First Phase: The Tower

As it was originally constructed, the Dema tower stood by itself on the hilltop, surrounded by the enclosure wall. There is no evidence that any other structure adjoined the tower or stood elsewhere within the enclosure. The rubble enclosure wall was probably not very high and cannot in any event have been designed as a defensive perimeter, for it is too insubstantial and encloses only a small area with no significant natural strength to recommend it as a defensive position. The wall probably served simply to define the precinct of the tower, the area within which men detailed to the tower were bivouacked. The enclosure wall probably never stood to a height of much more than 0.60 meters in stone. It probably had a superstructure, more likely a brushwood charax than a mud-brick wall, which would suffice to keep grazing animals out of the bivouac area.

The mud-brick and roof-file debris found around the base of the tower originally came from the walls and roof of the tower. The erection of mud-brick walls atop a solid rubble base some 2 meters high suggests that the tower was intended to be as lofty as was practicable using the simplest and most economical of building techniques. A mud-brick structure standing two stories above the base seems likely. A third story is possible, especially if timbers were used to reinforce the mud-brick walls, but this is certainly the maximum that could be allowed for mud-brick walls standing on loosely joined rubble. We may imagine, there-fore, that the Dema tower originally stood close to 8 meters in height if it had two stories, or as much as 11 meters if it had three.^[25]

Entrance to the tower and interior communication between its stories were likely provided by wooden ladders, since no trace of stone steps was found built into or against its tall rubble base. If we introduce evidence from the Hymettos tower, the remains of which are similar to those of

[25] DEMA 173 suggests that a structure of one or two smiles stood atop the base. The height of each story of the Dema tower can be approximated at 2.50 to 3.00 meters, to judge by other examples of room heights. This is the suggested height of the ground-floor rooms of the Vail house ("Vari House" 442), and the smiles above ground level in two towers built of cut stone, one on Andros and one on Keos, fall within this range; see Young 1956, 137. DEMA 174 note 52, followed by McCredie 1966, 119 note 3, suggests that the Dema tower was as high as 15-20 meters, which is clearly impossible. The towers of that height cited by DEMA to support this conjecture are all built of well-cut masonry to their full heights and should not be considered suitable evidence for the height of a freestanding mud-brick tower like the Dema tower.

the Dema tower (see figure 34), then it is likely that the Dema tower had a massive central pier taking up as much as half the space within the first story, providing a firm bedding for a central post that probably sup-ported the roof and that might have continued upward as a mast to carry signal flags. The topmost story must have been well provided with windows for observation and to allow flags to be sent up the mast.^[26]

The round shape of this tower gave it more stability than a rubble and mud-brick tower with corners would have had. The roof, however, must have been square in plan, since the roof tiles used in it were canonical rectangular tiles. It may have been single pitched, double pitched (gabled), or pyramidal in form. Each possibility raises questions about how a square roof was erected above a round wall. What little evidence there is suggests that a pyramidal roof was used.

A pyramidal roof would have had the advantage of having all edges of the roof at the same level, that is, with no raking cornices, and would have been the easiest type of roof to erect on a round building. A pyramidal roof is in fact the simplest approximation, in a roof of rectangular files, of a conical roof, which would have required specially shaped tiles.^[27] Even so, a pyramidal roof would have required some of its rectangular tiles to be modified in shape. Pan tiles running down the lateral edge of each of the four triangular facets of the roof would need to be trimmed along a 45-degree line in order for each facet to fit flush against the adjacent facets. Since Lakonian-type tiles have distinct upper and lower ends, the trimmed upper portions would have been discarded. This may well have provided the source of the fide fragments used for the incised gameboards, nos. 21 and 22, and possibly 23, for it is note-worthy that no. 21 is a gameboard incised on a tile fragment with one edge broken along a 45-degree angle to the original sides of the tile. The find contexts of these reused tile fragments do not associate them with any particular phase of activity on the tower site, but it would be most appropriate to assume that they were incised by the idle hands of men on long and uneventful shifts of duty at the tower in its first phase.

Of the pottery finds, only nos. 1 and 2 can be associated with the first phase of the Dema tower. All fragments of both were found in area 2S, where the three fragments of the gameboard no. 21 were also found. The cup no. 1 is represented by two handle fragments, one found on the surface of the ground beneath rubble and the other found a few

[26] The remains of the Hymettos tower with its platform are described in Munn 1983, 406-10; cf. also McCredie 1966, 117-19, and Ober 1985a, 132-33. Textual and archaeological evidence for the use of signal flags at outposts like the Dema tower will be presented in a separate study.

[27] For examples of the special tile systems required for well-built round buildings, see Thompson 1940, 65-73; Roux 1952, 442-83.

meters away, buried in the mud-brick stratum near where no. 2 was found. The handles are from a type of cup produced no later than the end of the fifth century, very probably within the last third of that century. The three fragments of no. 2 were found close by the deposit comprising nos. 3, 14, and 18. The vessel form of no. 2 is not precisely identifiable from the preserved portion, but enough remains to show that this was a type of cup manufactured ca. 420-400. The interval of more than a half-century between the dates of nos. 1 and 2 and the datable pottery associated with the secondary structures clearly separates these pieces from the second phase of the site. Since these cups are unlikely to be merely stray pieces left on this barren hilltop before any structure existed here to make the spot a focus of human activity, these pieces can be identified as debris from the original use of the tower. They provide evidence that the first phase of the Dema tower is to be placed no earlier than ca. 425 and probably no later than a generation after the end of the fifth century, that is, within the range ca. 425-375.^[28]

The evidence of these sherds fits closely with the terminus post quem for the date of the Dema wall established by the Dema wall saltcellar. Taken together, the saltcellar and cups 1 and 2 establish a fairly narrow chronological range within which the construction of the Dema defenses is to be placed. The saltcellar indicates that a date within the last quarter of the fifth century is possible but that a date after the beginning of the fourth century is more likely (appendix I), and the cups nos. 1 and 2, as

[28] As is the case with the sherds assigned to the second phase of use of the tower, especially the best-preserved examples, nos. 3 and 5, an allowance of about a quarter-century beyond the dates assigned to parallel pieces in Agora XII for the period of use and final deposition of individual pieces is appropriate when extrapolating from parallel pieces to the date of an archaeological context. The dates assigned to pieces in Agora XII represent an amalgamation of context, i.e., deposit date, as determined especially by the presence of figured fragments, and an estimation of the place of an individual shape in a stylistic series, i.e., by comparison with parallel pieces from other deposits (Agora XII, 46). As Sparkes and Talcott point out, this system of dating is founded on the "cumulative and interlocking" nature of the evidence of the deposits, and considering the number of de-posits and volume of material consulted by them, it should inspire confidence that the resulting dates are reasonably accurate. It should be noted, however, that since the foundations of their chronology are, first, the chronology of figured pottery and, second, judgments about the nature and rate of stylistic evolution of vessel shapes (a process of dating comparable to the stylistic criteria applied to vase painting), the entire system of dating is strongly influenced by criteria related to the date of manufacture of individual pieces. Al-though, with the manufacture, the period of use of any given piece may well lie within the time range, often of a quarter-century, assigned in Agora XII, it may extend beyond it by as much as a generation. So Sparkes and Talcott caution: "There is often no clear indication of the lag between the time when a vase was made and the time when it was broken and thrown away," and "Ordinary tableware and domestic pottery may  . . . lay claim to a considerable life simply because it remained quite good enough for everyday use" (Agora XII, 46 and note 129).

first-phase occupation debris from the tower, indicate that a date any later than ca. 375 is unlikely.

This conclusion corresponds well with the terminus ante quem provided by the material from the second phase of the site. After its initial use, the tower was abandoned for a sufficient length of time that it fell into ruin, and its debris was reused by a beekeeper to provide shelter for his stands of beehives. The most probable period for this reoccupation of the tower site, ca. 340-300, indicates that original construction of the tower should probably be placed before the middle of the fourth century. A construction date within the first quarter of the fourth century would allow an adequate interval for the tower to have become dilapidated and clearly useless as a watchtower by the time the beekeeper arrived.

Excavation of the Dema tower has thus yielded evidence, corroborated by the Dema wall saltcellar, indicating that the construction of the Dema wall and tower might be placed as early as the last quarter of the fifth century and more probably is to be placed within the first quarter of the fourth century. This agrees well with the analysis of other archaeological criteria and with the evidence for the tactical plan of the wall discussed in chapter 2. This conclusion is based in part on the assumption that the Dema wall and tower were built at the same time as part of the same defensive scheme. The demonstration of this assumption depends upon an understanding of the function of the Dema tower and of its fundamental importance to the defensive scheme embodied by the Dema wall.