Sounds Like a Word: a phonological choice test online
Two made-up words sit side by side, such as klowd and klowb. Neither is spelled right, but one of them sounds like a real English word when you say it in your head. Pick that one. 32 pairs, no clock and no sound.
Start Sounds Like a WordWhen the 32nd pair is answered you see how many sound-alikes you found and your median time, and a free account keeps them.
How it works
4 practice pairs with feedback, then 32 scored pairs with none. No pair has a time limit, and most people are done in about 3 minutes.
- 1 · First
4 practice pairs
After each one the screen names the sound-alike and the word it sounds like. When fewer than half are right, the 4 pairs come back once with a one-line reminder. Practice is never scored.
- 2 · Each pair
Say both silently
A cross shows for 300 ms, then the pair appears under the question Which one sounds like a real word? It stays until you answer. Sound out both, the way you would read a new name.
- 3 · Your answer
Left or right
Tap or click the sound-alike. With a keyboard, F or the Left arrow picks the left string and J or the Right arrow the right one. A 300 ms blank separates pairs.
- 4 · Your score
Sound-alikes found
Every pair counts once, so the score runs from 0 to 32. Your time on each correct pair feeds the speed and accuracy figure.
Every pair is made the same way. One string is a pseudohomophone: misspelled, yet read aloud it is a real word, as klowd is cloud. The other is that string with one letter changed, so it looks almost identical and sounds like no word at all. The sound-alike is on the left in 16 pairs and on the right in the other 16, so always choosing one side gets about half.
The design follows the phonological choice task of Olson, Forsberg, Wise and Rack (1994). It takes the place of a spoken-sound task: you need no speakers, headphones or microphone, only your inner voice. Form A and Form B hold different pairs in a fixed order, and a retake on the same device moves to the form you did not take last time.
What it measures
The task isolates decoding: turning letters into sounds in a string you have never seen, then hearing whether those sounds make a word you know.
Sounding out
- Applying letter and sound rules to an unfamiliar string, since neither option is something you could have stored by sight.
- Holding the sounds in mind long enough to hear the word they build.
- Knowing the real word behind the sound-alike, which is why the pairs are built from familiar words.
Telling the pair apart
- Catching that one changed letter turns cloud into nothing.
- Ignoring appearance: both strings look equally odd, so memory for spellings gives no clue.
- Working at your own pace, with time recorded but never forced.
Decoding tends to stay the hardest reading skill for adults with dyslexia. Pennington and colleagues (1990) found weaker phonological skills, such as sounding out made-up words, in adults with dyslexia long after school. Callens, Tops and Brysbaert (2012) saw the same kind of gap in students starting university. Familiar words can be learned one at a time by sight over the years; a string you have never seen cannot be read that way.
English and Spanish need different care here. English allows several spellings for one sound and its vowels shift with accent, so the English pairs avoid sounds that change from one accent to another. The Spanish pairs use only spellings that sound the same in every Spanish dialect, such as b and v or a silent h.
A test, not a diagnosis.
32 choices describe how one attempt went. They cannot diagnose dyslexia or rule it out.
What you get
The result appears the moment you answer the 32nd pair.
On the result screen
- How many of the 32 pairs you got right, and that number as a percentage.
- Your median time on correct answers.
- Speed and accuracy: that median time divided by your share of correct answers. Lower is better.
- A line saying no comparison group exists yet, so only your own numbers are shown.
- For clinicians: median time on all answers, pairs answered, the form and the device type, and a CSV download.
When you keep it
- Saved to a free account, ready to reopen on another device.
- Part of the dyslexia check, where it is read with Word or Not, Right Spelling and a few questions.
- Continue brings you back to the tests hub.
How this version is scored
The record of each pair holds both strings, the side the sound-alike was on, your choice and your time in milliseconds.
- Correct
- You picked the pseudohomophone, the string that sounds like a real word. There is no partial credit and no penalty for taking your time.
- Response time
- From the first frame that shows the pair to your tap or key press. The median uses correct answers between 300 and 8,000 ms, leaving out stray taps and long pauses.
- Speed and accuracy
- speed and accuracy = median correct RT ÷ share of pairs right. A careful reader who is rarely wrong and a quick one who often is can end up far apart on this figure even with similar times.
- Chance level
- With 2 options, guessing lands near 16 of 32. An attempt with fewer than 20 right, under 60%, is marked as not giving a clear reading.
Time is recorded, not demanded. There is no deadline, so a careful pace costs nothing on the number right. Time matters only through the speed and accuracy figure, and only correct answers count toward it.
Device type matters. A phone adds a small, steady delay to every tap. Times are only compared with attempts made on the same kind of device.
No percentile yet. No published table exists for these pairs. Percentiles will be added after 300 qualifying people per language and device type have taken this version; for now, only the dyslexia check uses provisional cut-offs.
Before you start
With no clock running, the main thing is to read each pair properly.
For a cleaner attempt
- Say each string silently from start to end; the first letters of both are often the same.
- Keep a steady pace instead of racing, since every answer is timed.
- Silence notifications: a long pause to read a message stretches your times.
- Wear your glasses if you use them, so a single changed letter is easy to see.
Interpret cautiously when
- English is not your first reading language, or your accent makes a sound-alike sound off.
- You did not know the real word a sound-alike stands for.
- You were tired or distracted, or answered by guessing a side.
- You changed devices between attempts.
Questions
What is a pseudohomophone?
A made-up spelling that sounds like a real word when read aloud, such as klowd for cloud. It looks unfamiliar, so you cannot recognise it by sight. To find the word inside it you have to sound it out, which is the skill this task looks at.
Why is there no audio?
Everything happens in your head as you read, so there is nothing to play or record. That means no speakers, headphones or microphone, and you can take it in a quiet room or on a train with the sound off.
I read well. Why did this feel slow?
Many fluent readers recognise familiar words by sight and rarely sound anything out. Made-up strings take that shortcut away, so the task can feel slow even for strong readers. A slow but accurate attempt is common and is not a finding on its own.
Can Sounds Like a Word diagnose dyslexia?
No. It looks at one reading skill in one sitting. Only a qualified professional can diagnose dyslexia, after a full assessment. The dyslexia check on this site uses this task with two others and some questions about reading, and it is a check, not a diagnosis.
Does it work on a phone?
Yes. The 2 strings fit side by side on a phone held upright, and each is a large tap target. Taps on a phone add a little delay, so compare a phone attempt only with other phone attempts.
Sources
- Pennington, Van Orden, Smith, Green and Haith (1990): Phonological processing skills and deficits in adult dyslexics, Child Development 61(6)
- Callens, Tops and Brysbaert (2012): Cognitive profile of students who enter higher education with an indication of dyslexia, PLoS ONE 7(6)
- Suárez-Coalla and Cuetos (2015): Reading difficulties in Spanish adults with dyslexia, Annals of Dyslexia 65