Writing the Sasi Dictionary: Some Personal Recollections

Introduction

Writing the Sasi dictionary was difficult. The purpose of this piece is to explain to a general audience (of non-linguists) why I found it so difficult. From this, they might get an idea of the challenges that a field linguist faces in their day-to-day work.

Sasi is the name of a language spoken in Botswana. As of 2024, there were around15 elderly speakers remaining. They all live in two small villages in southeastern Botswana (Mokgenene, Poloka), and surrounding cattle posts (Khibitswane). In 1996, I began working with the Sasi speakers. Nearly thirty years later (now in 2024), I have written a small dictionary of the language.

Just to be clear, the term ‘dictionary’ is not really an accurate term for what I wrote. In a real dictionary, there is a definition, as the following one from the Merriam-Webster dictionary:

dog noun

a carnivorous mammal (Canis familiaris) closely related to the gray wolf that has long been domesticated as a pet, occurs in a variety of sizes, colors, and coat types, and is sometimes trained to perform special tasks (such as herding, guarding, or acting as a service animal)

The definition in the Oxford English dictionary below has different details, but is organized in a similar fashion:

dog noun

A domesticated carnivorous mammal, Canis familiaris (or C. lupus familiaris), which typically has a long snout, an acute sense of smell, non-retractile claws, and a barking, howling, or whining voice, widely kept as a pet or for hunting, herding livestock, guarding, or other utilitarian purposes.

These definitions are in essence arbitrary. Both include the scientific name (Canis familiaris). But then each dictionary picks out some set of traits to focus on. One talks about the relation to the gray wolf, the other about the acute sense of smell. There is no single agreed upon definition of a ‘dog’. But at least there is some descriptive material to help the reader to understand what a dog is. 

In comparison, my dictionary entry for ‘dog’ is the following:

teema [LH.L] n. dog (Sets. ntša) teema-si dana ‘The dog is small.’

The Sasi word in this case is teema, and its translation into English is ‘dog’. Its translation into Setswana is ntša (pronounced ncha). The notation [LH.L] indicates the tone of the word (low tone, high tone, low tone). The period in the sequence LH.L is a syllable boundary: LH on the first syllable, L on the second syllable. Most of the entries in the dictionary have an example to show how the word is used. In the lexical entry above, the example is in italics. 

Missing from my dictionary entry is a general definition like the ones in the two dictionaries noted above. In fact, there is no definition at all, just a short translation into English and Setswana. If the reader did not know what a dog was in English, they would be completely lost as to what teema is in Sasi.

A better phrase for what I wrote would be ‘translated word list’. But nobody ever uses such a phrase, so I stick to the conventional term ‘dictionary’.

In my dictionary, there are a total of nearly 1,400 lexical entries like the one above, which is rather small number for dictionary standards. Other dictionaries of related Khoisan languages have thousands of lexical entries. In addition to the main dictionary, I appended a small reverse dictionary, going from English to Sasi. There is also a short introduction (11 pages) giving some background on the language. With the addition of the introduction and the reverse dictionary (English-Sasi), there are a total of 115 pages in the entire document. That is a very small contribution to the field of linguistics. Most linguists would agree that it is not much to show for nearly thirty years of work.

Chronology

Let me start off by giving a brief chronological account of my work on the Sasi language before I dive into the difficulties that I had in writing the dictionary.

I had originally stumbled upon the Sasi by pursuing a reference in one of Tony Traill’s papers. He talked about a group called the Tshasi, and gave Artesia (in eastern Botswana) as a location. In 1996, I took my truck to that location, but did not find any people fitting Traill’s description. However, I managed to talk to an elderly Botswana statesman named Greek Ruele, who actually knew a few words of Sasi. He had learned them working on the cattle post when he was a young boy. He was kind enough to sit down and talk with me. He directed me to the next village up the road, Dibete, where he said there were still some native speakers of Sasi left.

I left Greek Ruele in Artesia, and drove a bit down the A1 to find Dibete. As usual in this kind of situation, I started asking around, and eventually people directed me to Bakane, who lived just off the side of the main highway. I started working immediately with Bakane, who became my first Sasi consultant. At first, I was just trying to gather a few lexical items, and to figure out how to write Sasi down. These were just preliminary steps. Eventually, I brought Bakane to Molepolole, where I lived, and got him together with Titi, my ǂHoan consultant. I found that they could understand each other. So that meant that even though Sasi and ǂHoan were clearly different languages, they were mutually intelligible. That was a significant finding. And it brought up all kinds of questions about how the two languages were related and when they split up historically. That is, in what year did the two languages diverge and why did they diverge? The mystery deepened over the years as I realized that the Sasi people (located in the central district) had never heard of the ǂHoan, and vice versa.

I returned again to work with the Bakane in the summer 1999. This time, I was able to compare Sasi to ǂHoan in more detail and found out that syntactically (grammar) they were largely the same. The differences were most lexical (words) and phonological (sounds). I spent a lot of time recording, transcribing, glossing and translating seven short oral texts, that gave me an overall picture for how Sasi worked. The names of these oral texts are: Korhaan, Guinea Fowl Eggs, Ostrich Eggs, Wild Dogs, Lion Story, Working with Cows and Leopard. You can see from the titles how Bakane chose to talk about how he gathered food, and the wild animals he had hunted. I have yet to publish these, or any other, Sasi oral texts.

Other projects then came up and got in the way. In particular, I starting working another highly endangered Khoisan language N|uu during the summers of 2004, 2005 and 2007. This was a joint research project between me Amanda Miller and Bonny Sands. It involved going to Upington, South Africa on three separate occasions. Also, I changed jobs from Cornell to NYU in 2005, which was a huge undertaking, since I had to make sure my children and wife were settled in and happy in their new environment. One thing led to another, and I did not return to Botswana until 2011, when I applied for teaching/research Fulbright scholarship to teach at the University of Botswana.

Arriving back in Botswana in 2011-2012 (on another Fulbright), I tried to find Bakane to resume my study of Sasi. But when I went to Dibete, I found out that he had passed away. From his wife, I learned that he had two sisters in Mokgenene, which is around 60 kilometers up a gravel road from Dibete. So I went there. I found his sisters and put together a nice team of four speakers, all elderly ladies in their 70s. Our first sessions were in the summer of 2012. Since I was working as a teacher at the University of Botswana at the time, I would drive out to the village, and bring the consultants back to Gaborone where we worked when I had some free time. They loved visiting Gaborone, which they called Gabs. They loved being in the big city, watching television at night, having easily accessible water, and getting some time away from village life and its many responsibilities. All four have since passed away (as of 2024).

Starting in 2016, I started staying for short periods in Mokgenene, ranging from two weeks to a month. Then after each expedition, I would return back to Gaborone to think over my results and write things up.

During these years, I had the help of various different people: BK (a native speaker of Shua) and MC were my main translators from Setswana to English. My old NYU student, ZW participated in one of the field expeditions, but also helped me extensively with the technical aspects of using FLEx to write a dictionary. AC participated in some of the sessions in the summer of 2014, helping me with Setswana translations. He also proofread an early version of the dictionary in 2016.

In 2014, I published a Grammar of ǂHoan (with Jeff Gruber), which is closely related to Sasi. In that grammar, I included frequent comparisons to Sasi. Then in 2016, I self-published a ‘Sasi Spelling Primer’ It had a series of pictures, each with three words underneath it: Sasi, Setswana and English. These two documents ultimately made it possible for me to write the dictionary of Sasi, since by this time (2014-2016) I had a pretty good grasp of the basic syntax and phonology of the language. 

During expeditions to Mokgenene in 2016, I tried to focus on increasing the number of words in the dictionary. But it was very difficult to find new words. I combed through dictionaries of all the Khoisan languages of the region, and all the Setswana dictionaries I could find, and many books on animals, trees, flowers and birds. But finding new vocabulary was nearly impossible.

Oral Texts

I got the idea that in order to increase the size of the dictionary it would be a good strategy to collect oral more texts, in addition to the ones I already collected from Bakane. An oral text can be a personal anecdote, a historical account, instructions about how to do something, a fable, a conversation between two people, or just about anything where one or more people speak. I applied for an ELDP grant (2020) with the sole purpose of gathering oral texts in order to increase the size of the dictionary that I was writing. In part because of the pandemic, the whole process of working with ELDP/ELAR was extremely stressful for me. It was not a pleasant experience at all. But that is a topic for another blog post. 

Although I transcribed, glossed and translated over 3.5 hours of oral texts for the ELDP grant, the total number of new Sasi words resulting from the oral texts was less than 100. There were lots of new words in the oral texts, but they were all Setswana. When the Sasi speak their language, they code switch massively into Setswana, with nearly every sentence having several Setswana words in it.

But the oral texts served another equally important purpose for the creation of the dictionary. Initially, as I was writing the dictionary, I had the consultants create their own examples. For each word, I had them create an example or two. For example, for the word ‘hunt’, the example was ‘I am hunting a duiker.’ Then, I used those examples in the dictionary. But I wanted even more examples. I realized what a great source the oral texts were for dictionary examples. At that point (around Fall 2023), I went back over the entire dictionary, and added oral text examples wherever appropriate. As it turns out, the software I was using (SIL FLEx), makes such additions easy. For any word, you can simply search all the places in your oral texts where it appears (using ‘Show Entry in Concordance’). Then you can choose the best example sentence, and copy and paste it directly into the dictionary. That is a no-brainer, but it took me a long time to figure out the whole system.

Oral texts also provide a great way to check the definitions obtained by elicitation. By looking over all the various sentences where a word is used, it is sometimes possible to refine its definition to make it more much more accurate.

NSF Grant

In 2018, I received an NSF grant to train students to work on the Khoisan languages. Although the target of the grant was the Kalahari Khoe languages (which do not include Sasi), every time I came to Botswana I made some time to study Sasi as well. The grant was supposed to end in 2022, but because of the pandemic, it extended all the way until 2024, taking up an unexpectedly large portion of my life.

By around 2023, nearing the end of my NSF grant, I started to realize that my time doing research on Khoisan was coming to an end. My hair was turning grey, and my body was slowing down. And the pandemic had taken a huge mental and emotional toll on me. It was time to bring all of my Khoisan projects to an end. More than anything I wanted to get back to Togo where I started out in the Peace Corps, and to pass my final days teaching and doing research there, speaking Ewe to the Togolese. I had great plans for a Kpelegbe Ewe grammar and an interactive Ewe Dialect Map. I started to worry that with advancing age, I would never bring these dreams to fruition. 

So, starting in Fall 2023, my plan was to spend one more sabbatical year in Botswana, finishing up all my outstanding Khoisan projects (Kua grammatical sketch, Sasi dictionary, the investigation of Cua and Tshila as part of my NSF grant). I am happy to report that as of July 2024, I have indeed finished all those projects, including publication of the Kua grammatical sketch and Sasi dictionary.

In the goal of finishing up the Sasi dictionary, I was seriously inspired when my friend and colleague Bonny Sands, who worked together with Kerry Jones of South Africa and a large team of assistant editors, finished up the N|uu dictionary for publication. During Spring 2022, I worked every weekend with Bonny on transcription and grammar issues, helping her get the dictionary to its final stage. I also contributed a few dozen grammatical morphemes to the N|uu dictionary from the N|uu grammar that I had written with Levi Namaseb. My contribution to the N|uu grammar was small in comparison to the massive effort that Bonny and Kerry made. More importantly, the dogged determination and persistence that Bonny and Kerry showed inspired me to do the same for the Sasi dictionary. Without their example, there is a chance the Sasi dictionary would have never been published.

Finishing Up

When I arrived in Botswana during Fall 2023 for the last year of my NSF grant, I started the process by asking my old NYU student ZW to help me proofread the dictionary. ZW is a genius in using FLEx, the software program with which the Sasi dictionary was written. He also has a marvelous ability to focus on detail. So I paid him for 50 hours to proofread the entire dictionary and get it into shape for publication. He did a great job, pointing out many inconsistencies and possible errors. He pointed out hundreds of words that needed attention in one way or the other. Using his knowledge of FLEx, he also designed both the format of the dictionary and the format of the reverse dictionary, which now appear in print

Around the same time, I paid two separate Setswana speakers to read through the Setswana translations and to check the spelling and the Setswana translations. Both of these competent people had finished form five (high school), plus a few years of post-secondary training. One trained as an accountant, and the other as a chef in South Africa! They dug up about 20 Setswana errors (all very minor), which I corrected. I tried to hire a Setswana school teacher from a local secondary school to help me with this task. But once they knew a lekgoa ‘white person’ was involved, the price became exorbitant. One teacher I asked lived in Phakalane, the fanciest part of Gaborone. Over the phone, she seemed enthusiastic about the project and ready to meet with me, but after she talked to her wealthy husband, she said that she would charge me 500 Pula just for the petrol to drive to my house (a ridiculous demand), and that she wanted to charge me by the page for the work. Since the Botswana middle-class make comfortable salaries, it is not really possible to get them to work for the amount of money I had available. Unfortunately, this theme of radically overcharging lekgoa for goods and services was something I encountered regularly in Botswana.

As a final check on the dictionary, in October 2023 and February 2024, I brought in two Sasi speaking ladies that I had worked with previously, but who had not recorded the words in the dictionary. I will refer to them as MG and KK. My idea was to record every word in the dictionary with them, forcing me to carefully recheck transcriptions and definitions. In that way, I rechecked the whole dictionary, adding a few new entries and correcting many old ones. The entire process took over a month. I greatly enjoyed those last sessions with Sasi, those last few days working on a language that I had been in contact with since 1996.

What I found in doing the last-minute checking is that in many cases the new team had forgotten the words or never knew them in the first place. In a large number of other cases, they disagreed with the previous definitions. By the way, there is no chance of regional dialectal variation here. The whole Sasi community has always lived very closely together, at least in living memory. When the new team did not know a word, I added a note to the dictionary entry saying “Not everybody knows.” When they disagreed with a definition, I added a note to the dictionary entry along the lines of “Some people say….” indicating the difference. Adding all these cases together (forgetting, not knowing, disagreeing), I estimate that they accounted for at least 5% of the entire dictionary.

Writing the dictionary brought up the very basic question: When is a project finished? At what point do I say ‘This dictionary is complete.’ For a dictionary, there does not seem to be any natural stopping place. One can always add more words, more examples, more detailed definitions, etc. At some point, I reached the point of diminishing returns. Every new word took hours and hours of time, and there were often disagreements amongst the speakers. I speculate that the total vocabulary of the Sasi speakers is rather small because of their extended contact with the Batswana. 

I can even think of ways of getting more words now, if I had time. I could record little mini-conversations in Sasi, and ask them about new words as they came up. I am sure if I did this for a month (some conversations every day), I would find a few new words. It would be fun. But there is a point of diminishing returns. 

Until the very end of the process, when I sent the manuscript off to the publisher, I was horrified to discover further corrections of various kinds, including mistakes in transcribing the sounds and the tones, but also in the definitions. The issue of perfectionism raised its ugly head. Would it really be worth it to delay the publication of this small volume another few years, just to correct a few tones? Or should I let the community have what I created, in spite of its imperfections. For every project, every situation, every phase of life, there must be a beginning and an end. I needed to fix the endpoint of the Sasi dictionary project somewhere.

Difficulties

So why would a short dictionary be so hard for me to write? I enumerate some of the reasons below.

There was no Sasi speaking community.

Let me start by asking the most basic of all questions: How do you find words for a dictionary?

The best way by far is to live and work in the community, talking with people all day long, in day-to-day activities. For example, if you live in the community, you can go to the market, and buy your groceries, learning a lot of related vocabulary. You can learn the words for what you want to buy. You can learn the words for counting (as in the money you spend). You can learn the words for financial transactions (‘How much does it cost?’ ‘Please reduce the price.’). But that is just the tip of the iceberg. If you live in a community, you will strike up all kinds of conversations throughout the day. You will be immersed in the language you are studying. And being immersed like this will give you lots of opportunity to learn new words, and also refine your definitions of existing words. When I was a Peace Corps volunteer in Togo in 1985-1987, this is exactly how I learned both French and Ewe. In Togo, nobody except the other Peace Corps volunteers knew English. So I had to learn French. It was either sink or swim. But there were also plenty of opportunities to practice my Ewe in the community. That is why today, I am a fluent speaker of both French and Ewe. Because, I was required to speak these languages in a community on a day-to-day basis.

Unfortunately, nothing like this was possible with Sasi for the simple reason that nobody really uses the language anymore. The Sasi consultants never speak Sasi with their children, because their children only know Setswana (and in rare cases, English too). The children might know a few words here and there. They might even be able to understand some of what their parents say. But they do not ever use the language themselves.

Even when they consultants are together, they generally do not speak Sasi. The speak Setswana because the others around them speak Setswana. So there is no real community of Sasi speakers. I cannot just go to the market and try to purchase something using my Sasi, nor can I ask for directions. All uses of Sasi are confined narrowly to our work sessions. I can have conversations with my consultants then. That is where I learned to speak Sasi. But because Sasi is not spoken by a community, I have had no opportunity to hear it being used in many different day-to-day contexts, increasing my vocabulary.

Every word was like resurrecting the dead.

As a corollary, the Sasi speakers often had trouble recollecting even the simplest words. Although I started working on the dictionary in earnest in 2011, I did not find a word for mosquito until May 2016. I had asked the consultants many times before then, but they could not remember what the word was. Finally, out of nowhere, they finally remembered it: similiki. After I got the first word, alternative words started pouring in: minimini, similimili, similimiliku, siminimini. Everybody seemed to have their own word. It often took the consultants a long time to remember words, even after several years of trying.

As another indication of the same issue (difficulty in recollecting vocabulary), there was a large amount of disagreement between the consultants. In many cases, some consultants would remember a words, but the others would never remember it. In those cases, I added the note ‘Note everybody knows’. That note appears over 55 times in the dictionary, indicating that for those words, some of the consultants could simply not remember the word. 

Yet a third indication of this issue is the massive amount of code switching found in the Sasi oral texts. Whenever a speaker does not know a Sasi word, they just switch over to Setswana, which they speak fluently.

My consultants were elderly and illiterate

The community of Sasi speakers was very interested in the project. They always participated in a patient and professional way. And they were eager to see the results. The problem is there were only a handful of speakers.

In addition, my consultants were all elderly and illiterate. This had an immediate effect on how they were able to help me. There is a vast difference between a young literate speaker, who is generally meta-linguistically aware (aware of language as an object of study) and elderly illiterate speakers. But there are no young literate speakers of Sasi.

I am not a natural lexicographer.

To compound the difficulties, writing a dictionary is not a natural task for me. I am not a natural born lexicographer. To be a lexicographer, you need first to transcribe accurately the way people say the word. You need to be an expert on the sounds of language, because you never know what kinds of sounds the language will throw at you. Then you also need to be skilled at working out the meaning of individual words. You also need to be very good at grammar, in order to add the appropriate grammatical information e.g., that a word is noun or a verb or an adjective. Adding one single word to the dictionary requires many different interlocking skills, and a precise set of detailed tasks.

You then need to repeat these tasks thousands of times, once for each word, making sure that you stick to exactly the same conventions for each word. The task of assembling a large number of words into a document does not resonate with me. It is possible to make a dozen different mistakes in formulating a single lexical entry. In putting together a small dictionary, it is literally possible to make thousands of different errors. 

Different people have different intellectual strengths, and lexicography is not one of mine.  If I had to say what my strengths were, they would all revolve around syntactic analysis. For example, I love writing grammars. I do not find it difficult. Rather, it is like a giant puzzle, where all the pieces fall into place, and a few simple principles determine how they fit together. I enjoy working out a syntactic puzzle. Writing a dictionary requires a completely different mentality. People who do something similar might be accountants, map designers or book editors. These are people who have to keep track of thousands of complicated details, to be able to detect inconsistencies and connections of various sorts, people who have incredible attention to detail, people who enjoy new and unexpected cases and revel in their variety. That person is not me.

I had no idea was I was doing.

Another problem that I faced in writing the dictionary is that I had no idea of what I was doing. In graduate school, I had never taken a course on lexicography. What I knew about lexicography came uniquely from looking at other dictionaries for related languages, and trying to figure out how the authors got their information. For example, Patrick Dickens Ju|’hoansi dictionary was an important model for me. I felt like I was mostly just making things up as I went along. 

Here is an example of how ignorant I am. 

At some point, I realized that I had translated most of the example sentences with ‘he’ instead of ‘she’. But there was no need for that, since the third person singular pronoun jà in Sasi is not specified for gender, it can mean either ‘he’ or ‘she’ or ‘it’, depending on context. So I went back into the dictionary and changed the translations of all the examples to ‘he/she’. After making this change, I decided to write ‘s/he’ instead, which is a bit more streamlined. But then I realized that would also have to write ‘his/her’ and ‘him/her’. The examples started to look awkward and unnatural. I also looked at other dictionaries to see what they did in this case, and none of them used ‘he/she’ or ‘s/he’. So I went back into the dictionary yet again, and randomly changed all the instances of ‘s/he’ to either ‘he’ or ‘she’, making sure that I used each pronoun about half of the time. All this took quite a bit of time and thought. 

There were hundreds of decisions like this. Decisions about how to write something down which affected hundreds of words in the dictionary. So any such decision led to massive dictionary edits to ensure uniformity of the words. Since I am a perfectionist, these kinds of errors drove me up the wall. I break into a cold sweat and become anxious just thinking about them. Once again, it is clear I am not a natural born lexicographer.

But I do not regret writing the Sasi dictionary. 

The community of speakers has been especially supportive all along, talking about the ‘big book’ we are working on, and referring to our sessions and ‘Sasi school’.  In a sense, I am doing this work for them, since a small descriptive dictionary like this will not attract very much attention in the field of linguistics. But in my heart, I know that it was the right thing to do, since I was the only one in a position to do it. I was the linguist on whom this task landed. I was compelled to take it up, and I did.

What are the scientific benefits of this work?

There are lots of possible applications of this dictionary to historical research on Khoisan languages. 

A general question concerning the Khoisan languages is how they are related to one another. What does the historical tree of Khoisan languages look like? How did these languages interact historically? Did they borrow words from one another? What do the answers to these questions tell us about the historical movement of the various populations and their social connections? 

For example, Sasi has many words that are clearly from Khoe-Kwadi. The question is which Khoe-Kwadi languages did Sasi borrow the words from? And at which time period did it borrow those words? And what does this borrowing tell us about the history of the region.

As I mentioned, ǂHoan and Sasi are quite similar, and both are distantly related to Juǀ’hoansi. There has been serious historical work comparing ǂHoan and Juǀ’hoansi (leading to the language family Kx’a), but this work left out Sasi since materials were not available yet. So a feasible project at this moment would be to reconstruct proto-Kx’a partly on the basis of the new Sasi dictionary. Such a reconstruction could also give us information about the interaction of proto-Kx’a (northern Khoisan) with proto-Khoe-Kwadi (central Khoisan) and proto-Tuu (southern Khoisan).

Another project is to write up a Sasi grammatical sketch. There are so many example sentences in the dictionary and oral texts that they would provide excellent material on which to base a grammatical sketch.

A last project is to look into the Sasi click inventory more deeply. The speakers of Sasi have a tendency to let the tongue hit the bottom of their mouth during the production of the clicks. I call this ‘flopping’. I have not heard this with other speakers of Khoisan languages that I have looked at (e.g., Nǀuu, ǂHoan, Kua, Cua, Tshila). So it would be interesting to look at the phenomenon more systematically.