Can a Lui Chew speaker understand Hainanese? A word-by-word test

Someone in your family has probably said it: we are basically Hainanese. Or the opposite, said with more certainty than the speaker had: not at all, completely different. Both answers arrive without evidence, and this page is an attempt to hand you some.

Here is the short version. Take seventeen everyday words. In thirteen of them, the everyday Lui Chew form and its Hainanese counterpart still share their consonants and their vowel, or differ at the opening by a single shift that runs right through the language. One pair, 食 “to eat”, is identical down to the pitch. A fourteenth word, water, lines up only in the formal reading-aloud form, not the kitchen one. So the words really do carry across the strait. What does not carry is the tune. Both languages sort their syllables by pitch pattern, called tone, where changing the pitch changes the word, and of the seven pitch slots the two languages share, only two are pitched the same way. That gap is the difference between catching a word and following a sentence.

The short answer, and what it actually rests on

The two languages are close kin, and the Chinese-language account of the Leizhou people is careful about how. It traces today’s Leizhou people to the Putian area of Fujian, says that in the late Southern Song those migrants scattered across the Leizhou Peninsula and along parts of the Hainan coast, and says that the peninsula stream merged with the local population to become the Leizhou Min people. About the Hainanese it says only that the two peoples share origins. It does not trace one to the other, and neither do we.

The Leizhou City government’s own dialect page puts the result in one word: 姐妹语系, sister languages, naming Hainanese alongside Putian and Teochew. A few lines later the same page adds the part that matters here, that Leizhou’s vocabulary, pronunciation and grammar “have gradually developed differences” from its mother language and its sister languages.

Our comparison of the two peoples reaches for Spanish and Italian to describe that result. The analogy is ours, not a linguist’s finding, and it is a good one. It is still an analogy, and this page exists to put words in its place.

One term to have ready. A cognate is a word two languages both inherited from the same ancestor, which is why it still looks alike in both. The table below is seventeen candidates. Cognates are also exactly the trap: they make two languages feel mutually understandable when they are not, because recognising a word is not the same as following a sentence built out of them.

The test: seventeen everyday words

Every reading below comes from one place, English Wiktionary, which gives a Lui Chew reading (it labels it “Leizhou”) and a Hainanese reading inside the same entry. That matters more than it sounds. Two separate dictionaries would write the same sound differently, and the comparison would be worthless. One source writing both means the two columns can honestly be read against each other.

It is a community-compiled dictionary rather than a primary source. What gives it weight here is that the pattern it produces is not its own: the differences it shows between the two languages are the ones Wikipedia’s article on Hainanese lists.

The sounds are written in the IPA, the phonetic alphabet where one symbol always means one sound. The small raised numbers are the pitch.

Most of these words come in two forms, an everyday one you would use at a kitchen table and a formal one for reading a character aloud. Wiktionary marks which is which unevenly, so this table only marks a reading where the dictionary itself does. On the Lui Chew side it marks eleven of the seventeen words. On the Hainanese side it marks one, 我 “I”. Both columns print every reading the entry gives for the word named in the first column, in the order the entry prints them, and the right-hand column says which readings line up. Where a Hainanese reading carries no label, we do not guess at one.

The Lui Chew readings here are the homeland variety, the dialect as spoken on the Leizhou Peninsula. The Malaysian dialect differs from it in places, and the pitch table further down rests on a Malaysian source instead.

What the table shows, counted

Read the right-hand column downward and the seventeen words sort into six groups, every word named.

• One pair is identical, pitch and all: 食, to eat.
• Five share their consonants and their vowel and differ only in pitch: person, sea, eye, wind, fish.
• Seven have an opening consonant that differs by one of the regular shifts below, the kind that runs through a whole language rather than landing on one word: moon, I, tea, big, head, heart, mountain. In moon the vowel is written a shade apart as well, uɛ against ue.
• One has a reading that is identical on both sides, but not the kitchen one: water. /sui³¹/ appears in both columns; on the Lui Chew side the dictionary marks it as the formal, reading-aloud form. On the Hainanese side it marks nothing, so which of the two Hainanese readings a speaker would actually use is not something this source can tell us.
• Two share an opening and then part ways: sun and meat. In both, the vowel shifts, the syllable ends on a different consonant, and the pitch differs too.
• One is different from its first sound: cow.

So thirteen of seventeen keep a recognisable shape in the everyday form, and a fourteenth gets there only through the formal one. That is a high hit rate, and it is why the “we are basically the same” instinct in families is not foolish. It is also why it misleads. Thirteen recognisable words are thirteen pauses in a sentence you are otherwise not following.

The changes you can hear

The seven words in the middle group are not scattered exceptions. Wikipedia’s article on Hainanese lists five regular changes in one bulleted passage. Four of them turn up in everyday words, and those are the four below; the fifth, an old h becoming breathier in Wenchang, does not appear in this word set.

The source scopes these carefully, and the scopes are not the same for all four. The first change below is given for the more innovative varieties, naming Wenchang and Haikou. The second is given for the more innovative varieties with no names attached. The third is given for Haikou and Banqiao but specifically not Wenchang. The fourth carries no variety label, with an exception noted in the conservative Banqiao dialect. Haikou is the variety in the table above, and Wenchang is the one the article calls the prestige dialect, used as its reference point.

The plain stops were pulled inward. An implosive is a consonant made by drawing air in rather than pushing it out, written /ʔd/ here; it sounds like a d with a small swallow in front of it. Hainanese turned its plain t into one. You can hear it in tea, /tɛ²²/ against /ʔdɛ³¹/, and again in big.

The puffed consonants lost their puff. A consonant is aspirated when a breath of air follows it, the p of “pin” rather than the p of “spin”. Hainanese thinned those breaths into fricatives, sounds made by friction. That is why head is /tʰau²²/ on one side of the strait and /hau³¹/ on the other: the puff itself became the whole consonant.

An old b softened into a v. Moon and I both show it, and moon shows it most clearly, because the pitch matches on both sides and the opening consonant is the main thing that moved.

Last of the four, an old s hardened into a t. Heart is the clearest case, /sim³⁵/ against /tim²³/, and mountain repeats it with the same pitch step, /sua³⁵/ against /tua²³/.

Why the tunes do not travel

In both languages, changing the pitch of a syllable changes the word, and each sorts its syllables into eight pitch slots where Mandarin has four. The slots carry inherited names from the same family of categories, which is what lets the two charts be laid side by side at all. They are not the same eight, and the table shows where they part.

Linguists write a pitch pattern as its shape on a five-point scale, 1 lowest and 5 highest. So 24 means the voice starts low and rises to mid-high; 21 means it starts low and drops; 33 means it sits flat in the middle. The slot names, yin ping (阴平) and so on, say where a pattern came from historically rather than how it sounds.

Two sources, two languages, and neither is the source behind the word table. The Lui Chew column is Chen Limao’s 2023 field study, which measured the dialect as it is spoken in Malaysia, from five named speakers over three field trips. The Hainanese column is the tone chart in Wikipedia’s article, which the article labels for Hainanese generally rather than for any one variety, and which carries no date and no citation. The Leizhou Min article used for the cross-check below carries a general “needs more citations” tag and is filed as a stub, so neither Wikipedia chart is on the footing Chen’s paper is.

The numbers in this section and the numbers in the word table are not on one scale. The word table’s pitch digits are Wiktionary’s, and they do not agree with either chart here; Wiktionary writes Haikou pitches as 31, 23 and 213 where the chart below writes 21, 24 and 211. So read the pitch digits against the other column of the same table, and not against a number from the other table. That is also why the word table’s evidence is the consonants and vowels, and the pitch argument is made here instead.

Seven slots appear on both sides. Two of those seven are pitched the same way, one more is close enough to survive a noisy room, and the remaining four have drifted. One has drifted into reverse: the same historic slot falls in Lui Chew and rises in Hainanese. On top of that, each language keeps one slot the other’s chart has no place for.

That is the mechanism. You recognise the consonants and the vowel, which is why individual words keep jumping out. The pitch you have no reliable grip on, and in these languages pitch is part of the word rather than decoration. Thirteen recognisable skeletons set to the wrong music do not add up to a conversation.

One check worth having, and it is a different operation from the one just warned against, because it sets two descriptions of the same language side by side rather than two languages: Chen’s Malaysian measurements sit close to the chart Wikipedia publishes for the China dialect. Four of the eight slots match digit for digit. Of the remaining four, three differ only because one chart writes a level pitch with two digits and the other with one, and the fourth, yin shang (阴上), is the same falling shape one step apart at the top. So the Lui Chew column is not an overseas outlier.

What we do not know yet

Nobody appears to have tested it. We could not find a published comprehension study of these two languages against each other, the kind that plays a recorded passage to speakers of the other language and counts what gets through. Nothing of the sort is cited or described in either Wikipedia article or their reference lists, in Chen’s 2023 paper or its ten references, in the Leizhou City government’s dialect pages, in the Chinese-language account of the Leizhou people, in Baidu Baike’s entry on the dialect, in Glottolog, or in the Ethnologue entries for the two languages, as of August 2026. The strongest institutional statement available is that the two carry separate international language codes, luh and hnm, and separate catalogue entries, which also file both inside one Chinese macrolanguage. Those entries record how the catalogues classify the two. Neither says anything about whether their speakers understand each other.

So the flat verdict is an inference, including ours. Six of our own pages state that the languages are not mutually intelligible, in the register of a quotation. None attaches a citation to the sentence; they route to our sources page, whose section on this language lists Wikipedia’s Leizhou Min article, the Ethnologue entry and the Chinese-language account of the Leizhou people among its entries, alongside the Hainanese article and Chen 2023 elsewhere on the same page. Read them and this is what is there:

• The article on Leizhou Min says the language has “low intelligibility with other Southern Min varieties”. It names no variety at all.
• The article on Hainanese says Hainanese is “mutually unintelligible with Southern Min varieties such as Hokkien and Teochew”. It names two, and neither is Leizhou Min.
• The Chinese-language account and the Ethnologue entry carry no intelligibility statement of any kind.

Neither of the first two names the other language across the strait. And under the classification each sentence is reporting, Yuan Jiahua’s, each language sits inside the other’s unnamed “Southern Min varieties”, which is exactly why a statement about an unnamed class cannot be quoted as a verdict on this particular pair. Our own pages are on that list because the honest version of this answer is worth more than a tidy one, and because the table above is a better argument than the sentence it replaces.

And the evidence still points the same way. The verdict is very probably right. It has just never been demonstrated, so it belongs among the things the record strongly suggests rather than the things the record states.

Try it yourself

Three things you can do this week, in rising order of usefulness to everybody.

Say the table out loud. Start with the five words that differ only in pitch: person, sea, eye, wind, fish. Say each one with the Lui Chew pattern, then the Hainanese one. That is the whole experience of hearing the other language, compressed into a syllable at a time.

Listen to both. Search for a song sung in Leizhou dialect, then for Hainan opera. Play a minute of each back to back and listen for the rhythm and the consonants first, which is where the two are close, then the pitch, which is where they part. You are not looking for comprehension. You are checking whether the second one sounds like a language you half know.

Record an elder. This is the one that outlives the other two. Chen Limao’s paper says that before his own work, the study of Chinese dialects in Malaysia had not touched the Leizhou dialect; it gives the number of Leizhou people in Malaysia as only a few thousand, calls the dialect extremely endangered, and says the case for rescuing and investigating it is more urgent still. A phone recording of a grandparent counting to ten, naming the things in a kitchen, or telling one story is a primary source that, so far as this project can establish, sits in no public archive. If you make one, it is worth more than this page. Our page on how few speakers are left says why.

Sources for the word readings, the tone charts, the sound changes and the classification history are listed on our sources page, and more of the language sits on our Living Culture hub. The Lui Chew pitch measurements are from Chen Limao, "On the Phonology Characteristics of Leizhou Dialect in Malaysia", Open Journal of Applied Sciences 13(9), 2023. The per-word readings are from English Wiktionary, retrieved 31 August 2026, and are marked as community-compiled there and here. See also how the language works as a whole on Leizhou Min, and the two peoples set beside each other on Lui Chew vs Hainanese.