Figure 1. Model of a house. China, Eastern Han dynasty, 1st century CE. Earthenware with traces of slip and pigment; H x W x D (overall): 14 x 13.4 x 10.5 cm. The Dr. Paul Singer Collection of Chinese Art of the Arthur M. Sackler Gallery, Smithsonian Institution; a joint gift of the Arthur M. Sackler Foundation, Paul Singer, the AMS Foundation for the Arts, Sciences, and Humanities, and the Children of Arthur M. Sackler, Arthur M. Sackler Gallery, S2012.9.3315.
Introduction: Chinese Origins
Architectural ornaments known as chimi 鴟尾 in Korea adorn the two ends of the main roof ridge of traditional wooden buildings.1 Literally meaning “owl’s tail,” the term chimi first appears in the “Biography of Yu Wenkai (Ch. Yu Wenkai zhuan 宇文愷傳)” in the History of the Northern Dynasties (Ch. Beishi 北史) compiled during the Tang 唐 dynasty (618–907).2 However, since elements resembling chimi appear in Han 漢 dynasty (206 BCE–220 CE) pictorial stone reliefs and burial objects inspired by actual buildings, China seems to have had a tradition of adorning the ends of the main roof ridge of buildings much earlier (Figure 1). Thus, a record in the Spring and Autumn Annals of Wu and Yue (Ch. Wu Yue chunqiu 吳越春秋) written by Zhao Ye 趙燁 during the Latter Han 後漢 dynasty (25–220 CE) states “On the southern gate tower of the inner city, the roof was raised to form two nimiao 鯢鱙 resembling the horns of a dragon.”3 The roof decoration mentioned here may be understood as roof ridge ornaments shown in Han dynasty stone reliefs and burial objects.4
Figure 1. 
Figure 2.
Architectural details carved in the Northern Wei 北魏 period (386–534) Buddhist cave shrines at Yungang 雲崗 and Maijishan 麥積山 more closely resemble chimi found in Korea. In particular, carvings in the antechamber of Cave 10 at Yungang, the entrance to the antechamber of Cave 43 at Maijishan, and the south wall mural of Cave 257 at Dunhuang 敦煌 depict ornaments at both ends of the main roof ridge that look quite similar to chimi found on the Korean peninsula (Figure 2). Classic examples still in place in China can be found on the roof of the main hall of Nanchansi 南禪寺 on Mount Wutai 五臺山, which was built during the Tang dynasty. The extant building, associated with a rededication inscription dating to the third year of the Jianzhong era 建中 (782), is decorated with roof ridge ornaments at both ends of the main peak of the hip-and-gable design (Figures 3, 4). Based upon all this evidence, the special architectural ornaments that embellish the ends of the main roof ridge of traditional wooden buildings in China, Korea, and Japan must have first appeared in China by the Han dynasty at the latest. Roof ridge ornaments similar to those unearthed in Korea must have been produced in China by the Northern and Southern dynasties (317–589) and continued in the subsequent Tang.
Figure 3. 
Figure 4.
In Korea, chimi have been installed on the tiled roofs of grand wooden buildings like palaces and Buddhist temples throughout the peninsula since the Three Kingdoms period (57 BCE–676 CE). Most ancient examples have been unearthed as fragments at archaeological sites. The largest number of fragmentary chimi come from excavations carried out in Buyeo and Iksan, capitals of the Baekje 百濟 kingdom, and in Gyeongju, the capital of the Silla 新羅 kingdom. In North Korea, a chimi related to the kingdom of Goguryeo 高句麗 was discovered at the Anhak palace 安鶴宮 site. Finally, a green-glazed chimi associated with Balhae 渤海 (698–926) was found within the boundaries of the Chinese city of Ning’an 寧安 in Heilongjiang 黑龍江 province.
Chimi of the Goguryeo Kingdom
The earliest datable Korean chimi may be those represented in numerous murals painted on the walls of ancient Goguryeo tombs. These intriguing images, created during the fourth and fifth centuries, seem to show wooden buildings with chimi at both ends of their main roof ridge. Images including such details have been found in the murals of Tombs 1 and 3 at Anak 安岳 and in a tomb at Yaksu-ri 藥水里. The mural in Tomb 3 at Anak, datable to 357, features chimi-like elements with sharp tips at both ends of the main roof ridge of a building that seems to be a mill (Figure 5). If these images indeed illustrate roof ridge ornaments, chimi or architectural ornaments similar to chimi were already in use by the mid-fourth century in the Goguryeo kingdom.
Figure 5. 
Figure 6.
In addition to the painted depictions, one chimi survives from the Goguryeo kingdom. It is said to have been excavated at the Anhak palace site in Pyeongyang (currently North Korea). Measuring 210 cm in height, it was rebuilt from fragments (Figure 6). Since it has been restored, it is difficult to verify the accuracy of its size, but the reconstruction is significantly larger than the one excavated from the site of Hwangnyongsa 皇龍寺 temple in Gyeongju which is 182 cm tall. Although the main body is completely undecorated, the pair of wings projecting at the back from the base to the tip are marked by a stepped pattern that resembles pleating. A belt between the body and wing is decorated with beaded patterns. A unique feature of the Anhak palace site chimi is the elongated tip, which is circular in cross section and sharply pointed.5 These details are quite different from extant works produced in the Baekje and Silla kingdoms but resemble the images of chimi with pointed tips that are painted on the walls of the Goguryeo tombs already mentioned. The overall form and decoration resemble a green-glazed chimi unearthed at the Balhae 渤海 site of Shangjing 上京, which is currently housed in the Korean Central History Museum in Pyeongyang; the beaded pattern on the belt and the undecorated body of this restored example are also reminiscent of the Goguryeo chimi.6
Surviving evidence from Goguryeo—chiefly the painted tomb murals—suggest that chimi embellished a range of building types, from impressive halls to lesser structures like mills and stables (Figure 7). Further research is required to examine if painted tomb images illustrating chimi reflect pictorial conventions or actual architectural practice, signifying that they were used more widely—and in a broader range of building types—than in the Silla and Baekje kingdoms.
Chimi of the Baekje Kingdom
Chimi are not yet known from the early Baekje period remains at Hanseong 漢城 and Ungjin 熊津 but have been unearthed in some number at later Baekje sites in and around present-day Buyeo and Iksan. Although some twenty chimi dating to the Baekje period have been excavated thus far,7 it has been difficult to reconstruct their original forms with confidence since a significant number of examples were found both broken and incomplete. Fortunately, the relatively whole sets of fragments unearthed from the temple sites at Mount Buso 扶蘇山,8 Mireuksa 彌勒寺, and the recently discovered Wangheungsa 王興寺9 have facilitated the reconstruction of their overall form thus contributing significantly to our understanding of the distinct characteristics of Baekje chimi.10 The restored height of the chimi from the temple site at Mount Buso measures around 90.9 cm and the overall color ranges from gray to grayish white. Its two projecting wings are ornamented with a dramatic saw tooth pattern that extends all the way from the base to the curved tip. The body is marked by a stepped pattern resembling pleats composed of units of regular size. The two zones are separated by a belt that is defined by two parallel applied bands, both circular in cross section. Within these raised borders, incised lines are linked to the saw teeth of the wings. The raised spine, which is elliptical in cross section, extends upward from above the base before turning sharply to become an extended tip. The smooth, flat surface between the two projecting wings is adorned with a decorative medallion designed as an eight-petal lotus. The motif measures 12.8 cm in diameter and the tips of the petals are slightly raised above the surface. The center of the blossom contains seven lotus seeds (Figures 8, 9).
Figure 8. 
Figure 9.
Although fragments of several chimi have been excavated at the temple site of Mireuksa in Iksan, only one has been restored to show its original form. Like the example from the Wangheungsa temple site, which will be discussed below, the overall form is divided into two separate units that meet at a sloping join. The restored height measures around 99 cm and the overall color is gray or grayish white. Lacking the dramatic openwork of the wings on the chimi from the temple site at Mount Buso, the vertical edges of the wings here have a stepped pattern composed of units of roughly similar size extending from the base to the tip across the join of the two units. The end of each simulated fold is marked by at least one pierced hole. While the lower portion of the body is adorned with seven or eight stepped “pleats,” the upper portion is left undecorated. The boundary between body and wings is not particularly pronounced but is indicated by a separate curving band containing stepped units with different spacing. A raised spine curves up from above the base to the tip. Unfortunately, damage near the tip makes it difficult to know its original form in detail. The smooth, flat surface between the two projecting wings is unadorned. A circular hole in the vertical wall directly below the raised spine must have been used to secure the chimi to the building structure at the main roof ridge (Figure 10).
The final Baekje chimi to be discussed here were unearthed at the Wangheungsa temple site in Buyeo and represent reconstructions of the upper and lower portions of two different examples. The lower unit was unearthed as fragments at the northern end of a building site on the east side of the complex while the upper unit was reconstructed from pieces found at the southern end of the same area. Considering the overall form, decorative motifs, and character of the clay, these pieces are thought to have served as the upper and lower portions of a pair of chimi marking the ends of the main roof ridge. When placed one above the other, the restored height measures around 123 cm, which is much taller than those unearthed at Mount Buso and Mireuksa addressed above (Figure 11). The color of the clay ranges from gray to reddish brown.
The wings are marked by a stepped pattern composed of units of roughly similar size extending from the base to the tip across the join of the upper and lower sections, not unlike the chimi from the Mireuksa site, but the end of every other simulated fold is punctuated with two small holes.11 The pattern on the body, which also continues across the two halves, features roughly regular simulated pleats superimposed at the center with a raised cloud motif constructed from arcs and spirals. The body and wings are separated by a belt that is defined by two parallel applied bands both circular in cross section. A total of six decorative medallions with small eight-petal lotus designs are placed at regular intervals within the belt on both sides of the object. Incised floral motifs fill the space between the lotus medallions (Figure 12). The raised spine rises upward from above the base before turning sharply to become a short tip. The smooth, flat surface between the two projecting wings is adorned with an incised vegetal design flanking an applied eight-petal lotus medallion (Figure 13). As the lotus design on the chimi resembles those found on roof tile ends produced when the temple was first established, this chimi must have been produced and used around the same time (Figure 14).
Figure 11. 
Figure 12.
Figure 13. 
Figure 14.
These Baekje works reveal several features that can be compared with aspects of the chimi associated with other kingdoms. First, the examples unearthed at the temple sites of Mount Buso and Wangheungsa exhibit applied lotus medallions on the smooth, flat surface between their projecting wings. A related motif is used differently on the chimi found at the Hwangnyongsa temple site in Gyeongju, illustrating an important link between the chimi of Baekje and Silla (Figure 15).
Figure 15. 
Figure 16.
Second, the chimi unearthed from the Mireuksa and Wangheungsa temple sites are constructed from upper and lower halves. This is also true of chimi found in the ancient Silla capital of Gyeongju. Representative examples include chimi excavated at the Hwangnyongsa temple site, the Cheongun-dong 千軍洞 temple site, and the Wolji 月池 palace site. The chimi reconstructed from fragments discovered at the Hwangnyongsa temple site is clearly composed of the upper and lower units, too. While those excavated from the Cheongun-dong and Wolji sites are single pieces, it is believed that they, too, were designed to have two separate elements considering the surviving portions lack features that would allow them to be anchored to the main roof ridge.
It is also possible to note some details shared by the chimi excavated from the temple site at Mount Buso and the one associated with Goguryeo. Although it is not as sharply pointed as the Goguryeo chimi from the Anhak palace site, both examples have tips that are circular in cross section and protrude beyond the wings. Despite these similarities, however, the design of the wings, the decorative treatment, and other features are quite different.
Chimi of the Silla Kingdom
There are no written records indicating when chimi were first used in the Silla kingdom. Although several have been discovered in Gyeongju, location of the ancient Silla capital, none are dated. Based upon style, however, most scholars believe that the one from the temple site of Hwangnyongsa dates to the Three Kingdoms period, thus making it the earliest surviving Silla example. All remaining chimi—including those excavated from the temple sites of Sacheonwangsa 四天王寺, Inwang-dong 仁旺洞, Bunhwangsa 芬皇寺, and Cheongun-dong as well as the palace site of Wolji—are regarded as later examples postdating the Silla unification of the peninsula in the late seventh century.
During an archaeological survey in 1978, the shattered fragments of the Hwangnyongsa chimi were found in the northeastern sector of the remains of the lecture hall site.12 The fragments were used to reconstruct the original form, which measures around 182 cm in height, making it the largest Silla example known (Figure 17). The dominant color is close to gray but there are also reddish-brown areas. The restored configuration is composed of two units that are nearly equal in height. The wings are neatly decorated with a stepped pattern composed of units of roughly similar size extending from the base to the tip across the join of the upper and lower halves. Although the front section of the body is undecorated, small holes, six on each side, punctuate the top edge of the lower portion and the bottom edge of the upper portion; they may have been used to join the two parts.
This chimi has an unusual, notched base, a feature yet to be found on any other example excavated in Korea. A vertical, arcing belt is outlined by pairs of applied strips of clay that are rectangular in cross section. Between these borders, horizontal bands also consisting of two applied strips of clay divide the space into seven sections, three on the lower half and four on the upper. Although the bottom and top units of the series are undecorated, the others bear alternating designs of a lotus blossom and a human face. Unlike the others discussed so far, this chimi lacks a raised spine, and the wings extend all the way to the front. On the back, between the wings, strips of clay identical to those on the two sides create an irregular frame that is subdivided into quarters. A total of fourteen lotus and human face motifs are arranged in an orderly manner within and around the framed space (Figure 18).
Figure 17. 
Figure 18.
Although the Hwangnyongsa chimi is the only extant Silla example dating to the Three Kingdoms period, some of its characteristics help to reveal the origins and implications of Silla designs. The notched shape of the lower portion must have been crucial for installing and securing the large, heavy form to the wooden structure of the main roof ridge or the stacked roof tiles that protected it. Since it is not found on examples dating to the Unified Silla period, it may be a distinctive characteristic unique to this particular work or its period. The subdivision of the form into upper and lower sections was already observed among the Baekje chimi from the temple sites of Wangheungsa and Mireuksa; as with those examples, it may have been done to ease firing or installation.13 The design and location of the surface decoration compare favorably with Baekje chimi from Wangheungsa and Mount Buso. In particular, the composition of the lotus motifs and their placement on the sides and back of the Hwangnyongsa and Wangheungsa chimi is quite similar. These characteristics suggest that the Hwangnyongsa chimi may have evolved from Baekje models. Nevertheless, the distinctive shape of the lower portion of the body and human face motifs can be regarded as characteristics unique to Silla (Figure 19). Notably, simple, unrefined representations of the human face may be related to the humble aesthetics observed among Silla clay figurines. (Figure 20).
Figure 19. 
Figure 20.
Since it was found with discarded roof tiles and other objects of various dates, it is difficult to specify the date and original location of the Hwangnyongsa chimi. Such a large example must not have been installed on a minor building but more likely used for the lecture or main hall, which were not far from the excavation site of the fragments. Two dates for the object have been proposed. A written record stating that a new image hall was constructed in the sixth year of King Jinpyeong 眞平 (584, r. 579–632) has been cited by researchers arguing for a date in the late sixth century.14 More recently, however, a stylistic comparison with the lotus designs found on roof-end tiles suggests a mid-seventh century dating.15 Since the shape and design of the Hwangnyongsa chimi are quite sophisticated, it is plausible that earlier Silla examples were made. It is hoped that further excavations at sites like the Wolseong 月城 palace or early Buddhist temple sites of the Three Kingdoms period will provide these antecedents.
