As a native speaker of American English, when I was listening to the difference sounds in this IPA chart, I was really surprised when I realized that I could not differentiate between p/b, t/d, and k/g. (I think I've always been distinguishing the pairs based on whether or not the consonant is aspirated.)

I know the difference has to do with vibrations of the vocal chords, but I am not quite sure what to listen for.

  • I found I distinguish voicing differently with headphones and without headphones, with the same sound sample
    – Ming-Tang
    Commented Sep 14, 2011 at 1:15
  • 1
    @SHiNKiROU: this may be because the main acoustic difference between the voiced and voiceless stops are in a very concentrated part of the acoustic spectrum, and if one of your playback devices has difficulty transmitting that portion, you may lose the distinction.
    – Steven
    Commented Sep 14, 2011 at 1:24
  • See voice onset time. Commented Sep 14, 2011 at 3:17
  • 3
    The speakers by the link indeed do not correctly pronounce "d". They pronounce it voiceless. The p/b and k/g pairs on the other hand, are OK and easily distinguishable.
    – Anixx
    Commented May 25, 2013 at 1:53
  • 1
    Sapolsky of Stanford has a nice anecdote showing how important this distinction can be. Commented Aug 2, 2015 at 19:25

7 Answers 7


Since you asked about that particular chart and the accompany voice samples, I took the samples and laid them out in Praat, a tool designed for speech analysis.

The top half of the visual representation you see is a waveform, a visualization of air pressure on the y-axis and time on the x-axis. This is the actual energy your ear picks up and interprets as sound.

The bottom half is a spectrogram, which is a mathematical transformation of the waveform into its constituent frequencies. On the y-axis is the frequency (0 Hz to 5000 Hz) and on the x-axis is still time. A dark region at 3000 Hz at some particular time X means that the acoustic signal at that moment had strong "energy" in the 3000 Hz region.

Red and blue lines and regions are explained later.

A note to those who haven't heard the actual samples in question. They are in the environment /a/ /a_a/.

Here are the results (annotations as we go along):




Corresponding speech samples:

The important difference between [b] above and [p] below is the relatively low-amplitude vibration before the red lines. This vibration is seen in the waveform as a typical wave and seen in the spectrogram as a "band" of energy in the lower frequencies. This vibration is the result of the vocal chords vibrating.

In both cases, the time leading up to the red line is characterized by a complete closure of the vocal tract by the lips (hence the term "bilabial stop"). In [b], during this time, the vocal chords are vibrating. In [p], during this time, they aren't.




Corresponding speech samples:

You see the exact same pattern here. The "stop" portions (the period immediately before the red lines) have flat waveforms and no voicing bar on the spectrograms.



enter image description here

Corresponding speech samples:

You might have expected the pattern to continue here. I did too. I believe that the speaker flubbed the pronunciation of the [d] here. As you can see, there is no voicing in the stop periods.

I voiced my own version of [d] in the same frame, and this is the resulting spectrogram and sound:


Sound: [d] version 2

Voiceless aspirated stops

Here are three voiceless aspirated stops (what we're used to seeing in English as /p t k/ when they occur as simple onsets at the beginning of stressed syllables):




Notice in all these examples, there is considerable delay after the red line to the blue area.

Corresponding sound samples:

The spectrum

So you won't have to pick apart samples from above, here's a side by side comparison between [p] aspirated, [p] unaspirated, and [b]:


Onset of voicing (blue) occurs well after stop release (red).


Onset of voicing coincides with stop release.


Onset of voicing occurs well before stop release and also fills up intervocalic (between vowels) stop duration.

Corresponding speech samples:


So what does this mean for you. As a native English speaker, you have only ever been exposed to [p] aspirated and [p] unaspirated. You understand the former to be /p/ and the latter to be /b/.

When you get presented with a sound you've never heard before in that particular phonetic environment, your brain just associates it to the nearest match. Your brain disregards the voicing before the stop release and notices that there is no aspiration duration, concluding immediately that it is also /b/. This is helped in large part in that in even slightly rapid speech, the English sound /b/ appears not as an unaspirated [p] but actually a fully voiced [b] (e.g., in the word tabs).

And that's why you may not be able to tell.

It certainly didn't help that their [d] was undervoiced.

  • The first d sound (the one from the IPA chat) is not prevoiced, whereas others (g, b) are. In languages like Russian, French, Spanish, etc, /b,d,g/ are fully voiced (or prevoiced or with negative VOT); in English, they are said to be partially voiced when intial, voiced when intervocalic.
    – RainDoctor
    Commented Jan 6, 2014 at 1:48
  • 1
    "you have only ever been exposed to [p] aspirated and [p] unaspirated." I would be very careful with that statement. This is certainly not true. Both [p] and [b] are allophones of /b/, whereas the usual realisation of /p/ is [pʰ]. As you wrote yourself "This is helped in large part in that in even slightly rapid speech, the English sound /b/ appears not as an unaspirated [p] but actually a fully voiced [b] (e.g., in the word tabs)." So, they definitely have been exposed to [b]. Commented Dec 27, 2018 at 14:17

You've mostly answered it yourself. To native English ears, aspiration is very relevant to the consonants in question. In various other languages, it's not. As aspiration is considered an additional feature in the IPA, your sound samples avoid it.

Mostly, distinguishing the sounds just takes practice. Your ears are trained to check for aspiration to find unvoiced consonants, so you'll have to "untrain" that. Also keep in mind that the vowels connected to them in the examples are always voiced, so you're really listening for whether there is a break in voicing or not.

  • I believe there are languages where voicing and aspiration are on separate axes. I am pretty sure some languages have three series of stops varying by aspiration and voicing, but I don't know if there are any which have four full series to distinguish all four possible combinations of +/- voice and +/- aspiration. Commented Sep 13, 2011 at 23:32
  • @hippietrail: To the extent that it is useful to pick from the acoustic signal voicing and aspiration "features", you may be right about them operating independently given the case of the breathy voiced stops (e.g., in Hindi).
    – Steven
    Commented Sep 13, 2011 at 23:58
  • Besides their common aerodynamic component, I would think you'd need an example of a phonological alternation selecting both voiced and voiceless [±aspirated] to convincingly label it a feature.
    – Alek Storm
    Commented Sep 14, 2011 at 0:22
  • 3
    @Alek Storm: by some analyses, Hindi has all four: voiceless aspirated, voiceless unaspirated, voiced aspirated, and voiceless unaspirated.
    – Steven
    Commented Sep 14, 2011 at 1:09
  • 1
    @dainichi: Good point. I don't know if +voiced +aspirated is possible. Time to read up on that ... OK Wikipedia seems to agree: The word 'aspiration' and the aspiration diacritic are sometimes used with voiced plosives, such as ⟨dʰ⟩. However, such voiced aspiration, also known as breathy voice or murmur, is less ambiguously transcribed with dedicated diacritics, as ⟨d̤⟩ or ⟨dʱ⟩. Commented Feb 20, 2012 at 13:44

If you're still having trouble distinguishing between voiced and voiceless consonants by sound alone, you can try a simple trick I learned in one of my introductory linguistics classes.

Practice producing the voiced/voiceless pairs (I think it works best if you say them in the context of a word) while pressing your fingers to your throat. You'll actually be able to feel the vibrations from your vocal chords when you utter the voiced consonant. I've found that it works pretty well for differentiating between more similar sounds, which are kind of hard to hear and aren't necessarily expressed in the orthography, such as the dental fricatives in English (voiceless θ, as in thistle' and voiced ð, as in 'father').

  • 2
    This is perhaps better as a comment than an answer but yes, it is a good trick :)
    – kaleissin
    Commented Sep 15, 2011 at 8:24
  • That is a good tip for voiced fricatives and other sounds that are usually voiced in English. The problem is that you can't count on the English plosives to be voiced. But this reminds me of my trick for the plosives: Use your voice during the closure (during a bilabial closure you can feel your cheeks getting bloated/puffed) and keep voicing for half a second or a second before you release the plosive. That is a true voiced plosive. (You don't have to hold the closure for half a second but that just helps you to learn) Commented Dec 27, 2018 at 15:29

You are right - it is the vibration of the vocal cord that adds "voice". Distinguishing them, however, depends on the language - whether there can be two words that differ only by that sound. If this is not the case, then the two consonants are allophones

In English, the voiced and unvoiced consonants are not allophones. Aspiration was suggested as a possible reason for not distinguishing them, but I'm not sure if this is the case - aspirated and unaspirated voiceless consonants are allophones in English, but they differ from voiced ones. For example you have:

crap - crab
bad - bat
knack - nag

So it might be the recordings that are causing the problem (I don't have quicktime to hear them)

  • 1
    The trick for checking whether a stop is aspirated or not is to speak towards a burning candle. With aspiration (and ejectives) the flame might actually go out all together. A piece of thin paper can also be used, if the stop is aspirated it will snap away from you.
    – kaleissin
    Commented Sep 15, 2011 at 8:19
  • Or put your hand in front of your mouth - if you're pronouncing aspirated plosives correctly you should feel a small current of air in your hand. For us Spanish native speakers it's the opposite problem, we find it hard both to pronounce and distinguish in speech aspirated vs non aspirted...
    – Joe Pineda
    Commented Dec 29, 2013 at 17:15
  • No, voiced [b d g] and unvoiced [p t k] plosives are allophones of the so-called voiced plosives /b d g/. Commented Dec 27, 2018 at 14:23
  • If an Indian person speaks English they usually use unaspirated plosives. You can clearly hear a difference to English spoken and transmitted by native speakers. And I wouldn't be surprised if you'd perceive their /p t k/ as /b d g/. Commented Dec 27, 2018 at 14:32
  • About the unaspirated allophones of /p t k/: there are unreleased allophones of the voiceless stops /p t k/ (unaspirated because they are unreleased). There may also be sometimes unaspirated final plosives, but that's not as common as the other alternatives, in my experience. And it is usually claimed that the unaspirated plosives in the onset after /s/ are allophones of /p t k/. Commented Dec 27, 2018 at 14:35

Normally, English "k", "t" and "p" are always aspirated, so they have a very slight "h" sound right after them. But, here's the catch: not always! When they're preceded by an "s" they're pronounced pure. Native speakers almost never notice the difference, but it's crucial if you want to pronounce correctly some languages.

Practice reading slowly "It's Pete's " versus "it spits" and you'll notice the difference in the way the "p" sounds: 1st one sounds almost as a soda can being opened at the same time the "p" is uttered, such small "h" sound is the aspirated which is missing in the second one.

Likewise with "It's Top Cat!" versus "it stops".

  • I always wonder what this analysis is based on. What's the synchronic evidence that it isn't a /sb sd sg/ cluster? I think a large part of it is due to the speakers being used to the spellings ⟨sp, st, sk⟩. Commented Dec 27, 2018 at 15:22

In a language like Russian, the voiced stops have a continued vibration of the vocal folds, even while the mouth is closed and no air is escaping. The unvoiced stops have no vibration during the time the mouth is closed and no air is escaping.

  • Interesting. So for the voiced stops, if no air is escaping, I guess that mean that the air is stored in the mouth instead?
    – dainichi
    Commented Feb 20, 2012 at 4:01

See the following link, ( and try by putting your hand on your throat) :


  • 1
    Welcome to Linguistics! Whilst this may theoretically answer the question, it would be preferable to include the essential parts of the answer here, and provide the link for reference. Commented Dec 27, 2018 at 15:57
  • The site to which you have linked is a professional essay writing service. They have a quantity of essays which are public on their site specifically so that people will find them and link to the site, thus doing their advertising for them (and so they can make posts similar to this one for their own advertising). There's no way for us to know if you linked to this knowingly, but as written, this is just spam.
    – Makyen
    Commented Dec 28, 2018 at 7:05
  • In addition, the essay you've linked is stated to be the work of a student, which is uncredited to the actual author. Is there any reason to believe that this student's work is reasonable to cite (e.g. that it's well researched)?
    – Makyen
    Commented Dec 28, 2018 at 7:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.