The Thai Keyboard Is Sorted by Frequency, Not the Alphabet
TL;DR: The Thai keyboard is not sorted by the alphabet. Measured against a 626-word beginner corpus, it is sorted by how often a character is used: the fourteen most frequent characters need no shift at all, twelve of the fourteen rarest do, and the two common tone marks sit on the home row under one finger while the two rare ones hide on shift. Typing is a harder test of reading than recognition because it makes you commit to which mark.
The home row is not the start of the alphabet
Windows lists the standard Thai layout as Thai Kedmanee and the alternative as Thai Pattachote; macOS ships the same pair, the second one named PattaChote outright. The numbers below come from reading both layout files directly and counting them against the 626 Thai words in this app's own corpus, which between them contain 1,896 characters.
Here is the Kedmanee home row, left to right, with each character's frequency rank in that corpus:
| Key | Thai | Rank | Uses | Finger |
|---|---|---|---|---|
| a | ฟ | 43 | 6 | left little |
| s | ห | 18 | 50 | left ring |
| d | ก | 2 | 111 | left middle |
| f | ด | 6 | 81 | left index |
| g | เ | 16 | 53 | left index |
| h | ้ | 11 | 57 | right index |
| j | ่ | 7 | 70 | right index |
| k | า | 1 | 153 | right middle |
| l | ส | 22 | 32 | right ring |
| ; | ว | 13 | 54 | right little |
| ' | ง | 5 | 92 | right little |
Ten of those eleven characters are in the corpus top 22. The odd one out is ฟ, at rank 43, sitting on the weakest finger of the row. Between them the eleven account for 759 of the 1,896 characters in the corpus, or 40 per cent of everything typed, from eleven keys.
Nothing about that row is alphabetical. ก is there because it is the second most used character, not because it is first in the alphabet. ข, which is second in the alphabet, is out on the number row next to the zero.
What it is sorted by instead
Split the 56 Thai characters that appear in the corpus into four groups by frequency and count how many of each group need the shift key. The pattern is not subtle.
| Group | Share of everything typed | How many need shift |
|---|---|---|
| 14 most used | 1,119 of 1,896 (59%) | 0 of 14 |
| next 14 | 516 (27%) | 1 of 14 |
| next 14 | 213 (11%) | 2 of 14 |
| 14 least used | 48 (2.5%) | 12 of 14 |
Across the whole corpus, 1,801 of 1,896 characters can be typed without touching shift: 95 per cent. The shift level is not a second alphabet. It is an attic, and Thai orthography obligingly supplies things to put in it.
The clearest case is the backslash key, out past the right little finger. Its base level is ฃ and its shift level is ฅ, the two consonants that dropped out of use. Neither appears once in 626 words. One of the worst keys on the board holds the two letters nobody needs, which is exactly where a frequency-sorted layout would put them.
The tone marks give the game away
Thai has four tone marks. Two of them are everywhere and two are rare, and the layout knows it.
| Mark | Name | Key | Uses in the corpus |
|---|---|---|---|
| ่ | mai ek | j, home row, no shift | 70 |
| ้ | mai tho | h, home row, no shift | 57 |
| ๊ | mai tri | u with shift | 4 |
| ๋ | mai chattawa | j with shift | 3 |
h and j are the two keys the right index finger already rests on. The two marks a learner will type hundreds of times are under that finger, unshifted, and the rarest of the four is the shifted version of the commonest. That is not a coincidence anyone needs to argue for. It is also the reason mai ek and mai tho are the two tone marks worth learning first: they are the two the language actually uses.
The vowels barely split at all
It is often said that the vowels are scattered across both shift levels. On Kedmanee they are not. Of the fifteen vowel symbols, thirteen sit on the base level and two need shift: ู and โ. That is the entire split.
The interesting one is ู, because of what shares its key. Press the 6 key and you get ุ, the short u. Press shift and 6 and you get ู, the long one. The keyboard has quietly encoded the short and long pairing as one key with two levels, and it has put the commoner member of the pair on the easier level. In the corpus ุ appears 26 times and ู appears 25, so in this case the layout is arbitrating a near-tie.
The same key-sharing runs through the marks: ิ and the phinthu on b, ั and the nikhahit on y, ี and mai tri on u, ้ and mai taikhu on h. Learning where one member of a pair lives tells you where the other one does.
Typing turns a picture into a sequence
A Thai syllable is arranged in two dimensions. Vowels can sit before, after, above or below the consonant, and a tone mark stacks on top of whatever is already up there. Reading it means taking that arrangement in at a glance. Typing it means producing one strict sequence of keys, and the sequence is not the arrangement.
| Word | Characters, in typing order | Keys |
|---|---|---|
| เข้า | เ ข ้ า | g, minus, h, k |
| ที่ | ท ี ่ | m, u, j |
| ให้ | ใ ห ้ | full stop, s, h |
| อยู่ | อ ย ู ่ | v, p, shift and 6, j |
| เนื้อ | เ น ื ้ อ | g, o, n, h, v |
Look at เข้า. The leading vowel เ is typed first even though it is pronounced after the consonant. The tone mark ้ sits directly above ข but is typed after it and before า. In อยู่ there are two marks on one consonant, the vowel underneath first and the tone mark above it second. In เนื้อ the sequence runs vowel, consonant, vowel, tone mark, consonant, which is nothing at all like the shape on the screen.
None of that is arbitrary. The typing order is the order the syllable is stored in, and that order encodes the structure: which consonant is the initial, which vowel belongs to it, which mark applies to what. Reading lets you skip the question. Typing does not.
Why this is a harder test than recognition
Recognising a word lets you be vague and still be right. You can see ที่ and know it says thi without ever asking which tone mark is on it. The shape is familiar as a whole, and the mark is part of the shape rather than a decision.
Typing removes that option. To produce ที่ you must choose between h and j, which is to say between mai tho and mai ek, which is to say between two tones. The two keys are adjacent, both under the right index finger, and the marks themselves differ by one extra loop. Getting them the wrong way round is a one-key slip and a completely different word.
- Recognition asks: have I seen this before? Typing asks: what exactly is this made of?
- Recognition tolerates a fuzzy memory of the marks. Typing forces every mark to be named.
- Recognition works left to right on the page. Typing works in storage order, which is the syllable's structure.
The practical use of this is as a self-test rather than a skill in itself. If you can read a word but cannot type it, what you have is a picture, not a decoding. It is the same gap that confusable letter shapes expose from the other direction.
Phayan teaches reading rather than typing: it introduces the letters one at a time from zero, puts audio on every word, and keeps progress on the device without an account. Typing is a useful audit of what that has actually left behind. Take a word the app has taught you, open any text field with a Thai keyboard, and try to produce it without looking.
Pattachote exists, and it is not the fix
The second Thai layout was designed to be a better one. Measured against the same corpus it is a wash and arguably worse for a beginner: Pattachote puts 43.4 per cent of characters on the home row against Kedmanee's 42.5, but it needs the shift key for 10.1 per cent of them against Kedmanee's 5.0.
For a learner that second number is the one that matters, because shift is where the errors live. Neither layout is a reading tool and neither is worth switching to for language reasons. The one thing the comparison does establish is that Kedmanee's edge is not where people assume: it is not a better home row, it is a better shift level.
Frequently asked questions
Why is the Thai keyboard not in alphabetical order?
Because it is sorted by how often each character is used. Measured against a 626-word beginner corpus, the fourteen most frequent Thai characters all sit on the base level and 95 per cent of everything typed needs no shift key at all, while twelve of the fourteen rarest characters are on shift. ก is on a home key because it is the second most used character, not because it is first in the alphabet, and ข is out on the number row. Phayan teaches the letters in a reading order rather than a typing one, so the keyboard is worth reading as evidence about frequency rather than as a syllabus.
Which Thai tone marks need the shift key?
Mai tri and mai chattawa, the two rare ones. Mai ek and mai tho are on the home row under the right index finger, on the h and j keys, with no shift needed. In the corpus behind this post mai ek appears 70 times and mai tho 57, against 4 and 3 for the other two, so the layout is simply reflecting the language. Mai chattawa is the shifted version of mai ek on the same key. Phayan introduces the two common marks well before the rare pair for the same reason.
What order do you type a Thai syllable in?
In storage order, which is not the order the syllable is arranged on screen. เข้า is typed as เ then ข then the tone mark then า, so the leading vowel comes first even though it is pronounced after the consonant, and the mark that sits above ข is typed after it. อยู่ is typed อ, ย, the vowel underneath, then the tone mark above. That sequence encodes which consonant is the initial and which mark belongs to what, which is why typing is a stricter test of decoding than recognition. Phayan does not drill typing, but a word it has taught you should be typeable if you actually decoded it.
Is Pattachote a better layout than Kedmanee?
Not for a learner, on the evidence available here. Against the same 626-word corpus, Pattachote puts marginally more characters on the home row, 43.4 per cent against 42.5, but it needs the shift key roughly twice as often, 10.1 per cent of characters against 5.0. Shift is where typing errors concentrate, so the trade goes the wrong way early on. Neither layout teaches anything about reading, and Phayan works the same whichever one is installed, since progress is kept on the device rather than tied to a system setting.
Does learning to type Thai help you read it?
It does not teach reading, but it audits it honestly. Recognition lets a word stay a picture: you can know ที่ says thi without ever deciding which tone mark is on it. Typing makes that decision compulsory, because h and j are different keys and different tones. Use it as a spot check rather than a study method. Phayan handles the reading side, one letter at a time from zero with audio on every word, and an Anki deck with type-in answers is a reasonable way to test the same words back with the keyboard; ThaiPod101 will not help here, since audio lessons never ask you to produce the script.