(It is produced in American English and the word is one syllable. The time scale is arbitrary so feel free to ignore it.) I was wondering what cues everyone would look for when reading this, and perhaps possible suggestions for a word that produces this spectrogram.

For example, I am guessing that the beginning of the word is a stop because of the closure and burst between (1) and (2) (I am not sure if this is right), but I cannot figure out what type of stop it is.

I am also confused as to how to tell if the movements of F2 and F3 are due to diphthongs or due to glides/other things, and how to detect the presence of nasals since I am not sure I can spot a cue.

Any input would be much appreciated!

  • This feels like a homework problem or similar; I'd recommend making it broader, and instead of asking for the IPA transcription of a particular word, just ask what cues people look at to figure out the different segments on the spectrogram (like the second half of your current question). That'll make the question more useful to others and less likely to get closed.
    – Draconis
    Jan 5 at 18:09
  • (For example, looking at this, I see an /s/, a diphthong vowel, and a nasal (not in that order); if you broadened the question, I and others could explain what makes us think this and what cues you can look at to figure out which diphthong it is etc.)
    – Draconis
    Jan 5 at 18:12
  • @Draconis thank you so much for your suggestions! I'm still relatively new to spectrograms, so I am not quite sure how to spot for nasals and determining whether F2/F3 movements are due to consonants or vowels.
    – Jay
    Jan 5 at 19:41

The question of glide vs. diphthong is unanswerable from a spectrogram, since that is a claim about phonological analysis and is not a measurable acoustic property. The fact that this is English means that guesswork can get you pretty far, for example you don't have to consider the possibility that it contains [ɖ], [ʕ] or [ɗ]. Voicoids are the low-hanging fruit which you might as well pluck first, so you should estimate values for F1 and F2 at some points, and see what vocoids best fit those numbers. The difference between "continuously changing" versus "relatively level" tells you something (namely, the articulators aren't moving). Vocal fold vibrations are also within reach – you can count roughly how many times the vocal folds bang into each other during this utterance. Since this is English, you have a limited number of possible things that can be at the beginning of the utterance.

A brute-force solution is to test reasonable hypotheses, by recording and spectrogrammifying (a non-standard verb) various utterances in English and comparing what you say with what you see. To get a clue as to the first segment, you might test the sample against "sock, talk, dock, knock, stock". That procedure should rule out a number of possibilities, and in doing this you will develop a starter theory of acoustic cues (focusing on major class properties of initial consonants). I suggest also including "auk". The same kind of comparison method can be used to pin down the place of articulation of a consonant, by looking at the pattern of formant changes into and out of consonants. For instance, "bug, dug, jug, gug" (it doesn't have to be an actual word) and "big, dig, jig, gig" are fodder for a theory of consonant place cues.


From about 1.5 to 2 on this timescale, there's a lot of periodicity (the vertical stripes), and a strong first formant (the dark bar at the bottom), but not much energy above that. We don't see any other especially strong formants, for example, or any energy up at high frequencies. This usually indicates a sonorant consonant, such as /n/ or /l/.

From 2 to 7 or so, we see that same periodicity, but now with multiple strong formants. This generally indicates a vowel or semivowel, like /a/ or /j/. Furthermore, those formants change significantly over the course of the vowel, which suggests a diphthong.

The start of the diphthong has a very high second formant, up around 2500 Hz, which tends to indicate a front vowel. High vowels tend to be more fronted than low vowels, so I'm guessing this is an /i/ or a /j/.

The end of the diphthong has a second formant somewhere around 1250 Hz, which isn't especially front or especially back, so I'm guessing that's a schwa (/ə/).

Finally, from 7 onward, we see a lot of noise, which tends to indicate a fricative. There's a lot of it at very high frequencies, which usually means a sibilant. And at the same time there's still periodicity (the regular vertical bars at the bottom), which indicates that it's voiced.

Putting it all together, I'm guessing that this is a word like "nears" in a non-rhotic accent (/niəz/). The next step would be to record myself saying "nears" and comparing the spectrogram of my recording against the given one, seeing which parts look similar and which parts look different. Personally I've never been good at distinguishing sonorants, so I'd probably also try "leers" and "rears", and see which one looks closest.

(P.S. This recording doesn't seem to have any stops in it, but you can identify those by a period of almost total silence followed by a distinct release.)

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