Solution to Last Month's Mystery Spectrogram

Solution for January 2007

labeled spectrogram
"We have two dogs."

So, it's January, and we're starting with basics. The main exercise here is to recognize my 'point' vowels, the vowels that are the furthest apart in my vowel space. These are [i], [æ], [ɑ], [u]. And I worked pretty hard at producing an actual back round [u], in spite of the consonant context. You'll see that I failed, but I blame that on context. More below.

[w], IPA 170
Lower Case W
Much as I, as a phonetician, prefer to avoid the term 'glide', it's worth noting the term's facility in describing the appearance of these pre-vocalic, (semi)vowel-like things, which are most reasonably regarded not as steady-state 'segments' on their own but as beginning (or ending) points of transitions. That is, they 'glide' into (or out of) the following vowel. So when you see this kind of movement, especially where the transitions are 'straight', i.e. not accelerating or decelerating, think of the semivocalic approximants. This one has an absurdly low F2 starting point indicating something very back and round. If yout ake the three (or so) pulses before the F2 and F3 kick on (about 20-25 msec around the 100 msec tickmark) as a 'moment' of steady state, even though you can't see the F2, it has to be really really low. The F3 starts around 2500 Hz, which is fairly neutral, not greatly lowered as it would be for [ɹ], and not at all raised as it would be for a lateral. So that pretty much leaves [w].

[i], IPA 301
Lower-case I
Our first vowel. Abstracting away from the transition, we've got a fairly steady F2 from 200 msec to almost 350 msec. The F1 goes from around 250 msec and moderates a little before you get to 300 msec, but it's still well below neutral (around 500 Hz for F1), so this must be a fairly high vowel. F2 is absurdly high, up around 2300 Hz, which is about as high as my F2 ever gets. It's practically into my F3 range. So this is very front. About as front as you can get. High and front. And fairly long. Must be [i].

[ɦ], IPA 147
Hooktop H
Okay, this may be the first time I've used this symbol—it's certainly the first time I've used it in a long time. Hooktop-h is the voiced counterpart to [h], and if you're wondering how you can voice an [h], think "breathy voice". So you can see the striations at the bottom. That's voicing. Above that, however, the energy is mostly noisy rather than periodic. That's a fricative. So we've got a voiced fricative. What makes it glottal, or rather epiglottal (don't ask), is that the noise is clearly organized into formant-like resonances, and continuous with thevowel resonances. So the F1 is around 900 Hz, right in line with the following vowel. F2 is transitioning between very high to, well, falling (it kind of levels out, a little, for a while in the middle, but you can't really tell that from the formant trace). The F3 also transitions, a little less, toward neutral. I don't know what's going on in the F4, but then I don't care. So what we have here is something with the oral articulation of a vowel, or semivowel, or something vowel-like, with resonances and no real closure, but lots of friction echoing around the oral cavity. And voiced. So this is an /h/. Or rather, [ɦ].

[æ], IPA 325
Ash
Now we have another vowel. Again, F1 is fairly flat for most of it, until the last quarter or so. It't high, around 800 Hz or so, which is about as high as my F1 ever gets. Well, not quite, but close. So this is very, very low vowel, relatively speaking. F2, while moving, kidn of flattens out, briefly, sort of, around 1500 Hz, maybe a little lower, which is essentially neutral. But by no means back. And it's hard for very low vowels to be very front, so you have to interpet this carefully. So very low, which in my voice means only [æ], [ɑ] or in a pinch [ʌ]. And it's not back enough to be [ɑ]. Which limits the possibilities. And once you clue in to the 'pointiness' of the vowels in this spectrogram, you arrive at [æ]. But note the fallinng formants in the last 25-50 msec or so.

[v], IPA 129
Lower-case V
Falling transitions in all formants like that almost always mean labial, if not bilabial. There's a short (probably voiced) gap for about 30 msec or so around 650 msec, but then it's noisy, but without a release burst or anything. So this is probably more fricative than anything else. Possibly tightened to stoppiness, but since there's no plosion, I'm hoping you'll believe me if I suggest that it just takes that amount of time to recover airflow and generate enough pressure to get this noise. That and I had to fix so much in the transcription so many times that I just don't have it in me to go back and add a raising diacritic. So we've got a voiced labial fricative and in English that can only be labiodental. Moving on.

[tʰ], IPA 103 + 404
Lower-case T + Right Superscript H
So there's a real gap from 700 to about 775 msec, with a nice sharp release burst and about 50 msec of aspiration. By which I mean [h]-like noise between release and the onset of voicing. So we have an aspirated plosive here. The burst noise is oddly concentrated in the F2/F3 range, and depending on how you interpret it it could be F2/F3 'pinchy'. But that would be inconsistent with the noise in the aspiration, which is [s]-shaped—a single broad band, centered in the highest frequencies. So on balance this looks alveolar.

[ʉ], IPA 318
Barred U
Another vowel, this time with the F1 really, really low, below 250 Hz or so. So a really high vowel. F2 is quite low, almost as low as it ever gets in my voice, certainly in a real vowel. So we have a backish and mostly unround vowel. Or possibly a centralish and fairly round vowel. Or something. I was trying hard to make this round, and not at all front, which my post-alveolar /u/s tend to be (due to the western US merger of post-coronal /u/ and /ju/). I'm quite proud that there isn't a distinct frontish on-glide to this. So not as back as it could be, but fairly round.

[d], IPA 104
Lower-case D
Okay, here's that big red herring. The transitions into and out of this are exceedingly misleading. But to details. Voiced, weakly, but you can see it down there at the bottom. F2 transitions up to near 1500 Hz, and then down again on the other side. So definitely not labial. Now for an alveolar we'd hope to see something even higher than that, around 1700 to 1800 Hz or so. But you can't have everything. The real scary part is the F3 and F4 transitions, which are just, well, misleading. Usually, downward pointing transitions, particularly upper transitions, are indicative of a labial. Not here. At least not 'officially'. I think this just indicates the extent of rounding in the preceding vowel. So given the uppy-pointing F2 and the downy-pointing F3, this looks like velar pinch. Except there's no reason for F4 to do that for a velar. So on the balance, alveolar is probably the best guess, but there's no shame in being misled. This spectrogram is about the vowels anyway....

[ɑ], IPA 305
Script A
The formant trace places the F1 of this around 700 Hz or so, but I'd actually place it a little higher. I think what it's finding is the a fairly strong harmonic at the bottom of a fairly diffuse band, and I'd put F1 up around 800 Hz. Similarly I think the F2 is placed between the 'true' F2 and the 'true' F1, so the F2 should be up around 1100 Hz or so. But even if you trust the formant trace, we've got something on the lower end of mid, and something in the very backish/roundish space. The only thing that's ever back there in my speech is some kind of reflex of the "LOT"/"THOUGHT" vowel, which are merged in my dialect. The [ɑ] character is clearer if you place the formants where I place them, but given this is my voice, even something lower-mid and that back could only be a /ɑ/ just because there's nothing else back there for it to be.

[ɡ], IPA 110
Lower-case G
Now here's a nice example of velar pinch. Even though it's not very 'pinchy' the transitions into this gap have to be velar. F2 rising and F3 falling, with F4 rising or hanging out. In some models F2 and F4 are coupled (as the first and second resonances of one cavity or other) especially approaching velars. That combination of rising F2 and falling F3, in the absence of any other evidence, indicates velarity. So this is a velar plosive. At the bottom, we've got fairly strong striations that last most of the closure, so this must be voiced. Purely perseverative voicing into a voiceless closure usually doesn't last more than a 1/4 of the vowel duration or so, and is usually weaker and dies off more sharply.

[z̥], IPA 133 + 402
Lower-case Z + Under-Ring
And finally, we have a nice fricative. It seems to be strongest off the top of the spectrogram, so its center seems to be above 4400 Hz, which is where my spectrograms typically cut off. The only thing that centers that high is an alveolar fricative. While not strictly speaking 'voiced', (no striations) this is too weak to be a fully voiceless [s]. A voiceless [z] is produced with passive devoicing (i.e. not enough pressure-drop across the glottis, with adducted vocal folds) as opposed to a truly voiceless [s], which would have high flow across the glottis and abducted vocal folds.

So, how'd you do?