The Dovecotes of Karanis
By Elinor M. Husselman
University of Michigan
Published in American Philological Association 84 (1953): 81-91.
(Note: To properly view the Greek text, you must have either the Athenian,
Attika, or Kadmos font installed.)
The study of the granaries of Karanis presented last
year at the annual meeting of the American Philological Association1 was the first of a projected series in which it is planned to correlate
the archaeological evidence from the excavations of the University of
Michigan at the site with the written evidence from the papyri of Greco-Roman
Egypt. Since nowhere else has such exhaustive work been carried on at
the site of an Egyptian village of this period, it is to be hoped that
the archaeological findings may illuminate some of the doubtful papyrus
passages, and that papyri may sometimes provide answers to questions raised
by the excavations.
The importance of grain in the economy of Egypt made it natural that
the granaries for its storage should be a prominent feature of the Egyptian
village. Although not of comparable importance economically and of relative
insignificance from the administrative point of view, the raising of pigeons,
both for food and for the production of manure, was widespread in ancient
Egypt, as it is in Egypt today, and the dovecotes at Karanis were both
large and numerous.
The remains of the pigeon houses probably do not indicate the actual
number that were in use. Since the common practice was to build the dovecotes
in the upper story of a house or tower, they would naturally be the first
part of the house to fall to ruin. Yet there still remain five extensive
dovecotes and part of a sixth, from which we can learn the details of
their structure. Two of these, B9 from the second level of excavation
and C37 from the third, were briefly described in the published report
of the excavations2.
Another, which was constructed in the forecourt of the large granary C65,
was mentioned in the same report3,
but had not at that time been fully excavated. One of the largest and
one of the most interesting of the pigeon houses was D301, which was probably
built in the first century B.C., since it is in the fourth level of occupation.
It continued in use, however, into the second level, and a few scanty
remains of another were found in the top level, which in general appears
to have been occupied in the fourth and fifth centuries.
The two largest dovecotes are C37 (to which probably the rooms designated
C35 also belong) and D301. From their size and the number of nesting places
provided for the pigeons it is certain that they were establishments either
operated commercially or carried on as an adjunct to a large estate or
C35 and C37 comprise a complex of rooms, corridors and courtyards surrounding
three massive towers in which the pigeons were housed. The overall dimensions
of the building were 14.5 by 11.75 meters, and the towers, although irregular
in construction, were roughly 4.5 meters square at the base. The walls
were built of sun-dried brick, with a slight batter, and were approximately
1-1.5 meters thick at the base and from .75-1 meter thick at the top.
The basement of each tower was used for storage, and in two of them this
space was divided into bins for grain such as we find in the large granaries.
These semi-underground storage rooms were covered with vaulted ceilings
about 2.5 meters high. Since the tops of the towers have fallen to ruin,
it is not possible to determine their original height, or to state from
positive evidence whether or not they were roofed. It is safe, however,
to assume that they were left unroofed to allow the entrance and egress
of the pigeons, since in all the structures the nests face the inside
of the tower.
Within the tower the walls are lined with pots set horizontally in the
masonry with the mouth serving as the entrance and the body of the pot
as the nesting place. The pots were of special construction, usually 45
cm. in height and 25 cm. across the largest part of the body, with a round
mouth of about 15 cm. in diameter. They were made on the wheel, and the
small opening at the bottom of the pot, where it rested on the peg of
the wheel, was not closed, since this was unnecessary for the purpose
for which it was to be used. The edge of the hole was sometimes reinforced
by an added ridge of clay. Many of these pots were found both in situ
and elsewhere throughout the excavations4.
Beneath the rows of pots were either two or three rows of small rectangular
niches, 20 cm. deep and having an opening 15 cm. square. These niches
were a regular feature of all the large dovecotes and their purpose is
uncertain. The most reasonable assumption seems to be that they were used
for nests for the squabs and young pigeons before they reached breeding
age. The earliest age at which the squabs would be ready for the market
would be from three to four weeks5.
If, as Aelian states (Var. Hist. 1.15), the Egyptian pigeons
produced twelve broods a year, the squabs would be barely two weeks old
when the new brood hatched. Moreover the parent birds are said to expel
the young birds from the nest when the pigeon "milk" begins to form for
the next brood, which occurs several days before the eggs are hatched.
It is clear, therefore, that the young pigeons would have to be fed for
at least a week before they would be large enough to be sold for food.
Even if the pigeons did not breed as prolifically as Aelian believes,
the squabs were probably kept till they were older and were specially
fed to fatten them for market. Cato (De agriculture 90) gives
formulas to be used in the "cramming" of pigeons, and Varro (Res
rusticae 3.7.9-10) and Clumella (De re rustica 8.8)
both give directions for the fattening of squabs to increase their purchase
price. The papyri offer no evidence on the fattening of pigeons, but they
do show that other fowl were so treated6.
The rectangular niches would be a convenient place to keep the birds that
were to be specially fed, since they were low and open and hence more
accessible than the deep, narrow-necked nesting pots.
The suggestion that the niches were used as a place to strew food for
the pigeons is scarcely tenable, since they are too small and too deep
for the purpose. Varro says that pigeons' food should be placed in troughs
running around the walls; and, according to Columella, it may be simple
scattered along the walls, the only place where it would not be contaminated
with dung. Nothing remains in the Karanis dovecotes to suggest how the
pigeons were fed and watered, and, if special vessels were provided for
the purpose, they have completely disappeared.
At a rough estimate there were at least 1000 nesting places in the part
of C37 that remains standing, and it is not unlikely that the total number
was 1500, or even more, depending on the original height of the towers.
It is obvious that an establishment of this size was a commercial enterprise,
since it was much too extensive for the needs of a single family or even
group of families.
The remainder of the house contained courtyards and rooms used either
as offices or as living quarters for the owners. Additional storage space
for grain was provided in underground vaults at the southern end of the
building. The ground on which the house was erected sloped down on the
north, and a door on this level opened on the street that ran along the
northern side. There are some indications that the building, aside from
the towers, originally had only one story. A stairway from the underground
floor appears to have been a later addition, and there is some evidence
that suggests that the second floor had also either been built on at a
later date or undergone extensive reconstruction. Another pigeon loft
was built against the northern wall of the northwest dovecote with additional
nesting places on the north and west, and this too appears to have been
added to the earlier structure.
The two dovecotes, C91 and C65, are similar in their plan and arrangement
to C37, although neither is so extensive. C91 was probably built in the
early second century, but continued to be used for the raising of pigeons
after alterations made in the early third century. The house itself, 12.5
meters long and 10.8 meters wide, with an ell extension on the northeast
corner measuring 3.5 by 7 meters consisted of a series of one-story rooms
and courtyards surrounding a central three-story tower that housed the
dovecote. The tower was the same size as those in C37, 4.5 meters square
at its base. The underground floor was used as a storage vault for grain,
with the usual division into bins. The arrangement of nesting pots and
square niches is the same as in C37.
The dovecote which was built in the forecourt of the large granary C65
was similar in construction to C37 and C91, but, instead of being a square
tower, surrounded by other rooms of the building, it was an independent
rectangular structure, 7.5 by 2.5 meters in size. The lower floor was
made up of two vaulted storage rooms, each divided into bins. The second
floor, in which the pigeon nests were built, was divided into two parts
by a wall, .5 meter thick, in which there was a doorway, 1.75 meters high
and a little over .5 meter wide. The lintel was made of branches covered
with mud brick and plaster, and a large branch was also set in across
the doorway, .5 meter below the lintel. Inset in the walls of the two
rooms, and in the doorway as well, were the rows of nesting pots. Six
rows of pots remain, either in whole or in part, and the total number
could not have been less than 300. It is not unlikely that the tower originally
extended considerably higher and that the number of nesting places was
correspondingly greater. The additional strength provided by the cross
wall would have served to brace the long and otherwise unsupported side
The only entrance to this dovecote was a door at the southern end, nearly
three meters above ground level. No stairs led to the door, so it is apparent
that access could only have been by means of a ladder. Where the dovecote
tower formed an integral part of the building, access might have been
from a second-story room.
Nothing further need be added to the published description of the dovecote
in house B9, which followed the pattern of C37, C65, and C91. D301, however,
is a more elaborate structure that deserves special discussion. As stated
above it dates from the fourth level of occupation, probably first century
B.C. It was an independent building, with the nests built down to ground
level, and there was no basement storage room beneath the pigeon house
The building was in the form of a square, 9.7 meters on each side. The
mud brick walls were no less than 1.5 meters thick. Entrance was through
a door on the south into a corridor 7.5 meters long and almost a meter
wide. Opening from this central corridor and at right angles to it were
three transverse corridors extending almost three meters to each side.
On the north side of the north transverse corridor were three niches,
.5 meter deep, extending to floor level, and there was a corresponding
arrangement on the south wall of the southern corridor, although here
in the middle niche was in part taken up by the entrance passage.
The upper part of this dovecote was destroyed when house C401 was built
above it, and it is impossible to estimate its original size with accuracy.
Judging by the arrangement of the other pigeon houses, we may be sure
that at least 1250 nesting places were provided, and close to 200 of the
small square niches.
A few traces of a dovecote appeared in the A level, that of the latest
occupation in the fourth or fifth century. The fragmentary remains of
three walls disclose the same square structure, with nesting pots embedded
in the walls and opening toward the inside.
Several ancient Roman writers on agriculture have dealt with the subject
of pigeon breeding, and have given us a reasonably complete picture of
the methods employed in Italy during the early centuries of our era. Varro,
who wrote his Res rusticae about 30 B.C., and Columella,
whose De re rustica dates from the first century of the Christian
era, have been cited above7.
Palladius, who composed a similar work in the fourth century, adds little
to the earlier accounts. The peristeron as described by Varro
is a large building with a vaulted roof. It has one door and either "Punic"
windows, which we may assume from Palladius (De re rustica 1.24) were
small and narrow, or wider windows covered with lattice work to exclude
birds of prey. The walls are smoothly plastered, so that lizards and mice
cannot get in. Round nests are provided for each pair of pigeons, and
they are set side by side in rows running up to the vaulted roof. Each
nest has an opening just large enough to allow the entrance and exit of
a single pigeon, but the interior measures three palms (ca. 22 cm.) in
all directions. Before each row a board, two palms wide, is fixed to serve
as an entrance to the nests and to provide a place for the pigeons to
walk. To this description of a properly constructed dovecote Columella
adds little, although he does mention the use of pottery nests, fictilia
columbaria. He also specifies that the walls of the dovecote and the nests
themselves should be whitewashed, because as he says, eo colore praecipue
delectatur hoc genus avium. Both Columella and Palladius state that the
pigeon house should be in tower or elevated place.
The dovecotes at Karanis follow closely the precepts of Varro and his
successors, and such differences as may be noted are due to the difference
in climatic conditions between Italy and Egypt, where protection from
the weather was a negligible factor, the top of the dovecote could be
left unroofed. As a result the windows, which were designed to give the
birds entrance and also to provide sunlight and ventilation, could be
and were dispensed with. The fictilia columbaria of Columella are the
pottery jars set in the dovecote walls for nests, and they meet the specifications
of Varro in that they have narrow mouths but a body approximately 25 cm.
in diameter. The walls and the rectangular niches are plastered with the
usual yellowish brown mud plaster, so common in the buildings of Karanis,
but there is not trace of whitewash. Perhaps the Egyptian pigeons, unlike
those in Italy, were not partial to white. The boards to be used as walk-ways
for the birds are lacking. That such boards would have disappeared in
the course of centuries might perhaps be expected, since wood was scarce
in Egypt and not a great deal remains that is not a structural part of
the buildings. Nevertheless it should be noted that there is not indication
that such walk-ways were provided, and no remaining evidence that anything
was attached to the walls below the rows of nestng pots.
The papyrus evidence regarding the raising of pigeons in Egypt was briefly
discussed by Schnebel8,
and was subsequently more fully examined by Maria Cobianichi9.
The evidence regarding the pigeon houses was also treated by Luckhard10.
It is not necessary here to go into this evidence in detail, but only
to point out how it fits in with the actual construction of the dovecotes
as we find them at Karanis.
The dovecotes were often situated in an outlying farmstead, where they
either occupied part of some outbuilding, or stood alone, frequently adjacent
to a vineyard or garden. The latter was a convenient location, since pigeon
dung was largely used to fertilize land used for these purposes11.
The construction of the dovecotes was no doubt the same whether they were
built in the villages or outside, but a self-contained building like D301
would probably be the type used when the pigeon house was built at a distance
from the other farm building. But it is not surprising that we should
find many pigeon houses in the villages. Except where there were large
farm holdings with many buildings and large households, the owners and
farmers of the land probably lived in the villages, going out by day to
cultivate their fields in the adjacent country-side and returning to their
homes at night. Their property would receive greater protection within
the confines of the village, and it may be for this reason that we find
granaries, dovecotes, and animal pens in such numbers in Karanis.
There is some indication in the papyri as to the size of the dovecotes.
In PTebt. 1.62, which is a list of owners of temple and cleruchic land
of the second century B.C., an area of one aroura is recorded in line
49, in which 19/32 were taken up by a shrine to Isis, 12/32 by a garden,
and 1/32 by a pigeon house. The pigeon house therefore occupied an area
of slightly over 86 square meters, and would be comparable in size to
D301, which was 9.7 meters square and covered a little more than 94 square
meters. Another of the many documents dealing with the land survey in
Ptolemaic Egypt published in the first volume of Tebtunis papyri, PTebt.
1.86.15, lists a dovecote occupying the same amount of space. Still another
entry, in a register of cleruchs of ca. 148 B.C., PTebt. 1.79.71, designates
a dovecote of twice that size, 1/16 of an aroura. The entire building
C35 and C37, with its three towers and adjacent rooms and courtyards,
occupied over 170 square meters, or almost 1/16 of an aroura (172.25 square
Only one document mentions the nesting pots in the dovecotes. It is PTebt.
1.84, part of the land survey of Kerkeosiris made in 118 B.C. In lines
8-10 property is listed which includes dovecotes with 1000 pots (AGGEIA).
The number of dovecotes is not given, but we have noted that the individual
towers C37 and C91 probably contained from 300 to 400 nests, and we might
deduce therefore that this property in Kerkeosiris contained three such
The rentals stipulated in leases of pigeon houses are frequently to be
paid in part in young pigeons12,
and, if we had any knowledge of the percentage of the total flock that
this payment represented, we would have some basis for computing the size
of the dovecotes that housed them. In no lease, however, do we have any
information given with regard to the size of the dovecote of the number
of pigeons that were included in the transaction.
The construction of the pigeon house required the services of a professional
builder. In PGrenf. 1.21 (Mitteis 2.302), a will of the year 126 B.C.,
a half-finished dovecote is included in the property of the testator,
and the heirs, a son and five daughters, are required to share equally
in the expense of a builder who is to be hired to complete it. A papyrus
of the Byzantine period, BGU 3.962, is an order to an oil dealer to pay
six xestai of oil to the carpenter and sawyers engaged on the
building of a dovecote; the employment of the woodworkers would seem to
indicate a type of construction different from at in Karanis, where the
use of wood was minimal.
Another document throws an interesting sidelight on the dovecotes and
fits in neatly with the archaeological evidence. It was noted above13
that there are no stairs leading to me second-story dovecote associated
with the granary C65, and that a ladder would have been necessary for
anyone to enter it to tend the pigeons or to clean out the manure. POxy.
8.1127 (A.D. 183) is the lease of an upper room in which there is a dovecote
described in the following terms: TON UPERVON
TOPON TH% UPARXOU%H% AUTV EN MOUXINUR OIKIA% KAI ON EXEI EKEI PERI%TEREVNA
%UN TH TOUTOU KLEIMAKI JULINH. Such a ladder would not only be
used to enter the dovecote, but would also serve to reach the higher nests,
if the young pigeons were to be removed to the lower nests for special
feeding and fattening after a subsequent brood was hatched.
The same document also stipulates that at the termination of the lease
the lessee is to return the pigeon house in the condition in which he
received it, with the two doors and one key. Although generally the doors
of the dovecotes in Karanis have disappeared, there is little doubt that
they once had them, and that the doors could be locked. A small opening
at the base of the dovecote tower in the courtyard of house B914
was closed by a wooden door, which remained in its original position.
The size of the opening and its location make it probable that it was
used for the removal of the manure from the bottom of the dovecote.
A third century document from Oxyrhynchus, PFlor. 1.10, is a lease of
two dovecotes and a KELLA. The KELLA
is probably a storeroom, such as we find in connection with most of the
Karanis dovecotes, and it would most naturally be used to store feed for
In view of the number and size of the Karanis dovecotes, it is perhaps
surprising that few papyri found during the excavations mention either
pigeons or pigeon houses, and that of those found none can be associated
with the actual structures. P.Mich.Inv. 2933, an unpublished property
declaration dated 53 A.D., reports the ownership of an OIKIA
KAI AULH KAI ELAIOURGI[ON] KAI PERI%TEREVN. Another unpublished
document, P.Mich.Inv. 2915, dated in a year of Domitian, is a fragmentary
contract for the division among a brother and two sisters of several pieces
of inherited property, including a YH%AURON KAI
PERI%TEREVNA. One of the sisters is a Minucia Thermoutharion, the
mother of the Valeria Diodora who, in PMich. 6.428 (154 A.D.), sells to
Gaius Julius Niger a house which stands to the north of a dovecote which
she also owns. These three documents come from an area on the northwest
of the excavations where no dovecotes were found.
The Karanis tax rolls15
supply some additional information regarding the raising of pigeons in
that area in the records of the payments of the dovecote tax, TRITH
PERI%TEREVNO%. Four persons pay this tax in the register of the
twelfth year of Marcus Aurelius Antoninus, one of whom also make a payment
in the thirteenth year. Eight other persons are listed as paying the tax
in the thirteenth year, and one payment is recorded in the fourteenth
Aside from the pigeon houses excavated at Karanis we have one other bit
of archaeological evidence on the construction of the dovecotes in Egypt.
This is the frequently cited Palestrina mosaic16,
which depicts scenes along the Nile. The dovecote shown in the mosaic
differs considerably from those at Karanis. It is a circular tower, surmounted
by a conical dome in which the nests are placed in rows opening to the
outside. Around the top of the tower beneath the dome branches have been
inserted to provide a place for the pigeons to light before entering the
Although there are variations in details of construction, the dovecote
in Egypt today remains essentially what it was 2000 years ago17.
It may either occupy a separate building, more or less towerlike in form,
or be built in an upper story of a house or outbuilding, frequently a
granary. The nests are still made of clay pots set in masonry. Branches
are generally embedded in the outer walls to provide roosting places for
the birds, a feature that appears to be lacking in the dovecotes of Karanis,
but which is present in the representation on the Palestrina mosaic. The
nesting pots open either to the outside or to the inside of the structure,
and again, although the nests in all the pigeon houses in Karanis open
to the inside, the mosaic shows that both types were used in ancient Egypt.
These modern pigeon houses, when compared with those at Karanis, give
clear evidence of the continuity of the basic types of Egyptian living
from remotest times to the present.
to the publications page