A rhyme game becomes reliable only when it grades pairs by their actual sound endings rather than by how the words are spelled on the page. Rhyme Match runs five four-choice rounds, asks one clear question each round, and compares candidates using the CMUdict ARPAbet transcription rather than letter patterns. Each correct answer is worth 200 points, and finishing all five rounds yields exactly 1,000 points. The five verified pairs in the game are light and night, cake and lake, moon and spoon, train and rain, and blue and glue. Because the comparison starts at the last stressed vowel and runs to the end of the word, words like rough and cough, which look like they rhyme, are correctly rejected, while words like light and late, which share only a final T, are also correctly rejected. The interface shows the target word and four neutral candidates, accepts touch, pointer, keyboard 1 to 4, arrow keys with Space, and Enter for submit, and only reveals the verified pronunciation tail after a correct submission. There is no audio playback, no account flow, and no dependency to install.
Why Spelling-Based Rhyming Games Mislead You
English spelling is a poor guide to English sound. The letters o-u-g-h encode several distinct pronunciations across common words: rough /rʌf/, cough /kɒf/, though /ðoʊ/, through /θruː/, thought /θɔːt/, plough /plaʊ/, and borough /bɜːroʊ/. A rhyming game that asks the player to match words by eye will tag rough and cough as a clean pair, even though their stressed vowel and final consonant both differ. The same trap catches light and late, where the trailing T is shared but the vowel tail, AY1 T versus EY1 T, is not. Moon and moan share a final N, yet UW1 N is not the same phoneme string as OW1 N. Players who trust spelling feel tricked when these pairs are marked wrong. A rhyme game that grades by sound tells the truth every round, and that distinction matters for anyone rebuilding their sense of what really rhymes, including kids still learning the alphabet and adults who have carried bad habits from the page.
What This Rhyming Game Actually Tests
Rhyme Match asks one clean question per round: which of the four candidate words shares the target word's ending sound, measured from the last stressed vowel through the very last phoneme? Each word is first looked up in the CMU Pronouncing Dictionary, which stores every entry as an ARPAbet string such as L AY1 T for light. The game scans that string backward until it hits the phoneme marked primary stress (digit 1) or secondary stress (digit 2). Everything from that vowel onward becomes the stressed tail of the word. Two words rhyme only when their stressed tails are character-identical and their initial consonants, the onset before that vowel, are different. That single rule explains why blue and glue pass, why moon and spoon pass, and why an identical pronunciation paired with itself is excluded. Spelling, accent, and prior word lists are not consulted. The rule is narrow and explicit, which is exactly the point of Rhyme Match, and it pairs naturally with the deeper walkthrough in Match Rhyming Words by Their Stressed Sound Tail.
How to Play Rhyme Match in Five Rounds
- Read the target word at the top of the round. The interface shows the target plus four candidate words arranged in a two-by-two grid.
- Pick the candidate whose spoken ending matches the target from the last stressed vowel onward. Trust the sound of the word, not its spelling.
- Select your choice using touch, pointer, keys 1 through 4, the arrow keys with Space, or by clicking the on-screen option.
- Press Submit match or hit Enter to lock in your answer. The round closes, and a single 200-point award is added for a correct match.
- Repeat for the next four rounds. After round five, the total reaches exactly 1,000 points when every match is correct.
Each round contains exactly one correct answer. The correct option positions rotate through the grid so that hammering one corner never clears the game: round one's correct answer sits at index 1, round two at index 3, round three at index 0, round four at index 2, and round five at index 1, using zero-based indexing. The score is exact, not rounded: five verified rounds times 200 points each equals 1,000 points.
Controls for Touch, Pointer, and Keyboard Players
Mobile users get four answer buttons plus a full-width Submit button, every one of them at least 44 pixels high. On narrow screens the four options stack into a single column; on wider screens they form a clean two-by-two grid with breakable text and no forced horizontal scroll. Pointer users can click directly. Keyboard users navigate the option grid with the arrow keys, toggle a highlighted option with Space, jump straight to options 1 through 4 with the number keys, and submit the highlighted choice with Enter. Wrong attempts are recorded as a signature built from the current round and the chosen option index. Repeating the same wrong signature is treated as an atomic no-op and does nothing. Picking a second different wrong answer closes the run, and you must hit Restart to begin again. Invalid option indexes, submitting with no selection, and any move past the option grid all preserve the current state without breaking the run.
The Five Verified Rhyme Pairs in the Game
The five pairs shipped in the game are not random pairs that happen to share letters. They are audited pairs whose CMUdict transcriptions agree from the last stressed vowel through the end of the word. The table lists each target, its correct match, the verified stressed tail, and the round in which the pair appears.
| Round | Target word | Correct match | Shared stressed tail |
|---|---|---|---|
| 1 | light | night | AY1 T |
| 2 | cake | lake | EY1 K |
| 3 | moon | spoon | UW1 N |
| 4 | train | rain | EY1 N |
| 5 | blue | glue | UW |
Each round also carries three deliberate distractors. A distractor fails the rhyme rule for a specific phonetic reason, such as a different stressed vowel or a different final consonant, so the rejected choice is informative rather than arbitrary. The same-shape trap (rough and cough) and the same-final-consonant trap (light and late) are exactly the kind of pair a sound-based rhyme game needs to expose.
Limits the Game Stays Honest About
Pronunciation varies across regions, generations, and individual speakers. This game makes a narrow, auditable claim: the five answer pairs rhyme under the cited CMUdict entries and the disclosed stressed-tail rule. It does not promise that every English accent realizes every vowel identically, that spelling predicts speech, or that the five-round set is a complete rhyming dictionary. The interface states that limitation in plain language beside the rule. Wrong submissions do not reveal which candidate would have been correct, so guessing by elimination does not turn into a free answer key. The game runs entirely in the browser, loads no audio file, no external asset, no API call, and adds no new dependency. Every word on screen is text the player reads, not sound the player hears.
Where the Pronunciation Data Comes From
The primary source is the CMU Pronouncing Dictionary, a long-maintained open project whose raw data file and source repository both returned HTTP 200 when the game was implemented. Each of the ten played words plus the spelling-trap pair rough and cough has an independent Cambridge Dictionary page that also returned HTTP 200, providing a second-source cross-check on every word in the game. A focused test fixture and a separately written tail oracle verify the five positive pairs and the spelling and vowel boundary negatives, and the test does not import runtime answer tables to decide what rhymes. The accessible page exposes only the target word, the four candidate words, cursor state, and selection state. It does not store the current ARPAbet strings, the shared tail, a correct-answer data attribute, a correct CSS class, or an answer-bearing ARIA description in the DOM. After a correct submission, a review note may show the previous pair and its verified CMUdict tail, drawn from sources such as the Cambridge page for cake. Wrong submissions keep the answer hidden so each round stays a sound-based test instead of a DOM inspection exercise.
Related reading: Shadow Match for Kids: Five Original Objects to Match.