing, but
fundamentally the press rooms were the domain of the servants who
worked there. In some ways the model is attractive. It presents a tangible
expression of how intrusive a servant must have felt at a grand dinner
party. Clearly there were parts of the Roman house which were very
much the home of the owner, and others which were very much the
The major problem with the model is the assumptions it makes of the
archaeology. We have no reason to assume that every richly decorated
room belonged to the owner or that only servants worked in rooms such
as the kitchen and the workshops. For example a Greek pedagogue might
well have a richly decorated bedroom in the house of an aristocrat, while
the gentleman fanner might well have his own private workshop for
making a variety of implements or fermenting a favourite brew.
Equally the evidence for assigning a room to a particular class may be
lacking. A reserve bedroom might be poorly painted. A servant might
Jive in a room whose decoration was relatively rich, but unfashionable.
The Laurentine house of the Younger Pliny contained a suite of rooms
which he says were suitable either for servants and freedmen or for receiving guests (Letter 2.17.9-10). This suite moreover is close to his
personal library, which contained all the books he read most frequently,
and which we would expect to be one of the most intimate rooms in the
house. It is not important here to question Carandini’s particular interpretation of Sette Finestre. It is enough to have demonstrated that it is not
possible to identify the functions of rooms sufficiently to divide them into
the two classes he suggested, and that there is considerable reason to
doubt that some rooms could be assigned to one or other social class.
Wallace-Hadrill’s theory of circulation is much more subtle than that of
Carandini. He immediately points out the wide-ranging nature of the
Roman household and clientele. He dismisses the idea of separate public
and private parts of the house. Instead he proposes a rather amorphous
schema in which particular rooms are more suitable for certain ranks of
guests or household. He illustrates this with a diagram (1988, 55): THEORIES OF CIRCULA nON IN ROMAN HOUSES 79
Though this is an excellent study of social history it is difficult to see
how this diagram translates into the archaeological record. The impression given by Wallace-Hadrill is that there was a greater distinction
between the circulation patterns of servant and household than there was
between the circulation patterns of household and visitor.
In examining the distinction between servants and household he identifies groups of secluded, poorly decorated rooms as servants’ quarters in
some Pompeian houses (the House of the Vetii and the House of the
Lovers – Wallace-Hadrill 1988, 79 and 82). Without disputing these
particular interpretations there are reasons to question such a clear distinction. As we have seen, such secluded rooms were used by Pliny for
servants or guests. Servants slept in many ill-defined parts of the house,
often right outside the rooms of their masters (Veyne 1987, 73; Treggiari
When discussing the relationship between household and visitors
Wallace-Hadrill makes the important point that Roman aristocrats would
receive honoured guests in their bedrooms for a Louis XIV style levee.
This is developed into an argument concerning a broad progression from
the more public areas of important paths of circulation, through open
reception areas, to more private areas such as the bedroom (WallaceHadrill 1988, 59 and 85). None of these rooms are exclusively public or
private, but they represent a continuum – the horizontal axis of his
model. The distinction between the levels of access given to various
visitors of different social standing is well drawn out Unfortunately, it is
not clear how these distinctions between visitors would be worked out
on the ground. As a result Wallace-Hadrill’s theory is one about the
function of houses in general within Roman society rather than a theory
about circulation and access between various rooms. 80 s. ELLIS
THE THEORETICAL MODEL OF HILLIER AND HANSON
Hillier and Hanson (1984) use a model of geometrical structure to analyse
the plans of both houses and settlements. Grahame has applied their approach to Pompeian houses elsewhere in this volume. Because Hillier
and Hanson’s theory assesses the articulation of spaces within the house,
and their links with each other, it is in many ways the only theory that we
have considered that can genuinely be called a theory of circulation.
Hillier and Hanson elaborate their theory along two principal dimensions, which they describe as “relative asymmetry” (1984,
151-152) and “relative ringiness” (1984, 102, 154). Relative asymmetry
relates the depth of a room from the street to the number of rooms in the
house. In other words relative assymetry takes the size of the house into
account when assessing the “depth” of a room. If for example the
position of the main reception room is assessed, a smaller figure indicates
a house whose reception room lies in a relatively shallow setting in
proportion to the size of the house, while a larger figure indicates a
reception room that is set back further in the building and is preceeded
by a more elaborate set of suites.
Relative ringiness assesses the extent to which circulation within a
house is constrained by limiting access to certain rooms. It is calculated
by counting the number of ways of passing through a room – rings – and
comparing this with the maximum possible number of rings for the
number of rooms in a house. The results for Roman houses suggest that
overall circulation literally revolved around the central peristyle, while
circulation between other rooms was rather limited. It is interesting to
note that town houses have a greater relative asymmetry and are more
“ringy” than villas. I presume this is because the courts of villas are often
large yards rather than “rings”, and because the larger space available for
villa building allows rooms to be strung out along the porticoes rather
than communicating directly with one another.
The Hillier and Hanson formula can only be used where the house
plan is complete and the functions of the main rooms are known. It is
easier to use in places such as Pompeii where the preservation of wall
paintings and other furnishings helps in the identification of room functions. The theory is much harder to apply on provincial houses. When
representing movement, or links between rooms, no account is taken for
a potential locked door, or the curtains and partitions that commonly divided Roman reception rooms (see discussion of the letters of Pliny and
Sidonius below, and Thebert 1987, Because of its somewhat THEORIES OF CIRCULA nON IN ROMAN HOUSES 81
mechanistic nature the Hillier and Hanson model provides a broad brush
for comparing very different types of house, but the results then need to
be overlaid with a considerable degree of social interpretation.
Conventional studies of Roman houses concentrate on activities that
took place in the major rooms, especially the cubiculum, and the oecus, or
triclinium. Most of the time they run up against the impossibility of identifying the functions of all the rooms in a Roman house. Many rooms may
have had more than one function. Pliny mentions one room in his Laurentine house that could be used as either a bedroom or a dining room
“vel cubiculum grande vel modica cenatio” (Pliny Letter 2.17.10). This is a
clear statement of the difficulty of identifying room functions on archaeological sites.
Many of our interpretations of room function, as with other ideas of
Roman domestic architecture, originate from Pompeii. I have often had a
nightmare in which I see the citizens of Pompeii on the night of the eruption. They are saying lilt’s such a lovely red sky! Let’s have dinner in the
bedroom tonight! /I No one would suggest this as a serious interpretation,
but it serves to illustrate the danger of accepting too wholeheartedly the
permanence of Roman furnishings. If at all possible we must try to move
away from explanations of Roman houses that depend on dubious identifications of room function towards theories that assess patterns of
circulation in different parts of the house, and which depend on the more
permanent aspects of the architecture and decor.
Carandini’s theory of circulation within Roman houses suffers from too
great a degree of segregation between areas devoted to servants and
areas devoted to the owner’s family. Wallace-Hadrill’s theory does not
provide enough clues on the precise limits of circulation of the social
categories that he identifies. Hillier and Hanson’s theory avoids all the
difficulties of identifying room function by abstracting the house to the
level where it becomes a series of chain links with few social meanings. A
more pragmatic approach to studying the monuments and drawing
value from all these ideas is necessary.
Architecture, decor, and the reception of visitors
The decor and architecture of the house was used by the owner to prevent the visitor from entering inappropriate rooms. If the visitor strayed
the atmosphere created by the room would immediately tell him or her
that a wrong turn had been taken.
The lowliest visitor was the cliens. He was often of significantly lower
social status than his patron. Outside high society in Rome itself he was 82 s. ELus
often impoverished. The architecture and decor used to receive him was
therefore rich. It usually incorporated stucco decorations in a lofty reception room. The visitor of similar status, or higher status, to the owner received better treatment The aim was to impress with superiority in the
case of the former, or devotion in the case of the latter. Such a vis tor was
normally treated to dinner in the main reception room The atmosphere
here spoke of the achievements of the host These could be expressed
through liberality in the banquet (Ellis 1991; Slater 1991), by honorific inscriptions on the walls or the floor (e.g. Apamea; Balty 1969), in the rich
style of the architecture, or by artefacts reflecting the host’s learning, such
as a painting of a rare myth.
This idea was considered by WalIace-HadriII (1988), but it has been
more fully developed by a collection of articles edited by Gazda (1991). In
particular Bettina Bergman (1991, 49-70) studies Statius’s Silvae 2.2, which
describes a visit to a first-century BC Neapolitan villa. The reader is taken
on a walk through the villa during which attention is drawn to various
landscapes that are seen from the windows and a variety of painted landscapes on the walls of the room. As the visitor moves through the building the paintings in the rooms help to establish their atmosphere with
views of peaceful harbours (Bergman 1991, 65) or grand porticoes. The
Roman house was clearly designed so that on walking through it the eye
would be led to an impressive view or painting. These provided clues to
the observer as to what, or who, was appropriate in any particular room.
An important example of how painting, architecture, and reality combined to provoke an impression on the visitor is the use of the view from
the main reception room. Pliny (Letter 2.17.5) especially prided himself on
his dining room from which he could look out on one of three seas by rearranging the partitions. In every Roman house considerable attention
was devoted to the view from the main reception room Conventionally
the reception room had a triple entrance with the largest central door
framing the view. The use of water was important. A villa might overlook the sea or a river. If the reception room faced a peristyle, there
would usually be a fountain in the court (see both South Shields and
Neapolis discussed below). Subtle adjustments would sometimes be
made to the architecture to ensure that these conditions were met
A literary tour around Roman villas
The view from the triclinium is one element that is stressed by ancient
authors. Literary descriptions of villas also provide a good idea of the ele-THEORIES OF CIRCULA nON IN ROMAN HOUSES 83
ments of the house that an owner singled out to impress visitors, and to
some extent the patterns of circulation within houses.
Early theories of Roman housing made much of Pliny’s Letters 2.17 and
5.6 in which he describes his houses in great detail. They have been used
unsuccessfully to reconstruct the physical layout of the buildings. But
looking at them from the point of view of circulation allows us to see
them and other texts in a new light. The following discussion will also include a poem by Statius on the Sorrentan villa of Pollius Felix (Silvae 2.2).
In the early fifth century AD Sidonius Apollinaris wrote a poem about
the villa of Pontius Leontius (Carmen 22) and a letter describing his own
house (Letter 2.2). To some extent Sidonius may be copying the earlier
authors, but there are sufficient details about the houses he describes to
show that they represent contemporary buildings rather than literary allusions. Sidonius’s work will also be considered.
Two points indicate that these literary works are concerned with circulation. The purpose of all four works is to emphasise the marvellous
nature of the houses concerned, so the authors pick out the rooms that
would most impress visitors. Secondly the four works are organised as a
visitor’s itinerary. An imaginary visitor is lead up the driveway through
the front door, into the house, culminating in the most intimate rooms
that a guest is allowed to see.
Many of the points picked out by the three ancient authors are ones
that have already been mentioned above. Throughout the works the nature of the decor is mentioned, especially rich marbles. As the guests
arrive they are greeted by magnificent scenery and by the roof of the
villa. In particular the attention of the guest is first drawn to the private
bath house. The baths are singled out because of their distinctive roofs
outside the villa, and again because of their magnificent architecture once
the guests have entered them. Pliny leaves the description of the baths
until late in his narrative. But even here the emphasis is enough to
demonstrate that the baths were public rooms.
After leaving the baths the visitors pass a number of utilitarian rooms,
the “atrium etiam ex more veterum” of Pliny’s Tuscan house (Letter 5.6.16),
storeroom “cella penaria 1/ (Sidonius Letter 2.2.9), and horrea (Sidonius
Carmen 22.169). In both Pliny’s Laurentine villa (Letter 2.17.4) and the
villa of Pontius Leontius (Sidonius Carmen 22.150-155) these rooms are
reached across a semi-circular peristyle. At this level of penetration into
the building the first reception rooms are encountered. In Pliny’s villas
(Letters 2.17.5, 5.6.19) it is the main triclinium, while in Sidonius’s villa it
is the triclinium matronalis (Letter 2.2.9). These rooms contain movable 84 S. ELUS
partitions “valvis”, which can be adjusted to give views across the
surrounding area The views of the countryside from outside the building
are now matched by vistas opening up through windows.
From this reception area the visitors proceed to the more intimate
rooms. It is in this section that Pliny mentions the bedrooms, library, and
servants’/guests’ quarters of his Laurentine villa (Letter 2.17.7-10). In his
Tuscan villa there is the tric1inium where he receives his personal friends
“cotidiana amicorumque cenatio” (Letter 5.6.21) presumably for his private
recitations. Sidonius’s poem is more oblique at this point, talking about
the winter quarters of the owner “dominis hibema domus”, warm hearths,
and rich decor (Carmen 22.187-193).
Finally the culmination of the visit is always the most splendid reception room in the house. In Pliny’s Laurentine villa it is his heliocaminus, a
sun room with folding partitions disclosing views over the sea or the
woods (Letter 2.17.20-21). Adjacent are two bedrooms from which Pliny
explicitly states that the noise of his household cannot be heard (Letter
2.17.22-25). In Pliny’s Tuscan villa there is an open dining area and
cubiculum. They are built in marble “marmore splendet” (Letter 5.6.38), with
a stibadium couch and again folding partitions with views over the
garden. In Sidonius’s villa there is his cenatiunculam with dining couch
and shining (Le. marble) dinner table “stibadium et nitens abacus” (Letters
2.2.11). From here you look across the adjacent lake with your drink in
your hand At the top “alta volubilibus” of the villa of Pontius Leontius
it is a “cenatio” whose folding partitions disclose mountainside views
(Sidonius Carmen 22. 207-216). Similarly at the top of the villa of Pollius
Felix is the most richly decorated room in the house with a view overlooking Naples (Statius Silvae 2.2.83-97).
Though I have not discussed every room in each villa the general pattern is clear. The owners emphasise the setting of the house in its
landscape and the views from its rooms. At each house there is an
entrance with more functional reception rooms – the “old-fashioned”
atrium of Pliny and the women’s tric1inium of Sidonius. This area also
includes the first porticoes. The visitor then passes to more intimate
rooms. The culmination of the visit is a special intimate reception room
with views over the whole area This room has an atmosphere of
intimacy and tranqUility. The views over the domain, which the visitors
and owners share, are the hosts’ way of saying to the guests – all this is
It is noteworthy that Pliny stresses that he built the marble garden suite
of the Laurentine villa (Letter 2.17.20). Sidonius also mentions a founda-THEORIES OF CIRCULATION IN ROMAN HOUSES 85
tion stone at the entrance of the villa of Pontius Leontius which names
the owner (Cannm 22.142-144). Such pride often lead the owners of villas
to place inscriptions at the entrance to their main reception rooms (Balty
1%9; Ellis 1991).
lHEORY INfO PRACTICE: EXAMPLES OF ROMAN HOUSE CIRCULATION
We have now examined most of the theoretical aproaches to circulation
in Roman housing, and some of the key literary sources. I think that the
above critique has demonstrated that the most appropriate theory to describe circulation patterns in Roman houses will be one that combines the
sensitivity to social distinction of Wallace-Hadrill with the more structured approach of Hillier and Hanson. I would suggest that this may be
achieved by relating the circulation patterns within the house to the decor
of the rooms.
The Roman house, provincial or Italian, was not generally the kind of
house that relied on long dark corridors for circulation between rooms.
There are only two exceptions. One was the cenaculum, the metropolitan
apartment known from Ostia and Rome, where constraints on space led
to high rise blocks that sometimes had no central court (Frier 1980). Village housing in the eastern provinces from Syria to Egypt, where houses
were often two to three storeys high, had no central court and were
served (as in the cenaculum) by a dark stairwell (Ellis 1992).
Most Roman house designs made strenuous efforts to overcome restrictions of space and provide a central court In the case of rural
housing where there is no central courtyard the rooms often front a single
portico. Circulation directly between one room and another was rare.
Even though the Roman house made extensive use of porticoes and
courtyards it would have been possible for the rooms to have secondary
doors linking adjacent rooms together. However, it is common to find a
long line of rooms all opening off a portico, but with no doors between
each other. These two observations have important consequences for
patterns of circulation within the house. Despite the fact that the Roman
house was turned in on itself, facing its central court rather than the
street, it was an open house. As has been mentioned the Romans liked a
view, either through a portico over the countryside or across a court In
either case people in the portico, or outside it, could be seen, be they
family, servants, or visitors. Each of these groups could not be completely
isolated from the other. Servants and visitors would be conspicuous as
they moved around the house. 86 s. ELus
The lack of direct connection between rooms also encouraged openness. People could not pass from one room to another without passing
through the area of general circulation, usually the portico. There was
thus a strong chance that in crossing the portico between rooms one
would meet another member of the household, or a visitor. We have already mentioned the passage of Pliny’s letters in which he says rooms
near his favourite lIbrary were suitable for servants or guests. This seems
to indicate that the adjacent porticoes could be peopled by family, guests,
Variations upon this nonnal pattern of circulation within Roman
houses indicate that measures were sometimes taken to moderate this
intermingling of people within a house.
There are a number of houses in which there are rear entrances or even
corridors behind the triclinium or oecus. It is most likely that this was a
device to allow servants to bring food into the room, or to prepare the
table, without being seen. Four houses at Volubilis have corridors which
enter the tric1inium from the rear (The “Palace of Gordion”, the House of
the Cistern, the House of the Golden Coin, and the House of the Sundial
– Rebuffat 1969, 667-668). The House of the Hunt at Hippo (Rebuffat
1969,674) has a long corridor behind the triclinium. It serves several other
rooms and has apses at each end. This corridor thus seems rather more
than a simple service access. At his Tuscan villa Pliny mentions (Letter
5.6.30) a staircase which was used for a secret entrance during dinners,
presumably by servants: “scalae convivio utilia secretiore ambitu suggerunt”.
At this juncture it may be well to examine some of the physical remains
of Roman housing. The examples will bring out many of the lines of enquiry suggested above. They will demonstrate how the owner used the
decor of the main reception suite to control his social encounters. They
will also show how the articulation of the rooms and their decor indicated to a visitor the pattern of circulation within the house. I wish to take
two town houses as examples, one from third-century North Africa
(House of the Nymphs, Nabeul), and one from fourth-century Asia
Minor (House above the theatre, Ephesus).
The third century was probably the main floruit of Roman Africa, and
the House of the Nymphs (fig. 1) at Neapolis (modern Nabeul in Tunisia)
represents a typical middle-range house of that period (Darmon 1980). At
first sight the peristyle seems quite close to the street, but in fact on
Hillier and Hanson’s model it lies deep within the house five rooms away
from the street From the front porch (level 1) the visitor turns left into an
anteroom (2) and then right into a first reception room (3). The latter THEoRIES OF CIRCULA nON IN ROMAN HOUSES 87
Figure 1: House of the Nymphs, Nabeul, Tunisia.
opens onto an atrium (4) on the left, after which one enters the main peristyle (5). The ambience of the first reception room was developed by its
view on to the atrium and two pilasters at the middle of the room. The
view through the atrium provided glimpses of the peristyle, suggesting
the grandeur of the house. The pilasters added to the architectural
standing of the room and formed a frame to separate one half of the room 88 S. ELLIS
from the other. They may have been used to separate the clients from the
patron during the morning greeting ceremony, the salutatio.
But once the peristyle was entered, the articulation of the surrounding
rooms left little distinction between servants, owners, and guests. Four
cubicula can be identified on the east side of the peristyle. They all have
the characteristic plain mosaic panel that was the setting for the couch.
The presence of the panel presumably indicates that the rooms were in
relatively permanent use for this function. Anyone sleeping in these
rooms would not have much privacy from the rest of the household since
to pass to any other location they would have to go through the peristyle.
It may be presumed that the owner himself slept in a separate apartment
in the north corner of the house. Here two similar cubicula either side of a
vestibule are isolated from the main house by an atrium symmetrical to
that at the main entrance.
An alternative design can be seen at AcholIa in the House of Neptune
(Rebuffat 1974, 454). Here three cubicula, identified on the basis of the
mosaic panels, layoff the south-western corner of the peristyle and were
served by a 4m-Iong corridor. It is most probable that cubicula that were
segregated in this way were used by the immediate family of the owner.
It is possible that the two atria of the House of the Nymphs, which lay
to either side of the main tric1inium (identified by the U-shaped plain
mosaic panel for three couches), had symmetrical functions as well as a
symmetrical location. The atrium near the front entrance had a central
role in the reception of poorer clients. The other atrium, next to the
owners’ bedroom, could then have played a part in the reception of
visitors at their morning levee.
The peristyle itself appears to be unremarkable, but in fact it is the
most subtly designed of all the rooms of the house (Darmon 1980, 74-75).
The overall plan of the house was determined by its urban setting, and
presumably by the remains of earlier structures on the site. The architect
discovered that he could not design the canonical perfect rectangular
peristyle. The court instead had to be trapezoidal. However, by creating a
subtle curve in the southern portico he was able to make the south-eastern and south-western corners of the court into right angles. This gave
the occupants the illusion of a rectangular area. The architect was abetted
by the mosaicist who subtly varied the proportion and lines of the
mosaic, bending them towards the ideal of a rectangular form.
Ideally, the main reception room should be in the middle of one side of
the peristyle facing the centre of the opposite portico. The trapezoidal
peristyle made this impossible. However, the curve of the south portico TliEORIES OF CIRCULA lION IN ROMAN HOUSES 89
and the slight convergence of the southern ends of porticoes to the east
and west created the illusion of the centrality of the tric1inium when seen
from inside the room. Moreover, the doors of the rooms on each side of
the tric1inium almost fallon the lines of the eastern and western porticoes,
again suggesting its centrality. It would only be obvious that the tric1inium was not at the centre of the northern portico if the visitor was
viewing the room from the southern portico. However, the principal entrance to the house was in the north-west, and visitors would not often
have reason to view the room from the south.
The design of the peristyle of the House of the Nymphs illustrates how
important this part of the house was for circulation. The architect and
mosaicist went to extraordinary lengths to ensure that its design matched
the accepted standard. The arrangement they achieved also ensured that
an appropriately symmetric view would be obtained from the triclinium.
The house above the theatre at Ephesus (fig. 2) is a much grander
building than the house at Nabeul. It represents the town house of a
fourth-century nobleman of imperial rank (Ellis 1983, 1988). At this time
centralisation of the empire and increased taxes had devalued the political role of the provincial aristocracy. When a provincial needed to obtain
the help of the administration he tended to go to a local rich patron
whose pOSition as a senator of Rome or Constantinople would give him
considerable influence. This was a private visit to the patron’s home. Because of their increasing political role such patrons required extensive
reception facilities for clients. In material terms this resulted in a number
of houses that have an apsidal audience chamber close to the main entrance (Ellis 1988, 1991). The use of an apsidal room as an audience
chamber is of course well attested in Roman times. The best-preserved
example is the Constantinian basilica in Trier.
The function of the domestic room can be demonstrated from the fact
that it is close to the street, and preceeded by a vestibule. The vestibule of
the Palace of the Dux at Apollonia was lined by benches on which clients
awaiting admittance might rest (Ellis 1983). The function of the decor of
the audience chamber is not so much to impress guests as to overawe the
lowly client A location near the street ensured that the more humble
client did not disturb the privacy of the main house. Whereas in apsidal
tric1inia the entrance is from the opposite end of the room to the apse, in
the audience chamber there is a second door near the apse from which
the dominus could make a grand entrance.
In the house at Ephesus the audience chamber was really on a grand
scale and was the public front to a large town house. A secluded private 90 s. ELUS
o 10 ..
Figure 2: Houses with audience chambers: (1) the villa above the theatre, Ephesus; (2) the
“Governor’s Palace”, Aphrodisias; (3) the “Palace of the Dux”, Apollonia. A = audience
hall, T = triclinium. THEORIES OF CIRCULA nON IN ROMAN HOUSES 91
apartment with its own small peristyle is identifiable in the north-eastern
comer of the building, as at Nabeul. This apartment lay at a minimum of
eleven rooms away from the street, in Hillier and Hanson’s terms this
makes the house one of the “deepest” Roman houses known.
Discussion of the houses at Neapolis and Ephesus has demonstrated
the mechanisms used to control access and create an atmosphere with
architecture and decor. Sometimes the overall architecture such as the
“basilican” fonn could be used, at other times the dimensions of the
room might be altered to emphasise a particular axis of the house. The
use of columns to frame the end of a room and the view from the reception room door has been illustrated, and the use of appropriate floor
mosaics has been discussed. Circulation within the Roman house was of
course also controlled by locked doors and servants. However, as
archaeologists looking at the material evidence, we should direct our
attention to three particular areas: the overall form of a room, the way in
which the decor expresses this form, and the uses of decorative themes
that reflect liberality and education.
DOMESTIC ARCHITECTURE AND DECOR IN ROMAN BRITAIN
The above arguments prove that there was a symbolic link between the
architecture of Roman houses, the circulation patterns within them, the
decor of the rooms, and Roman customs for the reception of guests. This
structure has been shown in many provinces and at many periods.
In the context of the debate about Romanisation in Britain the existence
of the structure I have outlined would prove that the British elite
were able to use complex symbolic structure employed by the
provincial elites elsewhere in the Roman world;
had a detailed knowledge of Roman cultural symbols and
willingly adopted social practices of high Roman society (they could
not have been compelled to adopt such a complex symbolic model,
and it is unlikely that they would have based their housing on such a
complex model if they did not understand it).
Scott (1997) has already laid the ground for this interpretation in her
work She has demonstrated the use of complex architecture and decor in
the reception rooms of fourth-century villas in south-west Britain. I
would only like to develop her argument by examining the case for two
more houses. 92 S.ELLIS
• • • • I
• . –
( ) 0
Figure 3: Praetorium, South Shields.
The first house (fig. 3) lies outside the more “Romanised” southern
area of Britain. It is the recently excavated praeton’um at South Shields
(Frere 1992, 267-269). Here the design was constrained by the surrounding buildings (as indeed at Neapolis). The visitor entered a vestibule with
six columns or posts. This can be seen as a classical atrium or as a
derivative of Romano-British architecture. It may indeed have echoed
both, making visitors from the Mediterranean, and locals, feel at home.
Whatever its architectural derivation it clearly served as a reception area.
Another typical reception room – the “private” baths – was neatly
tucked into the adjoining comer of the house to the right For the Romans
bathing had always been a social institution, for conducting business and
meeting friends. Thebert (1987, 380-381) has suggested that the development of private baths in late Roman urban housing may reflect the desire
of the increasingly autocratic aristocrat to avoid showing himself naked
to his more lowly clients, The preferred setting for the lesser client was
the audience chamber, such as that at the Ephesus house. In this context
the South Shields baths may have been used with fellow officers or VIP
civilians. Two rooms to the left of the vestibule could have been a porter’s THEORIES OF ClRCULA nON IN ROMAN HOUSES 93
lodge. From the vestibule one entered a portico leading to the main
triclinium at the far end of the house, but there was no portico along the
south-east range of the house, and thus no peristyle. Constraints of space
had prevented this. Nevertheless, in the open centre of the court onto
which the portico fronted there were a large rectangular tank and a semicircular fountain. Both these elements are typical ornaments of a classical
At South Shields the absence of a true peristyle was countered by
decorative elements suggestive of one. As at Neapolis, the view from the
reception room was important in creating this impression. The semicircular fountain was placed so as to be seen through the columns in
front of the main triclinium. The main portico in the house was that
entered on leaving the main vestibule. It may be no coincidence that
figured wall plaster was found in the portico, where it would have
formed a rich decor to impress guests. I would also suggest that the blank
wall on the south-east side of the court would have been painted with a
trompe l’oeil colonnade. Between the columns would have been painted
garden or countryside scenes. This design would fit the ambience of the
court, it would conform to what we know of classical wall painting, and
it would represent the traditional “rural” view from the reception room.
The praetorium of an officer on the north-west frontier may not seem
the appropriate place for such Mediterranean allusions, but by any interpretation it is an unusual house. Whether it had the desired effect on visitors or not I would consider it is another example of the way in which
the architecture and decor of Roman houses was used to create ambience
and to indicate where a visitor was permitted and where he was out of
place. To pass from the vestibule to the portico at South Shields was to
pass away from the mores of the frontier to the mores of the
Mediterranean. A visitor in the vestibule might expect to be treated in a
strict military manner. A visitor who entered the “peristyle” had been
warned to act in a more elitist cultural vein.
The second house that I wish to examine is the fourth-century Roman
villa at Keynsham (fig. 4) (Bulleid and Horne 1926). The house was not
well preserved when it was excavated. An extremely large central
courtyard was discovered, together with some fourteen rooms in the
north-west of the site. Other than this there was only a small complex of
three rooms preserved in the south-west comer of the site.
The main room in this south-west complex was a hexagonal room (W)
with three rectangular niches, and two semi-circular niches near to the
entrance from the adjacent west portico of the courtyard. At the opposite 94 S.ELLIS
Figure 4: House at Keynsham.
end of the west portico was a baths complex that also centred around a
hexagonal room OJ. The architecture of the baths was thus quite elaborate
by British standards and one can imagine that it was one of the talking
points of an aristocratic owner.
The southern hexagonal room is a formal dining room. It was common
to place a very elaborate apsed dining room off one comer of the peristyle in late Roman houses. Examples (fig. 5) can be found in the Mediterranean area at Ravenna in Italy, Mediana in Serbia, and Djernila in
Algeria (Ellis 1991). Another example from Roman Britain is afforded by
Littlecote, consisting of a triapsidal room set off the corner of the peristyle, in the normal Mediterranean manner. Semi-circular dining couches
(the stibadium of Sidonius cited above) were placed in each apse. Literary
accounts of such halls, and dinners served in them, exist from Rome and THEORIES OF CIRCULA nON IN ROMAN HOUSES 95
Figure 5: Houses with multi-apsed grand dining halls: (1) House oj Bacchus, Djemila; (2)
the “Palace of Theodoric”, Ravenna; (3) villa at Mediana, near G = grand dining
hal/, T = triclinium. 96 S.ELLIS
Constantinople. The Keynsham example is closest to that from Mediana
(Drca et al. 1979), which has a similar hexagonal central room and three
apses. The two small niches at the entrance to the Keynsham room are
also found at Djemila and Constantinople (Krautheimer 1966). They
probably held statues.
In the context of the present discussion the important point is the
architectual parallelism of two hexagonal suites of major public rooms at
each end of the main portico of the villa. The whole west wing of the villa
was available for clients and friends sometimes passing no doubt direct
from the baths to the dining suite. In between the two suites one would
not be surprised to find a large reception hall, like that at Woodchester,
which could be used for larger receptions.
The sophistication of the dining suite is matched by that of its mosaic.
It was damaged but contained panels showing Europa and the bull and
a scene considered to be from the story of Marsyas (Stupperich 1980,
293-296). Whatever interpretation is adopted it is clear that the mosaic
depicts scenes that none but the most educated would appreciate. The
architecture and the decor both seem to indicate that the owner of the
Keynsham villa had a large number of very important friends. He was
thus happy to devote a large proportion of his house to major public
The houses discussed by Scott (1997) and the discussion of Keynsham
indicate a considerable degree of interest in Mediterranean fashion
shown by the aristocracy of Roman Britain. The house at South Shields
indicates that the army was one channel for purveying the latest Mediterranean domestic style and shows that a high-ranking officer was
detennined to keep up appearances even on the northern limit of the
BaIty, J. 1969. La grande mosafque de chasse du triclinos, Apamee de Syrie.
Brussels: Musee Royale (Miscellanea 2).
Bergmann, B. 1991. Painted perspectives of a villa visit landscape as status
and metaphor. In E. Gazda (ed.) 1991, 49-70.
Berti, F. 1976. Mosaici antichi in Italia; Aemilia-Ravenna 1. Rome: Libreria
Branigan, K. 1973. Town and Country: Verulamium and the Roman Chiltems.
Bourne End: Spur.
Branigan, K. 1975. The Roman Villa in South-West England. Bradford-onAvon: Moonraker. THEORIES OF ClRCULA nON IN ROMAN HOUSES
Bulleid, A. and D. Home 1926. The Roman house at Keynsham, Somerset
Archaeologia 25, 109-138.
Carandini, A. et al. 1985. Sette Finestre: una Villa Schiavistica nell’Etruria
Romana, 3 voIs. Modena.
Darmon, I.-P. 1980. Nymfarum Domus. Leiden: Brill.
Drca, S. et al. 1979. Mediana. Ni.§: National Museum.
Duval, N. 1984. Les maisons d’Apamee et I’architecture “palatiale” de
l’antiquite tardive. In Colloque Apamee de Syn’e, 447-470. Brussels:
Dwyer, E. 1991. The Pompeian atrium house in theory and practice. In
Gazda (ed.) (1991),25-48.
Ellis, S. 1985. The palace of the dux at Apollonia; and related houses. In G.
Barker, J. Reynolds, and J. Lloyd (eds), Cyrenaica in Antiquity, 15-25.
Oxford: British Archaeological Reports (International Series 236).
Ellis, S. 1988. The end of the Roman house. American Journal of Archaeology
Ellis, S. 1991. Power, architecture, and decor: how the late Roman aristocrat
appeared to his guests. In Gazda (ed.) (1991), 117-134.
Ellis, S. 1992. Graeco-Roman Egypt. Aylesbury: Shire.
Frier, B. 1980. Landlords and Tenants in Imperial Rome. Princeton: Princeton
Frere, S. (ed.) 1992. Roman Britain in 1991: sites explored. Britannia 23,
Gazda, E. (ed.) 1991. Roman Art in the Private Sphere. Ann Arbor: University
Hillier, B. and J. Hanson 1984. The Social Logic of Space. Cambridge:
Cambridge University Press.
Krautheimer, R. 1966. Die Decanneacubita in Konstantinopel. R6mische
Quartalschrift suppl. 3, 195-199.
Lavin,1. 1962. The house of the lord. Art Bulletin 44, 1-27.
Rebuffat, R. 1969. Maisons a peristyles d’Afrique du Nord, repertoire de
plans publies. Melanges Ecole de Rome 81, 659-724.
Rebuffat, R. 1970. Thamusida 2. Rome: Ecole de Rome.
Rebuffat, R. 1974. Maisons a peristyles d’Afrique du Nord, repertoire de
plans publies II. Melanges Ecole Franfaise de Rome 86, 445-499.
Saller, R. 1982. Personal Patronage in the Roman Empire. Cambridge:
Cambridge University Press.
Salza Prima Ricotti, E. 1987. The importance of water in Roman garden triclinia. In E. MacDougall (ed.), Ancient Roman Villa Gardens, 135-184.
Washington DC: Dumbarton Oaks.
97 98 S.Ews
Scott, S. 1997. The power of images in the late Roman house. In R.
Lawrence and A. Wallace-Hadrill (eds), Domestic Space in the Roman
World: Pompeii and Beyond, 53-68. Portsmouth: Oournal of Roman
Archaeology supp!. 22).
Slater, W. (ed.) 1991. Dining in a Classical Context. Ann Arbor: University of
Stevens, C. 1933. Sidonius Apollinaris and His Age. Oxford: Oxford
Stupperich, R. 1980. Some fourth century British mosaics. Britannia 11,
Thebert, Y. 1987. Private life and domestic architecture in Roman Africa. In
P. Veyne (ed.), A History of Private Life, 1: from Pagan Rome to Byzantium,
357-364. Cambridge: Belknap.
Toynbee, ]. 1981. Apollo, beasts, and seasons; some thoughts on the
Littlecote mosaic. Britannia 12, 1–6.
Treggiari, S. 1991. Roman Marriage. Oxford: Clarendon Press.
Veyne, P. 1987. The Roman empire. In P. Veyne (ed.), A History of Private
Life, 1: from Pagan Rome to Byzantium, 5-233. Cambridge, Belknap.
Walker, S. 1983. Women and housing in classical Greece: the archaeological
evidence. In A. Cameron and A. Kuhrt reds), Images of Women in
Antiquity, 81-89. London: Routledge.
Wallace-Hadrill, A. 1988. The social structure of the Roman house. Papers of
the British School of Rome 56, 43-97.