<html> <head> <meta http-equiv="content-type" content="text/html;charset=utf-16" /> <title>Solution to Last Month's Mystery Spectrogram - Rob Hagiwara</title> <meta name="keywords" content="Hagiwara, phonetics, linguistics, spectrogram, spectrogram reading, mystery spectrogram, solution" /> <meta name="description" content="Solution to last months mystery spectrogram" /> <meta name="author" content="Rob Hagiwara" /> <link href="../rzstyle.css" type="text/css" rel="stylesheet" /> </head> <body bgcolor="#CCCCCC"> <script type="text/javascript"><!--// copyright 1998 Idocs, Inc. http://www.idocs.com/tags/// Distribute this script freely, but please keep this // notice with the code.// var image information objectvar mloi=new Object();// set the image for swappingfunction setswap(){if (! document.images)return;var imgInfo=new Object();imgInfo.defaultImg = new Image();imgInfo.defaultImg.src = document.images[arguments[0]].src;imgInfo.opts = new Object();for (var i=1; i < arguments.length; i=i+2) { imgInfo.opts[arguments[i]]=new Image(); imgInfo.opts[arguments[i]].src = arguments[i+1]; }mloi[arguments[0]] = imgInfo;}// set up the link for swappingfunction swap(link,imgName,optName){if (! document.images)return;if (! link.swapReady) { link.imgName = imgName; link.onmouseout = swapBack; link.swapReady = true; }document.images[imgName].src=mloi[imgName].opts[optName].src;}function swapBack() {document.images[this.imgName].src=mloi[this.imgName].defaultImg.src}//--></script> <script type="text/javascript"><!--setswap("navbar", '2curm', 'hdrs/nav2curm.gif', '2sol', 'hdrs/nav2sol.gif', '2arch', 'hdrs/nav2arch.gif', '2how', 'hdrs/nav2how.gif', '2rsch', 'hdrs/nav2rsch.gif', '2cour', 'hdrs/nav2cour.gif', '2lab', 'hdrs/nav2lab.gif', '2dept', 'hdrs/nav2dept.gif', '2uofm', 'hdrs/nav2uofm.gif' );//--></script> <h2>Solution for December 2008</h2> <div class="spg"> <P align=center><img src="../spgs/sol0812.jpg" alt="Solution for December 2008" /><br /> <embed align="center" height="25" src="../archives/wav/wav00804.wav" controls="playbutton" width="25" autostart="false"> <noembed><a href="../archives/wav/wav0804.wav">Hear it!</noembed> &quot;The emblems are thistles and fish.&quot;</p> </div> <hr size="3" width="100%" /> <P>This started, way back before Christmas, as an attempt to resolve the problem of telling labiodental from an (inter)dental fricative. Well, we'll just see.</p> <P>[&#x00F0;], IPA 131 <BR /> Eth <!--dental fricative --> <BR /> So from about 100msec there's some weakish voicing, but the 'vowel' doesn't erally kick on until about 150 msec. The 'fuzzy' edge into the vowel makes it look less like a nasal and more like something else. The abrupt (sort of) and noisy couple of pulses in the higher frequencies (before the vowel fully kicks on) makes it look more like a fricative than an approximant. So given the odds, probably Eth.</p> <p>[&#x0069;], IPA 301 <BR /> Lower-case I <!-- cardinal 1 --> <BR /> On the other hand, the vowel here is clearly [i], which is the only thing that ever really gets an F2 around 2200 Hz or above.</p> <P>[&#x025B;], IPA 303 <BR /> Epsilon <br /> The long transition is actually another vowel. You can see this one has an F1 that is quite abit higher, rising to just above 500 Hz (let's say 600 Hz or so). So this is a middish kind of vowel, possibly lower-mid, certainly not higher mid or tense. The F2 is moving, but on average is still moderately front (I do this by looking at the middle, in between the moment the harmonics in between F1 and F2 kick in (about 250 msec) where the F2 is at about 2000 Hz or lower, and the moment the vowel kicks off, at about 350 msec (where the F2 is at about 1250 Hz or so), so the average is at about 1600 Hz or thereabouts, which is frontish for a mid-to-low vowel. So this is an [&#x025B;]. </p> <P>[&#x006D;], IPA 114 + 102<BR /> Lower-case M <!-- bilabial nasal --> + Right Superscript B<BR /> On the other hand, that low transition into this consonant is a dead giveaway that this is bilabial. If that weren't enough, this is pretty clearly a nasal (fully voiced, with weak, flat poles separated by clear zeroes), and the pole is at 1000 Hz (or close), which is also indicative of a bilabial. The sharp release indicates an oral component, which for these homorganic nasal-stop combos is all you ever get of an oral stop. Hence my transcription, which isn't strict IPA.</p> <P>[&#x026B;], IPA 209 <BR /> Tilde L (Dark L) <br /> Following the release, there's a shortish period (less that 50 msec) of 'something else' before we hit the vowel. Voiced, although weakly, with non-flat formant structure (so here the zeroes aren't zeroes, they're just stretches of weak harmonics unsupported by anything), and a little noisy (so we might be thinking fricative, but which?), this has a very low F2 (the resonances down there, weak as it is, clearly dips in the middle to below 1000 Hz)so whatever it is it's back (read: velarized). The F3 is fairly high, even 'raised' compared to the other F3s in the spectrogram, so this is a lateral.</p> <P>[&#x0259;], IPA 322 <BR /> Schwa <br /> Vowel. F1 is in the mid range, F2 is definitely backish. Schwa or something like IPA [&#x028C;] (not [&#x0250;] as in 'hut', at least for me, at least typically).</p> <P>[&#x006D;], IPA 114 <br /> Lower-case M <br /> Another nasal. Fairly clean edges with flat stuff in between. Good strong voicing bar, pole just above 1000 Hz. </p> <P>[&#x007A;], IPA 133 <BR /> Lower-case Z <br /> Now this is a fricative. The voicing sort of shuts off, but there are a few pulses in it. Noise concentrated in the high frequencies. Probably an alveolar sibilant. Short and potentially voiced. /z/.</p> <P>[&#x0268;], IPA 317 <BR /> Barred I <!-- high central unround --> <br /> Short little transitional vowel. I'm going to ignore it.</p> <P>[&#x0279;], IPA 151 <BR /> Turned R <br /> I'm glad I'm finally producing things. F3, at its minimum, is porbably at about 1850 Hz, starting at just about 700 msec. But it would be hard to identify that as the F3 if it weren't for the discontinuity in the F3 above it. One might be tempted to locate the F3 up there, in which case it would be above 2000 Hz. Which I've been telling people for years is not a problem (Hi S5!, keep meaning to track you down in Vermont or wherever you are now!), but here I think it's really lower. But this is a good example of the discontinuity, which suggests that something is not only producing a low resonance in the F2 range but something is anti-formanting the natural third resonance of the tube. (Consult your favourite discussion of [&#x0279;] in tube theory and one of those Guenther et al papers from the early 2000's.)</p> <P>[&#x03B8;], IPA 130 <BR /> Theta <br /> This spectrogram started out as an excuse to see [&#x03B8;] and [f] at the same time. So here's the [&#x03B8;]. Looks like plosive, whcih doesn't help. So the mystery of interdental fricatives remains. Meantime, this is clearly fortis-ized (my word--the more popular term is prosodic or initial 'strengthening' following Keating, among others), from fricative--via either extra force 'stopping' the airflow or deleting the frication in some other way--the result is something that looks and presumably sounds more plosive at a left prosodic boundary. So anyway, the F2 may be falling a little afterward (which eliminates labial as a major category) and the F3 is pretty flat afterward (suggesting coronal over velar). The lil bit o'noisey pulsey business at the edge of the folllowing vowel is what may be left over of the higher airflow. Must be coronal, can't be sibilant.<br /> </p> <P>[&#x026A;], IPA 319 <BR /> Small Capital I<br /> Vowel. Kinda short, but also kind of strong. Probably stressed so not safe to ignore it. So it's got an F1 just below 500 Hz, which probably indicates something fairly high. F2 is at about 1700 which isn't very, very front, but is fronter than most. F3 is not telling us much. So frontish and highish.</P> <P>[&#x0073;], IPA 132 <BR /> Lower-case S <br /> So here we have a sibilant. The noise is in one large band, centered off the top of the spectrogram, probably at least 6000 Hz or so. Very little formant-like shaping to the noise, indicating little coupling of the rear resonating chamber(s). Voiceless. Alveolar (due to the high cent(e)r(e).</p> <P>[&#x0259;], IPA 322 <BR /> Schwa <br /> Transitional vowel. Short, weak. Reduced.</p> <P>[&#x026B;], IPA 209 <BR /> Tilde L (Dark L) <br /> On the other hand, we want to talk about the fact taht the F3 in the schwa and here (in the weaker half of this syllable rime/coda) is distinctly raised. Raising of F3 is usually and indicator of lateralness, at least in North American English voices I'm familiar with. The low F2 is consistent with velarization, also expected if this is a lateral. (By the way, I don't mean the lateral configuration necessarily causes the high F3, only that there's a correlation. I'm waiting for Frank Guenther and Carol Espy-Wilson to solve this one too.)</p> <P>[&#x007A;], IPA 133 <BR /> Lower-case Z <br /> Okay, now thi sis important. Note that this stretch of whatever is voiced. Not strongly and not particularly consistently but voiced. So that's the first thing. Notice that there's very little to it except a) the voicing and b) the bi tof noise at the top. Now notice that this noise, such as it is, is in a single broad band, no formant-like shaping, and centered in the very-high frequencies. Spectrally, this looks like an [s], but weaker, and voiced. Guess what. It is like an [s], but weaker and voiced. And shorter, if it comes to that.</p> <P>[&#x0259;], IPA 322 <BR /> Schwa <br /> Moving on.</p> <P>[&#x006E;<sup>d</sup>], IPA 116 + 104 <BR /> Lower-case N + Right Superscript D<br /> k, this is gonna take some creative hand waving. Voiced, but weaker than a vowel. So we're talking really tight approximant or nasal. Sharpish edges and flattish resonances, so that argues for nasal. Which nasal. Can't be velar (no pinch on either side). F2 may be rising into it and is definintely not rising out of it, so this is unlikely to be bilabial. Which leaves alveolar. Except that the F3 and higher formants seem to be lowering. But not the F2--rounding/labialization usually affects all formants (go back and review perturbation theory), so this is unaccountable if it's labial, so again, we're thinking alveolar. too bad there isn't a nice little pole where the F2 seems to be pointing to be sure.... The sharp offset is the oral release--someone needs to explain the timing of these to me again, but I do this pretty consistently with homorganic nasal-stop sequences. In Steriade's aperture terms it would be a [nasal] closure with an [oral] release phase.</p> <P>[&#x0066;], IPA 128 <BR /> Lower-case F <br /> Wow. Okay, this looks really loud, and maybe I was blowing into the microphone. But check out the differerence. Clearly, the noise here is reverberating through my vocal tract. Could be an [h], but then I'd expect to see some F1. This seems to get stronger in the higher frequencies, but is too resonated (?) to be a nice tight [s] (and it's too long to be a weak [s] or unvoiced [z]). So this must be one of the weaker (voiceless fricatives), either [&#x03B8;] or [f]. This one has a nice rising F3 transition into the following vowel, but with no evidence of 'pinch'. SO if it can't be an [s], and it's probably not a [&#x03B8;], then I guess it's an [f]. Hmm. This comparison didn't work out as well as I'd hoped. But oh well.</p> <P>[&#x026A;], IPA 319 <BR /> Small Capital I <br /> Well, the F1 is more obviously mid-y here, but the F2 is too front to be anything but [&#x026A;], unless this si reduced, which given its length and its strength I'd say is unlikely. </p> <P>[&#x0283;], IPA 134 <BR /> Esh<BR /> And finally, here we have a nice long, fairly loud fricative. It's got some high frequency shaping, but appears to be centered in the F3/F4 range. Below F2, no energy to speak of. That's how we always describe [&#x0283;]. </p> <div class="footer"> <script type="text/javascript" language="JavaScript"><!-- document.write(" Last modified: " + document.lastModified) // --></script> <hr /> <div class="footer"> <table width="100%"> <tr> <td width="25%"><strong>Robert Hagiwara, Ph.D.</strong><br /> Dept. of Linguistics<br /> University of Manitoba<br /> Winnipeg, Manitoba<br /> CANADA R3T 5V5</td> <td width="50%"><center> <a href="http://home.cc.umanitoba.ca/~robh/">Current Mystery</a> - <a href="http://home.cc.umanitoba.ca/~robh/cursol.html">Solution</a> - <a href="http://home.cc.umanitoba.ca/~robh/archive.html">Past Mysteries</a><br /> <a href="http://home.cc.umanitoba.ca/~robh/howto.html">How To</a> - <a href="http://home.cc.umanitoba.ca/~robh/research.html">Research</a> - <a href="http://home.cc.umanitoba.ca/~robh/courses.html">Courses</a><br /> <a href="http://umanitoba.ca/faculties/arts/departments/linguistics/lab/">To the Lab</a> - <a href="http://umanitoba.ca/faculties/arts/departments/linguistics/">To the Department</a> - <a href="http://umanitoba.ca/">To the University</a> </center></td> <td width="25%"><img src="../images/bribbon.gif" alt="Support Free Speech" align="right" /></td> </tr> </table> </div> </div></body> </html>