The Alphabet That Will Save a People From Disappearing

Interesting article at the Atlantic on the creation of a new alphabet for the Fulani language.

Marred by a slight instance of cultural chauvinism: I snorted my milk at the following sentence: "But unlike Arabic, whose short vowels are written as diacritical marks above and below letters, the script assigned its five vowels proper letters." Arabic vowels may be different from Latin vowels; that hardly makes them "improper."

I also learned a useful marketing idea for those creating a new alphabet: Choose a catchy word with no repeating letters; make those letters the first of your alphabet; then take that same word to be the name of your alphabet (just as "alphabet"="Alpha"+"Beta", the first two letters of the Greek alphabet). Such is the case with "Adlam", the new alphabet described here.

Starting up a New Language: Some Case Studies

Photo by Ryan Buterbaugh 

How best to tackle a new language depends on many factors:

1. How difficult do you expect it to be?

2. Is a new script involved?

3. What's your immediate goal? Your ultimate goal?

4. What tools are available?

5. How much time do you have available?

Herewith some examples of how I chose to attack some new languages.

Swahili. (First of all, my answers to the questions above: 1: easy. 2: no. 3: just want to make incremental progress. 4: fair number of courses and books available. 5: almost none.)

The approach: I use the Language/30 Swahili course and an old edition of Teach Yourself Swahili, which gives a bare-bones presentation of grammar and vocabulary.  I use Audacity to clip audio sentences from the Language/30 course and put the sentences with text into Anki flashcards. I create additional cards with words and sentences from the Teach Yourself book. The Teach Yourself cards have no audio, unless and until I find a good source for ad-hoc Swahili audio.

But Swahili pronunciation is pretty easy. The Language/30 samples should provide ample pronunciation practice, as well as providing a storehouse of useful sentences "in the bank." So far, at the rate of one new card per day (one new word or sentence every two or three days) it takes less than a minute a day, plus occasional prep time to create new cards.

Thai. (1: moderate to difficult. 2: yes. 3: just want to make incremental progress. 4: fair number of courses and books available. 5: very little.)

The approach. First I designed some mnemonics to help me memorize the Thai alphabet. I spent about a minute a day for some months just practicing the letters. I made Anki cards from the first few reading lessons of the Thai Pimsleur course, mostly just nonsense syllables, but it helped me get a foothold on reading. I continued with the Teach Yourself Thai book (with accompanying CD's--some form of audio companion is a huge advantage). Sample sentences from the book go onto flashcards, while I download audio for individual words from Google translate. (Note: I don't do this for Swahili because Google Translate's Swahili voice is a weird robot which I have no desire to emulate.) I also picked up Stu Jay Raj's unusual book Cracking Thai Fundamentals and started reading through it and making flash cards as necessary (it has less memory-intensive content). And by dint of some effort, I devised a streamlined formulation of the rules for determining tones from the Thai script.

Generally speaking, for each word, phrase, or sentence I have as many as five cards:

(1 "English to Thai") Input: English. Output: Thai audio.
(2 "Reading") Input: Thai script. Output: English and incidentally Thai audio.
(3 "Listening") Input: Thai audio. Output: English and incidentally Thai script.
(4 "Spelling") Input: English and Thai audio. Output: Thai script.

In line with a principle that good flash cards should expect a single response, the term "incidentally" here indicates that the card presents the information as part of the response but I need not reproduce it as a "correct" response.

Only a few cards are of type 4. I am ramping up gradually to writing Thai. I use conditional compilation for Anki to generate these cards only when I specify.

On the other hand, I like the daily exposure to reading Thai.

I currently spend about four minutes a day working through my Thai Anki deck, adding new cards from the Teach Yourself book, Cracking Thai Fundamentals, or occasionally the Pimsleur reading lessons as necessary.

Starting up Khmer and Burmese was very similar, the main difference being a relative paucity of available resources.

Sumerian. (1: moderate. 2: yes. 3: finish Hayes book [see below]. 4. Chiefly books. 5. very little [see a trend here?])

Sumerian is a dead language, which entails both advantages and disadvantages.


(1) You need not be concerned about acquiring a perfect accent. For most dead languages, scholars are happy to inform you about the pronunciation with considerable detail and subtlety, but I'm always a little skeptical that they know quite as much as they think they do. Whether or not they really know what they're talking about, there's no harm in making things easy on yourself. This does not mean pronunciation can be ignored. I have yet to come across a writing system in which sound does not play some partial role at least. In the case of Sumerian, the writing is a combination of ideographic and phonetic symbols. Many words and phrases have alternate spellings. Having a crude idea of the pronunciation makes it much easier to recognize these.

(2) Likewise, you may choose to (or be forced to) downplay skills of writing and listening in favor of reading. Focusing on just one of the four standard language skills certainly simplifies matters.

For Sumerian, I depended on a single resource: John Hayes' book A Manual of Sumerian Grammar and TextsIncidentally I see the hardcover edition of the book is now going for $4000. which makes me feel strangely wealthy. The book provides many samples of Sumerian texts, with vocabulary lists and extensive discussions of each.

Sumerian is written in a cuneiform script, which my computer does not support. Certainly the computer would not reproduce the variety of sizes and shapes of the historical texts. Here Anki's ability to handle images facilitates greatly. By scanning a page, and clipping out samples of text, I can reproduce any word or phrase from the book.

Generally I use two Anki cards for each item of Sumerian text:

(1 "Reading") Input: Sumerian text. Output: the transliteration (phonetic reading) of the text.
(2 "English") Input: Sumerian text and transliteration. Output: the English translation.

I also use Anki Cloze cards for grammar rules, historical facts, etc.

This approach lets me work through at a steady (if slow) pace, while committing to heart the memorization-related content and reading some phrases of text (while reviewing Anki cards) every day.

"Nude" and "Flesh" Are Not Colors

A photo posted by Chyrstyn Mariah Fentroy (@chyrstynmariah) on

In the "institutional/pervasive racism" category for today, note the photo above, courtesy of Chyrstyn Mariah Fentroy, who, based on the evidence of the photo above: (1) is a ballet dancer, and (2) has a richly-hued complexion. The story told by the photo is that apparently it is impossible to buy ballet slippers in any color other than off-white. If your natural skin shade is something other than off-white, then you are doomed to hand-paint every new pair of ballet slippers that you buy. (Apparently ballet slippers that contrast with your skin are taboo.)

I sympathize with all ballet dancers who have to deal with this. It is something of a burden to have to do this all the time. More important the situation sends a message that you're the wrong color to be doing this. Sure, you can shrug it off, but at some point everyone experiences shrug fatigue.

I learned of this particular issue thanks to a Huffington Post article by Katherine Brooks.  Ms. Brooks is a serious contender for the Lack of Self-Awareness Awards. Note how the article becomes incomprehensible halfway through thanks to the profligate use of the term "flesh tone." What does this mean? Is it the pale peach one finds in a box of crayons? I literally could not follow the sense of the article once the term "flesh tone" is bandied about with reference to a diverse cast of characters.

Why are people still talking about "flesh tone"? I learned back in 1975 that this was uncool:

Marilyn Goes to Poland!

Marek P. has translated the Marilyn Method for memorizing the pronunciation of Chinese characters into Polish for his site Zyskiwanie Przewagi.

More Arabic Mnemonics

This is a companion to my earlier post on mnemonics for Arabic roots. An interesting feature of Arabic is that most words can be decomposed into a root and a pattern. The root consists strictly of consonants (usually three), while the pattern consists of vowels plus possible additional consonants.

In English we can see something of how this works by considering the words sing, sang, sung, song, singer. We could think of these composed of an underlying root s-ng overlaid on various patterns 1i2, 1a2, 1u2, 1o2, 1i2er.

We could combine some of these patterns with a different root dr-k to yield new words: drink, drank, drunk, drinker (but no dronk, unfortunately).

To repeat my earlier example, in Arabic, there is a root ك ت ب (k-t-b) which has to do generally with writing. Some of the words formed from this root by fitting it into different patterns would be:

كَتَبَ (kataba = k-t-b × 1a2a3a) "he wrote"
كِتَاب (kitaab = k-t-b × 1i2aa3) "book"
مَكْتَبَة (maktaba = k-t-b × ma12a3a) "library"

Most (not all) Arabic roots consist of exactly three consonants. The earlier post described a fairly complicated method for encoding these three consonants in an English word of phrase. This is not so simple because Arabic distinguishes many consonant sounds which English doesn't.

To be honest, the previous method is a sledgehammer, really important only for the most stubborn cases. Many roots (like the forementioned k-t-b) can stick in the mind with modest effortespecially if you remember just one of the derived words (such as "book").

In putting this method into practice, I realized that the puzzle needed another piecean idea I borrowed from Heisig's method for Chinese characters. This extra piece is an English name for the root, analogous to Heisig's keywords. In the case of Arabic, I don't call this name a "keyword" but rather a "syndrome." For example, the syndrome for the forementioned k-t-b root is "write."

One-by-one I peruse my Arabic dictionary (which groups together words sharing the same root), and do my best to choose an English syndrome which comes closest to capturing the spectrum of meanings associated with the root. I don't do this for every Arabic word I encounter, but only for those that I find particularly hard to remember.

For example, one pair of words that gave me trouble was:

طابِق (Taabiq), plural طَوابِق (Tawaabiq): a floor or story of a building, versus:طَبَق (Tabaq), plural أطباق ('aTbaaq): a dish or course of a meal.

Both share the root طبق (T-b-q), but I could see scant connection in the meanings. But perusing the various words associated with this root, it seems most are related to the idea of one thing on top of another. The floors of a building are arranged this way, as is the lid for a dish. The idea of a "dish" of a meal seems to be derived from the latter. I therefore chose the word "superpose" as the syndrome for the T-b-q root.

Here's a list of some of the syndromes I have selected thus far, along with a sample word for each.

هجر (h-j-r)
هاجَرَ haajara (migrated)
فصل (f-S-l)
فَصْل faSl (to dismiss, fire)
شرف (sh-r-f)
يُشْرِف yushrif (supervises)
حدد (H-d-d)
حّدِيِد Hadiid (iron)
fan out
نشر (n-sh-r)
نَشَرَ nashara (published)
خول (kh-w-l)
أَخْوال 'akhwaal (maternal uncles)

This leaves the issue of the pattern. How do I remember that طابِق (Taabiq) is a building story and طَبَق (Tabaq) is a dish rather than vice-versa? (This particular pair was a real issue for me.) These words have the same roots but different patterns. The Arabic grammarians had the clever idea of using the root فعل (f-`-l) "do" as a "neutral" root. They then described a given pattern by applying it to this "neutral" root. For طابِق (Taabiq), the pattern of vowels is (_aa_i_). For طَبَق (Tabaq), the pattern of vowels is (_a_a_). The Arabic terms for these patterns would be فّاعِل (faa`il) and فَعَل (fa`al), respectively.  

For the most part I handle this with a very simple device (also highly adaptable to many other situations). For a given pattern, I choose just one example of the pattern to use as a mnemonic hook for the pattern itself. For example consider the two patterns just mentioned: _aa_i_ and _a_a_. The former also appears in the word هاتِف haatif "telephone" and the latter in the word بَصَل baSal "onions". I use these words to represent their associated patterns. Thus the word طابِق (Taabiq) is broken down into superpose × telephone (root × pattern) while طَبَق (Tabaq) is broken down into superpose × onions. And my trouble in keeping this straight is resolved by creating whatever mental images serve to associate telephone with story of a building and onions with dish

In choosing a keyword to represent a given pattern, I try to choose something concrete, visualizable, and as distinct as possible from other keywords.

Here's a list of patterns and their associated keywords. Both this and the previous list are works in progress. I add words as I find I need them.

عَباءة `abaa'ah wool cloak
فَعَالَة _a_aa_ah
مُدُن  mudun towns
فُعُل _u_u_
صَحْن SaHn plate
فَعْل _a__
فُلُوس fuluus money
فُعُول _u_uu_
بِدَل bidal suits
فِعَل _i_a_
أَرْجُل 'arjul legs
أَفْعُل ’a__u_
مَسْرَح masraH theater
مَفْعَل ma__a_
بَصَل baSal onion
فَعَل _a_a_
قُفَّاز quffaaz gloves
فُعَّال _u_2aa_ (2 indicates double vowel)
رِسالة risaalah letter
فِعَالَة _i_aa_ah
سِجْن sijn jail
فِعْل _i__
خَلِيج khaliij gulf
فَعِيل _a_ii_
أَنْهَار 'anhaar rivers
أَفْعَال ’a__aa_
سَتائِر sataa'ir curtains
فَعَائِل _a_aa’i_
رِياح riyaaH winds
فِعَال _i_aa_
قُفْل qufl lock
فُعْل _u__
مَنْزِل manzil house
مَفْعِل ma__i_
فَواكِه fawaakih fruits
فَواعِل _awaa_i_
عَرُوس `aruus bride
فَعُول _a_uu_
هاتِف haatif telephone
فَاعِل _aa_i_

A couple more examples: طَوابِق (Tawaabiq), plural of طابِق (Taabiq), breaks down into superpose × fruits (T-b-q  × _awaa_i_), whereas أطباق ('aTbaaq), plural of طَبَق (Tabaq), breaks down into superpose × rivers (T-b-q × 'a__a_).

The Arabic verb forms get special treatment. Each verb form is a coordinated set of patterns describing the various verb forms. These are numbered from one I through ten X. Thus for a Form-IV verb, you know not only that the past-tense pattern is i__aa_a, but the present-tense pattern is yu__i_.  More information is available in an Arabic grammar book or various places on the Internet.

For the verb forms, rather than the example method I use for other patterns, I chose a code word for each form from II through X:

Form II
Form III
maiko, or mikado
Form IV
Form V
Form VI
Form VII
Form IX
Form X

I was guided by the Major system in choosing the code words, but that is not really important. Almost any set of visualizable code words works just as well once committed to memory. I also chose words with Japanese connotations. You might wonder why. Why not choose words with Arabic connotations—
sultan rather than sushi, for example? The reason is that sultan may well turn out to be a specific vocabulary item to be memorized. Using the same word as a code word creates a minor possibility for confusion. 

You notice also that I have two different code words for Form III. This is to distinguish the two versions of Form III that have different patterns for the "verbal noun". Maiko is used for those verbs with a verbal noun of pattern مُفَاعَلة (mu_aa_a_ah). Mikado is used for those verbs with a verbal noun of pattern فَعَالٌ (_a_aa_).

Form I presents a more complicated situation. Any of the three Arabic vowels a, i, u may appear in the past-tense form, and any of the three may also appear in the present-tense form (although not all combinations appear). The Form I code words are chosen to describe these vowels as well:

tatoo: past a, non-past u
taxi: past a, non-past i
tatami: past a, non-past a
titan: past i, non-past a
tutu: past u, non-past u

(BTW in researching this post I learned that there are actually fifteen verb forms, but forms XI to XV are extremely rare. The system is easy to extend in any case.)