Ghanaian interweaving in the nineteenth century: a new perspective on Ewe and Asante textile history.

Kente cloth is widely known in West Africa and beyond, especially in the United States. The name refers today to handwoven textiles, often made of rayon with bright color contrasts, which are composed of narrow strips sewn together edge to edge (Fig. 1). Most of the production of kente is currently concentrated in several villages around Kumasi (the Asante region) in the Twi-speaking area of Ghana; the Agotime area and coastal villages along the Keta Lagoon, both Ewe-speaking areas in Ghana and Togo; and in all major towns in this region. Many weaving workshops can also be found in Accra and Lome, and Ewe weavers, with their long tradition of migration, have settled in other major West African cities such as Lagos (Kraamer 2005a:72, 146-7, Clarke 1999:68-9, Klein 1998:37). (1)

[FIGURE 1 OMITTED]

The early history of kente cloth has received some scholarly attention (Lamb 1975:103-121; McLeod 1981:153-5; Ross 1998:78, 152), with a focus on textiles woven around Kumasi (Asante cloth). At the end of the 1990s, the history of these textiles was also debated in public discourse in Ghana, mainly in ethnic terms. The Ghanaian media focused especially on the supposed Ewe or Asante origin of kente (Kraamer 2005a:110-11). (2) In this article, I want to add to these discussions a new perspective on the nineteenth century histories of Ewe textiles in relation of Asante cloth, based on a combination of sources that have so far not been often studied. (3) It is possible to trace changes in precolonial African art traditions, even when extant objects from a specific period are limited in number. It is important to investigate this process of change, for although much of the literature in the last few decades has demonstrated over and over again that change lay at the core of many art traditions in Africa, African objects--especially pre-twentieth century works of art--are still often presented ahistorically (cf. Vansina 1984, Picton 1992, Ogbechie 2005:63).

One of the main characteristics of Ewe and Asante textiles is the alternation of weft- and warp-faced plain weave areas in one length of strip (Fig. 2). The use of two pairs of heddles, which enables this specific characteristic and is therefore crucial in the formal developments of both Ewe and Asante textiles, most likely, I argue, originated in the Agotime area. This technique spread to other weaving centers sometime in the eighteenth or early nineteenth century and was fully taken up by Asante and Ewe weavers at least by the middle of the nineteenth century, but probably earlier. The existence of different weaving centers in the Ewe-speaking area, along with the production of different types of textiles at certain points in time, complicates the history of Ewe textiles, in contrast to the more linear development of Asante textiles suggested by other writers (Lamb 1975:103-21; McLeod 1981:153-6; Ross 1998:78, 152). One indication of the several separate, though often interrelated paths at different weaving centers can be found in the current local linguistic terminology for classifying textiles in these places.

Techniques and Designs

In the Ewe-speaking area, the range of techniques and designs was already wide at the end of the nineteenth century, even though warp-faced plain-weave textiles, both from hand- and machine-spun cotton, outnumbered all other types (Figs. 3-4). Weavers from different centers east of the Volta river--including Agotime and the coastal area, the (Twi-speaking) Akwamu area around Anum, and the Peki region and further inland--produced a variety of textile types: warp-faced plain-weave textiles with warp-stripe patterns; weft-faced plain-weave textiles (Fig. 5); balanced plain-weave textiles with block patterns formed by different colored wefts (Fig. 6); supplementary weft-float figurative and nonfigurative motifs on a plain ground and framed by weft-faced blocks (Figs. 7-10); and weft-faced bands on a mainly warp-faced textile (Fig. 11). (4) It is unlikely for all of these to have been woven in any one place.

[FIGURES 3-11 OMITTED]

Until the first half of the nineteenth century, the range of techniques and designs woven in the Asante area is less clear. As there was only one weaving center, in and around Kumase, the variation was probably less than in the Ewe-speaking area. Asante weaving evolved from a warp-faced plain-weave tradition; there is no evidence of weft-faced plain-weave textiles. Asante weavers used red materials unravelled from other textiles (Romer 1965 [1760]:36; Bowdich 1966 [1819] :35), which they probably also used to weave balanced plain-weave textiles with weft- block and weft-stripe patterns formed by different colored wefts (Fig. 6). Some writers have suggested that at the beginning of the nineteenth century, Asante weavers could have used some supplementary wefts to make weft-faced bands or even weft-float motifs (Lamb 1975:95-5, Ross 1998:152), but there is no hard evidence for this theory except that one would expect it would take a tradition more than a few decades to develop the complicated cloth with weft-float motifs and weft blocks (hereafter called a warp-faced textile with weft blocks) that appear in Basel Mission photographs from the 1880s (see BMA QD--30.044.0041 and D-30.18.066; Kraamer 2005a:175-6, 534). (5)

Although Asante and Ewe textiles seem to have shared many characteristics in the nineteenth century, the weaving of supplementary weft-float figurative motifs and weft-faced plain-weave textiles and the use of plied yarn (i.e. the twisting of two or more colors in the weft or warp) were features particular to the Ewe area (Figs. 5, 7, and 11; Kraamer 2005b:299). From the end of the nineteenth or beginning of the twentieth century, elaborate Asante textiles were increasingly woven completely from silk or rayon, whereas Ewe textiles generally remained cotton; this is one of the reasons why the color arrangements in textiles that predate the mid-twentieth century are often so different from each other (Figs. 12-13).

[FIGURES 12-13 OMITTED]

Ross suggests that Ewe textiles followed the same path as Asante cloth before the twentieth century (Ross 1998:78), and Lamb implies that the early steps were the same but does not provide a detailed description of formal developments in Ewe textiles (Lamb 1975:207-15). They both propose that the history of textiles from southern Ghana developed from warp-faced plain-weave and the use of warp-striping into simple weft patterns produced with one pair of heddles. The use of solid color weft-faced bands, called bankuo, had started before weavers began unravelling foreign textiles to obtain silk (Ross 1998:78; Lamb 1975:110). (6) This unravelled, mostly red, material was initially selectively used for single color weft-faced designs. Only when foreign yarns became available in larger quantities did weavers begin to employ them for the warp. The first weft patterns would have been solid-color weft-faced designs, followed by solid blocks of weft-striped bands, called babadua in Twi, ampa among Agotime weavers, and novi among coastal Ewe weavers (Kraamer 2005a:98).

The last step would have been the introduction of patterns of overshot supplementary floating-weft in between these blocks, facilitated by the introduction of a second pair of heddles (Ross 1998:78, 152). Such a weft-float design is called adwen in Twi, and adanu or atiforfore in Ewe (Kraamer 2005a:160). Lamb and Ross suggest that this evolution took place long before the mid-nineteenth century, based on a regressive extrapolation from twentieth century textiles, some Basel Mission photographs, nineteenth century cloth and travel accounts, and linguistic terminology for textile parts in Twi (Ross 1998:78, 151-3; Lamb 1975:103-16). They assume, correctly, a development from simple to complicated techniques starting from warp-patterned textiles, because the oldest textiles from the Twi-speaking area are warp-faced plainweave and all textiles, up to today, have generally taken their names from the warp patterning--except for asasia cloth, which is woven with three pairs of heddles (Lamb 1975:110-21, 125-6; Ross 1998:78). (7) Though this seems a probable evolution for Asante textiles, albeit possibly over a shorter period of time than Lamb and Ross propose, Ewe cloth may have taken different paths. The use of two pairs of heddles and the subsequent framing of weft-float motifs between weft-faced blocks may have been a nineteenth century development.

Travel Accounts, Missionary Reports, Photographs, and Extant Cloths

Spinning and weaving were well established throughout the Ewe-speaking area by at least the eighteenth century, and weaving remained widespread throughout the nineteenth century (Kraamer 2005a:63-7, 165, 173-4). As cotton is a savannah plant, its appearance in the Ewe region is to be expected (in contrast to the forested Asante region). Yams, especially red ones, from unravelled European textiles have been used for the most expensive textiles from at least the eighteenth century (Isert 1992 [1788]:91-2). In the nineteenth century, coastal and Agotime weavers shifted somewhat to imported, machine-spun cotton (Spieth 1906:408; Kraamer 2005a:78-80), though silk seems also to have been used (BMA D-1.10 Afrika 1859, Nr. 6/III; Spieth 1906:230). In the Twi-speaking area, a strong preference for silk yarns developed at an early stage continuing until today, when many more-expensive textile types are made from silk or rayon.

The oldest references to weaving industries on the coast east of the Volta date, to my knowledge, from 1703 (van Dantzig 1978:95) and 1717 (NLNA:WIC 104), but these Dutch sources provide hardly any indication of the types of cloth woven at that time (Kraamer 2005a:161-3). Isert gives the first written account of a weaver in action in 1785. The loom with one pair of heddles that he describes seems to operate in the same way as current looms in this region, but it is unclear whether it was used to make warp-faced or weft-faced plain-weave textiles (Isert 1992 [1785]:91-2).

Only a few nineteenth century textiles thought to be from the Ewe-speaking area and even fewer from the Twi-speaking region can be found in museum collections: twelve in the National Museum of Denmark, Copenhagen (NMD), four in the Museum of Cultures, Basel (MKB), one in the British Museum (BM), and six in the World Museum Liverpool (WML). (8) The oldest is in the MKB and was presented to missionary Andreas Riis in Akropong ("Akuropon") in 1840 (Fig. 8). The only other mid-nineteenth century textiles are part of the Carstensen collection in the NMD, including three textiles that are pretty definitely from the Ewe-speaking area (Figs. 9, 15-6). (9) It is often not possible to determine whether the textiles in the MKB and NMD collections are Asante or Ewe. The textiles in the other museums are from the Ewe-speaking area (Ross 1998:153-7; Kraamer 2005a:168-72, 488-513). The oldest extant textile with an alternation of warp- and weft-faced plain-weave areas and supplementary weft-float designs is MKB 23.336 (see Ross 1998:156, fig. 10.10). This textile was collected from Kumase before 1888 and was thought to have been woven in Asante (Ross 1998:154), but could also have been produced in the Ewe-speaking area. (10) The extant nineteenth century textiles from the Ewe-speaking area demonstrate the weaving of weft-faced and warp-faced plain-weave textiles and therefore the use of two pairs of heddles and the weaving of supplementary weft-float motifs since at least the 1840s.

[FIGURE 15 OMITTED]

Basel and Bremen missionary photographs from the 1860s onwards sometimes provide clear evidence of techniques and designs, though many other photographs are variously interpretable. Most Bremen missionary photographs were taken in the Ewe-speaking area, but date and location were not always documented. Basel missionary photographs were mainly taken in the Krobo, Akyem, and Akyapim areas (just east of the Volta river), though some were also taken in the Asante area. Much more work has been done on their dating and location than on the Bremen photographs. Most cloths in these photographs are warp-faced plain-weave textiles. Textiles with an alternation of warp-faced and weft-faced plain-weave areas appear from the 1870s and with supplementary weft-float designs from the 1880s (Kraamer 2005a:174-8). All these photographs indicate that weaving with two pairs of heddles was firmly established by at least the end of the nineteenth century, probably in both the Ewe and Asante regions. Some photographs of weavers at work indicate that Ewe weavers (or those trained by Ewe weavers) were working outside the Ewe-speaking area. Of special interest is a photograph of two weavers at work in Abetifi, one at a six-pole Ewe loom, the other at a four-pole loom that resembles current Asante looms: Both use one pair of heddles (Fig. 16). This photograph is the only hard evidence of contact between weavers who were probably trained in different traditions. Weavers are, however, visually trained people with strong numbers skills. They do not need to have physical contact with weavers from other areas to adopt features of different cloth traditions; they only need to have access to textiles from other areas (Kraamer 2005a:147, 228-9, 263, 266-7).

[FIGURE 16 OMITTED]

Most nineteenth century written sources on Ewe textiles comment on flourishing weaving centers, including Peki and Agotime, and the wide trade in textiles (especially with Akyem, Accra, and Akyapim), but do not provide descriptions of cloth (see Kraamer 2005a:173-4). It is therefore likely that people from the Krobo, Akuapim, and Akyem areas, the people most photographed by Basel missionaries in the Gold Coast, used cloth from the Ewe-speaking area because of trade and proximity to the Ewe-speaking region. At least according to Austin Freeman, the weaving of elaborate Asante textiles was very limited in the 1880s (Freeman 1898:382), which suggests that these textiles were used just for the court and not for trade. The few descriptions of textiles come from Bremen and Basel missionary reports. They mainly refer to the narrow strip format (e.g. BMA D-1.4a Afrika 1851-1853, Nr. 14) and often include vague references such as "a piece of multicolored striped or checkered cotton gear" (Menge, MB 1852:78-9, my translation from the German) or "the finest woven fabrics on simple construction of sticks and cords" (Schlegel, MB 1858:378, my translation from the German).

Although influences on the formal qualities of textiles at different weaving centers in the Ewe- and Twi-speaking areas seem likely based on design, technology, and geographic proximity, there is insufficient evidence of actual contact. These influences are, however, supported by different sources, keeping in mind that weavers do not necessarily need physical contact with other weavers to be able to reproduce a sample. Apart from the documented trade in these textiles (Kraamer 2005a:161-2, 169, 173-4), the details of weaving processes and different counting systems for laying the warp also give indications of influences. Asante and Ewe weavers work at the loom with the wrong face uppermost, in contrast to Yoruba tradition where the right face is uppermost. These shared characteristics are only necessary for the weaving of supplementary weft-float motifs and thus indicate a common origin for Ewe and Asante weaving from at least this time. Counting systems (especially in relation to the threading of warp elements through the two pairs of heddles) for cotton textiles made in Asante, Agotime, and the coastal area were historically the same (Kraamer 2005a:92-3), also suggesting that Asante and Ewe weavers did not independently develop the use of two pairs of heddles. The counting system in the coastal area for warp-faced plain-weave textiles, which are not found in Agotime or Asante, might point to their independent development there or it may indicate a common origin for warp-faced weaving and other textile traditions, perhaps to the east. The current practices and linguistic evidence for loom construction and the historical manner of laying of the warp in the Ewe-speaking region share more with Yoruba than with Asante practices. This could point to a common origin or partly shared history of (some parts of) Ewe and Yoruba weaving, but this remains speculative. (11)

Agotime and the Use of Two Pairs of Heddles

Ewe textiles probably did not simply develop from warp-faced plain-weave textiles with warp patterns as was the case with Asante textiles; weft-faced plain-weave textiles have also a long tradition alongside warp-faced plain-weave cloth in the Ewe-speaking area (see Lamb 1975:180-81, Hiamey 1981:64). In local recollection, titriku, the common Ewe name for weft-faced plain-weave textiles, is the oldest textile type--or one of the oldest, depending on the weaving area (Fig. 17). The oldest extant weft-faced plain-weave textile, now in the British Museum, is said to be from 1865 (Fig. 5) (12) and the oldest possible mention of such a textile dates from the early 1850s (MB 1852:78-9; see Kraamer 2005a:172-3 for possible interpretations of this description of Peki textiles by the Bremen missionary Menge).

[FIGURE 17 OMITTED]

In the Ewe-speaking region, as in the Asante area, there are many terms for variant weave structures, and a textile takes its name from more than just the warp pattern (Ross 1998:78, 108), suggesting diversification over a longer time period (Kraamer 2005a:108-11,127-8,149-52). This diversification was probably at least partly the result of the wide spread of weaving in the whole Ewe-speaking area, particularly in the areas around Peki, Agotime, and the upper part of the Keta lagoon (Some and Anlo) in the nineteenth century (Kraamer 2005a:63-7).

A crucial development in both Ewe and Asante textiles was the alternation of weft- and warp-faced plain-weave areas in one length of strip through the use of two pairs of heddles. Evidence of the use of two pairs of heddles exists only from the mid-nineteenth century, in three of the textiles now in Copenhagen. One of these textiles is made of 28 textile strips with 148 figurative weft patterns and small weft bands (Fig. 7); the other two contain both entirely weft-faced strips and warp-faced plain weave strips. (13) Earlier written and visual descriptions of looms and textiles do not indicate the use of more than one pair of heddles. Both Isert and Bowdich make no mention of two pairs of heddles, when both had sufficient knowledge of textile technology to have noted them: Isert was the son of a German master weaver and Bowdich's comparison of the loom's operation to that of a European loom shows that he had a clear idea of how both types functioned (Bowdich 1966 [1819]:309-10). This suggests that the textiles were balanced plain-weave or supplementary handpicked weft-float bands or motifs. (14) Balanced plain-weave textiles were being woven at that time in the Gold Coast, as one such textile, now in the Museum of Cultures, Basel, was collected in 1840 (Fig. 6).

Bowdich writes about the textiles of Asante chiefs and the Asante royal family woven "from the costly foreign silks which had been unravelled to weave them in all the varieties of color, as well as pattern" (Bowdich 1966 [1819]:35, 330-43). Together with his drawings of a loom with one pair of heddles and a flute player wearing a warp-striped cloth with some weft-bands (Fig. 18; see also Bowdich 1966 [1819]) no. 3, in Ross 1998:152, fig. 10.5), this has been interpreted to indicate the weaving of supplementary weft-faced bands and even supplementary weft-float motifs (Lamb 1975:94-5; Ross 1998:152). There are other possible interpretations for the patterning he describes and illustrates, however. One possibility is balanced plain-weave cloth--just warp striping and weft bands--or even weft motifs in what the Ewe call dahume (Fig. 19). (15) Another would be a combination of warp stripes and handpicked wefts in weft-faced areas; weft-faced plain-weave for weft-bands; and perhaps supplementary wefts for weft-float patterns. Only one pair of heddles, possibly supplemented with a weaving sword, is needed for either possibility. A third possibility is that two pairs of heddles were used to make weft-faced bands and perhaps supplementary weft-float motifs, but the complications of illustrating such a loom led to one pair of heddles being left out of the picture. The date when two pairs of heddles were used simultaneously is thus unknown, but it is unlikely to be later than the first half of the nineteenth century.

[FIGURE 18 OMITTED]

There is also no hard evidence as to its place of introduction, but I would suggest it was in one of the weaving centers of the Ewe-speaking area where weavers were already familiar with both types of heddles, i.e., where both weft-faced and warp-faced plain-weave textiles were being woven. Linguistic terminology for heddles suggests that the second pair of heddles was positioned after the first (seen from the position of the weaver) in the Asante area, but not necessarily in the Ewe region. The word for 'heddle' is eno in Ewe and asatia in Twi. According to Christaller's 1881 dictionary, asatia is the combination of asa or asa dua, 'loom', with tia, 'to set a foot on, or to step in', thus 'loom part to set foot on' (Christaller 1933 [1881]:512). Christaller translates the word ntiamu (or ntiam') as 'treadle(s), e.g. of a loom' (Christaller 1933 [1881]:512) but does not give any translation for the name of the second pair of heddles, asanan (used to weave weft-faced plain-weave textiles or weft-float motifs). Asanan seems to be a contraction of asa and enan, 'foot" or 'hind-foot', thus referring to the heddle further from the weaver (Christaller 1933 [1881]:328). The word for 'foot', nsa, is not used to form a word for the first pair of heddles. Nan could also come from enan, 'four', a derivation given by Rattray. The etymologies he was given for other parts of the loom are slightly different from the ones Christaller gives: asa, 'to dance'; tia, 'short' or 'little'; nan, 'four', from the fact that an oba (four threads) or groups and fractions of an oba are generally passed through them (Rattray 1927:224, fn 2). This still indicates that asanan was introduced later than asatia.

In the Ewe-speaking region, the heddles are distinguished by calling the closer pair of heddles noga, 'big heddle' and the further pair novi, 'small heddle', which does not refer to the fact that one is used behind the other, but to the difference in the number of loops in the first and second pairs. The construction of both heddles is basically the same, but the loops in one are set further apart to facilitate a weft-faced plain weave. The weaving of cloth with alternating strips of entirely warp- or weft-faced plain weave was likely to have been a preliminary step toward the use of two pairs of heddles. (16) A textile in the Carstensen collection accessioned in 1844 is an example of this presumed intermediate step (Fig. 15). It was probably woven in the Ewe-speaking area because of the use of an entirely weft-faced plain-weave strip; furthermore, such a textile appears in a photograph taken in this area in the 1860s (Fig. 20).

[FIGURE 20 OMITTED]

This also provides evidence for the use of the two different pairs of heddles in at least one location in the Ewe-speaking area; (17) the available evidence suggests that Agotime was probably where this first happened. There the oldest type of textile is said to be titriku--used at commemorative events--which is weft-faced plain weave and woven with the novi. A variety of titriku textiles are woven in Agotime, while only one is produced in the coastal area, suggesting a longer tradition for this type in Agotime (Kraamer 2005a:113-16). Coastal textile terminology suggests that the basic type of weaving there was warp striped. Textiles with alternating weft-faced and warp-faced plain-weave areas in one strip are called novivor; the weft blocks are called novi, which is the also name of the second pair of heddles used to weave the weft-faced plain-weave structure. This implies that in the coastal area the novi heddles were introduced after the noga heddles. (18) In addition, the term for warp-faced plain-weave textiles, vutsatsa, 'shuttle-mat', suggests that this was the first type of cloth in the coastal area, evolving directly from mat-weaving. (19) In Agotime, there is no such proliferation of terminology. (20) This suggests that the use of weft-faced plain weave is a more recent development at the coast than in Agotime. (21)

It is possible that weft-faced plain-weave textiles were also woven alongside warp-faced plain-weave textiles in the Peki area at least from the middle of the nineteenth century. The remark by the Bremen missionary Menge on cloth production in Peki could refer to both types or to warp-faced and balanced plain-weave textiles (MB 1852:78-9; see also note 20). There is, however, no other archival or fieldwork evidence for weft-faced plain-weave textiles in the Peki area. (22) The nineteenth century evidence--consisting of travel accounts, extant cloth, missionary reports, and photographs--suggests that most weaving centers in the Ewe-speaking area were working in a warp-faced plain-weave tradition (Kraamer 2005a:168-78).

Arrangements of Textile Parts

Two other characteristics shared by many twentieth century Ewe and Asante textiles warrant historical investigation: the formats used and the weaving of supplementary weft-float motifs. In the Ewe-speaking region, especially in Agotime, weavers make a distinction between two formats: atisue and atitrala. In the atisue format a weft-faced block is alternated with a warp-faced block (Figs. 13 and 21); in the atitrala format a warp-faced plain-weave block is alternated with an elongated form grouping two or three weft blocks together (Fig. 2). Even though the use of two pairs of heddles seems to have begun in Agotime, it remains difficult to determine where and when these formats started and whether the use of a supplementary weft float had started before the introduction of the use of two pairs of heddles in one length of strip. (23) Balanced plain-weave textiles, which go hand in hand with the weaving of warp-faced plain-weave textiles, (24) might have played a role in the development of these formats. Supplementary weft-float motifs could have been woven without the second pair of heddles, though this second pair facilitated their production and therefore probably played a role in the weaving of many weft-float motifs.

[FIGURE 21 OMITTED]

Although hard evidence is lacking, it is possible to hypothesize the course of development of these formats and of the use of supplementary weft floats. The atitrala and atisue formats could have developed independently, or one format could have grown from the other: The atisue format is simpler than the atitrala, which could indicate that atitrala developed from the atisue, but the atisue uses more material which could indicate that the atisue format grew out of the atitrala, as the cost of materials has counted to a large degree for the cost of the entire textile. (The growing use of wefts in the course of the nineteenth century is therefore also attributable to the increasing availability of lower-cost machine-spun yarn.) Both formats may have developed out of experimentation with two pairs of heddles, while atisue may have developed under the influence of similar formats in weft-faced plain-weave textiles, and/or atitrala and atisue may have been influenced by textiles in balanced plain-weave.

There are five main pieces of evidence to consider for these different options. First, all extant nineteenth century textiles and dated photographs only display the atitrala format, suggesting that it was more woven than the atisue in (most) Ewe- and Twi-speaking areas (Figs. 9-10). Second, the atisue format appears to have a longer history in the Ewe than in the Asante region. Asante weavers only have a name for the atitrala format--susudua--but have no name for atisue. Susudua, 'measuring stick', refers to the format of two weft-faced blocks, each called babadua, framing a block with a supplementary weft-float design called adwen 'ornament or artifice'. Ewe weavers also use words that refer to the measuring stick but specify a difference between the two formats: atitrala means 'long stick', atisue 'short stick', (25) suggesting diversification over a longer period. Asante weavers make a distinction between textiles woven in two different techniques: topreko and faprenu. (26) Topreko, 'passed once', refers to a technique for weft-float designs that is also used in both atitrala and atisue formats by Ewe weavers (Fig. 13), but only in the atitrala (or susudua) format by Asante weavers. Asante weavers use another technique for weft-float designs; they call the resulting textiles faprenu, 'thrown twice', for textiles in the atisue format (although the format has no Asante name). Faprenu appears to be an early twentieth century development (Ross 1998:78) (27) and was copied by Ewe weavers in the 1950s (Kraamer 2005a:141-2).

Third, in Agotime, the atisue format was probably the first format woven after the introduction of two pairs of heddles. Weft-faced plain-weave textiles, titriku, are considered the oldest type and are still often woven to a width of about 5cm (2") instead of the usual 10cm (4") to achieve a chessboard pattern. One particular textile in the atisue format, trekeke, is in Agotime considered to be one of the oldest textile types and is often classified as part of the titriku family. It also is still woven with this smaller width of 5cm (Fig. 22). There are no existing textiles or nineteenth century photographs to support this theory; the first photograph showing a cloth in the atisue format was taken by a Bremen missionary at the end of the nineteenth or early in the twentieth century (Fig. 21). (28)

Fourth, one balanced plain-weave textile collected in Akropong in 1840 (now in the MKB) has the same aesthetic frame-work as the atitrala format, but is com posed of different colored weft blocks in the ground weft (Fig. 6). This is the oldest extant textile from the Gold Coast and could have been woven in the Ewe-or Twi-speaking area. Its ending strip, together with the wider trade in Ewe textiles and the existence of Ewe weavers in the Akuapim area, suggest an Ewe contribution. The cloth could, however, have been imported from Asante, as it could also fit into the evolution of Asante textiles and the Akuapim were in close contact with the Asante.

Fifth, one cloth in the Carstensen Collection (Fig. 14), which I believe--contrary to Ross (1998:153)--is more likely to be Ewe due to the incorporation of weft-faced plain-weave strips, has different balanced plain-weave strips: one resembling the atitrala format and including two weft-faced plain-weave blocks framing a supplementary weft-float design in the atitrala format; the other in the atisue format. In at least one Ewe-speaking production center, the use of a supplementary weft, the use of both formats, and the use of two pairs of heddles for plain-weave--each pair used independently for weft-faced and warp-faced plain weave and together in one length of strip for the alternation of these two weave structures--were known by the 1840s.

I would suggest that both formats developed independently from each other: In Agotime, the atisue format probably developed from the chessboard patterning of weft-faced plain-weave textiles (Figs. 5 and 17); in any other weaving center working in a warp-faced tradition in the Ewe-speaking area and/or Asante, the atitrala format could have been an attempt to reproduce a similar patterning of weft blocks in balanced plain-weave textiles in a textile with alternating weft- and warp-faced plain weave after the use of two pairs of heddles spread from Agotime (Figs. 6, 9, and 10). In Agotime, the weaving of the atisue format could have simply developed from experimentation with two pairs of heddles, but it was probably influenced by weft-faced plain-weave textiles, the titriku, which have a chessboard patterning in this area (a patterning common in many other West African weft-faced plain-weave traditions). (29) These textiles probably predated the use of two pairs of heddles because they are locally considered to be the oldest type of cloth woven in Agotime; the variety of chessboard patterning suggests the historical depth of these textiles; and the oldest existing Ewe weft-faced plain-weave cloth, from 1865, has such a patterning (Fig. 5). (30)

The lack of firm nineteenth century evidence of the atisue format in photographs or extant cloth could be because it was used only in Agotime. This would imply that although the use of two pairs of heddles spread from Agotime, this particular format did not spread--at least not as far as the Asante area, unless the atitrala format was also developed in Agotime. The fact that in Asante and perhaps several Ewe-speaking areas the use of two pairs of heddles seems to have been used initially just for the atitrala format could have been informed by an economical use of (costly) yarns, but I find it more likely that it was a response to the existence of a composition similar to the atitrala format in (almost) balanced plain-weave textiles, including vueve (Fig, 24), (31) those warp-faced plain-weave textiles woven with two shuttles. The existence of such a balanced plain-weave textile collected in 1840 supports this theory. (32) The development of this composition in balanced plain-weave textiles might have started in the coastal area, as weavers there made a wide variety of vueve and balanced plain-weave textiles, including dahume textiles (Fig. 19), which have discontinuous weft patterning. It also could have developed in Asante, where Rattray reported the weaving of dahume (1927:224), or in any other weaving center with a warp-faced tradition. (33)

[FIGURE 24 OMITTED]

The atitrala format could have been developed by Agotime weavers if they were familiar with balanced plain-weave textiles developed in other areas. (34) The use of two pairs of heddles would then have spread through or alongside the wider distribution of Agotime textiles woven in the atitrala format. The other possibility is only the use of two pairs of heddles spread from Agotime to other weaving areas. The two pairs of heddle were then used in these regions to weave textiles in the atitrala format based on balanced plain-weave textiles in a similar composition. Only later would this atitrala format have been added to the repertoire of Agotime weavers.

Weft-float Motifs

Even though the use of a supplementary weft to create weft-float motifs may have predated the use of two pairs of heddles--because both-faced weft-float designs could be produced with a weaving sword and one pair of heddles (the noga or asatia), a common practice in many West African warp-faced textile traditions--the widespread use and perhaps even the introduction of this supplementary weft probably postdated the introduction of two pairs of heddles, as the weaving of both-faced designs is facilitated by the use of the second pair. One of the two Carstensen textiles with weft-float designs also has weft-faced plain-weave areas (Fig. 14), suggesting the use of two pairs of heddles; the other has weft-faced bands which could have been woven without the second pair (Fig. 7). It is unclear which weaving center--Agotime, Peki, coastal region, or Asante--introduced the supplementary weft. (35)

If the use of a supplementary weft predated the use of two pairs of heddles, it might have begun at the time that unravelled red yarn was used. This may have happened in the Ewe-speaking area, leading to figurative designs on a plain warp-faced background, (36) and/or in the Asante region, leading initially to small weft bands and then to nonfigurative designs. The introduction of a supplementary weft probably occured independently from the introduction of two pairs of heddles--therefore, outside Agotime. The use of the second heddle pair for weaving a weft-float motif would have come from both the use of two pairs of heddles to make alternating weft- and warp-faced plain weave and the use of a supplementary weft float with a weaving sword, rather than that the use of a supplementary weft in Agotime led to the use of two pairs of heddles.

Zarma (or Djerma) textiles from Niger, Nigeria, and Burkina Faso may have influenced the use of a supplementary weft in the Ewe-speaking area, because one type of Zarma textile, identified as tera, (37) has supplementary figurative weft-float motifs. (38) The two oldest known tera were purchased in 1928 by the Newark Museum (see Roberts 1995:108, fig. 3) and by the British Museum in 1934 (see Picton and Mack 1989:107). (39) However, I would suggest instead that Ewe textiles influenced this type of Zarma cloth. There is no evidence of tera in the nineteenth century, and even though the weaving centers of Ewe and Zarma people seem remote from each other, Zarma traders moved between these areas from at least the end of the nineteenth century, when some of them also settled more permanently along the coast of Ghana and Togo, including the coastal area of the Volta region (Kraamer 2005a:185-6, 237-8). (40) If the introduction of a supplementary weft float postdated the use of two pairs of heddles, it may also have happened in one of the different Ewe and Asante weaving centers where both pairs were already in use.

Conclusions

I would suggest that, rather than a slow development, the combination of two pairs of heddles and the subsequent weaving of alternating weft-faced and warp-faced plain-weave areas was a revolutionary development in the late eighteenth or early nineteenth century that quickly spread through the Ewe- and Twi-speaking areas. The new invention developed in different ways, including blending with characteristic features of other areas. (41) In the evolution of Asante textiles, the use of unravelled material would then have first been used for the weft in balanced plain-weave textiles to form weft bands and perhaps already (elongated) chessboard patterns; the 1840 cloth in the Museum of Cultures, Basel could have originated from the Asante area (Fig. 6). It is possible that a supplementary weft was already being used before the mid nineteenth century, but there is lack of evidence. The production of more elaborate textiles including the use of silk was probably preserved for Asante nobility, but textiles from other areas could have been available. The use of two pairs of heddles and the alternation of weft- and warp-faced plain-weave in one strip might have been adopted from Ewe weavers sometime in the nineteenth century and rapidly developed in textiles in the particular Asante styles. A closer examination of nineteenth century sources on Asante textiles could provide more definite evidence for its evolution.

Examination of Basel missionary photographs suggests that this rapid dispersal could have been aided by both Ewe and Asante weavers who were weaving in areas between their main weaving centers: Akyem, Akuapim, Kwawu, Accra and Krobo. (42) Moreover, the possibility that textiles from Ewe weaving centers were also distributed in the Asante area should not be excluded, especially if we take into account that the more expensive and intricate range of Asante textiles was preserved for the Asante court, while Ewe textiles were available for anybody who could afford to buy them.

[This article was accepted for publication in May 2006.]

This article is a reworking of a chapter of my doctoral thesis. The research was made possible through funding from the Prins Bernhard Fund, Stichting Fonds Dr. C. van Tussenbroekfonds, and Wotro Reisfonds (The Netherlands); a postgraduate studentship of the British Academy; a postgraduate fellowship, a one-year language scholarship and postgraduate additional fieldwork award of SOAS (University of London); the Irvin Trust of the University of London; and a postgraduate fellowship of the Smithsonian Institution (United States). I would like to express my gratitude for these opportunities. I also wish to thank the institutions that have given me permission to reproduce objects and photographs from their collections.

Abbreviations

BM: British Museum

BMA: Basel Mission Archive

MB: Monatsblatt of the Norddeutsche Mission Gesellschaft (Bremen Mission Society)

MKB: Museum der Kulturen Basel (Museum of Cultures Basel)

NLNA: Nationaal Archief (National Archive of The Netherlands)

NMD: Nationalmuseet Danmark (National Museum of Denmark)

NMG: Norddeutsche Mission Gesellschaft (Bremen Mission Society)

RMV: Rijksmuseum voor Volkenkunde (National Museum of Ethnology, Leiden)

STAB: Staatsarchiv Bremen (State Archive Bremen)

WIC: West Indian Company

WML: World Museum Liverpool

Archival sources

BMA D-1.10 Afrika 1859, Nr. 6/III

BMA D-1.14a Afrika 1851-1853, Nr. 14

WIC 104

MB 1852:78-79

MB 1858:378

References cited

Amenumey, D.E.K. 1989. The Ewe Unification Movement: A Political History. Accra: Ghana University Press.

Bowdich, T.E. 1966 [1819]. A Mission of Cape Coast Castle to Ashantee. London: Cass.

Christaller, Rev. J.G. 1933 [1881]. Dictionary on the Asante and Fanti Language called Tshi (Twi). Basel: Basel Evangelical Missionary Society.

Clarke, Duncan. 1999. Aso Oke: the Evolving Tradition of Hand-Woven Textile Design among the Yoruba of South-Western Nigeria. PhD diss., School of Oriental and African Studies, University of London, London.

Collier, Katharine Alexandra. 2002. Ablode: Networks, Ideas, and Performance in Togoland Politics, 1950-2001. PhD diss. School of Historical Studies, Centre of West African Studies, University of Birmingham.

Dantzig, Albert van. 1978. The Dutch and the Guinea Coast, 1674-1742: A Collection of Documents from the General Archive at the Hague. Accra: GAAS.

Freeman, R. Austin. 1898. Travels and Life in Ashanti and Jaman. London: Constable.

Gabus, Jean. 1955. Au Sahara: les hommes et leurs outils. Neuchatel: A La Baconniere.

--. 1958. Au Sahara: arts et symboles. Neuchatel: A La Baconniere.

Hiamey, T.B. 1981. The Origin and Development of the Weaving Industry in Anlo-some Area of Ghana. Agbozume and Anlo-Afiadenyigba Weaving Industry: A Case Study. BA Thesis, College of Art, University of Science, and Technology, Kumase.

Isert, Paul Erdmann. 1992 [1788]. Letters on West Africa and the Slave Trade: Paul Erdmann Isert's Journey to Guinea and the Caribbean Islands in Columbia (1788). Trans. T.B.S.A. Winsnes. Union Academique Internationale. Fontes Historiae Africanae, serie varia 7. Oxford: Oxford University Press.

Kasfir, Sidney Littlefield. 1984. "One Tribe, One Style?: Paradigms in the Historiography of African Art." History in Africa 11:163-193.

Klein, Alex. 1998. Fishing without Formality: An Economic Anthropology of the Ewe on the Lagos-Badagry Seabeach, Nigeria. PhD diss., Department of Anthropology and Sociology, School of Oriental and African Studies, University of London, London.

Kraamer, Malika. 2005a. Colourful Changes. Two Hundred Years of Social and Design History in the Handwoven Textiles of the Ewe-speaking Regions of Ghana and Togo. PhD diss., School of Oriental and African Studies, University of London, London.

--. 2005b. "Contemporary Visual Arts in the Ewe-Speaking Area." In A Handbook of Eweland: The Ewes of Togo and Benin vol. 3, ed. B. Lawrence, pp. 293-310. Accra: Woeli Publishing Services.

Lamb, Venice. 1975. West African Weaving. London: Gerald Duckworth.

McLeod, Malcolm. 1981. The Asante. London: British Museum Publications.

Meyer, Birgit. 2002. "Christianity and the Ewe Nation: German Pietist Missionaries, Ewe Converts and the Politics of Culture." Journal of Religion in Africa 32 (2):167-99.

Nugent, Paul C. 1991. National Integration and the Vicissitudes of State Power in Ghana: The Political Integration of Likpe, a Border Community, 1945-1986. PhD diss. School of Oriental and African Studies, University of London.

Ogbechie, Sylvester Okwunodu. 2005. "The Historical Life of Objects." African Arts 38 (4):62-95.

Picton, John. 1991. "Africa and the Two Art Worlds." African Arts 24 (3):83-6.

--. 1992. "Tradition, Technology, and Lurex: Some Comments on Textile History and Design in West Africa." In History, Design, and Craft in West African Strip-Woven Cloth, pp. 13-52. Washington, D.C.: Smithsonian Institution.

Picton, John, and John Mack. 1989 [1979]. African Textiles. London: British Museum Press.

Rattray, R.S. 1927. Religion and Art in Ashanti. Oxford: Clarendon Press.

Roberts, Allen F. 1995. Animals in African Art: From the Familiar to the Marvelous. New York: Museum of African Art.

Romer, L.F, 1965 [1760]. The Coast of Guinea. Trans. K. Bertelsen. Legon: Institute of African Studies.

Ross, Doran H., ed. 1998. Wrapped in Pride. Ghanaian Kente and African American Heritage. UCLA Fowler Museum of Cultural History Textiles Series 2. Los Angeles: UCLA Fowler Museum of Cultural History.

Spieth, Jakob. 1906. Die Ewe-Stumme: Material zur Kunde des Ewe-Volkes in Deutch-Togo. Berlin: Dietrich Reimer (Ernst Volsen).

Sprigge, R.G.S. 1969. "Eweland's Adangbe: An Enquiry into an Oral Tradition." Transactions of the Historical Society of Ghana 10:87-128.

Vansina, J. 1984. Art History in Africa: An Introduction to Method. London: Longmans.

(1.) The use of the terms Ewe and Asante textiles or weavers is not unproblematic. As already demonstrated by Vanaina (1984:29-33) and Kasfir (1984:163-93), art traditions do not coincide neatly with ethnic and linguistic groups. Furthermore, the development of ethnic identities has its histories and changes over time. Ewe identity, for instance, only developed in the early twentieth century (Kraamer 2005a:60-63). The historical process of construction and dispersion of the notion of an Ewe identity has been widely discussed (e.g. Amenumey 1989, Nugent 1991, Collier 2002, and Meyer 2002). I will therefore refer to weavers and textiles from the "Ewe-speaking area," but also mention "Ewe weavers" and "Ewe textiles" for convenience. In the same way I will refer to weavers and textiles from the Twi-speaking area and Asante weavers and textiles. I will interchange these terms to remind the reader of the somewhat arbitrary nature of ethnically labeling people and art, the complex nature between art and ethnicity, the strong interrelationship between those textile traditions, and the cultural and linguistic background of weavers who produce these textiles.

(2.) I am currently preparing an article on this debate for Afrique: Archeologie & Arts.

(3.) Interlacing two sets of elements at right angles forms a textile. One set is called the warp, the other the weft. The process of weaving is made easier by mounting the warp in tension on a framework (the loom) and by introducing a mechanism (the shedding device; in the Ewe-speaking area a pair of heddles) to separate alternate warp elements so that the weft elements may pass through (Picton and Mack 1989:18, following Emery 1966). The loom used in this area is called the West African double-heddle loom or narrow-strip loom. "Double-heddle" indicates that both sets of warp elements are leashed to one or other of the heddles, and "narrow-strip" indicates the narrow web format of the strips in the textile. Ewe weavers make simple weaves (only one set each of warp and weft elements) and compound weaves (more than one set of either warps or wefts or both). Simple weaves can be described as balanced warp-faced or weft-faced plain weave. A balanced plain weave structure is one where the number of warps per centimeter is the same as the number of wefts (Fig. 8). However, this will change if the warps are packed together to make the weft invisible (warp-faced plain weave; Fig. 5-6); or if the warps are spread apart, thus allowing the wefts to hide the warps (weft-faced plain weave; Fig. 7). In the majority of warp-faced and weft-faced textiles the weft is continuous, i.e. worked across the full width of the cloth. An exception is the Ewe balanced-weave textile dahume, where the weft is not continuous but is worked back and forth across a limited area (Fig. 22). Ewe weavers also use compound weave structures to extend their design repertoires. The extra warp or weft is supplementary, In that it is not essential to the basic coherence of the textile. In the Ewe-speaking region, a supplementary weft is only used in the warp-faced areas of a textile. A pattern can only be created using a supplementary weft if it floats over two or more warp elements (Fig. 9-11). Ewe weavers make designs that are visible on one or both faces of the textile. Here these designs will be termed "one-faced" or "both-faced" patterns, following local usage. The use of a supplementary warp is common in the Ewe-speaking area at least since the early twentieth century, but rare in the rest of Africa. There is no nineteenth century evidence of this compound weave structure, though an eighteenth century African textile fragment with a supplementary warp exists in the Wisbech Museum (UK). Unfortunately it is unclear where in Africa this fabric was woven.

(4.) A reconstruction of the early history is not easy, due to the scattered nature of available written sources, the often undated or location-unspecified photographs of Basel and Bremen missionaries from the 1860s on, the small body of securely dated nineteenth century extant textiles from the Gold Coast in museum collections, and the uncertain origin of production centre of these fabrics. Other sources that I have used are the regressive extrapolation of existing techniques and designs from the early twentieth century back, evidence from naming and classifying textiles, and comparative evaluation with the formal developments in Asante textiles. Rather than discussing these sources and the way in which I used them for this reconstruction (which is a whole chapter in my PhD thesis and which will be the subject of a further article), I will focus in this article on the results of that analysis.

(5.) One such textile appears on a Basel missionary photograph possible taken in the 1870s (BMA QD-30.044.0041), but it is unclear whether the cloth was woven in the Ewe- or Twi-speaking areas (Kraamer 2005a:175, 533).

(6.) Menzel calls it bankoro (Menzel 1972, 3:ba-be). I have not come across an equivalent in Ewe, but textiles with this kind of weft-band do exist (for some examples, see Kraamer 2005a:438). All photos from the BMA may be viewed at www.bmpix.org.

(7.) They do not indicate when this third pair of heddles was introduced; there is no nineteenth century evidence for asasia textiles, but these textiles' strict use by the Asantehene and close kin could account for this.

(8.) Two strips in the National Museum of Ethnology, Leiden (RMV) are probably also from the Ewe-speaking area. One is a late-nineteenth century weft-faced plain weave strip (1248-4) assigned to the Asante, the other is a warp-faced plain-weave cloth with warp-stripes of twisted yarns (1318-1) assigned to an unidentified group the Bangi (see www.rmv.nl). I would be interested if other nineteenth century textiles from this region could be brought to my attention.

(9.) Eduard Carstensen, governor of the Danish possessions in the Gold Coast, collected twelve textiles between 1844 and 1850 (see Kraamer 2005a:168-70, 488-96 for a detailed discussion of the attribution of location of these textiles, in one instance, NMD Gd 1 in disagreement with Ross 1998:153).

(10.) Some of the weft-faced blocks were made from plied yarn of two colors, the oldest extant example of this technique. The location of collection favors an attribution to the Asante, but the extensive trade in textiles makes the identification of the production center on the basis of acquisition place insufficient as evidence. The use of plied yarn would favor an Ewe attribution, as there is no known example, to my knowledge, of a textile from the Twi-speaking area that uses plied yarn. Another feature pointing to an Ewe origin is the unequal spacing of the areas with weft-faced blocks and weft-float patterns in between one kind of warp background, a feature more common in the Ewe- than Twi-speaking area (based on the large body of textiles from the first half of the twentieth century).

(11.) Further study of the linguistic evidence of weaving, counting systems, and preparations of the warp of West African weaving traditions may provide answers to questions of origin and historical trajectories, but in the meantime the evidence suggests that what in due course became Yoruha, Ewe, and Asante weaving link into a system of trade and distribution at different points in place and/or in time.

(12.) This textile was apparently presented to the donator's father by the king of Dahomey around 1865. The format of the textile, alternating red and yellow blocks, which are transversed by two narrow blue stripes, of even lengths in and along each strip can still be found in Agotime. The width of the strips also suggest that it was woven in the Ewe-speaking area.

(13.) These are the only three textiles that are more-or-less definitely produced in the Ewe-speaking area, as the weaving of figurative weft-float motifs and entire weft-faced plain weave textiles are only characteristics of this region.

(14.) The only indication of weft-float motifs would be the interpretation of the word "pattern" as a discontinuous design. Even in that sense, it could indicate either supplementary weft-float motifs or the use of what the Ewe call the dahume.

(15.) Dahume textiles are more labor-intensive than warp-faced plain-weave cloth. They have different sorts of weft designs produced with the same pair of heddles as used for warp-faced plain weave textiles. By doubling the weft, hence reducing the visual predominance of the weft in the finished textile, relatively complex patterning can be made using wefts of different colors. Moreover, as a line begun in one color can be continued in another, it is possible to create patterns that are not limited to mere striping. It is unclear when dahume started being produced, but they are considered an older type of textile. They were woven abundantly before the 1950s, but their production then dwindled (Kraamer 2005a:121-2). Rattray is the only person who documented visually this technique for Asante weaving in the early decades of the twentieth century (Rattray 1927:236, 244), but there is no other evidence for its production at other times.

(16.) Another preliminary step could have been an Agotime response to weave weft-float motifs (if the use of a supplementary weft had been introduced before the use of two pairs of heddles in either the coastal or Asante areas). This response would then have led to the use of two pairs of heddles in one length of strip to facilitate the creation of the shed for inserting the supplementary weft and only in the second instance to be used to make weft-faced and warp-faced plain-weave areas in one length of strip. I find this to be technically less likely and there is no proof to support this hypothesis. It is unclear where a supplementary weft float was first introduced, and if its introduction pre- or post-dated the use of two pairs of heddles. I would rather suggest that the introduction of two pairs of heddles began in Agotime and the introduction of a supplementary weft in the coastal Ewe region or in Asante and that the mutual influence only took place when both techniques were already developed, or that the use of a supplementary weft post-dated the use of two pairs of heddles.

(17.) Although unlikely, the production of weft-faced plain-weave textiles could have developed after the Introduction of two pairs of heddles, which would make the original location of two pairs of heddles uncertain. As indicated before, there is no dated evidence before the mid-nineteenth century for this type of textile in the Ewe-speaking region. This suggestion would, however, go against local recollection and cannot be sustained by any available source.

(18.) If weft-faced plain-weave textiles had come before an alternation of weft-faced and warp-faced plain weave, one would expect that weft-faced plain-weave textiles would be called novivo or that all the different kinds of textiles and cloth parts woven with the novi would be termed differently (as is the case in Agotime). However, if warp-faced plain-weave textiles were made and the second pair of heddles, the novi, were introduced to produce alternating weft-faced and warp-faced plain-weave areas, then it follows logically that the weft-faced blocks are called after the weaving equipment that produce these blocks, the novi, and the whole textile after this new invention, noviva.

(19.) The term vutsatsa may date from a time when weaving, newly introduced, had to be distinguished from mat-making (another common coastal activity), thereby implying that the oldest form of weaving here is warp-faced plain-weave. Furthermore, in southern Togo, and formerly also in the coastal area, all hand-spun warp-faced plain-weave textiles were called lokpo, and in the coastal area one warp pattern, remembered as a very old one, is still called lokpo (Kraamer 2005a:117).

(20.) In Agotime, cloth with alternating weft- and warp-faced plain-weave is called ntsrim and adanuvc; the weft-faced plain-weave blocks are called ampa or afcdcli.

(21.) The particular origin of this weaving group (Dangme; see Sprigge 1969:87-128) could account for the fact that they brought a weft-faced plain-weave textile tradition Into a warp-faced plain-weave area. The evidence given by Spieth suggests that in Ho, next to Agotime, weavers only produced warp-faced plain-weave textiles (Spieth 1906:406, 408-10). Furthermore, the Agotime people seem to have encountered a group of people where they settled in the seventeenth century, perhaps one of the reasons why many of them speak Ewe, as reported in the nineteenth century (BMA D-1.10, Afrika 1859, Odumase 1859 Nr. 6/III; MB 1877:60). This group could well have been already producing warp-faced plain-weave textiles, like other groups in the area. However, the migration route of the Agotime is not well documented, and other Dangme groups do not have a (documented) weft-faced plain-weave textile tradition.

(22.) Lamb conducted fieldwork in 1975 in the Peki and surrounding areas (Lamb 1975:189-97), I conducted three group interviews in this area, with chiefs and elders in Kpando, March 2, 1999; with chiefs, elders, and others in Anfoega

Akukome, March 3, 1999; and with chiefs, the queen mother, elders, and others in Peki-Blengo, March 5, 1999).

(23.) The sources are few and scattered. Furthermore, the identification of the production centers of nineteenth century extant fabrics and textiles in Basel Mission photographs is complicated not only because most were collected or photographed in areas outside the Twi- and Ewe-speaking areas, but because trade in textiles, at least from the Ewe-speaking area, was recorded in several nineteenth century sources (Kraamer 2005a:173-4).

(24.) Balanced or almost-balanced plain-weave textiles are locally often classified as part of warp-faced plain-weave cloth, as they are woven with the same equipment and the there is only a small distinction between many warp-faced plain-weave textiles and balanced plain-weave fabrics (Kraamer 2005a:119-22).

(25.) These are the Agotime names, but coastal weavers make the same kind of division: atieve means 'two stick', atideke 'one stick'.

(26.) Frapenu uses much more material than topreko, and therefore is more expensive.

(27.) The format of faprenu may have been influenced by Ewe textiles at the beginning of the twentieth century, or by weft-faced plain-weave textiles with chessboard patterning from the north or from Agotime, or simply an internal invention.

(28.) Venice Lamb published a photograph from the collection of Mr. and Ms. Wilmarth, in which the chief in this image is shown in the same cloth accompanied by other people (Lamb 1975:181, fig. 266). One of them wears titriku. It is unclear where and when the photograph was taken, but the building in the image looks like a Bremen missionary structure.

(29.) John Picton has suggested a reaction to northern weft-faced blankets, khasa, which may have led to the experimental use of two pairs of heddles (Picton 1992:29). He stresses the speculative nature of this theory, pointing to the high desirability of the khasa in Asante ceremonies. This theory seems less likely to me, as these northern textiles traded into the Asante area were not those with a chessboard patterning but those in which similar weft-faced blocks are lined up.

(30.) This textile could have been woven in any of the Ewe-speaking weaving centers, as by that time the use of both pairs of heddles was already known in at least Agotime and the coastal area, though the chessboard patterning could suggest an Agotime origin (a patterning not used in coastal titriku of the twentieth century, see Fig. 26). As previously discussed (see note 19), in the coastal area, the weaving of weft-faced plain-weave textiles developed after the introduction of two pairs of heddles.

(31.) Another possibility could be the adoption of the atisue and atitraIa format in balanced plain-weave textiles as a cheaper imitation of the textiles woven with two pairs of heddles in these formats, as weft-faced plain-weave blocks used more material than such an imitative block in a balanced plain-weave strip, and the changing of pairs of heddles extends the production vueve, "two-shuttle" textiles, leading to the laying of warps that make an (almost) balanced plain-weave cloth, such as takpekpe le Anloga (see Fig. 28) and the 1840 textile now in the MKB (Fig. 8). The difficulty with this theory is that it would date the introduction of two pairs of heddles and the development of these textiles to well before 1840, the date of the oldest existing balanced plain-weave textile resembling the atitrala format, but there is no evidence. It is, of course, possible that the atitrala format was developed from a similar format in balanced plain-weave textiles in a particular area, and that the atitrala format was copied later in balanced-weave textiles as a cheaper imitation in other areas.

(32.) This is not to say that the use of two pairs of heddles in one length of strip necessarily occurred after the 1840s, as a certain kind of textile can be woven for a longer time.

(33.) Experimentation with chessboard variants in balanced plain-weave textiles through the use of different colored weft yarns could have been a response to attempts to reproduce the patterning of weft-faced textiles, from Agotime or the north, within a technology set up for a predominantly warp-faced fabric. There is however no evidence for this suggestion.

(34.) The atitrala format could have developed in Agotime without any influence from balanced plain-weave textiles, as a further experimentation with two pairs of heddles. This would cause however the same problem as discussed in note 32, and it then remains unsolved why the atitrala format would have spread to other weaving area but the atisue form would not have done so.

(35.) One could argue that it probably did not originate with coastal weavers, because their early twentieth century textiles with weft-float designs (akpedo) were less elaborate than the textiles from Agotime or Asante (Kraamer 2005a:133), but a simplification in design could also have been a local development. The same development happened in the second half of the twentieth century for a type of textile with designs visible mainly on one face of the cloth (see Fig. 18), invented in the 1930s in the coastal area (asidanuvo) (Kraamer 2005a:134-5).

(36.) A mid-nineteenth century NMD textile (see Fig. 9) and an early twentieth century LM cloth (see Fig. 10) could be part of such a tradition.

(37.) In the 1950s, J. Gabus identified such a cloth with figurative motifs as tera woven by nomadic Zarma weavers who live with the Hausa of Tahoua, 400km (249 miles) northeast of Niamey in Niger (1955:53, 1958:405 and inside covers).

(38.) Zarma weavers produce mostly balanced plain-weave textiles, which are interrupted at intervals by broad weft-faced stripes in various colors (see Fig. 29). This alternation of two simple weave structures is achieved by grouping the individual warp elements into units of four, by using two supplementary single-heddles, and by doubling the thickness of the weft. The same thick weft is also used in the warp-faced areas to form discontinuous overshot floating-weft patterns (Picton and Mack 1989:107). They therefore weave textiles with both-faced designs. Lamb and Picton have noted the similarities between Zarma cloth and Ewe textiles (Lamb 1975:210; Picton 1992:24-6).

(39.) This textile was collected in Obuasi, in the Central Region of Ghana, but probably woven by Zarma weavers in Dori, Burkina Faso (Picton and Mack 1989:107).

(40.) Together with the Hausa, they dominate the trade in Ewe cloth at the Agbozume textile market. Some live permanently in Agbozume-Zongo, others trade these textiles to all parts of West Africa, including Niger (Kraamer 2005a:238-43). Some of these traders own Zarma fabrics for private use, such as Seidou Ademu Tailor, one of the oldest inhabitants of Agbozume-Zongo, who has a 1930 textile, but not with figurative motifs (see Fig. 29; Kraamer 2005a:186).

(41.) In the Asante area, for instance, the introduction of two pairs of heddles was followed by the introduction of three pairs of heddles, creating the very rare asasia textiles in silk or rayon, the most elite of the royal textiles. These are characterized by their elaborate twill weaves. I have not come across any evidence of these textiles, or the use of three pairs of heddles, before the twentieth century. Lamb has seen an asasia textile belonging to the Ejisuhene, which, as she was told, "was given to the Stool at the time of the revolt against the British which led to the exile of Prempeh I in 1896" (Lamb 1975:126). In his extensive documentation of Asante patterns, Rattray does not mention any asasia pattern (Rattray 1927), but this might be because he did not have informants from Bonwire, where this type of cloth was woven (Lamb 1975:126).

(42.) The Asante wars in the Ewe-speaking region at the end of the 1860s could have caused relocation of inland weavers, including Agotime weavers, to the weaving towns in Asante; a possibility suggested locally in the Ewe-speaking area. However, the current Kentehene in Bonwire, Nana Yeboah, stated that no family homes in Bonwire trace an Ewe ancestry (Kraamer 2005a:187).