Solution to Last Month's Mystery Spectrogram

To properly view the phonetic symbols in the text below, you must have installed either SILDoulos IPA93 or Lucida Sans Unicode. If you are desperate to see phonetic symbols in SIL Sophia or SIL Manuscript, or some other kind of Unicode, drop me a line.

Solution for September 2002

"The galleys are due back next week."

Eth, IPA 131
[D], [ð]
Okay, there's no good way to tell this one. It's short to be a nasal, but nasals tend to get short initially. They also tend to not have much energy in them, and this doesn't even have a proper zero in it. So rule out Nasal. Could be some kind of weak aspiration to a stop. Could just be a really short fricative. For now, just notice something is going on between 60 and 80 msec that needs to be explained. Posit some kind of consonant, and get on to the vowel. And promise yourself to come back to this later when you have a better idea what this utterance is likely to be.

Barred I, IPA 317
[ö], [ɨ]
Well, I misled you. There's not a lot going on in this vowel either. It's just high of mid (judging from the F1) and just front of central (judging from the F2). It also seems to be moving frontwards, although you should probably notice that what it's moving to is a collision with F3. That should tell you something, at least about the following stop. But anyway, this is short. It's vaguely high and the F2 is closer to the F3 than the F1. Invoke Keating et al (1994) and call it barred-i.

Lower-case G, IPA 110
[g], [g]
Well, maybe it's me, maybe it's my microphone, maybe it's my noisy office, but I can't seem to make a good, no-nonsense stop anymore. This span is voiced, there's a sharp burst, and so, knowing how these go, I can convince myself that the low-amplitude noisy stuff in the middle of all this is worth ignoring. I really suppose I should go in fro an aerodynamic screening for a dyspraxia or dystonia or something, because something is leaking. But ignoring that, this looks like a good candidate for a stop. The F2-F3 pinch in the preceding vowel suggests velar, but the height of the pinch is boggling. Can only be a 'front velar'. Or a back palatal, or something like that. Don't get me started. Note the obvious VOT in spite of the stop being voiced for most of its duration. If I weren't so tired as I write this, I'd come up with something useful to say about aerodynamics and the duty cycle and pressure and such, but whatever.

Ash
IPA 325
[Q], [æ]
This is the ugliest vowel I've ever seen. It seems good and long, but you'd think something so long would have a hint of a steady state somewhere. Unless this is a diphthong. Which is is in a lot of dialects, and is sort of here, depending on your definition. Okay, here's what's noticeable. Ignoring the first 50 msec or so as transition, we've got something with a high F1 (or at least something moving to a high F1). The preceding velar can only be front, and this being English that suggests that this vowel should also be front. This is consistent with the starting position of the F2, but the F2 is diving so fast-and-furious-ly that it's hard to tell whether its initial frontness, is late backness, or its transitional mid-ness should be taken seriously as a feature. There being neither a steady state or a portion where the thing even slows down for a bit, take some point in the durational center and use that. I think the center of the formant at that moment is slightly high of the mid-range for F2, so let's call it front. Lowish and vaguely front. That only leaves a couple of possibilities.

Lower-case L + Mid-Tilde, IPA 155 + 428
[lò], [l̴]
I wish I'd asked Penny to record the Unicode number for this as a composed symbol, because concatenating the plain "l" with the diacritic is just ugly. (That's Penny Gilbert, who kindly produced the table of IPA, Unicode, and SIL ALT+keystrokes that I use to look up my characters. Yay Penny.) The good news is that the falling F2 into this and the rising F2 out of it point us at a nice low F2 target that coincides (not-so-coincidentally) with a nice steady F3 and F4, at a moment when the total amplitude seems to be going down. There's a target here folks! Okay, nobody can quite tell what's happening in the F1. The F2 is low low low suggesting back back back. The F3 (and F4) are high, suggesting ... what? Well, /l/. Dark /l/, cuz it's English, it's me, and it's got that low low low F2 suggesting backness. Velarization. Darkness.

Lower-case I, IPA 301
[i], [i]
Ugly ugly ugly. Okay, pick the moment where the least stuff seems to be happening. There are huge, herkin' transitions from the offset of the preceding /l/, at about 425 msec, to apoint where the F2 reaches a maximum and the F3 reaches a minimum, somewhere between 500 and 550 msec. Then there are transitions again. So why do you suppose there's this moment where the F2 and F3 seem to come together, only to be whisked apart again? There must be some reason, so let's assume there's some kind of target to be reached here, in spite of the forces acting to pull the formants in other directions. So the F1 at this moment looks a little low, suggesting a high vowel. The F2 is frickin' high, above 2000 Hz, which is just plain high for me. And the F3... wel he F3 is a little low, but not really so low as to suggest anything like /r/. So, low F1, high F2. Must be high and front. Very front. Fronter than front.

Lower-case Z, IPA 133
[z], [z]
There's a hint of voicing here, so this must be voiced. It's noisy, what there is of it, so it's probably a fricative of some kind. It's got some high frequency energy, certainly higher off the top of the spectrogram (we only see up to about 4200 Hz) than anywhere in the lower frequencies. Sounds /s/-like to me. But voiced.

Lower-case A + Rhoticity Sign, IPA 304 + 419
[aÕ], [a˞]
Yer gonna kill me here, but I cheated. I excised this vowel and listened to it again and again, and transcribed it the way I heard it. But it looks nothing like it should. Okay, I heard something that, while not strictly speaking low, was sort of low and front. Not really ash-like, not quite as front. But not strictly speaking low. It wasn't the vowel I usually transcribe as Turned V, which I don't do strictly IPA. I use it for the reflex of short U or O or whatever it is in 'hum' and 'hut'. It wasn't that vowel. This vowel, excised, sounded a lot like that Atlantic-Canada vowel I hear on 22 Minutes all the time--the one Rick Mercer and Mary Walsh have. So that's how I transcribed it. But the acoustics suggest a high vowel, if anything. Certainly at least mid. Definitely not low. The F2 tells us precious little, given what the F3 is doing, but it still looks a little front. Abd the lowering F3 is the rhoticity. But the vowel I seem to be describing is Small-Cap I, or one of those central vowels I never remember the symbols for. But it just doesn't sound like that. It's a vowel. It's /r/-colo(u)red. Moving on.

Turned R, IPA 151
[¨], [ɹ]
Well, at least the /r/-col(u)ring in the preceding vowel is come by honestly. I'd locate the 'moment' of this /r/ at the bottom of that F3 dip, between 725 and 750 msecs. It's worth noting that the F2 appears to be moving, slighlty, at that same moment, and it's low extremum is back before 700 msec. I'd say there was a vowel moment (around the F2 extremum) and an /r/ moment (around the F3 extremum), and so there's two segments here, or at least two different sets of targets separated in time. It just so happens that the F2 target is specified in one and the F3 is specified in the other, leaving the other to vary. My take on segmentalism.

Lower-case D, IPA 104
[d], [d]
Wonder of wonders, a stop that actually looks like I gap. If you don't pay too much attention to that stuff going on around F2.... Anyhoo, the sharp transient/burst thing suggests some pressure is being built up, so this must be a stop. If you went nasal, good spotting, but then you'd have to deal with the burst. If you went prenasalized stop, well, okay, but I'd expect some nasalization on the vowel preceding, and, well, just more resonance. Based on experience with my voice. F3 transitions point up (that is, up into it, and down out of it), F2 transitions are high of mid, so this is probably alveolar. Nice strong voicing bar for a good stretch of the gap in the higher frequencies, so voiced.

Barred I, IPA 317
[ö], [ɨ]
What would a spectrogram read be without a little controversy? This does, at least at the beginning, look a little like the previous barred I. But it's longer, and it's higher in pitch (note the closeness of the striations) so it's probably stressed. English doesn't have barred-i in stressed positions. Hmm. But look at that F2. Starts off drifting slightly downward, and picks up speed as it goes. Hmm. Okay, let's be systematic. This is definitely a high vowel. It's not quite as high as the /i/ from before, but it's still high. The F2 starts mostly front and then takes a nosedive. What makes F2 to that? Well, rounding and/or backing, of course. So we've got a high vowel, moving from central and slightly front to either backer or rounder. Hmm. Well, high and central is barred-i. But how do we account for the F2 change? We could be backing the vowel, but front-to-back high diphthongs are hard to come by in English. Maybe it's just rounding. But rounding doesn't do that to an F2. What's going on? Okay, the answer is that things are conspiring here to make this harder. It comes down to this. First, this is my high vowel that isn't /i/. It's usually regarded as /u/, but being a West Coast kind of individual, it's really more central, and it's usually completely unrounded. But, like most of my American colleagues, I don't maintain a difference between /u/ and /ju/ (i.e. 'do' vs. 'due'). Second, the preceding consonant is coronal, which for me selects a something closer to [ju] as an allophone. Third, the following consonant's place is also having it's effect on the F2 transition. Take a quick look at the transitions in the vowel on the other side of the following consonant and identify the consonant's place. Right, bilabial. So there's just a little labialization at the end of the vowel, which is doing this weird thing to F2. Just for my benefit, please notice that the closest thing to a steady state in this vowel is the beginning. And please notice for my benefit that the durational center of this vowel still has an F2 consistent with a frontish vowel. Hence my argument that this thing is just frickin' front. Get over it.

Lower-case B, IPA 102
[b], [b]
Well, we've already talked about this one. The transitions, even the incoming F3, all the outgoing ones, and well maybe not the incoming F1, but there you go, point to bilabial. Particularly those transitions into the following vowel. You can't get much more classic than that. Wow. Nice little gap, some decent voicing (maybe not quite enough to warrant the voicing label, but this is definitely unaspirated, even if there is a clear lag between the release transient and the first clear glottal pulse of the vowel.

Ash, IPA 325
[Q], [æ]
This looks like a diphthong, which isn't how I want it to look. But there it is. There's a sharp pitch change about halfway through this vowel, and there's a quality change there, sort of, as well. The first part, from 1025 to about 1100 msec, has pretty steady formants, although the F1 looks definitely like it's transitioning to the second part. But taking the middle of this bit, there's the F1 of a moderately low vowel, getting lower; an F2 just a hair front of central, and an F3 that isn't doing much. The second part has the F1 of something that's definitely low, the F2 of something central inching backward, and the F3 is falling. Weird. Okay so low or mid-to-low, mid or front-of-mid, followed by definitely low, possibly central. Sounds sort of like lower-case a or Ash, followed by schwa. Which is one of the available diphthong versions of /ae/, so we'd be right. What this does to my targets-in-time theory is open to discussion. Discuss.

Lower-case K, IPA 109
[k], [k]
Well, note the transitions moving into this gap. Lowering F3, and rising F2. Classic velar pinch. There's a little glotallization or breathiness moving into this stop, I'm sure because there's a major prosodic boundary following this word (I mean, this word is at the right edge of a fairly major prosodic constituent). The burst and aspiration, if that's what you want to call it, don't provide much useful information. THe lack of energy in the low frequencies in the aspiration might be misleading since it makes the upper frequencies look stronger. But it still doesn't look /s/ like, which is how it would look if this were a /t/. The burst itself is kind of long looking, but it doesn't look double, which would have been nice as a clue.

Lower-case N, IPA 116
[n], [n]
This is a pretty good nasal, considering it's initial in a phrase. Starting at 1350 msec or so, we see good voicing, but very little upper resonance. There's a definite zeroing of some of the energy, and evidence of resonance at about 1500. The higher resonance probably means something, but I don't know what. But this is definitely a nasal, what with the sharp difference between the zeroes in the nasal and the energy in the non-formant areas in the following vowel. The transitions in the following vowel are pretty alveolar looking--falling F3 and F2 rising from just above 1500.

Epsilon, IPA 303
[E], [ɛ]
It's not easy to tell what's happening with the F1 here, since the bandwidth seems so wide, but if the top edge of the band is around 1000 (I could go a little higher or a little lower depending on how I'm feeling) and the bottom band is definitely down near the bottom of the spectrogram, and there's something like the peak of the formant or at least a strong harmonic or something at about 500 Hz, I'd say there's a good chance this is mid. At least this isn't obviously very high or very low. The F2 is high, indicating frontness, which leaves us a choice between /e/ and /E/. You might thing with the rising F2 to go with [e(I)], but then you would have been misled by the transitions to the following consonant. This is short, and it looks like it's in a closed syllable. Epsilon is probably the safer bet, but partial credit either way.

Lower-case K, IPA 109
[k], [k]
Well, here's another one, and once again it's in a coda. The useful transitions are the ones moving into this stop. This one doesn't even have any useful release information. Sorry.

Lower-case S, IPA 116
[s], [s]
Well, this is short, but it has the right spectrum, noisy energy, strongest in the highest frequencies. No evidence of voicing, although there is some energy down there. This may have a little too much energy to be a devoiced [z], but it's hard to tell. It's awfully short and weak, but with that spectrum it can't help be anything but alveolar. If the onset of the noise were sharper, I might suspect this is just alveolar aspiration. But I'm rambling.

Lower-case T, IPA 103
[t], [t]
Well, it's not much of a gap. In fact it's not a gap. But it looks like it might have been trying to be a gap. But I don't think it's my imagination that something is happening in here. There's [s]. There's something gap like happens before the following [w] (which is what that is, even if it is partially devoiced. But I'm getting ahead of myself. Following the gap-like thing, there's something that looks a heckuva lot like the short, weakish [s] we were just talking about. Which is what prompted the comment about it looking like the aspiration (or just the release frication) of an alveolar. Thus we have the thematic payoff I strive for in my writing. (Did I mention how tired I am?)

Lower-case W, IPA 170
[w], [w]
Well, you can definitely see the F2 sweeping down in the 'tail' of the aspiration/release noise, and sweeping up again for most of the vowel. The more I look at [w]s, it looks like the F2 is w-colo(u)ring the vowel. But we don't hear it as colo(u)red. But anyway, low F2, back and/or round. No appreciable effect on F3, to some on F4, which is a little weird. But without the raised F3 or F4, this is unlikely to be a 'dark l'.

Lower-case I, IPA 301
[i], [i]
Well, if you're not spotting this incredibly high F2s by now, I don't know what more to say.

Lower-case K, IPA 109
[k], [k]
Gap, followed by some kind of release and some frication. I hesitated to transcribe this as an aspirated stop, since I'm not sure this is aspiration and not just frication. Richard Wright suggested it should be transcribed as [x] or something, but I was ambivalent about that idea. Anyway, there are signs of velar pinch going into it, and there is noise in the burst/release/frication/aspiration part that's also centered in the 'pinch' range. Oddly enough the two pinches don't happen at the same frequency. Probably due to the front vowel, it's a very front velar at its onset, but either the contact slides back to a more neutral position or the tongue body rolls backward during the closure. This happens. Pat Keating told me she saw the x-rays, but I don't remember where. The front is assimilation or coproduction or something like that with the front vowel. But since this is utterance final, there's no carticulatory pressure coming from the other side. We might expect the closure to not change, but obviously we'd be mistaken.

I'm really tired.