Introduction
Asaat script (Chinese: 阿剎字; Asaat: อ้าสาดชี้; Jyutping: aa3 saat3 zi6) is my attempt to adapt the Thai script for writing Cantonese. Why? There are a few reasons. The simplest reason is that Thai and Cantonese are my two favorite natural languages, and creating a script that connects them seemed both fun and funny. The other two reasons are more complex and can be broadly summarized as 1) creating a practical tool to aid in learning Cantonese, particularly its more challenging aspects, and 2) engaging in a bit of linguistic experimentation.
I am reasonably proficient in three tonal languages: Mandarin, Cantonese, and Thai. My proficiency ranking would be Mandarin > Thai > Cantonese, but if I were to rank the difficulty (learnability) of their tone systems, it would be Cantonese > Mandarin > Thai. This is due to three main factors: symmetry, distinguishability, and orthography. Among these, the Thai tone system stands out as being exceptionally learner-friendly.
Symmetry: The Thai tone system is a masterpiece of symmetry and one of the reasons I love the language. It features five tones: a flat mid tone (M), a high falling tone (F), a low rising tone (R), a high rising tone (H), and a low falling tone (L). In essence, there’s a central mid tone, and the remaining tones are like mirrored contours around it. This symmetry makes the system elegant and intuitive.
In contrast, the Mandarin and Cantonese tone systems are far less symmetrical. Mandarin has four main tones and a context-dependent neutral tone: a high flat tone (1), a rising tone (2), a low dipping tone (3), and a falling tone (4). While the Mandarin system is relatively straightforward, Cantonese is much more complex, with six tones: a high flat tone (1), a medium rising tone (2), a medium flat tone (3), a low falling tone (4), a low rising tone (5), and a low flat tone (6). The Cantonese tones lack symmetry, with their contours scattered and unbalanced, requiring each tone to be memorized individually. In contrast, the Thai system’s mirrored structure allows learners to internalize fewer patterns and infer the rest.
Distinguishability: Distinguishability measures how different tones are from one another. The greater the difference, the easier the tones are to distinguish and learn. Quantitatively, I calculate this as the mean pairwise difference in amplitude and slope between tones. The results are striking: 28.57 for Mandarin, 23.76 for Thai, and 13.86 for Cantonese.
Expand for details on calculating distinguishability.
First I represent each tone contour C as a function of pitch over time within a certain domain. In this case, I chose five levels of pitch p ∈ [0, 4] across five units of time t ∈ [0, 4]. For each language, you have a set of contours that constitute the tone system.
$$L = \{C_1, C_2, \ldots, C_n\}$$
Next, we want to calculate both the difference in pitch or amplitude between two contours as well as their difference in slope. The logic here being, the farther apart two tones are and the less similar their changes in pitch, the more distinguishable they are. We combine these measures and integrate over the whole domain, which sums these differences at every point, deriving the distinguishability measure D between any two tones.
$$\Delta C_{ij} = (C_i – C_j)^2$$
$$\Delta C_{ij}’ = (C_i’ – C_j’)^2$$
$$D_{ij} = \int_{t_{\text{min}}}^{t_{\text{max}}} [\Delta C_{ij} + \Delta C_{ij}’] \, dt$$
To get a single measure that represents the overall distinguishability of the entire set of contours in a given language, we sum all pairwise measures, where n represents the number of tones in the system, and take the mean, thus giving us the language-wide tone distinguishability measure.
$$D_{total} = \sum_{1 \leq i \leq j \leq n} D_{ij}$$
$$D_{mean} = \frac{D_{total}}{\binom{n}{2}}$$
Cantonese tones are far more similar to one another, making them harder to distinguish. When comparing more tonal languages (see below), not only does Cantonese rank near the bottom, but it also has the lowest minimum distinguishability score between two tones (1.13), tied only with Taishanese. Even native speakers often confuse certain tones, and recent studies (e.g., Fung and Lee 2019) highlight ongoing tone mergers in Cantonese. This inherent difficulty makes the learning of tones for Cantonese particularly challenging compared to Thai and Mandarin.
Expand to see the distinguishability values for languages.
The mean is provided for each language along with the range in parantheses. Note that I purposefully excluded languages with tonal systems that rely on phonation (e.g., Hmong) or that maintain the voicing split from Middle Chinese (e.g., Wu varieties of Chinese). Such languages rely on additional dimensions (beyond just pitch over time) which would naturally affect their distinguishability; although, the extent to which the languages below are purely pitch-reliant is also debatable.
| Language | Family | Tones | Distinguishability |
|---|---|---|---|
| Mandarin (Beijing) | Sino-Tibetan | 4 | 28.57 (11.61, 55.47) |
| Mandarin (Chengdu) | Sino-Tibetan | 4 | 32.61 (1.53, 55.47) |
| Mandarin (Taipei) | Sino-Tibetan | 4 | 15.81 (4.53, 24.53) |
| Yue (Cantonese) | Sino-Tibetan | 6 | 13.86 (1.13, 45.13) |
| Yue (Taishanese) | Sino-Tibetan | 5 | 15.15 (1.13, 45.13) |
| Hakka (Meixian) | Sino-Tibetan | 4 | 19.24 (3.20, 36.00) |
| Southern Min (Taipei) | Sino-Tibetan | 5 | 18.88 (3.20, 45.13) |
| Thai | Kra-Dai, Tai | 5 | 23.76 (5.33, 47.83) |
| Zhuang | Kra-Dai, Tai | 6 | 11.52 (3.2, 31.2) |
| Kam | Kra-Dai, Kam-Sui | 9 | 18.97 (1.80, 64.00) |
| Sui | Kra-Dai, Kam-Sui | 6 | 16.85 (4.53, 47.20) |
Orthography: For language learning, orthography plays a vital role in reinforcing phonetics, including tones. Mandarin and Cantonese, written in Chinese characters, provide minimal phonetic or tonal information. While input methods like pinyin or jyutping offer some phonetic encoding, they often neglect tones. Shape-based methods like Cangjie eliminate phonetic considerations entirely, while Zhuyin systems may or may not require tonal input. In short, typing or writing in Mandarin or Cantonese doesn’t consistently reinforce tones and their lexical associations.
Thai, on the other hand, deeply embeds tone in its orthography. Tone is encoded through a combination of consonants, vowels, and diacritics, creating a system where nearly every part of the word contributes to encoding the tone. This means that every word you write or type in Thai reinforces the tone associated with that word. When you visualize a word in Chinese characters, you think of its shape and meaning, but with Thai script, tone is an inseparable part of the word’s identity in the orthography. Admittedly, I sometimes forget what tone a morpheme has in Mandarin or Cantonese; but in Thai, this very rarely happens, because a very important part of learning words in Thai is learning how to spell them.
This feature of Thai orthography significantly enhances tone learning and retention. Adapting the Thai script for Cantonese, a language with a notoriously complex tone system, could therefore help learners associate tones with lexical items more effectively. The Asaat Script does this by combining Thai’s tone-encoding principles with a bit of historical Chinese linguistics.
Thai Script and Encoding Tone
In Old Thai, the script originally had a 1:1 sound-to-letter mapping excluding letters used to represent sounds in Pali that did not exist in Thai. Part of the phonology of Thai at that point involved a three-way tone distinction (written using tone diacritics), a two-way voiced vs. voiceless distinction among all fricative and sonorant consonants, and up to a four-way distinction among stops and affricates. Eventually, some of these voicing distinctions were lost and resulted in a new set of tonal distinctions, that is, a tone split. In other words, tones had different realizations dependent on whether the preceding consonant was originally voiced or voiceless. More specifically, plain voiced stops merged with voiceless aspirated stops, voiced fricatives merged with voiceless fricatives, and voiceless sonorants became voiced. The spelling of these consonants remained the same and instead came to be used to encode the new tonal distinctions, resulting in many letters representing the same consonant sound but belonging to different consonant classes depending on the tone encoding rules they elicit. This is a bit of an oversimplification, as later tone mergers and distinctions arising from vowel length complicate things, but this is more or less what ties the encoding of tone so closely to spelling in Thai.
Expand to see a flowchart detailing the rules for determining the tone of a Thai syllable.
The Four Tones of Middle Chinese and Tone Split
The modern tones in Cantonese are the result of a similar tone split that happened in Middle Chinese. Originally, Middle Chinese had four tones- level (平), rising (上), departing (去) and entering (入) – as well as a three-way voicing distinction for stops and affricates and a two-way voicing distinction for fricatives. Voiced stops/affricates and fricatives both underwent devoicing, the former merging with aspirated stops/affricates in level tone syllables and with unaspirated stops/affricates elsewhere. This resulted in a tone split where each of the four tones developed a dark (陰) variant and a light variant (陽). In addition to the split caused by voicing, Cantonese further split the dark entering tone according to vowel length. The table below details the tone splits leading to the modern Cantonese tone system.
| Middle Chinese | Cantonese | |||
| Tone | Initial | Nucleus | Tone (Historical) | Number (Phonetic) |
| Level | voiceless | dark level | 1 | |
| voiced | light level | 4 | ||
| Rising | voiceless | dark rising | 2 | |
| voiced | light rising | 5 | ||
| Departing | voiceless | dark departing | 3 | |
| voiced | light departing | 6 | ||
| Entering | voiceless | short | upper dark entering | 1 |
| long | lower dark entering | 3 | ||
| voiced | light entering | 6 | ||
Asaat Script
The Asaat script consists of 25 consonant letters representing 18 distinct consonant sounds, and 11 vowel symbol elements representing 8~11 distinct vowel sounds.
Consonants and Initials
Consonants are divided into two classes – dark (陰 jam1) and light (陽 joeng4), corresponding to the tone split of the four Middle Chinese tones. These class designations reflect phonetic qualities of the sounds that were present in Middle Chinese – namely whether the consonant was voiceless (corresponding to dark) or voiced (corresponding to light). Similarly to the Thai script, each consonant is associated with an acrophonic Cantonese word that starts with the same sound. The table below lists the 25 consonant letters along with their names, classes, and corresponding phonemes using the IPA.
| Symbol | Name | IPA | Class | |||
| Asaat | Jyutping | Chinese | Meaning | |||
ก | ก ไก | go1 gai1 | 雞 | chicken | /k/ | dark |
ข | ข ไข | ko1 kai1 | 溪 | stream | /kʰ/ | dark |
ค | ค คัม | ko4 kam4 | 琴 | zither | /kʰ~k/ | light |
ง | ง เงา | ngo4 ngau4 | 牛 | ox | /ŋ/ | light |
จ | จ จือ | zo1 zyu1 | 豬 | pig | /ts/ | dark |
ฉ | ฉ ฉืน | co1 cyun1 | 村 | village | /tsʰ/ | dark |
ช | ช ชา | co4 caa4 | 茶 | tea | /tsʰ~ts/ | light |
ซ | ซ แซ | so4 se4 | 蛇 | snake | /s/ | light |
ต | ต ตว | do1 dou1 | 刀 | knife | /t/ | dark |
ถ | ถ ถ้ว | to1 tou3 | 兔 | rabbit | /tʰ/ | dark |
ท | ท ทอย | to4 toi4 | 台 | table | /tʰ~t/ | light |
น | น เน่ย | no4 neoi5 | 女 | woman | /n/ | light |
ป | ป ปูย | bo1 bui1 | 杯 | cup | /p/ | dark |
ผ | ผ ผี้น | po1 pin3 | 片 | slice | /pʰ/ | dark |
ฝ | ฝ ฝู่ | fo1 fu2 | 虎 | tiger | /f/ | dark |
พ | พ พูน | po4 pun4 | 盆 | basin | /pʰ~p/ | light |
ฟ | ฟ ฟัต | fo4 fat6 | 佛 | Buddha | /f/ | light |
ม | ม ม่า | mo4 maa5 | 馬 | horse | /m/ | light |
ย | ย เยิง | jo4 joeng4 | 羊 | goat | /j/ | light |
ล | ล ลูง | lo4 lung4 | 龍 | dragon | /l/ | light |
ว | ว วอง | wo4 wong4 | 王 | king | /w/ | light |
ส | ส สื่อ | so1 syu2 | 鼠 | rat | /s/ | dark |
ห | ห หื่น | ho1 hyun2 | 犬 | dog | /h/ | dark |
อ | อ อูก | o1 uk1 | 屋 | house | /ʔ/ | dark |
ฮ | ฮ เฮา | ho4 hau4 | 猴 | monkey | /h/ | light |
Initial Consonant Digraphs
Initial glide consonants (i.e., ย and ว) can combine with อ (i.e., as อย and อว) to change their class from light to dark. This roughly corresponds to the glottal stop initial of Middle Chinese. For example, 日 jat6 ‘day’ written as ยัต corresponds to MC njit (the initial is voiced and therefore results in a light tone) while 一 jat1 ‘one’ written as อยัต corresponds to MC ʔjit (the initial is voiceless and therefore results in a dark tone).
Vowels and Finals
Vowel sounds and diphthongs, just like Thai script, are written using a mixture of vowel symbols on a consonant base. Vowels can go above, below, left or right of the consonant, or combinations of these places. The table below lists the various vowels and vowel combinations of the Asaat script as well as the spelling for all possible syllable finals in Cantonese.
| Main Vowel | Syllabic Consonant |
|||||||||
| /aː/ | /ɐ/ | /e/ | /ø/ | /o/ | /i/ | /y/ | /u/ | |||
| Monopthong | -า aː | แ- ɛː | เ-อ œː | -อ ɔː | -ี iː | -ือ yː | -ู uː | |||
| Dipthong | /i/ | -าย aːi | ไ- ɐi | เ- ei | เ-ย ɵy | -อย ɔːi | -ูย uːi |
|||
| /u/ | -าว aːu | เ-า ɐu | -ว ou | -ีว iːu | ||||||
| Nasal | /m/ | -าม aːm | -ัม ɐm | -ีม iːm | มุ m̩ |
|||||
| /n/ | -าน aːn | -ัน ɐn | เ-ิน ɵn | -อน ɔːn | -ีน iːn | -ืน yːn | -ูน uːn | |||
| /ŋ/ | -าง aːŋ | -ัง ɐŋ | แ-ง ɛːŋ | เ-ิง œːŋ | -อง ɔːŋ | เ-ง ɪŋ~eŋ | -ง ʊŋ~oŋ | งุ ŋ̩ |
||
| Stop | /p/ | -าป aːp | -ัป ɐp | -ีป iːp | ||||||
| /t/ | -าต aːt | -ัต ɐt | เ-ิต ɵt | -อต ɔːt | -ีต iːt | -ืต yːt | -ูต uːt |
|||
| /k/ | -าก aːk | -ัก ɐk | แ-ก ɛːk | เ-ิก œːk | -อก ɔːk | เ-ก ɪk~eŋ | -ก ʊk~ok |
|||
Tone Diacritics
There are three tone diacritics. Two of the tone diacritics correspond to the historical rising and departing tones of Middle Chinese. The third tone diacritic corresponds to changed tones, which is where a non-high level, non-mid rising tone (i.e. tones 3, 4, 5, and 6) transforms either into a mid-rising tone (tone 2) or a high level tone (tone 1). The exact realization is lexically dependent, typically occurring on compounds and reduplicated words. Additionally, changed tones are often used to mark something as diminutive (銀 งัน ngan4 ‘silver’ > งั๋น ngan2 ‘coin’), vocative (妹妹 มู้ยมู้ย mui6 mui6 > มูยมู๋ย mui4 mui2 ‘sister’), or – in the case of verbs – perfective aspect (去 เห้ย heoi3 ‘go’ > เห๋ย heoi2 ‘went’). Live syllables (i.e., open syllables and syllables ending in a sonorant) with no diacritic correspond to the level tone of Middle Chinese, while dead syllables (i.e., syllables ending in a stop) correspond to the entering tone of Middle Chinese. The table below outlines the tone diacritics and their tone realizations in different contexts.
| Symbol | Name | Syllable Type | Initial Consonant | |||
| Asaat | Jyutping | Chinese | Dark | Light | ||
| (มว) | (none) | (無) | live | 1 | 4 | |
| (มว) | (none) | (無) | dead short | 1 | 6 | |
| (มว) | (none) | (無) | dead long | 3 | 6 | |
่ | เซิ่งฟู | soeng5 fu4 | 上符 | live | 2 | 5 |
้ | เห้ยฟู | heoi3 fu4 | 去符 | live | 3 | 6 |
๋ | ปี้นฟู | bin3 fu4 | 變符 | live | 2/1 | 2/1 |
The flowchart below summarizes how to determine the tone of a Cantonese syllable written in Asaat.
Example Text: The Heart Sutra (心經)
觀自在菩薩，行深般若波羅蜜多時，照見五蘊皆空，度一切苦厄。舍利子，色不異空，空不異色，色即是空，空即是色，受想行識亦復如是。舍利子，是諸法空相，不生不滅，不垢不淨，不增不減。是故空中，無色，無受想行識，無眼耳鼻舌身意，無色聲香味觸法，無眼界乃至無意識界，無無明亦無無明盡，乃至無老死，亦無老死盡，無苦集滅道，無智亦無得。以無所得故，菩提薩埵依般若波羅蜜多故，心無罣礙；無罣礙故，無有恐怖，遠離顛倒夢想，究竟涅槃。三世諸佛依般若波羅蜜多故，得阿耨多羅三藐三菩提。故知般若波羅蜜多，是大神咒，是大明咒，是無上咒，是無等等咒，能除一切苦，真實不虛。故說般若波羅蜜多咒，即說咒曰：揭諦揭諦，波羅揭諦，波羅僧揭諦，菩提薩婆訶。