We found more than ten thousand pieces of glass in the excavations in the fort from the first through the sixth season. Observation of these, mainly the decorated glass, indicates that most are obviously dated between the 9^th and the 10^th centuries.¹ This is the transitional period from the traditional techniques of Roman glass to Islamic glass in the regions of Egypt, Palestine and Syria. Glass from show the nature of the changes in respect to composition, technique, decoration, and so on.

 As for the composition, as mentioned in a later chapter, there are both Group A whose composition is natron used in Roman glass and Group B, the so-called Islamic glass composition using plant-ash. The transition from Group A to Group B is supported by the context of the strata, because objects of the former group were unearthed from lower layers while those of the latter were from upper layers.²

 As for the relationship between the decoration and the composition, it was clarified through the previous studies that the majority of glass with incised, brown luster-stained, impressed and molded decoration belongs to Group A, and the glass with wheel-cut, orange or polychrome luster-stained decoration belongs to Group B.³ This shows the different date of the decorated glass and, at the same time, the different production sites from which the glassware was exported to . It is presumed that the pale bluish-green glass, making up the majority of the glass including undecorated glass, was a local production in places not distant from . The decorated glass in Group A, except for the glass with incised decoration on a colored background, is mainly made of natron. Whereas, the colorless glass with wheel-cut decoration that is completely decolored by the use of manganese might be manufactured in Egypt or Iraq. These products are regarded as high quality imported goods. The fact that colorless glass with wheel-cut decoration was found in the fort shows that high quality products were brought to the most powerful or richest person in the city. However, as for the low quality pale green or pale bluish-green glass, even if it has wheel-cut decoration, it can be thought of as a local product.

 In this season, 486 glass fragments were found in the fort. The number of pieces of decorated glass of Group A or glass with luster-stained, impressed or incised decoration (pls. 12-7, 34-18), is very small. A fragment with wheel-cut decoration of good quality was found in Room No. 3-1 (pls. 12-8 to 12-11, 34-19). Plate 12-8 is a part of round rim of a colorless cylindrical beaker, which has shallow, vertical-cut decoration. Plates 12-9 and 10 are parts of small bottles with cylindrical body or cup. Both have deep-cut decoration and complicated geometrical patterns. No. 9 is colorless and no. 10 is pale bluish-green. Plate 12-11 is a bottle with an outward-curved rim and the body is presumed to be cylindrical. It is colorless, and has vertical rectangular facet cut decoration on the neck.

 We found two small bottles. One of them is almost intact, 3.5 cm in height and colorless. The rim is folded inward and the body is rounded (pls. 12-12, 34-20). Another one is a small intact bottle 5.2 cm high typical of the 9^th and the 10^th centuries (pls. 12-13, 34-21). The body is shaped like a rectangular parallel-epiped and the neck is cylindrical, which shows a slight warp. The rim was reheated to make it round. The base is flat. The glass is colorless, and the chemical analysis of its composition reveals that it belongs to Group B.

　

 We started the excavation this season in the residential quarter, where we found 1,111 glass fragments. As mentioned in a later chapter, it is a distinguishing feature that most examples belong to Group A (before the 9^th century) in composition, and few objects of Group B were excavated from the upper layers of the fort. Consequently, among the unearthed decorated examples there is no type that is typical of Group B, such as glass with wheel-cut decoration and with orange luster-stain. Also, among the glass in Group A, there were only three pieces of the monochrome luster-stained glass.⁴

 Threaded decoration was one of the decorative techniques often used in late Roman glass, and it was used in the early Islamic period. Two pieces found in the residential quarter appear to be parts of small bottles with a thin wall and a spherical body (pls. 18-1 &2, 36-14 & 15). Plate 18-1 is a reconstruction drawn by jointing some fragments in comparison with some similar examples. The rim is slightly bent inward and the neck is short and cylindrical. The body is spherical. The base is presumed to be slightly convex. A thin thread the same color as the body encircles below the neck and continues in a zigzag way around the upper part of the body. A white stripe appears on the body, but it is probably attributable to impurities and not an intentional imbedded decoration, but this point is under consideration. Plate 18-2 is the neck of small bottle with an outward flared rim. In addition to the thin thread, a thick wavy brown thread encircles the neck. The thick wavy thread decoration was used in Roman glass. Plate 18-3 is a small fragment with impressed and projected decoration, characteristic of late Roman glass. These three fragments can be regarded as products of the early Islamic period which remain the techniques of the previous period.

 Plate 18-4 to 18-8 (pl. 36-16 & 17) are glass with impressed decoration, the so-called pincer decoration. The term "pincer decoration" is derived from the technique of applying the patterns on the glass. The glass wall is pinched by a pincer, an instrument like large tweezers used for shaping and modeling glass. In fact, other instruments were used for adding patterns.⁵ The dents are made by pressing an instrument both inside and outside the vessel wall and then pinching. The dents are characteristic of this technique. This type has the most examples among the excavated glassware from . Products with a combination of simple patterns were unearthed in the residential quarter. The patterns are composed of inverted angles; vertical broken lines (pl. 18-4); ellipses (pls. 18-5, 36-16): triple ellipses (pl. 18-6); double inverted drops (pl. 18-7); double diamonds (pls. 18-8, 36-17), etc. All are cylindrical beakers. As for color, only plate 18-4 is pale brown, while the others are pale bluish-green, which is the most popular type of glassware at .

 There are products with a complicated combination of vertical, horizontal and oblique lines in the glass with impressed decoration. Plate 18-9 is part of the rim of a dark green beaker. Plate 18-10 is a bowl of pale purple. The surface is weathered hard. A series of semicircles are formed below the rim, while are alternately filled with oblique lines. A braid pattern was added below the semicircles. Thin oblique lines surround the braid pattern. This exquisite pattern appears on bowls unearthed from and the Famen Temple in China, showing it belongs with those of the highest quality.

 Mosaic glass in opaque colors was also found (pls.18-11, 36-18). The mosaic technique flourished in the Hellenistic period. It declined after that because it required such great skill, but it was revived with the appearance of Islamic glass in the 9^th century. The object is a fragment of a rare vessel with linear and floral patterns on a yellow ground. The flower pattern has a red circle in the center, and alternate black and white lines around it, and is further edged with black circles. It is made of glass cane consisting of colors whose section appears flower-like. Then, the sliced glass cane is set in the mold. Examples made in this technique are very few in Islamic glass. However, this floral pattern was used in glass bracelets after the 13^th century, which shows that the technique was taken over for small-sized accessories like beads or bracelets after it had been discontinued for vessels.

 Beakers occupy a high percentage of the undecorated glass in the residential quater as well as in the fort (pls. 18-13, 36-19). Also there are examples with rims folded inward which are in good condition (pl. 18-12).

　

　

 Investigation in this season has revealed the mutual relationship of some motifs reported so far by furthering the joining work of painted plaster pieces. The present authors advance the reasonable assumption that the decorative paintings were rendered only on the qibla wall because many painted plaster pieces were excavated from the floor along the qibla wall. This study particularly has elucidated the configuration of the decoration that is considered to have been drawn roughly on the entire qibla wall. Moreover, further investigation of plaster pieces presumed to have been in contact with the ceiling board has revealed some architectural techniques used in construction of this mosque.

 Many small painted plaster pieces were newly excavated from under the floors of Rooms No. 3-8B, 3-12A and 10-1, which are separated by the street in front of the mosque, during the excavation in this season. Analysis proved that several of these pieces could be joined with those already discovered from the mosque. The painted plaster pieces that had initially collapsed inside the mosque were artificially carried outside, suggesting that more plaster pieces might be discovered by the progress of the excavation around the mosque. However, those found in this season had been subjected to severe salt damage. For that reason, many of them retain only a slight portion of the original painting: even their motifs are unclear.

 Progress of joining work will be described hereinafter for each motif. Subsequently, the assumed decorative motifs on the whole qibla wall will be discussed based on a comprehensive study.

　

　

　

　

　

　

　

 In this season, we investigated in detail the pieces of the reentrant corner excavated inside the mosque, some of which have an imprint by a flat plane. Two typical examples are presented in pl. 1-5 and pl. 2-6. Plate 1-5 shows the corner piece belonging to Group B (RB166 and RB169), which is a fragment with a fibrous indentation on a flat surface, suggesting that it had been in contact with a wooden surface. It is inferred to be a plaster piece that had been in contact with the wooden ceiling board. The other example shown in pl. 2-6 has a similar flat imprint and an indentation by a log on the reverse side (RB259 and PB306). The shape of the indentation suggests that it was made by a palm log that had been embedded in the wall. The palm log diameter can be estimated from the indentation curvature to be about 17 cm. This value does not differ considerably from the size of palm logs found in the fort of the site so far.

 Many of palm logs discovered in the fort had been split in half longitudinally. Their excavation situations and fabric materials unearthed together imply that they had been used as beams supporting the ceiling. However, the reentrant corner pieces with an indentation by a palm log found inside the mosque indicate that the palm log had been embedded inside the wall near the ceiling board. Namely, it seems rather reasonable in this case that this palm log was a corbel for the ceiling board embedded in the wall in order to place ceiling boards horizontally in place of a beam. The architectural technique around the ceiling edge is inferred from imprints as those shown in pl. 2-7.

　

 As the previous report pointed out, the gross area occupied by the excavated painted plaster pieces is quite small compared to the area of the entire qibla wall. Therefore, careful consideration is needed for restoration of the whole decorative painting. Nevertheless, it is considered that a presumption based on many issues resolved so far will allow its approximate restoration.

 Many plaster pieces were excavated along the qibla wall. Particularly, they were discovered intensively at the east corner of the mosque. The joining work of pieces found there could allow restoration of two decorative paintings of Groups A and B.

 A corner piece of the reentrant corner indicating that it had been in contact with the ceiling board was contained in Group B, whereas no such corner piece was contained in Group A. This fact is believed to provide an important key suggesting which motif was drawn on the upper qibla wall. At present, it seems reasonable to assume the following Group B, excavated with the corner piece near the ceiling, was probably drawn on the upper wall, below which the motif of Group A was rendered under some spacing to accommodate the restored leaf-shape plant pattern. Because it has been proven that a plain white ground had extended over the rosette figures drawn horizontally in the motif of Group B, it is inferred that a blank space had been left like a belt up to the ceiling board.

 The difference in the depth from which Groups A and B were excavated implies the collapse process of this part. Group A was excavated from the lower layer, and deposition was observed between excavation layers of Groups A and B. These facts indicate first that the decorative painting collapsed at the lower part of the wall and fell on the floor, subsequently, the decorative painting on the upper wall collapsed after a certain period of time.

 Furthermore, it is characteristic that the color of the painted plaster pieces discovered from the east corner of the mosque had been deteriorated less than that of other pieces. The reason may be that the ceiling at the east corner remained standing for a longer time; thereby, wind and rain could be partially avoided compared to other parts.

 In conclusion, based on the above discussion, the restored images of the interior decorative painting proposed at present are presented in pl. 2-8 and color pl. 2.