the intervocalic consonant seemed like a voiced stop to me. turns out it's a glottal one, masquerading under the phoneme /t/. pesky, pesky... my first guess was spot on with the last two portions forming an affricate. the /l/ that i guessed in the second guess was just there because i thought maybe there was an l intervocalicly- i didn't actually think there was such a sonorous sound at the end lol. i wasn't sure if the lowering of f2 for the first vowel was due to a diphthong or your typical segment boundary shenanigans. speaking of segment boundary shenanigans, i had a sneaking suspicion that the second vowel would have been preceeded by a bilabial because of how both f1 and f2 rise as it goes on, which is why i guessed /b/. little did i know that was from a nasal, not a stop, being before the vowel. as you can tell i had some trouble with the identify of the second vowel, but that's just how it be. oh and at the second guess i thought maybe there was a veeery brief stop before the first vowel, which is why i tried /t/ for it.
the equivalent of wordle but for linguists (gramle)
this has been my favourite game since i found it and i've actually been getting quite good at it:) (given that i never actually studied english phonetics)
lexical recall was not helping because idk what a minder is, even though at guess 4 as you can see i could tell it was phonologically plausible. i really got tripped up by the last vowel, which i thought was a high front one. what even is rhotacized schwa in terms of height? i wonder if it would be closer to [ɝ] anyways. it does sound like it the way the dude pronounces the last syllable. what even is the (perceptual!!) difference between [ɝ] and [ɚ] other than transcription conventions of not putting a schwa in a stressed syllable?
a couple weeks ago i finished my linguistics minor and i thought i'd do a very different kind of post and walk you guys through decoding a spectrogram! (mistakes and all)
northern arizona university created their own version of wordle for linguists, called 'gramle,' which gives the player a spectrogram and they have to guess the word being said.
the player types in IPA symbols (which will appear in this post and you can listen to here). they are then told if the sound is wrong, right but in the wrong place, or fully right.
here's a relatively easy one that i solved in 3 guesses:
(unfortunately, i didn't take a screenshot so you'll have to ignore the photo quality and the cursor on the screen. oops!)
for those unfamiliar with spectrograms, this looks pretty unintelligible. spectrograms are a graph of frequency over time, with intensity (or amplitude) of the waves shown via shading. to understand why they're helpful, we need to understand how air moves through your vocal tract:
the airwaves in your vocal tract are composed of many different frequencies of waves, all bouncing around. air moves through your vocal tract and hits different parts depending on the shape. different frequencies of waves are then amplified or dampened by the shape your mouth was in. the frequencies of waves that get amplified the most form dark bars of intensity called "formants." these formants can then give us clues about what shape the mouth was in when the speech sound was made!
in the picture above, we can see several different patterns of darkness and lightness. those are different sounds. in gramle, we are looking for 5 different sounds.
my first step is to identify the borders of the sounds
now that we have the boundaries between the sounds, we are going to try to identify the type of sound they are, aka their 'manner of articulation' (HOW the sounds are made).
here's what we can tell:
this word looks like it has two vowels in it, separated by a period of blankness. this means it will be two syllables.
the period of blankness is clearly a stop
the sound at the end is clearly a fricative
the sound at the beginning is also a fricative, but i guessed that it was an approximant first. (this was silly of me, and we'll get into why in a little bit)
now that we know HOW the sounds are made, we can look for clues about WHERE they're made (place of articulation):
again, here's what we can tell:
both vowels have a big distance between their first and second formant. F1 is low while F2 is high. this means they are high, front vowels. probably they are /i/ or /ɪ/
the formants on either side of the stop are bending downwards. this tells us the sound is probably bilabial, meaning it's produced with two lips. it's probably /p/ or /b/
there is also a completely blank period, meaning there's no voicing bar. (voicing is the vibration of the vocal folds and it shows up as a little bit of low-frequency noise). therefore, we can make an educated guess that the sound is /p/
the fricative has a ton of noise really high in frequency. this means it's a sibilant, a class of s-like sounds that make hissing noises. alveolar sibilants, /s/ and /z/, have higher pitched hissing than their postalveolar counterparts, /ʃ/ and/ʒ/.
(we're ignoring the first sound for now)
here's what i guessed:
(this is the word "lapis")
let's break this down:
because i could not think of a word that used vowel /i/, i went with /æ/ even though i knew it was wrong. this is still a game of wordle after all!
/ɪ/ was correct, but not in the right spot, which means the first vowel must be /ɪ/. because the second vowel is not /ɪ/, it must be /i/
/p/ was correct. yay!
/s/ was not right, which bummed me. I was pretty sure it was an alveolar fricative, but I didn't see a voicing bar. my two strategies were to change my assumptions about place and vocing, changing my guess to either /ʃ/ (voiceless post-alveolar), or /z/ (voiced alveolar)
I thought that the first sound was an approximant because the sound waves seemed more organized than they are for the final sound. approximants are made by resitrcitng airflow, but not so tighly and it does not become nearly as turbulent as the air does in a fricative. I tried /l/ without much strong place evidence.
with some new thoughts, it was time for my second guess. this time, i do not aim for a real word but instead throw some sounds on there to test them.
(this is not a real word, but if it were, it would be spelled ripiche)
we got the vowels confirmed!
/ʃ/ was not the ending. i guessed this because i thought there wasn't a voicing bar, but that can be tricky. a more experienced spectrogram reader would probably guess /z/ before /ʃ/, because the place evidence was stronger than the voicing evidence. but hey, i'm still learning :)
again, i try an approximant. i did not want to pick /w/ as that would create the same labial bend we saw for /p/, and /ɹ/ keeps F1 and F2 relatively in the same spot. the problem here is that it bends F3 a lot, and so I knew it wasn't quite right either. noticing that the sound is actually quite quiet by looking on the waveform, i realize that it's probably a fricative. the absence of a voicing bar also points to a fricative, as English approximants are voiced by default
now i can try again for a real guess:
(the word is "hippies")
i tried /z/ at the end this time and it was correct!
there was not a lot of evidence for the place of the fricative, so once i knew all the other sounds, i tried to see which sound would make a real english word. my best option was /h/, as 'fippies' and 'thipies' aren't words.
/h/ is produced in the glottis, which is the opening in your larynx. there is hardly any constriction when /h/ is made and it's place as a fricative is sometimes debated. because of this, it make sense that there weren't any strong place cues.
and with that we solved the gramle!
i hope to grow better at these and maybe be able to walk you all through a harder one! thank you for reading :)
I've started attempting gramles after I saw @shoesofthefishermanswife posting about them in the linguistics tag and while I enjoy deploying my (limited) phonetics skills, the American English definitely throws me, especially when it comes to the vowels. Just picture me going "Ra, Ra" imaging I'm American and you'll have a hint of what my housemates have been going through...
once again i ask myself what logic gramle applies when transcribing the vowels of unstressed syllables. i'm still not sure how to segment the first three segments. i correctly read the cues of a low f3 at the beginning as rounding, but it was hard to see where the f3 went after the first 0.15 seconds, though i luckily guessed that it increases (so the next vowel is not rounded). if anyone knows how to tell apart any of the sonorant consonants, lmk. also apparently english has not only final obstruent devoicing but also fricative devoicing too?
