Why Thai-Specific Spaced Repetition Matters

If you’ve read anything about learning vocabulary efficiently, you’ve met spaced repetition. It’s the closest thing language learning has to a sure bet: review each word just before you’d forget it, and retention climbs dramatically for a fraction of the study time. For a language with almost no cognates for English speakers, where nearly every word is new to you, it’s the best way to learn Thai vocabulary we know of.

So why do so many people grind through thousands of Thai flashcard reviews and still freeze the moment a street vendor talks back? They optimized the wrong half of the system: a brilliant scheduling algorithm pointed at broken cards. Thai spaced repetition only works when the content of each card respects what makes Thai hard—and generic SRS, built for languages like Spanish, doesn’t.

The Algorithm Is the Easy Part

Spaced repetition has earned its reputation, and the science is settled.

The engine runs on a single idea: memories fade on a predictable curve, and reviewing material right at the edge of forgetting strengthens it far more than reviewing while it’s still fresh. The landmark evidence is Cepeda and colleagues’ 2006 meta-analysis in Psychological Bulletin, which pooled 317 experiments and found that distributing your practice over time, rather than cramming, reliably improves long-term retention—on the order of a 10-point jump in recall (roughly 47% versus 37%)—for the same total study time. We unpack the mechanics in why spaced repetition works, and it pairs tightly with the testing effect: the act of retrieving a word, not re-reading it, is what does the work.

The algorithms that schedule those reviews have gotten very good. For years the standard was SM-2, the SuperMemo algorithm from the late 1980s that also powered Anki. As of late 2023, Anki built in FSRS—the Free Spaced Repetition Scheduler—which fits a statistical memory model to your own review history. Benchmarked across hundreds of millions of real reviews, FSRS hits the same retention with roughly 20–30% fewer reviews. That’s a real, measurable improvement.

But notice what every one of these algorithms takes as a given: the card. SM-2, FSRS, the lot—they all assume that whatever sits on the front and back of your flashcard is correct and complete. They schedule. They don’t teach. And that assumption is where Thai falls through the floor.

Garbage In, Efficiently Out

A spaced repetition system is an amplifier. Point it at high-quality cards and it carves accurate Thai into long-term memory with ruthless efficiency. Point it at flawed cards and it does the same thing—just to the flaws.

This is why a learner can do everything “right,” keep a 200-day streak, clear every review, and still be unintelligible in Thailand. FSRS doesn’t know that the word it just scheduled is missing its tone, paired with a robotic text-to-speech voice, or romanized three different ways across the deck. It just sees “card answered correctly” and pushes the next review out three weeks. The system is working perfectly. The input is broken.

So the real question for any Thai learner isn’t “which app has the best algorithm?” The algorithms are largely commoditized now. The question is: does this deck’s content carry the information a tonal, space-free language with almost no English cognates requires? For Thai, that comes down to four things generic SRS routinely drops.

Gap 1: The Recall Target Has to Include the Tone

In a Spanish deck, the answer to a card is a sequence of letters. Get the letters, get the word. Thai isn’t like that. The “same” syllable carries completely different meanings depending on pitch, so a card that only asks for the consonants and vowels is testing part of the word and calling it the whole thing.

Take the textbook example:

• màiLowใหม่new (low tone)
• mâiFallingไม่not (falling tone)
• mǎiRisingไหมquestion particle (rising tone)

Three different words, one bundle of consonants and vowels. A learner who “knows mai” knows nothing useful until they know which mai. The same trap is everywhere: ม้า(máa) versus หมา(mǎa), or ข้าว(khâao) versus ขาว(khǎao).

A Thai-aware card makes the tone part of the recall target, not a footnote. When you flip the card, “you got it” should mean you produced the right melody, not just the right consonants. Generic decks that store the gloss as “mai = new/not/question (depends)” are training you to be confidently wrong, and FSRS will happily lock that in. This is the single biggest reason we treat tone as core to every card rather than metadata—more on the full tone system in our complete guide to Thai tones.

Gap 2: Silent Flashcards Can’t Teach a Tonal Language

You cannot learn to produce a pitch contour from a romanized string. You have to hear it, many times, from real humans.

This is where the most popular tools fail Thai. A deck without audio—or with synthetic text-to-speech—asks you to imagine the tone, and your English-speaking brain imagines it wrong. Worse, it imagines it consistently wrong, and spaced repetition then rehearses that fabricated tone until it’s fossilized. Unlearning a fossilized tone takes far longer than learning it correctly the first time.

The research on tone perception backs this up: adults absolutely can learn to hear and produce lexical tones, but the method that works is high-variability phonetic training—repeated exposure to many native speakers producing the contrast, not silent reading. Software is excellent at delivering this—the kind of structured, on-demand native audio that’s hard to get from a busy partner or a weekly tutor. The catch is that the audio has to be there, on every card, recorded by Thai speakers. When it’s missing, no algorithm can fill the hole.

Gap 3: Reviews Have to Train Word Segmentation

Thai is written with no spaces between words. A sentence like ฉันไปกินข้าว(chǎn bpai gin khâao) is four words run together with nothing to mark where one ends and the next begins. Reading Thai isn’t just sounding out characters—it’s learning to see the boundaries that aren’t printed.

Generic SRS ignores this completely, because no European language demands it. A standard deck shows you isolated words on tidy little cards, so you drill ขอ(khǎw) in a vacuum and feel fluent—then hit a real menu where it’s wedged between ten other syllables and you can’t find it. The skill you practiced (recognize this word, alone, in isolation) isn’t the skill you need (find this word inside an unbroken stream).

Thai-specific review fixes this by deliberately mixing isolated-word cards with cards that make you parse a real chunk: spotting a known word inside a phrase, reading a short unspaced string, recognizing where เธอ(thoe) stops and the next word starts. It’s a small design choice with a big payoff, and it’s the kind of thing only a deck built for Thai bothers to include. We go deeper on this in our thai script complete guide.

Gap 4: One Romanization, Not Three

There’s no single official way to write Thai in Latin letters. The same greeting shows up as “sawatdee,” “sawasdee,” and “sà-wàt-dii” depending on the source. If your flashcards pull from mixed decks, you end up rehearsing several conflicting spellings of one word—and spaced repetition, doing its job, strengthens all of them. You’ve turned one memory into three competing ones.

It gets worse when the romanization drops tone information, which most casual systems do. A romanization without tone marks is just a worse version of the script: it hides the one piece of data that determines meaning. A single consistent system—one that actually marks the long vowels and the tones, so เสือ(sěua), เสื้อ(sêua), and เสื่อ(sèua) look as different on the page as they sound in the mouth—means every review reinforces one accurate mental model instead of three fuzzy ones. Across hundreds of words and thousands of reviews, that consistency compounds.

The Hidden One: Interference and Card Order

There’s a fifth issue that even good Thai decks often miss, and it’s subtle. Generic SRS treats every card as an independent island. FSRS schedules เสือ(sěua) and เสื้อ(sêua) by their individual forgetting curves, with no awareness that they’re a confusable minimal pair. When two near-identical words happen to come due in the same session, back to back, you don’t strengthen both—you blur them together. Memory researchers call this interference, and tonal minimal pairs are interference machines.

Smart Thai sequencing handles this two ways. Sometimes it separates confusable pairs so they’re not reviewed in the same breath. Other times it does the opposite on purpose—presenting them in deliberate contrast so you train the discrimination directly, which is precisely what high-variability tone training is good for. Either way, it’s a decision the deck has to make about Thai, using knowledge of which words collide. A language-agnostic scheduler can’t make it, because it doesn’t know เสือ(sěua) and เสื้อ(sêua) are neighbors. This is the same logic behind interleaving and active recall—the structure around the cards matters as much as the cards.

What Thai-Specific Spaced Repetition Looks Like

Put those gaps back together and you get a clear picture of what a Thai deck should do that a generic one won’t:

	Generic SRS card	Thai-specific SRS card
Recall target	Letters / gloss	Letters plus tone
Audio	Often none, or a synthetic voice	Native speaker, every card
Reading	Isolated words only	Isolated and in-context segmentation
Romanization	Whatever the deck used	One consistent system, with tone marks
Card order	Independent, by forgetting curve	Interference-aware for minimal pairs

This is the gap Jam Kham was built to close. Under the hood the scheduling is the same proven science everyone uses—a forgetting-curve algorithm spacing your reviews. The difference is the layer on top: tone baked into every recall target, native-speaker audio on every word, segmentation built into the reading practice, a single consistent romanization, and sequencing that knows which Thai words trip each other up. We describe the three-layer memory system and the broader evidence base elsewhere, but the short version is this: the algorithm gets you efficient reviews; the Thai-specific design makes sure you’re reviewing the right thing.

That’s also why “best app for Thai” rarely comes down to the scheduler. As we lay out in our comparison of Jam Kham, Duolingo, and Anki, Anki’s FSRS is arguably the best general-purpose algorithm in existence—and it’ll still teach you tone-blind Thai if you feed it a tone-blind deck. The algorithm was never the weak link; the cards are.

The Takeaway

Spaced repetition is the most efficient way to build Thai vocabulary. That part of the internet’s advice is right. What it usually leaves out is that the scheduling is the easy, solved part—the quality of the cards is what’s hard. A world-class scheduler pointed at tone-blind, silent, inconsistently romanized cards will efficiently make you fluent in a Thai that no Thai person speaks.

So when you choose a tool, look past the streaks and the scheduling claims. Ask whether the tone is part of the answer, whether every card has native audio, whether you’re being taught to read words inside real sentences, and whether the romanization is consistent enough to trust. Get those right, and the spacing science does the rest.

Try Thai-specific spaced repetition free → start with Read Thai. Native audio and tone training on every card, segmentation built into reading practice, and a scheduler that handles the timing—so the only thing you have to do is show up for twenty minutes a day.

Related reading: Why Spaced Repetition Works · The Testing Effect · Jam Kham vs Duolingo vs Anki

Frequently Asked Questions

What is the best way to learn Thai vocabulary?

Spaced repetition with Thai-specific cards. The spacing science—reviewing each word just before you’d forget it—is the most efficient method for memorizing vocabulary, and it matters even more for Thai because there are no cognates to lean on. The catch is that the cards have to carry the tone, native audio, and consistent romanization. Spaced repetition with generic, tone-blind cards is efficient at teaching you the wrong pronunciation.

Can’t I just use Anki with FSRS for Thai?

You can, and FSRS is an excellent scheduler—it hits the same retention as the old SM-2 algorithm with roughly 20–30% fewer reviews. But FSRS only schedules the cards you give it; it doesn’t check whether they’re good Thai cards. If your deck lacks native audio, omits tone from the answer, or mixes romanization systems, FSRS will efficiently drill those flaws into memory. Anki works for Thai only if you (or the deck’s author) build in tone, audio, and segmentation yourself.

Why does Thai need different flashcards than Spanish?

Because Thai breaks three assumptions baked into generic decks. First, meaning depends on tone, so the answer to a card has to include pitch, not just letters. Second, Thai is written with no spaces, so reading requires word-segmentation practice that a list of isolated words never builds. Third, there’s no standard romanization, so without one consistent system you rehearse conflicting spellings of the same word. Spanish decks don’t face any of these.

Does generic spaced repetition actually hurt, or just help less?

It can actively hurt. Spaced repetition strengthens whatever you rehearse, so a card with a missing or wrong tone doesn’t just fail to teach—it fossilizes an error that takes far longer to unlearn later. A silent card teaches you a confident, consistent mispronunciation. The efficiency that makes spaced repetition great is what makes bad Thai cards dangerous.

How much daily practice do I need for Thai spaced repetition to work?

About 15–20 focused minutes a day is plenty for steady progress, and consistency beats marathon sessions—short daily reviews line up with how memory consolidates. What matters more than the minutes is that those minutes are spent on tone-aware cards with native audio. Twenty good minutes a day will outpace an hour of tone-blind drilling every time.