Laura McPherson – Speech Surrogates

Senegalese sabar – Is it a drum language?

Today’s post is a guest post from Sofiya Ros, PhD student at University of Utrecht in the Netherlands. Learn more about Sofiya under the contributor bios page!

Senegalese drummers show the practice of playing drums in correlation to speech. These drummers are part of the social class of griots [Hale 1998, Tang 2012], and their most common drum is a single-headed drum known as sabar.

Although sabar drums are rarely used as a speech surrogate and their main function is to affect the listener rather than to convey a message, it is clear that the practices of playing the sabar involve a close connection to linguistic expressions. In personal interviews griots say that ‘the sabar can speak’ and utter spoken expressions in correlation to sabar rhythms they play [Winter 2014: 646].

To what extent do such correlations justify the claim that Sabar is a drum language?

Playing the sabar involves at least 9 different drum strokes (hand strokes, stick strokes or their combination), which can be seen as the basic phonemic units of the genre. These strokes compose different longer Sabar rhythms which can be correlated with spoken utterances in Wolof — the lingua-franca of Senegal.

Mapping between Sabar and Wolof is not as clear as in more well-known cases of speech surrogacy on drums (e.g. the Yoruba drum language). Yoruba has three contrastive tones: high, low and middle, and tone counters formed by a combination of two of the tones. The tones and counters of spoken Yoruba can be represented by the notes of the Yoruba drums. The drums literally mimic the spoken utterances, whereas Sabar works differently as Wolof is not tonal.

I am working with the data collected during previous expeditions to Senegal: bàkks (classical phrases in Sabar, not improvised on the spot) and improvisations in Sabar and their translations to Wolof. Our data has the advantage that it includes not only bàkks, which are more like fossils, previously generated phrases learnt by heart, but also improvisations, which attest that Sabar drumming is still productive and its performance is not restricted to an existing repertoire of traditional texts.

So, there they are, Senegalese griots, and here am I, having some recordings of their drumming with translations, trying to find out what lies behind the drumming. I am approaching this problem by finding out the rules of “translation” from Wolof to Sabar rhythms. Sabar exhibits certain rules on its own: rhythm production “involves grammatical operations different from those of the spoken language, and that meaningful sabar rhythms deserve to be studied as a separate object for linguistic research, a drum language referred to as Sabar” [Winter 2014: 645].

Since Sabar rhythms are clearly connected to Wolof, we should be interested in the rules and regularities that govern this connection.

My first step is to find the rules of the translation, the correlation between the rhythms and the spoken language. First, I test the hypothesis on phonological mapping between the two languages, meaning that each stroke of Sabar represents a syllable or a number of syllables in Wolof.

Surprisingly enough, there were enough correlations to assume that the phonological mapping can be a feature, for example, bi (‘the’) is always translated with the ‘gin’ stroke in 45 cases (and only once as ‘tan’):

(1) Adduna bi
tan tan gin gin
‘this world’

However, of course there are irregularities as well:

(2) Jëfee ndigël rekka wóor
turun gin tan tac tan rwan
‘to do what is commanded is the only true way’

(3) Jëf rekk gu baa-xa wóor
ce rwe gin pax gin gin
‘to do to do what is good in the only true way’

In 2 and 3 the same word wóor (‘true’) is drummed differently: as rwan and as gin.

My first statistical analysis has already shown some correlations: for example, ‘gin’ stroke (hand stroke at the edge of the drum) is used to represent short vowels in 1738 cases (87%) and long vowels in 252 cases (13%) and this distribution is significantly different from the general distribution of short and long vowels, so there are significantly more short vowels for gin than in general in the data (X² = 54.1531, p < .00001).

I am still working on the generalisations, nevertheless, inspire of some irregularities, I am already getting a shadowy feeling that there is a way out of this chaos.

-Sofiya Ros

References

Hale, Thomas A. 1998. Griots and griottes. Bloomington: Indiana University Press.
Tang, Patricia. 2012. The rapper as a modern griot: Reclaiming ancient traditions. Hip hop Africa: New African music in a globalizing world, ed. by Eric Charry, 79–91. Bloomington: Indiana University Press.
Winter Yoad (2014): “On the Grammar of a Senegalese Drum Language” – Language 90.3, pp. 644-668

How to decode a surrogate language, or, speech surrogates as wug testing

Let’s say in the course of doing fieldwork on a language, you come across a speech surrogate. What do you do? How do you begin to unravel the system and discover its inner workings?

As with most of my posts, I’ll focus on musical surrogates here, though I believe the methodology would be largely similar regardless of the modality.

With a musical surrogate language, chances are that there will be a preexisting repertoire of songs or commonly used phrases. While of course you should record as many of these as possible (more on the overall health of speech surrogates in another post), I wouldn’t recommend starting here to decode the surrogate unless the connection to speech is immediately apparent. The reason for this is that this material is likely to have been learned by rote, and quite possibly as music first before understanding its linguistic underpinnings. The repertoire may have been passed down largely unchanged from generation to generation, while the spoken language slowly evolved, creating a confusing distance between the two. It may not even be based on speech at all, but rather on a sung repertoire, as in the case of the Sambla balafon (McPherson 2018, McPherson and James ms). Finally, in most cultures, musical surrogate languages in their natural setting are full of proverbs and other figurative language, which may make it difficult to translate and compare spoken language to surrogate speech, especially if fieldwork is in its early stages or there is no thorough grammatical description available. Thus, even though the fixed repertoire may be the most natural, it is likely not the best place to start.

Instead, start with spontaneity. Ask how to say simple things. Treat it as you would regular elicitation for the spoken language. Depending upon the natural productivity of the surrogate language, this will be an easier or harder task for your musician consultants. In this post, I’ll describe my experience with the Sambla balafon, where the tradition remains highly productive. We’ll have other posts that describe cases where that may not be true, though I would maintain that testing productive phrases in this way is still an informative exercise regardless of how natural it is for the consultants.

It boils down to this: Elicitation with musical surrogate languages is essentially a kind of wug testing. Wug testing takes a morphological or phonological pattern observable in natural speech and asks speakers to apply it to novel forms. It reveals what speakers know about the rules and patterns in their language, rules that we infer from observing regular speech but in this context, speakers could simply be reproducing forms that they have heard rather than productively applying those rules. When we ask a musician to produce a novel phrase in the surrogate language, we are essentially asking, “What rules have they internalized from the repertoire or tradition and what can that reveal about how the relationship between the spoken and surrogate language?” Ideally it can go a step further and also reveal something about the way the spoken language works.

For the Sambla balafon, my initial meeting with Mamadou Diabate in Vienna provided me with a range of phrases on the balafon, most of which Mamadou himself had offered. This included some everyday phrases like “Where are you from?” or “What are you doing?” as well as a couple elicited phrases that I snuck in, like “I will buy a goat” vs. “I will buy goats”. Though some of the phrases were more complex than my level of Seenku understanding at the time, it was apparent that factors like tone and vowel length were playing a role in the surrogate language.

When I returned to Burkina Faso the following summer, I was determined to get to the bottom of how exactly tone and syllable structure worked, and whether there were any other phonological contrasts that I was missing. I spent a few days with Mamadou’s nephew Nigo Diabate focusing on elicitation. I began with a simple sentence whose tones I felt more or less confident about:

mó sḭ̌ səmâ nɛ̏
‘I am dancing’

Then, just as in regular spoken elicitation, I systematically varied one element, in this case, the subject pronoun, swapping out the High-toned 1sg mó for a Superhigh-toned 1pl mi̋:

mi̋ sḭ̌ səmâ nɛ̏
‘We are dancing’

Sure enough, only the beginning of the phrase differed on the balafon, with the pronoun mi̋ corresponding to a higher note on the balafon. Transcriptions of these two phrases are provided below, where the Seenku names of the notes of the balafon are abbreviated along the left-hand side and the square of the grid is filled in when that note is struck. (See Strand 2009 or McPherson 2018 for discussion of Sambla tuning and note names.)

After a few iterations with different subjects, I varied the verb, changing from səmâ ‘dance’ with a contour tone to kȍeewith a level extralow tone.

mó sḭ̌ kȍee nɛ̏
‘I am singing’

Sure enough, the notes corresponding to the verb dropped to the same level.

Of course, the relationship between the spoken and surrogate language may not always be straightforward or one-to-one. For the Sambla balafon, for instance, elicitation sometimes reveals free variation, such as the following two musical equivalents of ȁ sḭ̌ səmâ nɛ̏ ‘s/he is dancing’, with an extralow-toned subject pronoun:

Notice that in both of these cases, the səmâ nɛ̏ part is played lower on the instrument than it was with the other pronouns. I suspect this has to do with the initial extralow tone, but it remains a bit of a mystery, since when asked, a musician will also accept it played higher up.

Cases of variation like this highlight the importance of (1) recording the same phrase more than once, preferably on different occasions to avoid self-priming, and (2) collecting data from multiple consultants, if possible. This can help triage errors from true variation, identify which version might be most natural, and determine how consistent the surrogate language system is between practitioners. Much of this is common sense from wug testing, highlighting once again the connection between surrogate languages and these other types of experimental work.

Before wrapping up this post, it’s worth pointing out that different consultants will have differing levels of tolerance for this kind of elicitation. In my experience in Mali and Burkina Faso, classical wug testing in the spoken language is a near impossible task (“But that’s not a word…”), but the Sambla balafonists have been fairly tolerant of my bizarre requests to say unusual things on the instrument. Thus, elicitation work on speech surrogates offers a unique source of “external evidence” (Churma 1979) to probe the phonology of a language where more traditional methods like wug testing may fail.

Nevertheless, there are limits. The Sambla balafonists are much more willing to provide phrases than single words, since phrases offer some possibility of disambiguation and thus render the message more natural. For the same reasons, they are not too keen on meaningless frame sentences that would allow us to look for subtle differences in the musical rendition of single words. As Mamadou has told me, it just doesn’t mean anything and thus isn’t true balafon speech. This may depend on the tradition, since I have seen others like Samuel Akinbo working with Yorùbá dùndún drummers (Akinbo 2019) have more success eliciting single words.

Only once the productive rules are worked out through elicitation would I recommend tackling the fixed repertoire. You’ll be armed with the musicians’ metalinguistic knowledge of the system and use that to determine how close of a match the rote material is to its spoken translation.

References

Akinbo, Samuel. 2019. Representation of Yorùbá tones by a talking drum: An acoustic analysis. Langues et Linguistique Africaine.

Churma, Donald. 1979. Arguments from external evidence in phonology. PhD dissertation, Ohio State University.

McPherson, Laura. 2018. The talking balafon of the Sambla: Grammatical principles and documentary implications. Anthropological Linguistics 60.3: 255-294.

McPherson, Laura and Lucas James. Ms. Artistic adaptation of Seenku tone: Musical surrogates vs. vocal music. Submitted to Selected Proceedings of ACAL50.

Strand, Julie. 2009. The Sambla xylophone: Tradition and identity in Burkina Faso. PhD dissertation, Wesleyan University.

Why study musical surrogate languages? A linguist’s perspective

It may seem like an unusual choice for a linguist to study musical instruments, as if we’re dallying uninvited with the territory of ethnomusicology. However, linguistics is the study of human language in all of its myriad forms, and I would argue that musical surrogate languages represent one of these forms. Now, it’s not a primary form like spoken or signed language. No child grows up with a musical surrogate language as their mother tongue, used for all day-to-day interactions. But musical surrogate languages are linguistically grounded in a way that other forms of non-verbal self-expression, like dance or dress, are not. In other words, we can learn a lot about spoken language through musical surrogate languages.

Here, I’ll offer an anecdote about how I stumbled into this area, in the hopes that it will encourage other linguists to more purposefully join the quest to better understand these systems.

It was 2013. I was in Burkina Faso on my first major field trip to document Seenku, a small Mande language spoken near Bobo-Dioulasso (by an ethnicity known to outsiders as the Sambla). By this time, I had been working in West Africa for five years, having recently wrapped up a project in neighboring Mali documenting Tommo So, a member of the Dogon language family. A lifelong lover of music, I had spent the last year analyzing Tommo So vocal music to understand what happened to the tones of the language when it was set to melody. Even in my early days working with Seenku, it was clear that its tone system was more complex than that of Tommo So, so I wondered what those tones would do when sung.

I asked my two main consultants, Clement and Emma, if they could bring me any recordings of Seenku songs. A couple days later, Clement gave me a folder of mp3s from a Sambla musician, Mamadou Diabate. That night, I greedily pored over the music, thrilled to hear what this language would sound like in song. To my great disappointment, the majority of the recordings were instrumental, a delicious but impenetrable tangle of xylophones (known in West Africa as balafons). I managed to find a couple of sung phrases and brought them up with Clement and Emma the next day to translate.

tɛ́nɛ́ gba̰̋ kú dɔ̏ɔ-nɛ̏-fi̋ɛ kənű mḭ̏ɛ̰, kʊ́ lɛ́ í bɛ̋ í wó bṵ́ɛ̰ɛ-kɛ̰̋ɛ̰
‘Whoever smells the odor of second day millet beer, he will say he worships the fetish.’

I thought I had wrapped up the work when Clement said enigmatically, “I know what the balafon is saying.”

Until that moment, the thought had never crossed my mind that an instrument could speak. Sure, I had heard the phrase “talking drums” before, but never given it much thought. Clement told me how everyone in the village could understand when the balafon spoke to them and offered a few translations from the song we were listening to (Ji Te So, from Mamadou Diabate’s album Keneya). And then I turned my attention back to pressing matters of conjugating verbs and figuring out phrasal tone, tucking the balafon conversation away for a rainy day.

That day came the following year, when a winter trip to Burkina Faso was postponed due to revolution. I was headed to Europe for a workshop and remembered that Mamadou Diabate lived in Vienna. On a whim, I reached out to him to see if I could pay him a visit and try to unravel some of the mysteries of how the balafon speaks. He agreed, and before I knew it, there we were in his studio. He was sitting behind his balafon, a huge instrument of wood and gourds, dressed in a crisp white shirt and black slacks, ready for a concert later. He told me that anything you could say with your mouth, you could say on the balafon. I offered a few phrases I’d been learning (things you would probably never actually say on the balafon, but you theoretically could).

“I will buy a goat.”
“I will buy goats.”

And there it was, that tonal difference that distinguishes Seenku singulars from plurals, it was clear as day in the notes of the balafon. A pitch difference, subtle in the human voice, rendered categorically distinct by the keys of the instrument. Mamadou explicitly pointed out that “goat”, bî, had two tones, and he enunciated them: bí. ȉ. Then he played them on the balafon. I had been grappling with the representation of this word, wondering whether that slight pitch fall I heard was just a phonetic effect or a true phonological feature of the word, a question Mamadou all but answered for me in the space of thirty seconds.

Since that visit in November of 2014, it has become increasingly clear how much we as linguists can learn from musical surrogate languages. They are the musical embodiment of what speakers know about their language, whether consciously or subconsciously. Used productively, they can provide a window onto underlying forms, as in the case of the Sambla balafon where postlexical tone processes aren’t encoded when a musician plays. In rote lyrics or proverbs, they can reveal archaic language or crystallized encodings that suggest earlier stages of the grammar. Rhythmic measurements can shed light on inner speech and how closely it matches verbal production, even though the oral articulators aren’t involved. And since every linguistic contrast isn’t encoded in surrogate speech, the choice of what is encoded allows us to explore questions of salience, functional load, or the balance between the need for communication and aesthetic principles.

Most research on surrogate languages has been carried out by ethnomusicologists or anthropologists, who produce fascinating work on the musical and cultural underpinnings of the systems. But linguists have been slow to the table, leaving many questions unanswered and many details unquestioned.

With most musical surrogate systems falling into disuse, time is of the essence. Just like the call to document endangered languages in the 1990s, this is my call for us to document endangered musical surrogate languages. Otherwise we’ll never know what they could have taught us about human communication.

-Laura McPherson