Solution to Last Month's Mystery Spectrogram

Solution for June 2007

<A href=="../wav/wav0706.wav"> "Her pet cat's hip is bad."

[h], IPA 146
Lower-case H
We begin this month's solution with a fricative. We know it's a fricative because it's noisy. It's voiceless, because there's no indication of striated energy at the bottom (a 'voicing bar'). And it probably isn't an affricate because the stop portion of an affricate, particularly in sentence-initial position, is probably going to be strongly released, with a clear transient and burst noise, and the left edge of this fricative isn't like a plosive release at all. Instead, it seems to fade in (quickly) from nothing (instead of just 'starting' all at once, and very loudly). So this a fricative. The fact that most of the noise is in the region of the lower three formants suggests that the noise is being bounced around the resonating chambers of the vocal tract, so a) the tract is relatively open (let's say 'vowel like') and b) the noise is quite far from the open end (the lips) so that it can bounce around the resonating spaces and get transmitted to the outside (via the open end). So this is a glottal fricative.

[ɹ̩], IPA 151 + 431
Turned R + Syllabicity Mark
So, where's F1? Right! Down there below 500 Hz. Somewhere. It's hard to tell exactly when F1 is down there, since it runs into the voicing bar. But the top edge of the band is supported by a fairly strong harmonic around 500 Hz, so it can't be any higher than that. So where's F2? Right! Somewhere around or above that strong harmonic around 1200 Hz. So where's F3? BZZT. Well, I don't know, maybe you spoted it. There's another strongish looking thing rising across 1500 Hz just above the strong F2 harmonic. So while the two strongish bits are close enough together to be a single band, they don't move like harmonics--which will always be roughly parallel. The F2 band is flat, except when it starts to drop towards the end, and the F3 is rising throughout. Those aren't strong harmonics in a single strong band--these have to be two separate bands. So F3 is somewhere around 1500 Hz. And what do we call it when an F3 is that low? Right!

[p ʰ], IPA 101 + 404
Lower-case P + Right Superscript H
so we have a nice little gap, probably indicating a plosive. The residual noise in th elow frequencies is ignorable as a pulse of perseverative voicing. The gap is mostly voiceless. Further it's followed by a sharpish release burst at about 350 msec (that's a burst, cf the initial [h] onset), and about 50 msec or so of aspiration. See the noise? Very loud noise, actually, but it still has the resonances associated with aspiration rather than a fricative look all on its own. So we have a voiceless, aspirated plosive from about 275 msec to all the way past 400 msec. Place of articulation? Well, the F2 pointing down into the gap is a pretty dead giveaway for a bilabial. If you look at where the burst noise is in the release compared to where the formants end up, it looks like the F2 and F3 are rising into the aspiration as well, which is another pretty good indicator of bilabiality. I suppose we could see that low F3 as indicative of velar pinch (if we ignore the noise right in the burst and take the F2 to be basically falt), and then the energy in F2/F3 in the release would make sense. But I'm not convinced. Velar bursts are often (not always) doubled, and rarely have all that strong energy in the very, very low frequencies. Also the aspiration of velars tends to be a little longer. Which is a long way of saying that I'm guessing this is a bilabial, but if it really turned out to be a velar, I wouldn't be offended.

[ɛ], IPA 303
Epsilon
Vowel. From about 400 smec to about 500 msec. Basically flat. F1 is a little high of 500 Hz, so we're dealing with something mid-to-lower-mid probably. That is, certainly not in any way high, and probably not all that low. F2 is definitely front, being up around 1800 Hz or so. Frankly that's the F2 of a very front vowel in my experience. So we've got something monophthongal and lower-mid and front. There are a couple of vowels in that part of the space, and at least in my voice one of them tends to be diphthongized (that is, centralizing). This one isn't. So it's probably [ɛ].

[t], IPA 103
Lower-case T
Which brings us to another plosive. This one runs from about 500 msec when the vowel shuts off to at least that transient at 600 msec, which I take to be a release. Or a closure, considering there's more plosive on the other side. That duration (from 500 msec to way past 650 msec is way too long to be a simple, singleton plosive, so I'm taking that transient to be meaningful. The transitions in the preceding vowel are flat. Which doesn't tell us a great deal in terms of direct information. But not seeing velar pinch we rule out a velar, and likewise with bilabial so we're going to guess alveolar and move on. The transient thing could also be alveolar. It could also just be a wad of spit hitting something, so we'll just move on.

[kʰ], IPA 109 + 404
Lower-case K + Right Superscript H
So now we have this other plosive thing to contend with. The release burst is doubled, so we're really going to consider velar. F2 and F3 are both high, but arguably pinched together. And there's no sense in which we can see that F2 transition as rising (out of a bilabial) or for that matter pointing toward the alveolar locus, which we hope is somewhere around 1700 or 1800 Hz. It won't always (as we'll see) but we can hope. So this is a fronted velar. Voiceless and aspirated, of course. See what I mean about long aspiration?

[æ], IPA 325
Ash
Okay, so now we have a vowel, This one's moving, so we're going to have to talk about it's 'moments'. I don't know what else to call them. They're not 'spectral moments' but durational moments where 'something' seems to be going on. So F1 makes a nice little arc, with a low closure locus on either side and a maximum in the middle. That middle is important, since it's as close to the F1 'target' as we're going to get. And it's higher than the previous [ɛ], so I'm going to guess this is a low vowel. Relatively speaking. F2 is moving, so it has a tleast two movements. So taking those moments to be roughly when the F1 starts to level off and start to dive, more or less (soa bout 800 msec and about 50 msec later) we've got something that's still fairly front, and moving towards neutral. Now this is compounded by the front velar transition, but then since we mostly only get very front velars before very front vowels, I feel safe in suggesting this at least starts front. And moves centrally. F3 is flat. So we've got something that's lowish throughout (more or less) and basicaly front, with centralizing tendencies. So probably [æ]. Which makes the other one even more probably [ɛ].

[t], IPA 103
Lower-case T
So we've got another short plosive here. This one seems to have a transient which might be a release at about 925 msec or so, but the there's a delay of 10 msec or so before the frication clicks on in the lower frequencies. So probably that's a plosive release followed not by aspiration (which could come on more or less full strength with release) but by a separate fricative. Placoe of articulation? I'd say bilabial, given the F2 transition, but then I'd be wrong. Like I said, you can't always count on that locus thing, especially in codas. The F3 sin't really bilabial looking, but it's hard to tell that yet. Certainly not a velar. So could be bilabial, or (as it turns out) alveolar.

[s], IPA 132
Lower-case S
Okay, I mis-located the 'transition' between the fricative and the aspirate. I should have located the aspirate at the onset of F2, which leaves only the time from the release burst to about 15 or 20 msec later. Or something like that. So concentrating on that bit, there's no energy to speak of until you get up around F3. Then there's some fairly high frequency noise that looks like it could be a fricative. Trust me. And if it's a fricative, with that kind of spectral tilt (to the high frequences, in a single broad band) it has to be sibilant. Probably [s], or else there'd be more energy in the F2/F3 range. TMSAISTI.

[h], IPA 146
Lower-case H
So more interesting is the aspirate. Which we all should recognize at this point, being noisy (even possibly weakly voiced) and above all strongest in the formant bands. Look at that F2. You can follow it all the way from about 1700 Hz (yay!) until it reaches its 2200 Hz max around 1100 msec, and the falls again through the voiced section of the vowel. So this has to be an aspirate, i.e. some variety of /h/.

[ɪ], IPA 319
Small Capital I
Well, there's about 100 msec of vowel here. F1 indicates something mid to higher- mid--that strong harmonic or wahtever at 500 Hz being 'it' or the top edge of 'it'. F2 is very front, starting outrageously high and dropping. The fact that it's moving throughout is a little suspicious--usually there's some indication of a target approximation. So if we take the F2 extremum (the maximum just before the voicing sets in) it looks like a very front [i]. Not quite as high an F2 as we might expect, but plenty high enough. But then the movement down would have to be transitional. Which I suppose is possible. Or if we do the other thing, which is to take the middle of the voiced section of the vowel, we have an F2 around 2000 Hz. Which is very front--fronter than we might expect for an [ɪ], but is more compatible with the less-than-unequivocally-high vowel indicated by F1. If I can put it that way. So something highish and front. And relatively short, considering this is arguably the loudest vowel in the spectrogram, and probably the highest F0 (with the other possible hiF0 spot being the second syllable [ɛ]).

[p], IPA 101
Lower-case P
So here's another plosive of some kind. Same no perseverative voicing as before, but no aspiration to speak of. Well, there's a short VOT but not anywhere near enough to count. Barely the width of the hashmark at 1300 msec. So this is voicelss but unaspirated. So either it's phonologically voiced (or lax) and initial in its syllable, or it's phonologially voiceless (or tense) and final. Bear that in mind for a while. If you see the transitions on both sides, they point down into the closure, clearly indicating a bilabial. I love it when it's that easy.

[ɪ], IPA 319
Small Capital I
So basically here we have the same vowel as before except a) it's F1 is even lower (so either it's higher, or just as likely the harmonic that we're looking at is a little lower) and b) the F2 starts and ends lower. So we've got something that could be higher (but regardless it's still a highish vowel of some kind) and not quite as front. So again on the balance, I'd say this was [ɪ].

[z], IPA 133
Lower-case Z
There's a bit of very weak voicing at the bottom of the spectorgram, but nothing much above, except there's clearly some evidence of noise in the very high frequencies. But wait! When do we get noise that looks like that in the high frequencies? With nothing below it? With [s]! Imagine the noise were stronger. It would be a very broad band, loudest in the higher frequencies (higher than 4400 Hz or wherever the top of this spectrogram is) and trailing off as it gets lower in frequency. Looks like a weak [s], all right. But voiced of course.

[p], IPA 101
Lower-case P
Exactly waht's going on at the closure is not clear to me. You can see a nice clear moment where closure is reached in the higher frequencies, but that corresponds to a pulse or two of relatively strong voicing. I can only imagine that in closing the lips (and releasing the alveolar closure) there's a just enough change in the pressure that you can get just a little bit more air moving across the glottis until either a) pressure builds up and voicing ceases or b) the vocal folds move far enough apart for voicing to cease. So we've got a voiceless plosive, either way. The no aspiration/short VOT at about 1550 msec suggests something unaspirated, so once again we've got something that's either phonologically voiced (i.e. /b/) and initial in its syllable, or voiceless (/p/) and final. Since /zp/ final syllables are unlikely, I'd guess that way. Oh, I've already assumed bilabial because once again all the tranitions point down into the closure. On both sides, if you follow.

[æ], IPA 325
Ash
Do not be confused by the sudden change in voicing/amplitude in the last third of this vowel. That's just what happens to my voice when I hit an utterance final low boundary tone. I don't know why. Maybe someday I'll do an EGG study of it. But anyway, if you were thinking that's a final nasal or something, it ain't. No zeroes in the low frequencies and to much definition in the formants anyway. But it is odd that the formants all kind of level off at that point. Hmm. Okay, so let's see what's up. F1 starts fairly high and gets higher, reching it's maximum at about 800-900 Hz around 1700 msec, and flattening out from there (until you hit that final transition). Ignoring the transition at the beginning, the F2 starts at about 1750 Hz or so and transitions down to about 1500 Hz or a littl ehigher, and again flattens out. Look at that. Anywah, F3 is pretty stable and neutral throughout. So it starts fairly mid-low to low and gets lower, vaguely front and moves toward central, but ifyou know my voice, 1500 Hz is still characteristic of something fairly front. Even my unrounded/centralized [ɯ] has an F2 a little lower than that. So this is centralish-frontish. So we're probably looking at another [æ], withy maybe a schwa or [ɐ] (remember that I'm now using [ɐ] over [ʌ] for the STRUT vowel in my dialect, which isn't properly back at all) off glidey thing.

[d], IPA 104
Lower-case D
And finally, we've got another plosive. This one is clearly (and quite strongly) voiced through the closure duration, and even a little further. Obiously over-enunciated, but oh well. It was late, as I recall. F1 transitions down, which is consistent with approaching closure, F2 is basically just sitting there around or just above 1500 Hz, although it may be pointed just a wee bit upward from there. Note the corresponding bit of noise on the other side of the release (which may or may not be meaningful) is at 1700 or so. Hmm. That number sounds familiar? In what context does that number keep coming up? Hmm. The F3 transition in the preceding vowel is also not telling us much, being flat, but at least it isn't falling, which would be telling us something like velar or bilabial. So the flatness isn't telling us much, but it in not saying something specific, it's pointing us toward something else. That is, alveolar. Which is consisted with the high-frequency noise in the release burst, and that teeny bit of 1700 Hz resonance or whatever it is in that tail of voicing that follows the release.