Inspired by @kbmcgowan:
If I were queen of #phonetics for a day, I would remove #GlottalStop from the plosive row of the IPA chart. We could just put it it in the chaotic "Other Symbols" category and call it "glottalization segment" until we convincingly figure out what the damn thing actually is.
@lisa_b_davidson @kbmcgowan It's something containing a double 't' that by its own definition should never be pronounced. Schrödinger's phoneme.
@lisa_b_davidson @kbmcgowan I'm not familiar with this debate. What's the argument for it not being a stop?
@joshisanonymous @kbmcgowan As Maddieson and Ladefoged observed (as well as many others, like @cdicanio, Marc Garellek, etc), glottal stops are only rarely produced as actual closures. They range from a dip in intensity and/or F0 to a period of creaky voice to, at the most extreme, a full closure. Their realization can be conditioned by prosody & phonotactics. It's not like there's a raging debate right now about whether or not it should be classified as a stop, but imo maybe there should be.
@lisa_b_davidson @joshisanonymous @kbmcgowan There's this weird thing too - we hear closure when there is none. That makes them quite unlike stops, though affricates kind of come close - if you have enough of an amplitude dip, perhaps it's enough to fool our ears.
@cdicanio @joshisanonymous @kbmcgowan Christian, do we really hear *closure*? We certainly hear that thereʻs a glottal element, which can be cued by a full closure, a period of creaky voice, or even a dip in F0/amplitude. But I donʻt think we have any way of knowing if people perceive it as literal closure or not. At the same time, people are probably really bad at distinguishing between those phonetic implementations, because all of them point to the same phonological representation.
@cdicanio @joshisanonymous @kbmcgowan Thereʻs probably some kind of ABX or other discrimination task in here, Christian... Whenever you want to do it, you know where to find me 
!
@lisa_b_davidson @joshisanonymous @kbmcgowan Yes, it would be a neat study - probably a variant on Gerfen and Baker (2005) (itself based on Hillenbrand and Houde 1996)? The former clearly has a much lower threshold for "a glottalize vowel is present" for Mixtec listeners than Hillenbrand and Houde did for English listeners. Though, the language-specific JNDs have not been systematically investigated.
@lisa_b_davidson @joshisanonymous @kbmcgowan But it would be easy enough to present stimuli to Triqui listeners or Mixtec listeners and to then play the same experiment to English listeners in the lab. Experiment 1 would manipulate synthesized tokens and play them to listeners. Experiment 2 would use natural stimuli perhaps?
I'm clearly thinking about this too much now.