How Does Cantonese Grammar Work for Beginners?

Overview

How does Cantonese grammar work? An overview

Here's something most beginner content won't tell you plainly: Cantonese and Mandarin are not the same language wearing different tones. They share a writing system (mostly) and a linguistic family, but a Mandarin speaker cannot understand spoken Cantonese, and vice versa — different vocabulary, different grammar particles, a different sound inventory entirely, and in places a genuinely different word order preference. If you've been told Cantonese is "just Mandarin with more tones," that's a real disservice to how much separately-learned grammar this language actually requires.

The good news arrives from an unexpected direction: Cantonese verbs never conjugate. There's no equivalent of Spanish's six-person present tense, no equivalent of French's subjunctive mood, no equivalent of German's case-marked articles. 食sik6 (to eat) is 食 whether the subject is "I," "you," "she," "we," or "they," and whether the action happened yesterday or will happen tomorrow. That single fact removes an entire category of memorisation that trips up learners of most European languages.

What replaces conjugation is a different kind of precision, spread across several systems this guide covers in full: six lexical tones that distinguish words that would otherwise be pure homophones; an aspect system of small markers glued directly after the verb (咗zo2, 緊gan2, 住zyu6, 過gwo3, 開hoi1) that do the job tense endings do in English, but by describing whether an action is finished, ongoing, experienced before, or habitual rather than when on a calendar it happened; an obligatory measure-word system that requires a specific classifier between almost every number and the noun it counts; and a famously large set of sentence-final particles that carry the emotional shading English puts into intonation, hedge words, and tag questions ("...right?", "...I guess", "...or something").

Layered on top of all of that is a genuine, practical fork most textbooks gloss over: a real gap between colloquial spoken Cantonese, with its own particles, vocabulary, and word order habits, and Standard Written Chinese, the Mandarin-grammar prose that fills Hong Kong newspapers, official documents, and most textbook dialogues, simply read aloud with Cantonese pronunciation. Depending on whether your goal is speaking with people or reading formal text, you're studying two related but distinct grammatical systems — and this guide is built around the colloquial spoken grammar, because that's what carries a conversation.

None of this is a reason for discouragement about where you are right now. Confusing tone 4 for tone 6 well into intermediate study, needing to double-check which measure word goes with which noun, or feeling like sentence-final particles are an impossible intuition to develop is the normal shape of this specific hill — not a sign you're bad at languages, and not a sign the resource you're using is failing you. Every section below names exactly why a given system feels hard, then works through it with real Cantonese sentences you can hold onto.

Why Cantonese grammar specifically feels disorienting at first

Here's something most Cantonese content won't tell you plainly, and it's worth naming before anything else: the disorientation you feel in your first months isn't because you're processing this badly, it's because Cantonese asks you to build several genuinely new mental categories at once, and most learners have never needed any of them before. If your only prior language experience is English, French, Spanish, or German, you've never had to track lexical tone as a load-bearing part of a word's identity, never had to slot a mandatory classifier between every number and noun, and never had to hold an emotional-attitude particle system this dense in working memory while also managing word order and vocabulary retrieval in real time. That's not a small ask, and feeling stretched thin across all of it simultaneously in your first year is exactly what building those categories from scratch actually feels like from the inside — not a sign the language is uniquely cruel, and not a sign you're behind some invisible normal pace.

It also helps to notice which parts of this system are working in your favour, because it's easy to fixate on the six tones and the measure words and lose sight of how much heavier this could have been. No verb conjugation means an entire category of memorisation other language learners spend years on simply doesn't exist here. No grammatical gender means noun-adjective agreement, a genuine source of ongoing errors for learners of Spanish or German well into advanced study, isn't a concern at all. Pronouns don't change form by case. If you've felt like everything about Cantonese is hard, it's worth deliberately noticing the places where it's handing you something simpler than what you might have faced elsewhere — that noticing is not toxic positivity, it's an accurate read of where your effort is actually going to pay off fastest.

How to use this guide if you're at an early stage

If you're in your first few months, resist the urge to read this entire guide front to back in one sitting and try to hold all of it at once — that's a recipe for the specific kind of overwhelm that makes people quietly close the tab and not come back. The tones section, the jyutping section, and the written-vs-spoken section are the three genuinely foundational pieces worth reading first and returning to often, because almost everything else in this guide assumes you have at least a working, even if shaky, grip on those three. The measure-words, aspect, and particles sections are worth treating as living references you come back to repeatedly over months, not single-read material — nobody internalises the difference between 咗 and 過, or memorises even half the particle table, from one pass through a page. Bookmark this, revisit specific sections when a real sentence you're stuck on calls for it, and let repetition across many small return visits do the actual work that a single long read never could.

How to use this guide if you're past the early stage and feel stuck

If you've been studying Cantonese for a year or more and this guide's tables mostly confirm things you already sort of knew, the plateau you're likely feeling isn't about missing grammar facts anymore — it's the gap between recognising a pattern on a page and producing it unprompted, in real time, in an actual conversation. That gap is completely normal and it is not evidence that your studying so far hasn't worked. Receptive knowledge (recognising 過 when you read it, understanding a particle's flavour when you hear it) reliably runs ahead of productive knowledge (reaching for 過 correctly mid-sentence without stopping to think, choosing the right particle live rather than after the fact) for every learner of every language, and Cantonese's particle system in particular is a slow-burn skill that keeps deepening for years even among fairly advanced learners. If you can already follow most of this guide's content but still freeze up producing it in conversation, that's not a sign this guide (or your studying) failed — it's a sign you're exactly at the stage where sentence-level output practice, not more reference reading, is the lever that actually moves things forward.

Tones

The 6 tones (and the 9-tone framing)

The tone count is the first honest surprise for most learners: Cantonese runs on six tones where Mandarin has four. That's not a small step up — it's fifty percent more pitch contrasts to hear and produce accurately, layered on top of a completely different set of consonant and vowel sounds from Mandarin. Feeling slower to tell tones 2 and 5 apart, or 3 and 6, well into intermediate study is the normal shape of this specific hill, not a sign of a bad ear.

Because so many Cantonese syllables are genuinely distinguished only by tone, mishearing tone 4 for tone 6, or tone 2 for tone 5, is one of the most common real-world listening errors — and it's exactly the kind of error native speakers resolve instantly through context, the way an English speaker doesn't consciously distinguish homophones like "there/their/they're" because sentence context does the work. Building that same instant contextual resolution in Cantonese takes real listening volume over time, not a sharper tone chart.

Why some sources say Cantonese has 9 tones

You will run into claims that Cantonese has nine tones, and both framings are technically defensible — the confusion is worth resolving directly rather than leaving you unsure which number to trust. The extra three "tones" in the nine-tone count are not new pitch shapes; they are checked tones (入聲jap6 sing1, "entering tones"): syllables that end in a stopped consonant (-p, -t, or -k) rather than a vowel or -n/-ng, which cuts the syllable short and only ever carries the pitch of tones 1, 3, or 6.

Modern jyutping (the system this guide uses throughout, covered next) simply reuses the 1/3/6 numbers for checked syllables rather than assigning separate 7/8/9 digits, because the pitch really is identical — the only difference is that the syllable is clipped short by the final consonant. For practical study purposes: learn six tones, notice that syllables ending in -p/-t/-k always carry the pitch of 1, 3, or 6, and don't spend energy memorising a separate "tone 7/8/9" system that most contemporary teaching materials, dictionaries, and this guide don't use.

Tone pairs that cause the most listening trouble

• Tone 2 vs tone 5 — both rise, but tone 2 starts higher and rises further; tone 5 starts lower and stays a smaller rise. 史si2 (history) vs 市si5 (city) is the classic minimal pair.
• Tone 4 vs tone 6 — both are low, but tone 4 falls and often carries a breathy or creaky voice quality, while tone 6 stays level. 時si4 (time) vs 是si6 (to be) is the classic pair.
• Tone 1 vs tone 3 — both are level, but tone 1 sits noticeably higher than tone 3, which can feel subtle until you've heard both side by side repeatedly. 詩si1 (poem) vs 試si3 (to try).

Recording your own attempts and comparing them against native audio, syllable by syllable, is the single most effective habit for closing these gaps — reading a tone chart tells you the shapes exist, but only repeated production-and-comparison builds the muscle memory that lets you produce them without conscious thought mid-sentence.

Tone change: when a syllable's tone shifts in casual speech

One more genuinely important wrinkle, often left out of beginner tone charts entirely: certain syllables — especially common nouns, diminutive-feeling words, and words used casually and affectionately — regularly shift to a high rising tone (tone 2) in natural spoken Cantonese, regardless of what tone that syllable carries in its dictionary/citation form. Linguists call this tone change (變音bin3 jam1), and it's a real, systematic feature of colloquial speech rather than a random inconsistency between sources, which matters because otherwise it can look like your dictionary and the audio you're hearing simply disagree.

• 妹mui6 is citation-form tone 6 (mui6, "younger sister") but shifts to tone 2 (mui2) in the common casual address 妹妹mui2 mui2 or when referring to a girl affectionately.
• 女neoi5 is citation-form tone 5 (neoi5, "daughter/girl") but frequently surfaces as tone 2 (neoi2) in casual reference to "a girl/daughter," especially in a warm or diminutive context.
• 片pin3 is citation-form tone 3 (pin3, "a slice/film") but commonly shifts to tone 2 (pin2) in casual contexts like referring to a video clip.

Tone change doesn't follow a single mechanical rule you can apply to every word — some words shift consistently, some shift only in specific fixed expressions, and some never shift at all — which is exactly why it belongs in the "listen for it in real audio" category rather than the "memorise this rule" category. The practical takeaway for a learner: if a word you've drilled in its dictionary tone suddenly sounds like it's carrying a different, higher rising pitch in a native recording, that's very likely tone change at work rather than a mistake in your materials or your own mishearing, and it's worth logging the specific word rather than assuming your tone chart was wrong.

Jyutping

Jyutping: initials, finals, and tones

Jyutping (developed by the Linguistic Society of Hong Kong in 1993) is the modern standard for writing Cantonese pronunciation in Latin letters, spelling out the initial consonant, the vowel-plus-final, and a tone number 1 through 6 for every syllable — 你好nei5 hou2 (hello) becomes nei5 hou2. It's the system used across dictionaries, language-learning apps, and typing input methods, and it's the system this entire guide uses for every example.

If you already know pinyin from Mandarin, resist the urge to read jyutping the same way. The two systems assign different sounds to some of the same letters, and Cantonese has sound combinations — final stop consonants like -p, -t, -k that simply cut off the airflow rather than releasing into a vowel — that don't exist in Mandarin at all. Treat jyutping as its own system to learn from scratch, even coming from a Mandarin background: that background gives you real shortcuts, but they're vocabulary and character-recognition shortcuts, not pronunciation ones.

Decoding any jyutping syllable follows the same three-part read, left to right:

1. Read the initial consonant sound, if there is one — n, h, g, and so on.
2. Read the vowel and final consonant, if any — ei, ou, aa, or a stopped final like -p/-t/-k.
3. Read the trailing number 1 through 6, which tells you the tone to apply to the whole syllable.

Jyutping initials

Cantonese has 19 initial consonant sounds (plus syllables that start directly with a vowel, treated as having no initial). Several pairs that look similar on paper are genuinely distinct sounds worth drilling separately.

Jyutping finals (vowels and endings)

Finals combine a vowel (or diphthong) with an optional ending consonant. The endings split into two families: -n/-ng/-m (which let the tone ring out) and -p/-t/-k (which cut the syllable off sharply and always carry a checked tone).

Then the ending consonants layer on top of any vowel above:

The tone numbers, as a quick-reference table

A practical takeaway for reading jyutping fluently: if you see a syllable ending in -p, -t, or -k, its tone number will only ever be 1, 3, or 6 — never 2, 4, or 5 — because a clipped syllable physically can't carry a rising or a held-low-falling contour in the time available. Recognising that pattern removes a chunk of tone-guessing the moment you see the spelling.

Written vs spoken

Written vs spoken Cantonese: the split every learner needs to know about early

Here's the part that catches almost every learner off guard, often well after they've already invested months of study: most formal writing in Hong Kong is not written in Cantonese grammar. Newspapers, official documents, and many textbook dialogues are typically written in Standard Written Chinese (書面語syu1 min6 jyu5) — grammatically much closer to Mandarin — and then simply read aloud using Cantonese pronunciation. A sentence a Hong Kong news anchor reads on air often uses different word choices and different grammar from what that same anchor would say chatting with a friend five minutes later.

Genuinely colloquial Cantonese (口語hau2 jyu5) — with its own grammar, its own particles, and vocabulary that doesn't exist in Mandarin at all — is written using a specific set of Cantonese-only characters, some borrowed and repurposed from Classical Chinese, some genuinely unique to Cantonese. This colloquial written form shows up in text messages, comics, film subtitles, song lyrics, and casual online writing, but rarely in formal contexts. The table below is worth genuinely memorising, because these eight characters are probably the highest-frequency grammar words in everyday spoken Cantonese, and none of them behave the way their nearest Mandarin equivalent does.

For a learner, this means a real fork in the road depending on your goal: studying to read Hong Kong newspapers is closer to learning a Mandarin-grammar reading skill with Cantonese pronunciation layered on top, while studying to text a friend, watch Cantonese variety shows and dramas, or actually speak day to day requires the colloquial grammar and character set specifically. Most conversational Cantonese study — the kind aimed at speaking and understanding real people — rightly focuses on the colloquial form, and every section from here forward in this guide describes colloquial spoken Cantonese grammar, since that's the grammar that carries an actual conversation.

Why this split exists

The historical reason is straightforward: written Chinese has functioned for over a thousand years as a shared literary language across regions that speak mutually unintelligible varieties (Cantonese, Mandarin, Hokkien, Shanghainese, and more), because characters can represent meaning somewhat independently of how any given reader pronounces them. Modern Standard Written Chinese crystallised around Mandarin grammar specifically because Mandarin became the government-promoted national standard across mainland China and Taiwan in the 20th century, and Hong Kong's education and media largely adopted that same written standard even while the spoken street language stayed Cantonese. Colloquial written Cantonese survived and thrived anyway, mostly outside formal channels, precisely because Hong Kong retained day-to-day Cantonese as the dominant spoken language rather than shifting to Mandarin.

Practical signs you're reading Standard Written Chinese vs colloquial Cantonese

• Pronoun check — 他/她/它taa1 signals Standard Written Chinese; 佢keoi5 signals colloquial Cantonese.
• Negation check — 不bat1 signals Standard Written Chinese; 唔m4 signals colloquial Cantonese.
• Possessive check — 的dik1 signals Standard Written Chinese; 嘅ge3 signals colloquial Cantonese.
• Completion check — 了liu5 signals Standard Written Chinese; 咗zo2 signals colloquial Cantonese.
• Sentence-final particles present at all — a heavy particle presence (呀aa3, 㗎gaa3, 喎wo5, 咩me1, 囉lo1) is close to a guarantee you're reading colloquial Cantonese, since Standard Written Chinese uses these far more sparingly if at all.

Keep this checklist handy for your first months of reading practice — it turns "why doesn't this sentence sound like what I hear on TV?" from a confusing mystery into a quick, answerable diagnostic.

Word order

Word order: SVO and topic-comment

The frustrating thing about Cantonese word order for a beginner is that it's mostly reassuring — Subject-Verb-Object, the same backbone English uses — right up until it isn't, and the moments it isn't are exactly the moments a word-for-word translation from English produces a sentence that sounds distinctly foreign to a native ear. Both patterns matter, and knowing when each one is in play is a genuine skill, not a fixed rule you can apply blindly.

The default: Subject-Verb-Object

Most simple Cantonese sentences follow the same subject-verb-object order English does, which is exactly why Cantonese often feels approachable at the sentence-skeleton level even while the tones and particles feel foreign:

• 我食飯。ngo5 sik6 faan6. — I eat rice/food. [I - eat - rice]
• 佢鍾意呢隻貓。keoi5 zung1ji3 ni1 zek3 maau1. — They like this cat. [they - like - this cat]
• 我哋琴日睇咗場戲。ngo5 dei6 kam4jat6 tai2 zo2 coeng4 hei3. — We watched a movie yesterday. [we - yesterday - watched - a movie]

Notice the third example already shows a real Cantonese habit: time words (琴日kam4jat6, yesterday) sit right after the subject and before the verb, not tacked onto the end of the sentence the way "yesterday" often lands in casual English. This isn't optional flexibility — sentence-final time words are markedly less natural in Cantonese than subject-adjacent ones.

Topic-comment: the pattern that isn't in English

Where Cantonese genuinely diverges from a simple SVO expectation is its heavy use of the topic-comment structure: naming what you're about to talk about first (the topic), then saying something about it (the comment), even when that topic isn't the grammatical subject of the sentence's main verb. English does this occasionally ("That book, I already read it") but treats it as a marked, slightly unusual construction; Cantonese treats it as an entirely ordinary, extremely common sentence shape.

• 呢本書我睇過㗎喇。ni1 bun2 syu1 ngo5 tai2 gwo3 gaa3 laa3. — This book, I've [already] read it. [this book(topic) - I - have-read - particle]
• 今日啲天氣幾好。gam1jat6 di1 tin1hei3 gei2 hou2. — Today, the weather's pretty good. [today(topic) - weather - quite good]
• 呢個問題我都唔知點答。ni1 go3 man6tai4 ngo5 dou1 m4 zi1 dim2 daap3. — This problem, I don't even know how to answer [it]. [this problem(topic) - I - even - don't know - how to answer]

The functional test for whether you're looking at a topic-comment sentence rather than a plain SVO one: ask whether the first noun phrase is something the rest of the sentence is "about" rather than something literally performing the verb's action. In 呢本書我睇過ni1 bun2 syu1 ngo5 tai2 gwo3, the book isn't doing any "reading" — the speaker is — so 呢本書ni1 bun2 syu1 is a topic being commented on, not a grammatical subject.

Why this matters practically

Two concrete habits fall out of this for a learner building sentences:

1. You can front almost anything you want to emphasise. If a piece of information is the point of the sentence — what you're talking about — Cantonese lets you move it to the very front rather than burying it mid-clause, which is a genuinely useful tool once you notice native speakers doing it constantly.
2. Don't force an English "it" or dummy subject where Cantonese doesn't need one. Because the topic already establishes what's being discussed, Cantonese frequently drops a pronoun object that English would insist on: 睇咗未呀?tai2 zo2 mei6 aa3? literally "watched yet?" with no stated object at all, understood from context as "have you watched [it/that show/that movie]?"

Pro-drop: Cantonese omits pronouns constantly

Related to topic-comment structure, Cantonese is a heavily pro-drop language: subjects and objects that are clear from context are routinely dropped rather than restated, in a way that feels sparse to an English ear used to always naming "I," "you," "it." A text exchange like 食咗未?sik6 zo2 mei6? ("eaten yet?") / 食咗喇。sik6 zo2 laa3. ("[I've] eaten already") states no subject in either line, and no native speaker perceives anything missing. Resist the beginner instinct to insert 我ngo5 (I) and 佢keoi5 (it/they) into every sentence the way English grammar requires — overusing pronouns is one of the more reliable "textbook accent" tells in spoken Cantonese.

Pronouns

Pronouns

Cantonese pronouns are, refreshingly, one of the simplest corners of the grammar — there's no gendered "he" vs "she" in speech, no case system changing "I" to "me" to "my," and the plural is built with one predictable suffix rather than a separate word or an irregular form. The complexity that does exist is almost entirely about register (which pronoun sounds natural in which social context) rather than about grammatical machinery.

A few things worth internalising early about this table:

• 佢keoi5 covers "he," "she," "it," and singular "they" all at once in speech — spoken Cantonese does not distinguish gender in the pronoun the way English does, and does not distinguish animate from inanimate the way you might expect either (佢 is used for people, and can be used for an animal or occasionally an object treated with some personification, though 佢 for a plain inanimate object is less common than a demonstrative + measure word). The written-Cantonese-only characters 佢/佢哋keoi5 / keoi5 dei6 stay identical regardless of gender; only Standard Written Chinese splits 他/她/它taa1 by gender in writing (though all three are pronounced identically as taa1 even there).
• 哋 (dei6) is a genuinely productive plural suffix — it attaches to all three persons with zero irregularity, unlike English's mix of regular "-s" plurals and irregular pronoun-plural pairs (I→we, not "I-s").
• There's no case marking at all. 我ngo5 is "I," "me," and "my" depending on position, with a possessive particle (嘅ge3, covered in the relative-clauses section) doing the "my/your/their" job rather than a separate pronoun form: 我嘅書ngo5 ge3 syu1 (my book), 你嘅屋企nei5 ge3 uk1kei2 (your home).

Demonstratives: 呢ni1 and 嗰go2

Cantonese demonstratives ("this" and "that") always require a measure word between the demonstrative and the noun, unlike English's bare "this book" — the section on measure words below covers the full system, but the core pair is worth introducing here since it functions as a pronoun-adjacent building block used constantly in daily speech.

Question pronouns

Unlike English, Cantonese question words stay in their normal sentence position rather than moving to the front — 你食咗乜嘢?nei5 sik6 zo2 mat1je5? ("you ate what?") keeps 乜嘢mat1je5 exactly where the answer noun would go, rather than fronting it the way English "what did you eat?" does. This is a genuinely useful simplification: build the statement first, then swap in the question word where the unknown information belongs.

Reflexive and emphatic: 自己zi6gei2

自己zi6gei2 (self/oneself) attaches after any pronoun to build a reflexive or emphatic meaning, functioning the way English "-self" does but as a fully separate word rather than a suffix: 我自己嚟。ngo5 zi6gei2 lei4. (I'll come myself.) 佢自己整嘅。keoi5 zi6gei2 zing2 ge3. (they made it themselves.)

Numbers

Numbers

Cantonese numbers reward the effort you put in early: the counting system is almost entirely regular, follows a clean base-10 logic all the way into the millions, and — unlike English's irregular "eleven," "twelve," "twenty," "thirty" — never needs a separate word memorised for a "teens" range or a "tens" range. Once you know 1 through 10, the rest of the number system is mostly a small set of combination rules.

11 through 99: the fully regular pattern

Every number from 11 to 99 is built by combining the digits 1-10 above with almost no exceptions, following two simple templates:

• 11-19: 十 (ten) + digit. 十一sap6 jat1 (11), 十二sap6 ji6 (12), 十五sap6 ng5 (15), 十九sap6 gau2 (19).
• 20, 30, 40... 90: digit + 十 (ten). 二十ji6 sap6 (20), 三十saam1 sap6 (30), 九十gau2 sap6 (90).
• 21-99 (non-round): tens-digit + 十 + ones-digit. 二十一ji6 sap6 jat1 (21), 五十六ng5 sap6 luk6 (56), 九十九gau2 sap6 gau2 (99).

Contrast this directly with English, where "eleven," "twelve," "twenty," and "thirty" are all irregular words that don't obviously derive from "one," "two," "three": Cantonese never asks you to memorise a special word for any of these — 十一sap6 jat1 is transparently "ten-one," and 二十ji6 sap6 is transparently "two-ten."

100 and beyond

The genuinely disorienting part for English-speaking learners isn't Cantonese's fault — it's that Chinese number systems (Cantonese and Mandarin both) group digits in units of 10,000 (萬maan6) rather than English's units of 1,000. That means "one million" isn't a clean native unit the way it is in English; it's literally "one hundred ten-thousands" (一百萬jat1 baak3 maan6). Converting large numbers between English and Cantonese fluently, in your head, in real time, is a skill that takes deliberate separate practice even once your vocabulary is solid — budget real drilling time for this specifically rather than assuming it falls out naturally from knowing the digit words.

兩loeng5 vs 二ji6: the one genuine irregularity

The single number-system exception worth flagging clearly: when counting actual items with a measure word, "two" is almost always 兩loeng5, not 二ji6. 二 is used for the bare number itself (in counting sequences, phone numbers, math, dates, and the ones digit inside a bigger number), while 兩 is used specifically for "two of something": 兩個人loeng5 go3 jan4 (two people), 兩本書loeng5 bun2 syu1 (two books), but 二十ji6 sap6 (twenty) keeps 二 because it's not directly counting objects with a measure word — it's building the number twenty itself, and the two-ten interior combination doesn't switch to 兩.

Ordinal numbers

Ordinals ("first," "second," "third") are built by simply prefixing 第dai6 onto the cardinal number: 第一dai6 jat1 (first), 第二dai6 ji6 (second), 第十dai6 sap6 (tenth). No irregular forms exist the way English has "first/second/third" diverging from "one/two/three" — 第 plus any number, cardinal or compound, reliably produces the correct ordinal.

Telling time

Clock time follows its own compact set of conventions worth learning as a unit, since the words for "hour," "minute," and the five-minute increments don't derive transparently from the plain number system the way you might expect:

The 字 (zi6, "character," here repurposed as a five-minute counting unit) system is the genuinely distinctive part: 三點兩個字saam1 dim2 loeng5 go3 zi6 means "3:10" (three o'clock plus two five-minute blocks), a system with no direct English equivalent and worth deliberately drilling rather than assuming it'll click from the number system alone. 幾多點呀?gei2do1 dim2 aa3? (what time is it?) is the standard way to ask.

Dates and days

Note the same 一...十 counting system from earlier in this section carries straight through into months (一月jat1 jyut6 through 十二月sap6 ji6 jyut6, January through December) and most of the days of the week — genuinely one system reapplied across several everyday contexts, with Sunday as the sole exception worth memorising directly.

Money

蚊man1 is the colloquial unit for the Hong Kong dollar (formally 元jyun4, used in writing/formal contexts), and 毫hou4 or its casual form 毫子hou4zi2 covers the ten-cent unit: 五十蚊ng5sap6 man1 (fifty dollars), 三蚊半saam1 man1 bun3 (three dollars fifty — reusing the same 半, "half," pattern from telling time). 幾多錢?gei2do1 cin2? (how much [money]?) is the standard way to ask a price, alongside the near-synonym 幾錢?gei2 cin2? in fast casual speech.

Measure words

Measure words (量詞loeng6 ci4): the classifier system English doesn't have

This is the section that makes a lot of learners feel genuinely stuck, and for a legitimate reason: English lets you say "a book," "a car," "a dog" with the same bare article every time, so the very idea that a language requires a different, specific, memorised word between a number and almost every noun feels like an arbitrary extra burden rather than logical grammar. It isn't arbitrary, though — once you see the categories measure words track (shape, size, function, how something is bundled or contained), the system starts to feel like a genuinely useful extra layer of meaning rather than pointless overhead, closer to how English distinguishes "a herd of cattle" from "a flock of birds" from "a school of fish," except Cantonese requires this kind of distinction on nearly every noun, not just special cases.

The core rule: measure words are obligatory

Any time you have a number (or 呢/嗰, this/that) directly modifying a noun, a measure word (量詞, loeng6 ci4, also called a classifier) MUST sit between them. You cannot say "一書" the way English says "one book" — it must be 一本書jat1 bun2 syu1, with 本bun2 as the measure word specific to bound, page-based objects like books.

The most common measure words

There are dozens of measure words in active use, but a working set of around 20-25 covers the overwhelming majority of daily conversation. This table is worth treating as a genuine memorisation priority — pairing each measure word directly with its two or three most common nouns, rather than learning the measure words as an abstract list, is the approach that sticks.

個go3 is not a safe universal substitute

個go3 is genuinely the most common measure word and the closest thing Cantonese has to a "default," and it's true that using 個 when you don't know or can't recall the correct measure word is a real, usable fallback strategy that native speakers will understand. But it's worth being honest with yourself about what that strategy costs: substituting 個 everywhere (一個車jat1 go3 ce1, 一個貓jat1 go3 maau1, 一個書jat1 go3 syu1 instead of 一架車jat1 gaa3 ce1/一隻貓jat1 zek3 maau1/一本書jat1 bun2 syu1) is immediately noticeable to a native speaker as a clear non-native pattern, in roughly the same way a non-native English speaker saying "much people" instead of "many people" is understandable but marked. Treat 個 as your genuine emergency fallback, not your default habit, and prioritise building the specific-measure-word instinct for high-frequency nouns early.

The classifier-as-possessive construction

One of the more elegant, distinctly Cantonese constructions is using a bare measure word — with no number at all — directly as a possessive, dropping 嘅ge3 (the regular possessive particle covered in the relative-clauses section) entirely in casual, close-relationship speech: 我本書ngo5 bun2 syu1 ("my book," literally "I [+measure word] book") functions identically to the fuller 我嘅書ngo5 ge3 syu1, but sounds more casual and personal, almost like English's difference between "my book" and the more distant, formal "the book belonging to me." This pattern requires knowing the correct measure word for the specific noun (you couldn't substitute 個 here and sound natural), which is itself a good argument for learning nouns and their measure words as a single memorised unit rather than as two separate facts:

• 我隻貓好得意。ngo5 zek3 maau1 hou2 dak1ji3. — My cat is really cute. [I + MW(隻) + cat...]
• 佢架車幾新㗎。keoi5 gaa3 ce1 gei2 san1 gaa3. — Their car is pretty new. [they + MW(架) + car...]
• 你間屋好靚。nei5 gaan1 uk1 hou2 leng3. — Your house is really nice. [you + MW(間) + house...]

This construction is restricted to close, casual, personal-possession contexts — you wouldn't use it for something abstract or for a more distant/formal relationship, where the fuller 嘅 construction stays the natural, correct choice.

啲di1: the "some/plural" measure word

啲di1 deserves its own callout because it functions differently from the concrete-object measure words above: it marks an indefinite, uncounted quantity — "some" — and it is also the closest thing colloquial Cantonese has to marking a plural on an otherwise bare noun. 啲嘢di1 je5 (some things/stuff), 啲人di1 jan4 (some people/the people), 呢啲ni1 di1 (these), 嗰啲go2 di1 (those) are all extremely common in daily speech, and 啲 additionally shows up attached after an adjective as an intensifier/comparative marker (covered in the comparatives section) — the same character doing double duty across two genuinely different grammatical jobs is a pattern worth noticing rather than assuming is a typo or inconsistency.

Aspect

The aspect system: how Cantonese handles time without conjugation

Here's the reframe that makes this section click for a lot of learners: Cantonese isn't "missing" tense — it's using a genuinely different tool to do a related but distinct job. English tense (walked vs walk vs will walk) answers "when did this happen relative to now?" Cantonese aspect markers answer a different question: "what's the internal shape of this action — is it finished, ongoing, something that's happened before, a habit, or something you're doing a little bit of?" Once you stop expecting a one-to-one match with English tense and start asking the aspect question instead, the whole system becomes far more predictable, and you'll notice these markers attach directly after the verb with zero irregularity — no conjugation classes, no exceptions to memorise per verb.

咗zo2 (zo2) — perfective: the action is complete

咗 marks that an action has been completed, full stop — the single most common aspect marker you'll encounter, and the closest Cantonese equivalent to English's simple past for many everyday sentences, though it's genuinely about completion rather than "pastness" (a sentence can even use 咗 to describe a completed action that will happen in the future relative to some other future point, though that's a more advanced usage).

• 我食咗飯。ngo5 sik6 zo2 faan6. — I've eaten / I ate. [eat-completed]
• 佢走咗喇。keoi5 zau2 zo2 laa3. — They've left. [leave-completed + particle]
• 你做咗功課未呀?nei5 zou6 zo2 gung1fo3 mei6 aa3? — Have you done your homework yet? [do-completed + "yet" + particle]

咗 negates with 冇 (mou5, covered fully in the negation section), not 唔: 我冇食飯ngo5 mou5 sik6 faan6 (I didn't eat) — 咗 itself disappears in the negative rather than the sentence becoming "我唔食咗飯ngo5 m4 sik6 zo2 faan6," which is a genuine, memorisable exception worth flagging clearly rather than discovering by trial and error.

緊gan2 (gan2) — progressive: happening right now

緊 marks an action actively in progress at the moment of speaking (or at the moment being described), functioning like English's "-ing": 我食緊飯。ngo5 sik6 gan2 faan6. (I'm eating [right now].) 佢瞓緊覺,唔好嘈醒佢。keoi5 fan3 gan2 gaau3, m4 hou2 cou4 sing2 keoi5. (they're sleeping, don't wake them up.)

住zyu6 (zyu6) — continuous/sustained state

住 is subtly different from 緊 and genuinely trips learners up because English "-ing" covers both meanings that Cantonese splits apart: 住 marks a sustained state or an action being maintained/held, often layering on top of a resultative complement, rather than an action actively unfolding moment-to-moment. Compare 佢著住紅色件衫keoi5 zoek3 zyu6 hung4sik1 gin6 saam1 (they're wearing a red shirt — a maintained state of "wearing," not an ongoing act of putting it on) against 佢著緊衫keoi5 zoek3 gan2 saam1 (they're [in the middle of] getting dressed — an actual unfolding action). 攞住嗰本書。lo2 zyu6 go2 bun2 syu1. (hold onto that book) is another classic 住 use: a state of holding, sustained over time, not a momentary grabbing action.

過gwo3 (gwo3) — experiential: has happened at least once

過 marks that an action has been experienced at some point, without committing to exactly when — functionally close to English's present perfect "have done X [before]" rather than a specific-past simple tense: 我食過蛇。ngo5 sik6 gwo3 se4. (I've eaten snake before.) 你去過日本未呀?nei5 heoi3 gwo3 jat6bun2 mei6 aa3? (Have you been to Japan [before]?) The distinction between 咗 and 過 is genuinely important and easy to blur: 我食咗飯ngo5 sik6 zo2 faan6 states a specific, completed eating event (I ate [my meal]); 我食過蛇ngo5 sik6 gwo3 se4 states a category of life experience (I have [at some point] eaten snake), with no specific occasion necessarily in mind.

開hoi1 (hoi1) — habitual: something done as a regular habit

開 marks that an action is done habitually or as a matter of course, closer to English "usually do X" or "am used to doing X" than to any single tense: 我食開呢間嘅嘢。ngo5 sik6 hoi1 ni1 gaan1 ge3 je5. (I regularly eat food from this place.) 佢住開喺呢度。keoi5 zyu6 hoi1 hai2 ni1 dou6. (they normally/habitually live here.) This is a genuinely distinct concept from 緊 (happening right now) and from 過 (has happened before, unspecified frequency) — 開 specifically claims regularity or established habit.

The smaller but essential set: 埋maai4, 晒saai3, 返faan1

These three don't mark core aspect the way the five above do, but they're extremely high-frequency verb-suffixes that shape meaning in ways a learner needs early, so they belong in the same conceptual bucket:

Combining aspect markers with time words for full precision

Because none of these markers are tied to a calendar tense, Cantonese adds separate time words when the exact timing matters, stacking them naturally alongside the aspect marker rather than replacing it: 我尋日食咗飯。ngo5 cam4jat6 sik6 zo2 faan6. (I ate yesterday.) 我聽日會食緊飯嗰陣時你打嚟。ngo5 ting1jat6 wui5 sik6 gan2 faan6 go2 zam6si4 nei5 daa2 lei4. (Call me tomorrow while I'm [in the middle of] eating.) This is the real payoff of the system: aspect tells you the shape of the action, time words tell you when, and the two combine freely without any conjugation gymnastics.

Sentence-final particles

Sentence-final particles: where Cantonese puts its emotional nuance

If there's one system that separates "technically correct Cantonese" from "Cantonese that sounds like a real person said it," this is it — and it's genuinely one of the largest sentence-final particle inventories of any language studied by English speakers, which is exactly why this section runs long and comes with the biggest table in this guide. English conveys a huge amount of a sentence's emotional shading through intonation, filler words, and tag questions — "well, I mean, it's fine, I guess" carries real hesitation that a flat "it's fine" doesn't, and "you're coming, right?" carries a different social pressure than a bare "you're coming." Cantonese builds an unusually large share of that same nuance directly into small particle words tacked onto the very end of a sentence, often stacking two or three together in a single utterance.

It's worth being honest about the learning curve here: a list of particle "definitions" only gets you partway, because the actual feel of each one is closer to a tone of voice or a facial expression than a vocabulary word with one fixed meaning you can memorise once and be done with. That said, a solid working map of the most common particles — what mood or stance each one signals, roughly — genuinely accelerates both your listening comprehension and your ability to sound less flat when you speak. Treat the table below as your reference map, then build the real intuition through volume of real dialogue exposure.

The core particle reference table

Particles stack — and stacking order matters

Native speakers routinely combine two, sometimes three, particles in a single sentence, and the particles aren't randomly ordered — they follow a rough layering logic from "closer to the verb's grammar" out to "closer to pure tone/attitude," similar to how English adjective order isn't random (you'd say "a big old red barn," never "a red old big barn") even though no one consciously teaches that rule. 㗎喇gaa3 laa3, 㗎啦gaa3 laa1, and 㗎嘛gaa3 maa3 are all real, common combinations built from 㗎 plus a second particle, and each carries a genuinely distinct shade from either particle alone.

• 做完㗎喇。zou6 jyun4 gaa3 laa3. — It's done now, for real. [emphasis + change-of-state]
• 食多啲啦。sik6 do1 di1 laa1. — Eat some more, go on. [comparative 啲 + softening 啦]
• 佢唔嚟喎,聽講。keoi5 m4 lei4 wo5, ting1 gong2. — Apparently they're not coming, so I've heard. [reported information + explicit "so I've heard" reinforcing the same feel]

Why particles resist a "flashcard" approach

It's worth naming directly why this system feels harder to pin down than, say, memorising measure words: measure words map cleanly onto categories of physical objects, but particles map onto social and emotional stances — scepticism, resignation, gentle persuasion, reporting-secondhand-information — which shift subtly with context, speaker relationship, and tone of voice even while using the exact same particle. This isn't a flaw in how the language is taught to you; it's the actual shape of the system, the same way English speakers can't fully explain why "well, actually..." sounds condescending in one context and thoughtful in another — you know it from thousands of hours of exposure, not from a rule. Give yourself real permission to under-produce particles at first while over-indexing on listening for them; production catches up once the ear has logged enough real examples.

Negation

Negation: 唔m4, 冇mou5, and 未mei6

English gets away with one all-purpose negator — "not" — bolted onto almost anything ("I do not," "I did not," "I have not," "there is not"). Cantonese splits this single English job across three genuinely distinct words, each tied to a different kind of thing being negated, and picking the wrong one is one of the most common, most immediately noticeable beginner mistakes. The good news: the three-way split maps onto a real, learnable logic rather than arbitrary word choice, and once the logic clicks, the choice becomes close to automatic.

The core logic, spelled out

• 唔 is the default, general-purpose "not." Use it for describing states, opinions, habits, abilities, and future intentions: 我唔鍾意。ngo5 m4 zung1ji3. (I don't like [it].) 佢唔會嚟。keoi5 m4 wui5 lei4. (they won't come.)
• 冇 is specifically about absence and completed-past negation. It negates 有jau5 (to have/there is) directly — 冇 is historically and functionally 唔 + 有 fused together — and it's also the word that steps in whenever you'd otherwise negate a 咗-marked completed action, since 咗 itself simply disappears in the negative rather than being negated by 唔.
• 未 carries a built-in sense of "still pending." It implies the action genuinely hasn't happened but remains possible or expected — "not yet" rather than a flat "not." This is why 未 pairs naturally with 過 to mean "haven't ever [done this], though I still might" — both carry an open-ended, unresolved sense.

A direct three-way comparison

These three sentences look almost identical in English translation but mean genuinely different things in Cantonese, which is exactly the trap:

唔該m4 goi1 and 唔使m4 sai2: negation inside fixed expressions

唔 shows up baked into extremely common fixed phrases worth knowing as chunks rather than reparsing from scratch every time: 唔該m4 goi1 ("thank you" for a service/favour, or "excuse me") literally contains 唔 but functions as a polite formula rather than a literal negative statement; 唔使m4 sai2 ("no need to / don't have to") is the standard way to tell someone something isn't necessary: 唔使擔心。m4 sai2 daam1sam1. (no need to worry.)

Negating aspect markers directly

Beyond the three core negators, each aspect marker from the previous section has its own negation behaviour worth reviewing together in one place:

• 咗 (completed) → negate with 冇, and 咗 drops entirely: 食咗sik6 zo2 → 冇食mou5 sik6 (not 唔食咗).
• 緊 (progressive) → negate by rephrasing with 唔係...緊 or simply using 未: 佢未瞓緊覺keoi5 mei6 fan3 gan2 gaau3 is less natural than restating the fact plainly — in practice, progressive negation is often handled by negating the underlying situation rather than 緊 itself.
• 過 (experiential) → negate with 未: 食過sik6 gwo3 → 未食過mei6 sik6 gwo3 (haven't ever eaten [it]).
• 開 (habitual) → negate by describing the absence of the habit directly, often without 開 at all: 我唔食呢啲嘢ngo5 m4 sik6 ni1 di1 je5 (I don't eat this stuff) rather than forcing 開 into a negative frame.

Double negatives and 唔係唔m4 hai6 m4

Cantonese permits a deliberate double-negative construction for emphasis, similar to English's "it's not like I don't want to": 唔係唔想幫你,係冇時間。m4 hai6 m4 soeng2 bong1 nei5, hai6 mou5 si4gaan3. (It's not that I don't want to help you, it's [that I] don't have time.) This is a stylistic emphasis tool, not an accidental cancellation the way a careless double negative might read in formal English — used deliberately, it reads as natural and expressive rather than as an error.

Questions

Questions: sentence-final particles, A-not-A, and 定係ding6 hai6

English asks a yes-no question mostly by inverting word order ("You are coming" → "Are you coming?") or adding a helper verb ("Did you go?"). Cantonese does neither of those things by default — word order stays exactly the same as the statement — and instead offers three genuinely distinct question-forming strategies, each with a slightly different feel, all of which are worth having active in your toolkit rather than picking just one and using it everywhere.

Strategy 1: A-not-A, the most distinctly Cantonese pattern

The A-not-A construction builds a yes-no question by stating the verb (or adjective) in its positive form immediately followed by its negative form, back to back, with no separate question particle needed at all — the question is built entirely from the verb repetition itself. This pattern has no real English equivalent and is worth practising deliberately rather than translating from an English question template.

Notice that last row: 有jau5 (to have) is a genuine irregularity in the A-not-A pattern, because its negative isn't "唔有" but the fused word 冇 (mou5) — so the A-not-A question becomes 有冇jau5 mou5, not 有唔有jau5 m4 jau5. This single exception is worth memorising directly since 有冇 is one of the most frequent question patterns in daily speech (有冇興趣?jau5 mou5 hing3ceoi3?, are you interested? / do you have interest?).

For longer verbs (two-syllable verbs, or a verb plus complement), the A-not-A pattern often clips down to just the first syllable repeated, which is both faster to say and completely standard rather than a shortcut error: 鍾唔鍾意?zung1 m4 zung1ji3? (do you like it?) rather than the fuller-but-rarer 鍾意唔鍾意zung1ji3 m4 zung1ji3.

Strategy 2: question particles on a plain statement

The second strategy is simpler on the surface: take an ordinary statement and add a question-marking sentence-final particle, changing nothing about word order — very close in spirit to how a rising intonation alone can turn an English statement into a question ("You're coming?").

Strategy 3: 定係ding6 hai6 — offering a choice

定係ding6 hai6 ("or") builds an alternative question offering the listener a choice between two or more options, functionally like English "or" in "do you want tea or coffee?" — placed between the two choices, with the sentence otherwise built like a normal statement, no A-not-A or question particle needed since the "or" structure itself signals the question:

• 你想飲茶定係飲咖啡?nei5 soeng2 jam2 caa4 ding6hai6 jam2 gaa3fe1? — Do you want tea or coffee?
• 你去定係佢去?nei5 heoi3 ding6hai6 keoi5 heoi3? — Are you going or are they going?
• 今日定係聽日?gam1jat6 ding6hai6 ting1jat6? — Today or tomorrow?

定 alone (without 係) is also common in fast, casual speech as a contraction: 你想去邊度食飯定睇戲?nei5 soeng2 heoi3 bin1 dou6 sik6 faan6 ding6 tai2 hei3? (do you want to go eat or watch a movie?)

Question-word questions (who, what, where, when, why, how)

As covered in the pronouns section, Cantonese question words (邊個, 乜嘢/咩, 邊度, 幾時, 點解, 點) stay exactly where the answer would go in the sentence, rather than fronting to the start of the sentence the way English question words do. This is worth restating here as a question-forming strategy in its own right: build the statement, then swap in the appropriate question word at the position where the unknown information belongs, and no additional question particle or A-not-A construction is needed — the question word itself carries the full question force.

• 你食咗乜嘢?nei5 sik6 zo2 mat1je5? — What did you eat? [you - ate - what]
• 佢幾時返嚟?keoi5 gei2si4 faan1 lei4? — When are they coming back? [they - when - return-come]
• 你點解唔嚟?nei5 dim2gaai2 m4 lei4? — Why aren't you coming? [you - why - not-come]

Tag questions: 係咪hai6 mai6 and 啱唔啱aam1 m4 aam1

Cantonese has its own tag-question equivalents to English "...right?" / "...isn't it?", most commonly 係咪hai6 mai6 (a fused, faster form of 係唔係, "is it/isn't it") tacked onto the end of a statement, and 啱唔啱aam1 m4 aam1 ("correct or not," "right?") for checking agreement or accuracy: 你係香港人,係咪?nei5 hai6 hoeng1gong2jan4, hai6 mai6? (You're from Hong Kong, right?) 呢個做法啱唔啱?ni1 go3 zou6faat3 aam1 m4 aam1? (is this approach correct?)

Comparatives

Comparatives and superlatives

Comparatives are one of the more approachable grammar systems in Cantonese precisely because the core pattern is a single, consistent word (過, gwo3) doing the job English spreads across "-er than" and "more...than," with no irregular comparative forms to memorise (no equivalent of English "good→better," "bad→worse" breaking the regular "-er" pattern).

The basic comparative: adjective + 過

The standard comparative pattern places 過 (gwo3) directly after the adjective, followed by the thing being compared against — structurally "A [adjective]-than B," a mirror image of English's "A is [adjective]-er than B" with the comparison marker sitting right after the adjective instead of tacked onto its end as a suffix:

Superlatives: 最zeoi3 before the adjective

最zeoi3 ("most") sits before the adjective, the mirror image of 過's after-the-adjective position, and functions as a clean, regular superlative marker for any adjective with no exceptions:

• 佢最高。keoi5 zeoi3 gou1. — They're the tallest.
• 呢間餐廳最平。ni1 gaan1 caan1teng1 zeoi3 peng4. — This restaurant is the cheapest.
• 邊個跑得最快?bin1 go3 paau2 dak1 zeoi3 faai3? — Who runs the fastest? [combines 最 with the 得-complement pattern covered in the complements section]

啲di1: "a bit more" and direct two-way comparison

啲 (di1) attaches directly after an adjective to build a distinct comparative flavour from 過: rather than naming what's being compared against explicitly, 啲 signals "somewhat more [adjective]" as a general comparative degree, often used when the comparison point is already obvious from context or when comparing exactly two options presented together:

• 平啲。peng4 di1. — [It's] a bit cheaper.
• 呢個大啲,嗰個細啲。ni1 go3 daai6 di1, go2 go3 sai3 di1. — This one's a bit bigger, that one's a bit smaller. [direct two-way contrast, no 過 needed]
• 快啲!faai3 di1! — [Go] faster! [a bit more of the adjective, used as an imperative push]

The functional difference worth internalising: 過 explicitly states both sides of the comparison ("A [more than] B"), while 啲 states only the direction of the comparison ("[a bit] more X"), leaving the reference point implicit or already established by context. These two aren't interchangeable substitutes for each other — reach for 過 when you need to name what you're comparing against, and 啲 when the point of comparison is already clear or when you're contrasting two things side by side in the same breath.

Equal comparison: 一樣jat1joeng6

一樣jat1joeng6 ("the same") builds an equal-comparison ("as...as") pattern, typically with 咁gam3 ("so/this") marking the adjective that follows: 呢個同嗰個一樣貴。ni1 go3 tung4 go2 go3 jat1joeng6 gwai3. (this one is just as expensive as that one.) 佢同我一樣高。keoi5 tung4 ngo5 jat1joeng6 gou1. (they're the same height as me.) The negative — "not as...as" — uses 冇咁mou5 gam3 (literally "doesn't have so/this much"): 呢個冇嗰個咁貴。ni1 go3 mou5 go2 go3 gam3 gwai3. (this one isn't as expensive as that one.)

Comparative table, side by side

Complements

Resultative, directional, and potential complements

This is one of the genuinely hardest structural systems in Cantonese to get an intuitive feel for, precisely because English handles the same underlying ideas — did the action succeed, which direction did something move, is something even possible — through entirely different grammatical tools (separate verbs, adverbs, modal verbs), while Cantonese packs all three jobs into a small set of words glued directly onto the main verb. It's genuinely worth slowing down on this section rather than skimming it, because complements show up constantly in natural speech and getting them wrong (or avoiding them entirely, which learners often do without realising it) is a real gap between textbook Cantonese and spoken Cantonese.

Resultative complements: did the action actually work?

A resultative complement is a second verb or adjective attached directly after the main verb, stating the outcome or result of that action — something English usually needs a whole separate clause for ("I looked but couldn't find it" vs Cantonese's single-word-pair 揾唔到wan2 m4 dou2, "search-not-arrive," i.e., "searched and didn't find [it]"). The complement answers "and then what happened as a result?" rather than describing the action itself.

到dou2 ("arrive/reach") deserves special attention because it's the single most productive resultative complement, functioning as a general "successfully achieved/managed to" marker across a huge range of verbs, well beyond its literal "arrive" meaning: 睇到tai2 dou2 (managed to see/could see), 諗到nam2 dou2 (managed to think of/figured out), 買到maai5 dou2 (managed to buy/got hold of).

Directional complements: which way did it go?

A directional complement attaches a direction word directly after a motion verb to specify which way the action went, similar to English phrasal verbs like "come in," "go out," "bring back" — except Cantonese treats the direction word as a tightly bound complement rather than a loosely attached preposition.

These direction words frequently compound (入嚟jap6 lei4, "come in" = inward + towards-speaker; 出去ceot1 heoi3, "go out" = outward + away-from-speaker), and layering a resultative or aspect marker on top of a directional complement is completely normal in fluent speech: 攞返嚟咗未呀?lo2 faan1 lei4 zo2 mei6 aa3? (have you brought it back yet? — literally "take-back-towards-speaker-completed-not-yet-question").

Potential complements: is it even possible?

The potential complement is where 得dak1 and 唔 do their most distinctive work: inserting 得 or 唔 directly between the verb and its resultative/directional complement changes the meaning from "the action succeeded/failed as a completed event" to "the action is/isn't possible at all" — a genuinely different grammatical layer from the plain resultative pattern above.

The contrast worth holding onto directly: 揾唔到wan2 m4 dou2 (without an inserted structural slot — this is actually the negative resultative, "searched and didn't find") versus a genuine potential reading like 呢個名揾唔揾到㗎?ni1 go3 meng2 wan2 m4 wan2 dou2 gaa3? (is this name even findable?) shows how similar-looking strings can carry a plain-negative-result reading vs a genuine possibility-question reading depending on structure and context; in practice, 得/唔 + complement (揾得到wan2 dak1 dou2/揾唔到) is overwhelmingly read as the potential ("can/can't") sense in everyday speech, and that's the safe default reading to lead with as a learner.

The special case: 得 as "just/only" and as an ability marker

得 (dak1) does still more grammatical work beyond potential complements, which is worth flagging clearly since seeing the same character used three different ways can otherwise feel like an inconsistency rather than a genuinely multi-purpose word: as a bare word after a number or quantity, 得 means "only/just" (得三個dak1 saam1 go3, only three); as a standalone response, 得 means "okay/fine/that works" (得!得!dak1! dak1!, sure, sure!); and — covered fully in the next section — 識sik1 and 可以ho2ji5 are the more common way to express general ability, with 得 reserved mainly for the potential-complement and "only" uses just described.

Auxiliaries

Auxiliary and modal verbs: 得dak1, 識sik1, 可以ho2ji5, 要jiu3, 想soeng2, 應該jing1goi1

Cantonese modal verbs — the words for "can," "should," "want to," "must" — sit directly before the main verb, much like English modals, and don't conjugate at all, consistent with the rest of the verb system. The genuine difficulty here isn't structural; it's that Cantonese draws several finer distinctions than English does within a single English word like "can," splitting it across at least three different Cantonese words depending on what kind of "can" you mean.

The three-way "can": 識sik1, 可以ho2ji5, 會wui5

The clearest way to keep these three apart: ask whether you're describing a skill someone learned (識), a permission or situational green light (可以), or a prediction/likelihood (會). 我識彈鋼琴ngo5 sik1 taan4 gong3kam4 (I know how to play piano, a learned skill) is a genuinely different claim from 我可以彈鋼琴俾你聽ngo5 ho2ji5 taan4 gong3kam4 bei2 nei5 teng1 (I can/may play piano for you [right now, permission/circumstance allows it]) even though both would translate to English "I can play piano."

Desire and intention: 想soeng2, 要jiu3, 想要soeng2 jiu3

要 additionally functions as a plain transitive verb meaning "to want [a thing]" without a following verb at all: 要多杯咖啡。jiu3 do1 bui1 gaa3fe1. ([I'll] have another coffee.) — a genuinely common, simple pattern worth knowing separately from its modal "need to" use.

Obligation and expectation: 應該jing1goi1, 一定要jat1ding6 jiu3, 使sai2

得 as a permission/possibility marker in fixed patterns

Beyond its role in potential complements (previous section), 得 shows up as a standalone modal-adjacent word asking or confirming whether something is acceptable: 得唔得?dak1 m4 dak1? (is that okay? / does that work?) — one of the most common everyday questions in Cantonese, functioning almost like a catch-all "is this alright?" check, answerable with a simple 得dak1 (sure/that's fine) or 唔得m4 dak1 (that won't work/not okay).

Auxiliary verb ordering and stacking

Multiple auxiliaries can stack before the main verb, generally following a rough order of [likelihood/prediction] → [permission/ability] → [desire] → [main verb], though real speech is flexible about which combinations actually occur together: 佢應該可以嚟。keoi5 jing1goi1 ho2ji5 lei4. (they should be able to come — likelihood + permission/possibility.) 我一定要去買嘢食。ngo5 jat1ding6 jiu3 heoi3 maai5 je5 sik6. (I absolutely need to go buy food to eat — obligation + go + buy + eat, a chain of verbs in sequence rather than modal stacking, which the next section on colloquial patterns and serial verb constructions covers further.)

Adjectives and adverbs

Adjectives as stative verbs, and how adverbs attach

Here's a structural fact that quietly resolves a lot of early confusion: Cantonese adjectives behave grammatically like verbs, not like English adjectives, and once you internalise that, a whole category of "wait, where's the 'to be'?" moments disappears. English needs a linking verb before every adjective ("the soup is hot"); Cantonese adjectives function as their own predicate directly, with no separate "to be" verb required at all.

Adjectives don't need 係hai6 (to be)

係hai6 is reserved specifically for equating two nouns ("A is [the same as/identifies as] B": 佢係老師keoi5 hai6 lou5si1, "they are a teacher") or for emphatic confirmation ("yes, that's right" as a standalone reply), not for linking a subject to a plain adjective. This is exactly why 好hou2 ("good/very") shows up constantly in front of adjectives even when the sentence doesn't obviously need an intensifier — 好熱hou2 jit6 isn't necessarily "very hot," it's frequently just the natural, neutral way to predicate "hot" onto a subject, with 好 functioning partly as a grammatical linker-of-sorts as much as a true intensifier. Dropping 好 entirely (bare 湯熱tong1 jit6) is grammatical but reads as clipped or unusually blunt in most contexts, closer to a lab-report register than ordinary conversation.

Negating and questioning adjectives follows the verb pattern

Because adjectives function as stative verbs, they negate with 唔 exactly like ordinary verbs (唔熱m4 jit6, "not hot") and form A-not-A questions exactly like ordinary verbs (熱唔熱?jit6 m4 jit6?, "is it hot?") — no separate adjective-question or adjective-negation pattern to learn, because grammatically there isn't a separate category to begin with. This consistency is a genuine simplification once it clicks, even though the initial "adjectives are verbs" reframe takes a moment to settle in.

Reduplication for a vivid, descriptive intensity

Many one-syllable adjectives can reduplicate (repeat, often with a small tone shift on the repeated syllable) to add a vivid, descriptive, almost onomatopoeic intensity that a plain intensifier like 好 doesn't quite capture — closer to English's expressive doubling in "it was hot hot" said with emphasis, except fully grammaticalised in Cantonese rather than just a speech quirk: 圓圓哋jyun4 jyun4 dei2 ("roundish/nice and round"), 靜靜哋zing6 zing6 dei2 ("quiet, nice and calm"), 甜甜哋tim4 tim4 dei2 ("a little sweet, pleasantly so"). The trailing 哋dei2 (distinct from the plural-marking 哋 covered in the pronouns section — the same written character doing yet another separate grammatical job) softens the description into something gentle and observational rather than a flat statement of fact.

Common adverbs and where they attach

Adverbs in Cantonese sit before the verb or adjective they modify, the same relative position English uses, which makes this corner of the grammar comparatively approachable once the adjective-as-verb reframe above has settled in:

都dou1 deserves a specific callout because English speakers often misplace it: unlike English "also," which frequently sits at the end of a clause ("I want to go, too"), 都 must sit before the verb it modifies, never at the end — 我都想去ngo5 dou1 soeng2 heoi3, never a construction mirroring "我想去都." Internalising 都's fixed pre-verb position early avoids a recurring, easily fossilised word-order error.

Conjunctions

Conjunctions and connectives

Cantonese conjunctions link clauses in mostly familiar ways for an English speaker — "and," "but," "because," "if" all have direct Cantonese counterparts — but two genuine quirks are worth flagging early: Cantonese frequently uses paired conjunctions (one word in each clause, working together) where English gets by with a single connector, and casual spoken Cantonese drops connecting words far more often than written English ever would, relying on context and juxtaposition to signal the logical relationship instead.

Coordinating: and, or, but

Causal: because, so, therefore

Cantonese frequently uses a paired causal construction — 因為jan1wai6 ("because") in the reason clause AND 所以so2ji5 ("so/therefore") in the result clause, both present in the same sentence — where formal English would typically use only one connector and drop the other:

• 因為落雨,所以我哋冇去街。jan1wai6 lok6 jyu5, so2ji5 ngo5dei6 mou5 heoi3 gaai1. — Because it was raining, [so] we didn't go out. [both 因為 and 所以 present]
• 落緊雨,所以唔去喇。lok6 gan2 jyu5, so2ji5 m4 heoi3 laa3. — It's raining, so [we're] not going. [因為 dropped; 所以 alone still reads perfectly naturally]

Dropping either half individually is completely normal in casual speech — you'll hear 因為 alone stating a reason with the result left implicit, or 所以 alone stating a result with the cause already established by context — but when both appear, they appear as a genuine matched pair rather than redundant repetition.

Conditional: if...then

Like the causal pair, 如果...嘅話 frequently drops one half in casual speech — 落雨嘅話就唔去lok6 jyu5 ge3waa2 zau6 m4 heoi3 (if it rains, [then we] won't go) drops 如果 while keeping 嘅話 and 就, and this remains entirely natural. 就 deserves special attention as an extremely high-frequency word functioning as "then" in conditionals but also more generally as "precisely/exactly that" in other contexts (呢個就啱喇ni1 go3 zau6 aam1 laa3, this one is exactly right) — worth noticing the shared thread of 就 marking something falling into place as expected.

Sequence: first...then...after that

先 is worth a direct callout since it does double duty across this guide: as a sentence-adjacent adverb it means "first" (你先講。nei5 sin1 gong2., you speak first), and paired with 至/先 later in the same sentence it builds an "only after X, then Y" sequence (食完先至走sik6 jyun4 sin1 zi3 zau2, only leave after finishing eating) — the same underlying idea of "priority/precondition" showing up in both uses.

Purpose and contrast connectives

• 為咗wai6zo2 ("for the purpose of/in order to") — 為咗慳錢,佢冇買新嘢。wai6zo2 haan1 cin2, keoi5 mou5 maai5 san1 je5. (in order to save money, they didn't buy anything new.)
• 雖然...但係seoi1jin4 ... daan6hai6 ("although... but") — another paired construction: 雖然好貴,但係我都想買。seoi1jin4 hou2 gwai3, daan6hai6 ngo5 dou1 soeng2 maai5. (although it's expensive, I still want to buy it.)
• 反而faan2ji4 ("on the contrary/instead") — 佢反而幫咗手。keoi5 faan2ji4 bong1 zo2 sau2. (they, on the contrary, actually helped.)

Relative clauses

Relative clauses and 嘅ge3: the hardest-working particle in the language

嘅 (ge3) has already appeared twice in this guide — as a possessive particle in the written-vs-spoken section and as a sentence-final particle in the particles section — and this section covers its third, structurally most demanding job: turning a whole verb phrase into a modifier that describes a noun, doing the same work English relative clauses do with "who," "which," and "that," except positioned entirely differently in the sentence.

The core relative-clause pattern: description comes BEFORE the noun

English relative clauses trail after the noun they modify: "the book that I bought yesterday," "the person who called me." Cantonese does the reverse — the entire descriptive clause comes first, followed by 嘅, followed by the noun being described, mirroring the same "modifier-before-noun" logic that puts adjectives before nouns in both languages (English "a red book" already puts the modifier first; Cantonese simply extends that same ordering to full clauses, not just single adjectives).

Notice the bracketed portion in each Cantonese example functions exactly like an adjective sitting before its noun — however long and clause-like it gets, it still slots into the same "[modifier] + 嘅 + noun" template that a single adjective would use. This is genuinely one of the more mechanically satisfying patterns in Cantonese grammar once it clicks: there's no separate relative pronoun system to learn (no equivalent of choosing between "who" for people and "which" for things), just one particle doing the entire job regardless of what's being described.

嘅 as a straightforward possessive

The simplest, most frequent use of 嘅 links a possessor to the thing possessed, functioning like English's apostrophe-s or "of": 我嘅書ngo5 ge3 syu1 (my book), 老師嘅嘢lou5si1 ge3 je5 (the teacher's things), 香港嘅天氣hoeng1gong2 ge3 tin1hei3 (Hong Kong's weather). As covered in the measure-words section, this fuller possessive can compress into a bare measure word for close/casual possession (我本書 instead of 我嘅書) — but 嘅 remains the safe, universally correct default for any possessive relationship.

嘅 with plain adjectives

嘅 also links a descriptive adjective to its noun in many contexts, particularly with adjectives of two or more syllables or when the description is being stated with some emphasis, though single, common one-syllable adjectives often attach directly to the noun with no 嘅 at all: 靚嘅衫leng3 ge3 saam1 (pretty clothes) alongside the equally valid, shorter 靚衫leng3 saam1 (pretty clothes) — both are correct, with 嘅 adding a slightly more deliberate, spelled-out feel and its absence sounding more like a fixed, compound-noun-ish description.

When 嘅 is dropped

In fast, casual speech, especially with short, extremely common adjective-noun or possessive pairings, 嘅 is frequently dropped entirely, similar to how English speakers often drop "that" in casual relative clauses ("the book [that] I bought" is fine without "that"): 我媽ngo5 maa1 (my mum, instead of the fuller 我嘅媽媽ngo5 ge3 maa1maa1) is common in warm, familiar speech. There's no hard rule for exactly when dropping 嘅 sounds natural versus stilted — this is a genuine "listen for it in real speech" area rather than a memorisable rule, similar to how no textbook can fully specify when a native English speaker drops "that" versus keeps it.

嗰個/呢個 + 嘅: building "the one that..." phrases

Combining a relative clause with 嗰個go2 go3 ("that one") or 呢個ni1 go3 ("this one") builds the extremely common "the one that..." pattern, useful whenever you want to describe something without immediately naming it: 我要嗰個平啲嘅。ngo5 jiu3 go2 go3 peng4 di1 ge3. (I want the cheaper one.) 邊個係你噚日講嘅嗰個?bin1 go3 hai6 nei5 cam4jat6 gong2 ge3 go2 go3? (which one is the one you mentioned yesterday?)

Colloquial patterns

Common colloquial patterns and set constructions

Beyond the core systems already covered, colloquial Cantonese runs on a set of extremely frequent, semi-fixed sentence patterns that function less like "apply this abstract grammar rule" and more like recognisable templates you slot real content into — similar to how an English learner benefits from directly learning "I'd rather..." or "the thing is..." as ready-made chunks rather than deriving them fresh from first principles every time. This section works through the highest-value patterns for anyone building towards genuinely natural-sounding speech.

Serial verb constructions: chaining verbs with no connector

Cantonese frequently strings two or more verbs together in sequence with no connecting word at all, where English would insert "to," "and," or a subordinating structure: 我去買嘢食。ngo5 heoi3 maai5 je5 sik6. — literally "I go buy thing eat," meaning "I'm going to buy something to eat." Each verb in the chain describes a step in a single overall action or purpose, read left to right as the natural sequence of events:

• 佢攞本書嚟睇。keoi5 lo2 bun2 syu1 lei4 tai2. — They took a book [and came] to read [it]. [take-book-come-read, four verbs/verb-phrases in sequence]
• 我坐車去返工。ngo5 co5 ce1 heoi3 faan1 gung1. — I take a car/ride to go to work. [sit-vehicle-go-return-work]
• 行過去嗰邊坐低。haang4 gwo3 heoi3 go2 bin1 co5 dai1. — Walk over there and sit down. [walk-cross-go-there-side-sit-down]

The 有 existential pattern: "there is/are"

有jau5 ("to have / there is/are") builds existential sentences the way English "there is/are" does, placed at the very front: 有人喺度。jau5 jan4 hai2 dou6. (there's someone here.) 呢度有冇廁所?ni1 dou6 jau5 mou5 ci3so2? (is there a bathroom here?) 有啲人鍾意,有啲人唔鍾意。jau5 di1 jan4 zung1ji3, jau5 di1 jan4 m4 zung1ji3. (some people like it, some don't — 有啲 doing double duty here as "there are some [people who]...").

The 就快zau6 faai3 / 快啲faai3 di1 "about to / soon" pattern

就嚟zau6 lei4 or 就快zau6 faai3 placed before a verb signals an action is imminent, about to happen: 就嚟到喇。zau6 lei4 dou3 laa3. ([I'm] almost there.) 就快考試喇。zau6 faai3 haau2si3 laa3. (the exam is coming up soon.)

The 越...越jyut6 ... jyut6 "the more...the more" pattern

越jyut6 doubled across two clauses builds a direct parallel to English "the more X, the more Y": 越食越肥。jyut6 sik6 jyut6 fei4. — The more [you] eat, the fatter [you get]. 佢越嚟越叻。keoi5 jyut6 lei4 jyut6 lek1. — They're getting more and more capable/impressive [as time goes on].

The 又...又jau6 ... jau6 "both...and" pattern

又jau6 doubled across two adjectives or clauses expresses "both X and Y" simultaneously, usually of the same subject: 呢間嘢又平又好食。ni1 gaan1 je5 jau6 peng4 jau6 hou2sik6. — This place['s food] is both cheap and tasty. 佢又靚又叻。keoi5 jau6 leng3 jau6 lek1. — They're both pretty and smart.

The 一...就jat1 ... zau6 "as soon as...then" pattern

一 (jat1) in the first clause paired with 就 (zau6) in the second builds an "as soon as X, then Y immediately follows" meaning, emphasising immediacy: 一到就走。jat1 dou3 zau6 zau2. — As soon as [they] arrive, [they'll] leave [right away]. 佢一嬲就唔講嘢。keoi5 jat1 nau1 zau6 m4 gong2 je5. — As soon as they get angry, they stop talking [immediately].

The 嚟㗎lei4 gaa3 pattern: stating origin/nature emphatically

嚟 (lei4) plus a sentence-final particle after a noun phrase states what something fundamentally is or where it's from, with a note of clarifying/emphasising the fact: 呢個係日本嚟㗎。ni1 go3 hai6 jat6bun2 lei4 gaa3. — This is from Japan [that's the fact of it]. 佢係老師嚟嘅。keoi5 hai6 lou5si1 lei4 ge3. — They're a teacher [that's what they are].

啦, 喇, 㗎 combos in imperatives and suggestions

Building a natural-sounding suggestion or gentle command relies heavily on the right particle combination rather than a special verb form (Cantonese imperatives use the bare verb, with particles doing all the tone-softening work): 一齊去啦!jat1cai4 heoi3 laa1! (let's go together!) 唔好咁樣喇。m4 hou2 gam2joeng2 laa3. (don't do it like that [anymore].) 快啲喇,遲到喇!faai3 di1 laa3, ci4 dou3 laa3! (hurry up, [we're] going to be late!)

The 咁 "so / this much" intensifier

咁gam3 intensifies an adjective the way English "so" does in "so tired," "so expensive," and pairs constantly with comparison patterns already covered: 今日咁凍!gam1jat6 gam3 dung3! (today's so cold!) 你做咩咁忙?nei5 zou6 me1 gam3 mong4? (why are you so busy?)

Register

Register and politeness

Cantonese doesn't run on a formal honorific system the way Japanese or Korean does — there's no separate verb conjugation class for politeness, no dedicated humble-speech verb forms. Politeness and register instead live in vocabulary choice, particle choice, and a handful of specific formal substitutions, which is genuinely good news for a learner: you're not memorising an entire parallel grammar system, just a set of swaps and softening habits layered onto the grammar you already know.

你nei5 vs 您nin5: the (limited) formal pronoun

您 (nin5) exists as a formal "you," parallel to how some European languages distinguish formal and informal "you," but it's genuinely rare in spoken Hong Kong Cantonese — far rarer than its Mandarin equivalent 您 (nín) is in Mandarin — and shows up mainly in written formal address or exceptionally deferential speech rather than daily conversation. For virtually all practical spoken purposes, 你 (nei5) is used regardless of who you're addressing, with politeness signalled through other means (word choice, particles, and the softening patterns below) rather than pronoun swapping.

唔該m4 goi1 vs 多謝do1 ze6: two different "thank yous"

Mixing these up is a genuinely common learner slip worth avoiding deliberately: saying 多謝do1 ze6 to a waiter who just refilled your water (should be 唔該) reads slightly off in a way native speakers notice immediately, even though both phrases translate to English "thank you."

Softening requests and refusals

Direct requests and refusals feel noticeably blunter in Cantonese than the equivalent softened English phrasing, so several habitual softening tools carry real social weight:

• 唔該 + request — 唔該幫我攞返嗰個。m4 goi1 bong1 ngo5 lo2 faan1 go2 go3. (please help me get that back.) 唔該 functions here close to English "please," softening an otherwise bare imperative.
• 吓 for a small, low-pressure attempt — 睇吓先。tai2 haa5 sin1. (let me [just] take a look first.) 吓 signals a brief, non-committal try, lowering the stakes of the request or offer.
• 可唔可以 instead of a bare imperative — 可唔可以幫我?ho2 m4 ho2ji5 bong1 ngo5? (could you [possibly] help me?) uses the A-not-A form of 可以 (can/may) to turn a command into a polite request.
• 唔好意思 for a soft apology/interruption — 唔好意思,唔記得咗。m4hou2ji3si1, m4 gei3dak1 zo2. (sorry, I forgot.) — a lighter apology than the more formal/serious 對唔住deoi3 m4 zyu6 (I'm sorry, used for a more genuine fault or larger apology).

Colloquial vs Standard Written Chinese register in speech itself

Beyond the written-vs-spoken split covered earlier, spoken Cantonese itself has an internal register range: news broadcasters, formal presentations, and some older or more educated speakers occasionally borrow Standard-Written-Chinese-style vocabulary and slightly more formal grammar even while speaking Cantonese aloud, especially in prepared speech. A learner aiming for natural conversational fluency should prioritise the colloquial register covered throughout this guide, since it's what dominates daily speech, but recognising when a speaker shifts into a more formal register (typically signalled by reduced particle use and more Mandarin-adjacent vocabulary choices) helps make sense of news broadcasts and formal announcements once you're ready for that material.

Terms of address and honorific-adjacent vocabulary

Rather than grammatical honorifics, Cantonese politeness leans on specific respectful terms of address: adding 生saang1 ("Mr.") or 小姐siu2ze2 ("Miss") after a surname, or using a professional title directly as a form of address (老師lou5si1, "teacher," used to address an actual teacher directly, the way English rarely addresses someone as "Teacher!" but Cantonese does routinely) — 陳生can4 saang1 (Mr. Chan), 李小姐lei5 siu2ze2 (Miss Li). Among family and close friends, kinship terms substitute for names constantly even outside the immediate family, extending warmly to close family friends: calling an older woman 姨姨ji4ji4 ("auntie") or an older man 叔叔suk1suk1 ("uncle") as a respectful, warm form of address is completely standard, not literal claims of blood relation.

Why register in Cantonese doesn't map cleanly onto other languages you may already know

If you've studied Japanese or Korean before starting Cantonese, it's worth naming a specific expectation to let go of directly: there is no verb-ending politeness ladder here (nothing like Japanese's plain-form-versus-masu-form distinction, nothing like Korean's speech-level system baked into every verb conjugation), and reaching for a Cantonese equivalent of that system will send you looking for something that doesn't exist. That's not a gap in Cantonese, and it's not a simplification you need to feel suspicious of — it's a genuinely different design, where the same core grammar (the same aspect markers, the same measure words, the same sentence structure) serves every register, and the social calibration work happens entirely in word choice, particle selection, and the softening habits covered above. Coming from a language with grammaticalised politeness levels, this can initially feel like "too much freedom" or like you're missing a rule that should exist — trust that the freedom is real, and that native speakers are making the same word-choice and particle decisions you're learning to make, not silently applying some hidden conjugation you haven't been taught yet.

Reading the room: how much particle use signals register

One genuinely useful diagnostic, once you've absorbed the sentence-final particle system from earlier in this guide: particle density itself is a rough proxy for how casual a given stretch of speech is. A heavy particle presence — stacking 㗎啦gaa3 laa1, leaning on 囉lo1 and 咩me1 constantly — signals warm, casual, close-relationship speech; a noticeably lighter particle presence, closer to bare statements with only an occasional 呀aa3, signals a more careful, formal, or socially distant register, the kind you'd expect from a job interview, a formal presentation, or speaking with someone significantly senior to you in a professional context. This isn't a hard rule you can apply mechanically, but it's a genuinely reliable ear-training target: as you build listening volume, start noticing not just what a speaker said but how densely they packed particles into it, since that density alone tells you a great deal about the social register of the moment, often more than the actual word choices do.

Worked examples

Putting it together: worked example sentences across systems

Every section in this guide has covered one system at a time, which is the right way to learn each piece — but real Cantonese sentences rarely use just one system in isolation. A single ordinary sentence typically stacks a measure word, an aspect marker, and a sentence-final particle (or two) all at once, and seeing that stacking laid out explicitly, piece by piece, is often the moment the whole grammar stops feeling like a pile of separate rules and starts feeling like one coherent machine. This section walks through several complete sentences, breaking down exactly which system contributes which piece.

Example 1: a simple completed action with a particle

我噚日買咗兩本書喇。ngo5 cam4jat6 maai5 zo2 loeng5 bun2 syu1 laa3. — "I bought two books yesterday."

• 我ngo5 — pronoun, "I," no case marking needed regardless of sentence position.
• 噚日cam4jat6 — time word, "yesterday," sitting right after the subject as covered in the word-order section.
• 買maai5 — the main verb, "to buy," unconjugated.
• 咗zo2 — perfective aspect marker, confirming the purchase is a completed event.
• 兩loeng5 — "two," using the special counting form rather than 二, because it's directly counting objects.
• 本bun2 — the measure word specific to bound, page-based objects like books.
• 書syu1 — the noun, "book(s)."
• 喇laa3 — sentence-final particle marking a change of state / "now it's a settled fact," commonly paired with 咗 the way it is here.

Seven grammatical jobs, seven small pieces, each covered in a separate section of this guide, all cooperating in one twelve-syllable sentence with zero conjugation anywhere in it.

Example 2: a question stacking A-not-A with a directional complement

你今日返唔返嚟呀?nei5 gam1jat6 faan1 m4 faan1 lei4 aa3? — "Are you coming back today?"

• 你nei5 — pronoun, "you."
• 今日gam1jat6 — time word, "today," subject-adjacent as expected.
• 返...返faan1 ... faan1 — the A-not-A question pattern built directly from the directional/aspectual verb 返 ("to return"), repeated in positive and negative form.
• 唔m4 — the general negator, sitting inside the A-not-A repetition.
• 嚟lei4 — directional complement, "towards the speaker," clarifying that the return is specifically back to here (as opposed to 返去, returning to somewhere else).
• 呀aa3 — softening sentence-final particle, keeping the question conversational rather than clipped.

Example 3: a comparative sentence with a modal and a classifier-possessive

我覺得你嗰架車平過我架好多。ngo5 gok3dak1 nei5 go2 gaa3 ce1 peng4 gwo3 ngo5 gaa3 hou2 do1. — "I think your car was a lot cheaper than mine."

• 我覺得ngo5 gok3dak1 — "I feel/think that," a common opinion-framing opener.
• 你nei5 — "you," here functioning possessively without needing 嘅, since a measure word directly follows.
• 嗰go2 — demonstrative, "that," requiring the measure word slot right after it.
• 架gaa3 — the measure word for vehicles, doing double duty here as the possessive link between 你 and 車 (the classifier-as-possessive pattern from the measure-words section).
• 車ce1 — the noun, "car."
• 平過peng4 gwo3 — comparative structure, adjective + 過, "cheaper than."
• 我ngo5 — "I/me," here as the comparison target — and notice 架 doesn't need to repeat before 我, since the noun (car) is already established and understood.
• 好多hou2 do1 — "a lot," intensifying the degree of the comparison.

Example 4: a conditional with a paired conjunction and an experiential marker

如果你未去過嗰間餐廳嘅話,我哋聽日就去啦。jyu4gwo2 nei5 mei6 heoi3 gwo3 go2 gaan1 caan1teng1 ge3waa2, ngo5dei6 ting1jat6 zau6 heoi3 laa1. — "If you haven't been to that restaurant before, let's go tomorrow."

• 如果...嘅話jyu4gwo2 ... ge3waa2 — the paired conditional construction, "if...then," bracketing the condition clause.
• 你nei5 — "you," subject of the condition clause.
• 未mei6 — "not yet," the negator specifically for something that hasn't happened but still could.
• 去過heoi3 gwo3 — "have been [to]," the experiential aspect marker paired naturally with 未 to mean "haven't ever been."
• 嗰間餐廳go2 gaan1 caan1teng1 — demonstrative + measure word (間, for buildings/venues) + noun, "that restaurant."
• 我哋ngo5dei6 — "we," using the productive 哋 plural suffix.
• 聽日ting1jat6 — time word, "tomorrow."
• 就zau6 — "then," the conditional-result connective, marking the suggestion as the natural consequence of the stated condition.
• 去heoi3 — the main verb, "to go."
• 啦laa1 — softening sentence-final particle, turning the suggestion into a gentle "let's..." rather than a flat command.

Notice how much of this sentence is genuinely predictable once you know each individual piece — nothing here required a special exception or an irregular form once the condition clause, the experiential marker, the measure word, and the particle were each already familiar on their own. That's the real payoff of learning the systems in this guide separately: fluent Cantonese sentences are built by combining a manageable number of well-understood parts, not by memorising an unbounded number of whole-sentence templates.

Common mistakes

Common learner mistakes and L1 interference

Every mistake in this section shows up constantly and predictably, because each one traces back to a specific mismatch between how English (or, for some learners, Mandarin) handles a grammatical job and how Cantonese handles that same job. None of these mean you're bad at the language — they mean your brain is doing exactly what brains do, reaching for the pattern that worked in a language you already know. Naming the mismatch directly is usually enough to start noticing it in your own speech, which is the real first step towards correcting it.

Mistake 1: forcing a subject/object pronoun where Cantonese drops it

English requires an explicit subject in almost every sentence ("it's raining," not just "raining"), so English-speaking learners often over-supply 我, 你, 佢 where a native speaker would let context carry the information instead, as covered in the word-order section's pro-drop discussion.

• Overcorrected: 你食咗飯未呀?我食咗飯喇。nei5 sik6 zo2 faan6 mei6 aa3? ngo5 sik6 zo2 faan6 laa3. — grammatically fine, but restating 我 in the reply feels slightly stiff/over-explicit.
• More natural: 食咗飯未呀?食咗喇。sik6 zo2 faan6 mei6 aa3? sik6 zo2 laa3. — both subject and object dropped, understood entirely from context, and this is what a native speaker actually says.

Mistake 2: skipping the measure word, or defaulting to 個 everywhere

As covered in the measure-words section, English simply doesn't require this category of word at all, so the instinct to skip it entirely, or to reach for 個 as a universal fallback, is extremely common and extremely persistent even at intermediate levels.

• Wrong: 一車jat1 ce1 — missing the measure word entirely, not a valid Cantonese phrase.
• Marked/non-native-sounding: 一個車jat1 go3 ce1 — grammatically parseable but immediately flags as non-native, since 個 isn't the correct measure word for vehicles.
• Correct: 一架車jat1 gaa3 ce1 — 架 is the correct measure word for vehicles.

The fix that actually works long-term: stop learning nouns in isolation and start learning noun + measure word as a single vocabulary unit from day one (書syu1 is really 本書bun2 syu1, 貓maau1 is really 隻貓zek3 maau1), the same way English speakers just learn "a pair of scissors" as a fixed unit rather than deriving the "pair" part from a rule.

Mistake 3: confusing 唔, 冇, and 未

Covered in full in the negation section, but worth restating as the single most common negation error: using 唔 to negate a completed past action instead of 冇.

• Wrong: 我唔食咗飯。ngo5 m4 sik6 zo2 faan6. — mixes 唔 with 咗 in a way that doesn't match how native speakers negate a completed action.
• Correct: 我冇食飯。ngo5 mou5 sik6 faan6. — 冇 replaces both 唔 and 咗 for negating a completed past event.

Mistake 4: reversing relative-clause order (Mandarin/English word-order bleed)

Because English trails its relative clause after the noun, learners sometimes try to build a Cantonese sentence with the same after-the-noun order, producing a sentence that simply doesn't parse the way intended, or defaults into being read as two separate, disconnected statements instead of one modified noun phrase.

• English-order attempt (doesn't work as intended): attempting a literal "書我尋日買syu1 ngo5 cam4jat6 maai5" as a standalone description of "the book I bought yesterday" reads as an incomplete or oddly topic-fronted fragment, not a smoothly modified noun phrase.
• Correct Cantonese order: 我尋日買嘅書ngo5 cam4jat6 maai5 ge3 syu1 — the full descriptive clause comes first, 嘅 in the middle, noun last, as covered in the relative-clauses section.

Mistake 5: treating aspect markers as if they were tense

Learners coming from tensed languages sometimes try to use 咗 as a blanket "past tense" marker on every past-referring verb, even where the sentence calls for 過 (experiential) or where no marker is needed at all because a time word already does the job.

• Overused 咗: reaching for 咗 to describe general past experience — 我細個嗰陣時去咗好多次日本ngo5 sai3go3 go2zam6si4 heoi3 zo2 hou2 do1 ci3 jat6bun2 attempts to describe repeated childhood trips using 咗, when 過 (experiential, "have been [before], multiple times") fits the "life experience" framing better than 咗's "one specific completed event" framing.
• Better fit: 我細個嗰陣時去過好多次日本。ngo5 sai3go3 go2zam6si4 heoi3 gwo3 hou2 do1 ci3 jat6bun2. — 過 correctly frames this as accumulated life experience rather than one discrete completed event.

Mistake 6: skipping sentence-final particles entirely

Because English carries equivalent nuance through intonation and hedge words rather than a dedicated particle slot, learners sometimes build grammatically correct Cantonese sentences with zero sentence-final particles at all — technically fine, but landing noticeably flat and abrupt to a native ear, closer to how a string of unhedged, particle-free statements can come across as blunt or even slightly rude in English too, just without the tool most English speakers reach for automatically (a softening "...you know," a trailing "I guess," a rising "...right?").

• Flat/abrupt: 我唔想去。ngo5 m4 soeng2 heoi3. — grammatically complete, reads as a blunt, hard refusal with no particle attached.
• Softened, natural: 我唔想去喎。ngo5 m4 soeng2 heoi3 wo5. or 我唔想去呀。ngo5 m4 soeng2 heoi3 aa3. — particles soften the refusal into something more conversational and less confrontational.

Mistake 7: applying Mandarin pinyin reading habits to jyutping

For learners with a Mandarin background specifically, jyutping's letter-to-sound mapping genuinely diverges from pinyin in several places (covered directly in the jyutping section), and the single most common resulting error is misreading "j" as the pinyin "j" sound (like English "j" in "jump") instead of correctly reading it as an English "y" sound. 人jan4 spelled jan4 is pronounced with a "y," not a "j" — reading it with an English/Mandarin-pinyin "j" sound produces a mispronunciation a native Cantonese speaker will need a moment to parse.

Mistake 8: overusing 的-adjacent constructions from Mandarin instead of 嘅

Learners with Mandarin exposure sometimes default to 的dik1 in speech out of habit, since 的 is Mandarin's direct equivalent to 嘅. In colloquial spoken Cantonese, 的 sounds distinctly bookish/written, or flags the speaker as thinking in Mandarin-grammar terms rather than genuinely colloquial Cantonese — reach for 嘅 by default in spoken contexts, reserving 的 (or accepting it passively when reading) for Standard Written Chinese material specifically.

Mistake 9: building A-not-A questions incorrectly with 有/冇

Because most verbs follow the straightforward pattern (verb + negative-verb, e.g., 食唔食sik6 m4 sik6), learners sometimes over-regularise 有 into the same pattern, producing 有唔有jau5 m4 jau5 instead of the correct, irregular 有冇jau5 mou5, since 有's negative isn't a "唔有" construction at all but the fused word 冇.

A quick self-check habit

A genuinely useful practice, once you've read through this list: the next time you catch yourself translating an English sentence directly into Cantonese word for word, pause and ask which of these nine patterns might be quietly bleeding through. Most of these mistakes stop being conscious effort within a few months of deliberate noticing — the goal here isn't memorising a checklist forever, it's building the initial awareness that lets your ear start catching these patterns on its own in native audio, which is ultimately where the real correction happens.

What actually closes these gaps, and what doesn't

It's worth being honest about which study habits genuinely close these nine gaps and which ones only feel productive. Re-reading a grammar table (including this one) a second or third time rarely fixes a fossilised error on its own, because the mistake isn't a knowledge gap at that point — you already know the rule intellectually, the way you can recite it here — it's a retrieval-speed gap between knowing the rule and producing it correctly under the real-time pressure of an actual conversation. What does move the needle: deliberately noticing the error in your own output (catching yourself mid-sentence, or reviewing something you wrote and asking "did I use 個 there because I didn't know the real measure word, or because I forgot to check?"), and getting enough volume of native input that the correct pattern becomes the version your ear expects to hear, so that your own incorrect version starts to sound audibly "off" to you before anyone else has to point it out. That shift — from a rule you follow consciously to a pattern that simply sounds right or wrong to your own ear — is the actual finish line for each of these nine mistakes, and it arrives at a different pace for each one, which is completely normal rather than a sign you're progressing unevenly in some concerning way.

Applied practice

How can I practise Cantonese grammar instead of only reading rules?

The most useful answer to how can I practise Cantonese grammar is to work with short spoken exchanges that force a choice between real forms. Read the pair, cover the Cantonese line, produce it from the English intention, then change one detail. Use Jyutping to check sound and tones, but keep the characters visible so spoken grammar does not become a romanisation-only exercise.

A 30-minute Cantonese grammar workshop

1. Minutes 1–5: review one contrast, such as 唔 versus 冇, 咗 versus 過, or 呢 versus 嗰.
2. Minutes 6–12: read five examples aloud and underline the word that carries the contrast.
3. Minutes 13–20: hide the examples and answer five short prompts. Say the answer before checking Jyutping.
4. Minutes 21–26: change the time, person, object, or sentence-final particle while preserving the core structure.
5. Minutes 27–30: record two sentences, listen once without reading, and note whether the problem is tone, segmental sound, or grammar.

Keep the workshop spoken. Formal written Chinese can be valuable, but it does not substitute for producing colloquial Cantonese forms such as 佢, 冇, 喺, and sentence-final particles.

Aspect in dialogue

How do 咗, 緊, 住, and 過 change a Cantonese sentence?

Practise aspect by holding the verb constant and changing the speaker's view of the event. With 食 (sik6, eat), 食咗 presents a completed event, 食緊 presents eating in progress, 食住 presents a sustained state or action held while something else happens, and 食過 presents experience at least once.

你食咗飯未？ (nei5 sik6 zo2 faan6 mei6?)

Have you eaten yet? The question asks whether the meal is complete.

我食緊飯。 (ngo5 sik6 gan2 faan6.)

I am eating. The action is in progress now.

你坐住先。 (nei5 co5 zyu6 sin1.)

Sit for now. The position is maintained while the situation continues.

我食過嗰間餐廳。 (ngo5 sik6 gwo3 go2 gaan1 caan1 teng1.)

I have eaten at that restaurant before. The sentence reports experience, not a specific completed meal.

Add time words separately. 尋日 (yesterday) can locate a completed event, while 而家 (now) supports an action in progress. Cantonese aspect markers do not behave as simple replacements for English past, present, and perfect tense labels.

Particle clinic

How should I learn Cantonese sentence-final particles?

Learn a particle inside a recurring social action: asking for confirmation, softening a reminder, reporting new information, or urging a change. A one-word English translation is rarely stable enough. Compare the same proposition with different endings and listen for the speaker stance.

• 呀 (aa3/aa4): commonly marks questions, responses, or a softened tone; its exact effect depends on tone and context.
• 啦 (laa1/laa3): can mark a changed situation, suggestion, or urging, with tone affecting force.
• 喎 (wo3): often presents information as noteworthy, newly noticed, or contrary to an assumption.
• 啫 (ze1): can limit or downplay: “only,” “just,” or “that's all.”
• 嘅 (ge3): can add an explanatory or assertive stance as a final particle, separate from its possessive and relative-clause uses.

Record whole lines from reliable audio and imitate the intonation. Flashcards for particle names can support recognition, but varied mini-dialogues are the central answer to how can I practise Cantonese grammar when the meaning depends on relationship and discourse.

Exercises

Cantonese grammar practice: choose the form

1. You want to say “I have not eaten yet.” Choose 唔 or 未 and write the colloquial sentence.
2. You are watching someone read right now. Add the appropriate aspect marker to 佢睇書.
3. You want to ask whether a friend has ever been to Hong Kong. Use 過.
4. Complete “this book” with the demonstrative and classifier: ___ 本書.
5. Change 我有錢 (I have money) to “I do not have money.”
6. Write one sentence with 喎 that reports information the listener may not expect.

Answer key and reasons

1. 我未食飯。 (ngo5 mei6 sik6 faan6.) 未 means the expected action has not happened yet.
2. 佢睇緊書。 (keoi5 tai2 gan2 syu1.) 緊 marks the action in progress.
3. 你去過香港未呀？ (nei5 heoi3 gwo3 hoeng1 gong2 mei6 aa3?) 過 marks experience; 未 asks whether it has occurred yet.
4. 呢本書 (ni1 bun2 syu1), “this book.” 本 is the classifier for bound volumes.
5. 我冇錢。 (ngo5 mou5 cin2.) 冇 is the negative counterpart of 有.
6. Answers vary. A suitable line must present the information as noteworthy in context; the particle cannot be judged from the isolated sentence alone.

Repeat the exercise with new verbs and nouns. That substitution is how to practise Cantonese grammar as a productive system rather than memorise six completed answers.

Cantonese learner questions

Is Cantonese grammar harder than Mandarin grammar?

Both share broad features such as limited verb inflection, classifiers, topic-comment organisation, and aspect marking, but their forms and everyday usage differ. Cantonese has a rich system of sentence-final particles and a substantial gap between colloquial speech and Standard Written Chinese. Mandarin learning materials may feel more standardised, but familiarity and exposure strongly affect perceived difficulty.

Do not learn Cantonese as Mandarin with changed pronunciation. Study Cantonese pronouns, negation, aspect markers, particles, vocabulary, and spoken sentence patterns directly.

Should I learn spoken or written Cantonese first?

If conversation is the goal, prioritise colloquial spoken Cantonese and learn how it is represented in Cantonese characters and Jyutping. At the same time, recognise that formal writing in Hong Kong often follows Standard Written Chinese. The two varieties coexist and serve different situations.

Label sources clearly. A news article, formal notice, chat message, subtitle, and transcribed conversation may not use the same grammar or vocabulary. Reading skill eventually needs both the formal written standard and colloquial forms.

Are Cantonese tones part of grammar?

Tones primarily distinguish words and morphemes, so they are part of the sound system rather than grammar in the narrow sense. However, tone interacts with grammatical and discourse meaning in particles, changed tones, and the pronunciation of forms inside real speech. A particle with a different tone can express a different stance.

Practise tone and grammar together in complete short lines. A correct word order with the wrong lexical tone may change a word, while a correct sequence of words without the intended particle tone can change the social effect.

How do Cantonese classifiers work in everyday speech?

Classifiers appear between a number or demonstrative and a noun: 三本書 (saam1 bun2 syu1), three books; 呢個人 (ni1 go3 jan4), this person. The classifier also helps identify the noun when the noun is omitted: 我要嗰本 (ngo5 jiu3 go2 bun2), “I want that volume/one.”

Learn common noun–classifier pairs from useful phrases. 個 is frequent and sometimes substitutes in casual speech, but it is not the most natural classifier for every noun. Contrast pairs rather than memorising a list without objects.

Particle priorities

Do I need to learn every Cantonese sentence-final particle?

No. Begin with frequent particles and the communicative actions you need: forming questions, softening statements, marking changed situations, giving reminders, and presenting new information. Learn each from several recorded mini-dialogues rather than a large translation table.

Expand the inventory as listening exposes new contrasts. Some particles combine, change force with tone, or belong more strongly to particular registers. Recognition can grow before confident production. Accurate use of a smaller frequent set is more useful than inserting many particles without understanding speaker stance.

Tone progress

How long does it take to hear Cantonese tones clearly?

Progress depends on the listener's language background, audio exposure, and whether practice uses isolated syllables or real words. Begin with contrasts in familiar vocabulary, then identify tones inside short phrases and multiple speakers. Expect lexical recognition to improve before perfect tone-number labelling. Record confusion pairs and revisit them after a delay. The useful goal is understanding and producing the intended word reliably, not winning every abstract tone-identification test immediately.