Quantitative similarities between the banjo and a diverse collection of West African lutes

Abstract

The banjo entered world musical culture through the ingenuity of communities of enslaved Africans in the Caribbean. The banjo is rooted within the lute-playing traditions of West Africa, where several remarkably banjo-like instruments and playing styles exist today. The banjo is a creation of the Black diaspora, however, and has no obvious single ancestor among extant West African lutes. Understanding the relative similarities between extant West African lutes and the gourd banjo may shed light on the cultural context of banjo origins. This study examines structural similarities between the gourd banjo and 61 West African lutes using two quantitative approaches for measuring and representing similarity among entities (UGPMA and NeighborNet). The banjo groups with a cluster of lutes from peoples in the Senegambia region speaking Bakic languages, which includes the Jola ekonting, an instrument that has garnered considerable recent attention as a banjo relative, but also shows similarities to lutes from the Niger Basin. This suggests that the relatively egalitarian social context of lute playing seen in Bakic language-speaking cultures may have been especially influential on the development of the banjo among enslaved populations in the Caribbean, but that the banjo draws on heterogeneous cultural influences and that more attention should be paid to the influence of eastern Sahel musical cultures on the evolution of the instrument.

Introduction

The modern banjo is among the most famous American instruments and is strongly associated with American folk and country music. It has spread across the world and found surprising roles in musical cultures as disparate as those of Ireland, Tonga, and even West Africa, the land in which the banjo tree finds its roots (Dubois, 2016). The Black genesis of the banjo was well-known to early American commentators, including Thomas Jefferson (“The instrument proper to [Blacks] is the Banjer, which they brought hither from Africa…”) (Jefferson, 1787; Epstein, 1975, 1977; Conway, 1995; Dubois, 2016; Pestcoe and Adams, 2018a). The more recent association between the banjo and white, rural culture, particularly the 20th Century genre of commercial country music known as bluegrass, has obscured this history for many, however. The problem of identifying West African progenitors of the banjo is made more challenging by the intervening centuries of cultural change in both North America and West Africa and the ruptures of Colonialism and enslavement.

The banjo is a plucked lute with five, or, less commonly, four strings, and a resonator assembled from a round membrane stretched across a pot of wood, metal, or gourd (Fig. 1). Most modern banjos are fretted, although this is a relatively recent innovation, and some folk traditions prefer unfretted instruments (Conway, 1995). A short, unfretted drone string is a characteristic feature of five-string and early four-string banjos (modern four-string “tenor” and related banjos lack this string). The earliest banjos were built with resonators made from an open gourd (Cucurbita sp.) closed by a vibrating membrane formed from a stretched skin (“gourd banjos”). In the expansion of the Horbostel–Sachs system of organological classification proposed by Charry (1996), banjos are “plucked spike lutes” (Pestcoe, 2018a). The term “spike” refers to the way the neck of the lute, against which the strings are depressed, is attached to the resonator. In handle lutes, like a guitar or violin, the neck is either attached to the resonator externally or carved from the same piece of wood. In spike lutes, the neck penetrates the resonator body either partially or fully. This method of construction is seen in the earliest Mesopotamian lutes, but is relatively rare today, with the center of diversity appearing in West Africa (Charry, 1996).

Fig. 1: A modern, factory-produced banjo (photograph by the author).

A fretted neck is attached as a spike penetrating the internal diameter of a resonator constructed from a membrane stretched across a metal and plastic hoop. There are five strings, included a short, unfretted drone string.

The earliest reports of banjo-like instruments in the Western Hemisphere come from the 17th Century Caribbean and are closely associated with the musical culture of Africans enslaved on the plantations of the West Indies and the Guiana Shield (Dubois, 2016; Pestcoe, 2018b; Ross, 2018). The British naturalist Hans Sloane provided the earliest detailed description and illustration of a banjo in a European source, from an example collected on a plantation in Jamaica. His “strum strump” features a gourd resonator attached to a flat, wooden neck with friction-tuning pegs (Fig. 2) (Sloane, 1707). Another important early depiction comes from a watercolor produced in South Carolina in the late 18th Century, now known as the Old Plantation. It depicts an enslaved musician playing the banjo at a dance, perhaps celebrating a wedding. Physical examples of early banjos also exist, including an 1840 example collected in Haiti and housed in the Musee de la Musique in Paris (Ross, 2018). These records confirm that the distinctive features of the gourd banjo coalesced early in the history of the instrument. The first report of the instrument in the future United States comes from 1736 in New York City, at the time a major entrepôt for recently enslaved Africans who had been “acclimated” on West Indian plantations (Pestcoe and Adams, 2018b). The mid-Atlantic seaboard and New Orleans appear to become important centers for the diffusion of the banjo in the 18th and early 19th Centuries, and it is perhaps not coincidental that each region fosters important subsequent developments in American musical culture and Black–white musical exchange (Conway, 1995).

Fig. 2: Two early gourd banjos, labeled “strum strumps,” collected in Jamaica by Hans Sloan (Sloane, 1707).

Strum strumps in the foreground, illustrated with a harp and fibers used to produce lute strings.

It is likely that mountain whites, the population today most closely associated with the banjo, first encountered the instrument and certain characteristic regional playing styles from enslaved or free Black musicians living in their communities (Conway, 1995; Dubois, 2016; Gibson, 2018). White interest in the banjo may date from an early period, particularly in “backwoods” and frontier regions where social hierarchies may have been somewhat looser, as attested by reports of white men dancing to banjo music in Knoxville, Tennessee in 1798 and the white circus roustabout Archibald Ferguson, from (West) Virginia, who had mastered the instrument well enough to instruct early minstrel Dan Emmett by around 1840 (Conway, 1995; Gibson, 2018). The banjo was popularized for a mass white audience, however, by the minstrel stage. The Virginia Minstrels, featuring Emmett, Joel Sweeney, and Billy Whitlock, are credited (correctly or not) with several lasting innovations, including the addition of a fifth string to the originally four-string instrument (not, as sometimes believed, the fifth, short string, but another long string) and the replacement of gourd resonators with a membrane stretched across a manufactured cheese hoop (Conway, 1995). Popular skepticism of the authenticity of minstrelsy and efforts by banjo evangelists in the late 19th Century to “civilize” the instrument involved an erasure of the instrument’s Black origins and likely contributed to the white cultural associations of the instrument today (Epstein, 1975; Dubois, 2016).

The early banjo fits within a long tradition of West African plucked lutes, which, along with evidence from other instruments, supports the transmission of instrument-building and playing techniques across the Middle Passage to the Americas (Oliver, 1970; Conway, 1995; Charry, 1996; Dubois, 2016; Pestcoe, 2018a). Bowed lutes (fiddles) are also common in West Africa, and although enslaved Africans and their descendants in North America did not widely reproduce African bowed lutes, many became accomplished at the European violin and heavily influenced the development of American fiddling (Epstein, 1977; Jamison, 2015). Other important West African instrumental traditions (drum orchestras, thumb pianos, and xylophones, for example) were transmitted to other parts of the Americas, but the conditions of enslavement in North America were apparently unconducive to their wide persistence (Oliver, 1970; Epstein, 1977; Conway, 1995; Dubois, 2016). Drums were considered particularly threatening because of their ability to communicate across long distances and mobilize slave resistance and were banned across most of the slave states after the 1831 Nat Turner Rebellion (Epstein, 1977; Conway, 1995).

Despite the instrument’s clear African origins, even the earliest banjos show differences from any known African lute. West African lutes have round, unfretted necks with strings attached using hide tuning knots that must be loosened and tightened to adjust the pitch (Charry, 1996; Pestcoe, 2018a) (Fig. 3). Depictions of early banjos show flat necks with strings attached to European-style friction tuning pegs (Dubois, 2016; Pestcoe, 2018a, 2018b) (Fig. 2). The neck of the banjo passes the entire length of the resonator, making it a “full-spike” lute (Charry, 1996; Pestcoe, 2018a). This is a method of construction that appears largely confined to Senegambia (the region combining Senegal, The Gambia, Guinea-Bissau, and parts of Guinea and Mali). However, in Senegambian lutes, the neck passes over two indentations in the resonator and threads through the sound hole in the skin membrane, while in the earliest banjos the neck penetrates through the body of the resonator (Pestcoe, 2018a). This a construction technique associated with bowed and plucked lutes built by cultures in the Niger River basin, particularly the Hausa. Together, this evidence suggests that by the time the banjo garnered the notice of European observers it had emerged as an indigenous creation of enslaved Black communities, likely drawing on the instrument-building and playing traditions of multiple African cultures and European influences (Dubois, 2016; Pestcoe, 2018a; Pestcoe and Adams, 2018a). This creole origin was likely essential to its function as a tool for musical expression among the linguistically and culturally heterogenous slave societies of the Caribbean (Dubois, 2016). In this sense, there is no single “banjo ancestor.” However, identifying “banjo cousins” may enrich our understanding of the musical traditions and cultural contexts from which the banjo emerged.

Fig. 3: Hausa gurmi played by Salamatu Mai Gurmi, collected by Abdalla Ubu Adamu (still from YouTube video [Adamu, 2013]).

The gurmi is constructed by stretching a membrane across an open gourd resonator, which is closed by a vibrating skin membrane. The round neck fully penetrates the resonator, and the strings are attached to the neck with leather tuning knots. Note also the downstroking style of performance, also seen in early banjo playing.

A large diversity of apparently banjo-like lutes are played by people across West and Central Africa. Many of these instruments share quintessential banjo features of strings compressed against a neck and plucked in a down-stroking style, one or more strings played in an uncompressed droning style, and a vibrating skin membrane stretched across a gourd or carved resonator (Charry, 1996; Pestcoe, 2018a) (Fig. 3). Early interest in the African origins of African–American music focused on the equatorial, rainforest regions of West and Central Africa, close to the centers of slave exportation, where drum orchestras predominate (Oliver, 1970; Linford, 2019). In his speculations on the African origins of blues music, Oliver suggested instead that the plucked instrument traditions of peoples living in the Sahel and northern savanna regions to the north of the Slave Coast, particularly Senegambia, may have been disproportionally important in the development of North American Black music (Oliver, 1970). This has been supported by observations of lute playing styles from Senegambia that appear to closely resemble banjo down-stroking, likely the oldest banjo method (Conway, 1995; Adams and Levy, 2018). This region of West Africa is dominated by speakers of Mande languages, whose societies are traditionally structured around a caste system in which lute playing is the province of certain members of a hereditary musical class known as the jeli or griot (Oliver, 1970; Conway, 1995; Charry, 2000). Similar distinctions exist among neighboring Wolof and Fulbe peoples and may be the consequence of Mande influence and the shared cultural milieu of Islam (Charry, 2000). Griot lutes share certain structural features, including a carved wooden sound resonator and fan-shaped bridge attached to the neck through the sound hole (Charry, 1996, 2000). Other Senegambian peoples, particularly speakers of the Bakic subgroup of Atlantic languages, do not share this system of traditional social organization, and among them, lute playing is more widespread (Charry, 1996, 2000; Pestcoe, 2018a). The Jola of southern Senegal has garnered particular attention because of the close apparent similarity of their folk lute, the ekonting, to gourd banjos (Jatta, 2007; Adams and Levy, 2018; Linford, 2014, 2019; Pestcoe and Adams, 2018a).

Commentors on banjo origins have sometimes implicitly assumed that a single instrument could be found on the African continent that either through direct transportation or imitation gave rise to the gourd banjo (Pestcoe and Adams, 2018a). This is almost certainly impossible, given the structural differences between even the earliest banjos and West African lutes and the over 300 years separating the earliest records of Caribbean banjos from the living musical traditions of West Africa. In response to this challenge, Pestcoe and Adams proposed the framework of a banjo “genome” (2018a). This would capture the features of the gourd banjo that might point to its genealogical origins among African lutes. Building on the classificatory work of Charry (1996), Pestcoe and Adams proposed a set of four character states that might be especially important in clarifying banjo origins and used these to compose an organized list of known West African lutes (2018a, 2018c). This is a “typological” or “microtaxonomic” approach to instrument classification that attempts to work from informative character states that vary across populations of instruments to build a genealogical framework (Kartomi, 2001). There are connections between this approach and evolutionary approaches to understanding ethnomusicological change, particularly the framework of Alan Lomax emphasizing continuity (= inheritance), innovation (= variability), and selection (Lomax and Berkowitz, 1972; Savage, 2019). This assumes that instruments change through time, driven in part by the innovation and interests of performers, but that changes are likely to be influenced by the inherited character of folk instruments and inherited performance culture which shapes both background of instrument production and the received preferences of listeners and musicians. This non-teleological view of material cultural change is much closer to modern views of biological evolution than the progressivist and hierarchical understanding of instrument topologies that predominated in the early history of organology (Savage, 2019; Street et al., 2022).

The character list of Pestcoe and Adams (2018c) provides a potentially powerful typological framework in which to consider the genealogical kinship of the banjo among instruments from West Africa, but without statistically explicit analyses enumerations of character similarities can be misleading (Nunn, 2015). A quantitative clustering approach to banjo “genome” analysis may allow us to resolve clusters of West African instruments relatively similar to the gourd banjo, with implications for our understanding of the musical cultures that may have had the most influence on its development. Prior research on playing styles and instrument construction suggest that the banjo will show the closest similarities to Senegambian lutes, particularly folk lutes associated with the Jola and other peoples speaking Bakic languages (Jatta, 2007; Linford, 2014; Dubois, 2016; Pestcoe, 2018a; Adams and Levy, 2018).

Methods

West African lutes were coded into a taxon-character matrix from the classificatory descriptions provided by Pestcoe and Adams (2018c). These observations derive from personal communications and an exhaustive review of the primary and secondary literature on West African lutes assembled primarily by Shlomo Pestcoe from 2009 to 2015. Citations are included for each observation within Pestcoe and Adams (2018c). Pestcoe began with the list of lutes produced by Charry (1996) and tried to record any additional lutes potentially relevant to banjo origins. Instruments were classified by the language group associated with the people who play each instrument (Güldemann, 2018; Pestcoe and Adams, 2018c). These language groupings are intended to represent probable historical affinities among lute-playing peoples.

Six characters were used: player (griot or non-griot), bridge (bipedal, block, cylindrical, and fan), resonator (calabash, gourd, and wooden), neck (full-spike or semi-spike), the method of attachment of the neck to the resonator (passing over the rim or penetrating the body), and a number of strings (one to five). Four of these states (player, bridge, resonator, and neck) are listed by Pestcoe as significant in characterizing the “genome,” and hence genealogy, of West African plucked lutes (2018a). The method of attachment of the neck and number of strings are also provided by Pestcoe and Adams in their list of West African lutes and may provide additional discriminatory power (2018c). Multistate characters were used for instruments in which more than one configuration was listed. The gourd banjo was coded with a non-griot player, a bipedal bridge, a gourd resonator, a full spike neck, attachment by penetrating the body, and four strings, drawing on early descriptions of Caribbean instruments and the “Old Plantation” painting representing an 18th Century North American gourd banjo (Dubois, 2016; Pestcoe, 2018b, 2018c; Ross, 2018). Modern instruments with a pan-West African distribution or constructed from materials that would not have been available in the 17th Century (such as tin cans) were excluded, leaving 62 instrument “taxa” (Table 1).

Table 1 List of West African instruments incorporated into the cluster analyses.

Two approaches were taken in the quantitative analysis of instrument similarities. The first, unweighted pair group method analysis (UPGMA) joins taxa and groups of taxa using an algorithm that minimizes Euclidean distances among all clusters, starting with the cluster of the smallest distance. This produces a dendrogram of similarities among instruments that can be read like a phylogeny, although without a strong assumption that similar instruments share unique common ancestors defined by the acquisition of derived characters, as are used in the construction of a maximum parsimony phylogeny, which may not apply to the evolution of instruments (Kartomi, 2001; Nunn, 2015).

The second approach uses the distance matrix to create a network of weighted similarities called a NeighborNet, which builds on the Neighbor-Joining method of constructing dendrograms to represent similarities among taxa (Bryant and Moulton, 2004). Instead of creating a bifurcating tree, a NeighborNet represents multiple potential “splits,” or mutually exclusive bifurcations in the data, as parallel edges forming a reticulating network. This can account for uncertainty in the underlying reconstruction (similar to the use of a polytomy in a phylogeny or dendrogram) or model potential instances of horizontal information transfer (as is likely in the evolution of material culture objects like instruments) (Morrison, 2005). Phenetic similarity among tip taxa on a NeighborNet network can be measured as the minimum distances among taxa along each set of edges, with edge lengths proportional to the non-negative least squares weights assigned to each edge in determining the pattern of splits (Huson and Bryant, 2006).

These two methods produce complementary quantifications of similarity among taxa. UPGMA reveals nested patterns of similarity, while NeighborNet plotting reveals potential patterns of horizontal reticulation in the development of the banjo and a means of calculating continuous phenetic similarities among instruments on a quantitative morphospace that does not assume the relationships among the taxa to be modeled by a bifurcating tree. Analyses were performed using packages in the statistical programming language R. The matrix of character descriptions was converted into a Nexus format using functions in the R package “ape” and used to calculate a morphological Euclidean distance matrix using functions in the R package “Claddis.” This was used to construct a UPGMA tree and NeighborNet network using functions in the R package “phangorn” (Paradis et al., 2004; Schliep, 2011; Lloyd, 2015).

Results

The UPGMA dendrogram groups the gourd banjo as sister to six Senegambian plucked lutes (the bunchundo, busunde, ekonting, kisinta, kusunde, and nopata) associated with Bakic language-speaking peoples, themselves forming a cluster of zero distance (Fig. 4). The banjo shares with these Bakic origin instruments a non-griot player, a bipedal bridge, a gourd body, and a full-spike neck. It differs in the number of strings and method of attachment of the neck to the body. The sister group to the banjo + Bakic cluster is more diverse and includes instruments associated with Chadic-speaking peoples of the Omotic language group (Kilba, Kotoko, and Hausa), the Saharan Toubou, the Adamawa Bana, and the Atlantic Fulbe.

Fig. 4: Dendrogram by language.

UGPMA dendrogram of instruments, labeled by language group (Saharan-Songhai language groups and Berber-Semitic language groups each combined for plotting clarity). Tip labels follow the instrument names in the taxon character matrix (SI 1).

The NeighborNet network reveals a pattern of frequently conflicting splits among the NJ trees, which can be interpreted as evidence of horizontal transfer among instrument traditions or as reflecting a lack of data sufficient to fully resolve a dichotomous tree. Given the expectation that lute-playing traditions in West Africa have evolved in close geographic proximity and within a context of mutual cultural influence, it seems likely that substantial cultural transfer has occurred and that there is no true bifurcating tree of West African lutes. The banjo is visually grouped with the same set of Bakic language group lutes clustered with the banjo in the UPGMA analysis (Fig. 5). It is also close to a set of three lutes played by Omotic language speakers from the Niger River Basin (the Hausa girmi, Kilba gullum, and Kotoko gulum), which are actually separated from the banjo by a slightly lower minimum phenetic distance along possible edge configurations along the network (Fig. 6, Table 2).

Fig. 5: NeighborNet plot by language.

NeighborNet of instruments, labeled by language group (Saharan-Songhai language groups and Berber-Semitic language groups each combined for plotting clarity). Instrument labels as coded into taxon-character matrix: (1) banjo, (2) bappe, (3) bunchundo, (4) busunde, (5) diassare, (6) duru, (7) ekonting, (8) gambare, (9) garaya1, (10) garaya2, (11) gesere, (12) gullum, (13) gulum, (14) gurmi, (15) gurumi1, (16) gurumi2, (17) gurumi3, (18) gurumi4, (19) gzopoli, (20) jurkel1, (21) jurkel2, (22) juru_kelenni, (23) kaburu, (24) kakanza, (25) kambre, (26) keleli, (27) kerona, (28) keronaru, (29) kisinta, (30) kola_lemme, (31) kologo, (32) komo, (33) kona, (34) konde, (35) koni1, (36) koni2, (37) konou, (38) kontingo, (39) kook, (40) kuban, (41) kubru, (42) kusunde, (43) kwamsa, (44) molo1, (45) molo2, (46) molo3, (47) molo4, (48) molo5, (49) moolooru, (50) n_dere, (51) n_goni, (52) n_goni_ba, (53) n_goni_micin, (54) n_jurkel, (55) ngulan, (56) nopata, (57) teharden, (58) tidinit, (59) xalam, (60) xalam_gesere, (61) yakandi, and (62) yomshi.

Fig. 6: Boxbplot of NeighborNet phenetic distances by language group.

Boxplots of language NeighborNet phenetic distances, labeled by language group (Saharan-Songhai language groups and Berber-Semitic language groups each combined for plotting clarity). Boxes represent the interquartile range and whiskers extend to 1.5 times the interquartile range.

Table 2 Minimum edge-wise phenetic distances separating instruments from the banjo on the NeightborNet network.

Discussion

UPGMA cluster analysis supports a close resemblance between the American gourd banjo and a group of instruments, including the Jola ekonting, played by Bakic language speaking peoples living along the Atlantic coast of Senegambia. This is consistent with observed similarities in playing style and construction between the ekonting and early records of the banjo (Jatta, 2007; Linford, 2014, 2019; Ross, 2018). The banjo-Bakic cluster also shows affinity with seven instruments from the Niger River basin, all of which differ from Senegambian plucked lutes in possessing a neck that pierces the body of their resonators, as see in the gourd banjo (Pestcoe, 2018a). Three of these Niger Basin lutes, associated with Hausa, Kilba, and Kotoko musicians, are also closer to the banjo on the minimum path consistent with the NeightborNet than are the cluster of Bakic instruments (Fig. 5, Table 2). This underlines the heterogenous cultural influences that gave rise to the banjo and suggests that more attention should be paid to the potential influence on banjo origins and playing styles from Niger Basin lute traditions.

If enslaved Bakic language-speaking people were particularly important in the development of the American banjo, this has implications for our understanding of the social origins of banjo performance. Unlike the Mande, Wolof, and Fulbe-speaking peoples dominant in Senegambia, Bakic language speakers belong to traditionally egalitarian societies that, crucially, lack a griot class of hereditary musicians (Charry, 2000; Linford, 2019). The trauma of enslavement and transportation across the Atlantic would likely have disrupted the traditional class system of a people like the Mande and deprived griots of the ability to maintain their traditional lineage and occupation. It is also possible that griots, with musical skills considered highly valuable by West African courts, were less likely to be sold to trans-Atlantic slave traders when captured in warfare (Dubois, 2016). In this circumstance, the more informal lute-playing culture of the Bakic people may have been especially influential. Niger Basin musicians are more likely to belong to stratified traditional societies with at least some degree of hereditary musical specialization, but lute playing among Hausa, at least, may also be associated with less formal and more fluid musical contexts (Ames, 1973).

Linford’s ethnographic study of ekonting playing among the Jola of the Casamance region of southern Senegal and The Gambia is the most detailed description of the social context of a folk lute from among Bakic language-speaking cultures (Linford, 2019). The ekonting maintains as one of its functions the ability to accompany singing during celebrations and courtship. The instrument is not treated in a ceremonial way and is often stored out of doors near where it might be played by groups of men resting after work. The material of the ekonting, both symbolically and pragmatically, reflects the everyday working life of Jola men. Early gourd banjos were, like the ekonting, instruments that could be produced by non-specialists out of materials drawn from the natural environment and the tools of daily labor (Dubois, 2016). As other musical traditions were suppressed by overseers or colonial authorities or otherwise became difficult to maintain, plucked lutes became important to the expression of a newly salient creolized African identity among the enslaved and to accompany communal dances (Epstein, 1977; Dubois, 2016).

Suggestions for future study might incorporate more information about tuning and methods of performance in West African lutes than were available for this analysis as well as more information on the cultural conceptions of different lutes within West Africa. Additional West African instrumental traditions that have gone unnoticed by Western observers may be described by ethnomusicologists in the coming decades. A quantitative cluster analysis approach may prove useful in better understanding those instruments and how they may relate to other musical cultures within West Africa and to banjo origins. Finally, no attempt has been made to interpret the relationships recovered among West African instruments except as they relate to banjo origins, but this analysis may prove a profitable starting point for ethnomusicological analysis by Africanist scholars interested in West African lute traditions, particularly those with the ability to work in West Africa and African researchers who might be able to more fully contextualize these instrument traditions in situ.