PLEASE NOTE: I've become disillusioned with Lucida Sans Unicode--I've done some checking, and some of the diacritics are just wrong, and the linespacing is weird--some styles are lower than others. So enough of that. I'm now touting the virtues of Gentium, a freeware font by Victor Gaultney that (despite some minor kerning problems) is quite beautiful, legible, readable, etc. Failing that, I'm also testing the Beta of SIL's SILDoulosUnicodeIPA(all one word), the Unicode-compliant layout of their popular Doulos. The advantage is the Unicode organization. The disadvantage is that you *must* use Keyman or Insert Symbol or something to use it (or for that matter any Unicode special characters) into a word-processing document. And at least in Word (2000, ver 9.0.somethingsomething, SR something something), while the non-spacing characters look fine (even beautiful) on screen, for some reason when you print them, they come out spaced way too far to the left. It seems that Unicode 'smart font' features are getting outsmarted by Word.

On the other hand, these things look GORGEOUS on screen, and print fine for me off of the browser when coded in HTML. So, since this is coded in HTML, I'm switching to them. This month I'm leaving up the regular SILDoulos IPA93 as before, but have switched the defaults for the Unicode to Gentium (first) and SILDoulosUnicodeIPA. At some point I'll probably stop using the regular Doulos altogether. So please please please download either Gentium or SILDoulosUnicodeIPA from the above links, so you can see the symbols as they were meant to be seen.

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 May 2003


"The damage reduces the value."

Eth, IPA 131,
[D], [ð]
Okay, there's not much going on here, and now that I'm looking at the print of the spectrogram and not the original, there's not a lot of evidence of friction for a lot of it. Starting just sy of 100 msec there's 25-50 msec of voicing (three or four pulses), there's also a couple of very noisy pulses (look at the high frequencies) suggesting either frication, or the fricated release of a stop. Okay, so maybe I should have marked this as 'raised'. But at least it's fully voiced. No useful transition information, so go for coronal by default. Voiced, coronal, probably plosive or fricative. Moving on.

Schwa, IPA 322,
[], [ə]
Well, from 125 msec to 175 msec or so there's a vowel. The F1 is just a little low of mid, the F2 just a little high of central, and the F3 just plain high. Must have been particularly short that day, if you follow. But look at the distribution of the four visible formants. Dead even, all the way up. Must be schwa, although the F1 is a little low.

Lower-case D, IPA 104,
[d], [d]
Well, once again, not a lot of useful transition information, so if you're guessing, guess coronal. Definitely voiced for most of its duration, it's got a nice sharp release burst with some very high frequency frication. This is pretty classic for coronal plosives.

Ash, IPA 325,
[Q], [æ]
It's worth noticing that this vowel is pretty long, and things this long tend to be low vowels. So it's probably safe to ignore the F1 transition from the preceding consonant. on the other hand, the F2 is definitely diving down into the following consonant, so we have to factor that part of the F2 transition out. So we have a vowel whose F1 (excluding the first third or so as transition) has a *very, very* high F1, and whose F2 (ignoring the last third or so as transition) is also very, very high. F3 is not doing anything useful, so we'll ignore it. So the F1 says this vowel is low, and the F2 says this vowel is front. How many low, front vowels can you find in English? Well, maybe three, but ASH is the most obvious guess.

Lower-case M, IPA 114,
[m], [m]
So there's 50 msec (or so) between 400 and 500 msec where there's definitely strong voicing (note how it doesn't die out like in the earlier [d]), but of lower overall amplitude than either flanking vowel. So we definitely have a sonorant of some kind. Given the abruptness of its edges, not to mention the discontinuity in the first formant, I'm inclined to say 'nasal' over oral approximant, although it's not really perfect for a nasal. The F2 and F3 (and F4 if you're paying attention, and even the F1 if you look close) are all coming down into this segment, and rising out of it. When your formants are pointing down, that's usually labial, so this is probably [m]. Unfortunately, there's not a lot of zero/pole information in the middle to add to this guess.

Schwa, IPA 322,
[], [ə]
Once again, you have to ignore the transitions, but this is pretty classic schwa otherwise.

Lower-case D, IPA 104
[d], [d]
Sometimes I wonder about my plosives. Okay, there's a gap (ignoring the junk in the F2 range, which just seems to be me and phrase-internal plosives), with clear, if dying, voicing at the bottom. The transitions here (F2 and F3 at least) seem to point up, if they point anywhere, which is typical of coronals. Of course, it helps that the F2 of the flanking vowels is relatively low to the 1700-1800Hz 'locus' of alveolar transitions.

Yogh, IPA 135,
[Z], [ʒ]
Okay, this is the release of the preceding plosive, so phonologically we have an affricate here. Which narrows down the possibilities for English. But in case it wasn't English, or you weren't sure, this looks like a Yogh. It's fricative, it's quite high amplitude, it's very broad band, and it dies in the low frequencies. THe big difference between 's' and 'sh' is a) the frequency of the peak amplitude of the frication (high for 's' and in the F2-F3 or occasionally F3-F4 range for 'sh'), and the shape of the noise in the lower frequencies (sharp cut off below 1500 Hz or so for 'sh', where it's more gradual for 's'). I thought there was evidence of voicing during this, but now I don't see it. But knowing this is an affricate and believing the preceding plosive to have been voiced ....

Turned R, IPA 151,
[], [ɹ]
There's that low F3. There must be an /r/ in here somewhere.

Barred I, IPA 317,
[], [ɨ]
I went with Barred I, because I couldn't decide what else this would be. It's not really high (the F1 is low of mid, but not really Low). It's vaguely front (the F2 is high of central, but not really High--it doesn't help that the F3 is pushing down). The F3 is useless, since we know there's an /r/ in there somewhere. And listening to it (come back when this gets archived), it just sounds reduced (probably due to the /r/ colo(u)ring). Moving on.

Lower-case D, IPA 104,
[d], [d]
Again. Nothing new to add, really.

Barred U, IPA 318,
[], [ʉ]
I'm not sure I've ever used this symbol in a spectrogram before. This vowel was definitely round, contrary to all expectations (it's my western US speech, it follows a coronal). The vowel is moderately high (as high as vowels get in this stupid spectrogram). THe F2 is not high enough for this to be front, even if it is round. So call it central, and round, and move on.

Lower-case S, IPA 132,
[s], [s]
Classic, if slightly weak, [s]. The highest amplitude of the frication is in the very high frequencies, it's sort of broad band, and it dies (relatively) slowly as you go down in frequency. Okay, so it zeroes out abruptly at 1300 Hz or so, like the previous Yogh. But there's just no way to read the center of the energy in the F2/F3/F4 range in this one.

Schwa IPA 325,
[], [ə]
Another one of these too, albeit with more classical-looking formants.

Lower-case Z, IPA 133,
[z], [z]
Well, if the preceding consonant was [s]. this is just the same, but with voicing.

Eth + Raising SIgn, IPA 131 + 429,
[D3], [ð̝]
Well, this looks slightly voiced, but that might just be the background noise. But it's got a short VOT, whatever it is. Okay, so if you think it's a stop, fine. That's why I used the raising sign. It's not bilabial looking or velar looking, so it's probably coronal. It could be a [d], I guess, but it certainly isn't a /d/.

Schwa, IPA 322,
[], [ə]
I'm not sure I ever had this many reduced vowels in a spectrogram, let alone actual schwas. And this is a little long. But there you go.

Lower-case V, IPA 129,
[v], [v]
Well, it's voiced. There's stuff going on above, which looks like pulsing, but if you look really, really closely (and wish really, really hard), you can see that the apparent pulses in the middle of this don't line up at all with the pulses in the voicing bar. So either this is mid-frequency mushy stuff in the middle of one of my plosives, evidence of noise, or evidence of 'double clunking'--the kind of thing you get with velar releases (and occasionally 'distributed' coronal releases. Okay, so there's no release here (it's in the middle, after all), this may be some structure, like my fat lower lip, flapping against some other structure, like my upper lip or my upper incisors. You pick, based on what kinds of words you can get out of the strings.

Ash, IPA 325,
[Q], [æ]
I love it when the F1 and F2 steady states don't overlap at all. As in the previous [ae]. Well, here's another one, although the F2 isn't really front at all, and the F1 is tough to reckon given the apparent coarticulation with the following thing, whatever that is. SO it could be another schwa, but it's just too long and loud. And high pitched (if you can see the pulses). So something that ends up a low vowel, not at all back. Again, depending on what the voice is like you have your choice of three vowels. Again, you pick.

Lower-case L + Mid Tilde, IPA 155+428 (composed as 209),
[l], [ɫ]
Well, it would be easy to miss this, but there's this pretty abrupt 'thing' in the F2, around 1400 or 1425 msec, a discontinuity. This moment is when the F1 stops moving, the F2 lowers abruptly, and the amplitude of F3 suddenly drops. And something weird happens in F4, but I don't know what that's about. So something is happening here. F2 is low, so whatever it is, it's back. So it's either [w] or [l].

Lower-case J, IPA 153,
[j], [j]
Okay once and for all, see how the F2 of /j/ here is much much much higher than the F2 during the thing that turns out to be an [l]. So explain to me how this[l] can be anything but dark (velarized) when it's F2 is nowhere anywhere near front. Sorry, folks, while onset /l/ may not be as dark as coda /l/, US (and Canadian) English /l/s are never anything but velarized. In my experience. And certainly in my voice. Further more, the need for it to be very, very back, is so at odds with the need to produce a front glide into the following vowel, that the front glide gets displaced way way way later rather than risk fronting the [l]. At least that's my story.

Okay, so anyway, The F1 is pretty low (again, as low as it ever gets), the F2 clearly has a very high target, but no steady state. So this is probably a glide of some kind, and it must be front.

Barred U, IPA 318,
[], [ʉ]
Well, setting aside the end-of-utterance low amplitude problem, the 'targets' for this thing appear to be moderately high (lowish F1), and not front at all, but either sort of back but not round, or sort of round and not really far back. So I used the barred U again.