- 1 For the evidence and interpretation of setting lines in Greek monumental architecture of the Archa (...)
1Incised and painted lines were commonly used in stone masonry since the Archaic period to indicate the position of architectural elements on the top surface of their underlying course. While setting lines were originally meant to guide ancient builders, for modern scholars, they can reveal the place and dimensions of missing elements, the sequence of construction, or even mistakes and corrections that occurred in the building process.1 The case presented in this article is of particular interest, as it tells the story of an architect adjusting a building plan, while struggling with a very unusual task: adapting the Doric order to the interior design of a circular burial chamber.
2The Starosel tomb in central Bulgaria (fig. 1), is the largest and most elaborate funerary monument in pre‑Roman Thrace. Built upon a natural hill, and covered with an enormous earthen mound encompassed by a wall of granite ashlars (73–79 m in diameter), the tomb comprises an array of diverse plan components, structures, and decorative elements, masterfully blended into a unified composition. It consists of two underground chambers—a tholos and a rectangular antechamber—arranged in an axial sequence along an open dromos, and a monumental propylon with branching stairways connected to an elevated pathway around the mound. The outstanding craftmanship and skilful engineering indicate that the Starosel tomb was the work of an experienced architect and stone-carvers from mixed backgrounds, with extensive knowledge of monumental architecture from across the Aegean. It was built between 350 and 330 BC—most likely as a tomb and heroon for one of the last Odrysian rulers of Thrace, before Philip II of Macedon’s conquest of the region in 341 BC. The architecture of the Starosel tomb and its date have been discussed in detail elsewhere.2 In this article, I will focus on the setting lines and their relation to the planning of the tomb’s Doric tholos.
Fig. 1 — Location of Chetinyova Mogila, Starosel.
Ch. Tzochev.
- 3 Measured at the wall bottom, in the centre of each intercolumniation.
- 4 Measured from the stylobate level. The dome was reconstructed and some doubt remains about the exa (...)
- 5 An orthoimage of the stylobate top surface can be downloaded from https://journals.openedition.org (...)
3The Starosel tholos is an underground chamber, built of volcanic tuff ashlars upon granite foundations. It has a circular plan and is covered with a corbel dome (figs. 2–4). Its internal diameter is 5.38–5.39 m3 and its total interior height is 7.83 m.4 The walls of the tholos are articulated with ten engaged Doric columns (labelled A–J, counterclockwise), topped by a Doric entablature and an Ionic geison. The bottom wall course and the column drums (carved together with the blocks of this course), rest on a stylobate made of 25 volcanic tuff blocks. The visible upper surface of the stylobate was smoothed with a toothed chisel after the stylobate was assembled. This surface preserves the following incised lines (figs. 5–12):5
|
|
1) |
Lines indicating the bottom edge of the tholos wall. A line, which presumably makes an almost full circle on the stylobate, is visible at several places, where the bottom wall edge is broken off (fig. 5). It can also be seen to the right of column A, where it continues beyond the wall, to the end of the stylobate block (figs. 6, 7). On the same block, there is another line, parallel to, and slightly offset from the side of the tholos wall. A similar line probably existed on the now damaged stylobate surface on the opposite side of the entrance. |
|
|
2) |
Lines marking the position of the column drums on the stylobate (figs. 5–9). These semi-circular outlines are fully or partially visible around most of the 10 columns. The arrises of the bottom drums are aligned with the lines, or, in case of columns G, I, and possibly H, the drums are slightly offset from the lines. At column A, the curve is doubled by a second, shallower one, offset towards the entrance. |
Fig. 2 — Reconstructed plan and longitudinal section of the chambers.
Scale 1:90.
Drawing Ch. Tzochev.
Fig. 3 — Exploded axonometric section of the tholos.
Drawing Ch. Tzochev.
Fig. 4 — Unfolded view of the tholos interior: photogrammetric model and reconstruction.
Discrepancies in horizontal scale are due to “flattening” of elements in different planes.
Photo and drawing Ch. Tzochev.
Fig. 5 — Lines incised on the stylobate at column G.
Photo Ch. Tzochev.
Fig. 6 — Lines incised on the stylobate at column A.
Photo Ch. Tzochev.
Fig. 7 — Lines incised on the stylobate at column A.
Photo Ch. Tzochev.
Fig. 8 — Lines incised on the stylobate at column F.
Photo Ch. Tzochev.
Fig. 9 — Lines incised on the stylobate at column D.
Photo Ch. Tzochev.
|
|
3) |
Pairs of straight lines, incised on the stylobate radially, beginning at the edge and extending under each column (figs. 5, 7–10). One of the lines in each pair always passes through the column axis. The other line in the pair is slightly offset from the centre. The offset varies from column to column, being between 0.006 and 0.035 m. These lines are visible at all columns, except for E, H, and J, where the stylobate surface is damaged. In most cases, only the parts of the lines that project in front of the drums are visible. Where the bottom drum of column F is broken, one can clearly see that these lines continue under the columns, and probably extend to the edge of the wall (and maybe even further, under the wall blocks). |
Fig. 10 — Lines incised on the stylobate at column B.
Photo Ch. Tzochev.
- 6 The original position of the architrave has changed due to displacement in the tholos structure ca (...)
|
|
4) |
A curved line incised on the stylobate block on the left of column D (fig. 9). The curve is parallel to the edge of the stylobate and inset 0.070 m from it, having a radius of 2.491 m measured from the geometric centre of the stylobate. It coincides with the projection over the stylobate of the edge of the architrave above (radius 2.491–2.495 m).6 |
4In addition to these lines incised on the stylobate,
|
|
5) |
Each bottom drum has a notch in the middle of the central flute, aligned with the centred line in the pairs described in point 3 above (fig. 10). |
|
|
6) |
On one capital fragment, the top surface of the abacus plate preserves traces of two lines, about 3 mm thick, made with red paint (fig. 13). One of the lines is curved, and is parallel to the frontal curve of the abacus, 4.3 cm from the front edge of the crown plate. It indicates the line to which the edge of the architrave was aligned. The second line is perpendicular to the first one, and splits the capital exactly in the middle. This line indicated the column axis and the place of the joint between the architrave blocks. The fragment belongs to the capital of column A, C, or J—the remaining capitals have their central parts more or less preserved. Presumably, similar lines have been painted on the top surfaces of all capital blocks. |
Fig. 11 — Plan of the tholos stylobate with visible incised lines (in red).
Scale 1:50.
Drawing Ch. Tzochev.
Fig. 12 — Reconstruction of the lines incised on the tholos stylobate.
Scale 1:50.
Drawing Ch. Tzochev.
Fig. 13 — Painted setting lines on a capital fragment from the tholos: photograph and reconstruction.
Ch. Tzochev.
- 7 For incised lines marking column circumference and axis, wall face, and architrave joint, see e.g. (...)
5The purpose of most of these marks as setting lines for the bottom course of wall blocks and column drums (points 1, 2, 3 and 5), and for the architrave blocks (point 6) is self-evident, and they have parallels in other ancient buildings.7 The only exception is the curved line on the stylobate (4), which represented the projection of the architrave edge. Its purpose may have been to verify the position of the architrave block above by dropping a plumb line from it, in which case it had a similar function as the curved setting line painted on the capital top (fig. 14).
Fig. 14 — Reconstruction showing the proposed function of painted and incised lines at column D.
Drawing Ch. Tzochev.
6Among the marks on the stylobate, the pairs of straight, radially incised lines (3), require further explanation. If one of the lines in a pair indicates the column axis, what is the purpose of the other line? The following observations help us answer this question:
- Within a pair, where preserved, the offset line is always from the side of the tholos entrance, i.e. on the left of column axes A, B, C, and D, and on the right of column axes F, G, and I.
- The space between lines in a pair decreases from the entrance towards the back of the chamber (table 1, actual spaces).
- The intercolumniations in the tholos are not equal: the columns by the entrance are more distant, presumably to give wider space for the door, while the remaining intercolumniations are nearly equal (table 2).
Table 1 — Spaces between primary and secondary column-centring lines.
Column |
Actual spaces |
Theoretical spaces |
Difference actual/ theoretical spaces |
Angle between radii (°) |
Linear space on the stylobate edge (mm) |
Angle between radii (°) |
Linear space on the stylobate edge (mm) |
Angle between radii (°) |
Linear space on the stylobate edge (mm) |
A |
0.815 |
34 |
0.815 (actual) |
34 (actual) |
– |
– |
B |
0.573 |
24 |
0.634 |
26.4 |
0.061 |
2.4 |
C |
0.465 |
19 |
0.493 |
18.9 |
0.028 |
−0.1 |
D |
0.175 |
9 |
0.383 |
11.3 |
0.208 |
2.3 |
E |
– |
– |
0.298 |
3.8 |
– |
– |
F |
0.167 |
6 |
0.298 |
3.8 |
0.131 |
−2.2 |
G |
0.482 |
18 |
0.383 |
11.3 |
−0.099 |
−6.7 |
H |
– |
– |
0.493 |
18.9 |
- |
– |
I |
0.757 |
31 |
0.634 |
26.4 |
−0.123 |
−4.6 |
J |
– |
– |
0.815 |
34.0 |
– |
– |
The theoretical spaces reflect a wider intercolumniation J–A, and perfectly equal remaining intercolumniations.
Table 2 — Interaxial spaces in the tholos.
Inter- columniation |
Angle between radii (°) |
Linear space (wall curve, m) |
Linear space (wall chord, m) |
Value |
Standard deviation (mean 36°) |
Value |
Standard deviation (mean 1.692 m) |
Value |
Standard deviation (mean 1.665 m) |
A–B |
35.292 |
−0.708 |
1.660 |
−0.033 |
1.633 |
−0.032 |
B–C |
35.725 |
−0.275 |
1.678 |
−0.014 |
1.652 |
−0.013 |
C–D |
35.859 |
−0.141 |
1.684 |
−0.008 |
1.658 |
−0.007 |
D–E |
36.132 |
0.132 |
1.700 |
0.008 |
1.671 |
0.006 |
E–F |
36.070 |
0.070 |
1.696 |
0.003 |
1.668 |
0.003 |
F–G |
36.132 |
0.132 |
1.699 |
0.007 |
1.671 |
0.006 |
G–H |
36.125 |
0.125 |
1.697 |
0.005 |
1.670 |
0.005 |
H–I |
35.753 |
−0.247 |
1.680 |
−0.013 |
1.654 |
−0.011 |
I–J |
35.616 |
−0.384 |
1.676 |
−0.016 |
1.647 |
−0.018 |
J–A (entrance) |
37.317 |
1.317 |
1.754 |
0.062 |
1.723 |
0.058 |
Note that in a circular space the mean values are also the ideal values, as if spaces are evenly distributed.
- 8 The theoretical knowledge behind the division of a circle in ten equal parts was first described b (...)
- 9 As shown by Bousquet 1993, esp. p. 308.
7It appears that the pairs of radial setting lines reflect two stages of calculating the positions of the column axes. The lines on the side of the entrance, which are now offset from the centres of the columns, were made first, in order to divide the circle of the stylobate in ten equal parts. The exact method with which this division was achieved is unknown, but several practical solutions requiring only a straightedge and compass were available to 4th‑century-BC architects.8 One of these solutions involves finding the edges of an inscribed regular pentagon,9 and dividing it into halves to obtain the tenths. The same can be achieved by drawing a series of intersecting circles (fig. 15). Seven marks made during this initial division are now visible on the stylobate, allowing us to measure four out of the ten angles between the radii at the centre. The measurements given in table 3 show that the division was done with a deviation of up to 0.47° from the ideal 36° (1/10 of a circle), which, projected on the stylobate edge, gives an error of ca. 0.02 m. At least three of the currently visible circular outlines meant to indicate the position of column drums are aligned to this first set of lines—at columns A, G, and I.
Table 3 — Angles between the radii corresponding to the first set of lines.
Intercolumniation |
Angle between radii (°) |
Deviation from the ideal angle (36°) |
A–B |
35.534 |
−0.466 |
B–C |
35.833 |
−0.167 |
C–D |
36.149 |
0.149 |
D–E |
– |
– |
E–F |
– |
– |
F–G |
36.459 |
0.459 |
G–H |
– |
– |
H–I |
– |
– |
I–J |
– |
– |
J–A |
– |
– |
Fig. 15 — Method of dividing a circle in 10 equal parts using a straightedge and compass.
Drawing Ch. Tzochev.
8The decision to leave a wider intercolumniation at the entrance—1.317° wider than the ideal 36°—created a further complication, namely the need to equalize the remaining nine intercolumniations. This required shifting the initial dividing lines—actually marking a second set of lines, offset away from the door. To achieve an equal distribution, the offset had to decrease gradually as one moves away from the entrance. This is exactly the case with the offsets between the actual marks on the stylobate, where preserved. Distributing the spaces evenly and accurately requires that: (1) the columns flanking the entrance are offset symmetrically, and (2) that each offset further away from the entrance decreases by the amount with which the intercolumniation at the entrance was widened, divided by the number of contracted columns—whether calculated in terms of angle, or in linear space between the radii (fig. 16).
Fig. 16 — Principle of equalizing spaces between column axes after widening a single intercolumniation (offsets exaggerated for clarity).
Drawing Ch. Tzochev.
9Let me illustrate this process in practice. Since the initial division of ten equal parts was imperfect, we can take the space between the marks at column A, 34 mm, as a base, accepting that the intercolumniation at the entrance (J–A) was widened symmetrically by 34 × 2 = 68 mm. To equalize the spaces between the remaining nine marks, one needs to shift those at columns B and I away from the entrance with 26.4 mm (34 − 68/9) ; followed by a similar shift of the marks at columns C and H with 18.9 mm (26.4 − 68/9), etc. A comparison between theoretical and actual spaces (where preserved) shows that some are remarkably close, while others deviate up to 7 mm (table 1). Although the correction was not perfectly accurate, the figures suggest that the architect understood the principle and sought to equalize the intercolumniations.
- 10 To a small extent these offsets may be due to secondary deformation in the tholos structure, see n (...)
10The lines incised on the stylobate give a rare insight into the practical calculations made by an ancient architect to solve a problem caused by the geometry of a circular building plan. However, distributing the interaxial spaces was not the end of the problems for the architect and the stone-carvers. It was, in fact, where the real challenges began, as the adjusted positions of the column axes required corresponding adjustments in the Doric entablature above. In a Doric elevation with no corners and evenly spaced column axes, triglyphs and architrave joints would normally align with the column axes, and the widths of triglyphs and metopes would be consistent. Using uneven interaxial spaces required that one of these rules be broken in order to keep the rest in balance. A number of irregularities observable in the Starosel entablature testify to difficulties with making the necessary adjustments. The lengths of the individual architrave blocks vary so that most of the joints between them do not coincide with the column axes, but are offset from them by up to 0.04 m.10 Furthermore, the architrave block spanning columns A and B lacks a half-regula on the right (fig. 17) and has unusually wide spaces between the rest of the regulae (ca. 0.03 m wider than the mean). It seems that initially the block was carved longer and was then shortened from the right, which led to the removal of the half-regula. Reworking the block this way was a simple but short-sighted solution, as it caused discrepancy in the frieze above. The width of the triglyphs is relatively consistent, while the width of the metopes varies significantly, as fig. 4 and table 4 show. These variations were dictated by the positions of the regulae on the architrave blocks because a triglyph has to be aligned above a regula. Hence the abnormal architrave block A–B caused a problem: the frieze above has two unusually wide metopes, followed by two disproportionally narrow ones on the right. To avoid cramming triglyph T1 even closer to T2, the first was centred above the single half-regula. The result is a break in the Doric frieze system. Perhaps, working under time pressure, the builders sought to conceal their mistake, leaving triglyphs T1 and T2 unpainted (figs. 4, 17).
Table 4 — Widths of triglyphs and metopes in the tholos.
# |
Triglyph |
Metope |
Width (m) |
Standard deviation (mean 0.225 m) |
Width (m) |
Standard deviation (mean 0.296 m) |
1 |
0.225 |
0.000 |
0.230 |
−0.065 |
2 |
0.228 |
0.003 |
0.328 |
0.033 |
3 |
0.237 |
0.012 |
0.318 |
0.023 |
4 |
0.232 |
0.007 |
0.283 |
−0.013 |
5 |
0.228 |
0.003 |
0.290 |
−0.005 |
6 |
0.213 |
−0.012 |
0.310 |
0.015 |
7 |
0.224 |
−0.001 |
0.300 |
0.005 |
8 |
0.225 |
0.000 |
0.286 |
−0.009 |
9 |
0.228 |
0.003 |
0.298 |
0.003 |
10 |
0.225 |
0.000 |
0.288 |
−0.007 |
11 |
0.226 |
0.001 |
0.295 |
0.000 |
12 |
0.229 |
0.004 |
0.295 |
0.000 |
13 |
0.229 |
0.004 |
0.298 |
0.003 |
14 |
0.225 |
0.000 |
0.290 |
−0.005 |
15 |
0.223 |
−0.002 |
0.303 |
0.008 |
16 |
0.223 |
−0.002 |
0.294 |
−0.001 |
17 |
0.227 |
0.002 |
0.297 |
0.002 |
18 |
0.228 |
0.003 |
0.292 |
−0.003 |
19 |
0.221 |
−0.004 |
0.301 |
0.006 |
20 |
0.226 |
0.001 |
0.296 |
0.001 |
21 |
0.223 |
−0.002 |
0.298 |
0.003 |
22 |
0.229 |
0.004 |
0.283 |
−0.013 |
23 |
0.226 |
0.001 |
0.296 |
0.001 |
24 |
0.226 |
0.001 |
0.289 |
−0.006 |
25 |
0.226 |
0.001 |
0.297 |
0.002 |
26 |
0.212 |
−0.013 |
0.321 |
0.026 |
27 |
0.226 |
0.001 |
0.300 |
0.005 |
28 |
0.226 |
0.001 |
0.305 |
0.010 |
29 |
0.221 |
−0.004 |
0.302 |
0.007 |
30 |
0.226 |
0.001 |
0.282 |
−0.014 |
Fig. 17 — Detail of the entablature above column A, showing an omitted half-regula, and a heavily contracted metope between two unpainted triglyphs.
Photo Ch. Tzochev.
- 11 Coulton 1977, pp. 89‑90.
11Adapting the Doric order to the interior elevation of a circular building was a strikingly innovative choice, unparalleled in the 4th century BC. But the challenge that the Starosel architect faced with increasing the interaxial span between columns flanking an entrance was not uncommon in Greek monumental architecture. This kind of adjustment was practiced for aesthetic purposes, especially in small buildings. In Doric façades, it required an adjustment in the entablature, by varying the widths of the metopes, offsetting triglyphs and column axes, or adding an extra triglyph-metope pair. A good example is the treasury of the Athenians at Delphi. After the stylobate of the treasury was laid, the architect decided to slightly widen the central intercolumniation in the façade. The resulting discrepancy in the Doric entablature was solved with a longer architrave block in the centre, bringing the architrave joints in line with the column axes, while the triglyphs above were slightly offset.11 This way the width of the metopes and triglyphs remained consistent. The Starosel architect could have adopted this approach, or kept the triglyphs aligned over the column axes and varied the metope widths only (the extension above the door was too small to be filled with an additional triglyph-metope pair). Instead, in the Starosel entablature we can see both misaligned triglyphs and varying metopes widths, and these variations definitely do not look like carefully planned adjustments. The shortened architrave block A–B suggests that some of the elements of the entablature may have been carved in advance with a different layout in mind.
12To better understand the decisions made by the Starosel architect, we need to go back to the moment the tholos stylobate was laid, and ask why the entrance intercolumniation was widened. In Greek monumental buildings, it was not uncommon to make small adjustments for aesthetic purposes. Similar motives would not be surprising in the case of the Starosel tomb, where great care was given to details and other well-executed refinements are evident, such as column entasis, which were achieved purely for visual, rather than structural ends. However, there are indications that the architect was correcting a miscalculation rather than trying to refine the visual appearance of the tholos. At column A, left of the door, the stylobate preserves a circular setting line for the drum centred on the primary mark—the one made after the first division of the stylobate. This suggests that the division of 10 equal parts was not just a step towards calculating the final positions of the columns. An original plan with evenly distributed columns was actually meant to be implemented, but after drawing the first drum outline on the left of the door the architect decided to rework the plan with a wider intercolumniation at the entrance. The most likely explanation is that the initial division conflicted with the planned door width.
- 12 The antechamber and the tholos are independent structures and there is no clear evidence of the or (...)
13The width of the door, the diameter of the stylobate, the number of engaged columns and their diameters were interdependent parameters that were essential to the project and must have been decided before the construction of the tholos began. However, calculating these dimensions geometrically (derived from the intersection of the square antechamber and circular tholos) would have been difficult without full scale drawings. It was only when the stylobate was set and the layout of the columns drawn out in full scale that a miscalculation of a few centimetres would have become apparent. By that time, the tholos door may have already been carved. The tholos was equipped with a stone door leaf, which opened towards the interior, and was set in sockets cut in the stylobate and a lintel block that was a part of the tholos wall structure. This stone leaf, that weighed nearly 600 kg, had to be set in place before the top wall courses, meaning that it would have been designed and executed in advance. The leaf fitted in a richly decorated Ionic doorframe, which was integrated in the antechamber wall. Replacing or reworking the door leaf and frame would have been expensive and would have slowed down the construction, especially if the antechamber was built before the tholos.12 The choice to rework the tholos layout must have seemed a better option, despite the complications in the Doric frieze.
- 13 Evidence of burial(s) in the chambers, and indications that the propylon of the tomb has never bee (...)
- 14 Capelle 2020, p. 59.
14The Starosel setting lines give us a rare insight into the processes of planning the colonnaded interior of an ancient tholos, and solving unexpected challenges in the execution of an innovative design. The lines show an original plan of a circular stylobate divided in ten equal segments, one of which was consequently widened, and the rest adjusted accordingly. The revised layout affected the spacing of the Doric frieze above. The failure to address the resulting implications in a consistent way should not be seen as a lack of experience or understanding on the part of the architect. The Starosel tomb was an extraordinary project, involving complex logistics and work under time pressure.13 It likely required multiple teams of specialists operating simultaneously, and stone-carvers preparing architectural elements in advance. When mistakes occurred, they were corrected along the way. Avoiding errors in such complex and innovative building projects required a more advanced form of architectural drawings, namely blueprints, that allowed architects to visualize elevations, plans, and sections, and to test different building options beforehand. The earliest such drawings made on independent surfaces appeared around the time the Starosel tomb was built, probably in response to the increased variety of architectural forms in that period.14 Although in the case of Starosel there is no evidence that blueprints were used, the tomb helps us to understand why such drawings developed, and perhaps, how they developed: from setting lines and from the need to test and correct different versions of the plan.